diff --git a/.miniconfig b/.miniconfig new file mode 100644 index 0000000..5686e53 --- /dev/null +++ b/.miniconfig @@ -0,0 +1,89 @@ +#make allnoconfig KCONFIG_ALLCONFIG=miniconfig +CONFIG_X86_32=y +CONFIG_CLOCKSOURCE_WATCHDOG=y +CONFIG_LOCKDEP_SUPPORT=y +CONFIG_SEMAPHORE_SLEEPERS=y +CONFIG_MMU=y +CONFIG_GENERIC_ISA_DMA=y +CONFIG_GENERIC_HWEIGHT=y +CONFIG_DMI=y +CONFIG_INIT_ENV_ARG_LIMIT=32 +CONFIG_IKCONFIG=y +CONFIG_IKCONFIG_PROC=y +CONFIG_SYSFS_DEPRECATED=y +CONFIG_BLK_DEV_INITRD=y +CONFIG_SYSCTL=y +CONFIG_EMBEDDED=y +CONFIG_PRINTK=y +CONFIG_BASE_SMALL=1 +CONFIG_BLOCK=y +CONFIG_IOSCHED_NOOP=y +CONFIG_DEFAULT_IOSCHED="noop" +CONFIG_X86_GENERIC=y +CONFIG_X86_L1_CACHE_SHIFT=7 +CONFIG_GENERIC_CALIBRATE_DELAY=y +CONFIG_X86_WP_WORKS_OK=y +CONFIG_X86_BSWAP=y +CONFIG_X86_CMPXCHG64=y +CONFIG_X86_INTEL_USERCOPY=y +CONFIG_X86_TSC=y +CONFIG_PREEMPT_NONE=y +CONFIG_VM86=y +CONFIG_HIGHMEM=y +CONFIG_FLATMEM=y +CONFIG_MTRR=y +CONFIG_HZ_250=y +CONFIG_PHYSICAL_ALIGN=0x100000 +CONFIG_ARCH_ENABLE_MEMORY_HOTPLUG=y +CONFIG_PM=y +CONFIG_ACPI=y +CONFIG_ACPI_SLEEP=y +CONFIG_ACPI_BLACKLIST_YEAR=0 +CONFIG_ACPI_EC=y +CONFIG_ACPI_SYSTEM=y +CONFIG_PCI=y +CONFIG_PCI_GOANY=y +CONFIG_PCI_DIRECT=y +CONFIG_BINFMT_ELF=y +CONFIG_STANDALONE=y +CONFIG_BLK_DEV_LOOP=y +CONFIG_IDE=y +CONFIG_IDE_MAX_HWIFS=2 +CONFIG_BLK_DEV_IDE=y +CONFIG_BLK_DEV_IDEDISK=y +CONFIG_IDEDISK_MULTI_MODE=y +CONFIG_BLK_DEV_IDECD=y +CONFIG_IDE_GENERIC=y +CONFIG_INPUT_MOUSEDEV=y +CONFIG_INPUT_MOUSEDEV_SCREEN_X=1024 +CONFIG_INPUT_KEYBOARD=y +CONFIG_KEYBOARD_ATKBD=y +CONFIG_SERIO=y +CONFIG_VT=y +CONFIG_VT_CONSOLE=y +CONFIG_UNIX98_PTYS=y +CONFIG_VGA_CONSOLE=y +CONFIG_USB_ARCH_HAS_HCD=y +CONFIG_USB_ARCH_HAS_EHCI=y +CONFIG_EXT2_FS=y +CONFIG_DNOTIFY=y +CONFIG_ISO9660_FS=y +CONFIG_FAT_FS=y +CONFIG_VFAT_FS=y +CONFIG_FAT_DEFAULT_IOCHARSET="iso8859-1" +CONFIG_PROC_FS=y +CONFIG_PROC_SYSCTL=y +CONFIG_SYSFS=y +CONFIG_RAMFS=y +CONFIG_SQUASHFS=y +CONFIG_SQUASHFS_FRAGMENT_CACHE_SIZE=3 +CONFIG_MSDOS_PARTITION=y +CONFIG_NLS_DEFAULT="iso8859-1" +CONFIG_AUFS=y +CONFIG_AUFS_FAKE_DM=y +CONFIG_EARLY_PRINTK=y +CONFIG_DOUBLEFAULT=y +CONFIG_ZLIB_INFLATE=y +CONFIG_HAS_IOPORT=y +CONFIG_GENERIC_IRQ_PROBE=y +CONFIG_KTIME_SCALAR=y diff --git a/Makefile b/Makefile index d970cb1..a369204 100644 --- a/Makefile +++ b/Makefile @@ -188,7 +188,7 @@ CROSS_COMPILE ?= # Architecture as present in compile.h UTS_MACHINE := $(ARCH) -KCONFIG_CONFIG ?= .config +KCONFIG_CONFIG ?= .miniconfig # SHELL used by kbuild CONFIG_SHELL := $(shell if [ -x "$$BASH" ]; then echo $$BASH; \ diff --git a/arch/i386/boot/compressed/LzmaDecode.c b/arch/i386/boot/compressed/LzmaDecode.c new file mode 100644 index 0000000..21bf40b --- /dev/null +++ b/arch/i386/boot/compressed/LzmaDecode.c @@ -0,0 +1,588 @@ +/* + LzmaDecode.c + LZMA Decoder (optimized for Speed version) + + LZMA SDK 4.22 Copyright (c) 1999-2005 Igor Pavlov (2005-06-10) + http://www.7-zip.org/ + + LZMA SDK is licensed under two licenses: + 1) GNU Lesser General Public License (GNU LGPL) + 2) Common Public License (CPL) + It means that you can select one of these two licenses and + follow rules of that license. + + SPECIAL EXCEPTION: + Igor Pavlov, as the author of this Code, expressly permits you to + statically or dynamically link your Code (or bind by name) to the + interfaces of this file without subjecting your linked Code to the + terms of the CPL or GNU LGPL. Any modifications or additions + to this file, however, are subject to the LGPL or CPL terms. +*/ + +#include "LzmaDecode.h" + +#ifndef Byte +#define Byte unsigned char +#endif + +#define kNumTopBits 24 +#define kTopValue ((UInt32)1 << kNumTopBits) + +#define kNumBitModelTotalBits 11 +#define kBitModelTotal (1 << kNumBitModelTotalBits) +#define kNumMoveBits 5 + +#define RC_READ_BYTE (*Buffer++) + +#define RC_INIT2 Code = 0; Range = 0xFFFFFFFF; \ + { int i; for(i = 0; i < 5; i++) { RC_TEST; Code = (Code << 8) | RC_READ_BYTE; }} + +#ifdef _LZMA_IN_CB + +#define RC_TEST { if (Buffer == BufferLim) \ + { SizeT size; int result = InCallback->Read(InCallback, &Buffer, &size); if (result != LZMA_RESULT_OK) return result; \ + BufferLim = Buffer + size; if (size == 0) return LZMA_RESULT_DATA_ERROR; }} + +#define RC_INIT Buffer = BufferLim = 0; RC_INIT2 + +#else + +#define RC_TEST { if (Buffer == BufferLim) return LZMA_RESULT_DATA_ERROR; } + +#define RC_INIT(buffer, bufferSize) Buffer = buffer; BufferLim = buffer + bufferSize; RC_INIT2 + +#endif + +#define RC_NORMALIZE if (Range < kTopValue) { RC_TEST; Range <<= 8; Code = (Code << 8) | RC_READ_BYTE; } + +#define IfBit0(p) RC_NORMALIZE; bound = (Range >> kNumBitModelTotalBits) * *(p); if (Code < bound) +#define UpdateBit0(p) Range = bound; *(p) += (kBitModelTotal - *(p)) >> kNumMoveBits; +#define UpdateBit1(p) Range -= bound; Code -= bound; *(p) -= (*(p)) >> kNumMoveBits; + +#define RC_GET_BIT2(p, mi, A0, A1) IfBit0(p) \ + { UpdateBit0(p); mi <<= 1; A0; } else \ + { UpdateBit1(p); mi = (mi + mi) + 1; A1; } + +#define RC_GET_BIT(p, mi) RC_GET_BIT2(p, mi, ; , ;) + +#define RangeDecoderBitTreeDecode(probs, numLevels, res) \ + { int i = numLevels; res = 1; \ + do { CProb *p = probs + res; RC_GET_BIT(p, res) } while(--i != 0); \ + res -= (1 << numLevels); } + + +#define kNumPosBitsMax 4 +#define kNumPosStatesMax (1 << kNumPosBitsMax) + +#define kLenNumLowBits 3 +#define kLenNumLowSymbols (1 << kLenNumLowBits) +#define kLenNumMidBits 3 +#define kLenNumMidSymbols (1 << kLenNumMidBits) +#define kLenNumHighBits 8 +#define kLenNumHighSymbols (1 << kLenNumHighBits) + +#define LenChoice 0 +#define LenChoice2 (LenChoice + 1) +#define LenLow (LenChoice2 + 1) +#define LenMid (LenLow + (kNumPosStatesMax << kLenNumLowBits)) +#define LenHigh (LenMid + (kNumPosStatesMax << kLenNumMidBits)) +#define kNumLenProbs (LenHigh + kLenNumHighSymbols) + + +#define kNumStates 12 +#define kNumLitStates 7 + +#define kStartPosModelIndex 4 +#define kEndPosModelIndex 14 +#define kNumFullDistances (1 << (kEndPosModelIndex >> 1)) + +#define kNumPosSlotBits 6 +#define kNumLenToPosStates 4 + +#define kNumAlignBits 4 +#define kAlignTableSize (1 << kNumAlignBits) + +#define kMatchMinLen 2 + +#define IsMatch 0 +#define IsRep (IsMatch + (kNumStates << kNumPosBitsMax)) +#define IsRepG0 (IsRep + kNumStates) +#define IsRepG1 (IsRepG0 + kNumStates) +#define IsRepG2 (IsRepG1 + kNumStates) +#define IsRep0Long (IsRepG2 + kNumStates) +#define PosSlot (IsRep0Long + (kNumStates << kNumPosBitsMax)) +#define SpecPos (PosSlot + (kNumLenToPosStates << kNumPosSlotBits)) +#define Align (SpecPos + kNumFullDistances - kEndPosModelIndex) +#define LenCoder (Align + kAlignTableSize) +#define RepLenCoder (LenCoder + kNumLenProbs) +#define Literal (RepLenCoder + kNumLenProbs) + +#if Literal != LZMA_BASE_SIZE +StopCompilingDueBUG +#endif + +int LzmaDecodeProperties(CLzmaProperties *propsRes, const unsigned char *propsData, int size) +{ + unsigned char prop0; + if (size < LZMA_PROPERTIES_SIZE) + return LZMA_RESULT_DATA_ERROR; + prop0 = propsData[0]; + if (prop0 >= (9 * 5 * 5)) + return LZMA_RESULT_DATA_ERROR; + { + for (propsRes->pb = 0; prop0 >= (9 * 5); propsRes->pb++, prop0 -= (9 * 5)); + for (propsRes->lp = 0; prop0 >= 9; propsRes->lp++, prop0 -= 9); + propsRes->lc = prop0; + /* + unsigned char remainder = (unsigned char)(prop0 / 9); + propsRes->lc = prop0 % 9; + propsRes->pb = remainder / 5; + propsRes->lp = remainder % 5; + */ + } + + #ifdef _LZMA_OUT_READ + { + int i; + propsRes->DictionarySize = 0; + for (i = 0; i < 4; i++) + propsRes->DictionarySize += (UInt32)(propsData[1 + i]) << (i * 8); + if (propsRes->DictionarySize == 0) + propsRes->DictionarySize = 1; + } + #endif + return LZMA_RESULT_OK; +} + +#define kLzmaStreamWasFinishedId (-1) + +int LzmaDecode(CLzmaDecoderState *vs, + #ifdef _LZMA_IN_CB + ILzmaInCallback *InCallback, + #else + const unsigned char *inStream, SizeT inSize, SizeT *inSizeProcessed, + #endif + unsigned char *outStream, SizeT outSize, SizeT *outSizeProcessed) +{ + CProb *p = vs->Probs; + SizeT nowPos = 0; + Byte previousByte = 0; + UInt32 posStateMask = (1 << (vs->Properties.pb)) - 1; + UInt32 literalPosMask = (1 << (vs->Properties.lp)) - 1; + int lc = vs->Properties.lc; + + #ifdef _LZMA_OUT_READ + + UInt32 Range = vs->Range; + UInt32 Code = vs->Code; + #ifdef _LZMA_IN_CB + const Byte *Buffer = vs->Buffer; + const Byte *BufferLim = vs->BufferLim; + #else + const Byte *Buffer = inStream; + const Byte *BufferLim = inStream + inSize; + #endif + int state = vs->State; + UInt32 rep0 = vs->Reps[0], rep1 = vs->Reps[1], rep2 = vs->Reps[2], rep3 = vs->Reps[3]; + int len = vs->RemainLen; + UInt32 globalPos = vs->GlobalPos; + UInt32 distanceLimit = vs->DistanceLimit; + + Byte *dictionary = vs->Dictionary; + UInt32 dictionarySize = vs->Properties.DictionarySize; + UInt32 dictionaryPos = vs->DictionaryPos; + + Byte tempDictionary[4]; + + #ifndef _LZMA_IN_CB + *inSizeProcessed = 0; + #endif + *outSizeProcessed = 0; + if (len == kLzmaStreamWasFinishedId) + return LZMA_RESULT_OK; + + if (dictionarySize == 0) + { + dictionary = tempDictionary; + dictionarySize = 1; + tempDictionary[0] = vs->TempDictionary[0]; + } + + if (len == kLzmaNeedInitId) + { + { + UInt32 numProbs = Literal + ((UInt32)LZMA_LIT_SIZE << (lc + vs->Properties.lp)); + UInt32 i; + for (i = 0; i < numProbs; i++) + p[i] = kBitModelTotal >> 1; + rep0 = rep1 = rep2 = rep3 = 1; + state = 0; + globalPos = 0; + distanceLimit = 0; + dictionaryPos = 0; + dictionary[dictionarySize - 1] = 0; + #ifdef _LZMA_IN_CB + RC_INIT; + #else + RC_INIT(inStream, inSize); + #endif + } + len = 0; + } + while(len != 0 && nowPos < outSize) + { + UInt32 pos = dictionaryPos - rep0; + if (pos >= dictionarySize) + pos += dictionarySize; + outStream[nowPos++] = dictionary[dictionaryPos] = dictionary[pos]; + if (++dictionaryPos == dictionarySize) + dictionaryPos = 0; + len--; + } + if (dictionaryPos == 0) + previousByte = dictionary[dictionarySize - 1]; + else + previousByte = dictionary[dictionaryPos - 1]; + + #else /* if !_LZMA_OUT_READ */ + + int state = 0; + UInt32 rep0 = 1, rep1 = 1, rep2 = 1, rep3 = 1; + int len = 0; + const Byte *Buffer; + const Byte *BufferLim; + UInt32 Range; + UInt32 Code; + + #ifndef _LZMA_IN_CB + *inSizeProcessed = 0; + #endif + *outSizeProcessed = 0; + + { + UInt32 i; + UInt32 numProbs = Literal + ((UInt32)LZMA_LIT_SIZE << (lc + vs->Properties.lp)); + for (i = 0; i < numProbs; i++) + p[i] = kBitModelTotal >> 1; + } + + #ifdef _LZMA_IN_CB + RC_INIT; + #else + RC_INIT(inStream, inSize); + #endif + + #endif /* _LZMA_OUT_READ */ + + while(nowPos < outSize) + { + CProb *prob; + UInt32 bound; + int posState = (int)( + (nowPos + #ifdef _LZMA_OUT_READ + + globalPos + #endif + ) + & posStateMask); + + prob = p + IsMatch + (state << kNumPosBitsMax) + posState; + IfBit0(prob) + { + int symbol = 1; + UpdateBit0(prob) + prob = p + Literal + (LZMA_LIT_SIZE * + ((( + (nowPos + #ifdef _LZMA_OUT_READ + + globalPos + #endif + ) + & literalPosMask) << lc) + (previousByte >> (8 - lc)))); + + if (state >= kNumLitStates) + { + int matchByte; + #ifdef _LZMA_OUT_READ + UInt32 pos = dictionaryPos - rep0; + if (pos >= dictionarySize) + pos += dictionarySize; + matchByte = dictionary[pos]; + #else + matchByte = outStream[nowPos - rep0]; + #endif + do + { + int bit; + CProb *probLit; + matchByte <<= 1; + bit = (matchByte & 0x100); + probLit = prob + 0x100 + bit + symbol; + RC_GET_BIT2(probLit, symbol, if (bit != 0) break, if (bit == 0) break) + } + while (symbol < 0x100); + } + while (symbol < 0x100) + { + CProb *probLit = prob + symbol; + RC_GET_BIT(probLit, symbol) + } + previousByte = (Byte)symbol; + + outStream[nowPos++] = previousByte; + #ifdef _LZMA_OUT_READ + if (distanceLimit < dictionarySize) + distanceLimit++; + + dictionary[dictionaryPos] = previousByte; + if (++dictionaryPos == dictionarySize) + dictionaryPos = 0; + #endif + if (state < 4) state = 0; + else if (state < 10) state -= 3; + else state -= 6; + } + else + { + UpdateBit1(prob); + prob = p + IsRep + state; + IfBit0(prob) + { + UpdateBit0(prob); + rep3 = rep2; + rep2 = rep1; + rep1 = rep0; + state = state < kNumLitStates ? 0 : 3; + prob = p + LenCoder; + } + else + { + UpdateBit1(prob); + prob = p + IsRepG0 + state; + IfBit0(prob) + { + UpdateBit0(prob); + prob = p + IsRep0Long + (state << kNumPosBitsMax) + posState; + IfBit0(prob) + { + #ifdef _LZMA_OUT_READ + UInt32 pos; + #endif + UpdateBit0(prob); + + #ifdef _LZMA_OUT_READ + if (distanceLimit == 0) + #else + if (nowPos == 0) + #endif + return LZMA_RESULT_DATA_ERROR; + + state = state < kNumLitStates ? 9 : 11; + #ifdef _LZMA_OUT_READ + pos = dictionaryPos - rep0; + if (pos >= dictionarySize) + pos += dictionarySize; + previousByte = dictionary[pos]; + dictionary[dictionaryPos] = previousByte; + if (++dictionaryPos == dictionarySize) + dictionaryPos = 0; + #else + previousByte = outStream[nowPos - rep0]; + #endif + outStream[nowPos++] = previousByte; + #ifdef _LZMA_OUT_READ + if (distanceLimit < dictionarySize) + distanceLimit++; + #endif + + continue; + } + else + { + UpdateBit1(prob); + } + } + else + { + UInt32 distance; + UpdateBit1(prob); + prob = p + IsRepG1 + state; + IfBit0(prob) + { + UpdateBit0(prob); + distance = rep1; + } + else + { + UpdateBit1(prob); + prob = p + IsRepG2 + state; + IfBit0(prob) + { + UpdateBit0(prob); + distance = rep2; + } + else + { + UpdateBit1(prob); + distance = rep3; + rep3 = rep2; + } + rep2 = rep1; + } + rep1 = rep0; + rep0 = distance; + } + state = state < kNumLitStates ? 8 : 11; + prob = p + RepLenCoder; + } + { + int numBits, offset; + CProb *probLen = prob + LenChoice; + IfBit0(probLen) + { + UpdateBit0(probLen); + probLen = prob + LenLow + (posState << kLenNumLowBits); + offset = 0; + numBits = kLenNumLowBits; + } + else + { + UpdateBit1(probLen); + probLen = prob + LenChoice2; + IfBit0(probLen) + { + UpdateBit0(probLen); + probLen = prob + LenMid + (posState << kLenNumMidBits); + offset = kLenNumLowSymbols; + numBits = kLenNumMidBits; + } + else + { + UpdateBit1(probLen); + probLen = prob + LenHigh; + offset = kLenNumLowSymbols + kLenNumMidSymbols; + numBits = kLenNumHighBits; + } + } + RangeDecoderBitTreeDecode(probLen, numBits, len); + len += offset; + } + + if (state < 4) + { + int posSlot; + state += kNumLitStates; + prob = p + PosSlot + + ((len < kNumLenToPosStates ? len : kNumLenToPosStates - 1) << + kNumPosSlotBits); + RangeDecoderBitTreeDecode(prob, kNumPosSlotBits, posSlot); + if (posSlot >= kStartPosModelIndex) + { + int numDirectBits = ((posSlot >> 1) - 1); + rep0 = (2 | ((UInt32)posSlot & 1)); + if (posSlot < kEndPosModelIndex) + { + rep0 <<= numDirectBits; + prob = p + SpecPos + rep0 - posSlot - 1; + } + else + { + numDirectBits -= kNumAlignBits; + do + { + RC_NORMALIZE + Range >>= 1; + rep0 <<= 1; + if (Code >= Range) + { + Code -= Range; + rep0 |= 1; + } + } + while (--numDirectBits != 0); + prob = p + Align; + rep0 <<= kNumAlignBits; + numDirectBits = kNumAlignBits; + } + { + int i = 1; + int mi = 1; + do + { + CProb *prob3 = prob + mi; + RC_GET_BIT2(prob3, mi, ; , rep0 |= i); + i <<= 1; + } + while(--numDirectBits != 0); + } + } + else + rep0 = posSlot; + if (++rep0 == (UInt32)(0)) + { + /* it's for stream version */ + len = kLzmaStreamWasFinishedId; + break; + } + } + + len += kMatchMinLen; + #ifdef _LZMA_OUT_READ + if (rep0 > distanceLimit) + #else + if (rep0 > nowPos) + #endif + return LZMA_RESULT_DATA_ERROR; + + #ifdef _LZMA_OUT_READ + if (dictionarySize - distanceLimit > (UInt32)len) + distanceLimit += len; + else + distanceLimit = dictionarySize; + #endif + + do + { + #ifdef _LZMA_OUT_READ + UInt32 pos = dictionaryPos - rep0; + if (pos >= dictionarySize) + pos += dictionarySize; + previousByte = dictionary[pos]; + dictionary[dictionaryPos] = previousByte; + if (++dictionaryPos == dictionarySize) + dictionaryPos = 0; + #else + previousByte = outStream[nowPos - rep0]; + #endif + len--; + outStream[nowPos++] = previousByte; + } + while(len != 0 && nowPos < outSize); + } + } + RC_NORMALIZE; + + #ifdef _LZMA_OUT_READ + vs->Range = Range; + vs->Code = Code; + vs->DictionaryPos = dictionaryPos; + vs->GlobalPos = globalPos + (UInt32)nowPos; + vs->DistanceLimit = distanceLimit; + vs->Reps[0] = rep0; + vs->Reps[1] = rep1; + vs->Reps[2] = rep2; + vs->Reps[3] = rep3; + vs->State = state; + vs->RemainLen = len; + vs->TempDictionary[0] = tempDictionary[0]; + #endif + + #ifdef _LZMA_IN_CB + vs->Buffer = Buffer; + vs->BufferLim = BufferLim; + #else + *inSizeProcessed = (SizeT)(Buffer - inStream); + #endif + *outSizeProcessed = nowPos; + return LZMA_RESULT_OK; +} diff --git a/arch/i386/boot/compressed/LzmaDecode.h b/arch/i386/boot/compressed/LzmaDecode.h new file mode 100644 index 0000000..213062a --- /dev/null +++ b/arch/i386/boot/compressed/LzmaDecode.h @@ -0,0 +1,131 @@ +/* + LzmaDecode.h + LZMA Decoder interface + + LZMA SDK 4.21 Copyright (c) 1999-2005 Igor Pavlov (2005-06-08) + http://www.7-zip.org/ + + LZMA SDK is licensed under two licenses: + 1) GNU Lesser General Public License (GNU LGPL) + 2) Common Public License (CPL) + It means that you can select one of these two licenses and + follow rules of that license. + + SPECIAL EXCEPTION: + Igor Pavlov, as the author of this code, expressly permits you to + statically or dynamically link your code (or bind by name) to the + interfaces of this file without subjecting your linked code to the + terms of the CPL or GNU LGPL. Any modifications or additions + to this file, however, are subject to the LGPL or CPL terms. +*/ + +#ifndef __LZMADECODE_H +#define __LZMADECODE_H + +/* #define _LZMA_IN_CB */ +/* Use callback for input data */ + +/* #define _LZMA_OUT_READ */ +/* Use read function for output data */ + +/* #define _LZMA_PROB32 */ +/* It can increase speed on some 32-bit CPUs, + but memory usage will be doubled in that case */ + +/* #define _LZMA_LOC_OPT */ +/* Enable local speed optimizations inside code */ + +/* #define _LZMA_SYSTEM_SIZE_T */ +/* Use system's size_t. You can use it to enable 64-bit sizes supporting*/ + +#ifndef UInt32 +#ifdef _LZMA_UINT32_IS_ULONG +#define UInt32 unsigned long +#else +#define UInt32 unsigned int +#endif +#endif + +#ifndef SizeT +#ifdef _LZMA_SYSTEM_SIZE_T +#include +#define SizeT size_t +#else +#define SizeT UInt32 +#endif +#endif + +#ifdef _LZMA_PROB32 +#define CProb UInt32 +#else +#define CProb unsigned short +#endif + +#define LZMA_RESULT_OK 0 +#define LZMA_RESULT_DATA_ERROR 1 + +#ifdef _LZMA_IN_CB +typedef struct _ILzmaInCallback +{ + int (*Read)(void *object, const unsigned char **buffer, SizeT *bufferSize); +} ILzmaInCallback; +#endif + +#define LZMA_BASE_SIZE 1846 +#define LZMA_LIT_SIZE 768 + +#define LZMA_PROPERTIES_SIZE 5 + +typedef struct _CLzmaProperties +{ + int lc; + int lp; + int pb; + #ifdef _LZMA_OUT_READ + UInt32 DictionarySize; + #endif +}CLzmaProperties; + +int LzmaDecodeProperties(CLzmaProperties *propsRes, const unsigned char *propsData, int size); + +#define LzmaGetNumProbs(Properties) (LZMA_BASE_SIZE + (LZMA_LIT_SIZE << ((Properties)->lc + (Properties)->lp))) + +#define kLzmaNeedInitId (-2) + +typedef struct _CLzmaDecoderState +{ + CLzmaProperties Properties; + CProb *Probs; + + #ifdef _LZMA_IN_CB + const unsigned char *Buffer; + const unsigned char *BufferLim; + #endif + + #ifdef _LZMA_OUT_READ + unsigned char *Dictionary; + UInt32 Range; + UInt32 Code; + UInt32 DictionaryPos; + UInt32 GlobalPos; + UInt32 DistanceLimit; + UInt32 Reps[4]; + int State; + int RemainLen; + unsigned char TempDictionary[4]; + #endif +} CLzmaDecoderState; + +#ifdef _LZMA_OUT_READ +#define LzmaDecoderInit(vs) { (vs)->RemainLen = kLzmaNeedInitId; } +#endif + +int LzmaDecode(CLzmaDecoderState *vs, + #ifdef _LZMA_IN_CB + ILzmaInCallback *inCallback, + #else + const unsigned char *inStream, SizeT inSize, SizeT *inSizeProcessed, + #endif + unsigned char *outStream, SizeT outSize, SizeT *outSizeProcessed); + +#endif diff --git a/arch/i386/boot/compressed/Makefile b/arch/i386/boot/compressed/Makefile index a661217..fb40869 100644 --- a/arch/i386/boot/compressed/Makefile +++ b/arch/i386/boot/compressed/Makefile @@ -4,15 +4,16 @@ # create a compressed vmlinux image from the original vmlinux # -targets := vmlinux vmlinux.bin vmlinux.bin.gz head.o misc.o piggy.o \ - vmlinux.bin.all vmlinux.relocs +tragets := head.o lzma_misc.o piggy.o \ + vmlinux.bin.all vmlinux.relocs \ + vmlinux vmlinux.bin vmlinux.bin.gz EXTRA_AFLAGS := -traditional LDFLAGS_vmlinux := -T -CFLAGS_misc.o += -fPIC +CFLAGS_lzma_misc.o += -fPIC hostprogs-y := relocs -$(obj)/vmlinux: $(src)/vmlinux.lds $(obj)/head.o $(obj)/misc.o $(obj)/piggy.o FORCE +$(obj)/vmlinux: $(src)/vmlinux.lds $(obj)/head.o $(obj)/lzma_misc.o $(obj)/piggy.o FORCE $(call if_changed,ld) @: @@ -33,10 +34,10 @@ $(obj)/vmlinux.bin.all: $(vmlinux.bin.all-y) FORCE ifdef CONFIG_RELOCATABLE $(obj)/vmlinux.bin.gz: $(obj)/vmlinux.bin.all FORCE - $(call if_changed,gzip) + $(call if_changed,lzma) else $(obj)/vmlinux.bin.gz: $(obj)/vmlinux.bin FORCE - $(call if_changed,gzip) + $(call if_changed,lzma) endif LDFLAGS_piggy.o := -r --format binary --oformat elf32-i386 -T diff --git a/arch/i386/boot/compressed/lzma_misc.c b/arch/i386/boot/compressed/lzma_misc.c new file mode 100644 index 0000000..4f5f7f9 --- /dev/null +++ b/arch/i386/boot/compressed/lzma_misc.c @@ -0,0 +1,290 @@ +/* + * lzma_misc.c + * + * Decompress LZMA compressed vmlinuz + * Version 0.9 Copyright (c) Ming-Ching Tiew mctiew@yahoo.com + * Program adapted from misc.c for 2.6.20.1 kernel + * Please refer to misc.c for authorship and copyright. + * Date: 25 March 2007 + * Source released under GPL + */ + +#undef CONFIG_PARAVIRT +#include +#include +#include +#include +#include +#include + +/* WARNING!! + * This code is compiled with -fPIC and it is relocated dynamically + * at run time, but no relocation processing is performed. + * This means that it is not safe to place pointers in static structures. + */ + +#define OF(args) args +#define STATIC static + +#undef memset +#undef memcpy + +typedef unsigned char uch; +typedef unsigned short ush; +typedef unsigned long ulg; + +#define WSIZE 0x80000000 /* Window size must be at least 32k, + * and a power of two + * We don't actually have a window just + * a huge output buffer so I report + * a 2G windows size, as that should + * always be larger than our output buffer. + */ + +static uch *inbuf; /* input buffer */ +static uch *window; /* Sliding window buffer, (and final output buffer) */ + +static unsigned insize; /* valid bytes in inbuf */ +static unsigned inptr; /* index of next byte to be processed in inbuf */ + +/* gzip flag byte */ +#define ASCII_FLAG 0x01 /* bit 0 set: file probably ASCII text */ +#define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */ +#define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */ +#define ORIG_NAME 0x08 /* bit 3 set: original file name present */ +#define COMMENT 0x10 /* bit 4 set: file comment present */ +#define ENCRYPTED 0x20 /* bit 5 set: file is encrypted */ +#define RESERVED 0xC0 /* bit 6,7: reserved */ + +#define get_byte() (inptr < insize ? inbuf[inptr++] : fill_inbuf()) + +/* Diagnostic functions */ +#ifdef DEBUG +# define Assert(cond,msg) {if(!(cond)) error(msg);} +# define Trace(x) fprintf x +# define Tracev(x) {if (verbose) fprintf x ;} +# define Tracevv(x) {if (verbose>1) fprintf x ;} +# define Tracec(c,x) {if (verbose && (c)) fprintf x ;} +# define Tracecv(c,x) {if (verbose>1 && (c)) fprintf x ;} +#else +# define Assert(cond,msg) +# define Trace(x) +# define Tracev(x) +# define Tracevv(x) +# define Tracec(c,x) +# define Tracecv(c,x) +#endif + +static int fill_inbuf(void); +static void error(char *m); + +/* + * This is set up by the setup-routine at boot-time + */ +static unsigned char *real_mode; /* Pointer to real-mode data */ + +#define RM_EXT_MEM_K (*(unsigned short *)(real_mode + 0x2)) +#ifndef STANDARD_MEMORY_BIOS_CALL +#define RM_ALT_MEM_K (*(unsigned long *)(real_mode + 0x1e0)) +#endif +#define RM_SCREEN_INFO (*(struct screen_info *)(real_mode+0)) + +extern unsigned char input_data[]; +extern int input_len; + +static long bytes_out = 0; + +static void *memcpy(void *dest, const void *src, unsigned n); + +static void putstr(const char *); + +static unsigned long free_mem_ptr; +static unsigned long free_mem_end_ptr; + +#define HEAP_SIZE 0x3000 + +static char *vidmem = (char *)0xb8000; +static int vidport; +static int lines, cols; + +#ifdef CONFIG_X86_NUMAQ +void *xquad_portio; +#endif + +static void scroll(void) +{ + int i; + + memcpy ( vidmem, vidmem + cols * 2, ( lines - 1 ) * cols * 2 ); + for ( i = ( lines - 1 ) * cols * 2; i < lines * cols * 2; i += 2 ) + vidmem[i] = ' '; +} + +static void putstr(const char *s) +{ + int x,y,pos; + char c; + + x = RM_SCREEN_INFO.orig_x; + y = RM_SCREEN_INFO.orig_y; + + while ( ( c = *s++ ) != '\0' ) { + if ( c == '\n' ) { + x = 0; + if ( ++y >= lines ) { + scroll(); + y--; + } + } else { + vidmem [ ( x + cols * y ) * 2 ] = c; + if ( ++x >= cols ) { + x = 0; + if ( ++y >= lines ) { + scroll(); + y--; + } + } + } + } + + RM_SCREEN_INFO.orig_x = x; + RM_SCREEN_INFO.orig_y = y; + + pos = (x + cols * y) * 2; /* Update cursor position */ + outb_p(14, vidport); + outb_p(0xff & (pos >> 9), vidport+1); + outb_p(15, vidport); + outb_p(0xff & (pos >> 1), vidport+1); +} + +static void* memcpy(void* dest, const void* src, unsigned n) +{ + int i; + char *d = (char *)dest, *s = (char *)src; + + for (i=0;i ((-__PAGE_OFFSET-(512 <<20)-1) & 0x7fffffff)) + error("Destination address too large"); +#ifndef CONFIG_RELOCATABLE + if ((u32)output != LOAD_PHYSICAL_ADDR) + error("Wrong destination address"); +#endif + if( lzma_unzip(output) != 0 ) + { + error("inflate error\n"); + } + putstr("Ok, booting the kernel.\n"); + + return; +} diff --git a/arch/i386/boot/compressed/vmlinux.scr b/arch/i386/boot/compressed/vmlinux.scr index 707a88f..9d67263 100644 --- a/arch/i386/boot/compressed/vmlinux.scr +++ b/arch/i386/boot/compressed/vmlinux.scr @@ -3,8 +3,8 @@ SECTIONS .data.compressed : { input_len = .; LONG(input_data_end - input_data) input_data = .; + output_len = . + 5; *(.data) - output_len = . - 4; input_data_end = .; } } diff --git a/drivers/block/Kconfig b/drivers/block/Kconfig index 17ee97f..64b7bda 100644 --- a/drivers/block/Kconfig +++ b/drivers/block/Kconfig @@ -406,6 +406,47 @@ config BLK_DEV_RAM_BLOCKSIZE setups function - apparently needed by the rd_load_image routine that supposes the filesystem in the image uses a 1024 blocksize. +config LZMA_INITRD + boolean "Allow LZMA compression on initrd" + depends on BLK_DEV_INITRD=y + default "y" + help + Use lzma compression on initrd, example 'lzma e initrd initrd.7z -d16'. + If you have sufficient memory, you could compress using bigger dictionary size, + 'lzma e initrd initrd.7z'. + +config LZMA_INITRD_KMALLOC_ONLY + boolean "Use only kmalloc, do not use vmalloc on lzma initrd" + depends on LZMA_INITRD=y + default "n" + help + Set to y if you do not want to use vmalloc, ie use only kmalloc. + +config LZMA_INITRAM_FS + boolean "Allow LZMA compression on initramfs" + depends on BLK_DEV_RAM=y + default "y" + help + Use lzma compression on initramfs, example 'lzma e initramfs.cpio initramfs.cpio.lzma'. + +config LZMA_INITRAM_FS_SMALLMEM + boolean "Use lzma compression with small dictonary size." + depends on LZMA_INITRAM_FS=y + default "y" + help + Use lzma compression on initramfs with small dictionary size, example + 'lzma e initramfs.cpio initramfs.cpio.lzma -d16'. + Affects only the initramfs.cpio in the ~usr directory, which is compiled into + the kernel. If you prepared initramfs.cpio for use with bootloader, you would + need to specify the commandline options (-d16) yourself. + +config LZMA_INITRAM_FS_KMALLOC_ONLY + boolean "Use only kmalloc, do not use vmalloc on lzma initramfs" + depends on LZMA_INITRAM_FS=y + default "n" + help + Set to y if you do not want to use vmalloc, ie use only kmalloc. + config CDROM_PKTCDVD tristate "Packet writing on CD/DVD media" depends on !UML diff --git a/fs/Kconfig b/fs/Kconfig index 3c4886b..bdcc6fb 100644 --- a/fs/Kconfig +++ b/fs/Kconfig @@ -1371,6 +1371,71 @@ config CRAMFS If unsure, say N. +config SQUASHFS + tristate "SquashFS 3.2 - Squashed file system support" + select ZLIB_INFLATE + help + Saying Y here includes support for SquashFS 3.2 (a Compressed Read-Only File + System). Squashfs is a highly compressed read-only filesystem for Linux. + It uses zlib compression to compress both files, inodes and directories. + Inodes in the system are very small and all blocks are packed to minimise + data overhead. Block sizes greater than 4K are supported up to a maximum of 64K. + SquashFS 3.1 supports 64 bit filesystems and files (larger than 4GB), full + uid/gid information, hard links and timestamps. + + Squashfs is intended for general read-only filesystem use, for archival + use (i.e. in cases where a .tar.gz file may be used), and in embedded + systems where low overhead is needed. Further information and filesystem tools + are available from http://squashfs.sourceforge.net. + + If you want to compile this as a module ( = code which can be + inserted in and removed from the running kernel whenever you want), + say M here and read . The module + will be called squashfs. Note that the root file system (the one + containing the directory /) cannot be compiled as a module. + + If unsure, say N. + +config SQUASHFS_EMBEDDED + + bool "Additional options for memory-constrained systems" + depends on SQUASHFS + default n + help + Saying Y here allows you to specify cache sizes and how Squashfs + allocates memory. This is only intended for memory constrained + systems. + + If unsure, say N. + +config SQUASHFS_FRAGMENT_CACHE_SIZE + int "Number of fragments cached" if SQUASHFS_EMBEDDED + depends on SQUASHFS + default "3" + help + By default SquashFS caches the last 3 fragments read from + the filesystem. Increasing this amount may mean SquashFS + has to re-read fragments less often from disk, at the expense + of extra system memory. Decreasing this amount will mean + SquashFS uses less memory at the expense of extra reads from disk. + + Note there must be at least one cached fragment. Anything + much more than three will probably not make much difference. + +config SQUASHFS_VMALLOC + bool "Use Vmalloc rather than Kmalloc" if SQUASHFS_EMBEDDED + depends on SQUASHFS + default n + help + By default SquashFS uses kmalloc to obtain fragment cache memory. + Kmalloc memory is the standard kernel allocator, but it can fail + on memory constrained systems. Because of the way Vmalloc works, + Vmalloc can succeed when kmalloc fails. Specifying this option + will make SquashFS always use Vmalloc to allocate the + fragment cache memory. + + If unsure, say N. + config VXFS_FS tristate "FreeVxFS file system support (VERITAS VxFS(TM) compatible)" depends on BLOCK @@ -2057,3 +2122,4 @@ source "fs/dlm/Kconfig" endmenu +source "fs/aufs/Kconfig" diff --git a/fs/Makefile b/fs/Makefile index 9edf411..557766f 100644 --- a/fs/Makefile +++ b/fs/Makefile @@ -68,6 +68,7 @@ obj-$(CONFIG_JBD) += jbd/ obj-$(CONFIG_JBD2) += jbd2/ obj-$(CONFIG_EXT2_FS) += ext2/ obj-$(CONFIG_CRAMFS) += cramfs/ +obj-$(CONFIG_SQUASHFS) += squashfs/ obj-$(CONFIG_RAMFS) += ramfs/ obj-$(CONFIG_HUGETLBFS) += hugetlbfs/ obj-$(CONFIG_CODA_FS) += coda/ @@ -114,3 +115,4 @@ obj-$(CONFIG_HPPFS) += hppfs/ obj-$(CONFIG_DEBUG_FS) += debugfs/ obj-$(CONFIG_OCFS2_FS) += ocfs2/ obj-$(CONFIG_GFS2_FS) += gfs2/ +obj-$(CONFIG_AUFS) += aufs/ diff --git a/fs/aufs/Kconfig b/fs/aufs/Kconfig new file mode 100644 index 0000000..3a2121c --- /dev/null +++ b/fs/aufs/Kconfig @@ -0,0 +1,73 @@ +config AUFS + tristate "Another unionfs" + help + Aufs is a stackable unification filesystem such as Unionfs, + which unifies several directories and provides a merged single + directory. + In the early days, aufs was entirely re-designed and + re-implemented Unionfs Version 1.x series. After many original + ideas, approaches and improvements, it becomes totally + different from Unionfs while keeping the basic features. + See Unionfs for the basic features. + +if AUFS +comment "These options are generated automatically for "#UTS_RELEASE + +config AUFS_FAKE_DM + bool "Use simplified (fake) nameidata" + depends on AUFS + default y + help + Faking nameidata (VFS internal data), you can get better performance + in some cases. + +choice + prompt "Maximum number of branches" + depends on AUFS + default AUFS_BRANCH_MAX_127 + help + Specifies the maximum number of branches (or member directories) in a single aufs. The larger value consumes more system resources and has an impact to performance. +config AUFS_BRANCH_MAX_127 + bool "127" + help + Specifies the maximum number of branches (or member directories) in a single aufs. The larger value consumes more system resources and has an impact to performance. +config AUFS_BRANCH_MAX_511 + bool "511" + help + Specifies the maximum number of branches (or member directories) in a single aufs. The larger value consumes more system resources and has an impact to performance. +config AUFS_BRANCH_MAX_1023 + bool "1023" + help + Specifies the maximum number of branches (or member directories) in a single aufs. The larger value consumes more system resources and has an impact to performance. + +config AUFS_BRANCH_MAX_32767 + bool "32767" + help + Specifies the maximum number of branches (or member directories) in a single aufs. The larger value consumes more system resources and has an impact to performance. +endchoice +config AUFS_DEBUG + bool "Debug aufs" + depends on AUFS + default y + help + Enable this to compile aufs internal debug code. + The performance will be damaged. + +config AUFS_COMPAT + bool "Compatibility with Unionfs (obsolete)" + depends on AUFS + default n + help + This makes aufs compatible with unionfs-style mount options and some + behaviours. + The dirs= mount option and =nfsro branch permission flag are always + interpreted as br: mount option and =ro flag respectively. The + 'debug', 'delete' and 'imap' mount options are ignored. + If you disable this option, you will get, + - aufs issues a warning about the ignored mount options + - the default branch permission flag is set. RW for the first branch, + and RO for the rests. + - the name of a internal file which represents the directory is + 'opaque', becomes '.wh..wh..opq' + - the 'diropq=w' mount option is set by default +endif diff --git a/fs/aufs/Makefile b/fs/aufs/Makefile new file mode 100755 index 0000000..0ee3cd0 --- /dev/null +++ b/fs/aufs/Makefile @@ -0,0 +1,18 @@ +# AUFS Makefile for the Linux 2.6.16 and later +# $Id: Makefile,v 1.29 2007/04/23 00:59:50 sfjro Exp $ + +obj-$(CONFIG_AUFS) += aufs.o +aufs-y := module.o super.o sbinfo.o xino.o \ + branch.o cpup.o whout.o plink.o wkq.o dcsub.o vfsub.o \ + opts.o \ + dentry.o dinfo.o \ + file.o f_op.o finfo.o \ + dir.o vdir.o \ + inode.o i_op.o i_op_add.o i_op_del.o i_op_ren.o iinfo.o \ + misc.o +#xattr.o +aufs-$(CONFIG_AUFS_SYSAUFS) += sysaufs.o +aufs-$(CONFIG_AUFS_HINOTIFY) += hinotify.o +aufs-$(CONFIG_AUFS_EXPORT) += export.o +#aufs-$(CONFIG_DEBUGFS) += dbgfs.o +aufs-$(CONFIG_AUFS_DEBUG) += debug.o diff --git a/fs/aufs/aufs.h b/fs/aufs/aufs.h new file mode 100755 index 0000000..79b3b87 --- /dev/null +++ b/fs/aufs/aufs.h @@ -0,0 +1,64 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: aufs.h,v 1.24 2007/05/14 03:41:51 sfjro Exp $ */ + +#ifndef __AUFS_H__ +#define __AUFS_H__ + +#ifdef __KERNEL__ + +#include + +/* limited support before 2.6.16, curretly 2.6.15 only. */ +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,16) +#define atomic_long_t atomic_t +#define atomic_long_set atomic_set +#define timespec_to_ns(ts) ({(long long)(ts)->tv_sec;}) +#define D_CHILD d_child +#else +#define D_CHILD d_u.d_child +#endif + +/* ---------------------------------------------------------------------- */ + +#include "debug.h" + +#include "branch.h" +#include "cpup.h" +#include "dcsub.h" +#include "dentry.h" +#include "dir.h" +#include "file.h" +#include "inode.h" +#include "misc.h" +#include "module.h" +#include "opts.h" +#include "super.h" +#include "sysaufs.h" +#include "vfsub.h" +#include "whout.h" +#include "wkq.h" +//#include "xattr.h" + +#if defined(CONFIG_AUFS_MODULE) && !defined(CONFIG_AUFS_KSIZE_PATCH) +#define ksize(p) (-1U) +#endif + +#endif /* __KERNEL__ */ +#endif /* __AUFS_H__ */ diff --git a/fs/aufs/branch.c b/fs/aufs/branch.c new file mode 100755 index 0000000..f1ce008 --- /dev/null +++ b/fs/aufs/branch.c @@ -0,0 +1,818 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: branch.c,v 1.49 2007/05/14 03:38:23 sfjro Exp $ */ + +//#include +//#include +#include "aufs.h" + +static void free_branch(struct aufs_branch *br) +{ + TraceEnter(); + + if (br->br_xino) + fput(br->br_xino); + dput(br->br_wh); + dput(br->br_plink); + mntput(br->br_mnt); + DEBUG_ON(br_count(br) || atomic_read(&br->br_wh_running)); + kfree(br); +} + +/* + * frees all branches + */ +void free_branches(struct aufs_sbinfo *sbinfo) +{ + aufs_bindex_t bmax; + struct aufs_branch **br; + + TraceEnter(); + bmax = sbinfo->si_bend + 1; + br = sbinfo->si_branch; + while (bmax--) + free_branch(*br++); +} + +/* + * find the index of a branch which is specified by @br_id. + */ +int find_brindex(struct super_block *sb, aufs_bindex_t br_id) +{ + aufs_bindex_t bindex, bend; + + TraceEnter(); + + bend = sbend(sb); + for (bindex = 0; bindex <= bend; bindex++) + if (sbr_id(sb, bindex) == br_id) + return bindex; + return -1; +} + +/* + * test if the @br is readonly or not. + */ +int br_rdonly(struct aufs_branch *br) +{ + return ((br->br_mnt->mnt_sb->s_flags & MS_RDONLY) + || !br_writable(br->br_perm)) + ? -EROFS : 0; +} + +/* + * returns writable branch index, otherwise an error. + * todo: customizable writable-branch-policy + */ +static int find_rw_parent(struct dentry *dentry, aufs_bindex_t bend) +{ + int err; + aufs_bindex_t bindex, candidate; + struct super_block *sb; + struct dentry *parent, *hidden_parent; + + err = bend; + sb = dentry->d_sb; + parent = dget_parent(dentry); +#if 1 // branch policy + hidden_parent = au_h_dptr_i(parent, bend); + if (hidden_parent && !br_rdonly(stobr(sb, bend))) + goto out; /* success */ +#endif + + candidate = -1; + for (bindex = dbstart(parent); bindex <= bend; bindex++) { + hidden_parent = au_h_dptr_i(parent, bindex); + if (hidden_parent && !br_rdonly(stobr(sb, bindex))) { +#if 0 // branch policy + if (candidate == -1) + candidate = bindex; + if (!au_test_perm(hidden_parent->d_inode, MAY_WRITE)) + return bindex; +#endif + err = bindex; + goto out; /* success */ + } + } +#if 0 // branch policy + err = candidate; + if (candidate != -1) + goto out; /* success */ +#endif + err = -EROFS; + + out: + dput(parent); + return err; +} + +int find_rw_br(struct super_block *sb, aufs_bindex_t bend) +{ + aufs_bindex_t bindex; + + for (bindex = bend; bindex >= 0; bindex--) + if (!br_rdonly(stobr(sb, bindex))) + return bindex; + return -EROFS; +} + +int find_rw_parent_br(struct dentry *dentry, aufs_bindex_t bend) +{ + int err; + + err = find_rw_parent(dentry, bend); + if (err >= 0) + return err; + return find_rw_br(dentry->d_sb, bend); +} + +/* ---------------------------------------------------------------------- */ + +/* + * test if two hidden_dentries have overlapping branches. + */ +//todo: try is_subdir() +static int do_is_overlap(struct super_block *sb, struct dentry *hidden_d1, + struct dentry *hidden_d2) +{ + struct dentry *d; + + d = hidden_d1; + do { + if (unlikely(d == hidden_d2)) + return 1; + d = d->d_parent; // dget_parent() + } while (!IS_ROOT(d)); + + return (d == hidden_d2); +} + +#if defined(CONFIG_BLK_DEV_LOOP) || defined(CONFIG_BLK_DEV_LOOP_MODULE) +#include +static int is_overlap_loopback(struct super_block *sb, struct dentry *hidden_d1, + struct dentry *hidden_d2) +{ + struct inode *hidden_inode; + struct loop_device *l; + + hidden_inode = hidden_d1->d_inode; + if (MAJOR(hidden_inode->i_sb->s_dev) != LOOP_MAJOR) + return 0; + + l = hidden_inode->i_sb->s_bdev->bd_disk->private_data; + hidden_d1 = l->lo_backing_file->f_dentry; + if (unlikely(hidden_d1->d_sb == sb)) + return 1; + return do_is_overlap(sb, hidden_d1, hidden_d2); +} +#else +#define is_overlap_loopback(sb, hidden_d1, hidden_d2) 0 +#endif + +static int is_overlap(struct super_block *sb, struct dentry *hidden_d1, + struct dentry *hidden_d2) +{ + LKTRTrace("d1 %.*s, d2 %.*s\n", DLNPair(hidden_d1), DLNPair(hidden_d2)); + if (unlikely(hidden_d1 == hidden_d2)) + return 1; + return do_is_overlap(sb, hidden_d1, hidden_d2) + || do_is_overlap(sb, hidden_d2, hidden_d1) + || is_overlap_loopback(sb, hidden_d1, hidden_d2) + || is_overlap_loopback(sb, hidden_d2, hidden_d1); +} + +/* ---------------------------------------------------------------------- */ + +static int init_br_wh(struct super_block *sb, aufs_bindex_t bindex, + struct aufs_branch *br, int new_perm, + struct dentry *h_root, struct vfsmount *h_mnt) +{ + int err, old_perm; + struct inode *dir = sb->s_root->d_inode, + *h_dir = h_root->d_inode; + const int new = (bindex < 0); + + LKTRTrace("b%d, new_perm %d\n", bindex, new_perm); + + if (new) + hi_lock_parent(h_dir); + else + hdir_lock(h_dir, dir, bindex); + + br_wh_write_lock(br); + old_perm = br->br_perm; + br->br_perm = new_perm; + err = init_wh(h_root, br, au_do_nfsmnt(h_mnt), sb); + br->br_perm = old_perm; + br_wh_write_unlock(br); + + if (new) + i_unlock(h_dir); + else + hdir_unlock(h_dir, dir, bindex); + + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +/* + * returns a newly allocated branch. @new_nbranch is a number of branches + * after adding a branch. + */ +static struct aufs_branch *alloc_addbr(struct super_block *sb, int new_nbranch) +{ + struct aufs_branch **branchp, *add_branch; + int sz; + void *p; + struct dentry *root; + struct inode *inode; + struct aufs_hinode *hinodep; + struct aufs_hdentry *hdentryp; + + LKTRTrace("new_nbranch %d\n", new_nbranch); + SiMustWriteLock(sb); + root = sb->s_root; + DiMustWriteLock(root); + inode = root->d_inode; + IiMustWriteLock(inode); + + add_branch = kmalloc(sizeof(*add_branch), GFP_KERNEL); + //if (LktrCond) {kfree(add_branch); add_branch = NULL;} + if (unlikely(!add_branch)) + goto out; + + sz = sizeof(*branchp) * (new_nbranch - 1); + if (unlikely(!sz)) + sz = sizeof(*branchp); + p = stosi(sb)->si_branch; + branchp = au_kzrealloc(p, sz, sizeof(*branchp) * new_nbranch, + GFP_KERNEL); + //if (LktrCond) branchp = NULL; + if (unlikely(!branchp)) + goto out; + stosi(sb)->si_branch = branchp; + + sz = sizeof(*hdentryp) * (new_nbranch - 1); + if (unlikely(!sz)) + sz = sizeof(*hdentryp); + p = dtodi(root)->di_hdentry; + hdentryp = au_kzrealloc(p, sz, sizeof(*hdentryp) * new_nbranch, + GFP_KERNEL); + //if (LktrCond) hdentryp = NULL; + if (unlikely(!hdentryp)) + goto out; + dtodi(root)->di_hdentry = hdentryp; + + sz = sizeof(*hinodep) * (new_nbranch - 1); + if (unlikely(!sz)) + sz = sizeof(*hinodep); + p = itoii(inode)->ii_hinode; + hinodep = au_kzrealloc(p, sz, sizeof(*hinodep) * new_nbranch, + GFP_KERNEL); + //if (LktrCond) hinodep = NULL; // unavailable test + if (unlikely(!hinodep)) + goto out; + itoii(inode)->ii_hinode = hinodep; + return add_branch; /* success */ + + out: + kfree(add_branch); + TraceErr(-ENOMEM); + return ERR_PTR(-ENOMEM); +} + +/* + * test if the branch permission is legal or not. + */ +static int test_br(struct super_block *sb, struct inode *inode, int brperm, + char *path) +{ + int err; + + err = 0; + if (unlikely(br_writable(brperm) && IS_RDONLY(inode))) { + Err("write permission for readonly fs or inode, %s\n", path); + err = -EINVAL; + } + + TraceErr(err); + return err; +} + +/* + * retunrs,,, + * 0: success, the caller will add it + * plus: success, it is already unified, the caller should ignore it + * minus: error + */ +static int test_add(struct super_block *sb, struct opt_add *add, int remount) +{ + int err; + struct dentry *root; + struct inode *inode, *hidden_inode; + aufs_bindex_t bend, bindex; + + LKTRTrace("%s, remo%d\n", add->path, remount); + + root = sb->s_root; + if (unlikely(au_find_dbindex(root, add->nd.dentry) != -1)) { + err = 1; + if (!remount) { + err = -EINVAL; + Err("%s duplicated\n", add->path); + } + goto out; + } + + err = -ENOSPC; //-E2BIG; + bend = sbend(sb); + //if (LktrCond) bend = AUFS_BRANCH_MAX; + if (unlikely(AUFS_BRANCH_MAX <= add->bindex + || AUFS_BRANCH_MAX - 1 <= bend)) { + Err("number of branches exceeded %s\n", add->path); + goto out; + } + + err = -EDOM; + if (unlikely(add->bindex < 0 || bend + 1 < add->bindex)) { + Err("bad index %d\n", add->bindex); + goto out; + } + + inode = add->nd.dentry->d_inode; + DEBUG_ON(!inode || !S_ISDIR(inode->i_mode)); + err = -ENOENT; + if (unlikely(!inode->i_nlink)) { + Err("no existence %s\n", add->path); + goto out; + } + + err = -EINVAL; + if (unlikely(inode->i_sb == sb)) { + Err("%s must be outside\n", add->path); + goto out; + } + +#if 1 //ndef CONFIG_AUFS_ROBR + if (unlikely(au_is_aufs(inode->i_sb) + || !strcmp(au_sbtype(inode->i_sb), "unionfs"))) { + Err("nested " AUFS_NAME " %s\n", add->path); + goto out; + } +#endif + +#ifdef AuNoNfsBranch + if (unlikely(au_is_nfs(inode->i_sb))) { + Err(AuNoNfsBranchMsg ". %s\n", add->path); + goto out; + } +#endif + + err = test_br(sb, add->nd.dentry->d_inode, add->perm, add->path); + if (unlikely(err)) + goto out; + + if (unlikely(bend == -1)) + return 0; /* success */ + + hidden_inode = au_h_dptr(root)->d_inode; + if (unlikely(au_flag_test(sb, AuFlag_WARN_PERM) + && ((hidden_inode->i_mode & S_IALLUGO) + != (inode->i_mode & S_IALLUGO) + || hidden_inode->i_uid != inode->i_uid + || hidden_inode->i_gid != inode->i_gid))) + Warn("uid/gid/perm %s %u/%u/0%o, %u/%u/0%o\n", + add->path, + inode->i_uid, inode->i_gid, (inode->i_mode & S_IALLUGO), + hidden_inode->i_uid, hidden_inode->i_gid, + (hidden_inode->i_mode & S_IALLUGO)); + + err = -EINVAL; + for (bindex = 0; bindex <= bend; bindex++) + if (unlikely(is_overlap(sb, add->nd.dentry, + au_h_dptr_i(root, bindex)))) { + Err("%s is overlapped\n", add->path); + goto out; + } + err = 0; + + out: + TraceErr(err); + return err; +} + +int br_add(struct super_block *sb, struct opt_add *add, int remount) +{ + int err, sz; + aufs_bindex_t bend, add_bindex; + struct dentry *root; + struct aufs_iinfo *iinfo; + struct aufs_sbinfo *sbinfo; + struct aufs_dinfo *dinfo; + struct inode *root_inode; + unsigned long long maxb; + struct aufs_branch **branchp, *add_branch; + struct aufs_hdentry *hdentryp; + struct aufs_hinode *hinodep; + + LKTRTrace("b%d, %s, 0x%x, %.*s\n", add->bindex, add->path, + add->perm, DLNPair(add->nd.dentry)); + SiMustWriteLock(sb); + root = sb->s_root; + DiMustWriteLock(root); + root_inode = root->d_inode; + IMustLock(root_inode); + IiMustWriteLock(root_inode); + + err = test_add(sb, add, remount); + if (unlikely(err < 0)) + goto out; + if (unlikely(err)) + return 0; /* success */ + + bend = sbend(sb); + add_branch = alloc_addbr(sb, bend + 2); + err = PTR_ERR(add_branch); + if (IS_ERR(add_branch)) + goto out; + + err = 0; + rw_init_nolock(&add_branch->br_wh_rwsem); + add_branch->br_wh = add_branch->br_plink = NULL; + if (unlikely(br_writable(add->perm))) { + err = init_br_wh(sb, /*bindex*/-1, add_branch, add->perm, + add->nd.dentry, add->nd.mnt); + if (unlikely(err)) { + kfree(add_branch); + goto out; + } + } + add_branch->br_xino = NULL; + add_branch->br_mnt = mntget(add->nd.mnt); + atomic_set(&add_branch->br_wh_running, 0); + add_branch->br_id = new_br_id(sb); + add_branch->br_perm = add->perm; + atomic_set(&add_branch->br_count, 0); + + sbinfo = stosi(sb); + dinfo = dtodi(root); + iinfo = itoii(root_inode); + + add_bindex = add->bindex; + sz = sizeof(*(sbinfo->si_branch)) * (bend + 1 - add_bindex); + branchp = sbinfo->si_branch + add_bindex; + memmove(branchp + 1, branchp, sz); + *branchp = add_branch; + sz = sizeof(*hdentryp) * (bend + 1 - add_bindex); + hdentryp = dinfo->di_hdentry + add_bindex; + memmove(hdentryp + 1, hdentryp, sz); + hdentryp->hd_dentry = NULL; + sz = sizeof(*hinodep) * (bend + 1 - add_bindex); + hinodep = iinfo->ii_hinode + add_bindex; + memmove(hinodep + 1, hinodep, sz); + hinodep->hi_inode = NULL; + hinodep->hi_notify = NULL; + + sbinfo->si_bend++; + dinfo->di_bend++; + iinfo->ii_bend++; + if (unlikely(bend == -1)) { + dinfo->di_bstart = 0; + iinfo->ii_bstart = 0; + } + set_h_dptr(root, add_bindex, dget(add->nd.dentry)); + set_h_iptr(root_inode, add_bindex, igrab(add->nd.dentry->d_inode), 0); + if (!add_bindex) + au_cpup_attr_all(root_inode); + else + au_add_nlink(root_inode, add->nd.dentry->d_inode); + maxb = add->nd.dentry->d_sb->s_maxbytes; + if (sb->s_maxbytes < maxb) + sb->s_maxbytes = maxb; + + if (au_flag_test(sb, AuFlag_XINO)) { + struct file *base_file = stobr(sb, 0)->br_xino; + if (!add_bindex) + base_file = stobr(sb, 1)->br_xino; + err = xino_init(sb, add_bindex, base_file, /*do_test*/1); + if (unlikely(err)) { + DEBUG_ON(add_branch->br_xino); + Err("ignored xino err %d, force noxino\n", err); + err = 0; + au_flag_clr(sb, AuFlag_XINO); + } + } + + out: + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +/* + * test if the branch is deletable or not. + */ +static int test_children_busy(struct dentry *root, aufs_bindex_t bindex) +{ + int err, i, j, sigen; + struct au_dcsub_pages dpages; + + LKTRTrace("b%d\n", bindex); + SiMustWriteLock(root->d_sb); + DiMustWriteLock(root); + + err = au_dpages_init(&dpages, GFP_KERNEL); + if (unlikely(err)) + goto out; + err = au_dcsub_pages(&dpages, root, NULL, NULL); + if (unlikely(err)) + goto out_dpages; + + sigen = au_sigen(root->d_sb); + DiMustNoWaiters(root); + IiMustNoWaiters(root->d_inode); + di_write_unlock(root); + for (i = 0; !err && i < dpages.ndpage; i++) { + struct au_dpage *dpage; + dpage = dpages.dpages + i; + for (j = 0; !err && j < dpage->ndentry; j++) { + struct dentry *d; + + d = dpage->dentries[j]; + if (au_digen(d) == sigen) + di_read_lock_child(d, AUFS_I_RLOCK); + else { + di_write_lock_child(d); + err = au_reval_dpath(d, sigen); + if (!err) + di_downgrade_lock(d, AUFS_I_RLOCK); + else { + di_write_unlock(d); + break; + } + } + + if (au_h_dptr_i(d, bindex) + && (!S_ISDIR(d->d_inode->i_mode) + || dbstart(d) == dbend(d))) + err = -EBUSY; + di_read_unlock(d, AUFS_I_RLOCK); + if (err) + LKTRTrace("%.*s\n", DLNPair(d)); + } + } + di_write_lock_child(root); /* aufs_write_lock() calls ..._child() */ + + out_dpages: + au_dpages_free(&dpages); + out: + TraceErr(err); + return err; +} + +int br_del(struct super_block *sb, struct opt_del *del, int remount) +{ + int err, do_wh, rerr; + struct dentry *root; + struct inode *inode, *hidden_dir; + aufs_bindex_t bindex, bend, br_id; + struct aufs_sbinfo *sbinfo; + struct aufs_dinfo *dinfo; + struct aufs_iinfo *iinfo; + struct aufs_branch *br; + + LKTRTrace("%s, %.*s\n", del->path, DLNPair(del->h_root)); + SiMustWriteLock(sb); + root = sb->s_root; + DiMustWriteLock(root); + inode = root->d_inode; + IiMustWriteLock(inode); + + bindex = au_find_dbindex(root, del->h_root); + if (unlikely(bindex < 0)) { + if (remount) + return 0; /* success */ + err = -ENOENT; + Err("%s no such branch\n", del->path); + goto out; + } + LKTRTrace("bindex b%d\n", bindex); + + err = -EBUSY; + bend = sbend(sb); + br = stobr(sb, bindex); + if (unlikely(!bend || br_count(br))) { + LKTRTrace("bend %d, br_count %d\n", bend, br_count(br)); + goto out; + } + + do_wh = 0; + hidden_dir = del->h_root->d_inode; + if (unlikely(br->br_wh || br->br_plink)) { +#if 0 + /* remove whiteout base */ + err = init_br_wh(sb, bindex, br, AuBr_RO, del->h_root, + br->br_mnt); + if (unlikely(err)) + goto out; +#else + dput(br->br_wh); + dput(br->br_plink); + br->br_wh = br->br_plink = NULL; +#endif + do_wh = 1; + } + + err = test_children_busy(root, bindex); + if (unlikely(err)) { + if (unlikely(do_wh)) + goto out_wh; + goto out; + } + + err = 0; + sbinfo = stosi(sb); + dinfo = dtodi(root); + iinfo = itoii(inode); + + dput(au_h_dptr_i(root, bindex)); + aufs_hiput(iinfo->ii_hinode + bindex); + br_id = br->br_id; + free_branch(br); + + //todo: realloc and shrink memeory + if (bindex < bend) { + const aufs_bindex_t n = bend - bindex; + struct aufs_branch **brp; + struct aufs_hdentry *hdp; + struct aufs_hinode *hip; + + brp = sbinfo->si_branch + bindex; + memmove(brp, brp + 1, sizeof(*brp) * n); + hdp = dinfo->di_hdentry + bindex; + memmove(hdp, hdp + 1, sizeof(*hdp) * n); + hip = iinfo->ii_hinode + bindex; + memmove(hip, hip + 1, sizeof(*hip) * n); + } + sbinfo->si_branch[0 + bend] = NULL; + dinfo->di_hdentry[0 + bend].hd_dentry = NULL; + iinfo->ii_hinode[0 + bend].hi_inode = NULL; + iinfo->ii_hinode[0 + bend].hi_notify = NULL; + + sbinfo->si_bend--; + dinfo->di_bend--; + iinfo->ii_bend--; + if (!bindex) + au_cpup_attr_all(inode); + else + au_sub_nlink(inode, del->h_root->d_inode); + if (au_flag_test(sb, AuFlag_PLINK)) + half_refresh_plink(sb, br_id); + + if (sb->s_maxbytes == del->h_root->d_sb->s_maxbytes) { + bend--; + sb->s_maxbytes = 0; + for (bindex = 0; bindex <= bend; bindex++) { + unsigned long long maxb; + maxb = sbr_sb(sb, bindex)->s_maxbytes; + if (sb->s_maxbytes < maxb) + sb->s_maxbytes = maxb; + } + } + goto out; /* success */ + + out_wh: + /* revert */ + rerr = init_br_wh(sb, bindex, br, br->br_perm, del->h_root, br->br_mnt); + if (rerr) + Warn("failed re-creating base whiteout, %s. (%d)\n", + del->path, rerr); + out: + TraceErr(err); + return err; +} + +static int do_need_sigen_inc(int a, int b) +{ + return (br_whable(a) && !br_whable(b)); +} + +static int need_sigen_inc(int old, int new) +{ + return (do_need_sigen_inc(old, new) + || do_need_sigen_inc(new, old)); +} + +int br_mod(struct super_block *sb, struct opt_mod *mod, int remount, + int *do_update) +{ + int err; + struct dentry *root; + aufs_bindex_t bindex; + struct aufs_branch *br; + struct inode *hidden_dir; + + LKTRTrace("%s, %.*s, 0x%x\n", + mod->path, DLNPair(mod->h_root), mod->perm); + SiMustWriteLock(sb); + root = sb->s_root; + DiMustWriteLock(root); + IiMustWriteLock(root->d_inode); + + bindex = au_find_dbindex(root, mod->h_root); + if (unlikely(bindex < 0)) { + if (remount) + return 0; /* success */ + err = -ENOENT; + Err("%s no such branch\n", mod->path); + goto out; + } + LKTRTrace("bindex b%d\n", bindex); + + hidden_dir = mod->h_root->d_inode; + err = test_br(sb, hidden_dir, mod->perm, mod->path); + if (unlikely(err)) + goto out; + + br = stobr(sb, bindex); + if (unlikely(br->br_perm == mod->perm)) + return 0; /* success */ + + if (br_writable(br->br_perm)) { +#if 1 + /* remove whiteout base */ + //todo: mod->perm? + err = init_br_wh(sb, bindex, br, AuBr_RO, mod->h_root, + br->br_mnt); + if (unlikely(err)) + goto out; +#else + dput(br->br_wh); + dput(br->br_plink); + br->br_wh = br->br_plink = NULL; +#endif + + if (!br_writable(mod->perm)) { + /* rw --> ro, file might be mmapped */ + struct file *file, *hf; + +#if 1 // test here + DiMustNoWaiters(root); + IiMustNoWaiters(root->d_inode); + di_write_unlock(root); + + // no need file_list_lock() since sbinfo is locked + //file_list_lock(); + list_for_each_entry(file, &sb->s_files, f_u.fu_list) { + LKTRTrace("%.*s\n", DLNPair(file->f_dentry)); + fi_read_lock(file); + if (!S_ISREG(file->f_dentry->d_inode->i_mode) + || !(file->f_mode & FMODE_WRITE) + || fbstart(file) != bindex) { + FiMustNoWaiters(file); + fi_read_unlock(file); + continue; + } + + // todo: already flushed? + hf = au_h_fptr(file); + hf->f_flags = au_file_roflags(hf->f_flags); + hf->f_mode &= ~FMODE_WRITE; + FiMustNoWaiters(file); + fi_read_unlock(file); + } + //file_list_unlock(); + + /* aufs_write_lock() calls ..._child() */ + di_write_lock_child(root); +#endif + } + } + + *do_update |= need_sigen_inc(br->br_perm, mod->perm); + br->br_perm = mod->perm; + return err; /* success */ + + out: + TraceErr(err); + return err; +} diff --git a/fs/aufs/branch.h b/fs/aufs/branch.h new file mode 100755 index 0000000..2557836 --- /dev/null +++ b/fs/aufs/branch.h @@ -0,0 +1,235 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: branch.h,v 1.30 2007/05/14 03:41:51 sfjro Exp $ */ + +#ifndef __AUFS_BRANCH_H__ +#define __AUFS_BRANCH_H__ + +#ifdef __KERNEL__ + +#include +#include +#include +#include +#include "misc.h" +#include "super.h" + +/* protected by superblock rwsem */ +struct aufs_branch { + struct file *br_xino; + readf_t br_xino_read; + writef_t br_xino_write; + + aufs_bindex_t br_id; + + int br_perm; + struct vfsmount *br_mnt; + atomic_t br_count; + + /* whiteout base */ + struct aufs_rwsem br_wh_rwsem; + struct dentry *br_wh; + atomic_t br_wh_running; + + /* pseudo-link dir */ + struct dentry *br_plink; +}; + +/* ---------------------------------------------------------------------- */ + +/* branch permission and attribute */ +enum { + AuBr_RW, /* writable, linkable wh */ + AuBr_RO, /* readonly, no wh */ + AuBr_RR, /* natively readonly, no wh */ + + AuBr_RWNoLinkWH, /* un-linkable whiteouts */ + + AuBr_ROWH, + AuBr_RRWH, /* whiteout-able */ + + AuBr_Last +}; + +static inline int br_writable(int brperm) +{ + return (brperm == AuBr_RW + || brperm == AuBr_RWNoLinkWH); +} + +static inline int br_whable(int brperm) +{ + return (brperm == AuBr_RW + || brperm == AuBr_ROWH + || brperm == AuBr_RRWH); +} + +static inline int br_linkable_wh(int brperm) +{ + return (brperm == AuBr_RW); +} + +/* ---------------------------------------------------------------------- */ + +#define _AuNoNfsBranchMsg "NFS branch is not supported" +#if LINUX_VERSION_CODE == KERNEL_VERSION(2,6,15) +#define AuNoNfsBranch +#define AuNoNfsBranchMsg _AuNoNfsBranchMsg +#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,19) \ + && !defined(CONFIG_AUFS_LHASH_PATCH) +#define AuNoNfsBranch +#define AuNoNfsBranchMsg _AuNoNfsBranchMsg \ + ", try lhash.patch and CONFIG_AUFS_LHASH_PATCH" +#endif + +/* ---------------------------------------------------------------------- */ + +struct aufs_sbinfo; +void free_branches(struct aufs_sbinfo *sinfo); +int br_rdonly(struct aufs_branch *br); +int find_brindex(struct super_block *sb, aufs_bindex_t br_id); +int find_rw_br(struct super_block *sb, aufs_bindex_t bend); +int find_rw_parent_br(struct dentry *dentry, aufs_bindex_t bend); +struct opt_add; +int br_add(struct super_block *sb, struct opt_add *add, int remount); +struct opt_del; +int br_del(struct super_block *sb, struct opt_del *del, int remount); +struct opt_mod; +int br_mod(struct super_block *sb, struct opt_mod *mod, int remount, + int *do_update); + +/* ---------------------------------------------------------------------- */ + +static inline int br_count(struct aufs_branch *br) +{ + return atomic_read(&br->br_count); +} + +static inline void br_get(struct aufs_branch *br) +{ + atomic_inc(&br->br_count); +} + +static inline void br_put(struct aufs_branch *br) +{ + atomic_dec(&br->br_count); +} + +/* ---------------------------------------------------------------------- */ + +/* Superblock to branch */ +static inline aufs_bindex_t sbr_id(struct super_block *sb, aufs_bindex_t bindex) +{ + return stobr(sb, bindex)->br_id; +} + +static inline +struct vfsmount *sbr_mnt(struct super_block *sb, aufs_bindex_t bindex) +{ + return stobr(sb, bindex)->br_mnt; +} + +static inline +struct super_block *sbr_sb(struct super_block *sb, aufs_bindex_t bindex) +{ + return sbr_mnt(sb, bindex)->mnt_sb; +} + +#if 0 +static inline int sbr_count(struct super_block *sb, aufs_bindex_t bindex) +{ + return br_count(stobr(sb, bindex)); +} + +static inline void sbr_get(struct super_block *sb, aufs_bindex_t bindex) +{ + br_get(stobr(sb, bindex)); +} +#endif + +static inline void sbr_put(struct super_block *sb, aufs_bindex_t bindex) +{ + br_put(stobr(sb, bindex)); +} + +static inline int sbr_perm(struct super_block *sb, aufs_bindex_t bindex) +{ + return stobr(sb, bindex)->br_perm; +} + +static inline int sbr_is_whable(struct super_block *sb, aufs_bindex_t bindex) +{ + return br_whable(sbr_perm(sb, bindex)); +} + +/* ---------------------------------------------------------------------- */ + +#ifdef CONFIG_AUFS_LHASH_PATCH +static inline struct vfsmount *au_do_nfsmnt(struct vfsmount *h_mnt) +{ + if (!au_is_nfs(h_mnt->mnt_sb)) + return NULL; + return h_mnt; +} + +/* it doesn't mntget() */ +static inline +struct vfsmount *au_nfsmnt(struct super_block *sb, aufs_bindex_t bindex) +{ + return au_do_nfsmnt(sbr_mnt(sb, bindex)); +} +#else +static inline struct vfsmount *au_do_nfsmnt(struct vfsmount *h_mnt) +{ + return NULL; +} + +static inline +struct vfsmount *au_nfsmnt(struct super_block *sb, aufs_bindex_t bindex) +{ + return NULL; +} +#endif /* CONFIG_AUFS_LHASH_PATCH */ + +/* ---------------------------------------------------------------------- */ + +/* + * br_wh_read_lock, br_wh_write_lock + * br_wh_read_unlock, br_wh_write_unlock, br_wh_downgrade_lock + */ +SimpleRwsemFuncs(br_wh, struct aufs_branch *br, br->br_wh_rwsem); + +/* to debug easier, do not make them inlined functions */ +#define BrWhMustReadLock(br) do { \ + /* SiMustAnyLock(sb); */ \ + RwMustReadLock(&(br)->br_wh_rwsem); \ +} while (0) + +#define BrWhMustWriteLock(br) do { \ + /* SiMustAnyLock(sb); */ \ + RwMustWriteLock(&(br)->br_wh_rwsem); \ +} while (0) + +#define BrWhMustAnyLock(br) do { \ + /* SiMustAnyLock(sb); */ \ + RwMustAnyLock(&(br)->br_wh_rwsem); \ +} while (0) + +#endif /* __KERNEL__ */ +#endif /* __AUFS_BRANCH_H__ */ diff --git a/fs/aufs/cpup.c b/fs/aufs/cpup.c new file mode 100755 index 0000000..6636f40 --- /dev/null +++ b/fs/aufs/cpup.c @@ -0,0 +1,773 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: cpup.c,v 1.37 2007/05/14 03:41:52 sfjro Exp $ */ + +#include +#include "aufs.h" + +/* violent cpup_attr_*() functions don't care inode lock */ +void au_cpup_attr_timesizes(struct inode *inode) +{ + struct inode *hidden_inode; + + LKTRTrace("i%lu\n", inode->i_ino); + //IMustLock(inode); + hidden_inode = au_h_iptr(inode); + DEBUG_ON(!hidden_inode); + //IMustLock(!hidden_inode); + + inode->i_atime = hidden_inode->i_atime; + inode->i_mtime = hidden_inode->i_mtime; + inode->i_ctime = hidden_inode->i_ctime; + spin_lock(&inode->i_lock); + i_size_write(inode, i_size_read(hidden_inode)); + inode->i_blocks = hidden_inode->i_blocks; + spin_unlock(&inode->i_lock); +} + +void au_cpup_attr_nlink(struct inode *inode) +{ + struct inode *h_inode; + + LKTRTrace("i%lu\n", inode->i_ino); + //IMustLock(inode); + DEBUG_ON(!inode->i_mode); + + h_inode = au_h_iptr(inode); + inode->i_nlink = h_inode->i_nlink; + + /* + * fewer nlink makes find(1) noisy, but larger nlink doesn't. + * it may includes whplink directory. + */ + if (unlikely(S_ISDIR(h_inode->i_mode))) { + aufs_bindex_t bindex, bend; + bend = ibend(inode); + for (bindex = ibstart(inode) + 1; bindex <= bend; bindex++) { + h_inode = au_h_iptr_i(inode, bindex); + if (h_inode) + au_add_nlink(inode, h_inode); + } + } +} + +void au_cpup_attr_changable(struct inode *inode) +{ + struct inode *hidden_inode; + + LKTRTrace("i%lu\n", inode->i_ino); + //IMustLock(inode); + hidden_inode = au_h_iptr(inode); + DEBUG_ON(!hidden_inode); + + inode->i_mode = hidden_inode->i_mode; + inode->i_uid = hidden_inode->i_uid; + inode->i_gid = hidden_inode->i_gid; + au_cpup_attr_timesizes(inode); + + //?? + inode->i_flags = hidden_inode->i_flags; +} + +void au_cpup_igen(struct inode *inode, struct inode *h_inode) +{ + inode->i_generation = h_inode->i_generation; + itoii(inode)->ii_hsb1 = h_inode->i_sb; +} + +void au_cpup_attr_all(struct inode *inode) +{ + struct inode *hidden_inode; + + LKTRTrace("i%lu\n", inode->i_ino); + //IMustLock(inode); + hidden_inode = au_h_iptr(inode); + DEBUG_ON(!hidden_inode); + + au_cpup_attr_changable(inode); + if (inode->i_nlink > 0) + au_cpup_attr_nlink(inode); + + switch (inode->i_mode & S_IFMT) { + case S_IFBLK: + case S_IFCHR: + inode->i_rdev = hidden_inode->i_rdev; + } + inode->i_blkbits = hidden_inode->i_blkbits; + au_cpup_attr_blksize(inode, hidden_inode); + au_cpup_igen(inode, hidden_inode); +} + +/* ---------------------------------------------------------------------- */ + +/* Note: dt_dentry and dt_hidden_dentry are not dget/dput-ed */ + +/* keep the timestamps of the parent dir when cpup */ +void dtime_store(struct dtime *dt, struct dentry *dentry, + struct dentry *hidden_dentry) +{ + struct inode *inode; + + TraceEnter(); + DEBUG_ON(!dentry || !hidden_dentry || !hidden_dentry->d_inode); + + dt->dt_dentry = dentry; + dt->dt_h_dentry = hidden_dentry; + inode = hidden_dentry->d_inode; + dt->dt_atime = inode->i_atime; + dt->dt_mtime = inode->i_mtime; + //smp_mb(); +} + +// todo: remove extra parameter +void dtime_revert(struct dtime *dt, int h_parent_is_locked) +{ + struct iattr attr; + int err; + struct dentry *dentry; + + LKTRTrace("h_parent locked %d\n", h_parent_is_locked); + + attr.ia_atime = dt->dt_atime; + attr.ia_mtime = dt->dt_mtime; + attr.ia_valid = ATTR_FORCE | ATTR_MTIME | ATTR_MTIME_SET + | ATTR_ATIME | ATTR_ATIME_SET; + //smp_mb(); + dentry = NULL; + if (!h_parent_is_locked /* && !IS_ROOT(dt->dt_dentry) */) + dentry = dt->dt_dentry; + err = vfsub_notify_change(dt->dt_h_dentry, &attr, + need_dlgt(dt->dt_dentry->d_sb)); + if (unlikely(err)) + Warn("restoring timestamps failed(%d). ignored\n", err); +} + +/* ---------------------------------------------------------------------- */ + +static int cpup_iattr(struct dentry *hidden_dst, struct dentry *hidden_src, + int dlgt) +{ + int err; + struct iattr ia; + struct inode *hidden_isrc, *hidden_idst; + + LKTRTrace("%.*s\n", DLNPair(hidden_dst)); + hidden_idst = hidden_dst->d_inode; + //IMustLock(hidden_idst); + hidden_isrc = hidden_src->d_inode; + //IMustLock(hidden_isrc); + + ia.ia_valid = ATTR_FORCE | ATTR_MODE | ATTR_UID | ATTR_GID + | ATTR_ATIME | ATTR_MTIME + | ATTR_ATIME_SET | ATTR_MTIME_SET; + ia.ia_mode = hidden_isrc->i_mode; + ia.ia_uid = hidden_isrc->i_uid; + ia.ia_gid = hidden_isrc->i_gid; + ia.ia_atime = hidden_isrc->i_atime; + ia.ia_mtime = hidden_isrc->i_mtime; + err = vfsub_notify_change(hidden_dst, &ia, dlgt); + //if (LktrCond) err = -1; + if (!err) + hidden_idst->i_flags = hidden_isrc->i_flags; //?? + + TraceErr(err); + return err; +} + +/* + * to support a sparse file which is opened with O_APPEND, + * we need to close the file. + */ +static int cpup_regular(struct dentry *dentry, aufs_bindex_t bdst, + aufs_bindex_t bsrc, loff_t len) +{ + int err, i, sparse; + struct super_block *sb; + struct inode *hidden_inode; + enum {SRC, DST}; + struct { + aufs_bindex_t bindex; + unsigned int flags; + struct dentry *dentry; + struct file *file; + void *label, *label_file; + } *h, hidden[] = { + { + .bindex = bsrc, + .flags = O_RDONLY | O_NOATIME | O_LARGEFILE, + .file = NULL, + .label = &&out, + .label_file = &&out_src_file + }, + { + .bindex = bdst, + .flags = O_WRONLY | O_NOATIME | O_LARGEFILE, + .file = NULL, + .label = &&out_src_file, + .label_file = &&out_dst_file + } + }; + + LKTRTrace("dentry %.*s, bdst %d, bsrc %d, len %lld\n", + DLNPair(dentry), bdst, bsrc, len); + DEBUG_ON(bsrc <= bdst); + DEBUG_ON(!len); + sb = dentry->d_sb; + DEBUG_ON(test_ro(sb, bdst, dentry->d_inode)); + // bsrc branch can be ro/rw. + + h = hidden; + for (i = 0; i < 2; i++, h++) { + h->dentry = au_h_dptr_i(dentry, h->bindex); + DEBUG_ON(!h->dentry); + hidden_inode = h->dentry->d_inode; + DEBUG_ON(!hidden_inode || !S_ISREG(hidden_inode->i_mode)); + h->file = hidden_open(dentry, h->bindex, h->flags); + //if (LktrCond) + //{fput(h->file); sbr_put(sb, h->bindex); h->file = ERR_PTR(-1);} + err = PTR_ERR(h->file); + if (IS_ERR(h->file)) + goto *h->label; + err = -EINVAL; + if (unlikely(!h->file->f_op)) + goto *h->label_file; + } + + /* stop updating while we copyup */ + IMustLock(hidden[SRC].dentry->d_inode); + sparse = 0; + err = au_copy_file(hidden[DST].file, hidden[SRC].file, len, sb, + &sparse); + + /* sparse file: update i_blocks next time */ + if (unlikely(!err && sparse)) + d_drop(dentry); + + out_dst_file: + fput(hidden[DST].file); + sbr_put(sb, hidden[DST].bindex); + out_src_file: + fput(hidden[SRC].file); + sbr_put(sb, hidden[SRC].bindex); + out: + TraceErr(err); + return err; +} + +// unnecessary? +unsigned int au_flags_cpup(unsigned int init, struct dentry *parent) +{ + if (unlikely(parent && IS_ROOT(parent))) + init |= CPUP_LOCKED_GHDIR; + return init; +} + +/* return with hidden dst inode is locked */ +static int cpup_entry(struct dentry *dentry, aufs_bindex_t bdst, + aufs_bindex_t bsrc, loff_t len, unsigned int flags, + int dlgt) +{ + int err, isdir, symlen; + struct dentry *hidden_src, *hidden_dst, *hidden_parent, *parent; + struct inode *hidden_inode, *hidden_dir, *dir; + struct dtime dt; + umode_t mode; + char *sym; + mm_segment_t old_fs; + const int do_dt = flags & CPUP_DTIME; + struct super_block *sb; + + LKTRTrace("%.*s, i%lu, bdst %d, bsrc %d, len %Ld, flags 0x%x\n", + DLNPair(dentry), dentry->d_inode->i_ino, bdst, bsrc, len, + flags); + sb = dentry->d_sb; + DEBUG_ON(bdst >= bsrc || test_ro(sb, bdst, NULL)); + // bsrc branch can be ro/rw. + + hidden_src = au_h_dptr_i(dentry, bsrc); + DEBUG_ON(!hidden_src); + hidden_inode = hidden_src->d_inode; + DEBUG_ON(!hidden_inode); + + /* stop refrencing while we are creating */ + //parent = dget_parent(dentry); + parent = dentry->d_parent; + dir = parent->d_inode; + hidden_dst = au_h_dptr_i(dentry, bdst); + DEBUG_ON(hidden_dst && hidden_dst->d_inode); + //hidden_parent = dget_parent(hidden_dst); + hidden_parent = hidden_dst->d_parent; + hidden_dir = hidden_parent->d_inode; + IMustLock(hidden_dir); + + if (do_dt) + dtime_store(&dt, parent, hidden_parent); + + isdir = 0; + mode = hidden_inode->i_mode; + switch (mode & S_IFMT) { + case S_IFREG: + /* stop updating while we are referencing */ + IMustLock(hidden_inode); + err = vfsub_create(hidden_dir, hidden_dst, mode | S_IWUSR, NULL, + dlgt); + //if (LktrCond) {vfs_unlink(hidden_dir, hidden_dst); err = -1;} + if (!err) { + loff_t l = i_size_read(hidden_inode); + if (len == -1 || l < len) + len = l; + if (len) { + err = cpup_regular(dentry, bdst, bsrc, len); + //if (LktrCond) err = -1; + } + if (unlikely(err)) { + int rerr; + rerr = vfsub_unlink(hidden_dir, hidden_dst, + dlgt); + if (rerr) { + IOErr("failed unlinking cpup-ed %.*s" + "(%d, %d)\n", + DLNPair(hidden_dst), err, rerr); + err = -EIO; + } + } + } + break; + case S_IFDIR: + isdir = 1; + err = vfsub_mkdir(hidden_dir, hidden_dst, mode, dlgt); + //if (LktrCond) {vfs_rmdir(hidden_dir, hidden_dst); err = -1;} + if (!err) { + /* setattr case: dir is not locked */ + if (0 && ibstart(dir) == bdst) + au_cpup_attr_nlink(dir); + au_cpup_attr_nlink(dentry->d_inode); + } + break; + case S_IFLNK: + err = -ENOMEM; + sym = __getname(); + //if (LktrCond) {__putname(sym); sym = NULL;} + if (unlikely(!sym)) + break; + old_fs = get_fs(); + set_fs(KERNEL_DS); + err = symlen = hidden_inode->i_op->readlink + (hidden_src, (char __user*)sym, PATH_MAX); + //if (LktrCond) err = symlen = -1; + set_fs(old_fs); + if (symlen > 0) { + sym[symlen] = 0; + err = vfsub_symlink(hidden_dir, hidden_dst, sym, mode, + dlgt); + //if (LktrCond) + //{vfs_unlink(hidden_dir, hidden_dst); err = -1;} + } + __putname(sym); + break; + case S_IFCHR: + case S_IFBLK: + DEBUG_ON(!capable(CAP_MKNOD)); + /*FALLTHROUGH*/ + case S_IFIFO: + case S_IFSOCK: + err = vfsub_mknod(hidden_dir, hidden_dst, mode, + hidden_inode->i_rdev, dlgt); + //if (LktrCond) {vfs_unlink(hidden_dir, hidden_dst); err = -1;} + break; + default: + IOErr("Unknown inode type 0%o\n", mode); + err = -EIO; + } + + if (do_dt) + dtime_revert(&dt, flags & CPUP_LOCKED_GHDIR); + //dput(parent); + //dput(hidden_parent); + TraceErr(err); + return err; +} + +/* + * copyup the @dentry from @bsrc to @bdst. + * the caller must set the both of hidden dentries. + * @len is for trucating when it is -1 copyup the entire file. + */ +int cpup_single(struct dentry *dentry, aufs_bindex_t bdst, aufs_bindex_t bsrc, + loff_t len, unsigned int flags) +{ + int err, rerr, isdir, dlgt; + struct dentry *hidden_src, *hidden_dst, *parent;//, *h_parent; + struct inode *dst_inode, *hidden_dir, *inode, *src_inode; + struct super_block *sb; + aufs_bindex_t old_ibstart; + struct dtime dt; + + LKTRTrace("%.*s, i%lu, bdst %d, bsrc %d, len %Ld, flags 0x%x\n", + DLNPair(dentry), dentry->d_inode->i_ino, bdst, bsrc, len, + flags); + sb = dentry->d_sb; + DEBUG_ON(bsrc <= bdst); + hidden_dst = au_h_dptr_i(dentry, bdst); + DEBUG_ON(!hidden_dst || hidden_dst->d_inode); + //h_parent = dget_parent(hidden_dst); + //hidden_dir = h_parent->d_inode; + hidden_dir = hidden_dst->d_parent->d_inode; + IMustLock(hidden_dir); + hidden_src = au_h_dptr_i(dentry, bsrc); + DEBUG_ON(!hidden_src || !hidden_src->d_inode); + inode = dentry->d_inode; + IiMustWriteLock(inode); + + dlgt = need_dlgt(sb); + dst_inode = au_h_iptr_i(inode, bdst); + if (unlikely(dst_inode)) { + if (unlikely(!au_flag_test(sb, AuFlag_PLINK))) { + err = -EIO; + IOErr("i%lu exists on a upper branch " + "but plink is disabled\n", inode->i_ino); + goto out; + } + + if (dst_inode->i_nlink) { + hidden_src = lkup_plink(sb, bdst, inode); + err = PTR_ERR(hidden_src); + if (IS_ERR(hidden_src)) + goto out; + DEBUG_ON(!hidden_src->d_inode); + // vfs_link() does lock the inode + err = vfsub_link(hidden_src, hidden_dir, hidden_dst, dlgt); + dput(hidden_src); + goto out; + } else + /* udba work */ + au_update_brange(inode, 1); + } + + old_ibstart = ibstart(inode); + err = cpup_entry(dentry, bdst, bsrc, len, flags, dlgt); + if (unlikely(err)) + goto out; + dst_inode = hidden_dst->d_inode; + hi_lock_child2(dst_inode); + + //todo: test dlgt + err = cpup_iattr(hidden_dst, hidden_src, dlgt); + //if (LktrCond) err = -1; +#if 0 // xattr + if (0 && !err) + err = cpup_xattrs(hidden_src, hidden_dst); +#endif + isdir = S_ISDIR(dst_inode->i_mode); + if (!err) { + if (bdst < old_ibstart) + set_ibstart(inode, bdst); + set_h_iptr(inode, bdst, igrab(dst_inode), + au_hi_flags(inode, isdir)); + i_unlock(dst_inode); + src_inode = hidden_src->d_inode; + if (!isdir) { + if (src_inode->i_nlink > 1 + && au_flag_test(sb, AuFlag_PLINK)) + append_plink(sb, inode, hidden_dst, bdst); + else { + /* braces are added to stop a warning */ + ;//xino_write0(sb, bsrc, src_inode->i_ino); + /* ignore this error */ + } + } + //goto out; /* success */ + return 0; /* success */ + } + + /* revert */ + i_unlock(dst_inode); + parent = dget_parent(dentry); + //dtime_store(&dt, parent, h_parent); + dtime_store(&dt, parent, hidden_dst->d_parent); + dput(parent); + if (!isdir) + rerr = vfsub_unlink(hidden_dir, hidden_dst, dlgt); + else + rerr = vfsub_rmdir(hidden_dir, hidden_dst, dlgt); + //rerr = -1; + dtime_revert(&dt, flags & CPUP_LOCKED_GHDIR); + if (rerr) { + IOErr("failed removing broken entry(%d, %d)\n", err, rerr); + err = -EIO; + } + + out: + //dput(h_parent); + TraceErr(err); + return err; +} + +struct cpup_single_args { + int *errp; + struct dentry *dentry; + aufs_bindex_t bdst, bsrc; + loff_t len; + unsigned int flags; +}; + +static void call_cpup_single(void *args) +{ + struct cpup_single_args *a = args; + *a->errp = cpup_single(a->dentry, a->bdst, a->bsrc, a->len, a->flags); +} + +int sio_cpup_single(struct dentry *dentry, aufs_bindex_t bdst, + aufs_bindex_t bsrc, loff_t len, unsigned int flags) +{ + int err; + struct dentry *hidden_dentry; + umode_t mode; + + LKTRTrace("%.*s, i%lu, bdst %d, bsrc %d, len %Ld, flags 0x%x\n", + DLNPair(dentry), dentry->d_inode->i_ino, bdst, bsrc, len, + flags); + + hidden_dentry = au_h_dptr_i(dentry, bsrc); + mode = hidden_dentry->d_inode->i_mode & S_IFMT; + if ((mode != S_IFCHR && mode != S_IFBLK) + || capable(CAP_MKNOD)) + err = cpup_single(dentry, bdst, bsrc, len, flags); + else { + struct cpup_single_args args = { + .errp = &err, + .dentry = dentry, + .bdst = bdst, + .bsrc = bsrc, + .len = len, + .flags = flags + }; + au_wkq_wait(call_cpup_single, &args, /*dlgt*/0); + } + + TraceErr(err); + return err; +} + +/* + * copyup the @dentry from the first active hidden branch to @bdst, + * using cpup_single(). + */ +int cpup_simple(struct dentry *dentry, aufs_bindex_t bdst, loff_t len, + unsigned int flags) +{ + int err; + struct inode *inode; + aufs_bindex_t bsrc, bend; + + LKTRTrace("%.*s, bdst %d, len %Ld, flags 0x%x\n", + DLNPair(dentry), bdst, len, flags); + inode = dentry->d_inode; + DEBUG_ON(!S_ISDIR(inode->i_mode) && dbstart(dentry) < bdst); + + bend = dbend(dentry); + for (bsrc = bdst + 1; bsrc <= bend; bsrc++) + if (au_h_dptr_i(dentry, bsrc)) + break; + DEBUG_ON(!au_h_dptr_i(dentry, bsrc)); + + err = lkup_neg(dentry, bdst); + //err = -1; + if (!err) { + err = cpup_single(dentry, bdst, bsrc, len, flags); + if (!err) + return 0; /* success */ + + /* revert */ + set_h_dptr(dentry, bdst, NULL); + set_dbstart(dentry, bsrc); + } + + TraceErr(err); + return err; +} + +struct cpup_simple_args { + int *errp; + struct dentry *dentry; + aufs_bindex_t bdst; + loff_t len; + unsigned int flags; +}; + +static void call_cpup_simple(void *args) +{ + struct cpup_simple_args *a = args; + *a->errp = cpup_simple(a->dentry, a->bdst, a->len, a->flags); +} + +int sio_cpup_simple(struct dentry *dentry, aufs_bindex_t bdst, loff_t len, + unsigned int flags) +{ + int err, do_sio, dlgt; + //struct dentry *parent; + struct inode *hidden_dir, *dir; + + LKTRTrace("%.*s, b%d, len %Ld, flags 0x%x\n", + DLNPair(dentry), bdst, len, flags); + + //parent = dget_parent(dentry); + //dir = parent->d_inode; + dir = dentry->d_parent->d_inode; + hidden_dir = au_h_iptr_i(dir, bdst); + dlgt = need_dlgt(dir->i_sb); + do_sio = au_test_perm(hidden_dir, MAY_EXEC | MAY_WRITE, dlgt); + if (!do_sio) { + umode_t mode = dentry->d_inode->i_mode & S_IFMT; + do_sio = ((mode == S_IFCHR || mode == S_IFBLK) + && !capable(CAP_MKNOD)); + } + if (!do_sio) + err = cpup_simple(dentry, bdst, len, flags); + else { + struct cpup_simple_args args = { + .errp = &err, + .dentry = dentry, + .bdst = bdst, + .len = len, + .flags = flags + }; + au_wkq_wait(call_cpup_simple, &args, /*dlgt*/0); + } + + //dput(parent); + TraceErr(err); + return err; +} + +//todo: dcsub +/* cf. revalidate function in file.c */ +int cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst, struct dentry *locked) +{ + int err; + struct super_block *sb; + struct dentry *d, *parent, *hidden_parent; + unsigned int udba; + + LKTRTrace("%.*s, b%d, parent i%lu, locked %p\n", + DLNPair(dentry), bdst, parent_ino(dentry), locked); + sb = dentry->d_sb; + DEBUG_ON(test_ro(sb, bdst, NULL)); + parent = dentry->d_parent; + IiMustWriteLock(parent->d_inode); + if (unlikely(IS_ROOT(parent))) + return 0; + if (locked) { + DiMustAnyLock(locked); + IiMustAnyLock(locked->d_inode); + } + + /* slow loop, keep it simple and stupid */ + err = 0; + udba = au_flag_test(sb, AuFlag_UDBA_INOTIFY); + while (1) { + parent = dentry->d_parent; // dget_parent() + hidden_parent = au_h_dptr_i(parent, bdst); + if (hidden_parent) + return 0; /* success */ + + /* find top dir which is needed to cpup */ + do { + d = parent; + parent = d->d_parent; // dget_parent() + if (parent != locked) + di_read_lock_parent3(parent, !AUFS_I_RLOCK); + hidden_parent = au_h_dptr_i(parent, bdst); + if (parent != locked) + di_read_unlock(parent, !AUFS_I_RLOCK); + } while (!hidden_parent); + + if (d != dentry->d_parent) + di_write_lock_child3(d); + + /* somebody else might create while we were sleeping */ + if (!au_h_dptr_i(d, bdst) || !au_h_dptr_i(d, bdst)->d_inode) { + struct inode *h_dir = hidden_parent->d_inode, + *dir = parent->d_inode, + *h_gdir, *gdir; + + if (au_h_dptr_i(d, bdst)) + au_update_dbstart(d); + //DEBUG_ON(dbstart(d) <= bdst); + if (parent != locked) + di_read_lock_parent3(parent, AUFS_I_RLOCK); + h_gdir = gdir = NULL; + if (unlikely(udba && !IS_ROOT(parent))) { + gdir = parent->d_parent->d_inode; + h_gdir = hidden_parent->d_parent->d_inode; + hgdir_lock(h_gdir, gdir, bdst); + } + hdir_lock(h_dir, dir, bdst); + err = sio_cpup_simple(d, bdst, -1, + au_flags_cpup(CPUP_DTIME, + parent)); + //if (LktrCond) err = -1; + hdir_unlock(h_dir, dir, bdst); + if (unlikely(gdir)) + hdir_unlock(h_gdir, gdir, bdst); + if (parent != locked) + di_read_unlock(parent, AUFS_I_RLOCK); + } + + if (d != dentry->d_parent) + di_write_unlock(d); + if (unlikely(err)) + break; + } + +// out: + TraceErr(err); + return err; +} + +int test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst, + struct dentry *locked) +{ + int err; + struct dentry *parent; + struct inode *dir; + + parent = dentry->d_parent; + dir = parent->d_inode; + LKTRTrace("%.*s, b%d, parent i%lu, locked %p\n", + DLNPair(dentry), bdst, dir->i_ino, locked); + DiMustReadLock(parent); + IiMustReadLock(dir); + + if (au_h_iptr_i(dir, bdst)) + return 0; + + err = 0; + di_read_unlock(parent, AUFS_I_RLOCK); + di_write_lock_parent(parent); + if (au_h_iptr_i(dir, bdst)) + goto out; + + err = cpup_dirs(dentry, bdst, locked); + + out: + di_downgrade_lock(parent, AUFS_I_RLOCK); + TraceErr(err); + return err; +} diff --git a/fs/aufs/cpup.h b/fs/aufs/cpup.h new file mode 100755 index 0000000..86557aa --- /dev/null +++ b/fs/aufs/cpup.h @@ -0,0 +1,72 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: cpup.h,v 1.15 2007/05/14 03:41:52 sfjro Exp $ */ + +#ifndef __AUFS_CPUP_H__ +#define __AUFS_CPUP_H__ + +#ifdef __KERNEL__ + +#include +#include +#include + +static inline +void au_cpup_attr_blksize(struct inode *inode, struct inode *h_inode) +{ +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,19) + inode->i_blksize = h_inode->i_blksize; +#endif +} + +void au_cpup_attr_timesizes(struct inode *inode); +void au_cpup_attr_nlink(struct inode *inode); +void au_cpup_attr_changable(struct inode *inode); +void au_cpup_igen(struct inode *inode, struct inode *h_inode); +void au_cpup_attr_all(struct inode *inode); + +#define CPUP_DTIME 1 // do dtime_store/revert +// todo: remove this +#define CPUP_LOCKED_GHDIR 2 // grand parent hidden dir is locked +unsigned int au_flags_cpup(unsigned int init, struct dentry *parent); + +int cpup_single(struct dentry *dentry, aufs_bindex_t bdst, aufs_bindex_t bsrc, + loff_t len, unsigned int flags); +int sio_cpup_single(struct dentry *dentry, aufs_bindex_t bdst, + aufs_bindex_t bsrc, loff_t len, unsigned int flags); +int cpup_simple(struct dentry *dentry, aufs_bindex_t bdst, loff_t len, + unsigned int flags); +int sio_cpup_simple(struct dentry *dentry, aufs_bindex_t bdst, loff_t len, + unsigned int flags); + +int cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst, struct dentry *locked); +int test_and_cpup_dirs(struct dentry *dentry, aufs_bindex_t bdst, + struct dentry *locked); + +/* keep timestamps when copyup */ +struct dtime { + struct dentry *dt_dentry, *dt_h_dentry; + struct timespec dt_atime, dt_mtime; +}; +void dtime_store(struct dtime *dt, struct dentry *dentry, + struct dentry *h_dentry); +void dtime_revert(struct dtime *dt, int h_parent_is_locked); + +#endif /* __KERNEL__ */ +#endif /* __AUFS_CPUP_H__ */ diff --git a/fs/aufs/dcsub.c b/fs/aufs/dcsub.c new file mode 100755 index 0000000..6ec29d3 --- /dev/null +++ b/fs/aufs/dcsub.c @@ -0,0 +1,175 @@ +/* + * Copyright (C) 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: dcsub.c,v 1.3 2007/05/14 03:41:52 sfjro Exp $ */ + +#include "aufs.h" + +static void au_dpage_free(struct au_dpage *dpage) +{ + int i; + + TraceEnter(); + DEBUG_ON(!dpage); + + for (i = 0; i < dpage->ndentry; i++) + dput(dpage->dentries[i]); + free_page((unsigned long)dpage->dentries); +} + +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp) +{ + int err; + void *p; + + TraceEnter(); + + err = -ENOMEM; + dpages->dpages = kmalloc(sizeof(*dpages->dpages), gfp); + if (unlikely(!dpages->dpages)) + goto out; + p = (void*)__get_free_page(gfp); + if (unlikely(!p)) + goto out_dpages; + dpages->dpages[0].ndentry = 0; + dpages->dpages[0].dentries = p; + dpages->ndpage = 1; + return 0; /* success */ + + out_dpages: + kfree(dpages->dpages); + out: + TraceErr(err); + return err; +} + +void au_dpages_free(struct au_dcsub_pages *dpages) +{ + int i; + + TraceEnter(); + + for (i = 0; i < dpages->ndpage; i++) + au_dpage_free(dpages->dpages + i); + kfree(dpages->dpages); +} + +static int au_dpages_append(struct au_dcsub_pages *dpages, + struct dentry *dentry, gfp_t gfp) +{ + int err, sz; + struct au_dpage *dpage; + void *p; + + //TraceEnter(); + + dpage = dpages->dpages + dpages->ndpage - 1; + DEBUG_ON(!dpage); + sz = PAGE_SIZE/sizeof(dentry); + if (unlikely(dpage->ndentry >= sz)) { + LKTRLabel(new dpage); + err = -ENOMEM; + sz = dpages->ndpage * sizeof(*dpages->dpages); + p = au_kzrealloc(dpages->dpages, sz, + sz + sizeof(*dpages->dpages), gfp); + if (unlikely(!p)) + goto out; + dpage = dpages->dpages + dpages->ndpage; + p = (void*)__get_free_page(gfp); + if (unlikely(!p)) + goto out; + dpage->ndentry = 0; + dpage->dentries = p; + dpages->ndpage++; + } + + dpage->dentries[dpage->ndentry++] = dget(dentry); + return 0; /* success */ + + out: + //TraceErr(err); + return err; +} + +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root, + au_dpages_test test, void *arg) +{ + int err; + struct dentry *this_parent = root; + struct list_head *next; + struct super_block *sb = root->d_sb; + + TraceEnter(); + + err = 0; + spin_lock(&dcache_lock); + repeat: + next = this_parent->d_subdirs.next; + resume: + if (this_parent->d_sb == sb + && !IS_ROOT(this_parent) + && atomic_read(&this_parent->d_count) + && this_parent->d_inode + && (!test || test(this_parent, arg))) { + err = au_dpages_append(dpages, this_parent, GFP_ATOMIC); + if (unlikely(err)) + goto out; + } + + while (next != &this_parent->d_subdirs) { + struct list_head *tmp = next; + struct dentry *dentry = list_entry(tmp, struct dentry, D_CHILD); + next = tmp->next; + if (unlikely(/*d_unhashed(dentry) || */!dentry->d_inode)) + continue; + if (!list_empty(&dentry->d_subdirs)) { + this_parent = dentry; + goto repeat; + } + if (dentry->d_sb == sb + && atomic_read(&dentry->d_count) + && (!test || test(dentry, arg))) { + err = au_dpages_append(dpages, dentry, GFP_ATOMIC); + if (unlikely(err)) + goto out; + } + } + + if (this_parent != root) { + next = this_parent->D_CHILD.next; + this_parent = this_parent->d_parent; + goto resume; + } + out: + spin_unlock(&dcache_lock); +#if 0 + if (!err) { + int i, j; + j = 0; + for (i = 0; i < dpages->ndpage; i++) { + if ((dpages->dpages + i)->ndentry) + Dbg("%d: %d\n", i, (dpages->dpages + i)->ndentry); + j += (dpages->dpages + i)->ndentry; + } + if (j) + Dbg("ndpage %d, %d\n", dpages->ndpage, j); + } +#endif + TraceErr(err); + return err; +} diff --git a/fs/aufs/dcsub.h b/fs/aufs/dcsub.h new file mode 100755 index 0000000..0ba034b --- /dev/null +++ b/fs/aufs/dcsub.h @@ -0,0 +1,47 @@ +/* + * Copyright (C) 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: dcsub.h,v 1.2 2007/05/14 03:41:52 sfjro Exp $ */ + +#ifndef __AUFS_DCSUB_H__ +#define __AUFS_DCSUB_H__ + +#ifdef __KERNEL__ + +#include + +struct au_dpage { + int ndentry; + struct dentry **dentries; +}; + +struct au_dcsub_pages { + int ndpage; + struct au_dpage *dpages; +}; + +/* ---------------------------------------------------------------------- */ + +int au_dpages_init(struct au_dcsub_pages *dpages, gfp_t gfp); +void au_dpages_free(struct au_dcsub_pages *dpages); +typedef int (*au_dpages_test)(struct dentry *dentry, void *arg); +int au_dcsub_pages(struct au_dcsub_pages *dpages, struct dentry *root, + au_dpages_test test, void *arg); + +#endif /* __KERNEL__ */ +#endif /* __AUFS_DCSUB_H__ */ diff --git a/fs/aufs/debug.c b/fs/aufs/debug.c new file mode 100755 index 0000000..99d158b --- /dev/null +++ b/fs/aufs/debug.c @@ -0,0 +1,262 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: debug.c,v 1.27 2007/04/30 05:48:23 sfjro Exp $ */ + +#include "aufs.h" + +atomic_t aufs_cond = ATOMIC_INIT(0); + +#if defined(CONFIG_LKTR) || defined(CONFIG_LKTR_MODULE) +#define dpri(fmt, arg...) \ + do {if (LktrCond) printk(KERN_DEBUG fmt, ##arg);} while (0) +#else +#define dpri(fmt, arg...) printk(KERN_DEBUG fmt, ##arg) +#endif + +/* ---------------------------------------------------------------------- */ + +void au_dpri_whlist(struct aufs_nhash *whlist) +{ + int i; + struct hlist_head *head; + struct aufs_wh *tpos; + struct hlist_node *pos; + + for (i = 0; i < AUFS_NHASH_SIZE; i++) { + head = whlist->heads + i; + hlist_for_each_entry(tpos, pos, head, wh_hash) + dpri("b%d, %.*s, %d\n", + tpos->wh_bindex, + tpos->wh_str.len, tpos->wh_str.name, + tpos->wh_str.len); + } +} + +void au_dpri_vdir(struct aufs_vdir *vdir) +{ + int i; + union aufs_deblk_p p; + unsigned char *o; + + if (!vdir || IS_ERR(vdir)) { + dpri("err %ld\n", PTR_ERR(vdir)); + return; + } + + dpri("nblk %d, deblk %p %d, last{%d, %p}, ver %lu\n", + vdir->vd_nblk, vdir->vd_deblk, ksize(vdir->vd_deblk), + vdir->vd_last.i, vdir->vd_last.p.p, vdir->vd_version); + for (i = 0; i < vdir->vd_nblk; i++) { + p.deblk = vdir->vd_deblk[i]; + o = p.p; + dpri("[%d]: %p %d\n", i, o, ksize(o)); +#if 0 // verbose + int j; + for (j = 0; j < 8; j++) { + dpri("%p(+%d) {%02x %02x %02x %02x %02x %02x %02x %02x " + "%02x %02x %02x %02x %02x %02x %02x %02x}\n", + p.p, p.p - o, + p.p[0], p.p[1], p.p[2], p.p[3], + p.p[4], p.p[5], p.p[6], p.p[7], + p.p[8], p.p[9], p.p[10], p.p[11], + p.p[12], p.p[13], p.p[14], p.p[15]); + p.p += 16; + } +#endif + } +} + +static int do_pri_inode(aufs_bindex_t bindex, struct inode *inode) +{ + if (!inode || IS_ERR(inode)) { + dpri("i%d: err %ld\n", bindex, PTR_ERR(inode)); + return -1; + } + + /* the type of i_blocks depends upon CONFIG_LSF */ + BUILD_BUG_ON(sizeof(inode->i_blocks) != sizeof(unsigned long) + && sizeof(inode->i_blocks) != sizeof(u64)); + dpri("i%d: i%lu, %s, cnt %d, nl %u, 0%o, sz %Lu, blk %Lu," + " ct %Ld, np %lu, st 0x%lx, g %x\n", + bindex, + inode->i_ino, inode->i_sb ? au_sbtype(inode->i_sb) : "??", + atomic_read(&inode->i_count), inode->i_nlink, inode->i_mode, + i_size_read(inode), (u64)inode->i_blocks, + timespec_to_ns(&inode->i_ctime) & 0x0ffff, + inode->i_mapping ? inode->i_mapping->nrpages : 0, + inode->i_state, inode->i_generation); + return 0; +} + +void au_dpri_inode(struct inode *inode) +{ + struct aufs_iinfo *iinfo; + aufs_bindex_t bindex; + int err; + + err = do_pri_inode(-1, inode); + if (err || !au_is_aufs(inode->i_sb)) + return; + + iinfo = itoii(inode); + if (!iinfo) + return; + dpri("i-1: bstart %d, bend %d, gen %d\n", + iinfo->ii_bstart, iinfo->ii_bend, au_iigen(inode)); + if (iinfo->ii_bstart < 0) + return; + for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend; bindex++) + do_pri_inode(bindex, iinfo->ii_hinode[0 + bindex].hi_inode); +} + +static int do_pri_dentry(aufs_bindex_t bindex, struct dentry *dentry) +{ + if (!dentry || IS_ERR(dentry)) { + dpri("d%d: err %ld\n", bindex, PTR_ERR(dentry)); + return -1; + } + dpri("d%d: %.*s/%.*s, %s, cnt %d, flags 0x%x\n", + bindex, + DLNPair(dentry->d_parent), DLNPair(dentry), + dentry->d_sb ? au_sbtype(dentry->d_sb) : "??", + atomic_read(&dentry->d_count), dentry->d_flags); + do_pri_inode(bindex, dentry->d_inode); + return 0; +} + +void au_dpri_dentry(struct dentry *dentry) +{ + struct aufs_dinfo *dinfo; + aufs_bindex_t bindex; + int err; + + err = do_pri_dentry(-1, dentry); + if (err || !au_is_aufs(dentry->d_sb)) + return; + + dinfo = dtodi(dentry); + if (!dinfo) + return; + dpri("d-1: bstart %d, bend %d, bwh %d, bdiropq %d, gen %d\n", + dinfo->di_bstart, dinfo->di_bend, + dinfo->di_bwh, dinfo->di_bdiropq, au_digen(dentry)); + if (dinfo->di_bstart < 0) + return; + for (bindex = dinfo->di_bstart; bindex <= dinfo->di_bend; bindex++) + do_pri_dentry(bindex, dinfo->di_hdentry[0 + bindex].hd_dentry); +} + +static int do_pri_file(aufs_bindex_t bindex, struct file *file) +{ + char a[32]; + + if (!file || IS_ERR(file)) { + dpri("f%d: err %ld\n", bindex, PTR_ERR(file)); + return -1; + } + a[0] = 0; + if (bindex == -1 && ftofi(file)) + snprintf(a, sizeof(a), ", mmapped %d", au_is_mmapped(file)); + dpri("f%d: mode 0x%x, flags 0%o, cnt %d, pos %Lu%s\n", + bindex, file->f_mode, file->f_flags, file_count(file), + file->f_pos, a); + do_pri_dentry(bindex, file->f_dentry); + return 0; +} + +void au_dpri_file(struct file *file) +{ + struct aufs_finfo *finfo; + aufs_bindex_t bindex; + int err; + + err = do_pri_file(-1, file); + if (err || !file->f_dentry || !au_is_aufs(file->f_dentry->d_sb)) + return; + + finfo = ftofi(file); + if (!finfo) + return; + if (finfo->fi_bstart < 0) + return; + for (bindex = finfo->fi_bstart; bindex <= finfo->fi_bend; bindex++) { + struct aufs_hfile *hf; + //dpri("bindex %d\n", bindex); + hf = finfo->fi_hfile + bindex; + do_pri_file(bindex, hf ? hf->hf_file : NULL); + } +} + +static int do_pri_br(aufs_bindex_t bindex, struct aufs_branch *br) +{ + struct vfsmount *mnt; + struct super_block *sb; + + if (!br || IS_ERR(br) + || !(mnt = br->br_mnt) || IS_ERR(mnt) + || !(sb = mnt->mnt_sb) || IS_ERR(sb)) { + dpri("s%d: err %ld\n", bindex, PTR_ERR(br)); + return -1; + } + + dpri("s%d: {perm 0x%x, cnt %d}, " + "%s, flags 0x%lx, cnt(BIAS) %d, active %d, xino %p %p\n", + bindex, br->br_perm, br_count(br), + au_sbtype(sb), sb->s_flags, sb->s_count - S_BIAS, + atomic_read(&sb->s_active), br->br_xino, + br->br_xino ? br->br_xino->f_dentry : NULL); + return 0; +} + +void au_dpri_sb(struct super_block *sb) +{ + struct aufs_sbinfo *sbinfo; + aufs_bindex_t bindex; + int err; + struct vfsmount mnt = {.mnt_sb = sb}; + struct aufs_branch fake = { + .br_perm = 0, + .br_mnt = &mnt, + .br_count = ATOMIC_INIT(0), + .br_xino = NULL + }; + + atomic_set(&fake.br_count, 0); + err = do_pri_br(-1, &fake); + dpri("dev 0x%x\n", sb->s_dev); + if (err || !au_is_aufs(sb)) + return; + + sbinfo = stosi(sb); + if (!sbinfo) + return; + for (bindex = 0; bindex <= sbinfo->si_bend; bindex++) { + //dpri("bindex %d\n", bindex); + do_pri_br(bindex, sbinfo->si_branch[0 + bindex]); + } +} + +/* ---------------------------------------------------------------------- */ + +void DbgSleep(int sec) +{ + static DECLARE_WAIT_QUEUE_HEAD(wq); + Dbg("sleep %d sec\n", sec); + wait_event_timeout(wq, 0, sec * HZ); +} diff --git a/fs/aufs/debug.h b/fs/aufs/debug.h new file mode 100755 index 0000000..53f5f6a --- /dev/null +++ b/fs/aufs/debug.h @@ -0,0 +1,129 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: debug.h,v 1.31 2007/05/14 03:41:52 sfjro Exp $ */ + +#ifndef __AUFS_DEBUG_H__ +#define __AUFS_DEBUG_H__ + +#ifdef __KERNEL__ + +#include + +#ifdef CONFIG_AUFS_DEBUG +#define DEBUG_ON(a) BUG_ON(a) +extern atomic_t aufs_cond; +#define au_debug_on() atomic_inc(&aufs_cond) +#define au_debug_off() atomic_dec(&aufs_cond) +#define au_is_debug() atomic_read(&aufs_cond) +#else +#define DEBUG_ON(a) /* */ +#define au_debug_on() /* */ +#define au_debug_off() /* */ +#define au_is_debug() 0 +#endif + +#define MtxMustLock(mtx) DEBUG_ON(!mutex_is_locked(mtx)) + +/* ---------------------------------------------------------------------- */ + +/* debug print */ +#if defined(CONFIG_LKTR) || defined(CONFIG_LKTR_MODULE) +#include +#ifdef CONFIG_AUFS_DEBUG +#undef LktrCond +#define LktrCond unlikely((lktr_cond && lktr_cond()) || au_is_debug()) +#endif +#else +#define LktrCond au_is_debug() +#define LKTRDumpVma(pre, vma, suf) /* */ +#define LKTRDumpStack() /* */ +#define LKTRTrace(fmt, args...) do { \ + if (LktrCond) \ + Dbg(fmt, ##args); \ +} while (0) +#define LKTRLabel(label) LKTRTrace("%s\n", #label) +#endif /* CONFIG_LKTR */ + +#define TraceErr(e) do { \ + if (unlikely((e) < 0)) \ + LKTRTrace("err %d\n", (int)(e)); \ +} while (0) +#define TraceErrPtr(p) do { \ + if (IS_ERR(p)) \ + LKTRTrace("err %ld\n", PTR_ERR(p)); \ +} while (0) +#define TraceEnter() LKTRLabel(enter) + +/* dirty macros for debug print, use with "%.*s" and caution */ +#define LNPair(qstr) (qstr)->len,(qstr)->name +#define DLNPair(d) LNPair(&(d)->d_name) + +/* ---------------------------------------------------------------------- */ + +#define Dpri(lvl, fmt, arg...) \ + printk(lvl AUFS_NAME " %s:%d:%s[%d]: " fmt, \ + __func__, __LINE__, current->comm, current->pid, ##arg) +#define Dbg(fmt, arg...) Dpri(KERN_DEBUG, fmt, ##arg) +#define Warn(fmt, arg...) Dpri(KERN_WARNING, fmt, ##arg) +#define Warn1(fmt, arg...) do { \ + static unsigned char c; \ + if (!c++) Warn(fmt, ##arg); \ + } while (0) +#define Err(fmt, arg...) Dpri(KERN_ERR, fmt, ##arg) +#define Err1(fmt, arg...) do { \ + static unsigned char c; \ + if (!c++) Err(fmt, ##arg); \ + } while (0) +#define IOErr(fmt, arg...) Err("I/O Error, " fmt, ##arg) +#define IOErr1(fmt, arg...) do { \ + static unsigned char c; \ + if (!c++) IOErr(fmt, ##arg); \ + } while (0) +#define IOErrWhck(fmt, arg...) Err("I/O Error, try whck. " fmt, ##arg) + +/* ---------------------------------------------------------------------- */ + +#ifdef CONFIG_AUFS_DEBUG +struct aufs_nhash; +void au_dpri_whlist(struct aufs_nhash *whlist); +struct aufs_vdir; +void au_dpri_vdir(struct aufs_vdir *vdir); +void au_dpri_inode(struct inode *inode); +void au_dpri_dentry(struct dentry *dentry); +void au_dpri_file(struct file *filp); +void au_dpri_sb(struct super_block *sb); +#define DbgWhlist(w) do{LKTRTrace(#w "\n"); au_dpri_whlist(w);}while(0) +#define DbgVdir(v) do{LKTRTrace(#v "\n"); au_dpri_vdir(v);}while(0) +#define DbgInode(i) do{LKTRTrace(#i "\n"); au_dpri_inode(i);}while(0) +#define DbgDentry(d) do{LKTRTrace(#d "\n"); au_dpri_dentry(d);}while(0) +#define DbgFile(f) do{LKTRTrace(#f "\n"); au_dpri_file(f);}while(0) +#define DbgSb(sb) do{LKTRTrace(#sb "\n"); au_dpri_sb(sb);}while(0) +void DbgSleep(int sec); +#else +#define DbgWhlist(w) /* */ +#define DbgVdir(v) /* */ +#define DbgInode(i) /* */ +#define DbgDentry(d) /* */ +#define DbgFile(f) /* */ +#define DbgSb(sb) /* */ +#define DbgSleep(sec) /* */ +#endif + +#endif /* __KERNEL__ */ +#endif /* __AUFS_DEBUG_H__ */ diff --git a/fs/aufs/dentry.c b/fs/aufs/dentry.c new file mode 100755 index 0000000..2acb89b --- /dev/null +++ b/fs/aufs/dentry.c @@ -0,0 +1,946 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: dentry.c,v 1.41 2007/05/14 03:38:38 sfjro Exp $ */ + +//#include +//#include +#include "aufs.h" + +#ifdef CONFIG_AUFS_LHASH_PATCH + +#ifdef CONFIG_AUFS_DLGT +struct lookup_hash_args { + struct dentry **errp; + struct qstr *name; + struct dentry *base; + struct nameidata *nd; +}; + +static void call_lookup_hash(void *args) +{ + struct lookup_hash_args *a = args; + *a->errp = __lookup_hash(a->name, a->base, a->nd); +} +#endif /* CONFIG_AUFS_DLGT */ + +static struct dentry *lkup_hash(const char *name, struct dentry *parent, + int len, struct lkup_args *lkup) +{ + struct dentry *dentry; + char *p; + unsigned long hash; + struct qstr this; + unsigned int c; + struct nameidata tmp_nd; + + dentry = ERR_PTR(-EACCES); + this.name = name; + this.len = len; + if (unlikely(!len)) + goto out; + + p = (void*)name; + hash = init_name_hash(); + while (len--) { + c = *p++; + if (unlikely(c == '/' || c == '\0')) + goto out; + hash = partial_name_hash(c, hash); + } + this.hash = end_name_hash(hash); + + memset(&tmp_nd, 0, sizeof(tmp_nd)); + tmp_nd.dentry = dget(parent); + tmp_nd.mnt = mntget(lkup->nfsmnt); +#ifndef CONFIG_AUFS_DLGT + dentry = __lookup_hash(&this, parent, &tmp_nd); +#else + if (!lkup->dlgt) + dentry = __lookup_hash(&this, parent, &tmp_nd); + else { + struct lookup_hash_args args = { + .errp = &dentry, + .name = &this, + .base = parent, + .nd = &tmp_nd + }; + au_wkq_wait(call_lookup_hash, &args, /*dlgt*/1); + } +#endif + path_release(&tmp_nd); + + out: + TraceErrPtr(dentry); + return dentry; +} +#elif defined(CONFIG_AUFS_DLGT) +static struct dentry *lkup_hash(const char *name, struct dentry *parent, + int len, struct lkup_args *lkup) +{ + return ERR_PTR(-ENOSYS); +} +#endif + +#ifdef CONFIG_AUFS_DLGT +struct lookup_one_len_args { + struct dentry **errp; + const char *name; + struct dentry *parent; + int len; +}; + +static void call_lookup_one_len(void *args) +{ + struct lookup_one_len_args *a = args; + *a->errp = lookup_one_len(a->name, a->parent, a->len); +} +#endif /* CONFIG_AUFS_DLGT */ + +#if defined(CONFIG_AUFS_LHASH_PATCH) || defined(CONFIG_AUFS_DLGT) +/* cf. lookup_one_len() in linux/fs/namei.c */ +struct dentry *lkup_one(const char *name, struct dentry *parent, int len, + struct lkup_args *lkup) +{ + struct dentry *dentry; + + LKTRTrace("%.*s/%.*s, lkup{%p, %d}\n", + DLNPair(parent), len, name, lkup->nfsmnt, lkup->dlgt); + + if (!lkup->nfsmnt) { +#ifndef CONFIG_AUFS_DLGT + dentry = lookup_one_len(name, parent, len); +#else + if (!lkup->dlgt) + dentry = lookup_one_len(name, parent, len); + else { + struct lookup_one_len_args args = { + .errp = &dentry, + .name = name, + .parent = parent, + .len = len + }; + au_wkq_wait(call_lookup_one_len, &args, /*dlgt*/1); + } +#endif + } else + dentry = lkup_hash(name, parent, len, lkup); + + TraceErrPtr(dentry); + return dentry; +} +#endif + +struct lkup_one_args { + struct dentry **errp; + const char *name; + struct dentry *parent; + int len; + struct lkup_args *lkup; +}; + +static void call_lkup_one(void *args) +{ + struct lkup_one_args *a = args; + *a->errp = lkup_one(a->name, a->parent, a->len, a->lkup); +} + +/* + * returns positive/negative dentry, NULL or an error. + * NULL means whiteout-ed or not-found. + */ +static struct dentry *do_lookup(struct dentry *hidden_parent, + struct dentry *dentry, aufs_bindex_t bindex, + struct qstr *wh_name, int allow_neg, + mode_t type, int dlgt) +{ + struct dentry *hidden_dentry; + int wh_found, wh_able, opq; + struct inode *hidden_dir, *hidden_inode; + struct qstr *name; + struct super_block *sb; + struct lkup_args lkup = {.dlgt = dlgt}; + + LKTRTrace("%.*s/%.*s, b%d, allow_neg %d, type 0%o, dlgt %d\n", + DLNPair(hidden_parent), DLNPair(dentry), bindex, allow_neg, + type, dlgt); + DEBUG_ON(IS_ROOT(dentry)); + hidden_dir = hidden_parent->d_inode; + IMustLock(hidden_dir); + + wh_found = 0; + sb = dentry->d_sb; + wh_able = sbr_is_whable(sb, bindex); + lkup.nfsmnt = au_nfsmnt(sb, bindex); + name = &dentry->d_name; + if (unlikely(wh_able)) { +#if 0 //def CONFIG_AUFS_ROBR + if (strncmp(name->name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) + wh_found = is_wh(hidden_parent, wh_name, /*try_sio*/0, + &lkup); + else + wh_found = -EPERM; +#else + wh_found = is_wh(hidden_parent, wh_name, /*try_sio*/0, &lkup); +#endif + } + //if (LktrCond) wh_found = -1; + hidden_dentry = ERR_PTR(wh_found); + if (!wh_found) + goto real_lookup; + if (unlikely(wh_found < 0)) + goto out; + + /* We found a whiteout */ + //set_dbend(dentry, bindex); + set_dbwh(dentry, bindex); + if (!allow_neg) + return NULL; /* success */ + + real_lookup: + // do not superio. + hidden_dentry = lkup_one(name->name, hidden_parent, name->len, &lkup); + //if (LktrCond) {dput(hidden_dentry); hidden_dentry = ERR_PTR(-1);} + if (IS_ERR(hidden_dentry)) + goto out; + DEBUG_ON(d_unhashed(hidden_dentry)); + hidden_inode = hidden_dentry->d_inode; + if (!hidden_inode) { + if (!allow_neg) + goto out_neg; + } else if (wh_found + || (type && type != (hidden_inode->i_mode & S_IFMT))) + goto out_neg; + + if (dbend(dentry) <= bindex) + set_dbend(dentry, bindex); + if (dbstart(dentry) == -1 || bindex < dbstart(dentry)) + set_dbstart(dentry, bindex); + set_h_dptr(dentry, bindex, hidden_dentry); + + if (!hidden_inode || !S_ISDIR(hidden_inode->i_mode) || !wh_able) + return hidden_dentry; /* success */ + + hi_lock_child(hidden_inode); + opq = is_diropq(hidden_dentry, &lkup); + //if (LktrCond) opq = -1; + i_unlock(hidden_inode); + if (opq > 0) + set_dbdiropq(dentry, bindex); + else if (unlikely(opq < 0)) { + set_h_dptr(dentry, bindex, NULL); + hidden_dentry = ERR_PTR(opq); + } + goto out; + + out_neg: + dput(hidden_dentry); + hidden_dentry = NULL; + out: + TraceErrPtr(hidden_dentry); + return hidden_dentry; +} + +/* + * returns the number of hidden positive dentries, + * otherwise an error. + */ +int lkup_dentry(struct dentry *dentry, aufs_bindex_t bstart, mode_t type) +{ + int npositive, err, allow_neg, dlgt; + struct dentry *parent; + aufs_bindex_t bindex, btail; + const struct qstr *name = &dentry->d_name; + struct qstr whname; + struct super_block *sb; + + LKTRTrace("%.*s, b%d, type 0%o\n", LNPair(name), bstart, type); + DEBUG_ON(bstart < 0 || IS_ROOT(dentry)); + parent = dget_parent(dentry); + +#if 1 //ndef CONFIG_AUFS_ROBR + err = -EPERM; + if (unlikely(!strncmp(name->name, AUFS_WH_PFX, AUFS_WH_PFX_LEN))) + goto out; +#endif + + err = au_alloc_whname(name->name, name->len, &whname); + //if (LktrCond) {au_free_whname(&whname); err = -1;} + if (unlikely(err)) + goto out; + + sb = dentry->d_sb; + dlgt = need_dlgt(sb); + allow_neg = !type; + npositive = 0; + btail = dbtaildir(parent); + for (bindex = bstart; bindex <= btail; bindex++) { + struct dentry *hidden_parent, *hidden_dentry; + struct inode *hidden_inode; + struct inode *hidden_dir; + + hidden_dentry = au_h_dptr_i(dentry, bindex); + if (hidden_dentry) { + if (hidden_dentry->d_inode) + npositive++; + if (type != S_IFDIR) + break; + continue; + } + hidden_parent = au_h_dptr_i(parent, bindex); + if (!hidden_parent) + continue; + hidden_dir = hidden_parent->d_inode; + if (!hidden_dir || !S_ISDIR(hidden_dir->i_mode)) + continue; + + hi_lock_parent(hidden_dir); + hidden_dentry = do_lookup(hidden_parent, dentry, bindex, + &whname, allow_neg, type, dlgt); + // do not dput for testing + //if (LktrCond) {hidden_dentry = ERR_PTR(-1);} + i_unlock(hidden_dir); + err = PTR_ERR(hidden_dentry); + if (IS_ERR(hidden_dentry)) + goto out_wh; + allow_neg = 0; + + if (dbwh(dentry) != -1) + break; + if (!hidden_dentry) + continue; + hidden_inode = hidden_dentry->d_inode; + if (!hidden_inode) + continue; + npositive++; + if (!type) + type = hidden_inode->i_mode & S_IFMT; + if (type != S_IFDIR) + break; + else if (dbdiropq(dentry) != -1) + break; + } + + if (npositive) { + LKTRLabel(positive); + au_update_dbstart(dentry); + } + err = npositive; + + out_wh: + au_free_whname(&whname); + out: + dput(parent); + TraceErr(err); + return err; +} + +struct dentry *sio_lkup_one(const char *name, struct dentry *parent, int len, + struct lkup_args *lkup) +{ + struct dentry *dentry; + + LKTRTrace("%.*s/%.*s\n", DLNPair(parent), len, name); + IMustLock(parent->d_inode); + + if (!au_test_perm(parent->d_inode, MAY_EXEC, lkup->dlgt)) + dentry = lkup_one(name, parent, len, lkup); + else { + // ugly + int dlgt = lkup->dlgt; + struct lkup_one_args args = { + .errp = &dentry, + .name = name, + .parent = parent, + .len = len, + .lkup = lkup + }; + + lkup->dlgt = 0; + au_wkq_wait(call_lkup_one, &args, /*dlgt*/0); + lkup->dlgt = dlgt; + } + + TraceErrPtr(dentry); + return dentry; +} + +/* + * lookup @dentry on @bindex which should be negative. + */ +int lkup_neg(struct dentry *dentry, aufs_bindex_t bindex) +{ + int err; + struct dentry *parent, *hidden_parent, *hidden_dentry; + struct inode *hidden_dir; + struct lkup_args lkup; + + LKTRTrace("%.*s, b%d\n", DLNPair(dentry), bindex); + parent = dget_parent(dentry); + DEBUG_ON(!parent || !parent->d_inode + || !S_ISDIR(parent->d_inode->i_mode)); + hidden_parent = au_h_dptr_i(parent, bindex); + DEBUG_ON(!hidden_parent); + hidden_dir = hidden_parent->d_inode; + DEBUG_ON(!hidden_dir || !S_ISDIR(hidden_dir->i_mode)); + IMustLock(hidden_dir); + + lkup.nfsmnt = au_nfsmnt(dentry->d_sb, bindex); + lkup.dlgt = need_dlgt(dentry->d_sb); + hidden_dentry = sio_lkup_one(dentry->d_name.name, hidden_parent, + dentry->d_name.len, &lkup); + //if (LktrCond) {dput(hidden_dentry); hidden_dentry = ERR_PTR(-1);} + err = PTR_ERR(hidden_dentry); + if (IS_ERR(hidden_dentry)) + goto out; + if (unlikely(hidden_dentry->d_inode)) { + err = -EIO; + IOErr("b%d %.*s should be negative.%s\n", + bindex, DLNPair(hidden_dentry), + au_flag_test(dentry->d_sb, AuFlag_UDBA_INOTIFY) ? "" : + " Try udba=inotify."); + dput(hidden_dentry); + goto out; + } + + if (bindex < dbstart(dentry)) + set_dbstart(dentry, bindex); + if (dbend(dentry) < bindex) + set_dbend(dentry, bindex); + set_h_dptr(dentry, bindex, hidden_dentry); + err = 0; + + out: + dput(parent); + TraceErr(err); + return err; +} + +/* + * returns the number of found hidden positive dentries, + * otherwise an error. + */ +int au_refresh_hdentry(struct dentry *dentry, mode_t type) +{ + int npositive, pgen, new_sz, sgen, dgen; + struct aufs_dinfo *dinfo; + struct super_block *sb; + struct dentry *parent; + aufs_bindex_t bindex, parent_bend, parent_bstart, bwh, bdiropq, bend; + struct aufs_hdentry *p; + //struct nameidata nd; + + LKTRTrace("%.*s, type 0%o\n", DLNPair(dentry), type); + DiMustWriteLock(dentry); + sb = dentry->d_sb; + DEBUG_ON(IS_ROOT(dentry)); + parent = dget_parent(dentry); + pgen = au_digen(parent); + sgen = au_sigen(sb); + dgen = au_digen(dentry); + DEBUG_ON(pgen != sgen); + + npositive = -ENOMEM; + new_sz = sizeof(*dinfo->di_hdentry) * (sbend(sb) + 1); + dinfo = dtodi(dentry); + p = au_kzrealloc(dinfo->di_hdentry, sizeof(*p) * (dinfo->di_bend + 1), + new_sz, GFP_KERNEL); + //p = NULL; + if (unlikely(!p)) + goto out; + dinfo->di_hdentry = p; + + bend = dinfo->di_bend; + bwh = dinfo->di_bwh; + bdiropq = dinfo->di_bdiropq; + p += dinfo->di_bstart; + for (bindex = dinfo->di_bstart; bindex <= bend; bindex++, p++) { + struct dentry *hd, *hdp; + struct aufs_hdentry tmp, *q; + aufs_bindex_t new_bindex; + + hd = p->hd_dentry; + if (!hd) + continue; + hdp = dget_parent(hd); + if (hdp == au_h_dptr_i(parent, bindex)) { + dput(hdp); + continue; + } + + new_bindex = au_find_dbindex(parent, hdp); + dput(hdp); + DEBUG_ON(new_bindex == bindex); + if (dinfo->di_bwh == bindex) + bwh = new_bindex; + if (dinfo->di_bdiropq == bindex) + bdiropq = new_bindex; + if (new_bindex < 0) { // test here + hdput(p); + p->hd_dentry = NULL; + continue; + } + /* swap two hidden dentries, and loop again */ + q = dinfo->di_hdentry + new_bindex; + tmp = *q; + *q = *p; + *p = tmp; + if (tmp.hd_dentry) { + bindex--; + p--; + } + } + + // test here + dinfo->di_bwh = -1; + if (unlikely(bwh != -1 && bwh <= sbend(sb) && sbr_is_whable(sb, bwh))) + dinfo->di_bwh = bwh; + dinfo->di_bdiropq = -1; + if (unlikely(bdiropq != -1 && bdiropq <= sbend(sb) + && sbr_is_whable(sb, bdiropq))) + dinfo->di_bdiropq = bdiropq; + parent_bend = dbend(parent); + p = dinfo->di_hdentry; + for (bindex = 0; bindex <= parent_bend; bindex++, p++) + if (p->hd_dentry) { + dinfo->di_bstart = bindex; + break; + } + p = dinfo->di_hdentry + parent_bend; + //for (bindex = parent_bend; bindex > dinfo->di_bstart; bindex--, p--) + for (bindex = parent_bend; bindex >= 0; bindex--, p--) + if (p->hd_dentry) { + dinfo->di_bend = bindex; + break; + } + + npositive = 0; + parent_bstart = dbstart(parent); + if (type != S_IFDIR && dinfo->di_bstart == parent_bstart) + goto out_dgen; /* success */ + +#if 0 + nd.last_type = LAST_ROOT; + nd.flags = LOOKUP_FOLLOW; + nd.depth = 0; + nd.mnt = mntget(??); + nd.dentry = dget(parent); +#endif + npositive = lkup_dentry(dentry, parent_bstart, type); + //if (LktrCond) npositive = -1; + if (npositive < 0) + goto out; + + out_dgen: + au_update_digen(dentry); + out: + dput(parent); + TraceErr(npositive); + return npositive; +} + +static int h_d_revalidate(struct dentry *dentry, struct nameidata *nd, + int do_udba) +{ + int err, plus, locked, unhashed, is_root, h_plus, is_nfs; + struct nameidata fake_nd, *p; + aufs_bindex_t bindex, btail, bstart, ibs, ibe; + struct super_block *sb; + struct inode *inode, *first, *h_inode, *h_cached_inode; + umode_t mode, h_mode; + struct dentry *h_dentry; + int (*reval)(struct dentry *, struct nameidata *); + struct qstr *name; + + LKTRTrace("%.*s\n", DLNPair(dentry)); + inode = dentry->d_inode; + DEBUG_ON(inode && au_digen(dentry) != au_iigen(inode)); + //DbgDentry(dentry); + //DbgInode(inode); + + err = 0; + sb = dentry->d_sb; + plus = 0; + mode = 0; + first = NULL; + ibs = ibe = -1; + unhashed = d_unhashed(dentry); + is_root = IS_ROOT(dentry); + name = &dentry->d_name; + + /* + * Theoretically, REVAL test should be unnecessary in case of INOTIFY. + * But inotify doesn't fire some necessary events, + * IN_ATTRIB for atime/nlink/pageio + * IN_DELETE for NFS dentry + * Let's do REVAL test too. + */ + if (do_udba && inode) { + mode = (inode->i_mode & S_IFMT); + plus = (inode->i_nlink > 0); + first = au_h_iptr(inode); + ibs = ibstart(inode); + ibe = ibend(inode); + } + + btail = bstart = dbstart(dentry); + if (inode && S_ISDIR(inode->i_mode)) + btail = dbtaildir(dentry); + locked = 0; + if (nd) { + fake_nd = *nd; +#ifndef CONFIG_AUFS_FAKE_DM + if (dentry != nd->dentry) { + di_read_lock_parent(nd->dentry, 0); + locked = 1; + } +#endif + } + for (bindex = bstart; bindex <= btail; bindex++) { + h_dentry = au_h_dptr_i(dentry, bindex); + if (unlikely(!h_dentry)) + continue; + if (unlikely(do_udba + && !is_root + && (unhashed != d_unhashed(h_dentry) +#if 1 + || name->len != h_dentry->d_name.len + || memcmp(name->name, h_dentry->d_name.name, + name->len) +#endif + ))) { + LKTRTrace("unhash 0x%x 0x%x, %.*s %.*s\n", + unhashed, d_unhashed(h_dentry), + DLNPair(dentry), DLNPair(h_dentry)); + goto err; + } + + reval = NULL; + if (h_dentry->d_op) + reval = h_dentry->d_op->d_revalidate; + if (unlikely(reval)) { + //LKTRLabel(hidden reval); + p = fake_dm(&fake_nd, nd, sb, bindex); + DEBUG_ON(IS_ERR(p)); + err = !reval(h_dentry, p); + fake_dm_release(p); + if (unlikely(err)) { + //Dbg("here\n"); + goto err; + } + } + + if (unlikely(!do_udba)) + continue; + + /* UDBA tests */ + h_inode = h_dentry->d_inode; + if (unlikely(!!inode != !!h_inode)) { + //Dbg("here\n"); + goto err; + } + + h_plus = plus; + h_mode = mode; + h_cached_inode = h_inode; + is_nfs = 0; + if (h_inode) { + h_mode = (h_inode->i_mode & S_IFMT); + h_plus = (h_inode->i_nlink > 0); + } + if (inode && ibs <= bindex && bindex <= ibe) { + h_cached_inode = au_h_iptr_i(inode, bindex); + //is_nfs = au_is_nfs(h_cached_inode->i_sb); + } + + LKTRTrace("{%d, 0%o, %p}, h{%d, 0%o, %p}\n", + plus, mode, h_cached_inode, + h_plus, h_mode, h_inode); + if (unlikely(plus != h_plus || mode != h_mode + || (h_cached_inode != h_inode /* && !is_nfs */))) { + //Dbg("here\n"); + goto err; + } + continue; + + err: + err = -EINVAL; + break; + } +#ifndef CONFIG_AUFS_FAKE_DM + if (unlikely(locked)) + di_read_unlock(nd->dentry, 0); +#endif + +#if 0 + // some filesystem uses CURRENT_TIME_SEC instead of CURRENT_TIME. + // NFS may stop IN_DELETE because of DCACHE_NFSFS_RENAMED. +#if 0 + && (!timespec_equal(&inode->i_ctime, &first->i_ctime) + || !timespec_equal(&inode->i_atime, &first->i_atime)) +#endif + if (unlikely(!err && udba && first)) + au_cpup_attr_all(inode); +#endif + + TraceErr(err); + return err; +} + +static int simple_reval_dpath(struct dentry *dentry, int sgen) +{ + int err; + mode_t type; + struct dentry *parent; + struct inode *inode; + + LKTRTrace("%.*s, sgen %d\n", DLNPair(dentry), sgen); + SiMustAnyLock(dentry->d_sb); + DiMustWriteLock(dentry); + inode = dentry->d_inode; + DEBUG_ON(!inode); + + if (au_digen(dentry) == sgen) + return 0; + + parent = dget_parent(dentry); + di_read_lock_parent(parent, AUFS_I_RLOCK); + DEBUG_ON(au_digen(parent) != sgen); +#ifdef CONFIG_AUFS_DEBUG + { + struct dentry *d = parent; + while (!IS_ROOT(d)) { + DEBUG_ON(au_digen(d) != sgen); + d = d->d_parent; + } + } +#endif + type = (inode->i_mode & S_IFMT); + /* returns a number of positive dentries */ + err = au_refresh_hdentry(dentry, type); + if (err >= 0) + err = au_refresh_hinode(inode, dentry); + di_read_unlock(parent, AUFS_I_RLOCK); + dput(parent); + TraceErr(err); + return err; +} + +int au_reval_dpath(struct dentry *dentry, int sgen) +{ + int err; + struct dentry *d, *parent; + struct inode *inode; + + LKTRTrace("%.*s, sgen %d\n", DLNPair(dentry), sgen); + DEBUG_ON(!dentry->d_inode); + DiMustWriteLock(dentry); + + if (!stosi(dentry->d_sb)->si_failed_refresh_dirs) + return simple_reval_dpath(dentry, sgen); + + /* slow loop, keep it simple and stupid */ + /* cf: cpup_dirs() */ + err = 0; + while (au_digen(dentry) != sgen) { + d = dentry; + while (1) { + parent = d->d_parent; // dget_parent() + if (au_digen(parent) == sgen) + break; + d = parent; + } + + inode = d->d_inode; + if (d != dentry) { + //i_lock(inode); + di_write_lock_child(d); + } + + /* someone might update our dentry while we were sleeping */ + if (au_digen(d) != sgen) { + di_read_lock_parent(parent, AUFS_I_RLOCK); + /* returns a number of positive dentries */ + err = au_refresh_hdentry(d, inode->i_mode & S_IFMT); + //err = -1; + if (err >= 0) + err = au_refresh_hinode(inode, d); + //err = -1; + di_read_unlock(parent, AUFS_I_RLOCK); + } + + if (d != dentry) { + di_write_unlock(d); + //i_unlock(inode); + } + if (unlikely(err)) + break; + } + + TraceErr(err); + return err; +} + +/* + * THIS IS A BOOLEAN FUNCTION: returns 1 if valid, 0 otherwise. + * nfsd passes NULL as nameidata. + */ +static int aufs_d_revalidate(struct dentry *dentry, struct nameidata *nd) +{ + int valid, sgen, err, do_udba; + struct super_block *sb; + struct inode *inode; + + LKTRTrace("dentry %.*s\n", DLNPair(dentry)); + if (nd && nd->dentry) + LKTRTrace("nd %.*s\n", DLNPair(nd->dentry)); + //dir case: DEBUG_ON(dentry->d_parent != nd->dentry); + //remove failure case: DEBUG_ON(!IS_ROOT(dentry) && d_unhashed(dentry)); + DEBUG_ON(!dentry->d_fsdata); + //DbgDentry(dentry); + + err = -EINVAL; + inode = dentry->d_inode; + //DbgInode(inode); + sb = dentry->d_sb; + si_read_lock(sb); + sgen = au_sigen(sb); + if (au_digen(dentry) == sgen) + di_read_lock_child(dentry, !AUFS_I_RLOCK); + else { + DEBUG_ON(IS_ROOT(dentry)); +#ifdef ForceInotify + Dbg("UDBA or digen, %.*s\n", DLNPair(dentry)); +#endif + //i_lock(inode); + di_write_lock_child(dentry); + if (inode) + err = au_reval_dpath(dentry, sgen); + //err = -1; + di_downgrade_lock(dentry, AUFS_I_RLOCK); + //i_unlock(inode); + if (unlikely(err)) + goto out; + ii_read_unlock(inode); + DEBUG_ON(au_iigen(inode) != sgen); + } + + if (inode) { + if (au_iigen(inode) == sgen) + ii_read_lock_child(inode); + else { + DEBUG_ON(IS_ROOT(dentry)); +#ifdef ForceInotify + Dbg("UDBA or survived, %.*s\n", DLNPair(dentry)); +#endif + ii_write_lock_child(inode); + err = au_refresh_hinode(inode, dentry); + ii_downgrade_lock(inode); + if (unlikely(err)) + goto out; + DEBUG_ON(au_iigen(inode) != sgen); + } + } + +#if 0 // fix it + /* parent dir i_nlink is not updated in the case of setattr */ + if (S_ISDIR(inode->i_mode)) { + i_lock(inode); + ii_write_lock(inode); + au_cpup_attr_nlink(inode); + ii_write_unlock(inode); + i_unlock(inode); + } +#endif + + err = -EINVAL; + do_udba = !au_flag_test(sb, AuFlag_UDBA_NONE); + if (do_udba && inode && ibstart(inode) >= 0 + && au_test_higen(inode, au_h_iptr(inode))) + goto out; + err = h_d_revalidate(dentry, nd, do_udba); + //err = -1; + + out: + aufs_read_unlock(dentry, AUFS_I_RLOCK); + TraceErr(err); + valid = !err; + //au_debug_on(); + if (!valid) + LKTRTrace("%.*s invalid\n", DLNPair(dentry)); + //au_debug_off(); + return valid; +} + +static void aufs_d_release(struct dentry *dentry) +{ + struct aufs_dinfo *dinfo; + aufs_bindex_t bend, bindex; + + LKTRTrace("%.*s\n", DLNPair(dentry)); + DEBUG_ON(!d_unhashed(dentry)); + + dinfo = dentry->d_fsdata; + if (unlikely(!dinfo)) + return; + + /* dentry may not be revalidated */ + bindex = dinfo->di_bstart; + if (bindex >= 0) { + struct aufs_hdentry *p; + bend = dinfo->di_bend; + DEBUG_ON(bend < bindex); + p = dinfo->di_hdentry + bindex; + while (bindex++ <= bend) { + if (p->hd_dentry) + hdput(p); + p++; + } + } + kfree(dinfo->di_hdentry); + cache_free_dinfo(dinfo); +} + +#if 0 +/* it may be called at remount time, too */ +static void aufs_d_iput(struct dentry *dentry, struct inode *inode) +{ + struct super_block *sb; + + LKTRTrace("%.*s, i%lu\n", DLNPair(dentry), inode->i_ino); + + sb = dentry->d_sb; +#if 0 + si_read_lock(sb); + if (unlikely(au_flag_test(sb, AuFlag_PLINK) + && au_is_plinked(sb, inode))) { + ii_write_lock(inode); + au_update_brange(inode, 1); + ii_write_unlock(inode); + } + si_read_unlock(sb); +#endif + iput(inode); +} +#endif + +struct dentry_operations aufs_dop = { + .d_revalidate = aufs_d_revalidate, + .d_release = aufs_d_release + //.d_iput = aufs_d_iput +}; diff --git a/fs/aufs/dentry.h b/fs/aufs/dentry.h new file mode 100755 index 0000000..78049e3 --- /dev/null +++ b/fs/aufs/dentry.h @@ -0,0 +1,183 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: dentry.h,v 1.25 2007/05/14 03:41:52 sfjro Exp $ */ + +#ifndef __AUFS_DENTRY_H__ +#define __AUFS_DENTRY_H__ + +#ifdef __KERNEL__ + +#include +#include +#include "misc.h" + +struct aufs_hdentry { + struct dentry *hd_dentry; +}; + +struct aufs_dinfo { + atomic_t di_generation; + + struct aufs_rwsem di_rwsem; + aufs_bindex_t di_bstart, di_bend, di_bwh, di_bdiropq; + struct aufs_hdentry *di_hdentry; +}; + +struct lkup_args { + struct vfsmount *nfsmnt; + int dlgt; + //struct super_block *sb; +}; + +/* ---------------------------------------------------------------------- */ + +/* dentry.c */ +#if defined(CONFIG_AUFS_LHASH_PATCH) || defined(CONFIG_AUFS_DLGT) +struct dentry *lkup_one(const char *name, struct dentry *parent, int len, + struct lkup_args *lkup); +#else +static inline +struct dentry *lkup_one(const char *name, struct dentry *parent, int len, + struct lkup_args *lkup) +{ + return lookup_one_len(name, parent, len); +} +#endif + +extern struct dentry_operations aufs_dop; +struct dentry *sio_lkup_one(const char *name, struct dentry *parent, int len, + struct lkup_args *lkup); +int lkup_dentry(struct dentry *dentry, aufs_bindex_t bstart, mode_t type); +int lkup_neg(struct dentry *dentry, aufs_bindex_t bindex); +int au_refresh_hdentry(struct dentry *dentry, mode_t type); +int au_reval_dpath(struct dentry *dentry, int sgen); + +/* dinfo.c */ +int au_alloc_dinfo(struct dentry *dentry); +struct aufs_dinfo *dtodi(struct dentry *dentry); + +void di_read_lock(struct dentry *d, int flags, unsigned int lsc); +void di_read_unlock(struct dentry *d, int flags); +void di_downgrade_lock(struct dentry *d, int flags); +void di_write_lock(struct dentry *d, unsigned int lsc); +void di_write_unlock(struct dentry *d); +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir); +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir); +void di_write_unlock2(struct dentry *d1, struct dentry *d2); + +aufs_bindex_t dbstart(struct dentry *dentry); +aufs_bindex_t dbend(struct dentry *dentry); +aufs_bindex_t dbwh(struct dentry *dentry); +aufs_bindex_t dbdiropq(struct dentry *dentry); +struct dentry *au_h_dptr_i(struct dentry *dentry, aufs_bindex_t bindex); +struct dentry *au_h_dptr(struct dentry *dentry); + +aufs_bindex_t dbtail(struct dentry *dentry); +aufs_bindex_t dbtaildir(struct dentry *dentry); +aufs_bindex_t dbtail_generic(struct dentry *dentry); + +void set_dbstart(struct dentry *dentry, aufs_bindex_t bindex); +void set_dbend(struct dentry *dentry, aufs_bindex_t bindex); +void set_dbwh(struct dentry *dentry, aufs_bindex_t bindex); +void set_dbdiropq(struct dentry *dentry, aufs_bindex_t bindex); +void hdput(struct aufs_hdentry *hdentry); +void set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex, + struct dentry *h_dentry); + +void au_update_digen(struct dentry *dentry); +void au_update_dbstart(struct dentry *dentry); +int au_find_dbindex(struct dentry *dentry, struct dentry *h_dentry); + +/* ---------------------------------------------------------------------- */ + +static inline int au_digen(struct dentry *d) +{ + return atomic_read(&dtodi(d)->di_generation); +} + +#ifdef CONFIG_AUFS_HINOTIFY +static inline void au_digen_dec(struct dentry *d) +{ + atomic_dec(&dtodi(d)->di_generation); +} +#endif /* CONFIG_AUFS_HINOTIFY */ + +/* ---------------------------------------------------------------------- */ + +/* lock subclass for dinfo */ +enum { + AuLsc_DI_CHILD, /* child first */ + AuLsc_DI_CHILD2, /* rename(2), link(2), and cpup at hinotify */ + AuLsc_DI_CHILD3, /* copyup dirs */ + AuLsc_DI_PARENT, + AuLsc_DI_PARENT2, + AuLsc_DI_PARENT3 +}; + +/* + * di_read_lock_child, di_write_lock_child, + * di_read_lock_child2, di_write_lock_child2, + * di_read_lock_child3, di_write_lock_child3, + * di_read_lock_parent, di_write_lock_parent, + * di_read_lock_parent2, di_write_lock_parent2, + * di_read_lock_parent3, di_write_lock_parent3, + */ +#define ReadLockFunc(name, lsc) \ +static inline void di_read_lock_##name(struct dentry *d, int flags) \ +{di_read_lock(d, flags, AuLsc_DI_##lsc);} + +#define WriteLockFunc(name, lsc) \ +static inline void di_write_lock_##name(struct dentry *d) \ +{di_write_lock(d, AuLsc_DI_##lsc);} + +#define RWLockFuncs(name, lsc) \ + ReadLockFunc(name, lsc); \ + WriteLockFunc(name, lsc) + +RWLockFuncs(child, CHILD); +RWLockFuncs(child2, CHILD2); +RWLockFuncs(child3, CHILD3); +RWLockFuncs(parent, PARENT); +RWLockFuncs(parent2, PARENT2); +RWLockFuncs(parent3, PARENT3); + +#undef ReadLockFunc +#undef WriteLockFunc +#undef RWLockFunc + +/* to debug easier, do not make them inlined functions */ +#define DiMustReadLock(d) do { \ + SiMustAnyLock((d)->d_sb); \ + RwMustReadLock(&dtodi(d)->di_rwsem); \ +} while (0) + +#define DiMustWriteLock(d) do { \ + SiMustAnyLock((d)->d_sb); \ + RwMustWriteLock(&dtodi(d)->di_rwsem); \ +} while (0) + +#define DiMustAnyLock(d) do { \ + SiMustAnyLock((d)->d_sb); \ + RwMustAnyLock(&dtodi(d)->di_rwsem); \ +} while (0) + +#define DiMustNoWaiters(d) RwMustNoWaiters(&dtodi(d)->di_rwsem) + +#endif /* __KERNEL__ */ +#endif /* __AUFS_DENTRY_H__ */ diff --git a/fs/aufs/dinfo.c b/fs/aufs/dinfo.c new file mode 100755 index 0000000..6082149 --- /dev/null +++ b/fs/aufs/dinfo.c @@ -0,0 +1,419 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: dinfo.c,v 1.23 2007/05/07 03:43:36 sfjro Exp $ */ + +#include "aufs.h" + +int au_alloc_dinfo(struct dentry *dentry) +{ + struct aufs_dinfo *dinfo; + struct super_block *sb; + int nbr; + + LKTRTrace("%.*s\n", DLNPair(dentry)); + DEBUG_ON(dentry->d_fsdata); + + dinfo = cache_alloc_dinfo(); + //if (LktrCond) {cache_free_dinfo(dinfo); dinfo = NULL;} + if (dinfo) { + sb = dentry->d_sb; + nbr = sbend(sb) + 1; + if (unlikely(!nbr)) + nbr++; + dinfo->di_hdentry = kcalloc(nbr, sizeof(*dinfo->di_hdentry), + GFP_KERNEL); + //if (LktrCond) + //{kfree(dinfo->di_hdentry); dinfo->di_hdentry = NULL;} + if (dinfo->di_hdentry) { + rw_init_wlock_nested(&dinfo->di_rwsem, AuLsc_DI_PARENT); + dinfo->di_bstart = dinfo->di_bend = -1; + dinfo->di_bwh = dinfo->di_bdiropq = -1; + atomic_set(&dinfo->di_generation, au_sigen(sb)); + + dentry->d_fsdata = dinfo; + dentry->d_op = &aufs_dop; + return 0; /* success */ + } + cache_free_dinfo(dinfo); + } + TraceErr(-ENOMEM); + return -ENOMEM; +} + +struct aufs_dinfo *dtodi(struct dentry *dentry) +{ + struct aufs_dinfo *dinfo = dentry->d_fsdata; + DEBUG_ON(!dinfo + || !dinfo->di_hdentry + /* || stosi(dentry->d_sb)->si_bend < dinfo->di_bend */ + || dinfo->di_bend < dinfo->di_bstart + /* dbwh can be outside of this range */ + || (0 <= dinfo->di_bdiropq + && (dinfo->di_bdiropq < dinfo->di_bstart + /* || dinfo->di_bend < dinfo->di_bdiropq */)) + ); + return dinfo; +} + +/* ---------------------------------------------------------------------- */ + +static void do_ii_write_lock(struct inode *inode, unsigned int lsc) +{ + switch (lsc) { + case AuLsc_DI_CHILD: + ii_write_lock_child(inode); + break; + case AuLsc_DI_CHILD2: + ii_write_lock_child2(inode); + break; + case AuLsc_DI_CHILD3: + ii_write_lock_child3(inode); + break; + case AuLsc_DI_PARENT: + ii_write_lock_parent(inode); + break; + case AuLsc_DI_PARENT2: + ii_write_lock_parent2(inode); + break; + case AuLsc_DI_PARENT3: + ii_write_lock_parent3(inode); + break; + default: + BUG(); + } +} + +static void do_ii_read_lock(struct inode *inode, unsigned int lsc) +{ + switch (lsc) { + case AuLsc_DI_CHILD: + ii_read_lock_child(inode); + break; + case AuLsc_DI_CHILD2: + ii_read_lock_child2(inode); + break; + case AuLsc_DI_CHILD3: + ii_read_lock_child3(inode); + break; + case AuLsc_DI_PARENT: + ii_read_lock_parent(inode); + break; + case AuLsc_DI_PARENT2: + ii_read_lock_parent2(inode); + break; + case AuLsc_DI_PARENT3: + ii_read_lock_parent3(inode); + break; + default: + BUG(); + } +} + +void di_read_lock(struct dentry *d, int flags, unsigned int lsc) +{ + SiMustAnyLock(d->d_sb); + // todo: always nested? + rw_read_lock_nested(&dtodi(d)->di_rwsem, lsc); + if (d->d_inode) { + if (flags & AUFS_I_WLOCK) + do_ii_write_lock(d->d_inode, lsc); + else if (flags & AUFS_I_RLOCK) + do_ii_read_lock(d->d_inode, lsc); + } +} + +void di_read_unlock(struct dentry *d, int flags) +{ + SiMustAnyLock(d->d_sb); + if (d->d_inode) { + if (flags & AUFS_I_WLOCK) + ii_write_unlock(d->d_inode); + else if (flags & AUFS_I_RLOCK) + ii_read_unlock(d->d_inode); + } + rw_read_unlock(&dtodi(d)->di_rwsem); +} + +void di_downgrade_lock(struct dentry *d, int flags) +{ + SiMustAnyLock(d->d_sb); + rw_dgrade_lock(&dtodi(d)->di_rwsem); + if (d->d_inode && (flags & AUFS_I_RLOCK)) + ii_downgrade_lock(d->d_inode); +} + +void di_write_lock(struct dentry *d, unsigned int lsc) +{ + SiMustAnyLock(d->d_sb); + // todo: always nested? + rw_write_lock_nested(&dtodi(d)->di_rwsem, lsc); + if (d->d_inode) + do_ii_write_lock(d->d_inode, lsc); +} + +void di_write_unlock(struct dentry *d) +{ + SiMustAnyLock(d->d_sb); + if (d->d_inode) + ii_write_unlock(d->d_inode); + rw_write_unlock(&dtodi(d)->di_rwsem); +} + +void di_write_lock2_child(struct dentry *d1, struct dentry *d2, int isdir) +{ + struct dentry *d; + + TraceEnter(); + DEBUG_ON(d1 == d2 + || d1->d_inode == d2->d_inode + || d1->d_sb != d2->d_sb); + + if (isdir) + for (d = d1; d->d_parent != d; d = d->d_parent) // dget_parent() + if (d->d_parent == d2) { + di_write_lock_child(d1); + di_write_lock_child2(d2); + return; + } + + di_write_lock_child(d2); + di_write_lock_child2(d1); +} + +void di_write_lock2_parent(struct dentry *d1, struct dentry *d2, int isdir) +{ + struct dentry *d; + + TraceEnter(); + DEBUG_ON(d1 == d2 + || d1->d_inode == d2->d_inode + || d1->d_sb != d2->d_sb); + + if (isdir) + for (d = d1; d->d_parent != d; d = d->d_parent) // dget_parent() + if (d->d_parent == d2) { + di_write_lock_parent(d1); + di_write_lock_parent2(d2); + return; + } + + di_write_lock_parent(d2); + di_write_lock_parent2(d1); +} + +void di_write_unlock2(struct dentry *d1, struct dentry *d2) +{ + di_write_unlock(d1); + if (d1->d_inode == d2->d_inode) + rw_write_unlock(&dtodi(d2)->di_rwsem); + else + di_write_unlock(d2); +} + +/* ---------------------------------------------------------------------- */ + +aufs_bindex_t dbstart(struct dentry *dentry) +{ + DiMustAnyLock(dentry); + return dtodi(dentry)->di_bstart; +} + +aufs_bindex_t dbend(struct dentry *dentry) +{ + DiMustAnyLock(dentry); + return dtodi(dentry)->di_bend; +} + +aufs_bindex_t dbwh(struct dentry *dentry) +{ + DiMustAnyLock(dentry); + return dtodi(dentry)->di_bwh; +} + +aufs_bindex_t dbdiropq(struct dentry *dentry) +{ + DiMustAnyLock(dentry); + DEBUG_ON(dentry->d_inode + && dentry->d_inode->i_mode + && !S_ISDIR(dentry->d_inode->i_mode)); + return dtodi(dentry)->di_bdiropq; +} + +struct dentry *au_h_dptr_i(struct dentry *dentry, aufs_bindex_t bindex) +{ + struct dentry *d; + + DiMustAnyLock(dentry); + if (dbstart(dentry) < 0 || bindex < dbstart(dentry)) + return NULL; + DEBUG_ON(bindex < 0 + /* || bindex > sbend(dentry->d_sb) */); + d = dtodi(dentry)->di_hdentry[0 + bindex].hd_dentry; + DEBUG_ON(d && (atomic_read(&d->d_count) <= 0)); + return d; +} + +struct dentry *au_h_dptr(struct dentry *dentry) +{ + return au_h_dptr_i(dentry, dbstart(dentry)); +} + +aufs_bindex_t dbtail(struct dentry *dentry) +{ + aufs_bindex_t bend, bwh; + + bend = dbend(dentry); + if (0 <= bend) { + bwh = dbwh(dentry); + //DEBUG_ON(bend < bwh); + if (!bwh) + return bwh; + if (0 < bwh && bwh < bend) + return bwh - 1; + } + return bend; +} + +aufs_bindex_t dbtaildir(struct dentry *dentry) +{ + aufs_bindex_t bend, bopq; + + DEBUG_ON(dentry->d_inode + && dentry->d_inode->i_mode + && !S_ISDIR(dentry->d_inode->i_mode)); + + bend = dbtail(dentry); + if (0 <= bend) { + bopq = dbdiropq(dentry); + DEBUG_ON(bend < bopq); + if (0 <= bopq && bopq < bend) + bend = bopq; + } + return bend; +} + +aufs_bindex_t dbtail_generic(struct dentry *dentry) +{ + struct inode *inode; + + inode = dentry->d_inode; + if (inode && S_ISDIR(inode->i_mode)) + return dbtaildir(dentry); + else + return dbtail(dentry); +} + +/* ---------------------------------------------------------------------- */ + +// hard/soft set +void set_dbstart(struct dentry *dentry, aufs_bindex_t bindex) +{ + DiMustWriteLock(dentry); + DEBUG_ON(sbend(dentry->d_sb) < bindex); + /* */ + dtodi(dentry)->di_bstart = bindex; +} + +void set_dbend(struct dentry *dentry, aufs_bindex_t bindex) +{ + DiMustWriteLock(dentry); + DEBUG_ON(sbend(dentry->d_sb) < bindex + || bindex < dbstart(dentry)); + dtodi(dentry)->di_bend = bindex; +} + +void set_dbwh(struct dentry *dentry, aufs_bindex_t bindex) +{ + DiMustWriteLock(dentry); + DEBUG_ON(sbend(dentry->d_sb) < bindex); + /* dbwh can be outside of bstart - bend range */ + dtodi(dentry)->di_bwh = bindex; +} + +void set_dbdiropq(struct dentry *dentry, aufs_bindex_t bindex) +{ + DiMustWriteLock(dentry); + DEBUG_ON(sbend(dentry->d_sb) < bindex); + DEBUG_ON((bindex != -1 + && (bindex < dbstart(dentry) || dbend(dentry) < bindex)) + || (dentry->d_inode + && dentry->d_inode->i_mode + && !S_ISDIR(dentry->d_inode->i_mode))); + dtodi(dentry)->di_bdiropq = bindex; +} + +void hdput(struct aufs_hdentry *hd) +{ + dput(hd->hd_dentry); +} + +void set_h_dptr(struct dentry *dentry, aufs_bindex_t bindex, + struct dentry *h_dentry) +{ + struct aufs_hdentry *hd = dtodi(dentry)->di_hdentry + bindex; + DiMustWriteLock(dentry); + DEBUG_ON(bindex < dtodi(dentry)->di_bstart + || bindex > dtodi(dentry)->di_bend + || (h_dentry && atomic_read(&h_dentry->d_count) <= 0) + || (h_dentry && hd->hd_dentry) + ); + if (hd->hd_dentry) + hdput(hd); + hd->hd_dentry = h_dentry; +} + +/* ---------------------------------------------------------------------- */ + +void au_update_digen(struct dentry *dentry) +{ + //DiMustWriteLock(dentry); + DEBUG_ON(!dentry->d_sb); + atomic_set(&dtodi(dentry)->di_generation, au_sigen(dentry->d_sb)); +} + +void au_update_dbstart(struct dentry *dentry) +{ + aufs_bindex_t bindex, bstart = dbstart(dentry), bend = dbend(dentry); + struct dentry *hidden_dentry; + + DiMustWriteLock(dentry); + for (bindex = bstart; bindex <= bend; bindex++) { + hidden_dentry = au_h_dptr_i(dentry, bindex); + if (!hidden_dentry) + continue; + if (hidden_dentry->d_inode) { + set_dbstart(dentry, bindex); + return; + } + set_h_dptr(dentry, bindex, NULL); + } + //set_dbstart(dentry, -1); + //set_dbend(dentry, -1); +} + +int au_find_dbindex(struct dentry *dentry, struct dentry *hidden_dentry) +{ + aufs_bindex_t bindex, bend; + + bend = dbend(dentry); + for (bindex = dbstart(dentry); bindex <= bend; bindex++) + if (au_h_dptr_i(dentry, bindex) == hidden_dentry) + return bindex; + return -1; +} diff --git a/fs/aufs/dir.c b/fs/aufs/dir.c new file mode 100755 index 0000000..9afb1a9 --- /dev/null +++ b/fs/aufs/dir.c @@ -0,0 +1,564 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: dir.c,v 1.36 2007/05/14 03:38:52 sfjro Exp $ */ + +#include "aufs.h" + +static int reopen_dir(struct file *file) +{ + int err; + struct dentry *dentry, *hidden_dentry; + aufs_bindex_t bindex, btail, bstart; + struct file *hidden_file; + + dentry = file->f_dentry; + LKTRTrace("%.*s\n", DLNPair(dentry)); + DEBUG_ON(!S_ISDIR(dentry->d_inode->i_mode)); + + /* open all hidden dirs */ + bstart = dbstart(dentry); +#if 1 + for (bindex = fbstart(file); bindex < bstart; bindex++) + set_h_fptr(file, bindex, NULL); +#endif + set_fbstart(file, bstart); + btail = dbtaildir(dentry); +#if 1 + for (bindex = fbend(file); btail < bindex; bindex--) + set_h_fptr(file, bindex, NULL); +#endif + set_fbend(file, btail); + for (bindex = bstart; bindex <= btail; bindex++) { + hidden_dentry = au_h_dptr_i(dentry, bindex); + if (!hidden_dentry) + continue; + hidden_file = au_h_fptr_i(file, bindex); + if (hidden_file) { + DEBUG_ON(hidden_file->f_dentry != hidden_dentry); + continue; + } + + hidden_file = hidden_open(dentry, bindex, file->f_flags); + // unavailable + //if (LktrCond) {fput(hidden_file); + //br_put(stobr(dentry->d_sb, bindex));hidden_file=ERR_PTR(-1);} + err = PTR_ERR(hidden_file); + if (IS_ERR(hidden_file)) + goto out; // close all? + //cpup_file_flags(hidden_file, file); + set_h_fptr(file, bindex, hidden_file); + } + err = 0; + + out: + TraceErr(err); + return err; +} + +static int do_open_dir(struct file *file, int flags) +{ + int err; + aufs_bindex_t bindex, btail; + struct dentry *dentry, *hidden_dentry; + struct file *hidden_file; + + dentry = file->f_dentry; + LKTRTrace("%.*s, 0x%x\n", DLNPair(dentry), flags); + DEBUG_ON(!dentry->d_inode || !S_ISDIR(dentry->d_inode->i_mode)); + + err = 0; + set_fvdir_cache(file, NULL); + file->f_version = dentry->d_inode->i_version; + bindex = dbstart(dentry); + set_fbstart(file, bindex); + btail = dbtaildir(dentry); + set_fbend(file, btail); + for (; !err && bindex <= btail; bindex++) { + hidden_dentry = au_h_dptr_i(dentry, bindex); + if (!hidden_dentry) + continue; + + hidden_file = hidden_open(dentry, bindex, flags); + //if (LktrCond) {fput(hidden_file); + //br_put(stobr(dentry->d_sb, bindex));hidden_file=ERR_PTR(-1);} + if (!IS_ERR(hidden_file)) { + set_h_fptr(file, bindex, hidden_file); + continue; + } + err = PTR_ERR(hidden_file); + } + if (!err) + return 0; /* success */ + + /* close all */ + for (bindex = fbstart(file); !err && bindex <= btail; bindex++) + set_h_fptr(file, bindex, NULL); + set_fbstart(file, -1); + set_fbend(file, -1); + return err; +} + +static int aufs_open_dir(struct inode *inode, struct file *file) +{ + return au_do_open(inode, file, do_open_dir); +} + +static int aufs_release_dir(struct inode *inode, struct file *file) +{ + struct aufs_vdir *vdir_cache; + struct super_block *sb; + + LKTRTrace("i%lu, %.*s\n", inode->i_ino, DLNPair(file->f_dentry)); + + sb = file->f_dentry->d_sb; + si_read_lock(sb); + fi_write_lock(file); + vdir_cache = fvdir_cache(file); + if (vdir_cache) + free_vdir(vdir_cache); + fi_write_unlock(file); + au_fin_finfo(file); + si_read_unlock(sb); + return 0; +} + +static int fsync_dir(struct dentry *dentry, int datasync) +{ + int err; + struct inode *inode; + struct super_block *sb; + aufs_bindex_t bend, bindex; + + LKTRTrace("%.*s, %d\n", DLNPair(dentry), datasync); + DiMustAnyLock(dentry); + sb = dentry->d_sb; + SiMustAnyLock(sb); + inode = dentry->d_inode; + IMustLock(inode); + IiMustAnyLock(inode); + + err = 0; + bend = dbend(dentry); + for (bindex = dbstart(dentry); !err && bindex <= bend; bindex++) { + struct dentry *h_dentry; + struct inode *h_inode; + struct file_operations *fop; + + if (test_ro(sb, bindex, inode)) + continue; + h_dentry = au_h_dptr_i(dentry, bindex); + if (!h_dentry) + continue; + h_inode = h_dentry->d_inode; + if (!h_inode) + continue; + + /* cf. fs/nsfd/vfs.c and fs/nfsd/nfs4recover.c */ + //hdir_lock(h_inode, inode, bindex); + i_lock(h_inode); + fop = (void*)h_inode->i_fop; + err = filemap_fdatawrite(h_inode->i_mapping); + if (!err && fop && fop->fsync) + err = fop->fsync(NULL, h_dentry, datasync); + if (!err) + err = filemap_fdatawrite(h_inode->i_mapping); + //hdir_unlock(h_inode, inode, bindex); + i_unlock(h_inode); + } + + TraceErr(err); + return err; +} + +/* + * @file may be NULL + */ +static int aufs_fsync_dir(struct file *file, struct dentry *dentry, + int datasync) +{ + int err; + struct inode *inode; + struct file *hidden_file; + struct super_block *sb; + aufs_bindex_t bend, bindex; + + LKTRTrace("%.*s, %d\n", DLNPair(dentry), datasync); + inode = dentry->d_inode; + IMustLock(inode); + + err = 0; + sb = dentry->d_sb; + si_read_lock(sb); + if (file) { + err = au_reval_and_lock_finfo(file, reopen_dir, /*wlock*/1, + /*locked*/1); + //err = -1; + if (unlikely(err)) + goto out; + } else + di_read_lock_child(dentry, !AUFS_I_WLOCK); + + ii_write_lock_child(inode); + if (file) { + bend = fbend(file); + for (bindex = fbstart(file); !err && bindex <= bend; bindex++) { + hidden_file = au_h_fptr_i(file, bindex); + if (!hidden_file || test_ro(sb, bindex, inode)) + continue; + + err = -EINVAL; + if (hidden_file->f_op && hidden_file->f_op->fsync) { + // todo: try do_fsync() in fs/sync.c +#if 0 + DEBUG_ON(hidden_file->f_dentry->d_inode + != au_h_iptr_i(inode, bindex)); + hdir_lock(hidden_file->f_dentry->d_inode, inode, + bindex); +#else + i_lock(hidden_file->f_dentry->d_inode); +#endif + err = hidden_file->f_op->fsync + (hidden_file, hidden_file->f_dentry, + datasync); + //err = -1; +#if 0 + hdir_unlock(hidden_file->f_dentry->d_inode, + inode, bindex); +#else + i_unlock(hidden_file->f_dentry->d_inode); +#endif + } + } + } else + err = fsync_dir(dentry, datasync); + au_cpup_attr_timesizes(inode); + ii_write_unlock(inode); + if (file) + fi_write_unlock(file); + else + di_read_unlock(dentry, !AUFS_I_WLOCK); + + out: + si_read_unlock(sb); + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +static int aufs_readdir(struct file *file, void *dirent, filldir_t filldir) +{ + int err; + struct dentry *dentry; + struct inode *inode; + struct super_block *sb; + + dentry = file->f_dentry; + LKTRTrace("%.*s, pos %Ld\n", DLNPair(dentry), file->f_pos); + inode = dentry->d_inode; + IMustLock(inode); + + au_nfsd_lockdep_off(); + sb = dentry->d_sb; + si_read_lock(sb); + err = au_reval_and_lock_finfo(file, reopen_dir, /*wlock*/1, + /*locked*/1); + if (unlikely(err)) + goto out; + + ii_write_lock_child(inode); + err = au_init_vdir(file); + if (unlikely(err)) { + ii_write_unlock(inode); + goto out_unlock; + } + //DbgVdir(fvdir_cache(file));// goto out_unlock; + + /* nfsd filldir calls lookup_one_len(). */ + ii_downgrade_lock(inode); + err = au_fill_de(file, dirent, filldir); + //DbgVdir(fvdir_cache(file));// goto out_unlock; + + inode->i_atime = au_h_iptr(inode)->i_atime; + ii_read_unlock(inode); + + out_unlock: + fi_write_unlock(file); + out: + si_read_unlock(sb); + au_nfsd_lockdep_on(); +#if 0 // debug + if (LktrCond) + igrab(inode); +#endif + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +struct test_empty_arg { + struct aufs_nhash *whlist; + int whonly; + aufs_bindex_t bindex; + int err, called; +}; + +static int test_empty_cb(void *__arg, const char *__name, int namelen, + loff_t offset, filldir_ino_t ino, unsigned int d_type) +{ + struct test_empty_arg *arg = __arg; + char *name = (void*)__name; + + LKTRTrace("%.*s\n", namelen, name); + + arg->err = 0; + arg->called++; + //smp_mb(); + if (name[0] == '.' + && (namelen == 1 || (name[1] == '.' && namelen == 2))) + return 0; /* success */ + + if (namelen <= AUFS_WH_PFX_LEN + || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) { + if (arg->whonly && !test_known_wh(arg->whlist, name, namelen)) + arg->err = -ENOTEMPTY; + goto out; + } + + name += AUFS_WH_PFX_LEN; + namelen -= AUFS_WH_PFX_LEN; + if (!test_known_wh(arg->whlist, name, namelen)) + arg->err = append_wh(arg->whlist, name, namelen, arg->bindex); + + out: + //smp_mb(); + TraceErr(arg->err); + return arg->err; +} + +static int do_test_empty(struct dentry *dentry, struct test_empty_arg *arg) +{ + int err, dlgt; + struct file *hidden_file; + + LKTRTrace("%.*s, {%p, %d, %d}\n", + DLNPair(dentry), arg->whlist, arg->whonly, arg->bindex); + + hidden_file = hidden_open(dentry, arg->bindex, + O_RDONLY | O_NONBLOCK | O_DIRECTORY + | O_LARGEFILE); + err = PTR_ERR(hidden_file); + if (IS_ERR(hidden_file)) + goto out; + + dlgt = need_dlgt(dentry->d_sb); + //hidden_file->f_pos = 0; + do { + arg->err = 0; + arg->called = 0; + //smp_mb(); + err = vfsub_readdir(hidden_file, test_empty_cb, arg, dlgt); + if (err >= 0) + err = arg->err; + } while (!err && arg->called); + fput(hidden_file); + sbr_put(dentry->d_sb, arg->bindex); + + out: + TraceErr(err); + return err; +} + +struct do_test_empty_args { + int *errp; + struct dentry *dentry; + struct test_empty_arg *arg; +}; + +static void call_do_test_empty(void *args) +{ + struct do_test_empty_args *a = args; + *a->errp = do_test_empty(a->dentry, a->arg); +} + +static int sio_test_empty(struct dentry *dentry, struct test_empty_arg *arg) +{ + int err; + struct dentry *hidden_dentry; + struct inode *hidden_inode; + + LKTRTrace("%.*s\n", DLNPair(dentry)); + hidden_dentry = au_h_dptr_i(dentry, arg->bindex); + DEBUG_ON(!hidden_dentry); + hidden_inode = hidden_dentry->d_inode; + DEBUG_ON(!hidden_inode || !S_ISDIR(hidden_inode->i_mode)); + + hi_lock_child(hidden_inode); + err = au_test_perm(hidden_inode, MAY_EXEC | MAY_READ, + need_dlgt(dentry->d_sb)); + i_unlock(hidden_inode); + if (!err) + err = do_test_empty(dentry, arg); + else { + struct do_test_empty_args args = { + .errp = &err, + .dentry = dentry, + .arg = arg + }; + au_wkq_wait(call_do_test_empty, &args, /*dlgt*/0); + } + + TraceErr(err); + return err; +} + +int au_test_empty_lower(struct dentry *dentry) +{ + int err; + struct inode *inode; + struct test_empty_arg arg; + struct aufs_nhash *whlist; + aufs_bindex_t bindex, bstart, btail; + + LKTRTrace("%.*s\n", DLNPair(dentry)); + inode = dentry->d_inode; + DEBUG_ON(!inode || !S_ISDIR(inode->i_mode)); + + whlist = nhash_new(GFP_KERNEL); + err = PTR_ERR(whlist); + if (IS_ERR(whlist)) + goto out; + + bstart = dbstart(dentry); + arg.whlist = whlist; + arg.whonly = 0; + arg.bindex = bstart; + err = do_test_empty(dentry, &arg); + if (unlikely(err)) + goto out_whlist; + + arg.whonly = 1; + btail = dbtaildir(dentry); + for (bindex = bstart + 1; !err && bindex <= btail; bindex++) { + struct dentry *hidden_dentry; + hidden_dentry = au_h_dptr_i(dentry, bindex); + if (hidden_dentry && hidden_dentry->d_inode) { + DEBUG_ON(!S_ISDIR(hidden_dentry->d_inode->i_mode)); + arg.bindex = bindex; + err = do_test_empty(dentry, &arg); + } + } + + out_whlist: + nhash_del(whlist); + out: + TraceErr(err); + return err; +} + +int test_empty(struct dentry *dentry, struct aufs_nhash *whlist) +{ + int err; + struct inode *inode; + struct test_empty_arg arg; + aufs_bindex_t bindex, btail; + + LKTRTrace("%.*s\n", DLNPair(dentry)); + inode = dentry->d_inode; + DEBUG_ON(!inode || !S_ISDIR(inode->i_mode)); + + err = 0; + arg.whlist = whlist; + arg.whonly = 1; + btail = dbtaildir(dentry); + for (bindex = dbstart(dentry); !err && bindex <= btail; bindex++) { + struct dentry *hidden_dentry; + hidden_dentry = au_h_dptr_i(dentry, bindex); + if (hidden_dentry && hidden_dentry->d_inode) { + DEBUG_ON(!S_ISDIR(hidden_dentry->d_inode->i_mode)); + arg.bindex = bindex; + err = sio_test_empty(dentry, &arg); + } + } + + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +void au_add_nlink(struct inode *dir, struct inode *h_dir) +{ + DEBUG_ON(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode)); + dir->i_nlink += h_dir->i_nlink - 2; + if (unlikely(h_dir->i_nlink < 2)) + dir->i_nlink += 2; +} + +void au_sub_nlink(struct inode *dir, struct inode *h_dir) +{ + DEBUG_ON(!S_ISDIR(dir->i_mode) || !S_ISDIR(h_dir->i_mode)); + dir->i_nlink -= h_dir->i_nlink - 2; + if (unlikely(h_dir->i_nlink < 2)) + dir->i_nlink -= 2; +} + +/* ---------------------------------------------------------------------- */ + +#if 0 // comment +struct file_operations { + struct module *owner; + loff_t (*llseek) (struct file *, loff_t, int); + ssize_t (*read) (struct file *, char __user *, size_t, loff_t *); + ssize_t (*aio_read) (struct kiocb *, char __user *, size_t, loff_t); + ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *); + ssize_t (*aio_write) (struct kiocb *, const char __user *, size_t, loff_t); + int (*readdir) (struct file *, void *, filldir_t); + unsigned int (*poll) (struct file *, struct poll_table_struct *); + int (*ioctl) (struct inode *, struct file *, unsigned int, unsigned long); + long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long); + long (*compat_ioctl) (struct file *, unsigned int, unsigned long); + int (*mmap) (struct file *, struct vm_area_struct *); + int (*open) (struct inode *, struct file *); + int (*flush) (struct file *); + int (*release) (struct inode *, struct file *); + int (*fsync) (struct file *, struct dentry *, int datasync); + int (*aio_fsync) (struct kiocb *, int datasync); + int (*fasync) (int, struct file *, int); + int (*lock) (struct file *, int, struct file_lock *); + ssize_t (*readv) (struct file *, const struct iovec *, unsigned long, loff_t *); + ssize_t (*writev) (struct file *, const struct iovec *, unsigned long, loff_t *); + ssize_t (*sendfile) (struct file *, loff_t *, size_t, read_actor_t, void *); + ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int); + unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); + int (*check_flags)(int); + int (*dir_notify)(struct file *file, unsigned long arg); + int (*flock) (struct file *, int, struct file_lock *); +}; +#endif + +struct file_operations aufs_dir_fop = { + .read = generic_read_dir, + .readdir = aufs_readdir, + .open = aufs_open_dir, + .release = aufs_release_dir, + .flush = aufs_flush, + .fsync = aufs_fsync_dir, +}; diff --git a/fs/aufs/dir.h b/fs/aufs/dir.h new file mode 100755 index 0000000..3ddf309 --- /dev/null +++ b/fs/aufs/dir.h @@ -0,0 +1,125 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: dir.h,v 1.18 2007/05/14 03:41:52 sfjro Exp $ */ + +#ifndef __AUFS_DIR_H__ +#define __AUFS_DIR_H__ + +#ifdef __KERNEL__ + +#include +#include +#include + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,19) +#define filldir_ino_t u64 +#else +#define filldir_ino_t ino_t +#endif + +/* ---------------------------------------------------------------------- */ + +/* need to be faster and smaller */ + +#define AUFS_DEBLK_SIZE 512 // todo: changable +#define AUFS_NHASH_SIZE 32 // todo: changable +#if AUFS_DEBLK_SIZE < NAME_MAX || PAGE_SIZE < AUFS_DEBLK_SIZE +#error invalid size AUFS_DEBLK_SIZE +#endif + +typedef char aufs_deblk_t[AUFS_DEBLK_SIZE]; + +struct aufs_nhash { + struct hlist_head heads[AUFS_NHASH_SIZE]; +}; + +struct aufs_destr { + unsigned char len; + char name[0]; +} __attribute__ ((packed)); + +struct aufs_dehstr { + struct hlist_node hash; + struct aufs_destr *str; +}; + +struct aufs_de { + ino_t de_ino; + unsigned char de_type; + //caution: packed + struct aufs_destr de_str; +} __attribute__ ((packed)); + +struct aufs_wh { + struct hlist_node wh_hash; + aufs_bindex_t wh_bindex; + struct aufs_destr wh_str; +} __attribute__ ((packed)); + +union aufs_deblk_p { + unsigned char *p; + aufs_deblk_t *deblk; + struct aufs_de *de; +}; + +struct aufs_vdir { + aufs_deblk_t **vd_deblk; + int vd_nblk; + struct { + int i; + union aufs_deblk_p p; + } vd_last; + + unsigned long vd_version; + unsigned long vd_jiffy; +}; + +/* ---------------------------------------------------------------------- */ + +/* dir.c */ +extern struct file_operations aufs_dir_fop; +int au_test_empty_lower(struct dentry *dentry); +int test_empty(struct dentry *dentry, struct aufs_nhash *whlist); +void au_add_nlink(struct inode *dir, struct inode *h_dir); +void au_sub_nlink(struct inode *dir, struct inode *h_dir); + +/* vdir.c */ +struct aufs_nhash *nhash_new(gfp_t gfp); +void nhash_del(struct aufs_nhash *nhash); +void nhash_init(struct aufs_nhash *nhash); +void nhash_move(struct aufs_nhash *dst, struct aufs_nhash *src); +void nhash_fin(struct aufs_nhash *nhash); +int is_longer_wh(struct aufs_nhash *whlist, aufs_bindex_t btgt, int limit); +int test_known_wh(struct aufs_nhash *whlist, char *name, int namelen); +int append_wh(struct aufs_nhash *whlist, char *name, int namelen, + aufs_bindex_t bindex); +void free_vdir(struct aufs_vdir *vdir); +int au_init_vdir(struct file *file); +int au_fill_de(struct file *file, void *dirent, filldir_t filldir); + +/* ---------------------------------------------------------------------- */ + +static inline +unsigned int au_name_hash(const unsigned char *name, unsigned int len) +{ + return (full_name_hash(name, len) % AUFS_NHASH_SIZE); +} + +#endif /* __KERNEL__ */ +#endif /* __AUFS_DIR_H__ */ diff --git a/fs/aufs/export.c b/fs/aufs/export.c new file mode 100755 index 0000000..7b1c6ac --- /dev/null +++ b/fs/aufs/export.c @@ -0,0 +1,585 @@ +/* + * Copyright (C) 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: export.c,v 1.7 2007/05/14 03:38:24 sfjro Exp $ */ + +#include "aufs.h" + +extern struct export_operations export_op_default; +#define CALL(ops, func) (((ops)->func) ? ((ops)->func) : export_op_default.func) +#define is_anon(d) ((d)->d_flags & DCACHE_DISCONNECTED) + +union conv { +#if BITS_PER_LONG == 32 + __u32 a[1]; +#else + __u32 a[2]; +#endif + ino_t ino; +}; + +static ino_t decode_ino(__u32 *a) +{ + union conv u; + u.a[0] = a[0]; +#if BITS_PER_LONG == 64 + u.a[1] = a[1]; +#endif + return u.ino; +} + +static void encode_ino(__u32 *a, ino_t ino) +{ + union conv u; + u.ino = ino; + a[0] = u.a[0]; +#if BITS_PER_LONG == 64 + a[1] = u.a[1]; +#endif +} + +static void decode_br_id_sigen(__u32 a, aufs_bindex_t *br_id, + aufs_bindex_t *sigen) +{ + BUILD_BUG_ON((sizeof(*br_id) + sizeof(*sigen)) > sizeof(a)); + *br_id = a >> 16; + DEBUG_ON(*br_id < 0); + *sigen = a; + DEBUG_ON(*sigen < 0); +} + +static __u32 encode_br_id_sigen(aufs_bindex_t br_id, aufs_bindex_t sigen) +{ + DEBUG_ON(br_id < 0 || sigen < 0); + return (br_id << 16) | sigen; +} + +/* NFS file handle */ +enum { + /* support 64bit inode number */ + /* but untested */ + Fh_br_id_sigen, + Fh_ino1, +#if BITS_PER_LONG == 64 + Fh_ino2, +#endif + Fh_dir_ino1, +#if BITS_PER_LONG == 64 + Fh_dir_ino2, +#endif + Fh_h_ino1, +#if BITS_PER_LONG == 64 + Fh_h_ino2, +#endif + Fh_h_igen, + Fh_h_type, + Fh_tail, + + Fh_ino = Fh_ino1, + Fh_dir_ino = Fh_dir_ino1, + Fh_h_ino = Fh_h_ino1, +}; + +/* ---------------------------------------------------------------------- */ + +static struct dentry *decode_by_ino(struct super_block *sb, ino_t ino, + ino_t dir_ino) +{ + struct dentry *dentry; + struct inode *inode; + + LKTRTrace("i%lu, diri%lu\n", ino, dir_ino); + + dentry = NULL; + inode = ilookup(sb, ino); + if (unlikely(!inode)) + goto out; + + dentry = ERR_PTR(-ESTALE); + if (unlikely(is_bad_inode(inode))) + goto out_iput; + + dentry = NULL; + if (!S_ISDIR(inode->i_mode)) { + struct dentry *d; + spin_lock(&dcache_lock); + list_for_each_entry(d, &inode->i_dentry, d_alias) + if (!is_anon(d) + && d->d_parent->d_inode->i_ino == dir_ino) { + dentry = dget_locked(d); + break; + } + spin_unlock(&dcache_lock); + } else { + dentry = d_find_alias(inode); + if (dentry + && !is_anon(dentry) + && dentry->d_parent->d_inode->i_ino == dir_ino) + goto out_iput; /* success */ + + dput(dentry); + dentry = NULL; + } + + out_iput: + iput(inode); + out: + TraceErrPtr(dentry); + return dentry; +} + +/* ---------------------------------------------------------------------- */ + +struct find_name_by_ino { + int called, found; + ino_t ino; + char *name; + int namelen; +}; + +static int +find_name_by_ino(void *arg, const char *name, int namelen, loff_t offset, + filldir_ino_t ino, unsigned int d_type) +{ + struct find_name_by_ino *a = arg; + + a->called++; + if (a->ino != ino) + return 0; + + memcpy(a->name, name, namelen); + a->namelen = namelen; + a->found = 1; + return 1; +} + +static struct dentry *decode_by_dir_ino(struct super_block *sb, ino_t ino, + ino_t dir_ino) +{ + struct dentry *dentry, *parent; + struct inode *dir; + struct find_name_by_ino arg; + struct file *file; + int err; + + LKTRTrace("i%lu, diri%lu\n", ino, dir_ino); + + dentry = NULL; + dir = ilookup(sb, dir_ino); + if (unlikely(!dir)) + goto out; + + dentry = ERR_PTR(-ESTALE); + if (unlikely(is_bad_inode(dir))) + goto out_iput; + + dentry = NULL; + parent = d_find_alias(dir); + if (parent) { + if (unlikely(is_anon(parent))) { + dput(parent); + goto out_iput; + } + } else + goto out_iput; + + file = dentry_open(parent, NULL, au_dir_roflags); + dentry = (void*)file; + if (IS_ERR(file)) + goto out_iput; + + dentry = ERR_PTR(-ENOMEM); + arg.name = __getname(); + if (unlikely(!arg.name)) + goto out_fput; + arg.ino = ino; + arg.found = 0; + + do { + arg.called = 0; + //smp_mb(); + err = vfsub_readdir(file, find_name_by_ino, &arg, /*dlgt*/0); + } while (!err && !arg.found && arg.called); + dentry = ERR_PTR(err); + if (arg.found) { + /* do not call lkup_one(), nor dlgt */ + i_lock(dir); + dentry = lookup_one_len(arg.name, parent, arg.namelen); + i_unlock(dir); + TraceErrPtr(dentry); + } + + //out_putname: + __putname(arg.name); + out_fput: + fput(file); + out_iput: + iput(dir); + out: + TraceErrPtr(dentry); + return dentry; +} + +/* ---------------------------------------------------------------------- */ + +struct append_name { + int found, called, len; + char *h_path; + ino_t h_ino; +}; + +static int append_name(void *arg, const char *name, int len, loff_t pos, + filldir_ino_t ino, unsigned int d_type) +{ + struct append_name *a = arg; + char *p; + + a->called++; + if (ino != a->h_ino) + return 0; + + DEBUG_ON(len == 1 && *name == '.'); + DEBUG_ON(len == 2 && name[0] == '.' && name[1] == '.'); + a->len = strlen(a->h_path); + memmove(a->h_path - a->len - 1, a->h_path, a->len); + a->h_path -= a->len + 1; + p = a->h_path + a->len; + *p++ = '/'; + memcpy(p, name, a->len); + a->len += 1 + len; + a->found++; + return 1; +} + +static int h_acceptable(void *expv, struct dentry *dentry) +{ + return 1; +} + +static struct dentry* +decode_by_path(struct super_block *sb, aufs_bindex_t bindex, __u32 *fh, + int fh_len, void *context) +{ + struct dentry *dentry, *h_parent, *root, *h_root; + struct super_block *h_sb; + char *path, *p; + struct vfsmount *h_mnt; + struct append_name arg; + int len, err; + struct file *h_file; + struct nameidata nd; + struct aufs_branch *br; + + LKTRTrace("b%d\n", bindex); + SiMustAnyLock(sb); + + br = stobr(sb, bindex); + //br_get(br); + h_mnt = br->br_mnt; + h_sb = h_mnt->mnt_sb; + LKTRTrace("%s, h_decode_fh\n", au_sbtype(h_sb)); + h_parent = CALL(h_sb->s_export_op, decode_fh) + (h_sb, fh + Fh_tail, fh_len - Fh_tail, fh[Fh_h_type], + h_acceptable, /*context*/NULL); + dentry = h_parent; + if (unlikely(!h_parent || IS_ERR(h_parent))) { + Warn1("%s decode_fh failed\n", au_sbtype(h_sb)); + goto out; + } + dentry = NULL; + if (unlikely(is_anon(h_parent))) { + Warn1("%s decode_fh returned a disconnected dentry\n", + au_sbtype(h_sb)); + dput(h_parent); + goto out; + } + + dentry = ERR_PTR(-ENOMEM); + path = __getname(); + if (unlikely(!path)) { + dput(h_parent); + goto out; + } + + root = sb->s_root; + di_read_lock_parent(root, !AUFS_I_RLOCK); + h_root = au_h_dptr_i(root, bindex); + di_read_unlock(root, !AUFS_I_RLOCK); + arg.h_path = d_path(h_root, h_mnt, path, PATH_MAX); + dentry = (void*)arg.h_path; + if (unlikely(!arg.h_path || IS_ERR(arg.h_path))) + goto out_putname; + len = strlen(arg.h_path); + arg.h_path = d_path(h_parent, h_mnt, path, PATH_MAX); + dentry = (void*)arg.h_path; + if (unlikely(!arg.h_path || IS_ERR(arg.h_path))) + goto out_putname; + LKTRTrace("%s\n", arg.h_path); + if (len != 1) + arg.h_path += len; + LKTRTrace("%s\n", arg.h_path); + + /* cf. fs/exportfs/expfs.c */ + h_file = dentry_open(h_parent, NULL, au_dir_roflags); + dentry = (void*)h_file; + if (IS_ERR(h_file)) + goto out_putname; + + arg.found = 0; + arg.h_ino = decode_ino(fh + Fh_h_ino); + do { + arg.called = 0; + err = vfsub_readdir(h_file, append_name, &arg, /*dlgt*/0); + } while (!err && !arg.found && arg.called); + LKTRTrace("%s, %d\n", arg.h_path, arg.len); + + p = d_path(root, stosi(sb)->si_mnt, path, PATH_MAX - arg.len - 2); + dentry = (void*)p; + if (unlikely(!p || IS_ERR(p))) + goto out_fput; + p[strlen(p)] = '/'; + LKTRTrace("%s\n", p); + + err = path_lookup(p, LOOKUP_FOLLOW, &nd); + dentry = ERR_PTR(err); + if (!err) { + dentry = dget(nd.dentry); + if (unlikely(is_anon(dentry))) { + dput(dentry); + dentry = ERR_PTR(-ESTALE); + } + path_release(&nd); + } + + out_fput: + fput(h_file); + out_putname: + __putname(path); + out: + //br_put(br); + TraceErrPtr(dentry); + return dentry; +} + +/* ---------------------------------------------------------------------- */ + +static struct dentry* +aufs_decode_fh(struct super_block *sb, __u32 *fh, int fh_len, int fh_type, + int (*acceptable)(void *context, struct dentry *de), + void *context) +{ + struct dentry *dentry; + ino_t ino, dir_ino; + aufs_bindex_t bindex, br_id, sigen_v; + struct inode *inode, *h_inode; + + //au_debug_on(); + LKTRTrace("%d, fh{i%u, br_id_sigen 0x%x, hi%u}\n", + fh_type, fh[Fh_ino], fh[Fh_br_id_sigen], fh[Fh_h_ino]); + DEBUG_ON(fh_len < Fh_tail); + + si_read_lock(sb); + lockdep_off(); + + /* branch id may be wrapped around */ + dentry = ERR_PTR(-ESTALE); + decode_br_id_sigen(fh[Fh_br_id_sigen], &br_id, &sigen_v); + bindex = find_brindex(sb, br_id); + if (unlikely(bindex < 0 || au_sigen(sb) < sigen_v)) + goto out; + + /* is this inode still cached? */ + ino = decode_ino(fh + Fh_ino); + dir_ino = decode_ino(fh + Fh_dir_ino); + dentry = decode_by_ino(sb, ino, dir_ino); + if (IS_ERR(dentry)) + goto out; + if (dentry) + goto accept; + + /* is the parent dir cached? */ + dentry = decode_by_dir_ino(sb, ino, dir_ino); + if (IS_ERR(dentry)) + goto out; + if (dentry) + goto accept; + + /* lookup path */ + dentry = decode_by_path(sb, bindex, fh, fh_len, context); + if (IS_ERR(dentry)) + goto out; + if (unlikely(!dentry)) + goto out_stale; + if (unlikely(dentry->d_inode->i_ino != ino)) + goto out_dput; + + accept: + inode = dentry->d_inode; + h_inode = NULL; + ii_read_lock_child(inode); + if (ibstart(inode) <= bindex && bindex <= ibend(inode)) + h_inode = au_h_iptr_i(inode, bindex); + ii_read_unlock(inode); + if (h_inode + && h_inode->i_generation == fh[Fh_h_igen] + && acceptable(context, dentry)) + goto out; /* success */ + out_dput: + dput(dentry); + out_stale: + dentry = ERR_PTR(-ESTALE); + out: + lockdep_on(); + si_read_unlock(sb); + TraceErrPtr(dentry); + //au_debug_off(); + return dentry; +} + +/* ---------------------------------------------------------------------- */ + +static int aufs_encode_fh(struct dentry *dentry, __u32 *fh, int *max_len, + int connectable) +{ + int err; + struct super_block *sb, *h_sb; + struct inode *inode, *h_inode, *dir; + aufs_bindex_t bindex; + union conv u; + struct dentry *parent, *h_parent; + + //au_debug_on(); + BUILD_BUG_ON(sizeof(u.ino) != sizeof(u.a)); + LKTRTrace("%.*s, max %d, conn %d\n", + DLNPair(dentry), *max_len, connectable); + DEBUG_ON(is_anon(dentry)); + inode = dentry->d_inode; + DEBUG_ON(!inode); + parent = dentry->d_parent; + DEBUG_ON(is_anon(parent)); + + err = -ENOSPC; + if (unlikely(*max_len <= Fh_tail)) { + Warn1("NFSv2 client (max_len %d)?\n", *max_len); + goto out; + } + + sb = dentry->d_sb; + si_read_lock(sb); + di_read_lock_child(dentry, AUFS_I_RLOCK); + di_read_lock_parent(parent, AUFS_I_RLOCK); +#ifdef CONFIG_AUFS_DEBUG + if (unlikely(!au_flag_test(sb, AuFlag_XINO))) + Warn1("NFS-exporting requires xino\n"); +#if 0 + if (unlikely(au_flag_test(sb, AuFlag_UDBA_INOTIFY))) + Warn1("udba=inotify is not recommended when exporting\n"); +#endif +#endif + + err = -EPERM; + bindex = ibstart(inode); + h_sb = sbr_sb(sb, bindex); + if (unlikely(!h_sb->s_export_op)) { + Err1("%s branch is not exportable\n", au_sbtype(h_sb)); + goto out_unlock; + } + +#if 0 //def CONFIG_AUFS_ROBR + if (unlikely(SB_AUFS(h_sb))) { + Err1("aufs branch is not supported\n"); + goto out_unlock; + } +#endif + + /* doesn't support pseudo-link */ + if (unlikely(bindex < dbstart(dentry) + || dbend(dentry) < bindex + || !au_h_dptr_i(dentry, bindex))) { + Err("%.*s/%.*s, b%d, pseudo-link?\n", + DLNPair(dentry->d_parent), DLNPair(dentry), bindex); + goto out_unlock; + } + + fh[Fh_br_id_sigen] = encode_br_id_sigen(sbr_id(sb, bindex), + au_sigen(sb)); + encode_ino(fh + Fh_ino, inode->i_ino); + dir = parent->d_inode; + encode_ino(fh + Fh_dir_ino, dir->i_ino); + h_inode = au_h_iptr(inode); + encode_ino(fh + Fh_h_ino, h_inode->i_ino); + fh[Fh_h_igen] = h_inode->i_generation; + + /* it should be set at exporting time */ + if (unlikely(!h_sb->s_export_op->find_exported_dentry)) { + Warn("set default find_exported_dentry for %s\n", + au_sbtype(h_sb)); + h_sb->s_export_op->find_exported_dentry = find_exported_dentry; + } + + *max_len -= Fh_tail; + //LKTRTrace("Fh_tail %d, max_len %d\n", Fh_tail, *max_len); + h_parent = au_h_dptr_i(parent, bindex); + DEBUG_ON(is_anon(h_parent)); + err = fh[Fh_h_type] = CALL(h_sb->s_export_op, encode_fh) + (h_parent, fh + Fh_tail, max_len, connectable); + *max_len += Fh_tail; + if (err != 255) + err = 2; //?? + else + Warn1("%s encode_fh failed\n", au_sbtype(h_sb)); + + out_unlock: + di_read_unlock(parent, AUFS_I_RLOCK); + aufs_read_unlock(dentry, AUFS_I_RLOCK); + out: + TraceErr(err); + //au_debug_off(); + if (unlikely(err < 0)) + err = 255; + return err; +} + +/* ---------------------------------------------------------------------- */ + +#if 0 +struct export_operations { + struct dentry *(*decode_fh)(struct super_block *sb, __u32 *fh, int fh_len, int fh_type, + int (*acceptable)(void *context, struct dentry *de), + void *context); + int (*encode_fh)(struct dentry *de, __u32 *fh, int *max_len, + int connectable); + + /* the following are only called from the filesystem itself */ + int (*get_name)(struct dentry *parent, char *name, + struct dentry *child); + struct dentry * (*get_parent)(struct dentry *child); + struct dentry * (*get_dentry)(struct super_block *sb, void *inump); + + /* This is set by the exporting module to a standard helper */ + struct dentry * (*find_exported_dentry)( + struct super_block *sb, void *obj, void *parent, + int (*acceptable)(void *context, struct dentry *de), + void *context); +}; +#endif + +struct export_operations aufs_export_op = { + .decode_fh = aufs_decode_fh, + .encode_fh = aufs_encode_fh +}; diff --git a/fs/aufs/f_op.c b/fs/aufs/f_op.c new file mode 100755 index 0000000..3cd1081 --- /dev/null +++ b/fs/aufs/f_op.c @@ -0,0 +1,684 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: f_op.c,v 1.27 2007/05/14 03:38:24 sfjro Exp $ */ + +#include +#include +#include +#include +#include +#include "aufs.h" + +/* common function to regular file and dir */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) +#define FlushArgs hidden_file, id +int aufs_flush(struct file *file, fl_owner_t id) +#else +#define FlushArgs hidden_file +int aufs_flush(struct file *file) +#endif +{ + int err; + struct dentry *dentry; + aufs_bindex_t bindex, bend; + + dentry = file->f_dentry; + LKTRTrace("%.*s\n", DLNPair(dentry)); + + // aufs_read_lock_file() + si_read_lock(dentry->d_sb); + fi_read_lock(file); + di_read_lock_child(dentry, !AUFS_I_RLOCK); + + err = 0; + bend = fbend(file); + for (bindex = fbstart(file); !err && bindex <= bend; bindex++) { + struct file *hidden_file; + hidden_file = au_h_fptr_i(file, bindex); + if (hidden_file && hidden_file->f_op + && hidden_file->f_op->flush) + err = hidden_file->f_op->flush(FlushArgs); + } + + di_read_unlock(dentry, !AUFS_I_RLOCK); + fi_read_unlock(file); + si_read_unlock(dentry->d_sb); + TraceErr(err); + return err; +} +#undef FlushArgs + +/* ---------------------------------------------------------------------- */ + +static int do_open_nondir(struct file *file, int flags) +{ + int err; + aufs_bindex_t bindex; + struct super_block *sb; + struct file *hidden_file; + struct dentry *dentry; + struct inode *inode; + struct aufs_finfo *finfo; + + dentry = file->f_dentry; + LKTRTrace("%.*s, flags 0%o\n", DLNPair(dentry), flags); + FiMustWriteLock(file); + inode = dentry->d_inode; + DEBUG_ON(!inode || S_ISDIR(inode->i_mode)); + + err = 0; + finfo = ftofi(file); + finfo->fi_h_vm_ops = NULL; + sb = dentry->d_sb; + bindex = dbstart(dentry); + DEBUG_ON(!au_h_dptr(dentry)->d_inode); + /* O_TRUNC is processed already */ + BUG_ON(test_ro(sb, bindex, inode) && (flags & O_TRUNC)); + + hidden_file = hidden_open(dentry, bindex, flags); + //if (LktrCond) {fput(hidden_file); br_put(stobr(dentry->d_sb, bindex)); + //hidden_file = ERR_PTR(-1);} + if (!IS_ERR(hidden_file)) { + set_fbstart(file, bindex); + set_fbend(file, bindex); + set_h_fptr(file, bindex, hidden_file); + return 0; /* success */ + } + err = PTR_ERR(hidden_file); + TraceErr(err); + return err; +} + +static int aufs_open_nondir(struct inode *inode, struct file *file) +{ + return au_do_open(inode, file, do_open_nondir); +} + +static int aufs_release_nondir(struct inode *inode, struct file *file) +{ + struct super_block *sb = file->f_dentry->d_sb; + + LKTRTrace("i%lu, %.*s\n", inode->i_ino, DLNPair(file->f_dentry)); + + si_read_lock(sb); + au_fin_finfo(file); + si_read_unlock(sb); + return 0; +} + +/* ---------------------------------------------------------------------- */ + +static ssize_t aufs_read(struct file *file, char __user *buf, size_t count, + loff_t *ppos) +{ + ssize_t err; + struct dentry *dentry; + struct file *hidden_file; + struct super_block *sb; + struct inode *h_inode; + int dlgt; + + dentry = file->f_dentry; + LKTRTrace("%.*s, cnt %lu, pos %Ld\n", + DLNPair(dentry), (unsigned long)count, *ppos); + + sb = dentry->d_sb; + si_read_lock(sb); + err = au_reval_and_lock_finfo(file, au_reopen_nondir, /*wlock*/0, + /*locked*/0); + //if (LktrCond) {fi_read_unlock(file); err = -1;} + if (unlikely(err)) + goto out; + + /* support LSM and notify */ + dlgt = need_dlgt(sb); + hidden_file = au_h_fptr(file); + h_inode = hidden_file->f_dentry->d_inode; + if (!au_flag_test(sb, AuFlag_UDBA_INOTIFY)) + err = vfsub_read_u(hidden_file, buf, count, ppos, dlgt); + else { + struct inode *dir = dentry->d_parent->d_inode, + *h_dir = hidden_file->f_dentry->d_parent->d_inode; + aufs_bindex_t bstart = fbstart(file); + hdir_lock(h_dir, dir, bstart); + err = vfsub_read_u(hidden_file, buf, count, ppos, dlgt); + hdir_unlock(h_dir, dir, bstart); + } + memcpy(&file->f_ra, &hidden_file->f_ra, sizeof(file->f_ra)); //?? + dentry->d_inode->i_atime = hidden_file->f_dentry->d_inode->i_atime; + + fi_read_unlock(file); + out: + si_read_unlock(sb); + TraceErr(err); + return err; +} + +static ssize_t aufs_write(struct file *file, const char __user *__buf, + size_t count, loff_t *ppos) +{ + ssize_t err; + struct dentry *dentry; + struct inode *inode; + struct super_block *sb; + struct file *hidden_file; + char __user *buf = (char __user*)__buf; + struct inode *h_inode; + int dlgt; + + dentry = file->f_dentry; + LKTRTrace("%.*s, cnt %lu, pos %Ld\n", + DLNPair(dentry), (unsigned long)count, *ppos); + + inode = dentry->d_inode; + i_lock(inode); + sb = dentry->d_sb; + si_read_lock(sb); + err = au_reval_and_lock_finfo(file, au_reopen_nondir, /*wlock*/1, + /*locked*/1); + //if (LktrCond) {fi_write_unlock(file); err = -1;} + if (unlikely(err)) + goto out; + err = au_ready_to_write(file, -1); + //if (LktrCond) err = -1; + if (unlikely(err)) + goto out_unlock; + + /* support LSM and notify */ + dlgt = need_dlgt(sb); + hidden_file = au_h_fptr(file); + h_inode = hidden_file->f_dentry->d_inode; + if (!au_flag_test(sb, AuFlag_UDBA_INOTIFY)) + err = vfsub_write_u(hidden_file, buf, count, ppos, dlgt); + else { + struct inode *dir = dentry->d_parent->d_inode, + *h_dir = hidden_file->f_dentry->d_parent->d_inode; + aufs_bindex_t bstart = fbstart(file); + hdir_lock(h_dir, dir, bstart); + err = vfsub_write_u(hidden_file, buf, count, ppos, dlgt); + hdir_unlock(h_dir, dir, bstart); + } + ii_write_lock_child(inode); + au_cpup_attr_timesizes(inode); + ii_write_unlock(inode); + + out_unlock: + fi_write_unlock(file); + out: + si_read_unlock(sb); + i_unlock(inode); + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +#if 0 //def CONFIG_AUFS_ROBR +struct lvma { + struct list_head list; + struct vm_area_struct *vma; +}; + +static struct file *safe_file(struct vm_area_struct *vma) +{ + struct file *file = vma->vm_file; + struct super_block *sb = file->f_dentry->d_sb; + struct lvma *lvma, *entry; + struct aufs_sbinfo *sbinfo; + int found, warn; + + TraceEnter(); + DEBUG_ON(!SB_AUFS(sb)); + + warn = 0; + found = 0; + sbinfo = stosi(sb); + spin_lock(&sbinfo->si_lvma_lock); + list_for_each_entry(entry, &sbinfo->si_lvma, list) { + found = (entry->vma == vma); + if (unlikely(found)) + break; + } + if (!found) { + lvma = kmalloc(sizeof(*lvma), GFP_ATOMIC); + if (lvma) { + lvma->vma = vma; + list_add(&lvma->list, &sbinfo->si_lvma); + } else { + warn = 1; + file = NULL; + } + } else + file = NULL; + spin_unlock(&sbinfo->si_lvma_lock); + + if (unlikely(warn)) + Warn1("no memory for lvma\n"); + return file; +} + +static void reset_file(struct vm_area_struct *vma, struct file *file) +{ + struct super_block *sb = file->f_dentry->d_sb; + struct lvma *entry, *found; + struct aufs_sbinfo *sbinfo; + + TraceEnter(); + DEBUG_ON(!SB_AUFS(sb)); + + vma->vm_file = file; + + found = NULL; + sbinfo = stosi(sb); + spin_lock(&sbinfo->si_lvma_lock); + list_for_each_entry(entry, &sbinfo->si_lvma, list) + if (entry->vma == vma){ + found = entry; + break; + } + DEBUG_ON(!found); + list_del(&found->list); + spin_unlock(&sbinfo->si_lvma_lock); + kfree(found); +} + +#else + +static struct file *safe_file(struct vm_area_struct *vma) +{ + struct file *file; + + file = vma->vm_file; + if (file->private_data && au_is_aufs(file->f_dentry->d_sb)) + return file; + return NULL; +} + +static void reset_file(struct vm_area_struct *vma, struct file *file) +{ + vma->vm_file = file; + smp_mb(); +} +#endif /* CONFIG_AUFS_ROBR */ + +static struct page *aufs_nopage(struct vm_area_struct *vma, unsigned long addr, + int *type) +{ + struct page *page; + struct dentry *dentry; + struct file *file, *hidden_file; + struct inode *inode; + static DECLARE_WAIT_QUEUE_HEAD(wq); + struct aufs_finfo *finfo; + + TraceEnter(); + DEBUG_ON(!vma || !vma->vm_file); + wait_event(wq, (file = safe_file(vma))); + DEBUG_ON(!au_is_aufs(file->f_dentry->d_sb)); + dentry = file->f_dentry; + LKTRTrace("%.*s, addr %lx\n", DLNPair(dentry), addr); + inode = dentry->d_inode; + DEBUG_ON(!S_ISREG(inode->i_mode)); + + // do not revalidate, nor lock + finfo = ftofi(file); + hidden_file = finfo->fi_hfile[0 + finfo->fi_bstart].hf_file; + DEBUG_ON(!hidden_file || !au_is_mmapped(file)); + vma->vm_file = hidden_file; + //smp_mb(); + page = finfo->fi_h_vm_ops->nopage(vma, addr, type); + reset_file(vma, file); +#if 0 //def CONFIG_SMP + //wake_up_nr(&wq, online_cpu - 1); + wake_up_all(&wq); +#else + wake_up(&wq); +#endif + if (!IS_ERR(page)) { + //page->mapping = file->f_mapping; + //get_page(page); + //file->f_mapping = hidden_file->f_mapping; + //touch_atime(NULL, dentry); + //inode->i_atime = hidden_file->f_dentry->d_inode->i_atime; + } + TraceErrPtr(page); + return page; +} + +static int aufs_populate(struct vm_area_struct *vma, unsigned long addr, + unsigned long len, pgprot_t prot, unsigned long pgoff, + int nonblock) +{ + Err("please report me this application\n"); + BUG(); + return ftofi(vma->vm_file)->fi_h_vm_ops->populate + (vma, addr, len, prot, pgoff, nonblock); +} + +static struct vm_operations_struct aufs_vm_ops = { + //.open = aufs_vmaopen, + //.close = aufs_vmaclose, + .nopage = aufs_nopage, + .populate = aufs_populate, + //page_mkwrite(struct vm_area_struct *vma, struct page *page) +}; + +/* ---------------------------------------------------------------------- */ + +static int aufs_mmap(struct file *file, struct vm_area_struct *vma) +{ + int err, wlock, mmapped; + struct dentry *dentry; + struct super_block *sb; + struct file *h_file; + struct vm_operations_struct *vm_ops; + unsigned long flags; + + dentry = file->f_dentry; + LKTRTrace("%.*s, %lx, len %lu\n", + DLNPair(dentry), vma->vm_start, vma->vm_end - vma->vm_start); + DEBUG_ON(!S_ISREG(dentry->d_inode->i_mode)); + DEBUG_ON(down_write_trylock(&vma->vm_mm->mmap_sem)); + + mmapped = au_is_mmapped(file); + wlock = 0; + if (file->f_mode & FMODE_WRITE) { + flags = VM_SHARED | VM_WRITE; + wlock = ((flags & vma->vm_flags) == flags); + } + + sb = dentry->d_sb; + si_read_lock(sb); + err = au_reval_and_lock_finfo(file, au_reopen_nondir, + wlock | !mmapped, /*locked*/0); + //err = -1; + if (unlikely(err)) + goto out; + + if (wlock) { + err = au_ready_to_write(file, -1); + //err = -1; + if (unlikely(err)) + goto out_unlock; + } + + h_file = au_h_fptr(file); + vm_ops = ftofi(file)->fi_h_vm_ops; + if (unlikely(!mmapped)) { + // nfs uses some locks + lockdep_off(); + err = h_file->f_op->mmap(h_file, vma); + lockdep_on(); + if (unlikely(err)) + goto out_unlock; + vm_ops = vma->vm_ops; + DEBUG_ON(!vm_ops); + err = do_munmap(current->mm, vma->vm_start, + vma->vm_end - vma->vm_start); + if (unlikely(err)) { + IOErr("failed internal unmapping %.*s, %d\n", + DLNPair(h_file->f_dentry), err); + err = -EIO; + goto out_unlock; + } + } + DEBUG_ON(!vm_ops); + + err = generic_file_mmap(file, vma); + if (!err) { + file_accessed(h_file); + dentry->d_inode->i_atime = h_file->f_dentry->d_inode->i_atime; + vma->vm_ops = &aufs_vm_ops; + if (unlikely(!mmapped)) + ftofi(file)->fi_h_vm_ops = vm_ops; + } + + out_unlock: + if (!wlock && mmapped) + fi_read_unlock(file); + else + fi_write_unlock(file); + out: + si_read_unlock(sb); + TraceErr(err); + return err; +} + +// todo: try do_sendfile() in fs/read_write.c +static ssize_t aufs_sendfile(struct file *file, loff_t *ppos, + size_t count, read_actor_t actor, void *target) +{ + ssize_t err; + struct file *h_file; + const char c = current->comm[4]; + /* true if a kernel thread named 'loop[0-9].*' accesses a file */ + const int loopback = (current->mm == NULL + && '0' <= c && c <= '9' + && strncmp(current->comm, "loop", 4) == 0); + struct dentry *dentry; + struct super_block *sb; + + dentry = file->f_dentry; + LKTRTrace("%.*s, pos %Ld, cnt %lu, loopback %d\n", + DLNPair(dentry), *ppos, (unsigned long)count, loopback); + + sb = dentry->d_sb; + si_read_lock(sb); + err = au_reval_and_lock_finfo(file, au_reopen_nondir, /*wlock*/0, + /*locked*/0); + if (unlikely(err)) + goto out; + + err = -EINVAL; + h_file = au_h_fptr(file); + if (h_file->f_op && h_file->f_op->sendfile) { + if (/* unlikely */(loopback)) { + file->f_mapping = h_file->f_mapping; + smp_mb(); //?? + } + // nfs uses some locks + lockdep_off(); + err = h_file->f_op->sendfile + (h_file, ppos, count, actor, target); + lockdep_on(); + dentry->d_inode->i_atime = h_file->f_dentry->d_inode->i_atime; + } + fi_read_unlock(file); + + out: + si_read_unlock(sb); + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +/* copied from linux/fs/select.h, must match */ +#define DEFAULT_POLLMASK (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM) + +static unsigned int aufs_poll(struct file *file, poll_table *wait) +{ + unsigned int mask; + struct file *hidden_file; + int err; + struct dentry *dentry; + struct super_block *sb; + + dentry = file->f_dentry; + LKTRTrace("%.*s, wait %p\n", DLNPair(dentry), wait); + DEBUG_ON(S_ISDIR(dentry->d_inode->i_mode)); + + /* We should pretend an error happend. */ + mask = POLLERR /* | POLLIN | POLLOUT */; + sb = dentry->d_sb; + si_read_lock(sb); + err = au_reval_and_lock_finfo(file, au_reopen_nondir, /*wlock*/0, + /*locked*/0); + //err = -1; + if (unlikely(err)) + goto out; + + /* it is not an error of hidden_file has no operation */ + mask = DEFAULT_POLLMASK; + hidden_file = au_h_fptr(file); + if (hidden_file->f_op && hidden_file->f_op->poll) + mask = hidden_file->f_op->poll(hidden_file, wait); + fi_read_unlock(file); + + out: + si_read_unlock(sb); + TraceErr((int)mask); + return mask; +} + +static int aufs_fsync_nondir(struct file *file, struct dentry *dentry, + int datasync) +{ + int err, my_lock; + struct inode *inode; + struct file *hidden_file; + struct super_block *sb; + + LKTRTrace("%.*s, %d\n", DLNPair(dentry), datasync); + inode = dentry->d_inode; +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,17) + IMustLock(inode); + my_lock = 0; +#else + /* before 2.6.17, + * msync(2) calls me without locking i_sem/i_mutex, but fsync(2). + */ + my_lock = !i_trylock(inode); +#endif + + sb = dentry->d_sb; + si_read_lock(sb); + err = 0; //-EBADF; // posix? + if (unlikely(!(file->f_mode & FMODE_WRITE))) + goto out; + err = au_reval_and_lock_finfo(file, au_reopen_nondir, /*wlock*/1, + /*locked*/1); + //err = -1; + if (unlikely(err)) + goto out; + err = au_ready_to_write(file, -1); + //err = -1; + if (unlikely(err)) + goto out_unlock; + + err = -EINVAL; + hidden_file = au_h_fptr(file); + if (hidden_file->f_op && hidden_file->f_op->fsync) { + // todo: apparmor thread? + //file->f_mapping->host->i_mutex + ii_write_lock_child(inode); + hi_lock_child(hidden_file->f_dentry->d_inode); + err = hidden_file->f_op->fsync + (hidden_file, hidden_file->f_dentry, datasync); + //err = -1; + au_cpup_attr_timesizes(inode); + i_unlock(hidden_file->f_dentry->d_inode); + ii_write_unlock(inode); + } + + out_unlock: + fi_write_unlock(file); + out: + if (unlikely(my_lock)) + i_unlock(inode); + si_read_unlock(sb); + TraceErr(err); + return err; +} + +static int aufs_fasync(int fd, struct file *file, int flag) +{ + int err; + struct file *hidden_file; + struct dentry *dentry; + struct super_block *sb; + + dentry = file->f_dentry; + LKTRTrace("%.*s, %d\n", DLNPair(dentry), flag); + + sb = dentry->d_sb; + si_read_lock(sb); + err = au_reval_and_lock_finfo(file, au_reopen_nondir, /*wlock*/0, + /*locked*/0); + //err = -1; + if (unlikely(err)) + goto out; + + hidden_file = au_h_fptr(file); + if (hidden_file->f_op && hidden_file->f_op->fasync) + err = hidden_file->f_op->fasync(fd, hidden_file, flag); + fi_read_unlock(file); + + out: + si_read_unlock(sb); + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +#if 0 // comment +struct file_operations { + struct module *owner; + loff_t (*llseek) (struct file *, loff_t, int); + ssize_t (*read) (struct file *, char __user *, size_t, loff_t *); + ssize_t (*aio_read) (struct kiocb *, char __user *, size_t, loff_t); + ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *); + ssize_t (*aio_write) (struct kiocb *, const char __user *, size_t, loff_t); + int (*readdir) (struct file *, void *, filldir_t); + unsigned int (*poll) (struct file *, struct poll_table_struct *); + int (*ioctl) (struct inode *, struct file *, unsigned int, unsigned long); + long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long); + long (*compat_ioctl) (struct file *, unsigned int, unsigned long); + int (*mmap) (struct file *, struct vm_area_struct *); + int (*open) (struct inode *, struct file *); + int (*flush) (struct file *); + int (*release) (struct inode *, struct file *); + int (*fsync) (struct file *, struct dentry *, int datasync); + int (*aio_fsync) (struct kiocb *, int datasync); + int (*fasync) (int, struct file *, int); + int (*lock) (struct file *, int, struct file_lock *); + ssize_t (*readv) (struct file *, const struct iovec *, unsigned long, loff_t *); + ssize_t (*writev) (struct file *, const struct iovec *, unsigned long, loff_t *); + ssize_t (*sendfile) (struct file *, loff_t *, size_t, read_actor_t, void *); + ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int); + unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); + int (*check_flags)(int); + int (*dir_notify)(struct file *file, unsigned long arg); + int (*flock) (struct file *, int, struct file_lock *); +}; +#endif + +struct file_operations aufs_file_fop = { + .read = aufs_read, + .write = aufs_write, + .poll = aufs_poll, + .mmap = aufs_mmap, + .open = aufs_open_nondir, + .flush = aufs_flush, + .release = aufs_release_nondir, + .fsync = aufs_fsync_nondir, + .fasync = aufs_fasync, + .sendfile = aufs_sendfile, +}; diff --git a/fs/aufs/file.c b/fs/aufs/file.c new file mode 100755 index 0000000..857a4e8 --- /dev/null +++ b/fs/aufs/file.c @@ -0,0 +1,832 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: file.c,v 1.42 2007/05/14 03:39:09 sfjro Exp $ */ + +//#include +#include +//#include +//#include +#include "aufs.h" + +/* drop flags for writing */ +unsigned int au_file_roflags(unsigned int flags) +{ + flags &= ~(O_WRONLY | O_RDWR | O_APPEND | O_CREAT | O_TRUNC); + flags |= O_RDONLY | O_NOATIME; + return flags; +} + +/* common functions to regular file and dir */ +struct file *hidden_open(struct dentry *dentry, aufs_bindex_t bindex, int flags) +{ + struct dentry *hidden_dentry; + struct inode *hidden_inode; + struct super_block *sb; + struct vfsmount *hidden_mnt; + struct file *hidden_file; + struct aufs_branch *br; + loff_t old_size; + int udba; + + LKTRTrace("%.*s, b%d, flags 0%o\n", DLNPair(dentry), bindex, flags); + DEBUG_ON(!dentry); + hidden_dentry = au_h_dptr_i(dentry, bindex); + DEBUG_ON(!hidden_dentry); + hidden_inode = hidden_dentry->d_inode; + DEBUG_ON(!hidden_inode); + + sb = dentry->d_sb; + udba = au_flag_test(sb, AuFlag_UDBA_INOTIFY); + if (unlikely(udba)) { + // test here? + } + + br = stobr(sb, bindex); + br_get(br); + /* drop flags for writing */ + if (test_ro(sb, bindex, dentry->d_inode)) + flags = au_file_roflags(flags); + flags &= ~O_CREAT; + spin_lock(&hidden_inode->i_lock); + old_size = i_size_read(hidden_inode); + spin_unlock(&hidden_inode->i_lock); + + //DbgSleep(3); + + dget(hidden_dentry); + hidden_mnt = mntget(br->br_mnt); + hidden_file = dentry_open(hidden_dentry, hidden_mnt, flags); + //if (LktrCond) {fput(hidden_file); hidden_file = ERR_PTR(-1);} + + if (!IS_ERR(hidden_file)) { +#if 0 // remove this + if (/* old_size && */ (flags & O_TRUNC)) { + au_direval_dec(dentry); + if (!IS_ROOT(dentry)) + au_direval_dec(dentry->d_parent); + } +#endif + return hidden_file; + } + + br_put(br); + TraceErrPtr(hidden_file); + return hidden_file; +} + +static int do_coo(struct dentry *dentry, aufs_bindex_t bstart) +{ + int err; + struct dentry *parent, *h_parent, *h_dentry; + aufs_bindex_t bcpup; + struct inode *h_dir, *h_inode, *dir; + + LKTRTrace("%.*s\n", DLNPair(dentry)); + DEBUG_ON(IS_ROOT(dentry)); + DiMustWriteLock(dentry); + + parent = dentry->d_parent; // dget_parent() + di_write_lock_parent(parent); + bcpup = err = find_rw_parent_br(dentry, bstart); + //bcpup = err = find_rw_br(sb, bstart); + if (unlikely(err < 0)) { + err = 0; // stop copyup, it is not an error + goto out; + } + err = 0; + + h_parent = au_h_dptr_i(parent, bcpup); + if (!h_parent) { + err = cpup_dirs(dentry, bcpup, NULL); + if (unlikely(err)) + goto out; + h_parent = au_h_dptr_i(parent, bcpup); + } + + h_dir = h_parent->d_inode; + h_dentry = au_h_dptr_i(dentry, bstart); + h_inode = h_dentry->d_inode; + dir = parent->d_inode; + hdir_lock(h_dir, dir, bcpup); + hi_lock_child(h_inode); + DEBUG_ON(au_h_dptr_i(dentry, bcpup)); + err = sio_cpup_simple(dentry, bcpup, -1, + au_flags_cpup(CPUP_DTIME, parent)); + TraceErr(err); + i_unlock(h_inode); + hdir_unlock(h_dir, dir, bcpup); + + out: + di_write_unlock(parent); + TraceErr(err); + return err; +} + +int au_do_open(struct inode *inode, struct file *file, + int (*open)(struct file *file, int flags)) +{ + int err, coo; + struct dentry *dentry; + struct super_block *sb; + aufs_bindex_t bstart; + struct inode *h_dir, *dir; + + dentry = file->f_dentry; + LKTRTrace("i%lu, %.*s\n", inode->i_ino, DLNPair(dentry)); + + sb = dentry->d_sb; + si_read_lock(sb); + coo = 0; +#if 0 + switch (au_flag_test_coo(sb)) { + case AuFlag_COO_LEAF: + coo = !S_ISDIR(inode->i_mode); + break; + case AuFlag_COO_ALL: + coo = 1; + break; + } +#endif + err = au_init_finfo(file); + //if (LktrCond) {fi_write_unlock(file); fin_finfo(file); err = -1;} + if (unlikely(err)) + goto out; + + if (!coo) { + di_read_lock_child(dentry, AUFS_I_RLOCK); + bstart = dbstart(dentry); + } else { + di_write_lock_child(dentry); + bstart = dbstart(dentry); + if (test_ro(sb, bstart, dentry->d_inode)) { + err = do_coo(dentry, bstart); + if (err) { + di_write_unlock(dentry); + goto out_finfo; + } + bstart = dbstart(dentry); + } + di_downgrade_lock(dentry, AUFS_I_RLOCK); + } + + // todo: remove this extra locks + dir = dentry->d_parent->d_inode; + if (!IS_ROOT(dentry)) + ii_read_lock_parent(dir); + h_dir = au_h_iptr_i(dir, bstart); + hdir_lock(h_dir, dir, bstart); + err = open(file, file->f_flags); + //if (LktrCond) err = -1; + hdir_unlock(h_dir, dir, bstart); + if (!IS_ROOT(dentry)) + ii_read_unlock(dir); + di_read_unlock(dentry, AUFS_I_RLOCK); + + out_finfo: + fi_write_unlock(file); + if (unlikely(err)) + au_fin_finfo(file); + //DbgFile(file); + out: + si_read_unlock(sb); + TraceErr(err); + return err; +} + +int au_reopen_nondir(struct file *file) +{ + int err; + struct dentry *dentry; + aufs_bindex_t bstart, bindex, bend; + struct file *hidden_file, *h_file_tmp; + + dentry = file->f_dentry; + LKTRTrace("%.*s\n", DLNPair(dentry)); + DEBUG_ON(S_ISDIR(dentry->d_inode->i_mode) + || !au_h_dptr(dentry)->d_inode); + bstart = dbstart(dentry); + + h_file_tmp = NULL; + if (fbstart(file) == bstart) { + hidden_file = au_h_fptr(file); + if (file->f_mode == hidden_file->f_mode) + return 0; /* success */ + h_file_tmp = hidden_file; + get_file(h_file_tmp); + set_h_fptr(file, bstart, NULL); + } + DEBUG_ON(fbstart(file) < bstart + || ftofi(file)->fi_hfile[0 + bstart].hf_file); + + hidden_file = hidden_open(dentry, bstart, file->f_flags & ~O_TRUNC); + //if (LktrCond) {fput(hidden_file); br_put(stobr(dentry->d_sb, bstart)); + //hidden_file = ERR_PTR(-1);} + err = PTR_ERR(hidden_file); + if (IS_ERR(hidden_file)) + goto out; // close all? + err = 0; + //cpup_file_flags(hidden_file, file); + set_fbstart(file, bstart); + set_h_fptr(file, bstart, hidden_file); + memcpy(&hidden_file->f_ra, &file->f_ra, sizeof(file->f_ra)); //?? + + /* close lower files */ + bend = fbend(file); + for (bindex = bstart + 1; bindex <= bend; bindex++) + set_h_fptr(file, bindex, NULL); + set_fbend(file, bstart); + + out: + if (h_file_tmp) + fput(h_file_tmp); + TraceErr(err); + return err; +} + +/* + * copyup the deleted file for writing. + */ +static int cpup_wh_file(struct file *file, aufs_bindex_t bdst, loff_t len) +{ + int err; + struct dentry *dentry, *parent, *hidden_parent, *tmp_dentry; + struct dentry *hidden_dentry_bstart, *hidden_dentry_bdst; + struct inode *hidden_dir; + aufs_bindex_t bstart; + struct aufs_dinfo *dinfo; + struct dtime dt; + struct lkup_args lkup; + struct super_block *sb; + + dentry = file->f_dentry; + LKTRTrace("%.*s, bdst %d, len %Lu\n", DLNPair(dentry), bdst, len); + DEBUG_ON(S_ISDIR(dentry->d_inode->i_mode) + || !(file->f_mode & FMODE_WRITE)); + DiMustWriteLock(dentry); + parent = dentry->d_parent; + IiMustAnyLock(parent->d_inode); + hidden_parent = au_h_dptr_i(parent, bdst); + DEBUG_ON(!hidden_parent); + hidden_dir = hidden_parent->d_inode; + DEBUG_ON(!hidden_dir); + IMustLock(hidden_dir); + + sb = parent->d_sb; + lkup.nfsmnt = au_nfsmnt(sb, bdst); + lkup.dlgt = need_dlgt(sb); + tmp_dentry = lkup_whtmp(hidden_parent, &dentry->d_name, &lkup); + //if (LktrCond) {dput(tmp_dentry); tmp_dentry = ERR_PTR(-1);} + err = PTR_ERR(tmp_dentry); + if (IS_ERR(tmp_dentry)) + goto out; + + dtime_store(&dt, parent, hidden_parent); + dinfo = dtodi(dentry); + bstart = dinfo->di_bstart; + hidden_dentry_bdst = dinfo->di_hdentry[0 + bdst].hd_dentry; + hidden_dentry_bstart = dinfo->di_hdentry[0 + bstart].hd_dentry; + dinfo->di_bstart = bdst; + dinfo->di_hdentry[0 + bdst].hd_dentry = tmp_dentry; + dinfo->di_hdentry[0 + bstart].hd_dentry = au_h_fptr(file)->f_dentry; + err = cpup_single(dentry, bdst, bstart, len, + au_flags_cpup(!CPUP_DTIME, parent)); + //if (LktrCond) err = -1; + if (!err) + err = au_reopen_nondir(file); + //err = -1; + dinfo->di_hdentry[0 + bstart].hd_dentry = hidden_dentry_bstart; + dinfo->di_hdentry[0 + bdst].hd_dentry = hidden_dentry_bdst; + dinfo->di_bstart = bstart; + if (unlikely(err)) + goto out_tmp; + + DEBUG_ON(!d_unhashed(dentry)); + err = vfsub_unlink(hidden_dir, tmp_dentry, lkup.dlgt); + //if (LktrCond) err = -1; + if (unlikely(err)) { + IOErr("failed remove copied-up tmp file %.*s(%d)\n", + DLNPair(tmp_dentry), err); + err = -EIO; + } + dtime_revert(&dt, !CPUP_LOCKED_GHDIR); + + out_tmp: + dput(tmp_dentry); + out: + TraceErr(err); + return err; +} + +struct cpup_wh_file_args { + int *errp; + struct file *file; + aufs_bindex_t bdst; + loff_t len; +}; + +static void call_cpup_wh_file(void *args) +{ + struct cpup_wh_file_args *a = args; + *a->errp = cpup_wh_file(a->file, a->bdst, a->len); +} + +/* + * prepare the @file for writing. + */ +int au_ready_to_write(struct file *file, loff_t len) +{ + int err; + struct dentry *dentry, *parent, *hidden_dentry, *hidden_parent; + struct inode *hidden_inode, *hidden_dir, *inode, *dir; + struct super_block *sb; + aufs_bindex_t bstart, bcpup; + + dentry = file->f_dentry; + LKTRTrace("%.*s, len %Ld\n", DLNPair(dentry), len); + FiMustWriteLock(file); + + sb = dentry->d_sb; + bstart = fbstart(file); + DEBUG_ON(ftobr(file, bstart) != stobr(sb, bstart)); + + inode = dentry->d_inode; + ii_read_lock_child(inode); + LKTRTrace("rdonly %d, bstart %d\n", test_ro(sb, bstart, inode), bstart); + err = test_ro(sb, bstart, inode); + ii_read_unlock(inode); + if (!err && (au_h_fptr(file)->f_mode & FMODE_WRITE)) + return 0; + + /* need to cpup */ + parent = dentry->d_parent; // dget_parent() + di_write_lock_child(dentry); + di_write_lock_parent(parent); + bcpup = err = find_rw_parent_br(dentry, bstart); + //bcpup = err = find_rw_br(sb, bstart); + if (unlikely(err < 0)) + goto out_unlock; + err = 0; + + hidden_parent = au_h_dptr_i(parent, bcpup); + if (!hidden_parent) { + err = cpup_dirs(dentry, bcpup, NULL); + //if (LktrCond) err = -1; + if (unlikely(err)) + goto out_unlock; + hidden_parent = au_h_dptr_i(parent, bcpup); + } + + hidden_dir = hidden_parent->d_inode; + hidden_dentry = au_h_fptr(file)->f_dentry; + hidden_inode = hidden_dentry->d_inode; + dir = parent->d_inode; + hdir_lock(hidden_dir, dir, bcpup); + hi_lock_child(hidden_inode); + if (d_unhashed(dentry) || d_unhashed(hidden_dentry) + /* || !hidden_inode->i_nlink */) { + if (!au_test_perm(hidden_dir, MAY_EXEC | MAY_WRITE, + need_dlgt(sb))) + err = cpup_wh_file(file, bcpup, len); + else { + struct cpup_wh_file_args args = { + .errp = &err, + .file = file, + .bdst = bcpup, + .len = len + }; + au_wkq_wait(call_cpup_wh_file, &args, /*dlgt*/0); + } + //if (LktrCond) err = -1; + TraceErr(err); + } else { + if (!au_h_dptr_i(dentry, bcpup)) + err = sio_cpup_simple(dentry, bcpup, len, + au_flags_cpup(CPUP_DTIME, + parent)); + //if (LktrCond) err = -1; + TraceErr(err); + if (!err) + err = au_reopen_nondir(file); + //if (LktrCond) err = -1; + TraceErr(err); + } + i_unlock(hidden_inode); + hdir_unlock(hidden_dir, dir, bcpup); + + out_unlock: + di_write_unlock(parent); + di_write_unlock(dentry); +// out: + TraceErr(err); + return err; + +} + +/* ---------------------------------------------------------------------- */ + +/* + * after branch manipulating, refresh the file. + */ +static int refresh_file(struct file *file, int (*reopen)(struct file *file)) +{ + int err, new_sz; + struct dentry *dentry; + aufs_bindex_t bend, bindex, bstart, brid; + struct aufs_hfile *p; + struct aufs_finfo *finfo; + struct super_block *sb; + struct inode *inode; + struct file *hidden_file; + + dentry = file->f_dentry; + LKTRTrace("%.*s\n", DLNPair(dentry)); + FiMustWriteLock(file); + DiMustReadLock(dentry); + inode = dentry->d_inode; + IiMustReadLock(inode); + //au_debug_on(); + //DbgDentry(dentry); + //DbgFile(file); + //au_debug_off(); + + err = -ENOMEM; + sb = dentry->d_sb; + finfo = ftofi(file); + bstart = finfo->fi_bstart; + bend = finfo->fi_bstart; + new_sz = sizeof(*finfo->fi_hfile) * (sbend(sb) + 1); + p = au_kzrealloc(finfo->fi_hfile, sizeof(*p) * (finfo->fi_bend + 1), + new_sz, GFP_KERNEL); + //p = NULL; + if (unlikely(!p)) + goto out; + finfo->fi_hfile = p; + hidden_file = p[0 + bstart].hf_file; + + p = finfo->fi_hfile + finfo->fi_bstart; + brid = p->hf_br->br_id; + bend = finfo->fi_bend; + for (bindex = finfo->fi_bstart; bindex <= bend; bindex++, p++) { + struct aufs_hfile tmp, *q; + aufs_bindex_t new_bindex; + + if (!p->hf_file) + continue; + new_bindex = find_bindex(sb, p->hf_br); + if (new_bindex == bindex) + continue; + if (new_bindex < 0) { // test here + set_h_fptr(file, bindex, NULL); + continue; + } + + /* swap two hidden inode, and loop again */ + q = finfo->fi_hfile + new_bindex; + tmp = *q; + *q = *p; + *p = tmp; + if (tmp.hf_file) { + bindex--; + p--; + } + } + { + aufs_bindex_t s = finfo->fi_bstart, e = finfo->fi_bend; + finfo->fi_bstart = 0; + finfo->fi_bend = sbend(sb); + //au_debug_on(); + //DbgFile(file); + //au_debug_off(); + finfo->fi_bstart = s; + finfo->fi_bend = e; + } + + p = finfo->fi_hfile; + if (!au_is_mmapped(file) && !d_unhashed(dentry)) { + bend = sbend(sb); + for (finfo->fi_bstart = 0; finfo->fi_bstart <= bend; + finfo->fi_bstart++, p++) + if (p->hf_file) { + if (p->hf_file->f_dentry + && p->hf_file->f_dentry->d_inode) + break; + else + au_hfput(p); + } + } else { + bend = find_brindex(sb, brid); + //LKTRTrace("%d\n", bend); + for (finfo->fi_bstart = 0; finfo->fi_bstart < bend; + finfo->fi_bstart++, p++) + if (p->hf_file) + au_hfput(p); + //LKTRTrace("%d\n", finfo->fi_bstart); + bend = sbend(sb); + } + + p = finfo->fi_hfile + bend; + for (finfo->fi_bend = bend; finfo->fi_bend >= finfo->fi_bstart; + finfo->fi_bend--, p--) + if (p->hf_file) { + if (p->hf_file->f_dentry + && p->hf_file->f_dentry->d_inode) + break; + else + au_hfput(p); + } + //Dbg("%d, %d\n", finfo->fi_bstart, finfo->fi_bend); + DEBUG_ON(finfo->fi_bend < finfo->fi_bstart); + //DbgFile(file); + //DbgDentry(file->f_dentry); + + err = 0; +#if 0 // todo: + if (!au_h_dptr(dentry)->d_inode) { + au_update_figen(file); + goto out; /* success */ + } +#endif + + if (unlikely(au_is_mmapped(file) || d_unhashed(dentry))) + goto out_update; /* success */ + + again: + bstart = ibstart(inode); + if (bstart < finfo->fi_bstart + && au_flag_test(sb, AuFlag_PLINK) + && au_is_plinked(sb, inode)) { + struct dentry *parent = dentry->d_parent; // dget_parent() + struct inode *dir = parent->d_inode, *h_dir; + + if (test_ro(sb, bstart, inode)) { + di_read_lock_parent(parent, !AUFS_I_RLOCK); + bstart = err = find_rw_parent_br(dentry, bstart); + //bstart = err = find_rw_br(sb, bstart); + di_read_unlock(parent, !AUFS_I_RLOCK); + //todo: err = -1; + if (unlikely(err < 0)) + goto out; + } + di_read_unlock(dentry, AUFS_I_RLOCK); + di_write_lock_child(dentry); + if (bstart != ibstart(inode)) { // todo + /* someone changed our inode while we were sleeping */ + di_downgrade_lock(dentry, AUFS_I_RLOCK); + goto again; + } + + di_read_lock_parent(parent, AUFS_I_RLOCK); + err = test_and_cpup_dirs(dentry, bstart, NULL); + + // always superio. +#if 1 + h_dir = au_h_dptr_i(parent, bstart)->d_inode; + hdir_lock(h_dir, dir, bstart); + err = sio_cpup_simple(dentry, bstart, -1, + au_flags_cpup(CPUP_DTIME, parent)); + hdir_unlock(h_dir, dir, bstart); + di_read_unlock(parent, AUFS_I_RLOCK); +#else + if (!is_au_wkq(current)) { + struct cpup_pseudo_link_args args = { + .errp = &err, + .dentry = dentry, + .bdst = bstart, + .do_lock = 1 + }; + au_wkq_wait(call_cpup_pseudo_link, &args); + } else + err = cpup_pseudo_link(dentry, bstart, /*do_lock*/1); +#endif + di_downgrade_lock(dentry, AUFS_I_RLOCK); + if (unlikely(err)) + goto out; + } + + err = reopen(file); + //err = -1; + out_update: + if (!err) { + au_update_figen(file); + //DbgFile(file); + return 0; /* success */ + } + + /* error, close all hidden files */ + bend = fbend(file); + for (bindex = fbstart(file); bindex <= bend; bindex++) + set_h_fptr(file, bindex, NULL); + + out: + TraceErr(err); + return err; +} + +/* common function to regular file and dir */ +int au_reval_and_lock_finfo(struct file *file, int (*reopen)(struct file *file), + int wlock, int locked) +{ + int err, sgen, fgen, pseudo_link; + struct dentry *dentry; + struct super_block *sb; + aufs_bindex_t bstart; + + dentry = file->f_dentry; + LKTRTrace("%.*s, w %d, l %d\n", DLNPair(dentry), wlock, locked); + sb = dentry->d_sb; + SiMustAnyLock(sb); + + err = 0; + sgen = au_sigen(sb); + fi_write_lock(file); + fgen = au_figen(file); + di_read_lock_child(dentry, AUFS_I_RLOCK); + bstart = dbstart(dentry); + pseudo_link = (bstart != ibstart(dentry->d_inode)); + di_read_unlock(dentry, AUFS_I_RLOCK); + if (sgen == fgen && !pseudo_link && fbstart(file) == bstart) { + if (!wlock) + fi_downgrade_lock(file); + return 0; /* success */ + } + + LKTRTrace("sgen %d, fgen %d\n", sgen, fgen); + if (sgen != au_digen(dentry)) { + /* + * d_path() and path_lookup() is a simple and good approach + * to revalidate. but si_rwsem in DEBUG_RWSEM will cause a + * deadlock. removed the code. + */ + di_write_lock_child(dentry); + err = au_reval_dpath(dentry, sgen); + //if (LktrCond) err = -1; + di_write_unlock(dentry); + if (unlikely(err < 0)) + goto out; + DEBUG_ON(au_digen(dentry) != sgen); + } + + di_read_lock_child(dentry, AUFS_I_RLOCK); + err = refresh_file(file, reopen); + //if (LktrCond) err = -1; + di_read_unlock(dentry, AUFS_I_RLOCK); + if (!err) { + if (!wlock) + fi_downgrade_lock(file); + } else + fi_write_unlock(file); + + out: + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +// cf. aufs_nopage() +// for madvise(2) +static int aufs_readpage(struct file *file, struct page *page) +{ + TraceEnter(); + unlock_page(page); + return 0; +} + +// they will never be called. +#ifdef CONFIG_AUFS_DEBUG +static int aufs_prepare_write(struct file *file, struct page *page, + unsigned from, unsigned to) +{BUG();return 0;} +static int aufs_commit_write(struct file *file, struct page *page, + unsigned from, unsigned to) +{BUG();return 0;} +static int aufs_writepage(struct page *page, struct writeback_control *wbc) +{BUG();return 0;} +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,17) +static void aufs_sync_page(struct page *page) +{BUG();} +#else +static int aufs_sync_page(struct page *page) +{BUG(); return 0;} +#endif + +#if 0 // comment +static int aufs_writepages(struct address_space *mapping, + struct writeback_control *wbc) +{BUG();return 0;} +static int aufs_readpages(struct file *filp, struct address_space *mapping, + struct list_head *pages, unsigned nr_pages) +{BUG();return 0;} +static sector_t aufs_bmap(struct address_space *mapping, sector_t block) +{BUG();return 0;} +#endif + +static int aufs_set_page_dirty(struct page *page) +{BUG();return 0;} +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,17) +static void aufs_invalidatepage (struct page *page, unsigned long offset) +{BUG();} +#else +static int aufs_invalidatepage (struct page *page, unsigned long offset) +{BUG(); return 0;} +#endif +static int aufs_releasepage (struct page *page, gfp_t gfp) +{BUG();return 0;} +static ssize_t aufs_direct_IO(int rw, struct kiocb *iocb, + const struct iovec *iov, loff_t offset, + unsigned long nr_segs) +{BUG();return 0;} +static struct page* aufs_get_xip_page(struct address_space *mapping, + sector_t offset, int create) +{BUG();return NULL;} +//static int aufs_migratepage (struct page *newpage, struct page *page) +//{BUG();return 0;} +#endif + +#if 0 // comment +struct address_space { + struct inode *host; /* owner: inode, block_device */ + struct radix_tree_root page_tree; /* radix tree of all pages */ + rwlock_t tree_lock; /* and rwlock protecting it */ + unsigned int i_mmap_writable;/* count VM_SHARED mappings */ + struct prio_tree_root i_mmap; /* tree of private and shared mappings */ + struct list_head i_mmap_nonlinear;/*list VM_NONLINEAR mappings */ + spinlock_t i_mmap_lock; /* protect tree, count, list */ + unsigned int truncate_count; /* Cover race condition with truncate */ + unsigned long nrpages; /* number of total pages */ + pgoff_t writeback_index;/* writeback starts here */ + struct address_space_operations *a_ops; /* methods */ + unsigned long flags; /* error bits/gfp mask */ + struct backing_dev_info *backing_dev_info; /* device readahead, etc */ + spinlock_t private_lock; /* for use by the address_space */ + struct list_head private_list; /* ditto */ + struct address_space *assoc_mapping; /* ditto */ +} __attribute__((aligned(sizeof(long)))); + +struct address_space_operations { + int (*writepage)(struct page *page, struct writeback_control *wbc); + int (*readpage)(struct file *, struct page *); + void (*sync_page)(struct page *); + + /* Write back some dirty pages from this mapping. */ + int (*writepages)(struct address_space *, struct writeback_control *); + + /* Set a page dirty. Return true if this dirtied it */ + int (*set_page_dirty)(struct page *page); + + int (*readpages)(struct file *filp, struct address_space *mapping, + struct list_head *pages, unsigned nr_pages); + + /* + * ext3 requires that a successful prepare_write() call be followed + * by a commit_write() call - they must be balanced + */ + int (*prepare_write)(struct file *, struct page *, unsigned, unsigned); + int (*commit_write)(struct file *, struct page *, unsigned, unsigned); + /* Unfortunately this kludge is needed for FIBMAP. Don't use it */ + sector_t (*bmap)(struct address_space *, sector_t); + void (*invalidatepage) (struct page *, unsigned long); + int (*releasepage) (struct page *, gfp_t); + ssize_t (*direct_IO)(int, struct kiocb *, const struct iovec *iov, + loff_t offset, unsigned long nr_segs); + struct page* (*get_xip_page)(struct address_space *, sector_t, + int); + /* migrate the contents of a page to the specified target */ + int (*migratepage) (struct page *, struct page *); +}; +#endif + +struct address_space_operations aufs_aop = { + .readpage = aufs_readpage, +#ifdef CONFIG_AUFS_DEBUG + .writepage = aufs_writepage, + .sync_page = aufs_sync_page, + //.writepages = aufs_writepages, + .set_page_dirty = aufs_set_page_dirty, + //.readpages = aufs_readpages, + .prepare_write = aufs_prepare_write, + .commit_write = aufs_commit_write, + //.bmap = aufs_bmap, + .invalidatepage = aufs_invalidatepage, + .releasepage = aufs_releasepage, + .direct_IO = aufs_direct_IO, + .get_xip_page = aufs_get_xip_page, + //.migratepage = aufs_migratepage +#endif +}; diff --git a/fs/aufs/file.h b/fs/aufs/file.h new file mode 100755 index 0000000..f0fa448 --- /dev/null +++ b/fs/aufs/file.h @@ -0,0 +1,140 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: file.h,v 1.25 2007/05/14 03:41:52 sfjro Exp $ */ + +#ifndef __AUFS_FILE_H__ +#define __AUFS_FILE_H__ + +#ifdef __KERNEL__ + +#include +#include +#include +#include +#include "misc.h" + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) +// SEEK_xxx are defined in linux/fs.h +#else +enum {SEEK_SET, SEEK_CUR, SEEK_END}; +#endif + +/* ---------------------------------------------------------------------- */ + +struct aufs_branch; +struct aufs_hfile { + struct file *hf_file; + struct aufs_branch *hf_br; +}; + +struct aufs_vdir; +struct aufs_finfo { + atomic_t fi_generation; + + struct aufs_rwsem fi_rwsem; + struct aufs_hfile *fi_hfile; + aufs_bindex_t fi_bstart, fi_bend; + + union { + struct vm_operations_struct *fi_h_vm_ops; + struct aufs_vdir *fi_vdir_cache; + }; +}; + +/* ---------------------------------------------------------------------- */ + +/* file.c */ +extern struct address_space_operations aufs_aop; +unsigned int au_file_roflags(unsigned int flags); +struct file *hidden_open(struct dentry *dentry, aufs_bindex_t bindex, + int flags); +int au_do_open(struct inode *inode, struct file *file, + int (*open)(struct file *file, int flags)); +int au_reopen_nondir(struct file *file); +int au_ready_to_write(struct file *file, loff_t len); +int au_reval_and_lock_finfo(struct file *file, int (*reopen)(struct file *file), + int wlock, int locked); + +/* f_op.c */ +extern struct file_operations aufs_file_fop; +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) +int aufs_flush(struct file *file, fl_owner_t id); +#else +int aufs_flush(struct file *file); +#endif + +/* finfo.c */ +struct aufs_finfo *ftofi(struct file *file); +aufs_bindex_t fbstart(struct file *file); +aufs_bindex_t fbend(struct file *file); +struct aufs_vdir *fvdir_cache(struct file *file); +struct aufs_branch *ftobr(struct file *file, aufs_bindex_t bindex); +struct file *au_h_fptr_i(struct file *file, aufs_bindex_t bindex); +struct file *au_h_fptr(struct file *file); + +void set_fbstart(struct file *file, aufs_bindex_t bindex); +void set_fbend(struct file *file, aufs_bindex_t bindex); +void set_fvdir_cache(struct file *file, struct aufs_vdir *vdir_cache); +void au_hfput(struct aufs_hfile *hf); +void set_h_fptr(struct file *file, aufs_bindex_t bindex, struct file *h_file); +void au_update_figen(struct file *file); + +void au_fin_finfo(struct file *file); +int au_init_finfo(struct file *file); + +/* ---------------------------------------------------------------------- */ + +static inline int au_figen(struct file *f) +{ + return atomic_read(&ftofi(f)->fi_generation); +} + +static inline int au_is_mmapped(struct file *f) +{ + return !!(ftofi(f)->fi_h_vm_ops); +} + +/* ---------------------------------------------------------------------- */ + +/* + * fi_read_lock, fi_write_lock, + * fi_read_unlock, fi_write_unlock, fi_downgrade_lock + */ +SimpleRwsemFuncs(fi, struct file *f, ftofi(f)->fi_rwsem); + +/* to debug easier, do not make them inlined functions */ +#define FiMustReadLock(f) do {\ + SiMustAnyLock((f)->f_dentry->d_sb); \ + RwMustReadLock(&ftofi(f)->fi_rwsem); \ +} while (0) + +#define FiMustWriteLock(f) do { \ + SiMustAnyLock((f)->f_dentry->d_sb); \ + RwMustWriteLock(&ftofi(f)->fi_rwsem); \ +} while (0) + +#define FiMustAnyLock(f) do { \ + SiMustAnyLock((f)->f_dentry->d_sb); \ + RwMustAnyLock(&ftofi(f)->fi_rwsem); \ +} while (0) + +#define FiMustNoWaiters(f) RwMustNoWaiters(&ftofi(f)->fi_rwsem) + +#endif /* __KERNEL__ */ +#endif /* __AUFS_FILE_H__ */ diff --git a/fs/aufs/finfo.c b/fs/aufs/finfo.c new file mode 100755 index 0000000..1e09da8 --- /dev/null +++ b/fs/aufs/finfo.c @@ -0,0 +1,211 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: finfo.c,v 1.23 2007/04/30 05:45:21 sfjro Exp $ */ + +#include "aufs.h" + +struct aufs_finfo *ftofi(struct file *file) +{ + struct aufs_finfo *finfo = file->private_data; + DEBUG_ON(!finfo + || !finfo->fi_hfile + || (0 < finfo->fi_bend + && (/* stosi(file->f_dentry->d_sb)->si_bend + < finfo->fi_bend + || */ finfo->fi_bend < finfo->fi_bstart))); + return finfo; +} + +// hard/soft set +aufs_bindex_t fbstart(struct file *file) +{ + FiMustAnyLock(file); + return ftofi(file)->fi_bstart; +} + +aufs_bindex_t fbend(struct file *file) +{ + FiMustAnyLock(file); + return ftofi(file)->fi_bend; +} + +struct aufs_vdir *fvdir_cache(struct file *file) +{ + FiMustAnyLock(file); + return ftofi(file)->fi_vdir_cache; +} + +struct aufs_branch *ftobr(struct file *file, aufs_bindex_t bindex) +{ + struct aufs_finfo *finfo = ftofi(file); + struct aufs_hfile *hf; + + FiMustAnyLock(file); + DEBUG_ON(!finfo + || finfo->fi_bstart < 0 + || bindex < finfo->fi_bstart + || finfo->fi_bend < bindex); + hf = finfo->fi_hfile + bindex; + DEBUG_ON(hf->hf_br && br_count(hf->hf_br) <= 0); + return hf->hf_br; +} + +struct file *au_h_fptr_i(struct file *file, aufs_bindex_t bindex) +{ + struct aufs_finfo *finfo = ftofi(file); + struct aufs_hfile *hf; + + FiMustAnyLock(file); + DEBUG_ON(!finfo + || finfo->fi_bstart < 0 + || bindex < finfo->fi_bstart + || finfo->fi_bend < bindex); + hf = finfo->fi_hfile + bindex; + DEBUG_ON(hf->hf_file + && file_count(hf->hf_file) <= 0 + && br_count(hf->hf_br) <= 0); + return hf->hf_file; +} + +struct file *au_h_fptr(struct file *file) +{ + return au_h_fptr_i(file, fbstart(file)); +} + +void set_fbstart(struct file *file, aufs_bindex_t bindex) +{ + FiMustWriteLock(file); + DEBUG_ON(sbend(file->f_dentry->d_sb) < bindex); + ftofi(file)->fi_bstart = bindex; +} + +void set_fbend(struct file *file, aufs_bindex_t bindex) +{ + FiMustWriteLock(file); + DEBUG_ON(sbend(file->f_dentry->d_sb) < bindex + || bindex < fbstart(file)); + ftofi(file)->fi_bend = bindex; +} + +void set_fvdir_cache(struct file *file, struct aufs_vdir *vdir_cache) +{ + FiMustWriteLock(file); + DEBUG_ON(!S_ISDIR(file->f_dentry->d_inode->i_mode) + || (ftofi(file)->fi_vdir_cache && vdir_cache)); + ftofi(file)->fi_vdir_cache = vdir_cache; +} + +void au_hfput(struct aufs_hfile *hf) +{ + fput(hf->hf_file); + hf->hf_file = NULL; + DEBUG_ON(!hf->hf_br); + br_put(hf->hf_br); + hf->hf_br = NULL; +} + +void set_h_fptr(struct file *file, aufs_bindex_t bindex, struct file *val) +{ + struct aufs_finfo *finfo = ftofi(file); + struct aufs_hfile *hf; + + FiMustWriteLock(file); + DEBUG_ON(!finfo + || finfo->fi_bstart < 0 + || bindex < finfo->fi_bstart + || finfo->fi_bend < bindex); + DEBUG_ON(val && file_count(val) <= 0); + hf = finfo->fi_hfile + bindex; + DEBUG_ON(val && hf->hf_file); + if (hf->hf_file) + au_hfput(hf); + if (val) { + hf->hf_file = val; + hf->hf_br = stobr(file->f_dentry->d_sb, bindex); + } +} + +void au_update_figen(struct file *file) +{ + atomic_set(&ftofi(file)->fi_generation, au_digen(file->f_dentry)); +} + +void au_fin_finfo(struct file *file) +{ + struct aufs_finfo *finfo; + struct dentry *dentry; + aufs_bindex_t bindex, bend; + + dentry = file->f_dentry; + LKTRTrace("%.*s\n", DLNPair(dentry)); + SiMustAnyLock(dentry->d_sb); + + fi_write_lock(file); + bend = fbend(file); + bindex = fbstart(file); + if (bindex >= 0) + for (; bindex <= bend; bindex++) + set_h_fptr(file, bindex, NULL); + + finfo = ftofi(file); +#ifdef CONFIG_AUFS_DEBUG + if (finfo->fi_bstart >= 0) { + bend = fbend(file); + for (bindex = finfo->fi_bstart; bindex <= bend; bindex++) { + struct aufs_hfile *hf; + hf = finfo->fi_hfile + bindex; + DEBUG_ON(hf->hf_file || hf->hf_br); + } + } +#endif + + kfree(finfo->fi_hfile); + fi_write_unlock(file); + cache_free_finfo(finfo); + //file->private_data = NULL; +} + +int au_init_finfo(struct file *file) +{ + struct aufs_finfo *finfo; + struct dentry *dentry; + + dentry = file->f_dentry; + LKTRTrace("%.*s\n", DLNPair(dentry)); + DEBUG_ON(!dentry->d_inode); + + finfo = cache_alloc_finfo(); + if (finfo) { + finfo->fi_hfile = kcalloc(sbend(dentry->d_sb) + 1, + sizeof(*finfo->fi_hfile), GFP_KERNEL); + if (finfo->fi_hfile) { + rw_init_wlock(&finfo->fi_rwsem); + finfo->fi_bstart = -1; + finfo->fi_bend = -1; + atomic_set(&finfo->fi_generation, au_digen(dentry)); + + file->private_data = finfo; + return 0; /* success */ + } + cache_free_finfo(finfo); + } + + TraceErr(-ENOMEM); + return -ENOMEM; +} diff --git a/fs/aufs/hinotify.c b/fs/aufs/hinotify.c new file mode 100755 index 0000000..3bad3f7 --- /dev/null +++ b/fs/aufs/hinotify.c @@ -0,0 +1,536 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: hinotify.c,v 1.19 2007/05/14 03:39:21 sfjro Exp $ */ + +#include "aufs.h" + +static struct inotify_handle *in_handle; +static const __u32 in_mask = (IN_MOVE | IN_DELETE | IN_CREATE /* | IN_ACCESS */ + | IN_MODIFY | IN_ATTRIB + | IN_DELETE_SELF | IN_MOVE_SELF); + +int alloc_hinotify(struct aufs_hinode *hinode, struct inode *inode, + struct inode *hidden_inode) +{ + int err; + struct aufs_hinotify *hin; + s32 wd; + + LKTRTrace("i%lu, hi%lu\n", inode->i_ino, hidden_inode->i_ino); + + err = -ENOMEM; + hin = cache_alloc_hinotify(); + if (hin) { + DEBUG_ON(hinode->hi_notify); + hinode->hi_notify = hin; + hin->hin_aufs_inode = inode; + inotify_init_watch(&hin->hin_watch); + wd = inotify_add_watch(in_handle, &hin->hin_watch, hidden_inode, + in_mask); + if (wd >= 0) + return 0; /* success */ + + err = wd; + put_inotify_watch(&hin->hin_watch); + cache_free_hinotify(hin); + hinode->hi_notify = NULL; + } + + TraceErr(err); + return err; +} + +void do_free_hinotify(struct aufs_hinode *hinode) +{ + int err; + struct aufs_hinotify *hin; + + TraceEnter(); + + hin = hinode->hi_notify; + if (hin) { + err = 0; + if (atomic_read(&hin->hin_watch.count)) + err = inotify_rm_watch(in_handle, &hin->hin_watch); + + if (!err) { + cache_free_hinotify(hin); + hinode->hi_notify = NULL; + } else + IOErr1("failed inotify_rm_watch() %d\n", err); + } +} + +/* ---------------------------------------------------------------------- */ + +static void ctl_hinotify(struct aufs_hinode *hinode, const __u32 mask) +{ + struct inode *hi; + struct inotify_watch *watch; + + hi = hinode->hi_inode; + LKTRTrace("hi%lu, sb %p, 0x%x\n", hi->i_ino, hi->i_sb, mask); + if (0 && !strcmp(current->comm, "link")) + dump_stack(); + IMustLock(hi); + if (!hinode->hi_notify) + return; + + watch = &hinode->hi_notify->hin_watch; +#if 0 + { + u32 wd; + wd = inotify_find_update_watch(in_handle, hi, mask); + TraceErr(wd); + // ignore an err; + } +#else + watch->mask = mask; + smp_mb(); +#endif + LKTRTrace("watch %p, mask %u\n", watch, watch->mask); +} + +#define suspend_hinotify(hi) ctl_hinotify(hi, 0) +#define resume_hinotify(hi) ctl_hinotify(hi, in_mask) + +void do_hdir_lock(struct inode *h_dir, struct inode *dir, aufs_bindex_t bindex, + unsigned int lsc) +{ + struct aufs_hinode *hinode; + + LKTRTrace("i%lu, b%d, lsc %d\n", dir->i_ino, bindex, lsc); + DEBUG_ON(!S_ISDIR(dir->i_mode)); + hinode = itoii(dir)->ii_hinode + bindex; + DEBUG_ON(h_dir != hinode->hi_inode); + + hi_lock(h_dir, lsc); + if (1 /* unlikely(au_flag_test(dir->i_sb, AuFlag_UDBA_HINOTIFY) */) + suspend_hinotify(hinode); +} + +void hdir_unlock(struct inode *h_dir, struct inode *dir, aufs_bindex_t bindex) +{ + struct aufs_hinode *hinode; + + LKTRTrace("i%lu, b%d\n", dir->i_ino, bindex); + DEBUG_ON(!S_ISDIR(dir->i_mode)); + hinode = itoii(dir)->ii_hinode + bindex; + DEBUG_ON(h_dir != hinode->hi_inode); + + if (1 /* unlikely(au_flag_test(dir->i_sb, AuFlag_UDBA_HINOTIFY) */) + resume_hinotify(hinode); + i_unlock(h_dir); +} + +void hdir_lock_rename(struct dentry **h_parents, struct inode **dirs, + aufs_bindex_t bindex, int issamedir) +{ + struct aufs_hinode *hinode; + + LKTRTrace("%.*s, %.*s\n", DLNPair(h_parents[0]), DLNPair(h_parents[1])); + + vfsub_lock_rename(h_parents[0], h_parents[1]); + hinode = itoii(dirs[0])->ii_hinode + bindex; + DEBUG_ON(h_parents[0]->d_inode != hinode->hi_inode); + suspend_hinotify(hinode); + if (issamedir) + return; + hinode = itoii(dirs[1])->ii_hinode + bindex; + DEBUG_ON(h_parents[1]->d_inode != hinode->hi_inode); + suspend_hinotify(hinode); +} + +void hdir_unlock_rename(struct dentry **h_parents, struct inode **dirs, + aufs_bindex_t bindex, int issamedir) +{ + struct aufs_hinode *hinode; + + LKTRTrace("%.*s, %.*s\n", DLNPair(h_parents[0]), DLNPair(h_parents[1])); + + hinode = itoii(dirs[0])->ii_hinode + bindex; + DEBUG_ON(h_parents[0]->d_inode != hinode->hi_inode); + resume_hinotify(hinode); + if (!issamedir) { + hinode = itoii(dirs[1])->ii_hinode + bindex; + DEBUG_ON(h_parents[1]->d_inode != hinode->hi_inode); + resume_hinotify(hinode); + } + vfsub_unlock_rename(h_parents[0], h_parents[1]); +} + +void au_reset_hinotify(struct inode *inode, unsigned int flags) +{ + aufs_bindex_t bindex, bend; + struct inode *hi; + + LKTRTrace("i%lu, 0x%x\n", inode->i_ino, flags); + + bend = ibend(inode); + for (bindex = ibstart(inode); bindex <= bend; bindex++) { + hi = au_h_iptr_i(inode, bindex); + if (hi) { + //hi_lock(hi, AUFS_LSC_H_CHILD); + igrab(hi); + set_h_iptr(inode, bindex, NULL, 0); + set_h_iptr(inode, bindex, igrab(hi), flags); + iput(hi); + //i_unlock(hi); + } + } +} + +/* ---------------------------------------------------------------------- */ + +#ifdef CONFIG_AUFS_DEBUG +static char *in_name(u32 mask) +{ +#define test_ret(flag) if (mask & flag) return #flag; + test_ret(IN_ACCESS); + test_ret(IN_MODIFY); + test_ret(IN_ATTRIB); + test_ret(IN_CLOSE_WRITE); + test_ret(IN_CLOSE_NOWRITE); + test_ret(IN_OPEN); + test_ret(IN_MOVED_FROM); + test_ret(IN_MOVED_TO); + test_ret(IN_CREATE); + test_ret(IN_DELETE); + test_ret(IN_DELETE_SELF); + test_ret(IN_MOVE_SELF); + test_ret(IN_UNMOUNT); + test_ret(IN_Q_OVERFLOW); + test_ret(IN_IGNORED); + return ""; +#undef test_ret +} +#else +#define in_name(m) "??" +#endif + +static int dec_gen_by_name(struct inode *dir, const char *_name, u32 mask) +{ + int err; + struct dentry *parent, *child; + struct inode *inode; + struct qstr *dname; + char *name = (void*)_name; + unsigned int len; + + LKTRTrace("i%lu, %s, 0x%x %s\n", + dir->i_ino, name, mask, in_name(mask)); + + err = -1; + parent = d_find_alias(dir); + if (unlikely(!parent)) + goto out; + +#if 0 + if (unlikely(!memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN))) + name += AUFS_WH_PFX_LEN; +#endif + len = strlen(name); + spin_lock(&dcache_lock); + list_for_each_entry(child, &parent->d_subdirs, d_u.d_child) { + dname = &child->d_name; + if (len == dname->len && !memcmp(dname->name, name, len)) { + au_digen_dec(child); +#if 1 + //todo: why both are needed + if (mask & IN_MOVE) { + spin_lock(&child->d_lock); + __d_drop(child); + spin_unlock(&child->d_lock); + } +#endif + + inode = child->d_inode; + if (inode) + au_iigen_dec(inode); + err = !!inode; + + // todo: the i_nlink of newly created name by link(2) + // should be updated + // todo: some nfs dentry doesn't notified at deleteing + break; + } + } + spin_unlock(&dcache_lock); + dput(parent); + + out: + TraceErr(err); + return err; +} + +struct postproc_args { + struct inode *h_dir, *dir, *h_child_inode; + char *h_child_name; + u32 mask; +}; + +static void dec_gen_by_ino(struct postproc_args *a) +{ + struct super_block *sb; + aufs_bindex_t bindex, bend, bfound; + struct xino xino; + struct inode *cinode; + + TraceEnter(); + + sb = a->dir->i_sb; + DEBUG_ON(!au_flag_test(sb, AuFlag_XINO)); + + bfound = -1; + bend = ibend(a->dir); + for (bindex = ibstart(a->dir); bfound == -1 && bindex <= bend; bindex++) + if (au_h_iptr_i(a->dir, bindex) == a->h_dir) + bfound = bindex; + if (bfound < 0) + return; + + bindex = find_brindex(sb, itoii(a->dir)->ii_hinode[bfound + 0].hi_id); + if (bindex < 0) + return; + if (unlikely(xino_read(sb, bindex, a->h_child_inode->i_ino, &xino))) + return; + cinode = NULL; + if (xino.ino) + cinode = ilookup(sb, xino.ino); + if (cinode) { +#if 1 + if (1 || a->mask & IN_MOVE) { + struct dentry *child; + spin_lock(&dcache_lock); + list_for_each_entry(child, &cinode->i_dentry, d_alias) + au_digen_dec(child); + spin_unlock(&dcache_lock); + } +#endif + au_iigen_dec(cinode); + iput(cinode); + } +} + +static void reset_ino(struct postproc_args *a) +{ + aufs_bindex_t bindex, bend; + struct super_block *sb; + struct inode *h_dir; + + sb = a->dir->i_sb; + bend = ibend(a->dir); + for (bindex = ibstart(a->dir); bindex <= bend; bindex++) { + h_dir = au_h_iptr_i(a->dir, bindex); + if (h_dir && h_dir != a->h_dir) + xino_write0(sb, bindex, h_dir->i_ino); + /* ignore this error */ + } +} + +static void postproc(void *args) +{ + struct postproc_args *a = args; + struct super_block *sb; + struct aufs_vdir *vdir; + + //au_debug_on(); + LKTRTrace("mask 0x%x %s, i%lu, hi%lu, hci%lu\n", + a->mask, in_name(a->mask), a->dir->i_ino, a->h_dir->i_ino, + a->h_child_inode ? a->h_child_inode->i_ino : 0); + DEBUG_ON(!a->dir); +#if 0//def ForceInotify + Dbg("mask 0x%x %s, i%lu, hi%lu, hci%lu\n", + a->mask, in_name(a->mask), a->dir->i_ino, a->h_dir->i_ino, + a->h_child_inode ? a->h_child_inode->i_ino : 0); +#endif + + i_lock(a->dir); + sb = a->dir->i_sb; + si_read_lock(sb); // consider write_lock + ii_write_lock_parent(a->dir); + + /* make dir entries obsolete */ + vdir = ivdir(a->dir); + if (vdir) + vdir->vd_jiffy = 0; + a->dir->i_version++; + + /* + * special handling root directory, + * sine d_revalidate may not be called later. + * main purpose is maintaining i_nlink. + */ + if (unlikely(a->dir->i_ino == AUFS_ROOT_INO)) + au_cpup_attr_all(a->dir); + + if (a->h_child_inode && au_flag_test(sb, AuFlag_XINO)) + dec_gen_by_ino(a); + else if (a->mask & (IN_MOVE_SELF | IN_DELETE_SELF)) + reset_ino(a); + + ii_write_unlock(a->dir); + si_read_unlock(sb); + i_unlock(a->dir); + + au_mntput(a->dir->i_sb); + iput(a->h_child_inode); + iput(a->h_dir); + iput(a->dir); +#if 0 + if (atomic_dec_and_test(&stosi(sb)->si_hinotify)) + wake_up_all(&stosi(sb)->si_hinotify_wq); +#endif + kfree(a); + //au_debug_off(); +} + +static void aufs_inotify(struct inotify_watch *watch, u32 wd, u32 mask, + u32 cookie, const char *h_child_name, + struct inode *h_child_inode) +{ + struct aufs_hinotify *hinotify; + struct postproc_args *args; + int len; + char *p; + struct inode *dir; + //static DECLARE_WAIT_QUEUE_HEAD(wq); + + //au_debug_on(); + LKTRTrace("i%lu, wd %d, mask 0x%x %s, cookie 0x%x, hcname %s, hi%lu\n", + watch->inode->i_ino, wd, mask, in_name(mask), cookie, + h_child_name ? h_child_name : "", + h_child_inode ? h_child_inode->i_ino : 0); + //au_debug_off(); + //IMustLock(h_dir); +#if 0 //defined(ForceInotify) || defined(DbgInotify) + Dbg("i%lu, wd %d, mask 0x%x %s, cookie 0x%x, hcname %s, hi%lu\n", + watch->inode->i_ino, wd, mask, in_name(mask), cookie, + h_child_name ? h_child_name : "", + h_child_inode ? h_child_inode->i_ino : 0); +#endif + /* if IN_UNMOUNT happens, there must be another bug */ + if (mask & (IN_IGNORED | IN_UNMOUNT)) { + put_inotify_watch(watch); + return; + } + + switch (mask & IN_ALL_EVENTS) { + case IN_MODIFY: + case IN_ATTRIB: + if (h_child_name) + return; + break; + + case IN_MOVED_FROM: + case IN_MOVED_TO: + case IN_CREATE: + DEBUG_ON(!h_child_name || !h_child_inode); + break; + case IN_DELETE: + /* + * aufs never be able to get this child inode. + * revalidation should be in d_revalide() + * by checking i_nlink, i_generation or d_unhashed(). + */ + DEBUG_ON(!h_child_name); + break; + + case IN_DELETE_SELF: + case IN_MOVE_SELF: + DEBUG_ON(h_child_name || h_child_inode); + break; + + case IN_ACCESS: + default: + DEBUG_ON(1); + } + +#ifdef DbgInotify + WARN_ON(1); +#endif + + /* iput() will be called in postproc() */ + hinotify = container_of(watch, struct aufs_hinotify, hin_watch); + DEBUG_ON(!hinotify || !hinotify->hin_aufs_inode); + dir = hinotify->hin_aufs_inode; + + /* force re-lookup in next d_revalidate() */ + if (dir->i_ino != AUFS_ROOT_INO) + au_iigen_dec(dir); + len = 0; + if (h_child_name && dec_gen_by_name(dir, h_child_name, mask)) + len = strlen(h_child_name); + + //wait_event(wq, (args = kmalloc(sizeof(*args), GFP_KERNEL))); + args = kmalloc(sizeof(*args) + len + 1, GFP_KERNEL); + if (unlikely(!args)) { + Err("no memory\n"); + return; + } + args->mask = mask; + args->dir = igrab(dir); + args->h_dir = igrab(watch->inode); + args->h_child_inode = NULL; + if (len) { + if (h_child_inode) + args->h_child_inode = igrab(h_child_inode); + p = (void*)args; + args->h_child_name = p + sizeof(*args); + memcpy(args->h_child_name, h_child_name, len + 1); + } + //atomic_inc(&stosi(args->dir->i_sb)->si_hinotify); + /* prohibit umount */ + au_mntget(args->dir->i_sb); + au_wkq_nowait(postproc, args, /*dlgt*/0); +} + +#if 0 +void hinotify_flush(struct super_block *sb) +{ + atomic_t *p = &stosi(sb)->si_hinotify; + wait_event(stosi(sb)->si_hinotify_wq, !atomic_read(p)); +} +#endif + +static void aufs_inotify_destroy(struct inotify_watch *watch) +{ + return; +} + +static struct inotify_operations aufs_inotify_ops = { + .handle_event = aufs_inotify, + .destroy_watch = aufs_inotify_destroy +}; + +/* ---------------------------------------------------------------------- */ + +int __init au_inotify_init(void) +{ + in_handle = inotify_init(&aufs_inotify_ops); + if (!IS_ERR(in_handle)) + return 0; + TraceErrPtr(in_handle); + return PTR_ERR(in_handle); +} + +void au_inotify_fin(void) +{ + inotify_destroy(in_handle); +} diff --git a/fs/aufs/i_op.c b/fs/aufs/i_op.c new file mode 100755 index 0000000..1cd0453 --- /dev/null +++ b/fs/aufs/i_op.c @@ -0,0 +1,641 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: i_op.c,v 1.30 2007/04/23 00:55:05 sfjro Exp $ */ + +//#include +//#include +#include +#include +#include "aufs.h" + +#ifdef CONFIG_AUFS_DLGT +struct security_inode_permission_args { + int *errp; + struct inode *h_inode; + int mask; + struct nameidata *fake_nd; +}; + +static void call_security_inode_permission(void *args) +{ + struct security_inode_permission_args *a = args; + LKTRTrace("fsuid %d\n", current->fsuid); + *a->errp = security_inode_permission(a->h_inode, a->mask, a->fake_nd); +} +#endif + +static int hidden_permission(struct inode *hidden_inode, int mask, + struct nameidata *fake_nd, int brperm, int dlgt) +{ + int err, submask; + const int write_mask = (mask & (MAY_WRITE | MAY_APPEND)); + + LKTRTrace("ino %lu, mask 0x%x, brperm 0x%x\n", + hidden_inode->i_ino, mask, brperm); + + err = -EACCES; + if (unlikely(write_mask && IS_IMMUTABLE(hidden_inode))) + goto out; + + /* skip hidden fs test in the case of write to ro branch */ + submask = mask & ~MAY_APPEND; + if (unlikely((write_mask && !br_writable(brperm)) + || !hidden_inode->i_op + || !hidden_inode->i_op->permission)) { + //LKTRLabel(generic_permission); + err = generic_permission(hidden_inode, submask, NULL); + } else { + //LKTRLabel(h_inode->permission); + err = hidden_inode->i_op->permission(hidden_inode, submask, + fake_nd); + TraceErr(err); + } + +#if 1 + if (!err) { +#ifndef CONFIG_AUFS_DLGT + err = security_inode_permission(hidden_inode, mask, fake_nd); +#else + if (!dlgt) + err = security_inode_permission(hidden_inode, mask, + fake_nd); + else { + struct security_inode_permission_args args = { + .errp = &err, + .h_inode = hidden_inode, + .mask = mask, + .fake_nd = fake_nd + }; + au_wkq_wait(call_security_inode_permission, &args, + /*dlgt*/1); + } +#endif + } +#endif + + out: + TraceErr(err); + return err; +} + +static int silly_lock(struct inode *inode, struct nameidata *nd) +{ + int locked = 0; + struct super_block *sb = inode->i_sb; + + LKTRTrace("i%lu, nd %p\n", inode->i_ino, nd); + +#ifdef CONFIG_AUFS_FAKE_DM + si_read_lock(sb); + ii_read_lock_child(inode); +#else + if (!nd || !nd->dentry) { + si_read_lock(sb); + ii_read_lock_child(inode); + } else if (nd->dentry->d_inode != inode) { + locked = 1; + /* lock child first, then parent */ + si_read_lock(sb); + ii_read_lock_child(inode); + di_read_lock_parent(nd->dentry, 0); + } else { + locked = 2; + aufs_read_lock(nd->dentry, AUFS_I_RLOCK); + } +#endif + return locked; +} + +static void silly_unlock(int locked, struct inode *inode, struct nameidata *nd) +{ + struct super_block *sb = inode->i_sb; + + LKTRTrace("locked %d, i%lu, nd %p\n", locked, inode->i_ino, nd); + +#ifdef CONFIG_AUFS_FAKE_DM + ii_read_unlock(inode); + si_read_unlock(sb); +#else + switch (locked) { + case 0: + ii_read_unlock(inode); + si_read_unlock(sb); + break; + case 1: + di_read_unlock(nd->dentry, 0); + ii_read_unlock(inode); + si_read_unlock(sb); + break; + case 2: + aufs_read_unlock(nd->dentry, AUFS_I_RLOCK); + break; + default: + BUG(); + } +#endif +} + +static int aufs_permission(struct inode *inode, int mask, struct nameidata *nd) +{ + int err, locked, dlgt; + aufs_bindex_t bindex, bend; + struct inode *hidden_inode; + struct super_block *sb; + struct nameidata fake_nd, *p; + const int write_mask = (mask & (MAY_WRITE | MAY_APPEND)); + const int nondir = !S_ISDIR(inode->i_mode); + + LKTRTrace("ino %lu, mask 0x%x, nondir %d, write_mask %d, " + "nd %p{%p, %p}\n", + inode->i_ino, mask, nondir, write_mask, + nd, nd ? nd->dentry : NULL, nd ? nd->mnt : NULL); + + sb = inode->i_sb; + locked = silly_lock(inode, nd); + dlgt = need_dlgt(sb); + + if (nd) + fake_nd = *nd; + if (/* unlikely */(nondir || write_mask)) { + hidden_inode = au_h_iptr(inode); + DEBUG_ON(!hidden_inode + || ((hidden_inode->i_mode & S_IFMT) + != (inode->i_mode & S_IFMT))); + err = 0; + bindex = ibstart(inode); + p = fake_dm(&fake_nd, nd, sb, bindex); + /* actual test will be delegated to LSM */ + if (IS_ERR(p)) + DEBUG_ON(PTR_ERR(p) != -ENOENT); + else { + err = hidden_permission(hidden_inode, mask, p, + sbr_perm(sb, bindex), dlgt); + fake_dm_release(p); + } + if (write_mask && !err) { + err = find_rw_br(sb, bindex); + if (err >= 0) + err = 0; + } + goto out; + } + + /* non-write to dir */ + err = 0; + bend = ibend(inode); + for (bindex = ibstart(inode); !err && bindex <= bend; bindex++) { + hidden_inode = au_h_iptr_i(inode, bindex); + if (!hidden_inode) + continue; + DEBUG_ON(!S_ISDIR(hidden_inode->i_mode)); + + p = fake_dm(&fake_nd, nd, sb, bindex); + /* actual test will be delegated to LSM */ + if (IS_ERR(p)) + DEBUG_ON(PTR_ERR(p) != -ENOENT); + else { + err = hidden_permission(hidden_inode, mask, p, + sbr_perm(sb, bindex), dlgt); + fake_dm_release(p); + } + } + + out: + silly_unlock(locked, inode, nd); + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +static struct dentry *aufs_lookup(struct inode *dir, struct dentry *dentry, + struct nameidata *nd) +{ + struct dentry *ret, *parent; + int err, npositive; + struct inode *inode; + + LKTRTrace("dir %lu, %.*s\n", dir->i_ino, DLNPair(dentry)); + DEBUG_ON(IS_ROOT(dentry)); + IMustLock(dir); + + parent = dentry->d_parent; // dget_parent() + aufs_read_lock(parent, !AUFS_I_RLOCK); + err = au_alloc_dinfo(dentry); + //if (LktrCond) err = -1; + ret = ERR_PTR(err); + if (unlikely(err)) + goto out; + + err = npositive = lkup_dentry(dentry, dbstart(parent), /*type*/0); + //err = -1; + ret = ERR_PTR(err); + if (unlikely(err < 0)) + goto out_unlock; + inode = NULL; + if (npositive) { + inode = au_new_inode(dentry); + ret = (void*)inode; + } + if (!IS_ERR(inode)) { +#if 1 + /* d_splice_alias() also supports d_add() */ + ret = d_splice_alias(inode, dentry); + if (unlikely(IS_ERR(ret) && inode)) + ii_write_unlock(inode); +#else + d_add(dentry, inode); +#endif + } + + out_unlock: + di_write_unlock(dentry); + out: + aufs_read_unlock(parent, !AUFS_I_RLOCK); + TraceErrPtr(ret); + return ret; +} + +/* ---------------------------------------------------------------------- */ + +/* + * decide the branch and the parent dir where we will create a new entry. + * returns new bindex or an error. + * copyup the parent dir if needed. + */ +int wr_dir(struct dentry *dentry, int add_entry, struct dentry *src_dentry, + aufs_bindex_t force_btgt, int do_lock_srcdir) +{ + int err; + aufs_bindex_t bcpup, bstart, src_bstart; + struct dentry *hidden_parent; + struct super_block *sb; + struct dentry *parent, *src_parent = NULL; + struct inode *dir, *src_dir = NULL; + + LKTRTrace("%.*s, add %d, src %p, force %d, lock_srcdir %d\n", + DLNPair(dentry), add_entry, src_dentry, force_btgt, + do_lock_srcdir); + + sb = dentry->d_sb; + parent = dentry->d_parent; // dget_parent() + bcpup = bstart = dbstart(dentry); + if (force_btgt < 0) { + if (src_dentry) { + src_bstart = dbstart(src_dentry); + if (src_bstart < bstart) + bcpup = src_bstart; + } + if (test_ro(sb, bcpup, dentry->d_inode)) { + if (!add_entry) + di_read_lock_parent(parent, !AUFS_I_RLOCK); + bcpup = err = find_rw_parent_br(dentry, bcpup); + //bcpup = err = find_rw_br(sb, bcpup); + if (!add_entry) + di_read_unlock(parent, !AUFS_I_RLOCK); + //err = -1; + if (unlikely(err < 0)) + goto out; + } + } else { + DEBUG_ON(bstart <= force_btgt + || test_ro(sb, force_btgt, dentry->d_inode)); + bcpup = force_btgt; + } + LKTRTrace("bstart %d, bcpup %d\n", bstart, bcpup); + + err = bcpup; + if (bcpup == bstart) + goto out; /* success */ + + /* copyup the new parent into the branch we process */ + hidden_parent = au_h_dptr(dentry)->d_parent; // dget_parent() + if (src_dentry) { + src_parent = src_dentry->d_parent; // dget_parent() + src_dir = src_parent->d_inode; + if (do_lock_srcdir) + di_write_lock_parent2(src_parent); + } + + dir = parent->d_inode; + if (add_entry) { + au_update_dbstart(dentry); + IMustLock(dir); + DiMustWriteLock(parent); + IiMustWriteLock(dir); + } else + di_write_lock_parent(parent); + + err = 0; + if (!au_h_dptr_i(parent, bcpup)) + err = cpup_dirs(dentry, bcpup, src_parent); + //err = -1; + if (!err && add_entry) { + hidden_parent = au_h_dptr_i(parent, bcpup); + DEBUG_ON(!hidden_parent || !hidden_parent->d_inode); + hi_lock_parent(hidden_parent->d_inode); + err = lkup_neg(dentry, bcpup); + //err = -1; + i_unlock(hidden_parent->d_inode); + } + + if (!add_entry) + di_write_unlock(parent); + if (do_lock_srcdir) + di_write_unlock(src_parent); + if (!err) + err = bcpup; /* success */ + //err = -EPERM; + out: + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +static int aufs_setattr(struct dentry *dentry, struct iattr *ia) +{ + int err, isdir; + aufs_bindex_t bstart, bcpup; + struct inode *hidden_inode, *inode, *dir, *h_dir, *gh_dir, *gdir; + struct dentry *hidden_dentry, *parent; + unsigned int udba; + + LKTRTrace("%.*s, ia_valid 0x%x\n", DLNPair(dentry), ia->ia_valid); + inode = dentry->d_inode; + IMustLock(inode); + + aufs_read_lock(dentry, AUFS_D_WLOCK); + bstart = dbstart(dentry); + bcpup = err = wr_dir(dentry, /*add*/0, /*src_dentry*/NULL, + /*force_btgt*/-1, /*do_lock_srcdir*/0); + //err = -1; + if (unlikely(err < 0)) + goto out; + + /* crazy udba locks */ + udba = au_flag_test(dentry->d_sb, AuFlag_UDBA_INOTIFY); + parent = NULL; + gdir = gh_dir = dir = h_dir = NULL; + if ((udba || bstart != bcpup) && !IS_ROOT(dentry)) { + parent = dentry->d_parent; // dget_parent() + dir = parent->d_inode; + di_read_lock_parent(parent, AUFS_I_RLOCK); + h_dir = au_h_iptr_i(dir, bcpup); + } + if (parent) { + if (unlikely(udba && !IS_ROOT(parent))) { + gdir = parent->d_parent->d_inode; // dget_parent() + ii_read_lock_parent2(gdir); + gh_dir = au_h_iptr_i(gdir, bcpup); + hgdir_lock(gh_dir, gdir, bcpup); + } + hdir_lock(h_dir, dir, bcpup); + } + + isdir = S_ISDIR(inode->i_mode); + hidden_dentry = au_h_dptr(dentry); + hidden_inode = hidden_dentry->d_inode; + DEBUG_ON(!hidden_inode); + +#define HiLock(bindex) do {\ + if (!isdir) \ + hi_lock_child(hidden_inode); \ + else \ + hdir2_lock(hidden_inode, inode, bindex); \ + } while (0) +#define HiUnlock(bindex) do {\ + if (!isdir) \ + i_unlock(hidden_inode); \ + else \ + hdir_unlock(hidden_inode, inode, bindex); \ + } while (0) + + if (bstart != bcpup) { + loff_t size = -1; + + if ((ia->ia_valid & ATTR_SIZE) + && ia->ia_size < i_size_read(inode)) { + size = ia->ia_size; + ia->ia_valid &= ~ATTR_SIZE; + } + HiLock(bstart); + err = sio_cpup_simple(dentry, bcpup, size, + au_flags_cpup(CPUP_DTIME, parent)); + //err = -1; + HiUnlock(bstart); + if (unlikely(err || !ia->ia_valid)) + goto out_unlock; + + hidden_dentry = au_h_dptr(dentry); + hidden_inode = hidden_dentry->d_inode; + DEBUG_ON(!hidden_inode); + } + + HiLock(bcpup); + err = vfsub_notify_change(hidden_dentry, ia, need_dlgt(dentry->d_sb)); + //err = -1; + if (!err) + au_cpup_attr_changable(inode); + HiUnlock(bcpup); +#undef HiLock +#undef HiUnlock + + out_unlock: + if (parent) { + hdir_unlock(h_dir, dir, bcpup); + di_read_unlock(parent, AUFS_I_RLOCK); + } + if (unlikely(gdir)) { + hdir_unlock(gh_dir, gdir, bcpup); + ii_read_unlock(gdir); + } + out: + aufs_read_unlock(dentry, AUFS_D_WLOCK); + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +static int hidden_readlink(struct dentry *dentry, int bindex, + char __user * buf, int bufsiz) +{ + struct super_block *sb; + struct dentry *hidden_dentry; + + hidden_dentry = au_h_dptr_i(dentry, bindex); + if (unlikely(!hidden_dentry->d_inode->i_op + || !hidden_dentry->d_inode->i_op->readlink)) + return -EINVAL; + + sb = dentry->d_sb; + if (!test_ro(sb, bindex, dentry->d_inode)) { + touch_atime(sbr_mnt(sb, bindex), hidden_dentry); + dentry->d_inode->i_atime = hidden_dentry->d_inode->i_atime; + } + return hidden_dentry->d_inode->i_op->readlink + (hidden_dentry, buf, bufsiz); +} + +static int aufs_readlink(struct dentry *dentry, char __user * buf, int bufsiz) +{ + int err; + + LKTRTrace("%.*s, %d\n", DLNPair(dentry), bufsiz); + + aufs_read_lock(dentry, AUFS_I_RLOCK); + err = hidden_readlink(dentry, dbstart(dentry), buf, bufsiz); + //err = -1; + aufs_read_unlock(dentry, AUFS_I_RLOCK); + TraceErr(err); + return err; +} + +static void *aufs_follow_link(struct dentry *dentry, struct nameidata *nd) +{ + int err; + char *buf; + mm_segment_t old_fs; + + LKTRTrace("%.*s, nd %.*s\n", DLNPair(dentry), DLNPair(nd->dentry)); + + err = -ENOMEM; + buf = __getname(); + //buf = NULL; + if (unlikely(!buf)) + goto out; + + aufs_read_lock(dentry, AUFS_I_RLOCK); + old_fs = get_fs(); + set_fs(KERNEL_DS); + err = hidden_readlink(dentry, dbstart(dentry), (char __user *)buf, + PATH_MAX); + //err = -1; + set_fs(old_fs); + aufs_read_unlock(dentry, AUFS_I_RLOCK); + + if (err >= 0) { + buf[err] = 0; + /* will be freed by put_link */ + nd_set_link(nd, buf); + return NULL; /* success */ + } + __putname(buf); + + out: + path_release(nd); + TraceErr(err); + return ERR_PTR(err); +} + +static void aufs_put_link(struct dentry *dentry, struct nameidata *nd, + void *cookie) +{ + LKTRTrace("%.*s\n", DLNPair(dentry)); + __putname(nd_get_link(nd)); +} + +/* ---------------------------------------------------------------------- */ +#if 0 // comment +struct inode_operations { + int (*create) (struct inode *,struct dentry *,int, struct nameidata *); + struct dentry * (*lookup) (struct inode *,struct dentry *, struct nameidata *); + int (*link) (struct dentry *,struct inode *,struct dentry *); + int (*unlink) (struct inode *,struct dentry *); + int (*symlink) (struct inode *,struct dentry *,const char *); + int (*mkdir) (struct inode *,struct dentry *,int); + int (*rmdir) (struct inode *,struct dentry *); + int (*mknod) (struct inode *,struct dentry *,int,dev_t); + int (*rename) (struct inode *, struct dentry *, + struct inode *, struct dentry *); + int (*readlink) (struct dentry *, char __user *,int); + void * (*follow_link) (struct dentry *, struct nameidata *); + void (*put_link) (struct dentry *, struct nameidata *, void *); + void (*truncate) (struct inode *); + int (*permission) (struct inode *, int, struct nameidata *); + int (*setattr) (struct dentry *, struct iattr *); + int (*getattr) (struct vfsmount *mnt, struct dentry *, struct kstat *); + int (*setxattr) (struct dentry *, const char *,const void *,size_t,int); + ssize_t (*getxattr) (struct dentry *, const char *, void *, size_t); + ssize_t (*listxattr) (struct dentry *, char *, size_t); + int (*removexattr) (struct dentry *, const char *); + void (*truncate_range)(struct inode *, loff_t, loff_t); +}; +#endif + +struct inode_operations aufs_symlink_iop = { + .permission = aufs_permission, + .setattr = aufs_setattr, + + .readlink = aufs_readlink, + .follow_link = aufs_follow_link, + .put_link = aufs_put_link +}; + +//i_op_add.c +int aufs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev); +int aufs_symlink(struct inode *dir, struct dentry *dentry, const char *symname); +int aufs_create(struct inode *dir, struct dentry *dentry, int mode, + struct nameidata *nd); +int aufs_link(struct dentry *src_dentry, struct inode *dir, + struct dentry *dentry); +int aufs_mkdir(struct inode *dir, struct dentry *dentry, int mode); + +//i_op_del.c +int aufs_unlink(struct inode *dir, struct dentry *dentry); +int aufs_rmdir(struct inode *dir, struct dentry *dentry); + +// i_op_ren.c +int aufs_rename(struct inode *src_dir, struct dentry *src_dentry, + struct inode *dir, struct dentry *dentry); + +struct inode_operations aufs_dir_iop = { + .create = aufs_create, + .lookup = aufs_lookup, + .link = aufs_link, + .unlink = aufs_unlink, + .symlink = aufs_symlink, + .mkdir = aufs_mkdir, + .rmdir = aufs_rmdir, + .mknod = aufs_mknod, + .rename = aufs_rename, + + .permission = aufs_permission, + .setattr = aufs_setattr, + +#if 0 // xattr + .setxattr = aufs_setxattr, + .getxattr = aufs_getxattr, + .listxattr = aufs_listxattr, + .removexattr = aufs_removexattr +#endif +}; + +struct inode_operations aufs_iop = { + .permission = aufs_permission, + .setattr = aufs_setattr, + +#if 0 // xattr + .setxattr = aufs_setxattr, + .getxattr = aufs_getxattr, + .listxattr = aufs_listxattr, + .removexattr = aufs_removexattr +#endif +}; diff --git a/fs/aufs/i_op_add.c b/fs/aufs/i_op_add.c new file mode 100755 index 0000000..977d773 --- /dev/null +++ b/fs/aufs/i_op_add.c @@ -0,0 +1,621 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: i_op_add.c,v 1.37 2007/05/07 03:46:08 sfjro Exp $ */ + +//#include +//#include +#include "aufs.h" + +/* + * final procedure of adding a new entry, except link(2). + * remove whiteout, instantiate, copyup the parent dir's times and size + * and update version. + * if it failed, re-create the removed whiteout. + */ +static int epilog(struct dentry *wh_dentry, struct dentry *dentry) +{ + int err, rerr; + aufs_bindex_t bwh; + struct inode *inode, *dir; + struct dentry *wh; + struct lkup_args lkup; + + LKTRTrace("wh %p, %.*s\n", wh_dentry, DLNPair(dentry)); + + lkup.dlgt = need_dlgt(dentry->d_sb); + bwh = -1; + if (wh_dentry) { + bwh = dbwh(dentry); + err = au_unlink_wh_dentry(wh_dentry->d_parent->d_inode, + wh_dentry, dentry, lkup.dlgt); + //err = -1; + if (unlikely(err)) + goto out; + } + + inode = au_new_inode(dentry); + //inode = ERR_PTR(-1); + if (!IS_ERR(inode)) { + d_instantiate(dentry, inode); + dir = dentry->d_parent->d_inode; + /* or always cpup dir mtime? */ + if (ibstart(dir) == dbstart(dentry)) + au_cpup_attr_timesizes(dir); + dir->i_version++; + return 0; /* success */ + } + + err = PTR_ERR(inode); + if (!wh_dentry) + goto out; + + /* revert */ + lkup.nfsmnt = au_nfsmnt(dentry->d_sb, bwh); + wh = simple_create_wh(dentry, bwh, wh_dentry->d_parent, &lkup); + //wh = ERR_PTR(-1); + rerr = PTR_ERR(wh); + if (!IS_ERR(wh)) { + dput(wh); + goto out; + } + IOErr("%.*s reverting whiteout failed(%d, %d)\n", + DLNPair(dentry), err, rerr); + err = -EIO; + + out: + TraceErr(err); + return err; +} + +/* + * initial procedure of adding a new entry. + * prepare writable branch and the parent dir, lock it, + * lookup whiteout for the new entry. + */ +static struct dentry * +lock_hdir_lkup_wh(struct dentry *dentry, struct dtime *dt, + struct dentry *src_dentry, int do_lock_srcdir) +{ + struct dentry *wh_dentry, *parent, *hidden_parent; + int err; + aufs_bindex_t bstart, bcpup; + struct inode *dir, *h_dir; + struct lkup_args lkup; + + LKTRTrace("%.*s, src %p\n", DLNPair(dentry), src_dentry); + + parent = dentry->d_parent; + bstart = dbstart(dentry); + bcpup = err = wr_dir(dentry, 1, src_dentry, -1, do_lock_srcdir); + //err = -1; + wh_dentry = ERR_PTR(err); + if (unlikely(err < 0)) + goto out; + + dir = parent->d_inode; + hidden_parent = au_h_dptr_i(parent, bcpup); + h_dir = hidden_parent->d_inode; + hdir_lock(h_dir, dir, bcpup); + if (dt) + dtime_store(dt, parent, hidden_parent); + if (/* bcpup != bstart || */ bcpup != dbwh(dentry)) + return NULL; /* success */ + + lkup.nfsmnt = au_nfsmnt(parent->d_sb, bcpup); + lkup.dlgt = need_dlgt(parent->d_sb); + wh_dentry = lkup_wh(hidden_parent, &dentry->d_name, &lkup); + //wh_dentry = ERR_PTR(-1); + if (IS_ERR(wh_dentry)) + hdir_unlock(h_dir, dir, bcpup); + + out: + TraceErrPtr(wh_dentry); + return wh_dentry; +} + +/* ---------------------------------------------------------------------- */ + +enum {Mknod, Symlink, Creat}; +struct simple_arg { + int type; + union { + struct { + int mode; + struct nameidata *nd; + } c; + struct { + const char *symname; + } s; + struct { + int mode; + dev_t dev; + } m; + } u; +}; + +static int add_simple(struct inode *dir, struct dentry *dentry, + struct simple_arg *arg) +{ + int err, dlgt; + struct dentry *hidden_dentry, *hidden_parent, *wh_dentry, *parent; + struct inode *hidden_dir; + struct dtime dt; + + LKTRTrace("type %d, %.*s\n", arg->type, DLNPair(dentry)); + IMustLock(dir); + + aufs_read_lock(dentry, AUFS_D_WLOCK); + parent = dentry->d_parent; + di_write_lock_parent(parent); + wh_dentry = lock_hdir_lkup_wh(dentry, &dt, /*src_dentry*/NULL, + /*do_lock_srcdir*/0); + //wh_dentry = ERR_PTR(-1); + err = PTR_ERR(wh_dentry); + if (IS_ERR(wh_dentry)) + goto out; + + hidden_dentry = au_h_dptr(dentry); + hidden_parent = hidden_dentry->d_parent; + hidden_dir = hidden_parent->d_inode; + IMustLock(hidden_dir); + dlgt = need_dlgt(dir->i_sb); + +#if 1 // partial testing + switch (arg->type) { + case Creat: +#if 0 + if (arg->u.c.nd) { + struct nameidata fake_nd; + fake_nd = *arg->u.c.nd; + fake_nd.dentry = dget(hidden_parent); + fake_nd.mnt = sbr_mnt(dentry->d_sb, dbstart(dentry)); + mntget(fake_nd.mnt); + err = vfsub_create(hidden_dir, hidden_dentry, + arg->u.c.mode, &fake_nd, dlgt); + path_release(&fake_nd); + } else +#endif + err = vfsub_create(hidden_dir, hidden_dentry, + arg->u.c.mode, NULL, dlgt); + break; + case Symlink: + err = vfsub_symlink(hidden_dir, hidden_dentry, + arg->u.s.symname, S_IALLUGO, dlgt); + break; + case Mknod: + err = vfsub_mknod(hidden_dir, hidden_dentry, + arg->u.m.mode, arg->u.m.dev, dlgt); + break; + default: + BUG(); + } +#else + err = -1; +#endif + if (!err) + err = epilog(wh_dentry, dentry); + //err = -1; + + /* revert */ + if (unlikely(err && hidden_dentry->d_inode)) { + int rerr; + rerr = vfsub_unlink(hidden_dir, hidden_dentry, dlgt); + //rerr = -1; + if (rerr) { + IOErr("%.*s revert failure(%d, %d)\n", + DLNPair(dentry), err, rerr); + err = -EIO; + } + dtime_revert(&dt, !CPUP_LOCKED_GHDIR); + d_drop(dentry); + } + + hdir_unlock(hidden_dir, dir, dbstart(dentry)); + dput(wh_dentry); + + out: + if (unlikely(err)) { + au_update_dbstart(dentry); + d_drop(dentry); + } + di_write_unlock(parent); + aufs_read_unlock(dentry, AUFS_D_WLOCK); + TraceErr(err); + return err; +} + +int aufs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) +{ + struct simple_arg arg = { + .type = Mknod, + .u.m = {.mode = mode, .dev = dev} + }; + return add_simple(dir, dentry, &arg); +} + +int aufs_symlink(struct inode *dir, struct dentry *dentry, const char *symname) +{ + struct simple_arg arg = { + .type = Symlink, + .u.s.symname = symname + }; + return add_simple(dir, dentry, &arg); +} + +int aufs_create(struct inode *dir, struct dentry *dentry, int mode, + struct nameidata *nd) +{ + struct simple_arg arg = { + .type = Creat, + .u.c = {.mode = mode, .nd = nd} + }; + return add_simple(dir, dentry, &arg); +} + +/* ---------------------------------------------------------------------- */ + +struct link_arg { + aufs_bindex_t bdst, bsrc; + int issamedir, dlgt; + struct dentry *src_parent, *parent, *hidden_dentry; + struct inode *hidden_dir, *inode; +}; + +static int cpup_before_link(struct dentry *src_dentry, struct inode *dir, + struct link_arg *a) +{ + int err; + unsigned int flags; + struct inode *hi, *hdir = NULL, *src_dir; + + TraceEnter(); + + err = 0; + flags = au_flags_cpup(CPUP_DTIME, a->parent); + src_dir = a->src_parent->d_inode; + if (!a->issamedir) { + // todo: dead lock? + di_read_lock_parent2(a->src_parent, AUFS_I_RLOCK); + // this temporary unlock/lock is safe + hdir_unlock(a->hidden_dir, dir, a->bdst); + err = test_and_cpup_dirs(src_dentry, a->bdst, a->parent); + //err = -1; + if (!err) { + hdir = au_h_iptr_i(src_dir, a->bdst); + hdir_lock(hdir, src_dir, a->bdst); + flags = au_flags_cpup(CPUP_DTIME, a->src_parent); + } + } + + if (!err) { + hi = au_h_dptr(src_dentry)->d_inode; + hi_lock_child(hi); + err = sio_cpup_simple(src_dentry, a->bdst, -1, flags); + //err = -1; + i_unlock(hi); + } + + if (!a->issamedir) { + if (hdir) + hdir_unlock(hdir, src_dir, a->bdst); + hdir_lock(a->hidden_dir, dir, a->bdst); + di_read_unlock(a->src_parent, AUFS_I_RLOCK); + } + + TraceErr(err); + return err; +} + +static int cpup_or_link(struct dentry *src_dentry, struct link_arg *a) +{ + int err; + struct inode *inode, *h_inode, *h_dst_inode; + struct dentry *h_dentry; + aufs_bindex_t bstart; + struct super_block *sb; + + TraceEnter(); + + sb = src_dentry->d_sb; + inode = src_dentry->d_inode; + h_dentry = au_h_dptr(src_dentry); + h_inode = h_dentry->d_inode; + bstart = ibstart(inode); + h_dst_inode = NULL; + if (bstart <= a->bdst) + h_dst_inode = au_h_iptr_i(inode, a->bdst); + + if (!h_dst_inode) { + /* copyup src_dentry as the name of dentry. */ + set_dbstart(src_dentry, a->bdst); + set_h_dptr(src_dentry, a->bdst, dget(a->hidden_dentry)); + hi_lock_child(h_inode); + err = sio_cpup_single(src_dentry, a->bdst, a->bsrc, -1, + au_flags_cpup(!CPUP_DTIME, a->parent)); + //err = -1; + i_unlock(h_inode); + set_h_dptr(src_dentry, a->bdst, NULL); + set_dbstart(src_dentry, a->bsrc); + } else { + /* the inode of src_dentry already exists on a.bdst branch */ + h_dentry = d_find_alias(h_dst_inode); + if (h_dentry) { + err = vfsub_link(h_dentry, a->hidden_dir, + a->hidden_dentry, a->dlgt); + dput(h_dentry); + } else { + IOErr("no dentry found for i%lu on b%d\n", + h_dst_inode->i_ino, a->bdst); + err = -EIO; + } + } + + if (!err) + append_plink(sb, a->inode, a->hidden_dentry, a->bdst); + + TraceErr(err); + return err; +} + +int aufs_link(struct dentry *src_dentry, struct inode *dir, + struct dentry *dentry) +{ + int err, rerr; + struct dentry *hidden_parent, *wh_dentry, *hidden_src_dentry; + struct dtime dt; + struct link_arg a; + struct super_block *sb; + + LKTRTrace("src %.*s, i%lu, dst %.*s\n", + DLNPair(src_dentry), dir->i_ino, DLNPair(dentry)); + IMustLock(dir); + IMustLock(src_dentry->d_inode); + + aufs_read_and_write_lock2(dentry, src_dentry, /*isdir*/0); + a.src_parent = src_dentry->d_parent; + a.parent = dentry->d_parent; + a.issamedir = (a.src_parent == a.parent); + di_write_lock_parent(a.parent); + wh_dentry = lock_hdir_lkup_wh(dentry, &dt, src_dentry, !a.issamedir); + //wh_dentry = ERR_PTR(-1); + err = PTR_ERR(wh_dentry); + if (IS_ERR(wh_dentry)) + goto out; + + a.inode = src_dentry->d_inode; + a.hidden_dentry = au_h_dptr(dentry); + hidden_parent = a.hidden_dentry->d_parent; + a.hidden_dir = hidden_parent->d_inode; + IMustLock(a.hidden_dir); + + err = 0; + sb = dentry->d_sb; + a.dlgt = need_dlgt(sb); + + //todo: minor optimize, their sb may be same while their bindex differs. + a.bsrc = dbstart(src_dentry); + a.bdst = dbstart(dentry); + hidden_src_dentry = au_h_dptr(src_dentry); + if (unlikely(!au_flag_test(sb, AuFlag_PLINK))) { + /* + * copyup src_dentry to the branch we process, + * and then link(2) to it. + * gave up 'pseudo link by cpup' approach, + * since nlink may be one and some applications will not work. + */ + if (a.bdst < a.bsrc + /* && hidden_src_dentry->d_sb != a.hidden_dentry->d_sb */) + err = cpup_before_link(src_dentry, dir, &a); + if (!err) { + hidden_src_dentry = au_h_dptr(src_dentry); + err = vfsub_link(hidden_src_dentry, a.hidden_dir, + a.hidden_dentry, a.dlgt); + //err = -1; + } + } else { + if (a.bdst < a.bsrc + /* && hidden_src_dentry->d_sb != a.hidden_dentry->d_sb */) + err = cpup_or_link(src_dentry, &a); + else { + hidden_src_dentry = au_h_dptr(src_dentry); + err = vfsub_link(hidden_src_dentry, a.hidden_dir, + a.hidden_dentry, a.dlgt); + //err = -1; + } + } + if (unlikely(err)) + goto out_unlock; + if (wh_dentry) { + err = au_unlink_wh_dentry(a.hidden_dir, wh_dentry, dentry, + a.dlgt); + //err = -1; + if (unlikely(err)) + goto out_revert; + } + + dir->i_version++; + if (ibstart(dir) == dbstart(dentry)) + au_cpup_attr_timesizes(dir); + if (!d_unhashed(a.hidden_dentry) + /* || hidden_old_inode->i_nlink <= nlink */ + /* || SB_NFS(hidden_src_dentry->d_sb) */) { + dentry->d_inode = igrab(a.inode); + d_instantiate(dentry, a.inode); + a.inode->i_nlink++; + a.inode->i_ctime = dir->i_ctime; + } else + /* nfs case (< 2.6.15) */ + d_drop(dentry); +#if 0 + au_debug_on(); + DbgInode(a.inode); + au_debug_off(); + { + aufs_bindex_t i; + for (i = ibstart(a.inode); i <= ibend(a.inode); i++) { + struct xino xino; + struct inode *hi; + hi = au_h_iptr_i(a.inode, i); + if (hi) { + xino_read(sb, i, hi->i_ino, &xino); + Dbg("hi%lu, i%lu\n", hi->i_ino, xino.ino); + } + } + } +#endif + goto out_unlock; /* success */ + + out_revert: +#if 0 // remove + if (d_unhashed(a.hidden_dentry)) { + /* hardlink on nfs (< 2.6.15) */ + struct dentry *d; + const struct qstr *name = &a.hidden_dentry->d_name; + DEBUG_ON(a.hidden_dentry->d_parent->d_inode != a.hidden_dir); + // do not superio. + d = lkup_one(name->name, a.hidden_dentry->d_parent, name->len, + au_nfsmnt(sb, a.bdst)??, need_dlgt(sb)); + rerr = PTR_ERR(d); + if (IS_ERR(d)) + goto out_rerr; + dput(a.hidden_dentry); + a.hidden_dentry = d; + DEBUG_ON(!d->d_inode); + } +#endif + rerr = vfsub_unlink(a.hidden_dir, a.hidden_dentry, a.dlgt); + //rerr = -1; + if (!rerr) + goto out_dt; +// out_rerr: + IOErr("%.*s reverting failed(%d, %d)\n", DLNPair(dentry), err, rerr); + err = -EIO; + out_dt: + d_drop(dentry); + dtime_revert(&dt, !CPUP_LOCKED_GHDIR); + out_unlock: + hdir_unlock(a.hidden_dir, dir, a.bdst); + dput(wh_dentry); + out: + if (unlikely(err)) { + au_update_dbstart(dentry); + d_drop(dentry); + } + di_write_unlock(a.parent); + aufs_read_and_write_unlock2(dentry, src_dentry); + TraceErr(err); + return err; +} + +int aufs_mkdir(struct inode *dir, struct dentry *dentry, int mode) +{ + int err, rerr, diropq, dlgt; + struct dentry *hidden_dentry, *hidden_parent, *wh_dentry, *parent, + *opq_dentry; + struct inode *hidden_dir, *hidden_inode; + struct dtime dt; + aufs_bindex_t bindex; + struct super_block *sb; + + LKTRTrace("i%lu, %.*s, mode 0%o\n", dir->i_ino, DLNPair(dentry), mode); + IMustLock(dir); + + aufs_read_lock(dentry, AUFS_D_WLOCK); + parent = dentry->d_parent; + di_write_lock_parent(parent); + wh_dentry = lock_hdir_lkup_wh(dentry, &dt, /*src_dentry*/NULL, + /*do_lock_srcdir*/0); + //wh_dentry = ERR_PTR(-1); + err = PTR_ERR(wh_dentry); + if (IS_ERR(wh_dentry)) + goto out; + + sb = dentry->d_sb; + bindex = dbstart(dentry); + hidden_dentry = au_h_dptr(dentry); + hidden_parent = hidden_dentry->d_parent; + hidden_dir = hidden_parent->d_inode; + IMustLock(hidden_dir); + dlgt = need_dlgt(sb); + + err = vfsub_mkdir(hidden_dir, hidden_dentry, mode, dlgt); + //err = -1; + if (unlikely(err)) + goto out_unlock; + hidden_inode = hidden_dentry->d_inode; + + /* make the dir opaque */ + diropq = 0; + if (unlikely(wh_dentry || au_flag_test(sb, AuFlag_ALWAYS_DIROPQ))) { + hi_lock_child(hidden_inode); + opq_dentry = create_diropq(dentry, bindex, dlgt); + //opq_dentry = ERR_PTR(-1); + i_unlock(hidden_inode); + err = PTR_ERR(opq_dentry); + if (IS_ERR(opq_dentry)) + goto out_dir; + dput(opq_dentry); + diropq = 1; + } + + err = epilog(wh_dentry, dentry); + //err = -1; + if (!err) { + dir->i_nlink++; + goto out_unlock; /* success */ + } + + /* revert */ + if (unlikely(diropq)) { + LKTRLabel(revert opq); + hi_lock_child(hidden_inode); + rerr = remove_diropq(dentry, bindex, dlgt); + //rerr = -1; + i_unlock(hidden_inode); + if (rerr) { + IOErr("%.*s reverting diropq failed(%d, %d)\n", + DLNPair(dentry), err, rerr); + err = -EIO; + } + } + + out_dir: + LKTRLabel(revert dir); + rerr = vfsub_rmdir(hidden_dir, hidden_dentry, dlgt); + //rerr = -1; + if (rerr) { + IOErr("%.*s reverting dir failed(%d, %d)\n", + DLNPair(dentry), err, rerr); + err = -EIO; + } + d_drop(dentry); + dtime_revert(&dt, /*fake flag*/CPUP_LOCKED_GHDIR); + out_unlock: + hdir_unlock(hidden_dir, dir, bindex); + dput(wh_dentry); + out: + if (unlikely(err)) { + au_update_dbstart(dentry); + d_drop(dentry); + } + di_write_unlock(parent); + aufs_read_unlock(dentry, AUFS_D_WLOCK); + TraceErr(err); + return err; +} diff --git a/fs/aufs/i_op_del.c b/fs/aufs/i_op_del.c new file mode 100755 index 0000000..f29b204 --- /dev/null +++ b/fs/aufs/i_op_del.c @@ -0,0 +1,414 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: i_op_del.c,v 1.35 2007/05/14 03:41:52 sfjro Exp $ */ + +#include "aufs.h" + +/* returns, + * 0: wh is unnecessary + * plus: wh is necessary + * minus: error + */ +int wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup, + struct dentry *locked) +{ + int need_wh, err; + aufs_bindex_t bstart; + struct dentry *hidden_dentry; + struct super_block *sb; + + LKTRTrace("%.*s, isdir %d, *bcpup %d, locked %p\n", + DLNPair(dentry), isdir, *bcpup, locked); + sb = dentry->d_sb; + + bstart = dbstart(dentry); + LKTRTrace("bcpup %d, bstart %d\n", *bcpup, bstart); + hidden_dentry = au_h_dptr(dentry); + if (*bcpup < 0) { + *bcpup = bstart; + if (test_ro(sb, bstart, dentry->d_inode)) { + *bcpup = err = find_rw_parent_br(dentry, bstart); + //*bcpup = err = find_rw_br(sb, bstart); + //err = -1; + if (unlikely(err < 0)) + goto out; + } + } else { + /* braces are added to stop a warning */ + DEBUG_ON(bstart < *bcpup + || test_ro(sb, *bcpup, dentry->d_inode)); + } + LKTRTrace("bcpup %d, bstart %d\n", *bcpup, bstart); + + if (*bcpup != bstart) { + err = cpup_dirs(dentry, *bcpup, locked); + //err = -1; + if (unlikely(err)) + goto out; + need_wh = 1; + } else { + //struct nameidata nd; + aufs_bindex_t old_bend, new_bend, bdiropq = -1; + old_bend = dbend(dentry); + if (isdir) { + bdiropq = dbdiropq(dentry); + set_dbdiropq(dentry, -1); + } + err = need_wh = lkup_dentry(dentry, bstart + 1, /*type*/0); + //err = -1; + if (isdir) + set_dbdiropq(dentry, bdiropq); + if (unlikely(err < 0)) + goto out; + new_bend = dbend(dentry); + if (!need_wh && old_bend != new_bend) { + set_h_dptr(dentry, new_bend, NULL); + set_dbend(dentry, old_bend); +#if 0 + } else if (!au_h_dptr_i(dentry, new_bend)->d_inode) { + LKTRTrace("negative\n"); + set_h_dptr(dentry, new_bend, NULL); + set_dbend(dentry, old_bend); + need_wh = 0; +#endif + } + } + LKTRTrace("need_wh %d\n", need_wh); + err = need_wh; + + out: + TraceErr(err); + return err; +} + +static struct dentry * +lock_hdir_create_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup, + struct dtime *dt) +{ + struct dentry *wh_dentry; + int err, need_wh; + struct dentry *hidden_parent, *parent; + struct inode *dir, *h_dir; + struct lkup_args lkup; + + LKTRTrace("%.*s, isdir %d\n", DLNPair(dentry), isdir); + + err = need_wh = wr_dir_need_wh(dentry, isdir, bcpup, NULL); + //err = -1; + wh_dentry = ERR_PTR(err); + if (unlikely(err < 0)) + goto out; + + parent = dentry->d_parent; + dir = parent->d_inode; + hidden_parent = au_h_dptr_i(parent, *bcpup); + h_dir = hidden_parent->d_inode; + hdir_lock(h_dir, dir, *bcpup); + dtime_store(dt, parent, hidden_parent); + if (!need_wh) + return NULL; /* success, no need to create whiteout */ + + lkup.nfsmnt = au_nfsmnt(dentry->d_sb, *bcpup); + lkup.dlgt = need_dlgt(dentry->d_sb); + wh_dentry = simple_create_wh(dentry, *bcpup, hidden_parent, &lkup); + //wh_dentry = ERR_PTR(-1); + if (!IS_ERR(wh_dentry)) + goto out; /* success */ + /* returns with the parent is locked and wh_dentry is DGETed */ + + hdir_unlock(h_dir, dir, *bcpup); + + out: + TraceErrPtr(wh_dentry); + return wh_dentry; +} + +static int renwh_and_rmdir(struct dentry *dentry, aufs_bindex_t bindex, + struct aufs_nhash *whlist, struct inode *dir) +{ + int rmdir_later, err; + struct dentry *hidden_dentry; + + LKTRTrace("%.*s, b%d\n", DLNPair(dentry), bindex); + + err = rename_whtmp(dentry, bindex); + //err = -1; +#if 0 + //todo: bug + if (unlikely(err)) { + au_direval_inc(dentry->d_parent); + return err; + } +#endif + + hidden_dentry = au_h_dptr_i(dentry, bindex); + if (!au_is_nfs(hidden_dentry->d_sb)) { + const int dirwh = stosi(dentry->d_sb)->si_dirwh; + rmdir_later = (dirwh <= 1); + if (!rmdir_later) + rmdir_later = is_longer_wh(whlist, bindex, dirwh); + if (rmdir_later) + return rmdir_later; + } + + err = rmdir_whtmp(hidden_dentry, whlist, bindex, dir, dentry->d_inode); + //err = -1; + if (unlikely(err)) { + IOErr("rmdir %.*s, b%d failed, %d. ignored\n", + DLNPair(hidden_dentry), bindex, err); + err = 0; + } + TraceErr(err); + return err; +} + +static void epilog(struct inode *dir, struct dentry *dentry, + aufs_bindex_t bindex) +{ + d_drop(dentry); + dentry->d_inode->i_ctime = dir->i_ctime; + if (atomic_read(&dentry->d_count) == 1) { + set_h_dptr(dentry, dbstart(dentry), NULL); + au_update_dbstart(dentry); + } + if (ibstart(dir) == bindex) + au_cpup_attr_timesizes(dir); + dir->i_version++; +} + +static int do_revert(int err, struct dentry *wh_dentry, struct dentry *dentry, + aufs_bindex_t bwh, struct dtime *dt, int dlgt) +{ + int rerr; + + rerr = au_unlink_wh_dentry(wh_dentry->d_parent->d_inode, wh_dentry, + dentry, dlgt); + //rerr = -1; + if (!rerr) { + set_dbwh(dentry, bwh); + dtime_revert(dt, !CPUP_LOCKED_GHDIR); + return 0; + } + + IOErr("%.*s reverting whiteout failed(%d, %d)\n", + DLNPair(dentry), err, rerr); + return -EIO; +} + +/* ---------------------------------------------------------------------- */ + +int aufs_unlink(struct inode *dir, struct dentry *dentry) +{ + int err, dlgt; + struct inode *inode, *hidden_dir; + struct dentry *parent, *wh_dentry, *hidden_dentry, *hidden_parent; + struct dtime dt; + aufs_bindex_t bwh, bindex, bstart; + struct super_block *sb; + + LKTRTrace("i%lu, %.*s\n", dir->i_ino, DLNPair(dentry)); + IMustLock(dir); + inode = dentry->d_inode; + if (unlikely(!inode)) + return -ENOENT; // possible? + IMustLock(inode); + + aufs_read_lock(dentry, AUFS_D_WLOCK); + parent = dentry->d_parent; + di_write_lock_parent(parent); + + bstart = dbstart(dentry); + bwh = dbwh(dentry); + bindex = -1; + wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/0, &bindex, &dt); + //wh_dentry = ERR_PTR(-1); + err = PTR_ERR(wh_dentry); + if (IS_ERR(wh_dentry)) + goto out; + + sb = dir->i_sb; + dlgt = need_dlgt(sb); + hidden_dentry = au_h_dptr(dentry); + dget(hidden_dentry); + hidden_parent = hidden_dentry->d_parent; + hidden_dir = hidden_parent->d_inode; + + if (bindex == bstart) { + err = vfsub_unlink(hidden_dir, hidden_dentry, dlgt); + //err = -1; + } else { + DEBUG_ON(!wh_dentry); + hidden_parent = wh_dentry->d_parent; + DEBUG_ON(hidden_parent != au_h_dptr_i(parent, bindex)); + hidden_dir = hidden_parent->d_inode; + IMustLock(hidden_dir); + err = 0; + } + + if (!err) { + inode->i_nlink--; + epilog(dir, dentry, bindex); +#if 0 + xino_write0(sb, bstart, hidden_dentry->d_inode->i_ino); + /* ignore this error */ +#endif + goto out_unlock; /* success */ + } + + /* revert */ + if (wh_dentry) { + int rerr; + rerr = do_revert(err, wh_dentry, dentry, bwh, &dt, dlgt); + if (rerr) + err = rerr; + } + + out_unlock: + hdir_unlock(hidden_dir, dir, bindex); + dput(wh_dentry); + dput(hidden_dentry); + out: + di_write_unlock(parent); + aufs_read_unlock(dentry, AUFS_D_WLOCK); + TraceErr(err); + return err; +} + +int aufs_rmdir(struct inode *dir, struct dentry *dentry) +{ + int err, rmdir_later; + struct inode *inode, *hidden_dir; + struct dentry *parent, *wh_dentry, *hidden_dentry, *hidden_parent; + struct dtime dt; + aufs_bindex_t bwh, bindex, bstart; + struct rmdir_whtmp_arg *arg; + struct aufs_nhash *whlist; + struct super_block *sb; + + LKTRTrace("i%lu, %.*s\n", dir->i_ino, DLNPair(dentry)); + IMustLock(dir); + inode = dentry->d_inode; + if (unlikely(!inode)) + return -ENOENT; // possible? + IMustLock(inode); + + whlist = nhash_new(GFP_KERNEL); + err = PTR_ERR(whlist); + if (IS_ERR(whlist)) + goto out; + + err = -ENOMEM; + arg = kmalloc(sizeof(*arg), GFP_KERNEL); + //arg = NULL; + if (unlikely(!arg)) + goto out_whlist; + + aufs_read_lock(dentry, AUFS_D_WLOCK); + parent = dentry->d_parent; + di_write_lock_parent(parent); + err = test_empty(dentry, whlist); + //err = -1; + if (unlikely(err)) + goto out_arg; + + bstart = dbstart(dentry); + bwh = dbwh(dentry); + bindex = -1; + wh_dentry = lock_hdir_create_wh(dentry, /*isdir*/ 1, &bindex, &dt); + //wh_dentry = ERR_PTR(-1); + err = PTR_ERR(wh_dentry); + if (IS_ERR(wh_dentry)) + goto out_arg; + + hidden_dentry = au_h_dptr(dentry); + dget(hidden_dentry); + hidden_parent = hidden_dentry->d_parent; + hidden_dir = hidden_parent->d_inode; + + rmdir_later = 0; + if (bindex == bstart) { + IMustLock(hidden_dir); + err = renwh_and_rmdir(dentry, bstart, whlist, dir); + //err = -1; + if (err > 0) { + rmdir_later = err; + err = 0; + } + } else { + DEBUG_ON(!wh_dentry); + hidden_parent = wh_dentry->d_parent; + DEBUG_ON(hidden_parent != au_h_dptr_i(parent, bindex)); + hidden_dir = hidden_parent->d_inode; + IMustLock(hidden_dir); + err = 0; + } + + sb = dentry->d_sb; + if (!err) { + //aufs_bindex_t bi, bend; + + au_reset_hinotify(inode, /*flags*/0); + inode->i_nlink = 0; + set_dbdiropq(dentry, -1); + epilog(dir, dentry, bindex); + + if (rmdir_later) { + kick_rmdir_whtmp(hidden_dentry, whlist, bstart, dir, + inode, arg); + arg = NULL; + } + +#if 0 + bend = dbend(dentry); + for (bi = bstart; bi <= bend; bi++) { + struct dentry *hd; + hd = au_h_dptr_i(dentry, bi); + if (hd && hd->d_inode) + xino_write0(sb, bi, hd->d_inode->i_ino); + /* ignore this error */ + } +#endif + + goto out_unlock; /* success */ + } + + /* revert */ + LKTRLabel(revert); + if (wh_dentry) { + int rerr; + rerr = do_revert(err, wh_dentry, dentry, bwh, &dt, + need_dlgt(sb)); + if (rerr) + err = rerr; + } + + out_unlock: + hdir_unlock(hidden_dir, dir, bindex); + dput(wh_dentry); + dput(hidden_dentry); + out_arg: + di_write_unlock(parent); + aufs_read_unlock(dentry, AUFS_D_WLOCK); + kfree(arg); + out_whlist: + nhash_del(whlist); + out: + TraceErr(err); + return err; +} diff --git a/fs/aufs/i_op_ren.c b/fs/aufs/i_op_ren.c new file mode 100755 index 0000000..08137f9 --- /dev/null +++ b/fs/aufs/i_op_ren.c @@ -0,0 +1,637 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: i_op_ren.c,v 1.39 2007/05/14 03:41:52 sfjro Exp $ */ + +//#include +//#include +#include "aufs.h" + +enum {SRC, DST}; +struct rename_args { + struct dentry *hidden_dentry[2], *parent[2], *hidden_parent[2]; + struct aufs_nhash whlist; + aufs_bindex_t btgt, bstart[2]; + struct super_block *sb; + + unsigned int isdir:1; + unsigned int issamedir:1; + unsigned int whsrc:1; + unsigned int whdst:1; + unsigned int dlgt:1; +} __attribute__((aligned(sizeof(long)))); + +static int do_rename(struct inode *src_dir, struct dentry *src_dentry, + struct inode *dir, struct dentry *dentry, + struct rename_args *a) +{ + int err, need_diropq, bycpup, rerr; + struct rmdir_whtmp_arg *tharg; + struct dentry *wh_dentry[2], *hidden_dst, *hg_parent; + struct inode *hidden_dir[2]; + aufs_bindex_t bindex, bend; + unsigned int flags; + struct lkup_args lkup = {.dlgt = a->dlgt}; + + LKTRTrace("%.*s/%.*s, %.*s/%.*s, " + "hd{%p, %p}, hp{%p, %p}, wh %p, btgt %d, bstart{%d, %d}, " + "flags{%d, %d, %d, %d}\n", + DLNPair(a->parent[SRC]), DLNPair(src_dentry), + DLNPair(a->parent[DST]), DLNPair(dentry), + a->hidden_dentry[SRC], a->hidden_dentry[DST], + a->hidden_parent[SRC], a->hidden_parent[DST], + &a->whlist, a->btgt, + a->bstart[SRC], a->bstart[DST], + a->isdir, a->issamedir, a->whsrc, a->whdst); + hidden_dir[SRC] = a->hidden_parent[SRC]->d_inode; + hidden_dir[DST] = a->hidden_parent[DST]->d_inode; + IMustLock(hidden_dir[SRC]); + IMustLock(hidden_dir[DST]); + + /* prepare workqueue arg */ + hidden_dst = NULL; + tharg = NULL; + if (a->isdir && a->hidden_dentry[DST]->d_inode) { + err = -ENOMEM; + tharg = kmalloc(sizeof(*tharg), GFP_KERNEL); + //tharg = NULL; + if (unlikely(!tharg)) + goto out; + hidden_dst = dget(a->hidden_dentry[DST]); + } + + wh_dentry[SRC] = wh_dentry[DST] = NULL; + lkup.nfsmnt = au_nfsmnt(a->sb, a->btgt); + /* create whiteout for src_dentry */ + if (a->whsrc) { + wh_dentry[SRC] = simple_create_wh(src_dentry, a->btgt, + a->hidden_parent[SRC], &lkup); + //wh_dentry[SRC] = ERR_PTR(-1); + err = PTR_ERR(wh_dentry[SRC]); + if (IS_ERR(wh_dentry[SRC])) + goto out_tharg; + } + + /* lookup whiteout for dentry */ + if (a->whdst) { + struct dentry *d; + d = lkup_wh(a->hidden_parent[DST], &dentry->d_name, &lkup); + //d = ERR_PTR(-1); + err = PTR_ERR(d); + if (IS_ERR(d)) + goto out_whsrc; + if (!d->d_inode) + dput(d); + else + wh_dentry[DST] = d; + } + + /* rename dentry to tmpwh */ + if (tharg) { + err = rename_whtmp(dentry, a->btgt); + //err = -1; + if (unlikely(err)) + goto out_whdst; + set_h_dptr(dentry, a->btgt, NULL); + err = lkup_neg(dentry, a->btgt); + //err = -1; + if (unlikely(err)) + goto out_whtmp; + a->hidden_dentry[DST] = au_h_dptr_i(dentry, a->btgt); + } + + /* cpup src */ + if (a->hidden_dentry[DST]->d_inode && a->bstart[SRC] != a->btgt) { + flags = au_flags_cpup(!CPUP_DTIME, a->parent[SRC]); + hg_parent = a->hidden_parent[SRC]->d_parent; + if (!(flags & CPUP_LOCKED_GHDIR) + && hg_parent == a->hidden_parent[DST]) + flags |= CPUP_LOCKED_GHDIR; + + hi_lock_child(a->hidden_dentry[SRC]->d_inode); + err = sio_cpup_simple(src_dentry, a->btgt, -1, flags); + //err = -1; // untested dir + i_unlock(a->hidden_dentry[SRC]->d_inode); + if (unlikely(err)) + goto out_whtmp; + } + +#if 0 + /* clear the target ino in xino */ + LKTRTrace("dir %d, dst inode %p\n", a->isdir, a->hidden_dentry[DST]->d_inode); + if (a->isdir && a->hidden_dentry[DST]->d_inode) { + Dbg("here\n"); + err = xino_write(a->sb, a->btgt, + a->hidden_dentry[DST]->d_inode->i_ino, 0); + if (unlikely(err)) + goto out_whtmp; + } +#endif + + /* rename by vfs_rename or cpup */ + need_diropq = a->isdir + && (wh_dentry[DST] + || dbdiropq(dentry) == a->btgt + || au_flag_test(a->sb, AuFlag_ALWAYS_DIROPQ)); + bycpup = 0; + if (dbstart(src_dentry) == a->btgt) { + if (need_diropq && dbdiropq(src_dentry) == a->btgt) + need_diropq = 0; + err = vfsub_rename(hidden_dir[SRC], au_h_dptr(src_dentry), + hidden_dir[DST], a->hidden_dentry[DST], + a->dlgt); + //err = -1; + } else { + bycpup = 1; + flags = au_flags_cpup(!CPUP_DTIME, a->parent[DST]); + hg_parent = a->hidden_parent[DST]->d_parent; + if (!(flags & CPUP_LOCKED_GHDIR) + && hg_parent == a->hidden_parent[SRC]) + flags |= CPUP_LOCKED_GHDIR; + + hi_lock_child(a->hidden_dentry[SRC]->d_inode); + set_dbstart(src_dentry, a->btgt); + set_h_dptr(src_dentry, a->btgt, dget(a->hidden_dentry[DST])); + //DbgDentry(src_dentry); + //DbgInode(src_dentry->d_inode); + err = sio_cpup_single(src_dentry, a->btgt, a->bstart[SRC], -1, + flags); + //err = -1; // untested dir + if (unlikely(err)) { + set_h_dptr(src_dentry, a->btgt, NULL); + set_dbstart(src_dentry, a->bstart[SRC]); + } + i_unlock(a->hidden_dentry[SRC]->d_inode); + } + if (unlikely(err)) + goto out_whtmp; + + /* make dir opaque */ + if (need_diropq) { + struct dentry *diropq; + struct inode *h_inode; + + h_inode = au_h_dptr_i(src_dentry, a->btgt)->d_inode; + hdir_lock(h_inode, src_dentry->d_inode, a->btgt); + diropq = create_diropq(src_dentry, a->btgt, a->dlgt); + //diropq = ERR_PTR(-1); + hdir_unlock(h_inode, src_dentry->d_inode, a->btgt); + err = PTR_ERR(diropq); + if (IS_ERR(diropq)) + goto out_rename; + dput(diropq); + } + + /* remove whiteout for dentry */ + if (wh_dentry[DST]) { + err = au_unlink_wh_dentry(hidden_dir[DST], wh_dentry[DST], + dentry, a->dlgt); + //err = -1; + if (unlikely(err)) + goto out_diropq; + } + + /* remove whtmp */ + if (tharg) { + if (au_is_nfs(hidden_dst->d_sb) + || !is_longer_wh(&a->whlist, a->btgt, + stosi(a->sb)->si_dirwh)) { + err = rmdir_whtmp(hidden_dst, &a->whlist, a->btgt, dir, + dentry->d_inode); + if (unlikely(err)) + Warn("failed removing whtmp dir %.*s (%d), " + "ignored.\n", DLNPair(hidden_dst), err); + } else { + kick_rmdir_whtmp(hidden_dst, &a->whlist, a->btgt, dir, + dentry->d_inode, tharg); + dput(hidden_dst); + tharg = NULL; + } + } + err = 0; + goto out_success; + +#define RevertFailure(fmt, args...) do { \ + IOErrWhck("revert failure: " fmt " (%d, %d)\n", \ + ##args, err, rerr); \ + err = -EIO; \ + } while(0) + + out_diropq: + if (need_diropq) { + struct inode *h_inode; + + h_inode = au_h_dptr_i(src_dentry, a->btgt)->d_inode; + // i_lock simplly since inotify is not set to h_inode. + hi_lock_parent(h_inode); + //hdir_lock(h_inode, src_dentry->d_inode, a->btgt); + rerr = remove_diropq(src_dentry, a->btgt, a->dlgt); + //rerr = -1; + //hdir_unlock(h_inode, src_dentry->d_inode, a->btgt); + i_unlock(h_inode); + if (rerr) + RevertFailure("remove diropq %.*s", + DLNPair(src_dentry)); + } + out_rename: + if (!bycpup) { + struct dentry *d; + struct qstr *name = &src_dentry->d_name; + d = lkup_one(name->name, a->hidden_parent[SRC], name->len, + &lkup); + //d = ERR_PTR(-1); + rerr = PTR_ERR(d); + if (IS_ERR(d)) { + RevertFailure("lkup_one %.*s", DLNPair(src_dentry)); + goto out_whtmp; + } + DEBUG_ON(d->d_inode); + rerr = vfsub_rename + (hidden_dir[DST], au_h_dptr_i(src_dentry, a->btgt), + hidden_dir[SRC], d, a->dlgt); + //rerr = -1; + d_drop(d); + dput(d); + //set_h_dptr(src_dentry, a->btgt, NULL); + if (rerr) + RevertFailure("rename %.*s", DLNPair(src_dentry)); + } else { + rerr = vfsub_unlink(hidden_dir[DST], a->hidden_dentry[DST], + a->dlgt); + //rerr = -1; + set_h_dptr(src_dentry, a->btgt, NULL); + set_dbstart(src_dentry, a->bstart[SRC]); + if (rerr) + RevertFailure("unlink %.*s", + DLNPair(a->hidden_dentry[DST])); + } + out_whtmp: + if (tharg) { + struct dentry *d; + struct qstr *name = &dentry->d_name; + LKTRLabel(here); + d = lkup_one(name->name, a->hidden_parent[DST], name->len, + &lkup); + //d = ERR_PTR(-1); + rerr = PTR_ERR(d); + if (IS_ERR(d)) { + RevertFailure("lookup %.*s", LNPair(name)); + goto out_whdst; + } + if (d->d_inode) { + d_drop(d); + dput(d); + goto out_whdst; + } + DEBUG_ON(d->d_inode); + rerr = vfsub_rename(hidden_dir[DST], hidden_dst, + hidden_dir[DST], d, a->dlgt); + //rerr = -1; + d_drop(d); + dput(d); + if (rerr) { + RevertFailure("rename %.*s", DLNPair(hidden_dst)); + goto out_whdst; + } + set_h_dptr(dentry, a->btgt, NULL); + set_h_dptr(dentry, a->btgt, dget(hidden_dst)); + } + out_whdst: + dput(wh_dentry[DST]); + wh_dentry[DST] = NULL; + out_whsrc: + if (wh_dentry[SRC]) { + LKTRLabel(here); + rerr = au_unlink_wh_dentry(hidden_dir[SRC], wh_dentry[SRC], + src_dentry, a->dlgt); + //rerr = -1; + if (rerr) + RevertFailure("unlink %.*s", DLNPair(wh_dentry[SRC])); + } +#undef RevertFailure + d_drop(src_dentry); + bend = dbend(src_dentry); + for (bindex = dbstart(src_dentry); bindex <= bend; bindex++) { + struct dentry *hd; + hd = au_h_dptr_i(src_dentry, bindex); + if (hd) + d_drop(hd); + } + d_drop(dentry); + bend = dbend(dentry); + for (bindex = dbstart(dentry); bindex <= bend; bindex++) { + struct dentry *hd; + hd = au_h_dptr_i(dentry, bindex); + if (hd) + d_drop(hd); + } + au_update_dbstart(dentry); + if (tharg) + d_drop(hidden_dst); + out_success: + dput(wh_dentry[SRC]); + dput(wh_dentry[DST]); + out_tharg: + if (tharg) { + dput(hidden_dst); + kfree(tharg); + } + out: + TraceErr(err); + return err; +} + +/* + * test if @dentry dir can be rename destination or not. + * success means, it is a logically empty dir. + */ +static int may_rename_dstdir(struct dentry *dentry, aufs_bindex_t btgt, + struct aufs_nhash *whlist) +{ + LKTRTrace("%.*s\n", DLNPair(dentry)); + + return test_empty(dentry, whlist); +} + +/* + * test if @dentry dir can be rename source or not. + * if it can, return 0 and @children is filled. + * success means, + * - or, it is a logically empty dir. + * - or, it exists on writable branch and has no children including whiteouts + * on the lower branch. + */ +static int may_rename_srcdir(struct dentry *dentry, aufs_bindex_t btgt) +{ + int err; + aufs_bindex_t bstart; + + LKTRTrace("%.*s\n", DLNPair(dentry)); + + bstart = dbstart(dentry); + if (bstart != btgt) { + struct aufs_nhash *whlist; + + whlist = nhash_new(GFP_KERNEL); + err = PTR_ERR(whlist); + if (IS_ERR(whlist)) + goto out; + err = test_empty(dentry, whlist); + nhash_del(whlist); + goto out; + } + + if (bstart == dbtaildir(dentry)) + return 0; /* success */ + + err = au_test_empty_lower(dentry); + + out: + if (/* unlikely */(err == -ENOTEMPTY)) + err = -EXDEV; + TraceErr(err); + return err; +} + +int aufs_rename(struct inode *src_dir, struct dentry *src_dentry, + struct inode *dir, struct dentry *dentry) +{ + int err, do_dt_dstdir; + aufs_bindex_t bend, bindex; + struct inode *inode, *dirs[2]; + enum {PARENT, CHILD}; + /* reduce stack space */ + struct { + struct rename_args a; + struct dtime dt[2][2]; + } *p; + + LKTRTrace("i%lu, %.*s, i%lu, %.*s\n", + src_dir->i_ino, DLNPair(src_dentry), + dir->i_ino, DLNPair(dentry)); + IMustLock(src_dir); + IMustLock(dir); + /* braces are added to stop a warning */ + if (dentry->d_inode) { + IMustLock(dentry->d_inode); + } + + err = -ENOMEM; + BUILD_BUG_ON(sizeof(*p) > PAGE_SIZE); + p = kmalloc(sizeof(*p), GFP_KERNEL); + if (unlikely(!p)) + goto out; + + err = -ENOTDIR; + p->a.sb = src_dentry->d_sb; + inode = src_dentry->d_inode; + p->a.isdir = !!S_ISDIR(inode->i_mode); + if (unlikely(p->a.isdir && dentry->d_inode + && !S_ISDIR(dentry->d_inode->i_mode))) + goto out_free; + + aufs_read_and_write_lock2(dentry, src_dentry, p->a.isdir); + p->a.dlgt = !!need_dlgt(p->a.sb); + p->a.parent[SRC] = p->a.parent[DST] = dentry->d_parent; + p->a.issamedir = (src_dir == dir); + if (p->a.issamedir) + di_write_lock_parent(p->a.parent[DST]); + else { + p->a.parent[SRC] = src_dentry->d_parent; + di_write_lock2_parent(p->a.parent[SRC], p->a.parent[DST], + /*isdir*/1); + } + + /* which branch we process */ + p->a.bstart[DST] = dbstart(dentry); + p->a.btgt = err = wr_dir(dentry, 1, src_dentry, /*force_btgt*/-1, + /*do_lock_srcdir*/0); + if (unlikely(err < 0)) + goto out_unlock; + + /* are they available to be renamed */ + err = 0; + nhash_init(&p->a.whlist); + if (p->a.isdir && dentry->d_inode) { + set_dbstart(dentry, p->a.bstart[DST]); + err = may_rename_dstdir(dentry, p->a.btgt, &p->a.whlist); + set_dbstart(dentry, p->a.btgt); + } + p->a.hidden_dentry[DST] = au_h_dptr(dentry); + if (unlikely(err)) + goto out_unlock; + //todo: minor optimize, their sb may be same while their bindex differs. + p->a.bstart[SRC] = dbstart(src_dentry); + p->a.hidden_dentry[SRC] = au_h_dptr(src_dentry); + if (p->a.isdir) { + err = may_rename_srcdir(src_dentry, p->a.btgt); + if (unlikely(err)) + goto out_children; + } + + /* prepare the writable parent dir on the same branch */ + err = wr_dir_need_wh(src_dentry, p->a.isdir, &p->a.btgt, + p->a.issamedir ? NULL : p->a.parent[DST]); + if (unlikely(err < 0)) + goto out_children; + p->a.whsrc = !!err; + p->a.whdst = (p->a.bstart[DST] == p->a.btgt); + if (!p->a.whdst) { + err = cpup_dirs(dentry, p->a.btgt, + p->a.issamedir ? NULL : p->a.parent[SRC]); + if (unlikely(err)) + goto out_children; + } + + p->a.hidden_parent[SRC] = au_h_dptr_i(p->a.parent[SRC], p->a.btgt); + p->a.hidden_parent[DST] = au_h_dptr_i(p->a.parent[DST], p->a.btgt); + dirs[0] = src_dir; + dirs[1] = dir; + hdir_lock_rename(p->a.hidden_parent, dirs, p->a.btgt, p->a.issamedir); + + /* store timestamps to be revertible */ + dtime_store(p->dt[PARENT] + SRC, p->a.parent[SRC], + p->a.hidden_parent[SRC]); + if (!p->a.issamedir) + dtime_store(p->dt[PARENT] + DST, p->a.parent[DST], + p->a.hidden_parent[DST]); + do_dt_dstdir = 0; + if (p->a.isdir) { + dtime_store(p->dt[CHILD] + SRC, src_dentry, + p->a.hidden_dentry[SRC]); + if (p->a.hidden_dentry[DST]->d_inode) { + do_dt_dstdir = 1; + dtime_store(p->dt[CHILD] + DST, dentry, + p->a.hidden_dentry[DST]); + } + } + + err = do_rename(src_dir, src_dentry, dir, dentry, &p->a); + if (unlikely(err)) + goto out_dt; + hdir_unlock_rename(p->a.hidden_parent, dirs, p->a.btgt, p->a.issamedir); + + /* update dir attributes */ + dir->i_version++; + if (p->a.isdir) + au_cpup_attr_nlink(dir); + if (ibstart(dir) == p->a.btgt) + au_cpup_attr_timesizes(dir); + + if (!p->a.issamedir) { + src_dir->i_version++; + if (p->a.isdir) + au_cpup_attr_nlink(src_dir); + if (ibstart(src_dir) == p->a.btgt) + au_cpup_attr_timesizes(src_dir); + } + + // is this updating defined in POSIX? + if (unlikely(p->a.isdir)) { + //i_lock(inode); + au_cpup_attr_timesizes(inode); + //i_unlock(inode); + } + +#if 0 + d_drop(src_dentry); +#else + /* dput/iput all lower dentries */ + set_dbwh(src_dentry, -1); + bend = dbend(src_dentry); + for (bindex = p->a.btgt + 1; bindex <= bend; bindex++) { + struct dentry *hd; + hd = au_h_dptr_i(src_dentry, bindex); + if (hd) + set_h_dptr(src_dentry, bindex, NULL); + } + set_dbend(src_dentry, p->a.btgt); + + bend = ibend(inode); + for (bindex = p->a.btgt + 1; bindex <= bend; bindex++) { + struct inode *hi; + hi = au_h_iptr_i(inode, bindex); + if (hi) + set_h_iptr(inode, bindex, NULL, 0); + } + set_ibend(inode, p->a.btgt); +#endif + +#if 0 + //au_debug_on(); + //DbgDentry(dentry); + //DbgInode(dentry->d_inode); + //au_debug_off(); + inode = dentry->d_inode; + if (inode) { + aufs_bindex_t bindex, bend; + struct dentry *hd; + bend = dbend(dentry); + for (bindex = dbstart(dentry); bindex <= bend; bindex++) { + hd = au_h_dptr_i(dentry, bindex); + if (hd && hd->d_inode) + xino_write0(p->a.sb, bindex, hd->d_inode->i_ino); + /* ignore this error */ + } + } +#endif + + goto out_children; /* success */ + + out_dt: + dtime_revert(p->dt[PARENT] + SRC, + p->a.hidden_parent[SRC]->d_parent + == p->a.hidden_parent[DST]); + if (!p->a.issamedir) + dtime_revert(p->dt[PARENT] + DST, + p->a.hidden_parent[DST]->d_parent + == p->a.hidden_parent[SRC]); + if (p->a.isdir && err != -EIO) { + struct dentry *hd; + + hd = p->dt[CHILD][SRC].dt_h_dentry; + hi_lock_child(hd->d_inode); + dtime_revert(p->dt[CHILD] + SRC, 1); + i_unlock(hd->d_inode); + if (do_dt_dstdir) { + hd = p->dt[CHILD][DST].dt_h_dentry; + hi_lock_child(hd->d_inode); + dtime_revert(p->dt[CHILD] + DST, 1); + i_unlock(hd->d_inode); + } + } + hdir_unlock_rename(p->a.hidden_parent, dirs, p->a.btgt, p->a.issamedir); + out_children: + nhash_fin(&p->a.whlist); + out_unlock: + //if (unlikely(err /* && p->a.isdir */)) { + if (unlikely(err && p->a.isdir)) { + au_update_dbstart(dentry); + d_drop(dentry); + } + if (p->a.issamedir) + di_write_unlock(p->a.parent[DST]); + else + di_write_unlock2(p->a.parent[SRC], p->a.parent[DST]); + aufs_read_and_write_unlock2(dentry, src_dentry); + out_free: + kfree(p); + out: + TraceErr(err); + return err; +} diff --git a/fs/aufs/iinfo.c b/fs/aufs/iinfo.c new file mode 100755 index 0000000..9efbd38 --- /dev/null +++ b/fs/aufs/iinfo.c @@ -0,0 +1,286 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: iinfo.c,v 1.31 2007/05/14 03:41:52 sfjro Exp $ */ + +//#include +#include "aufs.h" + +struct aufs_iinfo *itoii(struct inode *inode) +{ + struct aufs_iinfo *iinfo; + + iinfo = &(container_of(inode, struct aufs_icntnr, vfs_inode)->iinfo); + /* bad_inode case */ + if (unlikely(!iinfo->ii_hinode)) + return NULL; + DEBUG_ON(!iinfo->ii_hinode + /* || stosi(inode->i_sb)->si_bend < iinfo->ii_bend */ + || iinfo->ii_bend < iinfo->ii_bstart); + return iinfo; +} + +aufs_bindex_t ibstart(struct inode *inode) +{ + IiMustAnyLock(inode); + return itoii(inode)->ii_bstart; +} + +aufs_bindex_t ibend(struct inode *inode) +{ + IiMustAnyLock(inode); + return itoii(inode)->ii_bend; +} + +struct aufs_vdir *ivdir(struct inode *inode) +{ + IiMustAnyLock(inode); + DEBUG_ON(!S_ISDIR(inode->i_mode)); + return itoii(inode)->ii_vdir; +} + +struct inode *au_h_iptr_i(struct inode *inode, aufs_bindex_t bindex) +{ + struct inode *hidden_inode; + + IiMustAnyLock(inode); + DEBUG_ON(bindex < 0 || ibend(inode) < bindex); + hidden_inode = itoii(inode)->ii_hinode[0 + bindex].hi_inode; + DEBUG_ON(hidden_inode && atomic_read(&hidden_inode->i_count) <= 0); + return hidden_inode; +} + +struct inode *au_h_iptr(struct inode *inode) +{ + return au_h_iptr_i(inode, ibstart(inode)); +} + +aufs_bindex_t itoid_index(struct inode *inode, aufs_bindex_t bindex) +{ + IiMustAnyLock(inode); + DEBUG_ON(bindex < 0 + || ibend(inode) < bindex + || !itoii(inode)->ii_hinode[0 + bindex].hi_inode); + return itoii(inode)->ii_hinode[0 + bindex].hi_id; +} + +// hard/soft set +void set_ibstart(struct inode *inode, aufs_bindex_t bindex) +{ + struct aufs_iinfo *iinfo = itoii(inode); + struct inode *h_inode; + + IiMustWriteLock(inode); + DEBUG_ON(sbend(inode->i_sb) < bindex); + iinfo->ii_bstart = bindex; + h_inode = iinfo->ii_hinode[bindex + 0].hi_inode; + if (h_inode) + au_cpup_igen(inode, h_inode); +} + +void set_ibend(struct inode *inode, aufs_bindex_t bindex) +{ + IiMustWriteLock(inode); + DEBUG_ON(sbend(inode->i_sb) < bindex + || bindex < ibstart(inode)); + itoii(inode)->ii_bend = bindex; +} + +void set_ivdir(struct inode *inode, struct aufs_vdir *vdir) +{ + IiMustWriteLock(inode); + DEBUG_ON(!S_ISDIR(inode->i_mode) + || (itoii(inode)->ii_vdir && vdir)); + itoii(inode)->ii_vdir = vdir; +} + +void aufs_hiput(struct aufs_hinode *hinode) +{ + if (unlikely(hinode->hi_notify)) + do_free_hinotify(hinode); + if (hinode->hi_inode) + iput(hinode->hi_inode); +} + +unsigned int au_hi_flags(struct inode *inode, int isdir) +{ + unsigned int flags; + struct super_block *sb = inode->i_sb; + + flags = 0; + if (au_flag_test(sb, AuFlag_XINO)) + flags = AUFS_HI_XINO; + if (unlikely(isdir && au_flag_test(sb, AuFlag_UDBA_INOTIFY))) + flags |= AUFS_HI_NOTIFY; + return flags; +} + +void set_h_iptr(struct inode *inode, aufs_bindex_t bindex, + struct inode *h_inode, unsigned int flags) +{ + struct aufs_hinode *hinode; + struct inode *hi; + struct aufs_iinfo *iinfo = itoii(inode); + + LKTRTrace("i%lu, b%d, hi%lu, flags 0x%x\n", + inode->i_ino, bindex, h_inode ? h_inode->i_ino : 0, flags); + IiMustWriteLock(inode); + hinode = iinfo->ii_hinode + bindex; + hi = hinode->hi_inode; + DEBUG_ON(bindex < ibstart(inode) || ibend(inode) < bindex + || (h_inode && atomic_read(&h_inode->i_count) <= 0) + || (h_inode && hi)); + + if (hi) + aufs_hiput(hinode); + hinode->hi_inode = h_inode; + if (h_inode) { + int err; + struct super_block *sb = inode->i_sb; + + if (bindex == iinfo->ii_bstart) + au_cpup_igen(inode, h_inode); + hinode->hi_id = sbr_id(sb, bindex); + if (flags & AUFS_HI_XINO) { + struct xino xino = { + .ino = inode->i_ino, + //.h_gen = h_inode->i_generation + }; + //WARN_ON(xino.h_gen == AuXino_INVALID_HGEN); + err = xino_write(sb, bindex, h_inode->i_ino, &xino); + if (unlikely(err)) { + IOErr1("failed xino_write() %d, force noxino\n", + err); + au_flag_clr(sb, AuFlag_XINO); + } + } + if (flags & AUFS_HI_NOTIFY) { + err = alloc_hinotify(hinode, inode, h_inode); + if (unlikely(err)) + IOErr1("alloc_hinotify() %d\n", err); + else { + /* braces are added to stop a warning */ + DEBUG_ON(!hinode->hi_notify); + } + } + } +} + +void au_update_iigen(struct inode *inode) +{ + //IiMustWriteLock(inode); + DEBUG_ON(!inode->i_sb); + atomic_set(&itoii(inode)->ii_generation, au_sigen(inode->i_sb)); +} + +/* it may be called at remount time, too */ +void au_update_brange(struct inode *inode, int do_put_zero) +{ + struct aufs_iinfo *iinfo; + + LKTRTrace("i%lu, %d\n", inode->i_ino, do_put_zero); + IiMustWriteLock(inode); + + iinfo = itoii(inode); + if (unlikely(!iinfo) || iinfo->ii_bstart < 0) + return; + + if (do_put_zero) { + aufs_bindex_t bindex; + for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend; + bindex++) { + struct inode *h_i; + h_i = iinfo->ii_hinode[0 + bindex].hi_inode; + if (h_i && !h_i->i_nlink) + set_h_iptr(inode, bindex, NULL, 0); + } + } + + iinfo->ii_bstart = -1; + while (++iinfo->ii_bstart <= iinfo->ii_bend) + if (iinfo->ii_hinode[0 + iinfo->ii_bstart].hi_inode) + break; + if (iinfo->ii_bstart > iinfo->ii_bend) { + iinfo->ii_bend = iinfo->ii_bstart = -1; + return; + } + + iinfo->ii_bend++; + while (0 <= --iinfo->ii_bend) + if (iinfo->ii_hinode[0 + iinfo->ii_bend].hi_inode) + break; +} + +/* ---------------------------------------------------------------------- */ + +int au_iinfo_init(struct inode *inode) +{ + struct aufs_iinfo *iinfo; + struct super_block *sb; + int nbr, i; + + sb = inode->i_sb; + DEBUG_ON(!sb); + iinfo = &(container_of(inode, struct aufs_icntnr, vfs_inode)->iinfo); + DEBUG_ON(iinfo->ii_hinode); + nbr = sbend(sb) + 1; + if (unlikely(!nbr)) + nbr++; + iinfo->ii_hinode = kcalloc(nbr, sizeof(*iinfo->ii_hinode), GFP_KERNEL); + //iinfo->ii_hinode = NULL; + if (iinfo->ii_hinode) { + for (i = 0; i < nbr; i++) + iinfo->ii_hinode[i].hi_id = -1; + atomic_set(&iinfo->ii_generation, au_sigen(sb)); + rw_init_nolock(&iinfo->ii_rwsem); + iinfo->ii_bstart = -1; + iinfo->ii_bend = -1; + iinfo->ii_vdir = NULL; + return 0; + } + return -ENOMEM; +} + +void au_iinfo_fin(struct inode *inode) +{ + struct aufs_iinfo *iinfo; + + iinfo = itoii(inode); + /* bad_inode case */ + if (unlikely(!iinfo)) + return; + + if (unlikely(iinfo->ii_vdir)) + free_vdir(iinfo->ii_vdir); + + if (iinfo->ii_bstart >= 0) { + aufs_bindex_t bend; + struct aufs_hinode *hi; + hi = iinfo->ii_hinode + iinfo->ii_bstart; + bend = iinfo->ii_bend; + while (iinfo->ii_bstart++ <= bend) { + if (hi->hi_inode) + aufs_hiput(hi); + hi++; + } + //iinfo->ii_bstart = iinfo->ii_bend = -1; + } + + kfree(iinfo->ii_hinode); + //iinfo->ii_hinode = NULL; +} diff --git a/fs/aufs/inode.c b/fs/aufs/inode.c new file mode 100755 index 0000000..f18b5d8 --- /dev/null +++ b/fs/aufs/inode.c @@ -0,0 +1,339 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: inode.c,v 1.22 2007/05/07 03:44:35 sfjro Exp $ */ + +#include "aufs.h" + +int au_refresh_hinode(struct inode *inode, struct dentry *dentry) +{ + int err, new_sz, update, isdir; + struct inode *first; + struct aufs_hinode *p, *q, tmp; + struct super_block *sb; + struct aufs_iinfo *iinfo; + aufs_bindex_t bindex, bend, new_bindex; + unsigned int flags; + + LKTRTrace("%.*s\n", DLNPair(dentry)); + IiMustWriteLock(inode); + + err = -ENOMEM; + sb = dentry->d_sb; + bend = sbend(sb); + new_sz = sizeof(*iinfo->ii_hinode) * (bend + 1); + iinfo = itoii(inode); + p = au_kzrealloc(iinfo->ii_hinode, sizeof(*p) * (iinfo->ii_bend + 1), + new_sz, GFP_KERNEL); + //p = NULL; + if (unlikely(!p)) + goto out; + + iinfo->ii_hinode = p; + err = 0; + update = 0; + p = iinfo->ii_hinode + iinfo->ii_bstart; + first = p->hi_inode; + for (bindex = iinfo->ii_bstart; bindex <= iinfo->ii_bend; + bindex++, p++) { + if (unlikely(!p->hi_inode)) + continue; + + new_bindex = find_brindex(sb, p->hi_id); + if (new_bindex == bindex) + continue; + if (new_bindex < 0) { + update++; + aufs_hiput(p); + p->hi_inode = NULL; + continue; + } + + if (new_bindex < iinfo->ii_bstart) + iinfo->ii_bstart = new_bindex; + if (iinfo->ii_bend < new_bindex) + iinfo->ii_bend = new_bindex; + /* swap two hidden inode, and loop again */ + q = iinfo->ii_hinode + new_bindex; + tmp = *q; + *q = *p; + *p = tmp; + if (tmp.hi_inode) { + bindex--; + p--; + } + } + + isdir = S_ISDIR(inode->i_mode); + flags = au_hi_flags(inode, isdir); + bend = dbend(dentry); + for (bindex = dbstart(dentry); bindex <= bend; bindex++) { + struct inode *hi; + struct dentry *hd; + + hd = au_h_dptr_i(dentry, bindex); + if (!hd || !hd->d_inode) + continue; + + if (iinfo->ii_bstart <= bindex && bindex <= iinfo->ii_bend) { + hi = au_h_iptr_i(inode, bindex); + if (hi) { + if (hi == hd->d_inode) + continue; + //Dbg("here\n"); + err = -ESTALE; + break; + } + } + if (bindex < iinfo->ii_bstart) + iinfo->ii_bstart = bindex; + if (iinfo->ii_bend < bindex) + iinfo->ii_bend = bindex; + set_h_iptr(inode, bindex, igrab(hd->d_inode), flags); + update++; + } + + bend = iinfo->ii_bend; + p = iinfo->ii_hinode; + for (bindex = 0; bindex <= bend; bindex++, p++) + if (p->hi_inode) { + iinfo->ii_bstart = bindex; + break; + } + p = iinfo->ii_hinode + bend; + for (bindex = bend; bindex > iinfo->ii_bstart; bindex--, p--) + if (p->hi_inode) { + iinfo->ii_bend = bindex; + break; + } + DEBUG_ON(iinfo->ii_bstart > bend || iinfo->ii_bend < 0); + + if (unlikely(err)) + goto out; + + if (1 || first != au_h_iptr(inode)) + au_cpup_attr_all(inode); + if (update && isdir) + inode->i_version++; + au_update_iigen(inode); + + out: + //au_debug_on(); + TraceErr(err); + //au_debug_off(); + return err; +} + +static int set_inode(struct inode *inode, struct dentry *dentry) +{ + int err, isdir; + struct dentry *hidden_dentry; + struct inode *hidden_inode; + umode_t mode; + aufs_bindex_t bindex, bstart, btail; + struct aufs_iinfo *iinfo; + unsigned int flags; + + LKTRTrace("i%lu, %.*s\n", inode->i_ino, DLNPair(dentry)); + DEBUG_ON(!(inode->i_state & I_NEW)); + IiMustWriteLock(inode); + hidden_dentry = au_h_dptr(dentry); + DEBUG_ON(!hidden_dentry); + hidden_inode = hidden_dentry->d_inode; + DEBUG_ON(!hidden_inode); + + err = 0; + isdir = 0; + bstart = dbstart(dentry); + mode = hidden_inode->i_mode; + switch (mode & S_IFMT) { + case S_IFREG: + btail = dbtail(dentry); + break; + case S_IFDIR: + isdir = 1; + btail = dbtaildir(dentry); + inode->i_op = &aufs_dir_iop; + inode->i_fop = &aufs_dir_fop; + break; + case S_IFLNK: + btail = dbtail(dentry); + inode->i_op = &aufs_symlink_iop; + break; + case S_IFBLK: + case S_IFCHR: + case S_IFIFO: + case S_IFSOCK: + btail = dbtail(dentry); + init_special_inode(inode, mode, hidden_inode->i_rdev); + break; + default: + IOErr("Unknown file type 0%o\n", mode); + err = -EIO; + goto out; + } + + flags = au_hi_flags(inode, isdir); + iinfo = itoii(inode); + iinfo->ii_bstart = bstart; + iinfo->ii_bend = btail; + for (bindex = bstart; bindex <= btail; bindex++) { + hidden_dentry = au_h_dptr_i(dentry, bindex); + if (!hidden_dentry) + continue; + DEBUG_ON(!hidden_dentry->d_inode); + set_h_iptr(inode, bindex, igrab(hidden_dentry->d_inode), flags); + } + au_cpup_attr_all(inode); + + out: + TraceErr(err); + return err; +} + +/* successful returns with iinfo write_locked */ +//todo: return with unlocked? +static int reval_inode(struct inode *inode, struct dentry *dentry, int *matched) +{ + int err; + struct inode *h_inode, *h_dinode; + aufs_bindex_t bindex, bend; + //const int udba = !au_flag_test(inode->i_sb, AuFlag_UDBA_NONE); + + LKTRTrace("i%lu, %.*s\n", inode->i_ino, DLNPair(dentry)); + + *matched = 0; + + /* + * before this function, if aufs got any iinfo lock, it must be only + * one, the parent dir. + * it can happen by UDBA and the obsoleted inode number. + */ + err = -EIO; + if (unlikely(inode->i_ino == parent_ino(dentry))) + goto out; + + h_dinode = au_h_dptr(dentry)->d_inode; + hi_lock_child(inode); // bad name, this is not a hidden inode. + ii_write_lock_new(inode); + bend = ibend(inode); + for (bindex = ibstart(inode); bindex <= bend; bindex++) { + h_inode = au_h_iptr_i(inode, bindex); + if (h_inode && h_inode == h_dinode) { + //&& (ibs != bstart || !au_test_higen(inode, h_inode))); + *matched = 1; + err = 0; + if (unlikely(au_iigen(inode) != au_digen(dentry))) + err = au_refresh_hinode(inode, dentry); + break; + } + } + i_unlock(inode); + if (unlikely(err)) + ii_write_unlock(inode); + + out: + TraceErr(err); + return err; +} + +/* successful returns with iinfo write_locked */ +//todo: return with unlocked? +struct inode *au_new_inode(struct dentry *dentry) +{ + struct inode *inode, *h_inode; + struct dentry *h_dentry; + ino_t h_ino; + struct super_block *sb; + int err, match; + aufs_bindex_t bstart; + struct xino xino; + + LKTRTrace("%.*s\n", DLNPair(dentry)); + sb = dentry->d_sb; + h_dentry = au_h_dptr(dentry); + DEBUG_ON(!h_dentry); + h_inode = h_dentry->d_inode; + DEBUG_ON(!h_inode); + + bstart = dbstart(dentry); + h_ino = h_inode->i_ino; + err = xino_read(sb, bstart, h_ino, &xino); + //err = -1; + inode = ERR_PTR(err); + if (unlikely(err)) + goto out; + new_ino: + if (!xino.ino) { + xino.ino = xino_new_ino(sb); + if (!xino.ino) { + inode = ERR_PTR(-EIO); + goto out; + } + } + + LKTRTrace("i%lu\n", xino.ino); + err = -ENOMEM; + inode = iget_locked(sb, xino.ino); + if (unlikely(!inode)) + goto out; + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out; + err = -ENOMEM; + if (unlikely(is_bad_inode(inode))) + goto out_iput; + + LKTRTrace("%lx, new %d\n", inode->i_state, !!(inode->i_state & I_NEW)); + if (inode->i_state & I_NEW) { + sb->s_op->read_inode(inode); + if (!is_bad_inode(inode)) { + ii_write_lock_new(inode); + err = set_inode(inode, dentry); + //err = -1; + } + unlock_new_inode(inode); + if (!err) + goto out; /* success */ + ii_write_unlock(inode); + goto out_iput; + } else { + err = reval_inode(inode, dentry, &match); + if (!err) + goto out; /* success */ + else if (match) + goto out_iput; + } + + Warn1("broken ino, b%d, %.*s/%.*s, hi%lu, i%lu. Try udba=inotify.\n", + bstart, DLNPair(dentry->d_parent), DLNPair(dentry), h_ino, + xino.ino); + xino.ino = 0; + err = xino_write0(sb, bstart, h_ino); + if (!err) { + iput(inode); + goto new_ino; + } + + out_iput: + iput(inode); + inode = ERR_PTR(err); + out: + TraceErrPtr(inode); + return inode; +} diff --git a/fs/aufs/inode.h b/fs/aufs/inode.h new file mode 100755 index 0000000..b001ac3 --- /dev/null +++ b/fs/aufs/inode.h @@ -0,0 +1,377 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: inode.h,v 1.32 2007/05/14 03:41:52 sfjro Exp $ */ + +#ifndef __AUFS_INODE_H__ +#define __AUFS_INODE_H__ + +#ifdef __KERNEL__ + +#include +#include +#include +#include +#include "misc.h" +#include "vfsub.h" + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) +#else +struct inotify_watch {}; +#endif + +/* ---------------------------------------------------------------------- */ + +struct aufs_hinotify { + struct inotify_watch hin_watch; + struct inode *hin_aufs_inode; /* no get/put */ +}; + +struct aufs_hinode { + struct inode *hi_inode; + aufs_bindex_t hi_id; + struct aufs_hinotify *hi_notify; +}; + +struct aufs_vdir; +struct aufs_iinfo { + atomic_t ii_generation; + struct super_block *ii_hsb1; /* no get/put */ + + struct aufs_rwsem ii_rwsem; + aufs_bindex_t ii_bstart, ii_bend; + struct aufs_hinode *ii_hinode; + struct aufs_vdir *ii_vdir; +}; + +struct aufs_icntnr { + struct aufs_iinfo iinfo; + struct inode vfs_inode; +}; + +/* ---------------------------------------------------------------------- */ + +/* inode.c */ +int au_refresh_hinode(struct inode *inode, struct dentry *dentry); +struct inode *au_new_inode(struct dentry *dentry); + +/* i_op.c */ +extern struct inode_operations aufs_iop, aufs_symlink_iop, aufs_dir_iop; +int wr_dir(struct dentry *dentry, int negative, struct dentry *src_dentry, + aufs_bindex_t force_btgt, int do_lock_srcdir); + +/* i_op_del.c */ +int wr_dir_need_wh(struct dentry *dentry, int isdir, aufs_bindex_t *bcpup, + struct dentry *locked); + +/* iinfo.c */ +struct aufs_iinfo *itoii(struct inode *inode); +aufs_bindex_t ibstart(struct inode *inode); +aufs_bindex_t ibend(struct inode *inode); +struct aufs_vdir *ivdir(struct inode *inode); +struct inode *au_h_iptr_i(struct inode *inode, aufs_bindex_t bindex); +struct inode *au_h_iptr(struct inode *inode); +aufs_bindex_t itoid_index(struct inode *inode, aufs_bindex_t bindex); + +void set_ibstart(struct inode *inode, aufs_bindex_t bindex); +void set_ibend(struct inode *inode, aufs_bindex_t bindex); +void set_ivdir(struct inode *inode, struct aufs_vdir *vdir); +void aufs_hiput(struct aufs_hinode *hinode); +#define AUFS_HI_XINO 1 +#define AUFS_HI_NOTIFY 2 +unsigned int au_hi_flags(struct inode *inode, int isdir); +void set_h_iptr(struct inode *inode, aufs_bindex_t bindex, + struct inode *h_inode, unsigned int flags); +void au_update_iigen(struct inode *inode); +void au_update_brange(struct inode *inode, int do_put_zero); + +int au_iinfo_init(struct inode *inode); +void au_iinfo_fin(struct inode *inode); + +/* plink.c */ +#ifdef CONFIG_AUFS_DEBUG +void au_list_plink(struct super_block *sb); +#else +static inline void au_list_plink(struct super_block *sb) +{ + /* nothing */ +} +#endif +int au_is_plinked(struct super_block *sb, struct inode *inode); +struct dentry *lkup_plink(struct super_block *sb, aufs_bindex_t bindex, + struct inode *inode); +void append_plink(struct super_block *sb, struct inode *inode, + struct dentry *h_dentry, aufs_bindex_t bindex); +void au_put_plink(struct super_block *sb); +void half_refresh_plink(struct super_block *sb, aufs_bindex_t br_id); + +/* ---------------------------------------------------------------------- */ + +/* lock subclass for hidden inode */ +/* default MAX_LOCKDEP_SUBCLASSES(8) is not enough */ +// todo: reduce it by dcsub. +enum { + AuLsc_Begin = I_MUTEX_QUOTA, + AuLsc_HI_GPARENT, /* setattr with inotify */ + AuLsc_HI_PARENT, /* hidden inode, parent first */ + AuLsc_HI_CHILD, + AuLsc_HI_PARENT2, /* copyup dirs */ + AuLsc_HI_CHILD2, + AuLsc_End +}; + +/* simple abstraction */ +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,16) +static inline void i_lock(struct inode *i) +{ + down(&i->i_sem); +} + +static inline void i_unlock(struct inode *i) +{ + up(&i->i_sem); +} + +static inline int i_trylock(struct inode *i) +{ + return down_trylock(&i->i_sem); +} + +static inline void hi_lock(struct inode *i, unsigned int lsc) +{ + i_lock(i); +} + +#define IMustLock(i) DEBUG_ON(!down_trylock(&(i)->i_sem)) +#else +static inline void i_lock(struct inode *i) +{ + mutex_lock(&i->i_mutex); +} + +static inline void i_unlock(struct inode *i) +{ + mutex_unlock(&i->i_mutex); +} + +static inline int i_trylock(struct inode *i) +{ + return mutex_trylock(&i->i_mutex); +} + +static inline void hi_lock(struct inode *i, unsigned int lsc) +{ + mutex_lock_nested(&i->i_mutex, lsc); +} + +#define IMustLock(i) MtxMustLock(&(i)->i_mutex) +#endif + +/* + * hi_lock_gparent, hi_lock_parent, hi_lock_parent2, hi_lock_child, + * hi_lock_child2, hi_lock_whplink + */ +#define LockFunc(name, lsc) \ +static inline void hi_lock_##name(struct inode *h_i) \ +{hi_lock(h_i, AuLsc_HI_##lsc);} + +LockFunc(gparent, GPARENT); +LockFunc(parent, PARENT); +LockFunc(parent2, PARENT2); +LockFunc(child, CHILD); +LockFunc(child2, CHILD2); +LockFunc(whplink, CHILD2); /* sharing lock-subclass */ + +#undef LockFunc + +/* ---------------------------------------------------------------------- */ + +/* tiny test for inode number */ +/* tmpfs generation is too rough */ +static inline int au_test_higen(struct inode *inode, struct inode *h_inode) +{ + //IiMustAnyLock(inode); + return !(itoii(inode)->ii_hsb1 == h_inode->i_sb + && inode->i_generation == h_inode->i_generation); +} + +static inline int au_iigen(struct inode *inode) +{ + return atomic_read(&itoii(inode)->ii_generation); +} + +#ifdef CONFIG_AUFS_HINOTIFY +static inline void au_iigen_dec(struct inode *inode) +{ + //Dbg("i%lu\n", inode->i_ino); + atomic_dec(&itoii(inode)->ii_generation); +} + +/* hinotify.c */ +int alloc_hinotify(struct aufs_hinode *hinode, struct inode *inode, + struct inode *h_inode); +void do_free_hinotify(struct aufs_hinode *hinode); +void do_hdir_lock(struct inode *h_dir, struct inode *dir, aufs_bindex_t bindex, + unsigned int lsc); +void hdir_unlock(struct inode *h_dir, struct inode *dir, aufs_bindex_t bindex); +void hdir_lock_rename(struct dentry **h_parents, struct inode **dirs, + aufs_bindex_t bindex, int issamedir); +void hdir_unlock_rename(struct dentry **h_parents, struct inode **dirs, + aufs_bindex_t bindex, int issamedir); +void au_reset_hinotify(struct inode *inode, unsigned int flags); +int __init au_inotify_init(void); +void au_inotify_fin(void); +#else +static inline +int alloc_hinotify(struct aufs_hinode *hinode, struct inode *inode, + struct inode *h_inode) +{ + return -EOPNOTSUPP; +} + +static inline void do_free_hinotify(struct aufs_hinode *hinode) +{ + /* nothing */ +} + +static inline +void do_hdir_lock(struct inode *h_dir, struct inode *dir, aufs_bindex_t bindex, + unsigned int lsc) +{ + hi_lock(h_dir, lsc); +} + +static inline +void hdir_unlock(struct inode *h_dir, struct inode *dir, aufs_bindex_t bindex) +{ + i_unlock(h_dir); +} + +static inline +void hdir_lock_rename(struct dentry **h_parents, struct inode **dirs, + aufs_bindex_t bindex, int issamedir) +{ + vfsub_lock_rename(h_parents[0], h_parents[1]); +} + +static inline +void hdir_unlock_rename(struct dentry **h_parents, struct inode **dirs, + aufs_bindex_t bindex, int issamedir) +{ + vfsub_unlock_rename(h_parents[0], h_parents[1]); +} + +static inline void au_reset_hinotify(struct inode *inode, unsigned int flags) +{ + /* nothing */ +} + +#define au_inotify_init() 0 +#define au_inotify_fin() /* */ +#endif /* CONFIG_AUFS_HINOTIFY */ + +static inline void free_hinotify(struct inode *inode, aufs_bindex_t bindex) +{ + do_free_hinotify(itoii(inode)->ii_hinode + bindex); +} + +/* + * hgdir_lock, hdir_lock, hdir2_lock + */ +#define LockFunc(name, lsc) \ +static inline \ +void name##_lock(struct inode *h_dir, struct inode *dir, aufs_bindex_t bindex) \ +{do_hdir_lock(h_dir, dir, bindex, AuLsc_HI_##lsc);} + +LockFunc(hgdir, GPARENT); +LockFunc(hdir, PARENT); +LockFunc(hdir2, PARENT2); + +#undef LockFunc + +/* ---------------------------------------------------------------------- */ + +/* lock subclass for iinfo */ +enum { + AuLsc_II_CHILD, /* child first */ + AuLsc_II_CHILD2, /* rename(2), link(2), and cpup at hinotify */ + AuLsc_II_CHILD3, /* copyup dirs */ + AuLsc_II_PARENT, + AuLsc_II_PARENT2, + AuLsc_II_PARENT3, + AuLsc_II_NEW /* new inode */ +}; + +/* + * ii_read_lock_child, ii_write_lock_child, + * ii_read_lock_child2, ii_write_lock_child2, + * ii_read_lock_child3, ii_write_lock_child3, + * ii_read_lock_parent, ii_write_lock_parent, + * ii_read_lock_parent2, ii_write_lock_parent2, + * ii_read_lock_parent3, ii_write_lock_parent3, + * ii_read_lock_new, ii_write_lock_new + */ +#define ReadLockFunc(name, lsc) \ +static inline void ii_read_lock_##name(struct inode *i) \ +{rw_read_lock_nested(&itoii(i)->ii_rwsem, AuLsc_II_##lsc);} + +#define WriteLockFunc(name, lsc) \ +static inline void ii_write_lock_##name(struct inode *i) \ +{rw_write_lock_nested(&itoii(i)->ii_rwsem, AuLsc_II_##lsc);} + +#define RWLockFuncs(name, lsc) \ + ReadLockFunc(name, lsc); \ + WriteLockFunc(name, lsc) + +RWLockFuncs(child, CHILD); +RWLockFuncs(child2, CHILD2); +RWLockFuncs(child3, CHILD3); +RWLockFuncs(parent, PARENT); +RWLockFuncs(parent2, PARENT2); +RWLockFuncs(parent3, PARENT3); +RWLockFuncs(new, NEW); + +#undef ReadLockFunc +#undef WriteLockFunc +#undef RWLockFunc + +/* + * ii_read_unlock, ii_write_unlock, ii_downgrade_lock + */ +SimpleUnlockRwsemFuncs(ii, struct inode *i, itoii(i)->ii_rwsem); + +/* to debug easier, do not make them inlined functions */ +#define IiMustReadLock(i) do { \ + SiMustAnyLock((i)->i_sb); \ + RwMustReadLock(&itoii(i)->ii_rwsem); \ +} while (0) + +#define IiMustWriteLock(i) do { \ + SiMustAnyLock((i)->i_sb); \ + RwMustWriteLock(&itoii(i)->ii_rwsem); \ +} while (0) + +#define IiMustAnyLock(i) do { \ + SiMustAnyLock((i)->i_sb); \ + RwMustAnyLock(&itoii(i)->ii_rwsem); \ +} while (0) + +#define IiMustNoWaiters(i) RwMustNoWaiters(&itoii(i)->ii_rwsem) + +#endif /* __KERNEL__ */ +#endif /* __AUFS_INODE_H__ */ diff --git a/fs/aufs/misc.c b/fs/aufs/misc.c new file mode 100755 index 0000000..32e0549 --- /dev/null +++ b/fs/aufs/misc.c @@ -0,0 +1,228 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: misc.c,v 1.31 2007/05/14 03:41:52 sfjro Exp $ */ + +//#include +//#include +//#include +//#include +#include "aufs.h" + +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp) +{ + void *q; + + LKTRTrace("p %p, nused %d, sz %d, ksize %d\n", + p, nused, new_sz, ksize(p)); + DEBUG_ON(new_sz <= 0); + if (new_sz <= nused) + return p; + if (new_sz <= ksize(p)) { + memset(p + nused, 0, new_sz - nused); + return p; + } + + q = kmalloc(new_sz, gfp); + //q = NULL; + if (unlikely(!q)) + return NULL; + memcpy(q, p, nused); + memset(q + nused, 0, new_sz - nused); + //smp_mb(); + kfree(p); + return q; +} + +/* ---------------------------------------------------------------------- */ + +// todo: make it inline +struct nameidata *fake_dm(struct nameidata *fake_nd, struct nameidata *nd, + struct super_block *sb, aufs_bindex_t bindex) +{ + LKTRTrace("nd %p, b%d\n", nd, bindex); + + if (!nd) + return NULL; + + fake_nd->dentry = NULL; + fake_nd->mnt = NULL; + +#ifndef CONFIG_AUFS_FAKE_DM + DiMustAnyLock(nd->dentry); + + if (bindex <= dbend(nd->dentry)) + fake_nd->dentry = au_h_dptr_i(nd->dentry, bindex); + if (fake_nd->dentry) { + dget(fake_nd->dentry); + fake_nd->mnt = sbr_mnt(sb, bindex); + DEBUG_ON(!fake_nd->mnt); + mntget(fake_nd->mnt); + } else + fake_nd = ERR_PTR(-ENOENT); +#endif + + TraceErrPtr(fake_nd); + return fake_nd; +} + +void fake_dm_release(struct nameidata *fake_nd) +{ +#ifndef CONFIG_AUFS_FAKE_DM + if (fake_nd) { + mntput(fake_nd->mnt); + dput(fake_nd->dentry); + } +#endif +} + +/* ---------------------------------------------------------------------- */ + +int au_copy_file(struct file *dst, struct file *src, loff_t len, + struct super_block *sb, int *sparse) +{ + int err, all_zero, dlgt; + unsigned long blksize; + char *buf; + /* reduce stack space */ + struct iattr *ia; + + LKTRTrace("%.*s, %.*s\n", + DLNPair(dst->f_dentry), DLNPair(src->f_dentry)); + DEBUG_ON(!(dst->f_mode & FMODE_WRITE)); + IMustLock(dst->f_dentry->d_parent->d_inode); + + err = -ENOMEM; + blksize = dst->f_dentry->d_sb->s_blocksize; + if (!blksize || PAGE_SIZE < blksize) + blksize = PAGE_SIZE; + LKTRTrace("blksize %lu\n", blksize); + buf = kmalloc(blksize, GFP_KERNEL); + //buf = NULL; + if (unlikely(!buf)) + goto out; + ia = kmalloc(sizeof(*ia), GFP_KERNEL); + if (unlikely(!ia)) + goto out_buf; + + dlgt = need_dlgt(sb); + err = all_zero = 0; + dst->f_pos = src->f_pos = 0; + while (len) { + size_t sz, rbytes, wbytes, i; + char *p; + + LKTRTrace("len %lld\n", len); + sz = blksize; + if (len < blksize) + sz = len; + + /* support LSM and notify */ + rbytes = 0; + while (!rbytes || err == -EAGAIN || err == -EINTR) + err = rbytes = vfsub_read_k(src, buf, sz, &src->f_pos, + dlgt); + if (unlikely(err < 0)) + break; + + all_zero = 0; + if (len >= rbytes && rbytes == blksize) { + all_zero = 1; + p = buf; + for (i = 0; all_zero && i < rbytes; i++) + all_zero = !*p++; + } + if (!all_zero) { + wbytes = rbytes; + p = buf; + while (wbytes) { + size_t b; + /* support LSM and notify */ + err = b = vfsub_write_k(dst, p, wbytes, + &dst->f_pos, dlgt); + if (unlikely(err == -EAGAIN || err == -EINTR)) + continue; + if (unlikely(err < 0)) + break; + wbytes -= b; + p += b; + } + } else { + loff_t res; + LKTRLabel(hole); + *sparse = 1; + err = res = vfsub_llseek(dst, rbytes, SEEK_CUR); + if (unlikely(res < 0)) + break; + } + len -= rbytes; + err = 0; + } + + /* the last block may be a hole */ + if (unlikely(!err && all_zero)) { + struct dentry *h_d = dst->f_dentry; + struct inode *h_i = h_d->d_inode; + + LKTRLabel(last hole); + do { + err = vfsub_write_k(dst, "\0", 1, &dst->f_pos, dlgt); + } while (err == -EAGAIN || err == -EINTR); + if (err == 1) { + ia->ia_size = dst->f_pos; + ia->ia_valid = ATTR_SIZE | ATTR_FILE; + ia->ia_file = dst; + hi_lock_child2(h_i); + err = vfsub_notify_change(h_d, ia, dlgt); + i_unlock(h_i); + } + } + + kfree(ia); + out_buf: + kfree(buf); + out: + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +int test_ro(struct super_block *sb, aufs_bindex_t bindex, struct inode *inode) +{ + int err; + + err = br_rdonly(stobr(sb, bindex)); + if (!err && inode) { + struct inode *hi = au_h_iptr_i(inode, bindex); + if (hi) + err = IS_IMMUTABLE(hi) ? -EROFS : 0; + } + return err; +} + +int au_test_perm(struct inode *hidden_inode, int mask, int dlgt) +{ + if (!current->fsuid) + return 0; + if (unlikely(au_is_nfs(hidden_inode->i_sb) + && (mask & MAY_WRITE) + && S_ISDIR(hidden_inode->i_mode))) + mask |= MAY_READ; /* force permission check */ + return vfsub_permission(hidden_inode, mask, NULL, dlgt); +} diff --git a/fs/aufs/misc.h b/fs/aufs/misc.h new file mode 100755 index 0000000..fea4a2c --- /dev/null +++ b/fs/aufs/misc.h @@ -0,0 +1,187 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: misc.h,v 1.25 2007/05/14 03:41:52 sfjro Exp $ */ + +#ifndef __AUFS_MISC_H__ +#define __AUFS_MISC_H__ + +#ifdef __KERNEL__ + +#include +#include +#include +#include +#include + +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,18) +#define I_MUTEX_QUOTA 0 +#define lockdep_off() /* */ +#define lockdep_on() /* */ +#define mutex_lock_nested(mtx, lsc) mutex_lock(mtx) +#define down_write_nested(rw, lsc) down_write(rw) +#define down_read_nested(rw, lsc) down_read(rw) +#endif + +/* ---------------------------------------------------------------------- */ + +struct aufs_rwsem { + struct rw_semaphore rwsem; +#ifdef CONFIG_AUFS_DEBUG + atomic_t rcnt; +#endif +}; + +#ifdef CONFIG_AUFS_DEBUG +#define DbgRcntInit(rw) atomic_set(&(rw)->rcnt, 0) +#define DbgRcntInc(rw) atomic_inc(&(rw)->rcnt) +#define DbgRcntDec(rw) WARN_ON(atomic_dec_return(&(rw)->rcnt) < 0) +#else +#define DbgRcntInit(rw) /* */ +#define DbgRcntInc(rw) /* */ +#define DbgRcntDec(rw) /* */ +#endif + +static inline void rw_init_nolock(struct aufs_rwsem *rw) +{ + DbgRcntInit(rw); + init_rwsem(&rw->rwsem); +} + +static inline void rw_init_wlock(struct aufs_rwsem *rw) +{ + rw_init_nolock(rw); + down_write(&rw->rwsem); +} + +static inline void rw_init_wlock_nested(struct aufs_rwsem *rw, unsigned int lsc) +{ + rw_init_nolock(rw); + down_write_nested(&rw->rwsem, lsc); +} + +static inline void rw_read_lock(struct aufs_rwsem *rw) +{ + down_read(&rw->rwsem); + DbgRcntInc(rw); +} + +static inline void rw_read_lock_nested(struct aufs_rwsem *rw, unsigned int lsc) +{ + down_read_nested(&rw->rwsem, lsc); + DbgRcntInc(rw); +} + +static inline void rw_read_unlock(struct aufs_rwsem *rw) +{ + DbgRcntDec(rw); + up_read(&rw->rwsem); +} + +static inline void rw_dgrade_lock(struct aufs_rwsem *rw) +{ + DbgRcntInc(rw); + downgrade_write(&rw->rwsem); +} + +static inline void rw_write_lock(struct aufs_rwsem *rw) +{ + down_write(&rw->rwsem); +} + +static inline void rw_write_lock_nested(struct aufs_rwsem *rw, unsigned int lsc) +{ + down_write_nested(&rw->rwsem, lsc); +} + +static inline void rw_write_unlock(struct aufs_rwsem *rw) +{ + up_write(&rw->rwsem); +} + +#if 0 // why is not _nested version defined +static inline int rw_read_trylock(struct aufs_rwsem *rw) +{ + int ret = down_read_trylock(&rw->rwsem); + if (ret) + DbgRcntInc(rw); + return ret; +} + +static inline int rw_write_trylock(struct aufs_rwsem *rw) +{ + return down_write_trylock(&rw->rwsem); +} +#endif + +#undef DbgRcntInit +#undef DbgRcntInc +#undef DbgRcntDec + +/* to debug easier, do not make them inlined functions */ +#define RwMustNoWaiters(rw) DEBUG_ON(!list_empty(&(rw)->rwsem.wait_list)) +#define RwMustAnyLock(rw) DEBUG_ON(down_write_trylock(&(rw)->rwsem)) +#ifdef CONFIG_AUFS_DEBUG +#define RwMustReadLock(rw) do { \ + RwMustAnyLock(rw); \ + DEBUG_ON(!atomic_read(&(rw)->rcnt)); \ +} while (0) +#define RwMustWriteLock(rw) do { \ + RwMustAnyLock(rw); \ + DEBUG_ON(atomic_read(&(rw)->rcnt)); \ +} while (0) +#else +#define RwMustReadLock(rw) RwMustAnyLock(rw) +#define RwMustWriteLock(rw) RwMustAnyLock(rw) +#endif + +#define SimpleLockRwsemFuncs(prefix, param, rwsem) \ +static inline void prefix##_read_lock(param) {rw_read_lock(&(rwsem));} \ +static inline void prefix##_write_lock(param) {rw_write_lock(&(rwsem));} +//static inline void prefix##_read_trylock(param) {rw_read_trylock(&(rwsem));} +//static inline void prefix##_write_trylock(param) {rw_write_trylock(&(rwsem));} +//static inline void prefix##_read_trylock_nested(param, lsc) +//{rw_read_trylock_nested(&(rwsem, lsc));} +//static inline void prefix##_write_trylock_nestd(param, lsc) +//{rw_write_trylock_nested(&(rwsem), nested);} + +#define SimpleUnlockRwsemFuncs(prefix, param, rwsem) \ +static inline void prefix##_read_unlock(param) {rw_read_unlock(&(rwsem));} \ +static inline void prefix##_write_unlock(param) {rw_write_unlock(&(rwsem));} \ +static inline void prefix##_downgrade_lock(param) {rw_dgrade_lock(&(rwsem));} + +#define SimpleRwsemFuncs(prefix, param, rwsem) \ + SimpleLockRwsemFuncs(prefix, param, rwsem); \ + SimpleUnlockRwsemFuncs(prefix, param, rwsem) + +/* ---------------------------------------------------------------------- */ + +typedef ssize_t (*readf_t)(struct file*, char __user*, size_t, loff_t*); +typedef ssize_t (*writef_t)(struct file*, const char __user*, size_t, loff_t*); + +void *au_kzrealloc(void *p, unsigned int nused, unsigned int new_sz, gfp_t gfp); +struct nameidata *fake_dm(struct nameidata *fake_nd, struct nameidata *nd, + struct super_block *sb, aufs_bindex_t bindex); +void fake_dm_release(struct nameidata *fake_nd); +int au_copy_file(struct file *dst, struct file *src, loff_t len, + struct super_block *sb, int *sparse); +int test_ro(struct super_block *sb, aufs_bindex_t bindex, struct inode *inode); +int au_test_perm(struct inode *h_inode, int mask, int dlgt); + +#endif /* __KERNEL__ */ +#endif /* __AUFS_MISC_H__ */ diff --git a/fs/aufs/module.c b/fs/aufs/module.c new file mode 100755 index 0000000..06c563e --- /dev/null +++ b/fs/aufs/module.c @@ -0,0 +1,334 @@ +/* + * Copyright (C) 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: module.c,v 1.9 2007/04/30 05:46:32 sfjro Exp $ */ + +//#include +//#include +#include +//#include +//#include +#include "aufs.h" + +/* ---------------------------------------------------------------------- */ + +/* + * aufs caches + */ +struct kmem_cache *aufs_cachep[AuCache_Last]; +static int __init create_cache(void) +{ +#define Cache(type) kmem_cache_create(#type, sizeof(struct type), 0, \ + SLAB_RECLAIM_ACCOUNT, NULL, NULL) + + if ((aufs_cachep[AuCache_DINFO] = Cache(aufs_dinfo)) + && (aufs_cachep[AuCache_ICNTNR] = Cache(aufs_icntnr)) + && (aufs_cachep[AuCache_FINFO] = Cache(aufs_finfo)) + //&& (aufs_cachep[AuCache_FINFO] = NULL) + && (aufs_cachep[AuCache_VDIR] = Cache(aufs_vdir)) + && (aufs_cachep[AuCache_DEHSTR] = Cache(aufs_dehstr)) + && (aufs_cachep[AuCache_HINOTIFY] = Cache(aufs_hinotify))) + return 0; + return -ENOMEM; + +#undef Cache +} + +static void destroy_cache(void) +{ + int i; + for (i = 0; i < AuCache_Last; i++) + if (aufs_cachep[i]) + kmem_cache_destroy(aufs_cachep[i]); +} + +/* ---------------------------------------------------------------------- */ + +char au_esc_chars[0x20 + 3]; /* 0x01-0x20, backslash, del, and NULL */ +int au_dir_roflags; +extern struct file_system_type aufs_fs_type; + +#ifdef DbgDlgt +#include +#include "dbg_dlgt.c" +#else +#define dbg_dlgt_init() 0 +#define dbg_dlgt_fin() /* */ +#endif + +/* + * functions for module interface. + */ +MODULE_LICENSE("GPL"); +MODULE_AUTHOR("Junjiro Okajima"); +MODULE_DESCRIPTION(AUFS_NAME " -- Another unionfs"); +MODULE_VERSION(AUFS_VERSION); + +/* it should be 'byte', but param_set_byte() prints by "%c" */ +short aufs_nwkq = AUFS_NWKQ_DEF; +MODULE_PARM_DESC(nwkq, "the number of workqueue thread, " AUFS_WKQ_NAME); +module_param_named(nwkq, aufs_nwkq, short, 0444); + +int sysaufs_brs = 0; +MODULE_PARM_DESC(brs, "use /fs/aufs/brs"); +module_param_named(brs, sysaufs_brs, int, 0444); + +static int __init aufs_init(void) +{ + int err, i; + char *p; + + //sbinfo->si_xino is atomic_long_t + BUILD_BUG_ON(sizeof(ino_t) != sizeof(long)); + +#ifdef CONFIG_AUFS_DEBUG + { + struct aufs_destr destr; + destr.len = -1; + DEBUG_ON(destr.len < NAME_MAX); + } + +#ifdef CONFIG_4KSTACKS + printk("CONFIG_4KSTACKS is defined.\n"); +#endif +#if 0 // verbose debug + { + union { + struct aufs_branch *br; + struct aufs_dinfo *di; + struct aufs_finfo *fi; + struct aufs_iinfo *ii; + struct aufs_hinode *hi; + struct aufs_sbinfo *si; + struct aufs_destr *destr; + struct aufs_de *de; + struct aufs_wh *wh; + struct aufs_vdir *vd; + } u; + + printk("br{" + "xino %d, readf %d, writef %d, " + "id %d, perm %d, mnt %d, count %d, " + "wh_sem %d, wh %d, run %d} %d\n", + offsetof(typeof(*u.br), br_xino), + offsetof(typeof(*u.br), br_xino_read), + offsetof(typeof(*u.br), br_xino_write), + offsetof(typeof(*u.br), br_id), + offsetof(typeof(*u.br), br_perm), + offsetof(typeof(*u.br), br_mnt), + offsetof(typeof(*u.br), br_count), + offsetof(typeof(*u.br), br_wh_rwsem), + offsetof(typeof(*u.br), br_wh), + offsetof(typeof(*u.br), br_wh_running), + sizeof(*u.br)); + printk("di{gen %d, rwsem %d, bstart %d, bend %d, bwh %d, " + "bdiropq %d, hdentry %d, reval %d} %d\n", + offsetof(typeof(*u.di), di_generation), + offsetof(typeof(*u.di), di_rwsem), + offsetof(typeof(*u.di), di_bstart), + offsetof(typeof(*u.di), di_bend), + offsetof(typeof(*u.di), di_bwh), + offsetof(typeof(*u.di), di_bdiropq), + offsetof(typeof(*u.di), di_hdentry), + offsetof(typeof(*u.di), di_reval), + sizeof(*u.di)); + printk("fi{gen %d, rwsem %d, hfile %d, bstart %d, bend %d, " + "h_vm_ops %d, vdir_cach %d} %d\n", + offsetof(typeof(*u.fi), fi_generation), + offsetof(typeof(*u.fi), fi_rwsem), + offsetof(typeof(*u.fi), fi_hfile), + offsetof(typeof(*u.fi), fi_bstart), + offsetof(typeof(*u.fi), fi_bend), + offsetof(typeof(*u.fi), fi_h_vm_ops), + offsetof(typeof(*u.fi), fi_vdir_cache), + sizeof(*u.fi)); + printk("ii{rwsem %d, bstart %d, bend %d, hinode %d, vdir %d} " + "%d\n", + offsetof(typeof(*u.ii), ii_rwsem), + offsetof(typeof(*u.ii), ii_bstart), + offsetof(typeof(*u.ii), ii_bend), + offsetof(typeof(*u.ii), ii_hinode), + offsetof(typeof(*u.ii), ii_vdir), + sizeof(*u.ii)); + printk("hi{inode %d, id %d, notify %d} %d\n", + offsetof(typeof(*u.hi), hi_inode), + offsetof(typeof(*u.hi), hi_id), + offsetof(typeof(*u.hi), hi_notify), + sizeof(*u.hi)); + printk("si{rwsem %d, gen %d, " + "failed_refresh %d, " + "bend %d, last id %d, br %d, " + "flags %d, " + "xino %d, " + "rdcache %d, " + "dirwh %d, " + "pl_lock %d, pl %d, " + "kobj %d} %d\n", + offsetof(typeof(*u.si), si_rwsem), + offsetof(typeof(*u.si), si_generation), + -1,//offsetof(typeof(*u.si), si_failed_refresh_dirs), + offsetof(typeof(*u.si), si_bend), + offsetof(typeof(*u.si), si_last_br_id), + offsetof(typeof(*u.si), si_branch), + offsetof(typeof(*u.si), si_flags), + offsetof(typeof(*u.si), si_xino), + offsetof(typeof(*u.si), si_rdcache), + offsetof(typeof(*u.si), si_dirwh), + offsetof(typeof(*u.si), si_plink_lock), + offsetof(typeof(*u.si), si_plink), + offsetof(typeof(*u.si), si_kobj), + sizeof(*u.si)); + printk("destr{len %d, name %d} %d\n", + offsetof(typeof(*u.destr), len), + offsetof(typeof(*u.destr), name), + sizeof(*u.destr)); + printk("de{ino %d, type %d, str %d} %d\n", + offsetof(typeof(*u.de), de_ino), + offsetof(typeof(*u.de), de_type), + offsetof(typeof(*u.de), de_str), + sizeof(*u.de)); + printk("wh{hash %d, bindex %d, str %d} %d\n", + offsetof(typeof(*u.wh), wh_hash), + offsetof(typeof(*u.wh), wh_bindex), + offsetof(typeof(*u.wh), wh_str), + sizeof(*u.wh)); + printk("vd{deblk %d, nblk %d, last %d, ver %d, jiffy %d} %d\n", + offsetof(typeof(*u.vd), vd_deblk), + offsetof(typeof(*u.vd), vd_nblk), + offsetof(typeof(*u.vd), vd_last), + offsetof(typeof(*u.vd), vd_version), + offsetof(typeof(*u.vd), vd_jiffy), + sizeof(*u.vd)); + } +#endif +#endif + + p = au_esc_chars; + for (i = 1; i <= ' '; i++) + *p++ = i; + *p++ = '\\'; + *p++ = '\x7f'; + *p = 0; + + au_dir_roflags = au_file_roflags(O_DIRECTORY | O_LARGEFILE); +#ifndef CONFIG_AUFS_SYSAUFS + sysaufs_brs = 0; +#endif + + err = -EINVAL; + if (unlikely(aufs_nwkq <= 0)) + goto out; + err = create_cache(); + if (unlikely(err)) + goto out; + err = sysaufs_init(); + if (unlikely(err)) + goto out_cache; + err = au_wkq_init(); + if (unlikely(err)) + goto out_kobj; + err = au_inotify_init(); + if (unlikely(err)) + goto out_wkq; + err = dbg_dlgt_init(); + if (unlikely(err)) + goto out_inotify; + err = register_filesystem(&aufs_fs_type); + if (unlikely(err)) + goto out_dlgt; + printk(AUFS_NAME " " AUFS_VERSION "\n"); + return 0; /* success */ + + out_dlgt: + dbg_dlgt_fin(); + out_inotify: + au_inotify_fin(); + out_wkq: + au_wkq_fin(); + out_kobj: + sysaufs_fin(); + out_cache: + destroy_cache(); + out: + TraceErr(err); + return err; +} + +static void __exit aufs_exit(void) +{ + unregister_filesystem(&aufs_fs_type); + dbg_dlgt_fin(); + au_inotify_fin(); + au_wkq_fin(); + sysaufs_fin(); + destroy_cache(); +} + +module_init(aufs_init); +module_exit(aufs_exit); + +/* ---------------------------------------------------------------------- */ + +// fake Kconfig +#if 1 +#ifdef CONFIG_AUFS_HINOTIFY +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,18) +#error CONFIG_AUFS_HINOTIFY is supported in linux-2.6.18 and later. +#endif +#ifndef CONFIG_INOTIFY +#error enable CONFIG_INOTIFY to use CONFIG_AUFS_HINOTIFY. +#endif +#endif + +#if AUFS_BRANCH_MAX > 511 && BITS_PER_LONG == 64 && PAGE_SIZE == 4096 +#warning For 4k pagesize and 64bit environment, \ + CONFIG_AUFS_BRANCH_MAX_511 or smaller is recommended. +#endif + +#ifdef CONFIG_AUFS_SYSAUFS +#ifndef CONFIG_SYSFS +#error CONFIG_AUFS_SYSAUFS requires CONFIG_SYSFS. +#endif +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,18) +#error CONFIG_AUFS_SYSAUFS requires linux-2.6.18 and later. +#endif +#endif + +#ifdef CONFIG_AUFS_EXPORT +#if !defined(CONFIG_EXPORTFS) && !defined(CONFIG_EXPORTFS_MODULE) +#error CONFIG_AUFS_EXPORT requires CONFIG_EXPORTFS +#endif +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,18) +#error CONFIG_AUFS_EXPORT requires linux-2.6.18 and later. +#endif +#if defined(CONFIG_EXPORTFS_MODULE) && defined(CONFIG_AUFS) +#error need CONFIG_EXPORTFS=y to link aufs statically with CONFIG_AUFS_EXPORT +#endif +#endif + +#ifdef CONFIG_DEBUG_PROVE_LOCKING +#if MAX_LOCKDEP_SUBCLASSES < AuLsc_End +#warning lockdep will not work since aufs uses deeper locks. +#endif +#endif + +#ifdef CONFIG_AUFS_COMPAT +#warning CONFIG_AUFS_COMPAT will be removed in the near future. +#endif + +#endif diff --git a/fs/aufs/module.h b/fs/aufs/module.h new file mode 100755 index 0000000..3769861 --- /dev/null +++ b/fs/aufs/module.h @@ -0,0 +1,60 @@ +/* + * Copyright (C) 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: module.h,v 1.8 2007/05/14 03:41:52 sfjro Exp $ */ + +#ifndef __AUFS_MODULE_H__ +#define __AUFS_MODULE_H__ + +#ifdef __KERNEL__ + +#include + +/* ---------------------------------------------------------------------- */ + +/* module parameters */ +extern short aufs_nwkq; +extern int sysaufs_brs; + +/* ---------------------------------------------------------------------- */ + +extern char au_esc_chars[]; +extern int au_dir_roflags; + +/* kmem cache */ +enum {AuCache_DINFO, AuCache_ICNTNR, AuCache_FINFO, AuCache_VDIR, + AuCache_DEHSTR, AuCache_HINOTIFY, AuCache_Last}; +extern struct kmem_cache *aufs_cachep[]; + +#define CacheFuncs(name, index) \ +static inline void *cache_alloc_##name(void) \ +{return kmem_cache_alloc(aufs_cachep[index], GFP_KERNEL);} \ +static inline void cache_free_##name(void *p) \ +{kmem_cache_free(aufs_cachep[index], p);} + +CacheFuncs(dinfo, AuCache_DINFO); +CacheFuncs(icntnr, AuCache_ICNTNR); +CacheFuncs(finfo, AuCache_FINFO); +CacheFuncs(vdir, AuCache_VDIR); +CacheFuncs(dehstr, AuCache_DEHSTR); +CacheFuncs(hinotify, AuCache_HINOTIFY); + +#undef CacheFuncs + +#endif /* __KERNEL__ */ +#endif /* __AUFS_MODULE_H__ */ diff --git a/fs/aufs/opts.c b/fs/aufs/opts.c new file mode 100755 index 0000000..c1a9445 --- /dev/null +++ b/fs/aufs/opts.c @@ -0,0 +1,1043 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: opts.c,v 1.34 2007/05/14 03:40:27 sfjro Exp $ */ + +#include // a distribution requires +#include +#include "aufs.h" + +enum { + Opt_br, + Opt_add, Opt_del, Opt_mod, Opt_append, Opt_prepend, + Opt_idel, Opt_imod, + Opt_dirwh, Opt_rdcache, Opt_deblk, Opt_nhash, + Opt_xino, Opt_zxino, Opt_noxino, + Opt_plink, Opt_noplink, Opt_list_plink, Opt_clean_plink, + Opt_udba, + Opt_diropq_a, Opt_diropq_w, + Opt_warn_perm, Opt_nowarn_perm, + Opt_findrw_dir, Opt_findrw_br, + Opt_coo, + Opt_dlgt, Opt_nodlgt, + Opt_tail, Opt_ignore, Opt_err +}; + +static match_table_t options = { + {Opt_br, "br=%s"}, + {Opt_br, "br:%s"}, + + {Opt_add, "add=%d:%s"}, + {Opt_add, "add:%d:%s"}, + {Opt_add, "ins=%d:%s"}, + {Opt_add, "ins:%d:%s"}, + {Opt_append, "append=%s"}, + {Opt_append, "append:%s"}, + {Opt_prepend, "prepend=%s"}, + {Opt_prepend, "prepend:%s"}, + + {Opt_del, "del=%s"}, + {Opt_del, "del:%s"}, + //{Opt_idel, "idel:%d"}, + {Opt_mod, "mod=%s"}, + {Opt_mod, "mod:%s"}, + //{Opt_imod, "imod:%d:%s"}, + + {Opt_dirwh, "dirwh=%d"}, + {Opt_dirwh, "dirwh:%d"}, + + {Opt_xino, "xino=%s"}, + {Opt_xino, "xino:%s"}, + {Opt_noxino, "noxino"}, + //{Opt_zxino, "zxino=%s"}, + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) + {Opt_plink, "plink"}, + {Opt_noplink, "noplink"}, +#ifdef CONFIG_AUFS_DEBUG + {Opt_list_plink, "list_plink"}, +#endif + {Opt_clean_plink, "clean_plink"}, +#endif + + {Opt_udba, "udba=%s"}, + + {Opt_diropq_a, "diropq=always"}, + {Opt_diropq_a, "diropq=a"}, + {Opt_diropq_w, "diropq=whiteouted"}, + {Opt_diropq_w, "diropq=w"}, + + {Opt_warn_perm, "warn_perm"}, + {Opt_nowarn_perm, "nowarn_perm"}, + +#ifdef CONFIG_AUFS_DLGT + {Opt_dlgt, "dlgt"}, + {Opt_nodlgt, "nodlgt"}, +#endif + + {Opt_rdcache, "rdcache=%d"}, + {Opt_rdcache, "rdcache:%d"}, +#if 0 + {Opt_findrw_dir, "findrw=dir"}, + {Opt_findrw_br, "findrw=br"}, + + {Opt_coo, "coo=%s"}, + + {Opt_deblk, "deblk=%d"}, + {Opt_deblk, "deblk:%d"}, + {Opt_nhash, "nhash=%d"}, + {Opt_nhash, "nhash:%d"}, +#endif + + {Opt_br, "dirs=%s"}, + {Opt_ignore, "debug=%d"}, + {Opt_ignore, "delete=whiteout"}, + {Opt_ignore, "delete=all"}, + {Opt_ignore, "imap=%s"}, + + {Opt_err, NULL} +}; + +/* ---------------------------------------------------------------------- */ + +#define RW "rw" +#define RO "ro" +#define WH "wh" +#define RR "rr" +#define NoLinkWH "nolwh" + +static match_table_t brperms = { + {AuBr_RR, RR}, + {AuBr_RO, RO}, + {AuBr_RW, RW}, + + {AuBr_RRWH, RR "+" WH}, + {AuBr_ROWH, RO "+" WH}, + {AuBr_RWNoLinkWH, RW "+" NoLinkWH}, + + {AuBr_ROWH, "nfsro"}, + {AuBr_RO, NULL} +}; + +static int br_perm_val(char *perm) +{ + int val; + substring_t args[MAX_OPT_ARGS]; + + DEBUG_ON(!perm || !*perm); + LKTRTrace("perm %s\n", perm); + val = match_token(perm, brperms, args); + TraceErr(val); + return val; +} + +int br_perm_str(char *p, unsigned int len, int brperm) +{ + struct match_token *bp = brperms; + + LKTRTrace("len %d, 0x%x\n", len, brperm); + + while (bp->pattern) { + if (bp->token == brperm) { + if (strlen(bp->pattern) < len) { + strcpy(p, bp->pattern); + return 0; + } else + return -E2BIG; + } + bp++; + } + + return -EIO; +} + +/* ---------------------------------------------------------------------- */ + +static match_table_t udbalevel = { + {AuFlag_UDBA_REVAL, "reval"}, +#ifdef CONFIG_AUFS_HINOTIFY + {AuFlag_UDBA_INOTIFY, "inotify"}, +#endif + {AuFlag_UDBA_NONE, "none"}, + {-1, NULL} +}; + +static int udba_val(char *str) +{ + substring_t args[MAX_OPT_ARGS]; + return match_token(str, udbalevel, args); +} + +au_parser_pattern_t udba_str(int udba) +{ + struct match_token *p = udbalevel; + while (p->pattern) { + if (p->token == udba) + return p->pattern; + p++; + } + BUG(); + return "??"; +} + +void udba_set(struct super_block *sb, unsigned int flg) +{ + au_flag_clr(sb, AuMask_UDBA); + au_flag_set(sb, flg); +} + +/* ---------------------------------------------------------------------- */ + +static match_table_t coolevel = { + {AuFlag_COO_LEAF, "leaf"}, + {AuFlag_COO_ALL, "all"}, + {AuFlag_COO_NONE, "none"}, + {-1, NULL} +}; + +#if 0 +static int coo_val(char *str) +{ + substring_t args[MAX_OPT_ARGS]; + return match_token(str, coolevel, args); +} +#endif + +au_parser_pattern_t coo_str(int coo) +{ + struct match_token *p = coolevel; + while (p->pattern) { + if (p->token == coo) + return p->pattern; + p++; + } + BUG(); + return "??"; +} +static void coo_set(struct super_block *sb, unsigned int flg) +{ + au_flag_clr(sb, AuMask_COO); + au_flag_set(sb, flg); +} + +/* ---------------------------------------------------------------------- */ + +static const int lkup_dirflags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; + +#ifdef CONFIG_AUFS_DEBUG +static void dump_opts(struct opts *opts) +{ + /* reduce stack space */ + union { + struct opt_add *add; + struct opt_del *del; + struct opt_mod *mod; + struct opt_xino *xino; + } u; + struct opt *opt; + + TraceEnter(); + + opt = opts->opt; + while (/* opt < opts_tail && */ opt->type != Opt_tail) { + switch (opt->type) { + case Opt_add: + u.add = &opt->add; + LKTRTrace("add {b%d, %s, 0x%x, %p}\n", + u.add->bindex, u.add->path, u.add->perm, + u.add->nd.dentry); + break; + case Opt_del: + case Opt_idel: + u.del = &opt->del; + LKTRTrace("del {%s, %p}\n", u.del->path, u.del->h_root); + break; + case Opt_mod: + case Opt_imod: + u.mod = &opt->mod; + LKTRTrace("mod {%s, 0x%x, %p}\n", + u.mod->path, u.mod->perm, u.mod->h_root); + break; + case Opt_append: + u.add = &opt->add; + LKTRTrace("append {b%d, %s, 0x%x, %p}\n", + u.add->bindex, u.add->path, u.add->perm, + u.add->nd.dentry); + break; + case Opt_prepend: + u.add = &opt->add; + LKTRTrace("prepend {b%d, %s, 0x%x, %p}\n", + u.add->bindex, u.add->path, u.add->perm, + u.add->nd.dentry); + break; + case Opt_dirwh: + LKTRTrace("dirwh %d\n", opt->dirwh); + break; + case Opt_rdcache: + LKTRTrace("rdcache %d\n", opt->rdcache); + break; + case Opt_xino: + u.xino = &opt->xino; + LKTRTrace("xino {%s %.*s}\n", + u.xino->path, DLNPair(u.xino->file->f_dentry)); + break; + case Opt_noxino: + LKTRLabel(noxino); + break; + case Opt_plink: + LKTRLabel(plink); + break; + case Opt_noplink: + LKTRLabel(noplink); + break; + case Opt_list_plink: + LKTRLabel(list_plink); + break; + case Opt_clean_plink: + LKTRLabel(clean_plink); + break; + case Opt_udba: + LKTRTrace("udba %d, %s\n", + opt->udba, udba_str(opt->udba)); + break; + case Opt_diropq_a: + LKTRLabel(diropq_a); + break; + case Opt_diropq_w: + LKTRLabel(diropq_w); + break; + case Opt_warn_perm: + LKTRLabel(warn_perm); + break; + case Opt_nowarn_perm: + LKTRLabel(nowarn_perm); + break; + case Opt_dlgt: + LKTRLabel(dlgt); + break; + case Opt_nodlgt: + LKTRLabel(nodlgt); + break; + case Opt_coo: + LKTRTrace("coo %d, %s\n", opt->coo, coo_str(opt->coo)); + break; + default: + BUG(); + } + opt++; + } +} +#else +#define dump_opts(opts) /* */ +#endif + +void au_free_opts(struct opts *opts) +{ + struct opt *opt; + + TraceEnter(); + + opt = opts->opt; + while (opt->type != Opt_tail) { + switch (opt->type) { + case Opt_add: + case Opt_append: + case Opt_prepend: + path_release(&opt->add.nd); + break; + case Opt_del: + case Opt_idel: + dput(opt->del.h_root); + break; + case Opt_mod: + case Opt_imod: + dput(opt->mod.h_root); + break; + case Opt_xino: + fput(opt->xino.file); + break; + } + opt++; + } +} + +static int opt_add(struct opt *opt, char *opt_str, struct super_block *sb, + aufs_bindex_t bindex) +{ + int err; + struct opt_add *add = &opt->add; + char *p; + + LKTRTrace("%s, b%d\n", opt_str, bindex); + + add->bindex = bindex; + add->perm = AuBr_RO; + if (!bindex && !(sb->s_flags & MS_RDONLY)) + add->perm = AuBr_RW; +#ifdef CONFIG_AUFS_COMPAT + add->perm = AuBr_RW; +#endif + add->path = opt_str; + p = strchr(opt_str, '='); + if (unlikely(p)) { + *p++ = 0; + if (*p) + add->perm = br_perm_val(p); + } + + // LSM may detect it + // do not superio. + err = path_lookup(add->path, lkup_dirflags, &add->nd); + //err = -1; + if (!err) { + opt->type = Opt_add; + goto out; + } + Err("lookup failed %s (%d)\n", add->path, err); + err = -EINVAL; + + out: + TraceErr(err); + return err; +} + +/* called without aufs lock */ +int au_parse_opts(struct super_block *sb, char *str, struct opts *opts) +{ + int err, n; + struct dentry *root; + struct opt *opt, *opt_tail; + char *opt_str; + substring_t args[MAX_OPT_ARGS]; + aufs_bindex_t bindex; + struct nameidata nd; + /* reduce stack space */ + union { + struct opt_del *del; + struct opt_mod *mod; + struct opt_xino *xino; + } u; + struct file *file; + + LKTRTrace("%s, nopts %d\n", str, opts->max_opt); + + root = sb->s_root; + err = 0; + bindex = 0; + opt = opts->opt; + opt_tail = opt + opts->max_opt - 1; + opt->type = Opt_tail; + while (!err && (opt_str = strsep(&str, ",")) && *opt_str) { + int token, skipped; + char *p; + err = -EINVAL; + token = match_token(opt_str, options, args); + LKTRTrace("%s, token %d, args[0]{%p, %p}\n", + opt_str, token, args[0].from, args[0].to); + + skipped = 0; + switch (token) { + case Opt_br: + err = 0; + while (!err && (opt_str = strsep(&args[0].from, ":")) + && *opt_str) { + err = opt_add(opt, opt_str, sb, bindex++); + //if (LktrCond) err = -1; + if (unlikely(!err && ++opt > opt_tail)) { + err = -E2BIG; + break; + } + opt->type = Opt_tail; + skipped = 1; + } + break; + case Opt_add: + if (unlikely(match_int(&args[0], &n))) { + Err("bad integer in %s\n", opt_str); + break; + } + bindex = n; + err = opt_add(opt, args[1].from, sb, bindex); + break; + case Opt_append: + case Opt_prepend: + err = opt_add(opt, args[0].from, sb, /*dummy bindex*/1); + if (!err) + opt->type = token; + break; + case Opt_del: + u.del = &opt->del; + u.del->path = args[0].from; + LKTRTrace("del path %s\n", u.del->path); + // LSM may detect it + // do not superio. + err = path_lookup(u.del->path, lkup_dirflags, &nd); + if (unlikely(err)) { + Err("lookup failed %s (%d)\n", u.del->path, err); + break; + } + u.del->h_root = dget(nd.dentry); + path_release(&nd); + opt->type = token; + break; +#if 0 + case Opt_idel: + u.del = &opt->del; + u.del->path = "(indexed)"; + if (unlikely(match_int(&args[0], &n))) { + Err("bad integer in %s\n", opt_str); + break; + } + bindex = n; + aufs_read_lock(root, !AUFS_I_RLOCK); + if (bindex < 0 || sbend(sb) < bindex) { + Err("out of bounds, %d\n", bindex); + aufs_read_unlock(root, !AUFS_I_RLOCK); + break; + } + err = 0; + u.del->h_root = dget(au_h_dptr_i(root, bindex)); + opt->type = token; + aufs_read_unlock(root, !AUFS_I_RLOCK); + break; +#endif + + case Opt_mod: + u.mod = &opt->mod; + u.mod->path = args[0].from; + p = strchr(u.mod->path, '='); + if (unlikely(!p)) { + Err("no permssion %s\n", opt_str); + break; + } + *p++ = 0; + u.mod->perm = br_perm_val(p); + LKTRTrace("mod path %s, perm 0x%x, %s\n", + u.mod->path, u.mod->perm, p); + // LSM may detect it + // do not superio. + err = path_lookup(u.mod->path, lkup_dirflags, &nd); + if (unlikely(err)) { + Err("lookup failed %s (%d)\n", u.mod->path, err); + break; + } + u.mod->h_root = dget(nd.dentry); + path_release(&nd); + opt->type = token; + break; +#if 0 + case Opt_imod: + u.mod = &opt->mod; + u.mod->path = "(indexed)"; + if (unlikely(match_int(&args[0], &n))) { + Err("bad integer in %s\n", opt_str); + break; + } + bindex = n; + aufs_read_lock(root, !AUFS_I_RLOCK); + if (bindex < 0 || sbend(sb) < bindex) { + Err("out of bounds, %d\n", bindex); + aufs_read_unlock(root, !AUFS_I_RLOCK); + break; + } + u.mod->perm = br_perm_val(args[1].from); + LKTRTrace("mod path %s, perm 0x%x, %s\n", + u.mod->path, u.mod->perm, args[1].from); + err = 0; + u.mod->h_root = dget(au_h_dptr_i(root, bindex)); + opt->type = token; + aufs_read_unlock(root, !AUFS_I_RLOCK); + break; +#endif + case Opt_xino: + u.xino = &opt->xino; + file = xino_create(sb, args[0].from, /*silent*/0, + /*parent*/NULL); + err = PTR_ERR(file); + if (IS_ERR(file)) + break; + err = -EINVAL; + if (unlikely(file->f_dentry->d_sb == sb)) { + fput(file); + Err("%s must be outside\n", args[0].from); + break; + } + err = 0; + u.xino->file = file; + u.xino->path = args[0].from; + opt->type = token; + break; + + case Opt_dirwh: + if (unlikely(match_int(&args[0], &opt->dirwh))) + break; + err = 0; + opt->type = token; + break; + + case Opt_rdcache: + if (unlikely(match_int(&args[0], &opt->rdcache))) + break; + err = 0; + opt->type = token; + break; + + case Opt_noxino: + case Opt_plink: + case Opt_noplink: + case Opt_list_plink: + case Opt_clean_plink: + case Opt_diropq_a: + case Opt_diropq_w: + case Opt_warn_perm: + case Opt_nowarn_perm: + case Opt_dlgt: + case Opt_nodlgt: + err = 0; + opt->type = token; + break; + + case Opt_udba: + opt->udba = udba_val(args[0].from); + if (opt->udba >= 0) { + err = 0; + opt->type = token; + } + break; + +#if 0 + case Opt_coo: + opt->coo = coo_val(args[0].from); + if (opt->coo >= 0) { + err = 0; + opt->type = token; + } + break; +#endif + + case Opt_ignore: +#ifndef CONFIG_AUFS_COMPAT + Warn("ignored %s\n", opt_str); +#endif + skipped = 1; + err = 0; + break; + case Opt_err: + Err("unknown option %s\n", opt_str); + break; + } + + if (!err && !skipped) { + if (unlikely(++opt > opt_tail)) { + err = -E2BIG; + opt--; + opt->type = Opt_tail; + break; + } + opt->type = Opt_tail; + } + } + + dump_opts(opts); + if (unlikely(err)) + au_free_opts(opts); + TraceErr(err); + return err; +} + +/* + * returns, + * plus: processed without an error + * zero: unprocessed + */ +static int au_do_opt_simple(struct super_block *sb, struct opt *opt, + int remount, unsigned int *given) +{ + int err; + struct aufs_sbinfo *sbinfo = stosi(sb); + + TraceEnter(); + + err = 1; /* handled */ + switch (opt->type) { + case Opt_udba: + udba_set(sb, opt->udba); + *given |= opt->udba; + break; + + case Opt_plink: + au_flag_set(sb, AuFlag_PLINK); + *given |= AuFlag_PLINK; + break; + case Opt_noplink: + if (au_flag_test(sb, AuFlag_PLINK)) + au_put_plink(sb); + au_flag_clr(sb, AuFlag_PLINK); + *given |= AuFlag_PLINK; + break; + case Opt_list_plink: + if (au_flag_test(sb, AuFlag_PLINK)) + au_list_plink(sb); + break; + case Opt_clean_plink: + if (au_flag_test(sb, AuFlag_PLINK)) + au_put_plink(sb); + break; + + case Opt_diropq_a: + au_flag_set(sb, AuFlag_ALWAYS_DIROPQ); + *given |= AuFlag_ALWAYS_DIROPQ; + break; + case Opt_diropq_w: + au_flag_clr(sb, AuFlag_ALWAYS_DIROPQ); + *given |= AuFlag_ALWAYS_DIROPQ; + break; + + case Opt_dlgt: + au_flag_set(sb, AuFlag_DLGT); + *given |= AuFlag_DLGT; + break; + case Opt_nodlgt: + au_flag_clr(sb, AuFlag_DLGT); + *given |= AuFlag_DLGT; + break; + + case Opt_warn_perm: + au_flag_set(sb, AuFlag_WARN_PERM); + *given |= AuFlag_WARN_PERM; + break; + case Opt_nowarn_perm: + au_flag_clr(sb, AuFlag_WARN_PERM); + *given |= AuFlag_WARN_PERM; + break; + + case Opt_coo: + coo_set(sb, opt->coo); + *given |= opt->coo; + break; + + case Opt_dirwh: + sbinfo->si_dirwh = opt->dirwh; + break; + + case Opt_rdcache: + sbinfo->si_rdcache = opt->rdcache * HZ; + break; + + default: + err = 0; + break; + } + + TraceErr(err); + return err; +} + +/* + * returns tri-state. + * plus: processed without an error + * zero: unprocessed + * minus: error + */ +static int au_do_opt_br(struct super_block *sb, struct opt *opt, int remount, + int *do_refresh) +{ + int err; + + TraceEnter(); + + err = 0; + switch (opt->type) { + case Opt_append: + opt->add.bindex = sbend(sb) + 1; + goto add; + case Opt_prepend: + opt->add.bindex = 0; + add: + case Opt_add: + err = br_add(sb, &opt->add, remount); + if (!err) + *do_refresh = err = 1; + break; + + case Opt_del: + case Opt_idel: + err = br_del(sb, &opt->del, remount); + if (!err) + *do_refresh = err = 1; + break; + + case Opt_mod: + case Opt_imod: + err = br_mod(sb, &opt->mod, remount, do_refresh); + if (!err) + err = 1; + break; + } + + TraceErr(err); + return err; +} + +static int au_do_opt_xino(struct super_block *sb, struct opt *opt, int remount, + struct opt_xino **opt_xino) +{ + int err; + + TraceEnter(); + + err = 0; + switch (opt->type) { + case Opt_xino: + err = xino_set(sb, &opt->xino, remount); + if (!err) + *opt_xino = &opt->xino; + break; + case Opt_noxino: + err = xino_clr(sb); + break; + } + + TraceErr(err); + return err; +} + +static int verify_opts(struct super_block *sb, int remount) +{ + int err; + aufs_bindex_t bindex, bend; + struct aufs_branch *br; + struct dentry *root; + struct inode *dir; + unsigned int do_plink; + + TraceEnter(); + + if (unlikely(!(sb->s_flags & MS_RDONLY) + && !br_writable(sbr_perm(sb, 0)))) + Warn("first branch should be rw\n"); + + err = 0; + root = sb->s_root; + dir = sb->s_root->d_inode; + do_plink = au_flag_test(sb, AuFlag_PLINK); + bend = sbend(sb); + for (bindex = 0; !err && bindex <= bend; bindex++) { + struct inode *h_dir; + int skip; + + skip = 0; + h_dir = au_h_iptr_i(dir, bindex); + br = stobr(sb, bindex); + br_wh_read_lock(br); + switch (br->br_perm) { + case AuBr_RR: + case AuBr_RO: + case AuBr_RRWH: + case AuBr_ROWH: + skip = (!br->br_wh && !br->br_plink); + break; + + case AuBr_RWNoLinkWH: + skip = !br->br_wh; + if (skip) { + if (do_plink) + skip = !!br->br_plink; + else + skip = !br->br_plink; + } + break; + + case AuBr_RW: + skip = !!br->br_wh; + if (skip) { + if (do_plink) + skip = !!br->br_plink; + else + skip = !br->br_plink; + } + break; + + default: + BUG(); + } + br_wh_read_unlock(br); + + if (skip) + continue; + + hdir_lock(h_dir, dir, bindex); + br_wh_write_lock(br); + err = init_wh(au_h_dptr_i(root, bindex), br, + au_nfsmnt(sb, bindex), sb); + br_wh_write_unlock(br); + hdir_unlock(h_dir, dir, bindex); + } + + TraceErr(err); + return err; +} + +int au_do_opts_mount(struct super_block *sb, struct opts *opts) +{ + int err, do_refresh; + struct inode *dir; + struct opt *opt; + unsigned int flags, given; + struct opt_xino *opt_xino; + aufs_bindex_t bend, bindex; + + TraceEnter(); + SiMustWriteLock(sb); + DiMustWriteLock(sb->s_root); + dir = sb->s_root->d_inode; + IiMustWriteLock(dir); + + err = 0; + given = 0; + opt_xino = NULL; + opt = opts->opt; + while (err >= 0 && opt->type != Opt_tail) + err = au_do_opt_simple(sb, opt++, /*remount*/0, &given); + if (err > 0) + err = 0; + else if (unlikely(err < 0)) + goto out; + + /* disable them temporary */ + flags = au_flag_test(sb, AuFlag_XINO | AuMask_UDBA | AuFlag_DLGT); + au_flag_clr(sb, AuFlag_XINO | AuFlag_DLGT); + udba_set(sb, AuFlag_UDBA_REVAL); + + do_refresh = 0; + opt = opts->opt; + while (err >= 0 && opt->type != Opt_tail) + err = au_do_opt_br(sb, opt++, /*remount*/0, &do_refresh); + if (err > 0) + err = 0; + else if (unlikely(err < 0)) + goto out; + + bend = sbend(sb); + if (unlikely(bend < 0)) { + err = -EINVAL; + Err("no branches\n"); + goto out; + } + + if (flags & AuFlag_XINO) + au_flag_set(sb, AuFlag_XINO); + opt = opts->opt; + while (!err && opt->type != Opt_tail) + err = au_do_opt_xino(sb, opt++, /*remount*/0, &opt_xino); + if (unlikely(err)) + goto out; + + //todo: test this error case. + err = verify_opts(sb, /*remount*/0); + DEBUG_ON(err); + if (unlikely(err)) + goto out; + + /* enable xino */ + if (au_flag_test(sb, AuFlag_XINO) && !opt_xino) { + struct file *xino_file = xino_def(sb); + err = PTR_ERR(xino_file); + if (IS_ERR(xino_file)) + goto out; + + err = 0; + for (bindex = 0; !err && bindex <= bend; bindex++) + err = xino_init(sb, bindex, xino_file, + /*do_test*/bindex); + fput(xino_file); + if (unlikely(err)) + goto out; + } + + /* restore hinotify */ + udba_set(sb, flags & AuMask_UDBA); + if (flags & AuFlag_UDBA_INOTIFY) + au_reset_hinotify(dir, au_hi_flags(dir, 1) & ~AUFS_HI_XINO); + + /* restore dlgt */ + if (flags & AuFlag_DLGT) + au_flag_set(sb, AuFlag_DLGT); + + out: + TraceErr(err); + return err; +} + +int au_do_opts_remount(struct super_block *sb, struct opts *opts, + int *do_refresh, unsigned int *given) +{ + int err, rerr; + struct inode *dir; + struct opt_xino *opt_xino; + struct opt *opt; + unsigned int dlgt; + + TraceEnter(); + SiMustWriteLock(sb); + DiMustWriteLock(sb->s_root); + dir = sb->s_root->d_inode; + IiMustWriteLock(dir); + //DEBUG_ON(au_flag_test(sb, AuFlag_UDBA_INOTIFY)); + + err = 0; + *do_refresh = 0; + *given = 0; + dlgt = au_flag_test(sb, AuFlag_DLGT); + opt_xino = NULL; + opt = opts->opt; + while (err >= 0 && opt->type != Opt_tail) { + err = au_do_opt_simple(sb, opt, /*remount*/1, given); + + /* disable it temporary */ + dlgt = au_flag_test(sb, AuFlag_DLGT); + au_flag_clr(sb, AuFlag_DLGT); + + if (!err) + err = au_do_opt_br(sb, opt, /*remount*/1, do_refresh); + if (!err) + err = au_do_opt_xino(sb, opt, /*remount*/1, &opt_xino); + + /* restore it */ + au_flag_set(sb, dlgt); + opt++; + } + if (err > 0) + err = 0; + TraceErr(err); + + /* go on if err */ + + //todo: test this error case. + au_flag_clr(sb, AuFlag_DLGT); + rerr = verify_opts(sb, /*remount*/1); + au_flag_set(sb, dlgt); + + /* they are handled by the caller */ + if (!*do_refresh) + *do_refresh = !!((*given & AuMask_UDBA) + || au_flag_test(sb, AuFlag_XINO)); + + TraceErr(err); + return err; +} diff --git a/fs/aufs/opts.h b/fs/aufs/opts.h new file mode 100755 index 0000000..16c1a6a --- /dev/null +++ b/fs/aufs/opts.h @@ -0,0 +1,96 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: opts.h,v 1.13 2007/05/14 06:27:18 sfjro Exp $ */ + +#ifndef __AUFS_OPTS_H__ +#define __AUFS_OPTS_H__ + +#ifdef __KERNEL__ + +#include +#include +#include +#include + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22) +typedef const char* au_parser_pattern_t; +#else +typedef char* au_parser_pattern_t; +#endif + +/* ---------------------------------------------------------------------- */ + +struct opt_add { + aufs_bindex_t bindex; + char *path; + int perm; + struct nameidata nd; +}; + +struct opt_del { + char *path; + struct dentry *h_root; +}; + +struct opt_mod { + char *path; + int perm; + struct dentry *h_root; +}; + +struct opt_xino { + char *path; + struct file *file; +}; + +struct opt { + int type; + union { + struct opt_xino xino; + struct opt_add add; + struct opt_del del; + struct opt_mod mod; + int dirwh; + int rdcache; + int deblk; + int nhash; + int udba; + int coo; + }; +}; + +struct opts { + struct opt *opt; + int max_opt; +}; + +/* ---------------------------------------------------------------------- */ + +int br_perm_str(char *p, unsigned int len, int brperm); +au_parser_pattern_t udba_str(int udba); +void udba_set(struct super_block *sb, unsigned int flg); +//au_parser_pattern_t coo_str(int coo); +void au_free_opts(struct opts *opts); +int au_parse_opts(struct super_block *sb, char *str, struct opts *opts); +int au_do_opts_mount(struct super_block *sb, struct opts *opts); +int au_do_opts_remount(struct super_block *sb, struct opts *opts, + int *do_refresh, unsigned int *given); + +#endif /* __KERNEL__ */ +#endif /* __AUFS_OPTS_H__ */ diff --git a/fs/aufs/plink.c b/fs/aufs/plink.c new file mode 100755 index 0000000..0e520af --- /dev/null +++ b/fs/aufs/plink.c @@ -0,0 +1,331 @@ +/* + * Copyright (C) 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: plink.c,v 1.4 2007/05/14 03:39:10 sfjro Exp $ */ + +#include "aufs.h" + +struct pseudo_link { + struct list_head list; + struct inode *inode; +}; + +#ifdef CONFIG_AUFS_DEBUG +void au_list_plink(struct super_block *sb) +{ + struct aufs_sbinfo *sbinfo; + struct list_head *plink_list; + struct pseudo_link *plink; + + TraceEnter(); + SiMustAnyLock(sb); + DEBUG_ON(!au_flag_test(sb, AuFlag_PLINK)); + + sbinfo = stosi(sb); + plink_list = &sbinfo->si_plink; + spin_lock(&sbinfo->si_plink_lock); + list_for_each_entry(plink, plink_list, list) + Dbg("%lu\n", plink->inode->i_ino); + spin_unlock(&sbinfo->si_plink_lock); +} +#endif + +int au_is_plinked(struct super_block *sb, struct inode *inode) +{ + int found; + struct aufs_sbinfo *sbinfo; + struct list_head *plink_list; + struct pseudo_link *plink; + + LKTRTrace("i%lu\n", inode->i_ino); + SiMustAnyLock(sb); + DEBUG_ON(!au_flag_test(sb, AuFlag_PLINK)); + + found = 0; + sbinfo = stosi(sb); + plink_list = &sbinfo->si_plink; + spin_lock(&sbinfo->si_plink_lock); + list_for_each_entry(plink, plink_list, list) + if (plink->inode == inode) { + found = 1; + break; + } + spin_unlock(&sbinfo->si_plink_lock); + return found; +} + +// 20 is max digits length of ulong 64 +#define PLINK_NAME_LEN ((20 + 1) * 2) + +static int plink_name(char *name, int len, struct inode *inode, + aufs_bindex_t bindex) +{ + int rlen; + struct inode *h_inode; + + LKTRTrace("i%lu, b%d\n", inode->i_ino, bindex); + DEBUG_ON(len != PLINK_NAME_LEN); + h_inode = au_h_iptr_i(inode, bindex); + DEBUG_ON(!h_inode); + rlen = snprintf(name, len, "%lu.%lu", inode->i_ino, h_inode->i_ino); + DEBUG_ON(rlen >= len); + return rlen; +} + +struct dentry *lkup_plink(struct super_block *sb, aufs_bindex_t bindex, + struct inode *inode) +{ + struct dentry *h_dentry, *h_parent; + struct aufs_branch *br; + struct inode *h_dir; + char tgtname[PLINK_NAME_LEN]; + int len; + struct lkup_args lkup; + + LKTRTrace("b%d, i%lu\n", bindex, inode->i_ino); + br = stobr(sb, bindex); + h_parent = br->br_plink; + DEBUG_ON(!h_parent); + h_dir = h_parent->d_inode; + DEBUG_ON(!h_dir); + + len = plink_name(tgtname, sizeof(tgtname), inode, bindex); + + // always superio. + lkup.nfsmnt = au_do_nfsmnt(br->br_mnt); + lkup.dlgt = need_dlgt(sb); + hi_lock_whplink(h_dir); + h_dentry = sio_lkup_one(tgtname, h_parent, len, &lkup); + i_unlock(h_dir); + return h_dentry; +} + +static int do_whplink(char *tgt, int len, struct dentry *h_parent, + struct dentry *h_dentry, struct vfsmount *nfsmnt, + struct super_block *sb) +{ + int err; + struct dentry *h_tgt; + struct inode *h_dir; + struct lkup_args lkup = { + .nfsmnt = nfsmnt, + .dlgt = need_dlgt(sb) + }; + + h_tgt = lkup_one(tgt, h_parent, len, &lkup); + err = PTR_ERR(h_tgt); + if (IS_ERR(h_tgt)) + goto out; + + err = 0; + h_dir = h_parent->d_inode; + if (unlikely(h_tgt->d_inode && h_tgt->d_inode != h_dentry->d_inode)) + err = vfsub_unlink(h_dir, h_tgt, lkup.dlgt); + if (!err && !h_tgt->d_inode) { + err = vfsub_link(h_dentry, h_dir, h_tgt, lkup.dlgt); + //inode->i_nlink++; + } + dput(h_tgt); + + out: + TraceErr(err); + return err; +} + +struct do_whplink_args { + int *errp; + char *tgt; + int len; + struct dentry *h_parent; + struct dentry *h_dentry; + struct vfsmount *nfsmnt; + struct super_block *sb; +}; + +static void call_do_whplink(void *args) +{ + struct do_whplink_args *a = args; + *a->errp = do_whplink(a->tgt, a->len, a->h_parent, a->h_dentry, + a->nfsmnt, a->sb); +} + +static int whplink(struct dentry *h_dentry, struct inode *inode, + aufs_bindex_t bindex, struct super_block *sb) +{ + int err, len; + struct aufs_branch *br; + struct dentry *h_parent; + struct inode *h_dir; + char tgtname[PLINK_NAME_LEN]; + + LKTRTrace("%.*s\n", DLNPair(h_dentry)); + br = stobr(inode->i_sb, bindex); + h_parent = br->br_plink; + DEBUG_ON(!h_parent); + h_dir = h_parent->d_inode; + DEBUG_ON(!h_dir); + + len = plink_name(tgtname, sizeof(tgtname), inode, bindex); + + // always superio. + hi_lock_whplink(h_dir); + if (!is_au_wkq(current)) { + struct do_whplink_args args = { + .errp = &err, + .tgt = tgtname, + .len = len, + .h_parent = h_parent, + .h_dentry = h_dentry, + .nfsmnt = au_do_nfsmnt(br->br_mnt), + .sb = sb + }; + au_wkq_wait(call_do_whplink, &args, /*dlgt*/0); + } else + err = do_whplink(tgtname, len, h_parent, h_dentry, + au_do_nfsmnt(br->br_mnt), sb); + i_unlock(h_dir); + + TraceErr(err); + return err; +} + +void append_plink(struct super_block *sb, struct inode *inode, + struct dentry *h_dentry, aufs_bindex_t bindex) +{ + struct aufs_sbinfo *sbinfo; + struct list_head *plink_list; + struct pseudo_link *plink; + int found, err, cnt; + + LKTRTrace("i%lu\n", inode->i_ino); + SiMustAnyLock(sb); + DEBUG_ON(!au_flag_test(sb, AuFlag_PLINK)); + + cnt = 0; + found = 0; + sbinfo = stosi(sb); + plink_list = &sbinfo->si_plink; + spin_lock(&sbinfo->si_plink_lock); + list_for_each_entry(plink, plink_list, list) { + cnt++; + if (plink->inode == inode) { + found = 1; + break; + } + } + + err = 0; + if (!found) { + struct pseudo_link *plink; + + plink = kmalloc(sizeof(*plink), GFP_ATOMIC); + if (plink) { + plink->inode = igrab(inode); + list_add(&plink->list, plink_list); + cnt++; + } else + err = -ENOMEM; + } + spin_unlock(&sbinfo->si_plink_lock); + + if (!err) + err = whplink(h_dentry, inode, bindex, sb); + + if (unlikely(cnt > 100)) + Warn1("unexpectedly many pseudo links, %d\n", cnt); + if (unlikely(err)) + Warn("err %d, damaged pseudo link. ignored.\n", err); +} + +static void do_put_plink(struct pseudo_link *plink, int do_del) +{ + TraceEnter(); + + iput(plink->inode); + if (do_del) + list_del(&plink->list); + kfree(plink); +} + +void au_put_plink(struct super_block *sb) +{ + struct aufs_sbinfo *sbinfo; + struct list_head *plink_list; + struct pseudo_link *plink, *tmp; + + TraceEnter(); + SiMustWriteLock(sb); + DEBUG_ON(!au_flag_test(sb, AuFlag_PLINK)); + + sbinfo = stosi(sb); + plink_list = &sbinfo->si_plink; + //spin_lock(&sbinfo->si_plink_lock); + list_for_each_entry_safe(plink, tmp, plink_list, list) + do_put_plink(plink, 0); + INIT_LIST_HEAD(plink_list); + //spin_unlock(&sbinfo->si_plink_lock); +} + +void half_refresh_plink(struct super_block *sb, aufs_bindex_t br_id) +{ + struct aufs_sbinfo *sbinfo; + struct list_head *plink_list; + struct pseudo_link *plink, *tmp; + struct inode *inode; + aufs_bindex_t bstart, bend, bindex; + int do_put; + + TraceEnter(); + SiMustWriteLock(sb); + DEBUG_ON(!au_flag_test(sb, AuFlag_PLINK)); + + sbinfo = stosi(sb); + plink_list = &sbinfo->si_plink; + //spin_lock(&sbinfo->si_plink_lock); + list_for_each_entry_safe(plink, tmp, plink_list, list) { + do_put = 0; + inode = igrab(plink->inode); + ii_write_lock_child(inode); + bstart = ibstart(inode); + bend = ibend(inode); + if (bstart >= 0) { + for (bindex = bstart; bindex <= bend; bindex++) { + if (!au_h_iptr_i(inode, bindex) + || itoid_index(inode, bindex) != br_id) + continue; + set_h_iptr(inode, bindex, NULL, 0); + do_put = 1; + break; + } + } else + do_put_plink(plink, 1); + + if (do_put) { + for (bindex = bstart; bindex <= bend; bindex++) + if (au_h_iptr_i(inode, bindex)) { + do_put = 0; + break; + } + if (do_put) + do_put_plink(plink, 1); + } + ii_write_unlock(inode); + iput(inode); + } + //spin_unlock(&sbinfo->si_plink_lock); +} diff --git a/fs/aufs/sbinfo.c b/fs/aufs/sbinfo.c new file mode 100755 index 0000000..55cb64c --- /dev/null +++ b/fs/aufs/sbinfo.c @@ -0,0 +1,173 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: sbinfo.c,v 1.30 2007/05/14 03:39:31 sfjro Exp $ */ + +#include "aufs.h" + +struct aufs_sbinfo *stosi(struct super_block *sb) +{ + struct aufs_sbinfo *sbinfo; + sbinfo = sb->s_fs_info; + //DEBUG_ON(sbinfo->si_bend < 0); + return sbinfo; +} + +aufs_bindex_t sbend(struct super_block *sb) +{ + SiMustAnyLock(sb); + return stosi(sb)->si_bend; +} + +struct aufs_branch *stobr(struct super_block *sb, aufs_bindex_t bindex) +{ + SiMustAnyLock(sb); + DEBUG_ON(bindex < 0 || sbend(sb) < bindex + || !stosi(sb)->si_branch[0 + bindex]); + return stosi(sb)->si_branch[0 + bindex]; +} + +int au_sigen(struct super_block *sb) +{ + SiMustAnyLock(sb); + return stosi(sb)->si_generation; +} + +int au_sigen_inc(struct super_block *sb) +{ + int gen; + + SiMustWriteLock(sb); + gen = ++stosi(sb)->si_generation; + au_update_digen(sb->s_root); + au_update_iigen(sb->s_root->d_inode); + sb->s_root->d_inode->i_version++; + return gen; +} + +int find_bindex(struct super_block *sb, struct aufs_branch *br) +{ + aufs_bindex_t bindex, bend; + + bend = sbend(sb); + for (bindex = 0; bindex <= bend; bindex++) + if (stobr(sb, bindex) == br) + return bindex; + return -1; +} + +/* ---------------------------------------------------------------------- */ + +/* dentry and super_block lock. call at entry point */ +void aufs_read_lock(struct dentry *dentry, int flags) +{ + si_read_lock(dentry->d_sb); + if (flags & AUFS_D_WLOCK) + di_write_lock_child(dentry); + else + di_read_lock_child(dentry, flags); +} + +void aufs_read_unlock(struct dentry *dentry, int flags) +{ + if (flags & AUFS_D_WLOCK) + di_write_unlock(dentry); + else + di_read_unlock(dentry, flags); + si_read_unlock(dentry->d_sb); +} + +void aufs_write_lock(struct dentry *dentry) +{ + //au_wkq_wait_nwtask(); + si_write_lock(dentry->d_sb); + di_write_lock_child(dentry); +} + +void aufs_write_unlock(struct dentry *dentry) +{ + di_write_unlock(dentry); + si_write_unlock(dentry->d_sb); + //au_wkq_wait_nwtask(); +} + +void aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int isdir) +{ + DEBUG_ON(d1 == d2 || d1->d_sb != d2->d_sb); + si_read_lock(d1->d_sb); + di_write_lock2_child(d1, d2, isdir); +} + +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2) +{ + DEBUG_ON(d1 == d2 || d1->d_sb != d2->d_sb); + di_write_unlock2(d1, d2); + si_read_unlock(d1->d_sb); +} + +/* ---------------------------------------------------------------------- */ + +aufs_bindex_t new_br_id(struct super_block *sb) +{ + aufs_bindex_t br_id; + + TraceEnter(); + SiMustWriteLock(sb); + + while (1) { + br_id = ++stosi(sb)->si_last_br_id; + if (br_id && find_brindex(sb, br_id) < 0) + return br_id; + } +} + +/* ---------------------------------------------------------------------- */ + +#ifdef CONFIG_AUFS_SYSAUFS +static int make_xino(struct seq_file *seq, struct sysaufs_args *args, + int *do_size) +{ + int err; + struct super_block *sb = args->sb; + aufs_bindex_t bindex, bend; + struct file *xf; + struct inode *xi; + + TraceEnter(); + DEBUG_ON(args->index != SysaufsSb_XINO); + SiMustReadLock(sb); + + *do_size = 0; + err = seq_printf(seq, "%d %lu\n", sizeof(struct xino), + atomic_long_read(&stosi(sb)->si_xino)); + bend = sbend(sb); + for (bindex = 0; !err && bindex <= bend; bindex++) { + xf = stobr(sb, bindex)->br_xino; + xi = xf->f_dentry->d_inode; + err = seq_printf(seq, "%d: %d, %Lux%d %Ld\n", + bindex, file_count(xf), + (u64)xi->i_blocks, 1 << xi->i_blkbits, + i_size_read(xi)); + } + return err; +} + +sysaufs_op au_si_ops[] = { + [SysaufsSb_XINO] = make_xino +}; +#endif diff --git a/fs/aufs/super.c b/fs/aufs/super.c new file mode 100755 index 0000000..c1123f8 --- /dev/null +++ b/fs/aufs/super.c @@ -0,0 +1,716 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: super.c,v 1.50 2007/05/14 03:39:42 sfjro Exp $ */ + +#include +#include +#include +#include "aufs.h" + +/* + * super_operations + */ +static struct inode *aufs_alloc_inode(struct super_block *sb) +{ + struct aufs_icntnr *c; + + TraceEnter(); + + c = cache_alloc_icntnr(); + //if (LktrCond) {cache_free_icntnr(c); c = NULL;} + if (c) { + inode_init_once(&c->vfs_inode); + c->vfs_inode.i_version = 1; //sigen(sb); + c->iinfo.ii_hinode = NULL; + return &c->vfs_inode; + } + return NULL; +} + +static void aufs_destroy_inode(struct inode *inode) +{ + LKTRTrace("i%lu\n", inode->i_ino); + au_iinfo_fin(inode); + cache_free_icntnr(container_of(inode, struct aufs_icntnr, vfs_inode)); +} + +//todo: how about merge with alloc_inode()? +static void aufs_read_inode(struct inode *inode) +{ + int err; + + LKTRTrace("i%lu\n", inode->i_ino); + + err = au_iinfo_init(inode); + //if (LktrCond) err = -1; + if (!err) { + inode->i_version++; + inode->i_op = &aufs_iop; + inode->i_fop = &aufs_file_fop; + inode->i_mapping->a_ops = &aufs_aop; + return; /* success */ + } + + LKTRTrace("intializing inode info failed(%d)\n", err); + make_bad_inode(inode); +} + +int au_show_brs(struct seq_file *seq, struct super_block *sb) +{ + int err; + aufs_bindex_t bindex, bend; + char a[16]; + struct dentry *root; + + TraceEnter(); + SiMustAnyLock(sb); + root = sb->s_root; + DiMustAnyLock(root); + + err = 0; + bend = sbend(sb); + for (bindex = 0; !err && bindex <= bend; bindex++) { + err = br_perm_str(a, sizeof(a), sbr_perm(sb, bindex)); + if (!err) + err = seq_path(seq, sbr_mnt(sb, bindex), + au_h_dptr_i(root, bindex), au_esc_chars); + if (err > 0) + err = seq_printf(seq, "=%s", a); + if (!err && bindex != bend) + err = seq_putc(seq, ':'); + } + + TraceErr(err); + return err; +} + +static int aufs_show_options(struct seq_file *m, struct vfsmount *mnt) +{ + int err, n; + struct super_block *sb; + struct aufs_sbinfo *sbinfo; + struct dentry *root; + struct file *xino; + + TraceEnter(); + + sb = mnt->mnt_sb; + root = sb->s_root; + aufs_read_lock(root, !AUFS_I_RLOCK); + if (au_flag_test(sb, AuFlag_XINO)) { + err = seq_puts(m, ",xino="); + if (unlikely(err)) + goto out; + xino = stobr(sb, 0)->br_xino; + err = seq_path(m, xino->f_vfsmnt, xino->f_dentry, au_esc_chars); + if (unlikely(err <= 0)) + goto out; + err = 0; + +#define Deleted "\\040(deleted)" + m->count -= sizeof(Deleted) - 1; + DEBUG_ON(memcmp(m->buf + m->count, Deleted, + sizeof(Deleted) - 1)); +#undef Deleted + } else + err = seq_puts(m, ",noxino"); + + n = au_flag_test(sb, AuFlag_PLINK); + if (unlikely(!err && (AuDefFlags & AuFlag_PLINK) != n)) + err = seq_printf(m, ",%splink", n ? "" : "no"); + n = au_flag_test_udba(sb); + if (unlikely(!err && (AuDefFlags & AuMask_UDBA) != n)) + err = seq_printf(m, ",udba=%s", udba_str(n)); + n = au_flag_test(sb, AuFlag_ALWAYS_DIROPQ); + if (unlikely(!err && (AuDefFlags & AuFlag_ALWAYS_DIROPQ) != n)) + err = seq_printf(m, ",diropq=%c", n ? 'a' : 'w'); + n = au_flag_test(sb, AuFlag_DLGT); + if (unlikely(!err && (AuDefFlags & AuFlag_DLGT) != n)) + err = seq_printf(m, ",%sdlgt", n ? "" : "no"); + n = au_flag_test(sb, AuFlag_WARN_PERM); + if (unlikely(!err && (AuDefFlags & AuFlag_WARN_PERM) != n)) + err = seq_printf(m, ",%swarn_perm", n ? "" : "no"); + + sbinfo = stosi(sb); + n = sbinfo->si_dirwh; + if (unlikely(!err && n != AUFS_DIRWH_DEF)) + err = seq_printf(m, ",dirwh=%d", n); + n = sbinfo->si_rdcache / HZ; + if (unlikely(!err && n != AUFS_RDCACHE_DEF)) + err = seq_printf(m, ",rdcache=%d", n); +#if 0 + n = au_flag_test_coo(sb); + if (unlikely(!err && (AuDefFlags & AuMask_COO) != n)) + err = seq_printf(m, ",coo=%s", coo_str(n)); +#endif + + if (!err && !sysaufs_brs) { +#ifdef CONFIG_AUFS_COMPAT + err = seq_puts(m, ",dirs="); +#else + err = seq_puts(m, ",br:"); +#endif + if (!err) + err = au_show_brs(m, sb); + } + + out: + aufs_read_unlock(root, !AUFS_I_RLOCK); + TraceErr(err); + if (err) + err = -E2BIG; + TraceErr(err); + return err; +} + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) +#define StatfsLock(d) aufs_read_lock((d)->d_sb->s_root, 0) +#define StatfsUnlock(d) aufs_read_unlock((d)->d_sb->s_root, 0) +#define StatfsArg(d) au_h_dptr((d)->d_sb->s_root) +#define StatfsHInode(d) (StatfsArg(d)->d_inode) +#define StatfsSb(d) ((d)->d_sb) +static int aufs_statfs(struct dentry *arg, struct kstatfs *buf) +#else +#define StatfsLock(s) si_read_lock(s) +#define StatfsUnlock(s) si_read_unlock(s) +#define StatfsArg(s) sbr_sb(s, 0) +#define StatfsHInode(s) (StatfsArg(s)->s_root->d_inode) +#define StatfsSb(s) (s) +static int aufs_statfs(struct super_block *arg, struct kstatfs *buf) +#endif +{ + int err; + + TraceEnter(); + + StatfsLock(arg); + err = vfsub_statfs(StatfsArg(arg), buf, need_dlgt(StatfsSb(arg))); + //if (LktrCond) err = -1; + StatfsUnlock(arg); + if (!err) { + //buf->f_type = AUFS_SUPER_MAGIC; + buf->f_type = 0; + buf->f_namelen -= AUFS_WH_PFX_LEN; + memset(&buf->f_fsid, 0, sizeof(buf->f_fsid)); + } + //buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1; + + TraceErr(err); + return err; +} + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) || defined(UbuntuEdgy17Umount18) +#define UmountBeginSb(mnt) (mnt)->mnt_sb +static void aufs_umount_begin(struct vfsmount *arg, int flags) +#else +#define UmountBeginSb(sb) sb +static void aufs_umount_begin(struct super_block *arg) +#endif +{ + struct super_block *sb = UmountBeginSb(arg); + + if (unlikely(!stosi(sb))) + return; + + //au_wkq_wait_nwtask(); + si_write_lock(sb); + if (au_flag_test(sb, AuFlag_PLINK)) { + au_put_plink(sb); + //kobj_umount(stosi(sb)); + } +#if 0 + if (unlikely(au_flag_test(sb, AuFlag_UDBA_INOTIFY))) + shrink_dcache_sb(sb); +#endif + si_write_unlock(sb); +} + +static void free_sbinfo(struct aufs_sbinfo *sbinfo) +{ + TraceEnter(); + DEBUG_ON(!sbinfo + || !list_empty(&sbinfo->si_plink)); + + free_branches(sbinfo); + kfree(sbinfo->si_branch); + kfree(sbinfo); +} + +/* final actions when unmounting a file system */ +static void aufs_put_super(struct super_block *sb) +{ + struct aufs_sbinfo *sbinfo; + + TraceEnter(); + + sbinfo = stosi(sb); + if (unlikely(!sbinfo)) + return; + + sysaufs_del(sbinfo); + +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,18) && !defined(UbuntuEdgy17Umount18) + // umount_begin() may not be called. + aufs_umount_begin(sb); +#endif + free_sbinfo(sbinfo); +} + +/* ---------------------------------------------------------------------- */ + +/* + * refresh directories at remount time. + */ +static int do_refresh_dir(struct dentry *dentry, unsigned int flags) +{ + int err; + struct dentry *parent; + struct inode *inode; + + LKTRTrace("%.*s\n", DLNPair(dentry)); + inode = dentry->d_inode; + DEBUG_ON(!inode || !S_ISDIR(inode->i_mode)); + + di_write_lock_child(dentry); + parent = dget_parent(dentry); + di_read_lock_parent(parent, AUFS_I_RLOCK); + err = au_refresh_hdentry(dentry, S_IFDIR); + if (err >= 0) { + err = au_refresh_hinode(inode, dentry); + if (!err) + au_reset_hinotify(inode, flags); + } + if (unlikely(err)) + Err("unrecoverable error %d\n", err); + di_read_unlock(parent, AUFS_I_RLOCK); + dput(parent); + di_write_unlock(dentry); + + TraceErr(err); + return err; +} + +static int test_dir(struct dentry *dentry, void *arg) +{ + return S_ISDIR(dentry->d_inode->i_mode); +} + +static int refresh_dir(struct dentry *root, int sgen) +{ + int err, i, j, ndentry; + const unsigned int flags = au_hi_flags(root->d_inode, /*isdir*/1); + struct au_dcsub_pages dpages; + struct au_dpage *dpage; + struct dentry **dentries; + + LKTRTrace("sgen %d\n", sgen); + SiMustWriteLock(root->d_sb); + DEBUG_ON(au_digen(root) != sgen); + DiMustWriteLock(root); + + err = au_dpages_init(&dpages, GFP_KERNEL); + if (unlikely(err)) + goto out; + err = au_dcsub_pages(&dpages, root, test_dir, NULL); + if (unlikely(err)) + goto out_dpages; + + DiMustNoWaiters(root); + IiMustNoWaiters(root->d_inode); + di_write_unlock(root); + for (i = 0; !err && i < dpages.ndpage; i++) { + dpage = dpages.dpages + i; + dentries = dpage->dentries; + ndentry = dpage->ndentry; + for (j = 0; !err && j < ndentry; j++) { + struct dentry *d; + d = dentries[j]; + DEBUG_ON(!S_ISDIR(d->d_inode->i_mode) + || IS_ROOT(d) + || au_digen(d->d_parent) != sgen); + if (au_digen(d) != sgen) + err = do_refresh_dir(d, flags); + } + } + di_write_lock_child(root); /* aufs_write_lock() calls ..._child() */ + + out_dpages: + au_dpages_free(&dpages); + out: + TraceErr(err); + return err; +} + +/* stop extra interpretation of errno in mount(8), and strange error messages */ +static int cvt_err(int err) +{ + TraceErr(err); + + switch (err) { + case -ENOENT: + case -ENOTDIR: + case -EEXIST: + case -EIO: + err = -EINVAL; + } + return err; +} + +/* protected by s_umount */ +static int aufs_remount_fs(struct super_block *sb, int *flags, char *data) +{ + int err, do_refresh; + struct dentry *root; + struct inode *inode; + struct opts opts; + unsigned int given, dlgt; + + //au_debug_on(); + LKTRTrace("flags 0x%x, data %s, len %d\n", + *flags, data ? data : "NULL", data ? strlen(data) : 0); + + err = 0; + if (unlikely(!data || !*data)) + goto out; /* success */ + + err = -ENOMEM; + memset(&opts, 0, sizeof(opts)); + opts.opt = (void*)__get_free_page(GFP_KERNEL); + //if (LktrCond) {free_page((unsigned long)opts.opt); opts.opt = NULL;} + if (unlikely(!opts.opt)) + goto out; + opts.max_opt = PAGE_SIZE / sizeof(*opts.opt); + + /* parse it before aufs lock */ + err = au_parse_opts(sb, data, &opts); + //if (LktrCond) {au_free_opts(&opts); err = -1;} + if (unlikely(err)) + goto out_opts; + + root = sb->s_root; + inode = root->d_inode; + i_lock(inode); + aufs_write_lock(root); + + //DbgSleep(3); + + /* au_do_opts() may return an error */ + do_refresh = 0; + given = 0; + err = au_do_opts_remount(sb, &opts, &do_refresh, &given); + //if (LktrCond) err = -1; + au_free_opts(&opts); + + if (do_refresh) { + int rerr; + struct aufs_sbinfo *sbinfo; + + dlgt = au_flag_test(sb, AuFlag_DLGT); + au_flag_clr(sb, AuFlag_DLGT); + au_sigen_inc(sb); + au_reset_hinotify(inode, au_hi_flags(inode, /*isdir*/1)); + sbinfo = stosi(sb); + sbinfo->si_failed_refresh_dirs = 0; + rerr = refresh_dir(root, au_sigen(sb)); + if (unlikely(rerr)) { + sbinfo->si_failed_refresh_dirs = 1; + Warn("Refreshing directories failed, ignores (%d)\n", + rerr); + } + au_cpup_attr_all(inode); + au_flag_set(sb, dlgt); + } + + aufs_write_unlock(root); + i_unlock(inode); + /* braces are added to stop a warning */ + if (do_refresh) { + sysaufs_notify_remount(); + } + + out_opts: + free_page((unsigned long)opts.opt); + out: + err = cvt_err(err); + TraceErr(err); + //au_debug_off(); + return err; +} + +static struct super_operations aufs_sop = { + .alloc_inode = aufs_alloc_inode, + .destroy_inode = aufs_destroy_inode, + .read_inode = aufs_read_inode, + //.dirty_inode = aufs_dirty_inode, + //.write_inode = aufs_write_inode, + //void (*put_inode) (struct inode *); + .drop_inode = generic_delete_inode, + //.delete_inode = aufs_delete_inode, + //.clear_inode = aufs_clear_inode, + + .show_options = aufs_show_options, + .statfs = aufs_statfs, + + .put_super = aufs_put_super, + //void (*write_super) (struct super_block *); + //int (*sync_fs)(struct super_block *sb, int wait); + //void (*write_super_lockfs) (struct super_block *); + //void (*unlockfs) (struct super_block *); + .remount_fs = aufs_remount_fs, + // depends upon umount flags. also use put_super() (< 2.6.18) + .umount_begin = aufs_umount_begin +}; + +/* ---------------------------------------------------------------------- */ + +/* + * at first mount time. + */ + +static int alloc_sbinfo(struct super_block *sb) +{ + struct aufs_sbinfo *sbinfo; + + TraceEnter(); + + sbinfo = kmalloc(sizeof(*sbinfo), GFP_KERNEL); + //if (LktrCond) {kfree(sbinfo); sbinfo = NULL;} + if (unlikely(!sbinfo)) + goto out; + sbinfo->si_branch = kzalloc(sizeof(*sbinfo->si_branch), GFP_KERNEL); + //if (LktrCond) {kfree(sbinfo->si_branch); sbinfo->si_branch = NULL;} + if (unlikely(!sbinfo->si_branch)) { + kfree(sbinfo); + goto out; + } + rw_init_wlock(&sbinfo->si_rwsem); + sbinfo->si_bend = -1; + atomic_long_set(&sbinfo->si_xino, AUFS_FIRST_INO); + spin_lock_init(&sbinfo->si_plink_lock); + INIT_LIST_HEAD(&sbinfo->si_plink); + init_lvma(sbinfo); + sbinfo->si_generation = 0; + sbinfo->si_last_br_id = 0; + sbinfo->si_failed_refresh_dirs = 0; + sbinfo->si_flags = 0; + sbinfo->si_dirwh = AUFS_DIRWH_DEF; + sbinfo->si_rdcache = AUFS_RDCACHE_DEF * HZ; + //atomic_set(&sbinfo->si_hinotify, 0); + //init_waitqueue_head(&sbinfo->si_hinotify_wq); + + sb->s_fs_info = sbinfo; + au_flag_set(sb, AuDefFlags); +#ifdef ForceInotify + udba_set(sb, AuFlag_UDBA_INOTIFY); +#endif +#ifdef ForceDlgt + au_flag_set(sb, AuFlag_DLGT); +#endif +#ifdef ForceNoPlink + au_flag_clr(sb, AuFlag_PLINK); +#endif + return 0; /* success */ + + out: + TraceErr(-ENOMEM); + return -ENOMEM; +} + +static int alloc_root(struct super_block *sb) +{ + int err; + struct inode *inode; + struct dentry *root; + + TraceEnter(); + + err = -ENOMEM; + inode = iget(sb, AUFS_ROOT_INO); + //if (LktrCond) {iput(inode); inode = NULL;} + if (unlikely(!inode)) + goto out; + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out; + err = -ENOMEM; + if (unlikely(is_bad_inode(inode))) + goto out_iput; + + root = d_alloc_root(inode); + //if (LktrCond) {igrab(inode); dput(root); root = NULL;} + if (unlikely(!root)) + goto out_iput; + err = PTR_ERR(root); + if (IS_ERR(root)) + goto out_iput; + + err = au_alloc_dinfo(root); + //if (LktrCond){rw_write_unlock(&dtodi(root)->di_rwsem);err=-1;} + if (!err) { + sb->s_root = root; + return 0; /* success */ + } + dput(root); + goto out; /* do not iput */ + + out_iput: + iput(inode); + out: + TraceErr(err); + return err; + +} + +static int aufs_fill_super(struct super_block *sb, void *raw_data, int silent) +{ + int err; + struct dentry *root; + struct inode *inode; + struct opts opts; + char *arg = raw_data; + + //au_debug_on(); + if (unlikely(!arg || !*arg)) { + err = -EINVAL; + Err("no arg\n"); + goto out; + } + LKTRTrace("%s, silent %d\n", arg, silent); + + err = -ENOMEM; + memset(&opts, 0, sizeof(opts)); + opts.opt = (void*)__get_free_page(GFP_KERNEL); + //if (LktrCond) {free_page((unsigned long)opts.opt); opts.opt = NULL;} + if (unlikely(!opts.opt)) + goto out; + opts.max_opt = PAGE_SIZE / sizeof(*opts.opt); + + err = alloc_sbinfo(sb); + //if (LktrCond) {si_write_unlock(sb);free_sbinfo(stosi(sb));err=-1;} + if (unlikely(err)) + goto out_opts; + SiMustWriteLock(sb); + /* all timestamps always follow the ones on the branch */ + sb->s_flags |= MS_NOATIME | MS_NODIRATIME; + sb->s_op = &aufs_sop; + au_init_export_op(sb); + //err = kobj_mount(stosi(sb)); + //if (err) + //goto out_info; + + err = alloc_root(sb); + //if (LktrCond) {rw_write_unlock(&dtodi(sb->s_root)->di_rwsem); + //dput(sb->s_root);sb->s_root=NULL;err=-1;} + if (unlikely(err)) { + DEBUG_ON(sb->s_root); + si_write_unlock(sb); + goto out_info; + } + root = sb->s_root; + DiMustWriteLock(root); + inode = root->d_inode; + inode->i_nlink = 2; + + /* + * actually we can parse options regardless aufs lock here. + * but at remount time, parsing must be done before aufs lock. + * so we follow the same rule. + */ + ii_write_lock_parent(inode); + aufs_write_unlock(root); + err = au_parse_opts(sb, arg, &opts); + //if (LktrCond) {au_free_opts(&opts); err = -1;} + if (unlikely(err)) + goto out_root; + + /* lock vfs_inode first, then aufs. */ + i_lock(inode); + inode->i_op = &aufs_dir_iop; + inode->i_fop = &aufs_dir_fop; + aufs_write_lock(root); + + sb->s_maxbytes = 0; + err = au_do_opts_mount(sb, &opts); + //if (LktrCond) err = -1; + au_free_opts(&opts); + if (unlikely(err)) + goto out_unlock; + DEBUG_ON(!sb->s_maxbytes); + + //DbgDentry(root); + aufs_write_unlock(root); + i_unlock(inode); + //DbgSb(sb); + goto out_opts; /* success */ + + out_unlock: + aufs_write_unlock(root); + i_unlock(inode); + out_root: + dput(root); + sb->s_root = NULL; + out_info: + free_sbinfo(stosi(sb)); + sb->s_fs_info = NULL; + out_opts: + free_page((unsigned long)opts.opt); + out: + TraceErr(err); + err = cvt_err(err); + TraceErr(err); + //au_debug_off(); + return err; +} + +/* ---------------------------------------------------------------------- */ + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) +static int aufs_get_sb(struct file_system_type *fs_type, int flags, + const char *dev_name, void *raw_data, + struct vfsmount *mnt) +{ + int err; + + /* all timestamps always follow the ones on the branch */ + //mnt->mnt_flags |= MNT_NOATIME | MNT_NODIRATIME; + err = get_sb_nodev(fs_type, flags, raw_data, aufs_fill_super, mnt); + if (!err) { + struct aufs_sbinfo *sbinfo = stosi(mnt->mnt_sb); + sbinfo->si_mnt = mnt; + sysaufs_add(sbinfo); + } + return err; +} +#else +static struct super_block *aufs_get_sb(struct file_system_type *fs_type, + int flags, const char *dev_name, + void *raw_data) +{ + return get_sb_nodev(fs_type, flags, raw_data, aufs_fill_super); +} +#endif + +struct file_system_type aufs_fs_type = { + .name = AUFS_FSTYPE, + .fs_flags = FS_REVAL_DOT, // for UDBA and NFS branch + .get_sb = aufs_get_sb, + .kill_sb = generic_shutdown_super, + //no need to __module_get() and module_put(). + .owner = THIS_MODULE, +}; diff --git a/fs/aufs/super.h b/fs/aufs/super.h new file mode 100755 index 0000000..56ddee1 --- /dev/null +++ b/fs/aufs/super.h @@ -0,0 +1,339 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: super.h,v 1.44 2007/05/14 03:39:54 sfjro Exp $ */ + +#ifndef __AUFS_SUPER_H__ +#define __AUFS_SUPER_H__ + +#ifdef __KERNEL__ + +#include +#include +#include +#include "misc.h" +#include "sysaufs.h" + +#ifdef CONFIG_AUFS_SYSAUFS +/* entries under sysfs per mount-point */ +enum {SysaufsSb_XINO, /* SysaufsSb_PLINK, */ SysaufsSb_Last}; +struct sysaufs_sbinfo { + au_subsys_t subsys; + struct sysaufs_entry array[SysaufsSb_Last]; +}; +extern sysaufs_op au_si_ops[]; +#else +struct sysaufs_sbinfo {}; +#endif + +struct aufs_sbinfo { + struct aufs_rwsem si_rwsem; + + /* branch management */ + /* wrap around attack by superuser? No. */ + int si_generation; + + /* + * set true when refresh_dirs() at remount time failed. + * then try refreshing dirs at access time again. + * if it is false, refreshing dirs at access time is unnecesary + */ + unsigned int si_failed_refresh_dirs:1; + + aufs_bindex_t si_bend; + aufs_bindex_t si_last_br_id; + struct aufs_branch **si_branch; + + /* mount flags */ + unsigned int si_flags; + + /* external inode number table */ + atomic_long_t si_xino; // time bomb + //struct file *si_xino_bmap; + + /* readdir cache time, max, in HZ */ + unsigned long si_rdcache; + + /* + * If the number of whiteouts are larger than si_dirwh, leave all of + * them after rename_whtmp to reduce the cost of rmdir(2). + * future fsck.aufs or kernel thread will remove them later. + * Otherwise, remove all whiteouts and the dir in rmdir(2). + */ + unsigned int si_dirwh; + + /* pseudo_link list */ // dirty + spinlock_t si_plink_lock; + struct list_head si_plink; + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) + /* super_blocks list is not exported */ + struct list_head si_list; + struct vfsmount *si_mnt; /* no get/put */ +#endif + + /* sysfs */ + struct sysaufs_sbinfo si_sysaufs; + +#ifdef CONFIG_AUFS_HINOTIFY + /* hinotify */ + //atomic_t si_hinotify; + //wait_queue_head_t si_hinotify_wq; +#endif + +#ifdef CONFIG_AUFS_ROBR + /* locked vma list for mmap() */ // very dirty + spinlock_t si_lvma_lock; + struct list_head si_lvma; +#endif +}; + +/* an entry in a xino file */ +struct xino { + ino_t ino; + //__u32 h_gen; +} __attribute__ ((packed)); + +//#define AuXino_INVALID_HGEN (-1) + +/* ---------------------------------------------------------------------- */ + +/* Mount flags */ +#define AuFlag_XINO 1 +#define AuFlag_ZXINO (1 << 1) +#define AuFlag_PLINK (1 << 2) +#define AuFlag_UDBA_NONE (1 << 3) +#define AuFlag_UDBA_REVAL (1 << 4) +#define AuFlag_UDBA_INOTIFY (1 << 5) +#define AuFlag_WARN_PERM (1 << 6) +#define AuFlag_COO_NONE (1 << 7) +#define AuFlag_COO_LEAF (1 << 8) +#define AuFlag_COO_ALL (1 << 9) +#define AuFlag_ALWAYS_DIROPQ (1 << 10) +#define AuFlag_DLGT (1 << 11) + +#define AuMask_UDBA (AuFlag_UDBA_NONE | AuFlag_UDBA_REVAL \ + | AuFlag_UDBA_INOTIFY) +#define AuMask_COO (AuFlag_COO_NONE | AuFlag_COO_LEAF \ + | AuFlag_COO_ALL) + +#ifdef CONFIG_AUFS_COMPAT +#define AuDefFlag_DIROPQ AuFlag_ALWAYS_DIROPQ +#else +#define AuDefFlag_DIROPQ 0 +#endif + +#define AuDefFlags_COMM (AuFlag_XINO | AuFlag_UDBA_REVAL | AuFlag_WARN_PERM \ + | AuFlag_COO_NONE | AuDefFlag_DIROPQ) +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16) +#define AuDefFlags (AuDefFlags_COMM | AuFlag_PLINK) +#else +#define AuDefFlags AuDefFlags_COMM +#endif + +/* ---------------------------------------------------------------------- */ + +/* flags for aufs_read_lock()/di_read_lock() */ +#define AUFS_D_WLOCK 1 +#define AUFS_I_RLOCK 2 +#define AUFS_I_WLOCK 4 + +/* ---------------------------------------------------------------------- */ + +/* super.c */ +int au_show_brs(struct seq_file *seq, struct super_block *sb); + +/* xino.c */ +struct file *xino_create(struct super_block *sb, char *fname, int silent, + struct dentry *parent); +ino_t xino_new_ino(struct super_block *sb); +int xino_write0(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino); +int xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino, + struct xino *xino); +int xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino, + struct xino *xino); +int xino_init(struct super_block *sb, aufs_bindex_t bindex, + struct file *base_file, int do_test); +struct opt_xino; +int xino_set(struct super_block *sb, struct opt_xino *xino, int remount); +int xino_clr(struct super_block *sb); +struct file *xino_def(struct super_block *sb); + +/* sbinfo.c */ +struct aufs_sbinfo *stosi(struct super_block *sb); +aufs_bindex_t sbend(struct super_block *sb); +struct aufs_branch *stobr(struct super_block *sb, aufs_bindex_t bindex); +int au_sigen(struct super_block *sb); +int au_sigen_inc(struct super_block *sb); +int find_bindex(struct super_block *sb, struct aufs_branch *br); + +void aufs_read_lock(struct dentry *dentry, int flags); +void aufs_read_unlock(struct dentry *dentry, int flags); +void aufs_write_lock(struct dentry *dentry); +void aufs_write_unlock(struct dentry *dentry); +void aufs_read_and_write_lock2(struct dentry *d1, struct dentry *d2, int isdir); +void aufs_read_and_write_unlock2(struct dentry *d1, struct dentry *d2); + +aufs_bindex_t new_br_id(struct super_block *sb); + +/* ---------------------------------------------------------------------- */ + +static inline const char *au_sbtype(struct super_block *sb) +{ + return sb->s_type->name; +} + +static inline int au_is_aufs(struct super_block *sb) +{ + return !strcmp(au_sbtype(sb), AUFS_FSTYPE); +} + +static inline int au_is_nfs(struct super_block *sb) +{ +#if defined(CONFIG_NFS_FS) || defined(CONFIG_NFS_FS_MODULE) + return !strcmp(au_sbtype(sb), "nfs"); +#else + return 0; +#endif +} + +static inline int au_is_remote(struct super_block *sb) +{ + return au_is_nfs(sb); +} + +#ifdef CONFIG_AUFS_EXPORT +static inline void au_init_export_op(struct super_block *sb) +{ + extern struct export_operations aufs_export_op; + sb->s_export_op = &aufs_export_op; +} + +static inline int au_is_nfsd(struct task_struct *tsk) +{ + return (!tsk->mm && !strcmp(tsk->comm, "nfsd")); +} + +static inline void au_nfsd_lockdep_off(void) +{ + /* braces are added to stop a warning */ + if (au_is_nfsd(current)) { + lockdep_off(); + } +} + +static inline void au_nfsd_lockdep_on(void) +{ + /* braces are added to stop a warning */ + if (au_is_nfsd(current)) { + lockdep_on(); + } +} +#else +static inline int au_is_nfsd(struct task_struct *tsk) +{ + return 0; +} +static inline void au_init_export_op(struct super_block *sb) +{ + /* nothing */ +} +#define au_nfsd_lockdep_off() /* */ +#define au_nfsd_lockdep_on() /* */ +#endif /* CONFIG_AUFS_EXPORT */ + +static inline void init_lvma(struct aufs_sbinfo *sbinfo) +{ +#ifdef CONFIG_AUFS_ROBR + spin_lock_init(&sbinfo->si_lvma_lock); + INIT_LIST_HEAD(&sbinfo->si_lvma); +#else + /* nothing */ +#endif +} + +/* limited support before 2.6.18 */ +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,18) +static inline void au_mntget(struct super_block *sb) +{ + mntget(stosi(sb)->si_mnt); +} + +static inline void au_mntput(struct super_block *sb) +{ + mntput(stosi(sb)->si_mnt); +} +#else +static inline void au_mntget(struct super_block *sb) +{ + /* empty */ +} + +static inline void au_mntput(struct super_block *sb) +{ + /* empty */ +} +#endif + +/* ---------------------------------------------------------------------- */ + +static inline void au_flag_set(struct super_block *sb, unsigned int flag) +{ + //SiMustWriteLock(sb); + stosi(sb)->si_flags |= flag; +} + +static inline void au_flag_clr(struct super_block *sb, unsigned int flag) +{ + //SiMustWriteLock(sb); + stosi(sb)->si_flags &= ~flag; +} + +static inline +unsigned int au_flag_test(struct super_block *sb, unsigned int flag) +{ + //SiMustAnyLock(sb); + return stosi(sb)->si_flags & flag; +} + +static inline unsigned int au_flag_test_udba(struct super_block *sb) +{ + return au_flag_test(sb, AuMask_UDBA); +} + +static inline unsigned int au_flag_test_coo(struct super_block *sb) +{ + return au_flag_test(sb, AuMask_COO); +} + +/* ---------------------------------------------------------------------- */ + +/* lock superblock. mainly for entry point functions */ +/* + * si_read_lock, si_write_lock, + * si_read_unlock, si_write_unlock, si_downgrade_lock + */ +SimpleRwsemFuncs(si, struct super_block *sb, stosi(sb)->si_rwsem); + +/* to debug easier, do not make them inlined functions */ +#define SiMustReadLock(sb) RwMustReadLock(&stosi(sb)->si_rwsem) +#define SiMustWriteLock(sb) RwMustWriteLock(&stosi(sb)->si_rwsem) +#define SiMustAnyLock(sb) RwMustAnyLock(&stosi(sb)->si_rwsem) + +#endif /* __KERNEL__ */ +#endif /* __AUFS_SUPER_H__ */ diff --git a/fs/aufs/sysaufs.c b/fs/aufs/sysaufs.c new file mode 100755 index 0000000..d686862 --- /dev/null +++ b/fs/aufs/sysaufs.c @@ -0,0 +1,620 @@ +/* + * Copyright (C) 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: sysaufs.c,v 1.6 2007/05/14 03:40:10 sfjro Exp $ */ + +#include +#include +#include +#include "aufs.h" + +/* ---------------------------------------------------------------------- */ + +/* super_blocks list is not exported */ +static DEFINE_MUTEX(aufs_sbilist_mtx); +static LIST_HEAD(aufs_sbilist); + +/* ---------------------------------------------------------------------- */ + +typedef ssize_t (*rwfunc_t)(struct kobject *kobj, char *buf, loff_t offset, + size_t sz, struct sysaufs_args *args); +static ssize_t sysaufs_read(struct kobject *kobj, char *buf, loff_t offset, + size_t sz, struct sysaufs_args *args); +static ssize_t sysaufs_free_write(struct kobject *kobj, char *buf, loff_t + offset, size_t sz, struct sysaufs_args *args); + +#define GFunc(name, _index, func) \ +static ssize_t name(struct kobject *kobj, char *buf, loff_t offset, size_t sz) \ +{ \ + struct sysaufs_args args = { \ + .index = (_index), \ + .mtx = &aufs_sbilist_mtx, \ + .sb = NULL \ + }; \ + return func(kobj, buf, offset, sz, &args); \ +} + +#define GFuncs(name, _index) \ + GFunc(read_##name, _index, sysaufs_read); \ + GFunc(write_##name, _index, sysaufs_free_write); + +static struct super_block *find_sb_locked(struct kobject *kobj) +{ + struct super_block *sb; + struct aufs_sbinfo *sbinfo; + + TraceEnter(); + MtxMustLock(&aufs_sbilist_mtx); + + sb = NULL; + list_for_each_entry(sbinfo, &aufs_sbilist, si_list) { + if (&au_subsys_to_kset(sbinfo->si_sysaufs.subsys).kobj != kobj) + continue; + sb = sbinfo->si_mnt->mnt_sb; + si_read_lock(sb); + break; + } + return sb; +} + +static ssize_t sb_func(struct kobject *kobj, char *buf, loff_t offset, + size_t sz, struct sysaufs_args *args, rwfunc_t func) +{ + ssize_t err; + + err = -ENOENT; + mutex_lock(&aufs_sbilist_mtx); + args->sb = find_sb_locked(kobj); + if (args->sb) { + err = func(kobj, buf, offset, sz, args); + si_read_unlock(args->sb); + } + mutex_unlock(&aufs_sbilist_mtx); + return err; +} + +#define SbFunc(name, _index, func) \ +static ssize_t name(struct kobject *kobj, char *buf, loff_t offset, size_t sz) \ +{ \ + struct sysaufs_args args = { \ + .index = (_index), \ + .mtx = NULL \ + }; \ + return sb_func(kobj, buf, offset, sz, &args, func); \ +} + +#define SbFuncs(name, index) \ + SbFunc(read_##name, index, sysaufs_read); \ + SbFunc(write_##name, index, sysaufs_free_write) + +static decl_subsys(aufs, NULL, NULL); +enum {Brs, Stat, Config, _Last}; +static struct sysaufs_entry g_array[_Last]; +GFuncs(brs, Brs); +GFuncs(stat, Stat); +GFuncs(config, Config); + +SbFuncs(xino, SysaufsSb_XINO); + +#define SetEntry(e, _name, init_size, _ops) \ + do { \ + (e)->attr.attr.name = #_name; \ + (e)->attr.attr.owner = THIS_MODULE; \ + (e)->attr.attr.mode = S_IRUGO | S_IWUSR; \ + (e)->attr.read = read_##_name; \ + (e)->attr.write = write_##_name; \ + (e)->allocated = init_size; \ + (e)->err = -1; \ + (e)->ops = _ops; \ + } while (0) + +#define Priv(e) (e)->attr.private +#define Allocated(e) (e)->allocated +#define Len(e) (e)->attr.size +#define Name(e) attr_name((e)->attr) + +/* ---------------------------------------------------------------------- */ + +static void free_entry(struct sysaufs_entry *e) +{ + MtxMustLock(&aufs_sbilist_mtx); + DEBUG_ON(!Priv(e)); + + if (Allocated(e) > 0) + kfree(Priv(e)); + else + free_pages((unsigned long)Priv(e), -Allocated(e)); + Priv(e) = NULL; + Len(e) = 0; +} + +static void free_entries(void) +{ + static int a[] = {Brs, -1}; + int *p = a; + + MtxMustLock(&aufs_sbilist_mtx); + + while (*p >= 0) { + if (Priv(g_array + *p)) + free_entry(g_array + *p); + p++; + } +} + +static int alloc_entry(struct sysaufs_entry *e) +{ + MtxMustLock(&aufs_sbilist_mtx); + DEBUG_ON(Priv(e)); + //Dbg("%d\n", Allocated(e)); + + if (Allocated(e) > 0) + Priv(e) = kmalloc(Allocated(e), GFP_KERNEL); + else + Priv(e) = (void*)__get_free_pages(GFP_KERNEL, -Allocated(e)); + if (Priv(e)) + return 0; + return -ENOMEM; +} + +/* ---------------------------------------------------------------------- */ + +static void unreg(au_subsys_t *subsys, struct sysaufs_entry *a, int n, + au_subsys_t *parent) +{ + int i; + + TraceEnter(); + + for (i = 0; i < n; i++, a++) + if (!a->err) { + sysfs_remove_bin_file + (&au_subsys_to_kset(*subsys).kobj, &a->attr); + if (Priv(a)) + free_entry(a); + } + + subsystem_unregister(subsys); + subsys_put(parent); +} + +static int reg(au_subsys_t *subsys, struct sysaufs_entry *a, int n, + au_subsys_t *parent) +{ + int err, i; + + TraceEnter(); + + subsys_get(parent); + kobj_set_kset_s(&au_subsys_to_kset(*subsys), *parent); + err = subsystem_register(subsys); + if (unlikely(err)) + goto out; + + for (i = 0; !err && i < n; i++) + err = a[i].err = sysfs_create_bin_file + (&au_subsys_to_kset(*subsys).kobj, &a[i].attr); + if (unlikely(err)) + unreg(subsys, a, n, parent); + + out: + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +#define SbSetEntry(index, name, init_size) \ + SetEntry(sa->array + index, name, init_size, au_si_ops); + +void sysaufs_add(struct aufs_sbinfo *sbinfo) +{ + int err; + struct sysaufs_sbinfo *sa = &sbinfo->si_sysaufs; + + TraceEnter(); + + mutex_lock(&aufs_sbilist_mtx); + list_add_tail(&sbinfo->si_list, &aufs_sbilist); + free_entries(); + + memset(sa, 0, sizeof(*sa)); + SbSetEntry(SysaufsSb_XINO, xino, 128); + err = kobject_set_name(&au_subsys_to_kset(sa->subsys).kobj, "%p", + sbinfo->si_mnt->mnt_sb); + if (!err) + err = reg(&sa->subsys, sa->array, ARRAY_SIZE(sa->array), + &aufs_subsys); + if (unlikely(err)) + Warn("failed adding sysfs (%d)\n", err); + + mutex_unlock(&aufs_sbilist_mtx); +} + +void sysaufs_del(struct aufs_sbinfo *sbinfo) +{ + struct sysaufs_sbinfo *sa = &sbinfo->si_sysaufs; + + TraceEnter(); + + mutex_lock(&aufs_sbilist_mtx); + unreg(&sa->subsys, sa->array, ARRAY_SIZE(sa->array), &aufs_subsys); + list_del(&sbinfo->si_list); + free_entries(); + mutex_unlock(&aufs_sbilist_mtx); +} + +void sysaufs_notify_remount(void) +{ + mutex_lock(&aufs_sbilist_mtx); + free_entries(); + mutex_unlock(&aufs_sbilist_mtx); +} + +/* ---------------------------------------------------------------------- */ + +static int make_brs(struct seq_file *seq, struct sysaufs_args *args, + int *do_size) +{ + int err; + struct aufs_sbinfo *sbinfo; + + TraceEnter(); + MtxMustLock(&aufs_sbilist_mtx); + DEBUG_ON(args->index != Brs); + + err = 0; + list_for_each_entry(sbinfo, &aufs_sbilist, si_list) { + struct super_block *sb; + struct dentry *root; + struct vfsmount *mnt; + + sb = sbinfo->si_mnt->mnt_sb; + root = sb->s_root; + aufs_read_lock(root, !AUFS_I_RLOCK); + mnt = sbinfo->si_mnt; + err = seq_escape + (seq, mnt->mnt_devname ? mnt->mnt_devname : "none", + au_esc_chars); + if (!err) + err = seq_putc(seq, ' '); + if (!err) + err = seq_path(seq, mnt, root, au_esc_chars); + if (err > 0) + err = seq_printf(seq, " %p br:", sb); + if (!err) + err = au_show_brs(seq, sb); + aufs_read_unlock(root, !AUFS_I_RLOCK); + if (!err) + err = seq_putc(seq, '\n'); + else + break; + } + + TraceErr(err); + return err; +} + +static int make_config(struct seq_file *seq, struct sysaufs_args *args, + int *do_size) +{ + int err; + + TraceEnter(); + DEBUG_ON(args->index != Config); + +#ifdef CONFIG_AUFS + err = seq_puts(seq, "CONFIG_AUFS=y\n"); +#else + err = seq_puts(seq, "CONFIG_AUFS=m\n"); +#endif + +#define puts(m, v) \ + if (!err) err = seq_puts(seq, "CONFIG_AUFS_" #m "=" #v "\n") +#define puts_bool(m) puts(m, y) +#define puts_mod(m) puts(m, m) + +#ifdef CONFIG_AUFS_FAKE_DM + puts_bool(FAKE_DM); +#endif +#ifdef CONFIG_AUFS_BRANCH_MAX_127 + puts_bool(BRANCH_MAX_127); +#elif defined(CONFIG_AUFS_BRANCH_MAX_511) + puts_bool(BRANCH_MAX_511); +#elif defined(CONFIG_AUFS_BRANCH_MAX_1023) + puts_bool(BRANCH_MAX_1023); +#elif defined(CONFIG_AUFS_BRANCH_MAX_32767) + puts_bool(BRANCH_MAX_32767); +#endif + puts_bool(SYSAUFS); +#ifdef CONFIG_AUFS_HINOTIFY + puts_bool(HINOTIFY); +#endif +#ifdef CONFIG_AUFS_EXPORT + puts_bool(EXPORT); +#endif +#ifdef CONFIG_AUFS_ROBR + puts_bool(ROBR); +#endif +#ifdef CONFIG_AUFS_DLGT + puts_bool(DLGT); +#endif +#ifdef CONFIG_AUFS_LHASH_PATCH + puts_bool(LHASH_PATCH); +#endif +#ifdef CONFIG_AUFS_KSIZE_PATCH + puts_bool(KSIZE_PATCH); +#endif +#ifdef CONFIG_AUFS_DEBUG + puts_bool(DEBUG); +#endif +#ifdef CONFIG_AUFS_COMPAT + puts_bool(COMPAT); +#endif + +#undef puts_bool +#undef puts + + TraceErr(err); + return err; +} + +static int make_stat(struct seq_file *seq, struct sysaufs_args *args, + int *do_size) +{ + int err, i; + + TraceEnter(); + DEBUG_ON(args->index != Stat); + + *do_size = 0; + err = seq_puts(seq, "wkq max_busy:"); + for (i = 0; !err && i < aufs_nwkq; i++) + err = seq_printf(seq, " %u", au_wkq[i].max_busy); + if (!err) + err = seq_printf(seq, ", %u(generic)\n", + au_wkq[aufs_nwkq].max_busy); + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +static int make(struct sysaufs_entry *e, struct sysaufs_args *args, + int *do_size) + +{ + int err; + struct seq_file *seq; + + TraceEnter(); + DEBUG_ON(Priv(e)); + MtxMustLock(&aufs_sbilist_mtx); + + err = -ENOMEM; + seq = kzalloc(sizeof(*seq), GFP_KERNEL); + if (unlikely(!seq)) + goto out; + + Len(e) = 0; + while (1) { + err = alloc_entry(e); + if (unlikely(err)) + break; + + //mutex_init(&seq.lock); + seq->buf = Priv(e); + seq->count = 0; + seq->size = Allocated(e); + if (unlikely(Allocated(e) <= 0)) + seq->size = PAGE_SIZE << -Allocated(e); + + err = e->ops[args->index](seq, args, do_size); + if (!err) { + Len(e) = seq->count; + break; /* success */ + } + + free_entry(e); + if (Allocated(e) > 0) { + Allocated(e) <<= 1; + if (unlikely(Allocated(e) >= (int)PAGE_SIZE)) + Allocated(e) = 0; + } else + Allocated(e)--; + //Dbg("%d\n", Allocated(e)); + } + kfree(seq); + + out: + TraceErr(err); + return err; +} + +/* why does sysfs pass my parent kobject? */ +static struct dentry *find_me(struct dentry *parent, struct sysaufs_entry *e) +{ +#if 1 + struct dentry *dentry; + const char *name = Name(e); + const unsigned int len = strlen(name); + + //Dbg("%.*s\n", DLNPair(parent)); + spin_lock(&dcache_lock); + list_for_each_entry(dentry, &parent->d_subdirs, D_CHILD) { + //Dbg("%.*s\n", DLNPair(dentry)); + if (len == dentry->d_name.len + && !strcmp(dentry->d_name.name, name)) { + spin_unlock(&dcache_lock); + return dentry; + } + } + spin_unlock(&dcache_lock); +#endif + return NULL; +} + +static ssize_t sysaufs_read(struct kobject *kobj, char *buf, loff_t offset, + size_t sz, struct sysaufs_args *args) +{ + ssize_t err; + loff_t len; + struct dentry *d; + struct sysaufs_entry *e; + int do_size; + + LKTRTrace("{%d, %p}, offset %Ld, sz %lu\n", + args->index, args->sb, offset, (unsigned long)sz); + + if (unlikely(!sz)) + return 0; + + err = 0; + d = NULL; + e = g_array + args->index; + if (args->sb) + e = stosi(args->sb)->si_sysaufs.array + args->index; + + do_size = 1; + if (args->mtx) + mutex_lock(args->mtx); + if (unlikely(!Priv(e))) { + err = make(e, args, &do_size); + DEBUG_ON(Len(e) > INT_MAX); + if (do_size) { + d = find_me(kobj->dentry, e); + if (d) + i_size_write(d->d_inode, Len(e)); + } + } + + if (!err) { + err = len = Len(e) - offset; + LKTRTrace("%Ld\n", len); + if (len > 0) { + if (len > sz) + err = sz; + memcpy(buf, Priv(e) + offset, err); + } + + if (!do_size) + free_entry(e); + } + if (args->mtx) + mutex_unlock(args->mtx); + + TraceErr(err); + return err; +} + +static ssize_t sysaufs_free_write(struct kobject *kobj, char *buf, + loff_t offset, size_t sz, + struct sysaufs_args *args) +{ + struct dentry *d; + int allocated, len; + struct sysaufs_entry *e; + + LKTRTrace("{%d, %p}\n", args->index, args->sb); + + e = g_array + args->index; + if (args->sb) + e = stosi(args->sb)->si_sysaufs.array + args->index; + + if (args->mtx) + mutex_lock(args->mtx); + if (Priv(e)) { + allocated = Allocated(e); + if (unlikely(allocated <= 0)) + allocated = PAGE_SIZE << -allocated; + allocated >>= 1; + len = Len(e); + + free_entry(e); + if (unlikely(len <= allocated)) { + if (Allocated(e) >= 0) + Allocated(e) = allocated; + else + Allocated(e)++; + } + + d = find_me(kobj->dentry, e); + if (d && i_size_read(d->d_inode)) + i_size_write(d->d_inode, 0); + } + if (args->mtx) + mutex_unlock(args->mtx); + + return sz; +} + +static sysaufs_op g_ops[] = { + [Brs] = make_brs, + [Stat] = make_stat, + [Config] = make_config +}; + +/* ---------------------------------------------------------------------- */ + +#define GSetEntry(index, name, init_size) \ + SetEntry(g_array + index, name, init_size, g_ops) + +int __init sysaufs_init(void) +{ + int err; + + GSetEntry(Brs, brs, 128); + GSetEntry(Stat, stat, 32); + GSetEntry(Config, config, 256); + err = reg(&aufs_subsys, g_array, ARRAY_SIZE(g_array), &fs_subsys); + TraceErr(err); + return err; +} + +void __exit sysaufs_fin(void) +{ + mutex_lock(&aufs_sbilist_mtx); + unreg(&aufs_subsys, g_array, ARRAY_SIZE(g_array), &fs_subsys); + mutex_unlock(&aufs_sbilist_mtx); +} + +/* ---------------------------------------------------------------------- */ + +#ifdef DbgDlgt +int is_branch(struct super_block *h_sb) +{ + int found = 0; + struct aufs_sbinfo *sbinfo; + + //Dbg("here\n"); + mutex_lock(&aufs_sbilist_mtx); + list_for_each_entry(sbinfo, &aufs_sbilist, si_list) { + aufs_bindex_t bindex, bend; + struct super_block *sb; + + sb = sbinfo->si_mnt->mnt_sb; + si_read_lock(sb); + bend = sbend(sb); + for (bindex = 0; !found && bindex <= bend; bindex++) + found = (h_sb == sbr_sb(sb, bindex)); + si_read_unlock(sb); + } + mutex_unlock(&aufs_sbilist_mtx); + return found; +} +#endif diff --git a/fs/aufs/sysaufs.h b/fs/aufs/sysaufs.h new file mode 100755 index 0000000..cf0247f --- /dev/null +++ b/fs/aufs/sysaufs.h @@ -0,0 +1,83 @@ +/* + * Copyright (C) 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: sysaufs.h,v 1.3 2007/05/14 06:27:18 sfjro Exp $ */ + +#ifndef __SYSAUFS_H__ +#define __SYSAUFS_H__ + +#ifdef __KERNEL__ + +#include +#include +#include + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22) +typedef struct kset au_subsys_t; +#define au_subsys_to_kset(subsys) (subsys) +#else +typedef struct subsystem au_subsys_t; +#define au_subsys_to_kset(subsys) ((subsys).kset) +#endif + +/* ---------------------------------------------------------------------- */ + +/* arguments for an entry under sysfs */ +struct sysaufs_args { + int index; + struct mutex *mtx; + struct super_block *sb; +}; + +typedef int (*sysaufs_op)(struct seq_file *seq, struct sysaufs_args *args, + int *do_size); + +/* an entry under sysfs */ +struct sysaufs_entry { + struct bin_attribute attr; + int allocated; /* zero minus means pages */ + int err; + sysaufs_op *ops; +}; + +/* ---------------------------------------------------------------------- */ + +struct aufs_sbinfo; +#ifdef CONFIG_AUFS_SYSAUFS +void sysaufs_add(struct aufs_sbinfo *sbinfo); +void sysaufs_del(struct aufs_sbinfo *sbinfo); +int __init sysaufs_init(void); +void sysaufs_fin(void); +void sysaufs_notify_remount(void); +#else +static inline void sysaufs_add(struct aufs_sbinfo *sbinfo) +{ + /* nothing */ +} + +static inline void sysaufs_del(struct aufs_sbinfo *sbinfo) +{ + /* nothing */ +} +#define sysaufs_init() 0 +#define sysaufs_fin() /* */ +#define sysaufs_notify_remount() /* */ +#endif /* CONFIG_AUFS_SYSAUFS */ + +#endif /* __KERNEL__ */ +#endif /* __SYSAUFS_H__ */ diff --git a/fs/aufs/vdir.c b/fs/aufs/vdir.c new file mode 100755 index 0000000..8e99b7d --- /dev/null +++ b/fs/aufs/vdir.c @@ -0,0 +1,802 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: vdir.c,v 1.22 2007/05/14 03:38:52 sfjro Exp $ */ + +#include "aufs.h" + +static int calc_size(int namelen) +{ + int sz; + + sz = sizeof(struct aufs_de) + namelen; + if (sizeof(ino_t) == sizeof(long)) { + const int mask = sizeof(ino_t) - 1; + if (sz & mask) { + sz += sizeof(ino_t); + sz &= ~mask; + } + } else { +#if 0 // remove + BUG(); + // this block will be discarded by optimizer. + int m; + m = sz % sizeof(ino_t); + if (m) + sz += sizeof(ino_t) - m; +#endif + } + + DEBUG_ON(sz % sizeof(ino_t)); + return sz; +} + +static int set_deblk_end(union aufs_deblk_p *p, union aufs_deblk_p *deblk_end) +{ + if (calc_size(0) <= deblk_end->p - p->p) { + p->de->de_str.len = 0; + //smp_mb(); + return 0; + } + return -1; // error +} + +/* returns true or false */ +static int is_deblk_end(union aufs_deblk_p *p, union aufs_deblk_p *deblk_end) +{ + if (calc_size(0) <= deblk_end->p - p->p) + return !p->de->de_str.len; + return 1; +} + +static aufs_deblk_t *last_deblk(struct aufs_vdir *vdir) +{ + return vdir->vd_deblk[vdir->vd_nblk - 1]; +} + +void nhash_init(struct aufs_nhash *nhash) +{ + int i; + for (i = 0; i < AUFS_NHASH_SIZE; i++) + INIT_HLIST_HEAD(nhash->heads + i); +} + +struct aufs_nhash *nhash_new(gfp_t gfp) +{ + struct aufs_nhash *nhash; + + nhash = kmalloc(sizeof(*nhash), gfp); + if (nhash) { + nhash_init(nhash); + return nhash; + } + return ERR_PTR(-ENOMEM); +} + +void nhash_del(struct aufs_nhash *nhash) +{ + nhash_fin(nhash); + kfree(nhash); +} + +void nhash_move(struct aufs_nhash *dst, struct aufs_nhash *src) +{ + int i; + + TraceEnter(); + + //DbgWhlist(src); + *dst = *src; + for (i = 0; i < AUFS_NHASH_SIZE; i++) { + struct hlist_head *h; + h = dst->heads + i; + if (h->first) + h->first->pprev = &h->first; + INIT_HLIST_HEAD(src->heads + i); + } + //DbgWhlist(src); + //DbgWhlist(dst); + //smp_mb(); +} + +/* ---------------------------------------------------------------------- */ + +void nhash_fin(struct aufs_nhash *whlist) +{ + int i; + struct hlist_head *head; + struct aufs_wh *tpos; + struct hlist_node *pos, *n; + + TraceEnter(); + + for (i = 0; i < AUFS_NHASH_SIZE; i++) { + head = whlist->heads + i; + hlist_for_each_entry_safe(tpos, pos, n, head, wh_hash) { + //hlist_del(pos); + kfree(tpos); + } + } +} + +int is_longer_wh(struct aufs_nhash *whlist, aufs_bindex_t btgt, int limit) +{ + int n, i; + struct hlist_head *head; + struct aufs_wh *tpos; + struct hlist_node *pos; + + LKTRTrace("limit %d\n", limit); + //return 1; + + n = 0; + for (i = 0; i < AUFS_NHASH_SIZE; i++) { + head = whlist->heads + i; + hlist_for_each_entry(tpos, pos, head, wh_hash) + if (tpos->wh_bindex == btgt && ++n > limit) + return 1; + } + return 0; +} + +/* returns found(true) or not */ +int test_known_wh(struct aufs_nhash *whlist, char *name, int namelen) +{ + struct hlist_head *head; + struct aufs_wh *tpos; + struct hlist_node *pos; + struct aufs_destr *str; + + LKTRTrace("%.*s\n", namelen, name); + + head = whlist->heads + au_name_hash(name, namelen); + hlist_for_each_entry(tpos, pos, head, wh_hash) { + str = &tpos->wh_str; + LKTRTrace("%.*s\n", str->len, str->name); + if (str->len == namelen && !memcmp(str->name, name, namelen)) + return 1; + } + return 0; +} + +int append_wh(struct aufs_nhash *whlist, char *name, int namelen, + aufs_bindex_t bindex) +{ + int err; + struct aufs_destr *str; + struct aufs_wh *wh; + + LKTRTrace("%.*s\n", namelen, name); + + err = -ENOMEM; + wh = kmalloc(sizeof(*wh) + namelen, GFP_KERNEL); + if (unlikely(!wh)) + goto out; + err = 0; + wh->wh_bindex = bindex; + str = &wh->wh_str; + str->len = namelen; + memcpy(str->name, name, namelen); + hlist_add_head(&wh->wh_hash, + whlist->heads + au_name_hash(name, namelen)); + //smp_mb(); + + out: + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +void free_vdir(struct aufs_vdir *vdir) +{ + aufs_deblk_t **deblk; + + TraceEnter(); + + deblk = vdir->vd_deblk; + while (vdir->vd_nblk--) { + kfree(*deblk); + deblk++; + } + kfree(vdir->vd_deblk); + cache_free_vdir(vdir); +} + +static int append_deblk(struct aufs_vdir *vdir) +{ + int err, sz, i; + aufs_deblk_t **o; + union aufs_deblk_p p, deblk_end; + + TraceEnter(); + + err = -ENOMEM; + sz = sizeof(*o) * vdir->vd_nblk; + o = au_kzrealloc(vdir->vd_deblk, sz, sz + sizeof(*o), GFP_KERNEL); + if (unlikely(!o)) + goto out; + vdir->vd_deblk = o; + p.deblk = kmalloc(sizeof(*p.deblk), GFP_KERNEL); + if (p.deblk) { + i = vdir->vd_nblk++; + vdir->vd_deblk[i] = p.deblk; + vdir->vd_last.i = i; + vdir->vd_last.p.p = p.p; + deblk_end.deblk = p.deblk + 1; + err = set_deblk_end(&p, &deblk_end); + DEBUG_ON(err); + } + + out: + TraceErr(err); + return err; +} + +static struct aufs_vdir *alloc_vdir(void) +{ + struct aufs_vdir *vdir; + int err; + + TraceEnter(); + + err = -ENOMEM; + vdir = cache_alloc_vdir(); + if (unlikely(!vdir)) + goto out; + vdir->vd_deblk = kzalloc(sizeof(*vdir->vd_deblk), GFP_KERNEL); + if (unlikely(!vdir->vd_deblk)) + goto out_free; + + vdir->vd_nblk = 0; + vdir->vd_version = 0; + vdir->vd_jiffy = 0; + err = append_deblk(vdir); + if (!err) + return vdir; /* success */ + + kfree(vdir->vd_deblk); + + out_free: + cache_free_vdir(vdir); + out: + vdir = ERR_PTR(err); + TraceErrPtr(vdir); + return vdir; +} + +static int reinit_vdir(struct aufs_vdir *vdir) +{ + int err; + union aufs_deblk_p p, deblk_end; + + TraceEnter(); + + while (vdir->vd_nblk > 1) { + kfree(vdir->vd_deblk[vdir->vd_nblk - 1]); + vdir->vd_deblk[vdir->vd_nblk - 1] = NULL; + vdir->vd_nblk--; + } + p.deblk = vdir->vd_deblk[0]; + deblk_end.deblk = p.deblk + 1; + err = set_deblk_end(&p, &deblk_end); + DEBUG_ON(err); + vdir->vd_version = 0; + vdir->vd_jiffy = 0; + vdir->vd_last.i = 0; + vdir->vd_last.p.deblk = vdir->vd_deblk[0]; + //smp_mb(); + //DbgVdir(vdir); + return err; +} + +/* ---------------------------------------------------------------------- */ + +static void free_dehlist(struct aufs_nhash *dehlist) +{ + int i; + struct hlist_head *head; + struct aufs_dehstr *tpos; + struct hlist_node *pos, *n; + + TraceEnter(); + + for (i = 0; i < AUFS_NHASH_SIZE; i++) { + head = dehlist->heads + i; + hlist_for_each_entry_safe(tpos, pos, n, head, hash) { + //hlist_del(pos); + cache_free_dehstr(tpos); + } + } +} + +/* returns found(true) or not */ +static int test_known(struct aufs_nhash *delist, char *name, int namelen) +{ + struct hlist_head *head; + struct aufs_dehstr *tpos; + struct hlist_node *pos; + struct aufs_destr *str; + + LKTRTrace("%.*s\n", namelen, name); + + head = delist->heads + au_name_hash(name, namelen); + hlist_for_each_entry(tpos, pos, head, hash) { + str = tpos->str; + LKTRTrace("%.*s\n", str->len, str->name); + if (str->len == namelen && !memcmp(str->name, name, namelen)) + return 1; + } + return 0; + +} + +static int append_de(struct aufs_vdir *vdir, char *name, int namelen, ino_t ino, + unsigned int d_type, struct aufs_nhash *delist) +{ + int err, sz; + union aufs_deblk_p p, *room, deblk_end; + struct aufs_dehstr *dehstr; + + LKTRTrace("%.*s %d, i%lu, dt%u\n", namelen, name, namelen, ino, d_type); + + p.deblk = last_deblk(vdir); + deblk_end.deblk = p.deblk + 1; + room = &vdir->vd_last.p; + DEBUG_ON(room->p < p.p || deblk_end.p <= room->p + || !is_deblk_end(room, &deblk_end)); + + sz = calc_size(namelen); + if (unlikely(sz > deblk_end.p - room->p)) { + err = append_deblk(vdir); + if (unlikely(err)) + goto out; + p.deblk = last_deblk(vdir); + deblk_end.deblk = p.deblk + 1; + //smp_mb(); + DEBUG_ON(room->p != p.p); + } + + err = -ENOMEM; + dehstr = cache_alloc_dehstr(); + if (unlikely(!dehstr)) + goto out; + dehstr->str = &room->de->de_str; + hlist_add_head(&dehstr->hash, + delist->heads + au_name_hash(name, namelen)); + + room->de->de_ino = ino; + room->de->de_type = d_type; + room->de->de_str.len = namelen; + memcpy(room->de->de_str.name, name, namelen); + + err = 0; + room->p += sz; + if (unlikely(set_deblk_end(room, &deblk_end))) + err = append_deblk(vdir); + //smp_mb(); + + out: + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +struct fillvdir_arg { + struct file *file; + struct aufs_vdir *vdir; + struct aufs_nhash *delist; + struct aufs_nhash *whlist; + aufs_bindex_t bindex; + int err; + int called; +}; + +static int fillvdir(void *__arg, const char *__name, int namelen, loff_t offset, + filldir_ino_t h_ino, unsigned int d_type) +{ + struct fillvdir_arg *arg = __arg; + char *name = (void*)__name; + aufs_bindex_t bindex, bend; + struct xino xino; + struct super_block *sb; + + LKTRTrace("%.*s, namelen %d, i%Lu, dt%u\n", + namelen, name, namelen, (u64)h_ino, d_type); + + sb = arg->file->f_dentry->d_sb; + bend = arg->bindex; + arg->err = 0; + arg->called++; + //smp_mb(); + if (namelen <= AUFS_WH_PFX_LEN + || memcmp(name, AUFS_WH_PFX, AUFS_WH_PFX_LEN)) { + for (bindex = 0; bindex < bend; bindex++) + if (test_known(arg->delist + bindex, name, namelen) + || test_known_wh(arg->whlist + bindex, name, + namelen)) + goto out; /* already exists or whiteouted */ + + arg->err = xino_read(sb, bend, h_ino, &xino); + if (!arg->err && !xino.ino) { + //struct inode *h_inode; + xino.ino = xino_new_ino(sb); + if (unlikely(!xino.ino)) + arg->err = -EIO; +#if 0 + //xino.h_gen = AuXino_INVALID_HGEN; + h_inode = ilookup(sbr_sb(sb, bend), h_ino); + if (h_inode) { + if (!is_bad_inode(h_inode)) { + xino.h_gen = h_inode->i_generation; + WARN_ON(xino.h_gen == AuXino_INVALID_HGEN); + } + iput(h_inode); + } +#endif + arg->err = xino_write(sb, bend, h_ino, &xino); + } + if (!arg->err) + arg->err = append_de(arg->vdir, name, namelen, xino.ino, + d_type, arg->delist + bend); + } else { + name += AUFS_WH_PFX_LEN; + namelen -= AUFS_WH_PFX_LEN; + for (bindex = 0; bindex < bend; bindex++) + if (test_known_wh(arg->whlist + bend, name, namelen)) + goto out; /* already whiteouted */ + arg->err = append_wh(arg->whlist + bend, name, namelen, bend); + } + + out: + if (!arg->err) + arg->vdir->vd_jiffy = jiffies; + //smp_mb(); + TraceErr(arg->err); + return arg->err; +} + +static int read_vdir(struct file *file, int may_read) +{ + int err, do_read, dlgt; + struct dentry *dentry; + struct inode *inode; + struct aufs_vdir *vdir, *allocated; + unsigned long expire; + struct fillvdir_arg arg; + aufs_bindex_t bindex, bend, bstart; + struct super_block *sb; + + dentry = file->f_dentry; + LKTRTrace("%.*s, may %d\n", DLNPair(dentry), may_read); + FiMustWriteLock(file); + inode = dentry->d_inode; + IMustLock(inode); + IiMustWriteLock(inode); + DEBUG_ON(!S_ISDIR(inode->i_mode)); + + err = 0; + allocated = NULL; + do_read = 0; + sb = inode->i_sb; + expire = stosi(sb)->si_rdcache; + vdir = ivdir(inode); + if (!vdir) { + DEBUG_ON(fvdir_cache(file)); + do_read = 1; + vdir = alloc_vdir(); + err = PTR_ERR(vdir); + if (IS_ERR(vdir)) + goto out; + err = 0; + allocated = vdir; + } else if (may_read + && (inode->i_version != vdir->vd_version + || time_after(jiffies, vdir->vd_jiffy + expire))) { + LKTRTrace("iver %lu, vdver %lu, exp %lu\n", + inode->i_version, vdir->vd_version, + vdir->vd_jiffy + expire); + do_read = 1; + err = reinit_vdir(vdir); + if (unlikely(err)) + goto out; + } + //DbgVdir(vdir); goto out; + + if (!do_read) + return 0; /* success */ + + err = -ENOMEM; + bend = fbend(file); + arg.delist = kmalloc(sizeof(*arg.delist) * (bend + 1), GFP_KERNEL); + if (unlikely(!arg.delist)) + goto out_vdir; + arg.whlist = kmalloc(sizeof(*arg.whlist) * (bend + 1), GFP_KERNEL); + if (unlikely(!arg.whlist)) + goto out_delist; + err = 0; + for (bindex = 0; bindex <= bend; bindex++) { + nhash_init(arg.delist + bindex); + nhash_init(arg.whlist + bindex); + } + + dlgt = need_dlgt(sb); + arg.file = file; + arg.vdir = vdir; + bstart = fbstart(file); + for (bindex = bstart; !err && bindex <= bend; bindex++) { + struct file *hf; + struct inode *h_inode; + + hf = au_h_fptr_i(file, bindex); + if (!hf) + continue; + + h_inode = hf->f_dentry->d_inode; + //hf->f_pos = 0; + arg.bindex = bindex; + do { + arg.err = 0; + arg.called = 0; + //smp_mb(); + err = vfsub_readdir(hf, fillvdir, &arg, dlgt); + if (err >= 0) + err = arg.err; + } while (!err && arg.called); + } + + for (bindex = bstart; bindex <= bend; bindex++) { + free_dehlist(arg.delist + bindex); + nhash_fin(arg.whlist + bindex); + } + kfree(arg.whlist); + + out_delist: + kfree(arg.delist); + out_vdir: + if (!err) { + //file->f_pos = 0; + vdir->vd_version = inode->i_version; + vdir->vd_last.i = 0; + vdir->vd_last.p.deblk = vdir->vd_deblk[0]; + if (allocated) + set_ivdir(inode, allocated); + } else if (allocated) + free_vdir(allocated); + //DbgVdir(vdir); goto out; + + out: + TraceErr(err); + return err; +} + +static int copy_vdir(struct aufs_vdir *tgt, struct aufs_vdir *src) +{ + int err, i, rerr, n; + + TraceEnter(); + DEBUG_ON(tgt->vd_nblk != 1); + //DbgVdir(tgt); + + err = -ENOMEM; + if (tgt->vd_nblk < src->vd_nblk) { + aufs_deblk_t **p; + p = au_kzrealloc(tgt->vd_deblk, sizeof(*p) * tgt->vd_nblk, + sizeof(*p) * src->vd_nblk, GFP_KERNEL); + if (unlikely(!p)) + goto out; + tgt->vd_deblk = p; + } + + n = tgt->vd_nblk = src->vd_nblk; + memcpy(tgt->vd_deblk[0], src->vd_deblk[0], AUFS_DEBLK_SIZE); + //tgt->vd_last.i = 0; + //tgt->vd_last.p.deblk = tgt->vd_deblk[0]; + tgt->vd_version = src->vd_version; + tgt->vd_jiffy = src->vd_jiffy; + + for (i = 1; i < n; i++) { + tgt->vd_deblk[i] = kmalloc(AUFS_DEBLK_SIZE, GFP_KERNEL); + if (tgt->vd_deblk[i]) + memcpy(tgt->vd_deblk[i], src->vd_deblk[i], + AUFS_DEBLK_SIZE); + else + goto out; + } + //smp_mb(); + //DbgVdir(tgt); + return 0; /* success */ + + out: + rerr = reinit_vdir(tgt); + BUG_ON(rerr); + TraceErr(err); + return err; +} + +int au_init_vdir(struct file *file) +{ + int err; + struct dentry *dentry; + struct inode *inode; + struct aufs_vdir *vdir_cache, *allocated; + + dentry = file->f_dentry; + LKTRTrace("%.*s, pos %Ld\n", DLNPair(dentry), file->f_pos); + FiMustWriteLock(file); + inode = dentry->d_inode; + IiMustWriteLock(inode); + DEBUG_ON(!S_ISDIR(inode->i_mode)); + + err = read_vdir(file, !file->f_pos); + if (unlikely(err)) + goto out; + //DbgVdir(ivdir(inode)); goto out; + + allocated = NULL; + vdir_cache = fvdir_cache(file); + if (!vdir_cache) { + vdir_cache = alloc_vdir(); + err = PTR_ERR(vdir_cache); + if (IS_ERR(vdir_cache)) + goto out; + allocated = vdir_cache; + } else if (!file->f_pos && vdir_cache->vd_version != file->f_version) { + err = reinit_vdir(vdir_cache); + if (unlikely(err)) + goto out; + } else + return 0; /* success */ + //err = 0; DbgVdir(vdir_cache); goto out; + + err = copy_vdir(vdir_cache, ivdir(inode)); + if (!err) { + file->f_version = inode->i_version; + if (allocated) + set_fvdir_cache(file, allocated); + } else if (allocated) + free_vdir(allocated); + + out: + TraceErr(err); + return err; +} + +static loff_t calc_offset(struct aufs_vdir *vdir) +{ + loff_t offset; + union aufs_deblk_p p; + + p.deblk = vdir->vd_deblk[vdir->vd_last.i]; + offset = vdir->vd_last.p.p - p.p; + offset += sizeof(*p.deblk) * vdir->vd_last.i; + return offset; +} + +/* returns true or false */ +static int seek_vdir(struct file *file) +{ + int valid, i, n; + struct dentry *dentry; + struct aufs_vdir *vdir_cache; + loff_t offset; + union aufs_deblk_p p, deblk_end; + + dentry = file->f_dentry; + LKTRTrace("%.*s, pos %Ld\n", DLNPair(dentry), file->f_pos); + vdir_cache = fvdir_cache(file); + DEBUG_ON(!vdir_cache); + //DbgVdir(vdir_cache); + + valid = 1; + offset = calc_offset(vdir_cache); + LKTRTrace("offset %Ld\n", offset); + if (file->f_pos == offset) + goto out; + + vdir_cache->vd_last.i = 0; + vdir_cache->vd_last.p.deblk = vdir_cache->vd_deblk[0]; + if (!file->f_pos) + goto out; + + valid = 0; + i = file->f_pos / AUFS_DEBLK_SIZE; + LKTRTrace("i %d\n", i); + if (i >= vdir_cache->vd_nblk) + goto out; + + n = vdir_cache->vd_nblk; + //DbgVdir(vdir_cache); + for (; i < n; i++) { + p.deblk = vdir_cache->vd_deblk[i]; + deblk_end.deblk = p.deblk + 1; + offset = i * AUFS_DEBLK_SIZE; + while (!is_deblk_end(&p, &deblk_end) && offset < file->f_pos) { + int l; + l = calc_size(p.de->de_str.len); + offset += l; + p.p += l; + } + if (!is_deblk_end(&p, &deblk_end)) { + valid = 1; + vdir_cache->vd_last.i = i; + vdir_cache->vd_last.p = p; + break; + } + } + + out: + //smp_mb(); + //DbgVdir(vdir_cache); + TraceErr(!valid); + return valid; +} + +int au_fill_de(struct file *file, void *dirent, filldir_t filldir) +{ + int err, l; + struct dentry *dentry; + struct aufs_vdir *vdir_cache; + struct aufs_de *de; + union aufs_deblk_p deblk_end; + + dentry = file->f_dentry; + LKTRTrace("%.*s, pos %Ld\n", DLNPair(dentry), file->f_pos); + vdir_cache = fvdir_cache(file); + DEBUG_ON(!vdir_cache); + //DbgVdir(vdir_cache); + + if (!seek_vdir(file)) + return 0; + + while (1) { + deblk_end.deblk + = vdir_cache->vd_deblk[vdir_cache->vd_last.i] + 1; + while (!is_deblk_end(&vdir_cache->vd_last.p, &deblk_end)) { + de = vdir_cache->vd_last.p.de; + LKTRTrace("%.*s, off%Ld, i%lu, dt%d\n", + de->de_str.len, de->de_str.name, + file->f_pos, de->de_ino, de->de_type); + err = filldir(dirent, de->de_str.name, de->de_str.len, + file->f_pos, de->de_ino, de->de_type); + if (unlikely(err)) { + TraceErr(err); + //return err; + //todo: ignore the error caused by udba. + return 0; + } + + l = calc_size(de->de_str.len); + vdir_cache->vd_last.p.p += l; + file->f_pos += l; + } + if (vdir_cache->vd_last.i < vdir_cache->vd_nblk - 1) { + vdir_cache->vd_last.i++; + vdir_cache->vd_last.p.deblk + = vdir_cache->vd_deblk[vdir_cache->vd_last.i]; + file->f_pos = sizeof(*vdir_cache->vd_last.p.deblk) + * vdir_cache->vd_last.i; + continue; + } + break; + } + + //smp_mb(); + return 0; +} diff --git a/fs/aufs/vfsub.c b/fs/aufs/vfsub.c new file mode 100755 index 0000000..8571d21 --- /dev/null +++ b/fs/aufs/vfsub.c @@ -0,0 +1,665 @@ +/* + * Copyright (C) 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: vfsub.c,v 1.5 2007/04/23 00:55:06 sfjro Exp $ */ +// I'm going to slightly mad + +#include "aufs.h" + +/* ---------------------------------------------------------------------- */ + +#ifdef CONFIG_AUFS_DLGT +struct permission_args { + int *errp; + struct inode *inode; + int mask; + struct nameidata *nd; +}; + +static void call_permission(void *args) +{ + struct permission_args *a = args; + *a->errp = do_vfsub_permission(a->inode, a->mask, a->nd); +} + +int vfsub_permission(struct inode *inode, int mask, struct nameidata *nd, + int dlgt) +{ + if (!dlgt) + return do_vfsub_permission(inode, mask, nd); + else { + int err; + struct permission_args args = { + .errp = &err, + .inode = inode, + .mask = mask, + .nd = nd + }; + au_wkq_wait(call_permission, &args, /*dlgt*/1); + return err; + } +} + +/* ---------------------------------------------------------------------- */ + +struct create_args { + int *errp; + struct inode *dir; + struct dentry *dentry; + int mode; + struct nameidata *nd; +}; + +static void call_create(void *args) +{ + struct create_args *a = args; + *a->errp = do_vfsub_create(a->dir, a->dentry, a->mode, a->nd); +} + +int vfsub_create(struct inode *dir, struct dentry *dentry, int mode, + struct nameidata *nd, int dlgt) +{ + if (!dlgt) + return do_vfsub_create(dir, dentry, mode, nd); + else { + int err; + struct create_args args = { + .errp = &err, + .dir = dir, + .dentry = dentry, + .mode = mode, + .nd = nd + }; + au_wkq_wait(call_create, &args, /*dlgt*/1); + return err; + } +} + +struct symlink_args { + int *errp; + struct inode *dir; + struct dentry *dentry; + const char *symname; + int mode; +}; + +static void call_symlink(void *args) +{ + struct symlink_args *a = args; + *a->errp = do_vfsub_symlink(a->dir, a->dentry, a->symname, a->mode); +} + +int vfsub_symlink(struct inode *dir, struct dentry *dentry, const char *symname, + int mode, int dlgt) +{ + if (!dlgt) + return do_vfsub_symlink(dir, dentry, symname, mode); + else { + int err; + struct symlink_args args = { + .errp = &err, + .dir = dir, + .dentry = dentry, + .symname = symname, + .mode = mode + }; + au_wkq_wait(call_symlink, &args, /*dlgt*/1); + return err; + } +} + +struct mknod_args { + int *errp; + struct inode *dir; + struct dentry *dentry; + int mode; + dev_t dev; +}; + +static void call_mknod(void *args) +{ + struct mknod_args *a = args; + *a->errp = do_vfsub_mknod(a->dir, a->dentry, a->mode, a->dev); +} + +int vfsub_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev, + int dlgt) +{ + if (!dlgt) + return do_vfsub_mknod(dir, dentry, mode, dev); + else { + int err; + struct mknod_args args = { + .errp = &err, + .dir = dir, + .dentry = dentry, + .mode = mode, + .dev = dev + }; + au_wkq_wait(call_mknod, &args, /*dlgt*/1); + return err; + } +} + +struct mkdir_args { + int *errp; + struct inode *dir; + struct dentry *dentry; + int mode; +}; + +static void call_mkdir(void *args) +{ + struct mkdir_args *a = args; + *a->errp = do_vfsub_mkdir(a->dir, a->dentry, a->mode); +} + +int vfsub_mkdir(struct inode *dir, struct dentry *dentry, int mode, int dlgt) +{ + if (!dlgt) + return do_vfsub_mkdir(dir, dentry, mode); + else { + int err; + struct mkdir_args args = { + .errp = &err, + .dir = dir, + .dentry = dentry, + .mode = mode + }; + au_wkq_wait(call_mkdir, &args, /*dlgt*/1); + return err; + } +} + +/* ---------------------------------------------------------------------- */ + +struct link_args { + int *errp; + struct inode *dir; + struct dentry *src_dentry, *dentry; +}; + +static void call_link(void *args) +{ + struct link_args *a = args; + *a->errp = do_vfsub_link(a->src_dentry, a->dir, a->dentry); +} + +int vfsub_link(struct dentry *src_dentry, struct inode *dir, + struct dentry *dentry, int dlgt) +{ + if (!dlgt) + return do_vfsub_link(src_dentry, dir, dentry); + else { + int err; + struct link_args args = { + .errp = &err, + .src_dentry = src_dentry, + .dir = dir, + .dentry = dentry + }; + au_wkq_wait(call_link, &args, /*dlgt*/1); + return err; + } +} + +struct rename_args { + int *errp; + struct inode *src_dir, *dir; + struct dentry *src_dentry, *dentry; +}; + +static void call_rename(void *args) +{ + struct rename_args *a = args; + *a->errp = do_vfsub_rename(a->src_dir, a->src_dentry, a->dir, + a->dentry); +} + +int vfsub_rename(struct inode *src_dir, struct dentry *src_dentry, + struct inode *dir, struct dentry *dentry, int dlgt) +{ + if (!dlgt) + return do_vfsub_rename(src_dir, src_dentry, dir, dentry); + else { + int err; + struct rename_args args = { + .errp = &err, + .src_dir = src_dir, + .src_dentry = src_dentry, + .dir = dir, + .dentry = dentry + }; + au_wkq_wait(call_rename, &args, /*dlgt*/1); + return err; + } +} + +struct rmdir_args { + int *errp; + struct inode *dir; + struct dentry *dentry; +}; + +static void call_rmdir(void *args) +{ + struct rmdir_args *a = args; + *a->errp = do_vfsub_rmdir(a->dir, a->dentry); +} + +int vfsub_rmdir(struct inode *dir, struct dentry *dentry, int dlgt) +{ + if (!dlgt) + return do_vfsub_rmdir(dir, dentry); + else { + int err; + struct rmdir_args args = { + .errp = &err, + .dir = dir, + .dentry = dentry + }; + au_wkq_wait(call_rmdir, &args, /*dlgt*/1); + return err; + } +} + +/* ---------------------------------------------------------------------- */ + +struct read_args { + ssize_t *errp; + struct file *file; + union { + void *kbuf; + char __user *ubuf; + }; + size_t count; + loff_t *ppos; +}; + +static void call_read_k(void *args) +{ + struct read_args *a = args; + LKTRTrace("%.*s, cnt %lu, pos %Ld\n", + DLNPair(a->file->f_dentry), (unsigned long)a->count, + *a->ppos); + *a->errp = do_vfsub_read_k(a->file, a->kbuf, a->count, a->ppos); +} + +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count, + loff_t *ppos, int dlgt) +{ + if (!dlgt) + return do_vfsub_read_u(file, ubuf, count, ppos); + else { + ssize_t err, read; + struct read_args args = { + .errp = &err, + .file = file, + .count = count, + .ppos = ppos + }; + + if (unlikely(!count)) + return 0; + + /* + * workaround an application bug. + * generally, read(2) or write(2) may return the value shorter + * than requested. But many applications don't support it, + * for example bash. + */ + err = -ENOMEM; + if (args.count > PAGE_SIZE) + args.count = PAGE_SIZE; + args.kbuf = kmalloc(args.count, GFP_KERNEL); + if (unlikely(!args.kbuf)) + goto out; + + read = 0; + do { + au_wkq_wait(call_read_k, &args, /*dlgt*/1); + if (unlikely(err > 0 + && copy_to_user(ubuf, args.kbuf, err))) { + err = -EFAULT; + goto out_free; + } else if (!err) + break; + else if (unlikely(err < 0)) + goto out_free; + count -= err; + /* do not read too much because of file i/o pointer */ + if (unlikely(count < args.count)) + args.count = count; + ubuf += err; + read += err; + } while (count); + smp_mb(); + err = read; + + out_free: + kfree(args.kbuf); + out: + return err; + } +} + +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count, loff_t *ppos, + int dlgt) +{ + if (!dlgt) + return do_vfsub_read_k(file, kbuf, count, ppos); + else { + ssize_t err; + struct read_args args = { + .errp = &err, + .file = file, + .count = count, + .ppos = ppos + }; + args.kbuf = kbuf; + au_wkq_wait(call_read_k, &args, /*dlgt*/1); + return err; + } +} + +struct write_args { + ssize_t *errp; + struct file *file; + union { + void *kbuf; + const char __user *ubuf; + }; + void *buf; + size_t count; + loff_t *ppos; +}; + +static void call_write_k(void *args) +{ + struct write_args *a = args; + LKTRTrace("%.*s, cnt %lu, pos %Ld\n", + DLNPair(a->file->f_dentry), (unsigned long)a->count, + *a->ppos); + *a->errp = do_vfsub_write_k(a->file, a->kbuf, a->count, a->ppos); +} + +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count, + loff_t *ppos, int dlgt) +{ + if (!dlgt) + return do_vfsub_write_u(file, ubuf, count, ppos); + else { + ssize_t err, written; + struct write_args args = { + .errp = &err, + .file = file, + .count = count, + .ppos = ppos + }; + + if (unlikely(!count)) + return 0; + + /* + * workaround an application bug. + * generally, read(2) or write(2) may return the value shorter + * than requested. But many applications don't support it, + * for example bash. + */ + err = -ENOMEM; + if (args.count > PAGE_SIZE) + args.count = PAGE_SIZE; + args.kbuf = kmalloc(args.count, GFP_KERNEL); + if (unlikely(!args.kbuf)) + goto out; + + written = 0; + do { + if (unlikely(copy_from_user(args.kbuf, ubuf, args.count))) { + err = -EFAULT; + goto out_free; + } + + au_wkq_wait(call_write_k, &args, /*dlgt*/1); + if (err > 0) { + count -= err; + if (count < args.count) + args.count = count; + ubuf += err; + written += err; + } else if (!err) + break; + else if (unlikely(err < 0)) + goto out_free; + } while (count); + err = written; + + out_free: + kfree(args.kbuf); + out: + return err; + } +} + +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count, loff_t *ppos, + int dlgt) +{ + if (!dlgt) + return do_vfsub_write_k(file, kbuf, count, ppos); + else { + ssize_t err; + struct write_args args = { + .errp = &err, + .file = file, + .count = count, + .ppos = ppos + }; + args.kbuf = kbuf; + au_wkq_wait(call_write_k, &args, /*dlgt*/1); + return err; + } +} + +struct readdir_args { + int *errp; + struct file *file; + filldir_t filldir; + void *arg; +}; + +static void call_readdir(void *args) +{ + struct readdir_args *a = args; + *a->errp = do_vfsub_readdir(a->file, a->filldir, a->arg); +} + +int vfsub_readdir(struct file *file, filldir_t filldir, void *arg, int dlgt) +{ + if (!dlgt) + return do_vfsub_readdir(file, filldir, arg); + else { + int err; + struct readdir_args args = { + .errp = &err, + .file = file, + .filldir = filldir, + .arg = arg + }; + au_wkq_wait(call_readdir, &args, /*dlgt*/1); + return err; + } +} +#endif /* CONFIG_AUFS_DLGT */ + +/* ---------------------------------------------------------------------- */ + +struct notify_change_args { + int *errp; + struct dentry *h_dentry; + struct iattr *ia; +}; + +static void call_notify_change(void *args) +{ + struct notify_change_args *a = args; + struct inode *h_inode; + + LKTRTrace("%.*s, ia_valid 0x%x\n", + DLNPair(a->h_dentry), a->ia->ia_valid); + h_inode = a->h_dentry->d_inode; + IMustLock(h_inode); + + *a->errp = -EPERM; + if (!IS_IMMUTABLE(h_inode) && !IS_APPEND(h_inode)) { + lockdep_off(); + *a->errp = notify_change(a->h_dentry, a->ia); + lockdep_on(); + } + TraceErr(*a->errp); +} + +int vfsub_notify_change(struct dentry *dentry, struct iattr *ia, int dlgt) +{ + int err; + struct notify_change_args args = { + .errp = &err, + .h_dentry = dentry, + .ia = ia + }; + +#ifndef CONFIG_AUFS_DLGT + call_notify_change(&args); +#else + if (!dlgt) + call_notify_change(&args); + else + au_wkq_wait(call_notify_change, &args, /*dlgt*/1); +#endif + + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +struct unlink_args { + int *errp; + struct inode *dir; + struct dentry *dentry; +}; + +static void call_unlink(void *args) +{ + struct unlink_args *a = args; + struct inode *h_inode; + const int stop_sillyrename = (au_is_nfs(a->dentry->d_sb) + && atomic_read(&a->dentry->d_count) == 1); + + LKTRTrace("%.*s, stop_silly %d, cnt %d\n", + DLNPair(a->dentry), stop_sillyrename, + atomic_read(&a->dentry->d_count)); + IMustLock(a->dir); + + if (!stop_sillyrename) + dget(a->dentry); + h_inode = a->dentry->d_inode; + if (h_inode) + atomic_inc(&h_inode->i_count); +#if 0 // partial testing + { + struct qstr *name = &a->dentry->d_name; + if (name->len == sizeof(AUFS_XINO_FNAME) - 1 + && !strncmp(name->name, AUFS_XINO_FNAME, name->len)) { + lockdep_off(); + *a->errp = vfs_unlink(a->dir, a->dentry); + lockdep_on(); + } else + err = -1; + } +#else + // vfs_unlink() locks inode + lockdep_off(); + *a->errp = vfs_unlink(a->dir, a->dentry); + lockdep_on(); +#endif + + if (!stop_sillyrename) + dput(a->dentry); + if (h_inode) + iput(h_inode); + + TraceErr(*a->errp); +} + +/* + * @dir: must be locked. + * @dentry: target dentry. + */ +int vfsub_unlink(struct inode *dir, struct dentry *dentry, int dlgt) +{ + int err; + struct unlink_args args = { + .errp = &err, + .dir = dir, + .dentry = dentry + }; + +#ifndef CONFIG_AUFS_DLGT + call_unlink(&args); +#else + if (!dlgt) + call_unlink(&args); + else + au_wkq_wait(call_unlink, &args, /*dlgt*/1); +#endif + return err; +} + +/* ---------------------------------------------------------------------- */ + +struct statfs_args { + int *errp; + void *arg; + struct kstatfs *buf; +}; + +static void call_statfs(void *args) +{ + struct statfs_args *a = args; + *a->errp = vfs_statfs(a->arg, a->buf); +} + +int vfsub_statfs(void *arg, struct kstatfs *buf, int dlgt) +{ + int err; + struct statfs_args args = { + .errp = &err, + .arg = arg, + .buf = buf + }; + +#ifndef CONFIG_AUFS_DLGT + call_statfs(&args); +#else + if (!dlgt) + call_statfs(&args); + else + au_wkq_wait(call_statfs, &args, /*dlgt*/1); +#endif + return err; +} diff --git a/fs/aufs/vfsub.h b/fs/aufs/vfsub.h new file mode 100755 index 0000000..52f15cc --- /dev/null +++ b/fs/aufs/vfsub.h @@ -0,0 +1,427 @@ +/* + * Copyright (C) 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: vfsub.h,v 1.8 2007/05/14 03:39:10 sfjro Exp $ */ + +#ifndef __AUFS_VFSUB_H__ +#define __AUFS_VFSUB_H__ + +#ifdef __KERNEL__ + +#include +#include +#include "wkq.h" + +/* ---------------------------------------------------------------------- */ + +/* simple abstractions, for future use */ +static inline +int do_vfsub_permission(struct inode *inode, int mask, struct nameidata *nd) +{ + LKTRTrace("i%lu, mask 0x%x, nd %p\n", inode->i_ino, mask, nd); +#if 0 +#else + return permission(inode, mask, nd); +#endif +} + +static inline +struct file *vfsub_filp_open(const char *path, int oflags, int mode) +{ + struct file *err; + + LKTRTrace("%s\n", path); + + lockdep_off(); + err = filp_open(path, oflags, mode); + lockdep_on(); + return err; +} + +static inline +int vfsub_path_lookup(const char *name, unsigned int flags, + struct nameidata *nd) +{ + int err; + + LKTRTrace("%s\n", name); + + //lockdep_off(); + err = path_lookup(name, flags, nd); + //lockdep_on(); + return err; +} + +/* ---------------------------------------------------------------------- */ + +static inline +int do_vfsub_create(struct inode *dir, struct dentry *dentry, int mode, + struct nameidata *nd) +{ + LKTRTrace("i%lu, %.*s, 0x%x\n", dir->i_ino, DLNPair(dentry), mode); +#if 0 +#else + return vfs_create(dir, dentry, mode, nd); +#endif +} + +static inline +int do_vfsub_symlink(struct inode *dir, struct dentry *dentry, + const char *symname, int mode) +{ + LKTRTrace("i%lu, %.*s, %s, 0x%x\n", + dir->i_ino, DLNPair(dentry), symname, mode); +#if 0 +#else + return vfs_symlink(dir, dentry, symname, mode); +#endif +} + +static inline +int do_vfsub_mknod(struct inode *dir, struct dentry *dentry, int mode, + dev_t dev) +{ + LKTRTrace("i%lu, %.*s, 0x%x\n", dir->i_ino, DLNPair(dentry), mode); +#if 0 +#else + return vfs_mknod(dir, dentry, mode, dev); +#endif +} + +static inline +int do_vfsub_link(struct dentry *src_dentry, struct inode *dir, + struct dentry *dentry) +{ + int err; + + LKTRTrace("%.*s, i%lu, %.*s\n", + DLNPair(src_dentry), dir->i_ino, DLNPair(dentry)); + + lockdep_off(); +#if 0 +#else + err = vfs_link(src_dentry, dir, dentry); +#endif + lockdep_on(); + return err; +} + +static inline +int do_vfsub_rename(struct inode *src_dir, struct dentry *src_dentry, + struct inode *dir, struct dentry *dentry) +{ + int err; + + LKTRTrace("i%lu, %.*s, i%lu, %.*s\n", + src_dir->i_ino, DLNPair(src_dentry), + dir->i_ino, DLNPair(dentry)); + + lockdep_off(); +#if 0 +#else + err = vfs_rename(src_dir, src_dentry, dir, dentry); +#endif + lockdep_on(); + return err; +} + +static inline +int do_vfsub_mkdir(struct inode *dir, struct dentry *dentry, int mode) +{ + LKTRTrace("i%lu, %.*s, 0x%x\n", dir->i_ino, DLNPair(dentry), mode); +#if 0 +#else + return vfs_mkdir(dir, dentry, mode); +#endif +} + +static inline int do_vfsub_rmdir(struct inode *dir, struct dentry *dentry) +{ + int err; + + LKTRTrace("i%lu, %.*s\n", dir->i_ino, DLNPair(dentry)); + + lockdep_off(); +#if 0 +#else + err = vfs_rmdir(dir, dentry); +#endif + lockdep_on(); + return err; +} + +/* ---------------------------------------------------------------------- */ + +static inline +ssize_t do_vfsub_read_u(struct file *file, char __user *ubuf, size_t count, + loff_t *ppos) +{ + ssize_t err; + + LKTRTrace("%.*s, cnt %lu, pos %Ld\n", + DLNPair(file->f_dentry), (unsigned long)count, *ppos); + + /* nfs uses some locks */ + lockdep_off(); +#if 0 +#else + err = vfs_read(file, ubuf, count, ppos); +#endif + lockdep_on(); + return err; +} + +// kernel_read() ?? +static inline +ssize_t do_vfsub_read_k(struct file *file, void *kbuf, size_t count, + loff_t *ppos) +{ + ssize_t err; + mm_segment_t oldfs; + + oldfs = get_fs(); + set_fs(KERNEL_DS); + err = do_vfsub_read_u(file, (char __user*)kbuf, count, ppos); + set_fs(oldfs); + return err; +} + +static inline +ssize_t do_vfsub_write_u(struct file *file, const char __user *ubuf, + size_t count, loff_t *ppos) +{ + ssize_t err; + + LKTRTrace("%.*s, cnt %lu, pos %Ld\n", + DLNPair(file->f_dentry), (unsigned long)count, *ppos); + + lockdep_off(); +#if 0 +#else + err = vfs_write(file, ubuf, count, ppos); +#endif + lockdep_on(); + return err; +} + +static inline +ssize_t do_vfsub_write_k(struct file *file, void *kbuf, size_t count, + loff_t *ppos) +{ + ssize_t err; + mm_segment_t oldfs; + + oldfs = get_fs(); + set_fs(KERNEL_DS); + err = do_vfsub_write_u(file, (const char __user*)kbuf, count, ppos); + set_fs(oldfs); + return err; +} + +static inline +int do_vfsub_readdir(struct file *file, filldir_t filldir, void *arg) +{ + int err; + + LKTRTrace("%.*s\n", DLNPair(file->f_dentry)); + + lockdep_off(); +#if 0 +#else + err = vfs_readdir(file, filldir, arg); +#endif + lockdep_on(); + return err; +} + +/* ---------------------------------------------------------------------- */ + +static inline loff_t vfsub_llseek(struct file *file, loff_t offset, int origin) +{ + loff_t err; + + LKTRTrace("%.*s\n", DLNPair(file->f_dentry)); + + lockdep_off(); +#if 0 +#else + err = vfs_llseek(file, offset, origin); +#endif + lockdep_on(); + return err; +} + +/* ---------------------------------------------------------------------- */ + +#ifdef CONFIG_AUFS_DLGT +static inline int need_dlgt(struct super_block *sb) +{ + return (au_flag_test(sb, AuFlag_DLGT) && !is_au_wkq(current)); +} + +int vfsub_permission(struct inode *inode, int mask, struct nameidata *nd, + int dlgt); + +int vfsub_create(struct inode *dir, struct dentry *dentry, int mode, + struct nameidata *nd, int dlgt); +int vfsub_symlink(struct inode *dir, struct dentry *dentry, const char *symname, + int mode, int dlgt); +int vfsub_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev, + int dlgt); +int vfsub_link(struct dentry *src_dentry, struct inode *dir, + struct dentry *dentry, int dlgt); +int vfsub_rename(struct inode *src_dir, struct dentry *src_dentry, + struct inode *dir, struct dentry *dentry, int dlgt); +int vfsub_mkdir(struct inode *dir, struct dentry *dentry, int mode, int dlgt); +int vfsub_rmdir(struct inode *dir, struct dentry *dentry, int dlgt); + +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count, + loff_t *ppos, int dlgt); +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count, loff_t *ppos, + int dlgt); +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count, + loff_t *ppos, int dlgt); +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count, loff_t *ppos, + int dlgt); +int vfsub_readdir(struct file *file, filldir_t filldir, void *arg, int dlgt); + +#else + +static inline int need_dlgt(struct super_block *sb) +{ + return 0; +} + +static inline +int vfsub_permission(struct inode *inode, int mask, struct nameidata *nd, + int dlgt) +{ + return do_vfsub_permission(inode, mask, nd); +} + +static inline +int vfsub_create(struct inode *dir, struct dentry *dentry, int mode, + struct nameidata *nd, int dlgt) +{ + return do_vfsub_create(dir, dentry, mode, nd); +} + +static inline +int vfsub_symlink(struct inode *dir, struct dentry *dentry, const char *symname, + int mode, int dlgt) +{ + return do_vfsub_symlink(dir, dentry, symname, mode); +} + +static inline +int vfsub_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev, + int dlgt) +{ + return do_vfsub_mknod(dir, dentry, mode, dev); +} + +static inline +int vfsub_link(struct dentry *src_dentry, struct inode *dir, + struct dentry *dentry, int dlgt) +{ + return do_vfsub_link(src_dentry, dir, dentry); +} + +static inline +int vfsub_rename(struct inode *src_dir, struct dentry *src_dentry, + struct inode *dir, struct dentry *dentry, int dlgt) +{ + return do_vfsub_rename(src_dir, src_dentry, dir, dentry); +} + +static inline +int vfsub_mkdir(struct inode *dir, struct dentry *dentry, int mode, + int dlgt) +{ + return do_vfsub_mkdir(dir, dentry, mode); +} + +static inline +int vfsub_rmdir(struct inode *dir, struct dentry *dentry, int dlgt) +{ + return do_vfsub_rmdir(dir, dentry); +} + +static inline +ssize_t vfsub_read_u(struct file *file, char __user *ubuf, size_t count, + loff_t *ppos, int dlgt) +{ + return do_vfsub_read_u(file, ubuf, count, ppos); +} + +static inline +ssize_t vfsub_read_k(struct file *file, void *kbuf, size_t count, loff_t *ppos, + int dlgt) +{ + return do_vfsub_read_k(file, kbuf, count, ppos); +} + +static inline +ssize_t vfsub_write_u(struct file *file, const char __user *ubuf, size_t count, + loff_t *ppos, int dlgt) +{ + return do_vfsub_write_u(file, ubuf, count, ppos); +} + +static inline +ssize_t vfsub_write_k(struct file *file, void *kbuf, size_t count, loff_t *ppos, + int dlgt) +{ + return do_vfsub_write_k(file, kbuf, count, ppos); +} + +static inline +int vfsub_readdir(struct file *file, filldir_t filldir, void *arg, int dlgt) +{ + return do_vfsub_readdir(file, filldir, arg); +} +#endif /* CONFIG_AUFS_DLGT */ + +/* ---------------------------------------------------------------------- */ + +static inline +struct dentry *vfsub_lock_rename(struct dentry *d1, struct dentry *d2) +{ + struct dentry *d; + + lockdep_off(); + d = lock_rename(d1, d2); + lockdep_on(); + return d; +} + +static inline void vfsub_unlock_rename(struct dentry *d1, struct dentry *d2) +{ + lockdep_off(); + unlock_rename(d1, d2); + lockdep_on(); +} + +/* ---------------------------------------------------------------------- */ + +int vfsub_notify_change(struct dentry *dentry, struct iattr *ia, int dlgt); +int vfsub_unlink(struct inode *dir, struct dentry *dentry, int dlgt); +int vfsub_statfs(void *arg, struct kstatfs *buf, int dlgt); + +#endif /* __KERNEL__ */ +#endif /* __AUFS_VFSUB_H__ */ diff --git a/fs/aufs/whout.c b/fs/aufs/whout.c new file mode 100755 index 0000000..b7f874c --- /dev/null +++ b/fs/aufs/whout.c @@ -0,0 +1,933 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: whout.c,v 1.14 2007/05/14 03:40:40 sfjro Exp $ */ + +#include +#include +#include +#include +#include "aufs.h" + +#define WH_MASK S_IRUGO + +/* If a directory contains this file, then it is opaque. We start with the + * .wh. flag so that it is blocked by lookup. + */ +static struct qstr diropq_name = { + .name = AUFS_WH_DIROPQ, + .len = sizeof(AUFS_WH_DIROPQ) - 1 +}; + +/* + * generate whiteout name, which is NOT terminated by NULL. + * @name: original d_name.name + * @len: original d_name.len + * @wh: whiteout qstr + * returns zero when succeeds, otherwise error. + * succeeded value as wh->name should be freed by au_free_whname(). + */ +int au_alloc_whname(const char *name, int len, struct qstr *wh) +{ + char *p; + + DEBUG_ON(!name || !len || !wh); + + if (unlikely(len > PATH_MAX - AUFS_WH_PFX_LEN)) + return -ENAMETOOLONG; + + wh->len = len + AUFS_WH_PFX_LEN; + wh->name = p = kmalloc(wh->len, GFP_KERNEL); + //if (LktrCond) {kfree(p); wh->name = p = NULL;} + if (p) { + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN); + memcpy(p + AUFS_WH_PFX_LEN, name, len); + //smp_mb(); + return 0; + } + return -ENOMEM; +} + +void au_free_whname(struct qstr *wh) +{ + DEBUG_ON(!wh || !wh->name); + kfree(wh->name); +#ifdef CONFIG_AUFS_DEBUG + wh->name = NULL; +#endif +} + +/* ---------------------------------------------------------------------- */ + +/* + * test if the @wh_name exists under @hidden_parent. + * @try_sio specifies the necessary of super-io. + */ +int is_wh(struct dentry *hidden_parent, struct qstr *wh_name, int try_sio, + struct lkup_args *lkup) +{ + int err; + struct dentry *wh_dentry; + struct inode *hidden_dir; + + LKTRTrace("%.*s/%.*s, lkup{%p, %d}\n", DLNPair(hidden_parent), + wh_name->len, wh_name->name, lkup->nfsmnt, lkup->dlgt); + hidden_dir = hidden_parent->d_inode; + DEBUG_ON(!S_ISDIR(hidden_dir->i_mode)); + IMustLock(hidden_dir); + + if (!try_sio) + wh_dentry = lkup_one(wh_name->name, hidden_parent, + wh_name->len, lkup); + else + wh_dentry = sio_lkup_one(wh_name->name, hidden_parent, + wh_name->len, lkup); + //if (LktrCond) {dput(wh_dentry); wh_dentry = ERR_PTR(-1);} + err = PTR_ERR(wh_dentry); + if (IS_ERR(wh_dentry)) + goto out; + + err = 0; + if (!wh_dentry->d_inode) + goto out_wh; /* success */ + + err = 1; + if (S_ISREG(wh_dentry->d_inode->i_mode)) + goto out_wh; /* success */ + + err = -EIO; + IOErr("%.*s Invalid whiteout entry type 0%o.\n", + DLNPair(wh_dentry), wh_dentry->d_inode->i_mode); + + out_wh: + dput(wh_dentry); + out: + TraceErr(err); + return err; +} + +/* + * test if the @hidden_dentry sets opaque or not. + */ +int is_diropq(struct dentry *hidden_dentry, struct lkup_args *lkup) +{ + int err; + struct inode *hidden_dir; + + LKTRTrace("dentry %.*s\n", DLNPair(hidden_dentry)); + hidden_dir = hidden_dentry->d_inode; + DEBUG_ON(!S_ISDIR(hidden_dir->i_mode)); + IMustLock(hidden_dir); + + err = is_wh(hidden_dentry, &diropq_name, /*try_sio*/1, lkup); + TraceErr(err); + return err; +} + +/* + * returns a negative dentry whose name is unique and temporary. + */ +struct dentry *lkup_whtmp(struct dentry *hidden_parent, struct qstr *prefix, + struct lkup_args *lkup) +{ +#define HEX_LEN 4 + struct dentry *dentry; + int len, i; + char defname[AUFS_WH_PFX_LEN * 2 + DNAME_INLINE_LEN_MIN + 1 + + HEX_LEN + 1], *name, *p; + static unsigned char cnt; + + LKTRTrace("hp %.*s, prefix %.*s\n", + DLNPair(hidden_parent), prefix->len, prefix->name); + DEBUG_ON(!hidden_parent->d_inode); + IMustLock(hidden_parent->d_inode); + + name = defname; + len = sizeof(defname) - DNAME_INLINE_LEN_MIN + prefix->len - 1; + if (unlikely(prefix->len > DNAME_INLINE_LEN_MIN)) { + dentry = ERR_PTR(-ENAMETOOLONG); + if (unlikely(len >= PATH_MAX)) + goto out; + dentry = ERR_PTR(-ENOMEM); + name = kmalloc(len + 1, GFP_KERNEL); + //if (LktrCond) {kfree(name); name = NULL;} + if (unlikely(!name)) + goto out; + } + + // doubly whiteout-ed + memcpy(name, AUFS_WH_PFX AUFS_WH_PFX, AUFS_WH_PFX_LEN * 2); + p = name + AUFS_WH_PFX_LEN * 2; + memcpy(p, prefix->name, prefix->len); + p += prefix->len; + *p++ = '.'; + DEBUG_ON(name + len + 1 - p <= HEX_LEN); + + for (i = 0; i < 3; i++) { + sprintf(p, "%.*d", HEX_LEN, cnt++); + dentry = sio_lkup_one(name, hidden_parent, len, lkup); + //if (LktrCond) {dput(dentry); dentry = ERR_PTR(-1);} + if (unlikely(IS_ERR(dentry) || !dentry->d_inode)) + goto out_name; + dput(dentry); + } + //Warn("could not get random name\n"); + dentry = ERR_PTR(-EEXIST); + Dbg("%.*s\n", len, name); + BUG(); + + out_name: + if (unlikely(name != defname)) + kfree(name); + out: + TraceErrPtr(dentry); + return dentry; +#undef HEX_LEN +} + +/* + * rename the @dentry of @bindex to the whiteouted temporary name. + */ +int rename_whtmp(struct dentry *dentry, aufs_bindex_t bindex) +{ + int err; + struct inode *hidden_dir; + struct dentry *hidden_dentry, *hidden_parent, *tmp_dentry; + struct super_block *sb; + struct lkup_args lkup; + + LKTRTrace("%.*s, b%d\n", DLNPair(dentry), bindex); + hidden_dentry = au_h_dptr_i(dentry, bindex); + DEBUG_ON(!hidden_dentry || !hidden_dentry->d_inode); + hidden_parent = hidden_dentry->d_parent; + hidden_dir = hidden_parent->d_inode; + IMustLock(hidden_dir); + + sb = dentry->d_sb; + lkup.nfsmnt = au_nfsmnt(sb, bindex); + lkup.dlgt = need_dlgt(sb); + tmp_dentry = lkup_whtmp(hidden_parent, &hidden_dentry->d_name, &lkup); + //if (LktrCond) {dput(tmp_dentry); tmp_dentry = ERR_PTR(-1);} + err = PTR_ERR(tmp_dentry); + if (!IS_ERR(tmp_dentry)) { + /* under the same dir, no need to lock_rename() */ + err = vfsub_rename(hidden_dir, hidden_dentry, + hidden_dir, tmp_dentry, lkup.dlgt); + //if (LktrCond) err = -1; //unavailable + TraceErr(err); + dput(tmp_dentry); + } + + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +int au_unlink_wh_dentry(struct inode *hidden_dir, struct dentry *wh_dentry, + struct dentry *dentry, int dlgt) +{ + int err; + + LKTRTrace("hi%lu, wh %.*s, d %p\n", hidden_dir->i_ino, + DLNPair(wh_dentry), dentry); + DEBUG_ON((dentry && dbwh(dentry) == -1) + || !wh_dentry->d_inode + || !S_ISREG(wh_dentry->d_inode->i_mode)); + IMustLock(hidden_dir); + + err = vfsub_unlink(hidden_dir, wh_dentry, dlgt); + //if (LktrCond) err = -1; // unavailable + if (!err && dentry) + set_dbwh(dentry, -1); + + TraceErr(err); + return err; +} + +static int unlink_wh_name(struct dentry *hidden_parent, struct qstr *wh, + struct lkup_args *lkup) +{ + int err; + struct inode *hidden_dir; + struct dentry *hidden_dentry; + + LKTRTrace("%.*s/%.*s\n", DLNPair(hidden_parent), LNPair(wh)); + hidden_dir = hidden_parent->d_inode; + IMustLock(hidden_dir); + + // au_test_perm() is already done + hidden_dentry = lkup_one(wh->name, hidden_parent, wh->len, lkup); + //if (LktrCond) {dput(hidden_dentry); hidden_dentry = ERR_PTR(-1);} + if (!IS_ERR(hidden_dentry)) { + err = 0; + if (hidden_dentry->d_inode) + err = vfsub_unlink(hidden_dir, hidden_dentry, + lkup->dlgt); + dput(hidden_dentry); + } else + err = PTR_ERR(hidden_dentry); + + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +static void clean_wh(struct inode *h_dir, struct dentry *wh) +{ + TraceEnter(); + if (wh->d_inode) { + int err = vfsub_unlink(h_dir, wh, /*dlgt*/0); + if (unlikely(err)) + Warn("failed unlink %.*s (%d), ignored.\n", + DLNPair(wh), err); + } +} + +static void clean_plink(struct inode *h_dir, struct dentry *plink) +{ + TraceEnter(); + if (plink->d_inode) { + int err = vfsub_rmdir(h_dir, plink, /*dlgt*/0); + if (unlikely(err)) + Warn("failed rmdir %.*s (%d), ignored.\n", + DLNPair(plink), err); + } +} + +static int test_linkable(struct inode *h_dir) +{ + if (h_dir->i_op && h_dir->i_op->link) + return 0; + return -ENOSYS; +} + +static int plink_dir(struct inode *h_dir, struct dentry *plink) +{ + int err; + + err = -EEXIST; + if (!plink->d_inode) { + int mode = S_IRWXU; + if (unlikely(au_is_nfs(plink->d_sb))) + mode |= S_IXUGO; + err = vfsub_mkdir(h_dir, plink, mode, /*dlgt*/0); + } else if (S_ISDIR(plink->d_inode->i_mode)) + err = 0; + else + Err("unknown %.*s exists\n", DLNPair(plink)); + + return err; +} + +/* + * initialize the whiteout base file/dir for @br. + */ +int init_wh(struct dentry *h_root, struct aufs_branch *br, + struct vfsmount *nfsmnt, struct super_block *sb) +{ + int err; + struct dentry *wh, *plink; + struct inode *h_dir; + static struct qstr base_name[] = { + {.name = AUFS_WH_BASENAME, .len = sizeof(AUFS_WH_BASENAME) - 1}, + {.name = AUFS_WH_PLINKDIR, .len = sizeof(AUFS_WH_PLINKDIR) - 1} + }; + struct lkup_args lkup = { + .nfsmnt = nfsmnt, + .dlgt = 0 // always no dlgt + }; + const int do_plink = au_flag_test(sb, AuFlag_PLINK); + + LKTRTrace("nfsmnt %p\n", nfsmnt); + BrWhMustWriteLock(br); + SiMustWriteLock(sb); + h_dir = h_root->d_inode; + IMustLock(h_dir); + + // doubly whiteouted + wh = lkup_wh(h_root, base_name + 0, &lkup); + //if (LktrCond) {dput(wh); wh = ERR_PTR(-1);} + err = PTR_ERR(wh); + if (IS_ERR(wh)) + goto out; + DEBUG_ON(br->br_wh && br->br_wh != wh); + + plink = lkup_wh(h_root, base_name + 1, &lkup); + err = PTR_ERR(plink); + if (IS_ERR(plink)) + goto out_dput_wh; + DEBUG_ON(br->br_plink && br->br_plink != plink); + + dput(br->br_wh); + dput(br->br_plink); + br->br_wh = br->br_plink = NULL; + + err = 0; + switch (br->br_perm) { + case AuBr_RR: + case AuBr_RO: + case AuBr_RRWH: + case AuBr_ROWH: + clean_wh(h_dir, wh); + clean_plink(h_dir, plink); + break; + + case AuBr_RWNoLinkWH: + clean_wh(h_dir, wh); + if (do_plink) { + err = test_linkable(h_dir); + if (unlikely(err)) + goto out_nolink; + + err = plink_dir(h_dir, plink); + if (unlikely(err)) + goto out_err; + br->br_plink = dget(plink); + } else + clean_plink(h_dir, plink); + break; + + case AuBr_RW: + /* + * for the moment, aufs supports the branch filesystem + * which does not support link(2). + * testing on FAT which does not support i_op->setattr() fully either, + * copyup failed. + * finally, such filesystem will not be used as the writable branch. + */ + err = test_linkable(h_dir); + if (unlikely(err)) + goto out_nolink; + + err = -EEXIST; + if (!wh->d_inode) + err = vfsub_create(h_dir, wh, WH_MASK, NULL, /*dlgt*/0); + else if (S_ISREG(wh->d_inode->i_mode)) + err = 0; + else + Err("unknown %.*s/%.*s exists\n", + DLNPair(h_root), DLNPair(wh)); + if (unlikely(err)) + goto out_err; + + if (do_plink) { + err = plink_dir(h_dir, plink); + if (unlikely(err)) + goto out_err; + br->br_plink = dget(plink); + } else + clean_plink(h_dir, plink); + br->br_wh = dget(wh); + break; + + default: + BUG(); + } + + out_dput: + dput(plink); + out_dput_wh: + dput(wh); + out: + TraceErr(err); + return err; + out_nolink: + Err("%.*s doesn't support link(2), use noplink and rw+nolwh\n", + DLNPair(h_root)); + goto out_dput; + out_err: + Err("an error(%d) on the writable branch %.*s(%s)\n", + err, DLNPair(h_root), au_sbtype(h_root->d_sb)); + goto out_dput; +} + +struct reinit_br_wh { + struct super_block *sb; + struct aufs_branch *br; +}; + +static void reinit_br_wh(void *arg) +{ + int err; + struct reinit_br_wh *a = arg; + struct inode *hidden_dir, *dir; + struct dentry *hidden_root; + aufs_bindex_t bindex; + + TraceEnter(); + DEBUG_ON(!a->br->br_wh || !a->br->br_wh->d_inode || current->fsuid); + + err = 0; + /* big lock */ + si_write_lock(a->sb); + if (unlikely(!br_writable(a->br->br_perm))) + goto out; + bindex = find_brindex(a->sb, a->br->br_id); + if (unlikely(bindex < 0)) + goto out; + + dir = a->sb->s_root->d_inode; + hidden_root = a->br->br_wh->d_parent; + hidden_dir = hidden_root->d_inode; + DEBUG_ON(!hidden_dir->i_op || !hidden_dir->i_op->link); + hdir_lock(hidden_dir, dir, bindex); + br_wh_write_lock(a->br); + err = vfsub_unlink(hidden_dir, a->br->br_wh, /*dlgt*/0); + //if (LktrCond) err = -1; + dput(a->br->br_wh); + a->br->br_wh = NULL; + if (!err) + err = init_wh(hidden_root, a->br, au_do_nfsmnt(a->br->br_mnt), + a->sb); + br_wh_write_unlock(a->br); + hdir_unlock(hidden_dir, dir, bindex); + + out: + atomic_dec(&a->br->br_wh_running); + br_put(a->br); + si_write_unlock(a->sb); + au_mntput(a->sb); + kfree(arg); + if (unlikely(err)) + IOErr("err %d\n", err); +} + +static void kick_reinit_br_wh(struct super_block *sb, struct aufs_branch *br) +{ + int do_dec; + struct reinit_br_wh *arg; + + do_dec = 1; + if (atomic_inc_return(&br->br_wh_running) != 1) + goto out; + + // ignore ENOMEM + arg = kmalloc(sizeof(*arg), GFP_KERNEL); + if (arg) { + // dec(wh_running), kfree(arg) and br_put() in reinit function + arg->sb = sb; + arg->br = br; + br_get(br); + /* prohibit umount */ + au_mntget(sb); + au_wkq_nowait(reinit_br_wh, arg, /*dlgt*/0); + do_dec = 0; + } + + out: + if (do_dec) + atomic_dec(&br->br_wh_running); +} + +/* + * create the whiteoute @wh. + */ +static int link_or_create_wh(struct dentry *wh, struct super_block *sb, + aufs_bindex_t bindex) +{ + int err, dlgt; + struct aufs_branch *br; + struct dentry *hidden_parent; + struct inode *hidden_dir; + + LKTRTrace("%.*s\n", DLNPair(wh)); + SiMustReadLock(sb); + hidden_parent = wh->d_parent; + hidden_dir = hidden_parent->d_inode; + IMustLock(hidden_dir); + + dlgt = need_dlgt(sb); + br = stobr(sb, bindex); + br_wh_read_lock(br); + if (br->br_wh) { + err = vfsub_link(br->br_wh, hidden_dir, wh, dlgt); + if (!err || err != -EMLINK) + goto out; + + // link count full. re-initialize br_wh. + kick_reinit_br_wh(sb, br); + } + + // return this error in this context + err = vfsub_create(hidden_dir, wh, WH_MASK, NULL, dlgt); + + out: + br_wh_read_unlock(br); + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +/* + * create or remove the diropq. + */ +static struct dentry *do_diropq(struct dentry *dentry, aufs_bindex_t bindex, + int do_create, int dlgt) +{ + struct dentry *opq_dentry, *hidden_dentry; + struct inode *hidden_dir; + int err; + struct super_block *sb; + struct lkup_args lkup; + + LKTRTrace("%.*s, bindex %d, do_create %d\n", DLNPair(dentry), + bindex, do_create); + hidden_dentry = au_h_dptr_i(dentry, bindex); + DEBUG_ON(!hidden_dentry); + hidden_dir = hidden_dentry->d_inode; + DEBUG_ON(!hidden_dir || !S_ISDIR(hidden_dir->i_mode)); + IMustLock(hidden_dir); + + // already checked by au_test_perm(). + sb = dentry->d_sb; + lkup.nfsmnt = au_nfsmnt(sb, bindex); + lkup.dlgt = dlgt; + opq_dentry = lkup_one(diropq_name.name, hidden_dentry, diropq_name.len, + &lkup); + //if (LktrCond) {dput(opq_dentry); opq_dentry = ERR_PTR(-1);} + if (IS_ERR(opq_dentry)) + goto out; + + if (do_create) { + DEBUG_ON(opq_dentry->d_inode); + err = link_or_create_wh(opq_dentry, sb, bindex); + //if (LktrCond) {vfs_unlink(hidden_dir, opq_dentry); err = -1;} + if (!err) { + set_dbdiropq(dentry, bindex); + goto out; /* success */ + } + } else { + DEBUG_ON(/* !S_ISDIR(dentry->d_inode->i_mode) + * || */!opq_dentry->d_inode); + err = vfsub_unlink(hidden_dir, opq_dentry, lkup.dlgt); + //if (LktrCond) err = -1; + if (!err) + set_dbdiropq(dentry, -1); + } + dput(opq_dentry); + opq_dentry = ERR_PTR(err); + + out: + TraceErrPtr(opq_dentry); + return opq_dentry; +} + +struct do_diropq_args { + struct dentry **errp; + struct dentry *dentry; + aufs_bindex_t bindex; + int do_create, dlgt; +}; + +static void call_do_diropq(void *args) +{ + struct do_diropq_args *a = args; + *a->errp = do_diropq(a->dentry, a->bindex, a->do_create, a->dlgt); +} + +struct dentry *sio_diropq(struct dentry *dentry, aufs_bindex_t bindex, + int do_create, int dlgt) +{ + struct dentry *diropq, *hidden_dentry; + + LKTRTrace("%.*s, bindex %d, do_create %d\n", + DLNPair(dentry), bindex, do_create); + + hidden_dentry = au_h_dptr_i(dentry, bindex); + if (!au_test_perm(hidden_dentry->d_inode, MAY_EXEC | MAY_WRITE, dlgt)) + diropq = do_diropq(dentry, bindex, do_create, dlgt); + else { + struct do_diropq_args args = { + .errp = &diropq, + .dentry = dentry, + .bindex = bindex, + .do_create = do_create, + .dlgt = dlgt + }; + au_wkq_wait(call_do_diropq, &args, /*dlgt*/0); + } + + TraceErrPtr(diropq); + return diropq; +} + +/* ---------------------------------------------------------------------- */ + +/* + * lookup whiteout dentry. + * @hidden_parent: hidden parent dentry which must exist and be locked + * @base_name: name of dentry which will be whiteouted + * returns dentry for whiteout. + */ +struct dentry *lkup_wh(struct dentry *hidden_parent, struct qstr *base_name, + struct lkup_args *lkup) +{ + int err; + struct qstr wh_name; + struct dentry *wh_dentry; + + LKTRTrace("%.*s/%.*s\n", DLNPair(hidden_parent), LNPair(base_name)); + IMustLock(hidden_parent->d_inode); + + err = au_alloc_whname(base_name->name, base_name->len, &wh_name); + //if (LktrCond) {au_free_whname(&wh_name); err = -1;} + wh_dentry = ERR_PTR(err); + if (!err) { + // do not superio. + wh_dentry = lkup_one(wh_name.name, hidden_parent, wh_name.len, + lkup); + au_free_whname(&wh_name); + } + TraceErrPtr(wh_dentry); + return wh_dentry; +} + +/* + * link/create a whiteout for @dentry on @bindex. + */ +struct dentry *simple_create_wh(struct dentry *dentry, aufs_bindex_t bindex, + struct dentry *hidden_parent, + struct lkup_args *lkup) +{ + struct dentry *wh_dentry; + int err; + struct super_block *sb; + + LKTRTrace("%.*s/%.*s on b%d\n", DLNPair(hidden_parent), + DLNPair(dentry), bindex); + + sb = dentry->d_sb; + wh_dentry = lkup_wh(hidden_parent, &dentry->d_name, lkup); + //au_nfsmnt(sb, bindex), need_dlgt(sb)); + //if (LktrCond) {dput(wh_dentry); wh_dentry = ERR_PTR(-1);} + if (!IS_ERR(wh_dentry) && !wh_dentry->d_inode) { + IMustLock(hidden_parent->d_inode); + err = link_or_create_wh(wh_dentry, sb, bindex); + if (!err) + set_dbwh(dentry, bindex); + else { + dput(wh_dentry); + wh_dentry = ERR_PTR(err); + } + } + + TraceErrPtr(wh_dentry); + return wh_dentry; +} + +/* ---------------------------------------------------------------------- */ + +/* Delete all whiteouts in this directory in branch bindex. */ +static int del_wh_children(struct aufs_nhash *whlist, + struct dentry *hidden_parent, aufs_bindex_t bindex, + struct lkup_args *lkup) +{ + int err, i; + struct qstr wh_name; + char *p; + struct inode *hidden_dir; + struct hlist_head *head; + struct aufs_wh *tpos; + struct hlist_node *pos; + struct aufs_destr *str; + + LKTRTrace("%.*s\n", DLNPair(hidden_parent)); + hidden_dir = hidden_parent->d_inode; + IMustLock(hidden_dir); + DEBUG_ON(IS_RDONLY(hidden_dir)); + //SiMustReadLock(??); + + err = -ENOMEM; + wh_name.name = p = __getname(); + //if (LktrCond) {__putname(p); wh_name.name = p = NULL;} + if (unlikely(!wh_name.name)) + goto out; + memcpy(p, AUFS_WH_PFX, AUFS_WH_PFX_LEN); + p += AUFS_WH_PFX_LEN; + + // already checked by au_test_perm(). + err = 0; + for (i = 0; !err && i < AUFS_NHASH_SIZE; i++) { + head = whlist->heads + i; + hlist_for_each_entry(tpos, pos, head, wh_hash) { + if (tpos->wh_bindex != bindex) + continue; + str = &tpos->wh_str; + if (str->len + AUFS_WH_PFX_LEN <= PATH_MAX) { + memcpy(p, str->name, str->len); + wh_name.len = AUFS_WH_PFX_LEN + str->len; + err = unlink_wh_name(hidden_parent, &wh_name, + lkup); + //if (LktrCond) err = -1; + if (!err) + continue; + break; + } + IOErr("whiteout name too long %.*s\n", + str->len, str->name); + err = -EIO; + break; + } + } + __putname(wh_name.name); + + out: + TraceErr(err); + return err; +} + +struct del_wh_children_args { + int *errp; + struct aufs_nhash *whlist; + struct dentry *hidden_parent; + aufs_bindex_t bindex; + struct lkup_args *lkup; +}; + +static void call_del_wh_children(void *args) +{ + struct del_wh_children_args *a = args; + *a->errp = del_wh_children(a->whlist, a->hidden_parent, a->bindex, + a->lkup); +} + +/* ---------------------------------------------------------------------- */ + +/* + * rmdir the whiteouted temporary named dir @hidden_dentry. + * @whlist: whiteouted children. + */ +int rmdir_whtmp(struct dentry *hidden_dentry, struct aufs_nhash *whlist, + aufs_bindex_t bindex, struct inode *dir, struct inode *inode) +{ + int err; + struct inode *hidden_inode, *hidden_dir; + struct lkup_args lkup; + struct super_block *sb; + + LKTRTrace("hd %.*s, b%d, i%lu\n", + DLNPair(hidden_dentry), bindex, dir->i_ino); + IMustLock(dir); + IiMustAnyLock(dir); + hidden_dir = hidden_dentry->d_parent->d_inode; + IMustLock(hidden_dir); + + sb = inode->i_sb; + lkup.nfsmnt = au_nfsmnt(sb, bindex); + lkup.dlgt = need_dlgt(sb); + hidden_inode = hidden_dentry->d_inode; + DEBUG_ON(hidden_inode != au_h_iptr_i(inode, bindex)); + hdir2_lock(hidden_inode, inode, bindex); + if (!au_test_perm(hidden_inode, MAY_EXEC | MAY_WRITE, lkup.dlgt)) + err = del_wh_children(whlist, hidden_dentry, bindex, &lkup); + else { + // ugly + int dlgt = lkup.dlgt; + struct del_wh_children_args args = { + .errp = &err, + .whlist = whlist, + .hidden_parent = hidden_dentry, + .bindex = bindex, + .lkup = &lkup + }; + + lkup.dlgt = 0; + au_wkq_wait(call_del_wh_children, &args, /*dlgt*/0); + lkup.dlgt = dlgt; + } + hdir_unlock(hidden_inode, inode, bindex); + + if (!err) { + err = vfsub_rmdir(hidden_dir, hidden_dentry, lkup.dlgt); + //d_drop(hidden_dentry); + //if (LktrCond) err = -1; + } + + if (!err) { + if (ibstart(dir) == bindex) { + au_cpup_attr_timesizes(dir); + //au_cpup_attr_nlink(dir); + dir->i_nlink--; + } + return 0; /* success */ + } + + Warn("failed removing %.*s(%d), ignored\n", + DLNPair(hidden_dentry), err); + return err; +} + +static void do_rmdir_whtmp(void *arg) +{ + int err; + struct rmdir_whtmp_arg *a = arg; + struct super_block *sb; + + LKTRTrace("%.*s, b%d, dir i%lu\n", + DLNPair(a->h_dentry), a->bindex, a->dir->i_ino); + + i_lock(a->dir); + sb = a->dir->i_sb; + si_read_lock(sb); + err = test_ro(sb, a->bindex, NULL); + if (!err) { + struct inode *hidden_dir = a->h_dentry->d_parent->d_inode; + + ii_write_lock_child(a->inode); + ii_write_lock_parent(a->dir); + hdir_lock(hidden_dir, a->dir, a->bindex); + err = rmdir_whtmp(a->h_dentry, &a->whlist, a->bindex, + a->dir, a->inode); + hdir_unlock(hidden_dir, a->dir, a->bindex); + ii_write_unlock(a->dir); + ii_write_unlock(a->inode); + } + dput(a->h_dentry); + nhash_fin(&a->whlist); + iput(a->inode); + si_read_unlock(sb); + au_mntput(sb); + i_unlock(a->dir); + iput(a->dir); + kfree(arg); + if (unlikely(err)) + IOErr("err %d\n", err); +} + +void kick_rmdir_whtmp(struct dentry *hidden_dentry, struct aufs_nhash *whlist, + aufs_bindex_t bindex, struct inode *dir, + struct inode *inode, struct rmdir_whtmp_arg *arg) +{ + LKTRTrace("%.*s\n", DLNPair(hidden_dentry)); + IMustLock(dir); + + // all post-process will be done in do_rmdir_whtmp(). + arg->h_dentry = dget(hidden_dentry); + nhash_init(&arg->whlist); + nhash_move(&arg->whlist, whlist); + arg->bindex = bindex; + arg->dir = igrab(dir); + arg->inode = igrab(inode); + /* prohibit umount */ + au_mntget(dir->i_sb); + + au_wkq_nowait(do_rmdir_whtmp, arg, /*dlgt*/0); +} diff --git a/fs/aufs/whout.h b/fs/aufs/whout.h new file mode 100755 index 0000000..d44c3cd --- /dev/null +++ b/fs/aufs/whout.h @@ -0,0 +1,87 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: whout.h,v 1.8 2007/05/14 03:41:52 sfjro Exp $ */ + +#ifndef __AUFS_WHOUT_H__ +#define __AUFS_WHOUT_H__ + +#ifdef __KERNEL__ + +#include +#include + +int au_alloc_whname(const char *name, int len, struct qstr *wh); +void au_free_whname(struct qstr *wh); + +struct lkup_args; +int is_wh(struct dentry *h_parent, struct qstr *wh_name, int try_sio, + struct lkup_args *lkup); +int is_diropq(struct dentry *h_dentry, struct lkup_args *lkup); + +struct dentry *lkup_whtmp(struct dentry *h_parent, struct qstr *prefix, + struct lkup_args *lkup); +int rename_whtmp(struct dentry *dentry, aufs_bindex_t bindex); +int au_unlink_wh_dentry(struct inode *h_dir, struct dentry *wh_dentry, + struct dentry *dentry, int dlgt); + +struct aufs_branch; +int init_wh(struct dentry *h_parent, struct aufs_branch *br, + struct vfsmount *nfsmnt, struct super_block *sb); + +struct dentry *sio_diropq(struct dentry *dentry, aufs_bindex_t bindex, + int do_create, int dlgt); + +struct dentry *lkup_wh(struct dentry *h_parent, struct qstr *base_name, + struct lkup_args *lkup); +struct dentry *simple_create_wh(struct dentry *dentry, aufs_bindex_t bindex, + struct dentry *h_parent, + struct lkup_args *lkup); + +/* real rmdir the whiteout-ed dir */ +struct rmdir_whtmp_arg { + struct dentry *h_dentry; + struct aufs_nhash whlist; + aufs_bindex_t bindex; + struct inode *dir, *inode; +}; + +struct aufs_nhash; +int rmdir_whtmp(struct dentry *h_dentry, struct aufs_nhash *whlist, + aufs_bindex_t bindex, struct inode *dir, struct inode *inode); +void kick_rmdir_whtmp(struct dentry *h_dentry, struct aufs_nhash *whlist, + aufs_bindex_t bindex, struct inode *dir, + struct inode *inode, struct rmdir_whtmp_arg *arg); + +/* ---------------------------------------------------------------------- */ + +static inline +struct dentry *create_diropq(struct dentry *dentry, aufs_bindex_t bindex, + int dlgt) +{ + return sio_diropq(dentry, bindex, 1, dlgt); +} + +static inline +int remove_diropq(struct dentry *dentry, aufs_bindex_t bindex, int dlgt) +{ + return PTR_ERR(sio_diropq(dentry, bindex, 0, dlgt)); +} + +#endif /* __KERNEL__ */ +#endif /* __AUFS_WHOUT_H__ */ diff --git a/fs/aufs/wkq.c b/fs/aufs/wkq.c new file mode 100755 index 0000000..b5ab023 --- /dev/null +++ b/fs/aufs/wkq.c @@ -0,0 +1,283 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: wkq.c,v 1.14 2007/05/14 03:39:10 sfjro Exp $ */ + +#include +#include "aufs.h" + +struct au_wkq *au_wkq; + +struct au_cred { +#ifdef CONFIG_AUFS_DLGT + uid_t fsuid; + gid_t fsgid; + kernel_cap_t cap_effective, cap_inheritable, cap_permitted; + //unsigned keep_capabilities:1; + //struct user_struct *user; + //struct fs_struct *fs; + //struct nsproxy *nsproxy; +#endif +}; + +struct au_wkinfo { + struct work_struct wk; + + unsigned int wait:1; + unsigned int dlgt:1; + struct au_cred cred; + + au_wkq_func_t func; + void *args; + + atomic_t *busyp; + struct completion *comp; +}; + +/* ---------------------------------------------------------------------- */ + +#ifdef CONFIG_AUFS_DLGT +static void cred_store(struct au_cred *cred) +{ + cred->fsuid = current->fsuid; + cred->fsgid = current->fsgid; + cred->cap_effective = current->cap_effective; + cred->cap_inheritable = current->cap_inheritable; + cred->cap_permitted = current->cap_permitted; +} + +static void cred_revert(struct au_cred *cred) +{ + DEBUG_ON(!is_au_wkq(current)); + current->fsuid = cred->fsuid; + current->fsgid = cred->fsgid; + current->cap_effective = cred->cap_effective; + current->cap_inheritable = cred->cap_inheritable; + current->cap_permitted = cred->cap_permitted; +} + +static void cred_switch(struct au_cred *old, struct au_cred *new) +{ + cred_store(old); + cred_revert(new); +} +#endif + +/* ---------------------------------------------------------------------- */ + +static void update_busy(struct au_wkq *wkq, struct au_wkinfo *wkinfo) +{ +#ifdef CONFIG_AUFS_SYSAUFS + unsigned int new, old; + + do { + new = atomic_read(wkinfo->busyp); + old = wkq->max_busy; + if (new <= old) + break; + } while (cmpxchg(&wkq->max_busy, old, new) == old); +#endif +} + +static int enqueue(struct au_wkq *wkq, struct au_wkinfo *wkinfo) +{ + wkinfo->busyp = &wkq->busy; + update_busy(wkq, wkinfo); + if (wkinfo->wait) + return !queue_work(wkq->q, &wkinfo->wk); + else + return !schedule_work(&wkinfo->wk); +} + +static void do_wkq(struct au_wkinfo *wkinfo) +{ + unsigned int idle, n; + int i, idle_idx; + + TraceEnter(); + + while (1) { + if (wkinfo->wait) { + idle_idx = 0; + idle = UINT_MAX; + for (i = 0; i < aufs_nwkq; i++) { + n = atomic_inc_return(&au_wkq[i].busy); + if (n == 1 && !enqueue(au_wkq + i, wkinfo)) + return; /* success */ + + if (n < idle) { + idle_idx = i; + idle = n; + } + atomic_dec(&au_wkq[i].busy); + } + } else + idle_idx = aufs_nwkq; + + atomic_inc(&au_wkq[idle_idx].busy); + if (!enqueue(au_wkq + idle_idx, wkinfo)) + return; /* success */ + + /* impossible? */ + Warn1("failed to queue_work()\n"); + yield(); + } +} + +static AuWkqFunc(wkq_func, wk) +{ + struct au_wkinfo *wkinfo = container_of(wk, struct au_wkinfo, wk); + + LKTRTrace("wkinfo{%u, %u, %p, %p, %p}\n", + wkinfo->wait, wkinfo->dlgt, wkinfo->func, wkinfo->busyp, + wkinfo->comp); +#ifdef CONFIG_AUFS_DLGT + if (!wkinfo->dlgt) + wkinfo->func(wkinfo->args); + else { + struct au_cred cred; + cred_switch(&cred, &wkinfo->cred); + wkinfo->func(wkinfo->args); + cred_revert(&cred); + } +#else + wkinfo->func(wkinfo->args); +#endif + atomic_dec(wkinfo->busyp); + if (wkinfo->wait) + complete(wkinfo->comp); + else { + kfree(wkinfo); + module_put(THIS_MODULE); + } +} + +void au_wkq_run(au_wkq_func_t func, void *args, int dlgt, int do_wait) +{ + DECLARE_COMPLETION_ONSTACK(comp); + struct au_wkinfo _wkinfo = { + .wait = 1, + .dlgt = !!dlgt, + .func = func, + .args = args, + .comp = &comp + }, *wkinfo = &_wkinfo; + + LKTRTrace("dlgt %d, do_wait %d\n", dlgt, do_wait); + DEBUG_ON(is_au_wkq(current)); + + if (unlikely(!do_wait)) { + static DECLARE_WAIT_QUEUE_HEAD(wq); + /* + * never fail. + * wkq_func() must free this wkinfo. + * it highly depends upon the implementation of workqueue. + */ + wait_event(wq, (wkinfo = kmalloc(sizeof(*wkinfo), GFP_KERNEL))); + wkinfo->wait = 0; + wkinfo->dlgt = !!dlgt; + wkinfo->func = func; + wkinfo->args = args; + wkinfo->comp = NULL; + __module_get(THIS_MODULE); + } + + AuInitWkq(&wkinfo->wk, wkq_func); +#ifdef CONFIG_AUFS_DLGT + if (dlgt) + cred_store(&wkinfo->cred); +#endif + do_wkq(wkinfo); + if (do_wait) + wait_for_completion(wkinfo->comp); +} + +#if 0 +void au_wkq_wait_nwtask(void) +{ + static DECLARE_WAIT_QUEUE_HEAD(wq); + wait_event(wq, !atomic_read(&au_wkq[aufs_nwkq].busy)); +} +#endif + +void au_wkq_fin(void) +{ + int i; + + TraceEnter(); + + for (i = 0; i < aufs_nwkq; i++) + if (au_wkq[i].q && !IS_ERR(au_wkq[i].q)) + destroy_workqueue(au_wkq[i].q); + kfree(au_wkq); +} + +int __init au_wkq_init(void) +{ + int err, i; + struct au_wkq *nowaitq; + + LKTRTrace("%d\n", aufs_nwkq); + + /* '+1' is for accounting of nowait queue */ + err = -ENOMEM; + au_wkq = kcalloc(aufs_nwkq + 1, sizeof(*au_wkq), GFP_KERNEL); + if (unlikely(!au_wkq)) + goto out; + + err = 0; + for (i = 0; i < aufs_nwkq; i++) { + au_wkq[i].q = create_singlethread_workqueue(AUFS_WKQ_NAME); + if (au_wkq[i].q && !IS_ERR(au_wkq[i].q)) { + atomic_set(&au_wkq[i].busy, 0); + au_wkq[i].max_busy = 0; + continue; + } + + err = PTR_ERR(au_wkq[i].q); + au_wkq_fin(); + break; + } + + /* nowait accounting */ + nowaitq = au_wkq + aufs_nwkq; + atomic_set(&nowaitq->busy, 0); + nowaitq->max_busy = 0; + nowaitq->q = NULL; + +#if 0 // test accouting + if (!err) { + static void f(void *args) { + DbgSleep(1); + } + int i; + //au_debug_on(); + LKTRTrace("f %p\n", f); + for (i = 0; i < 10; i++) + au_wkq_nowait(f, NULL, 0); + for (i = 0; i < aufs_nwkq; i++) + au_wkq_wait(f, NULL, 0); + DbgSleep(11); + //au_debug_off(); + } +#endif + + out: + TraceErr(err); + return err; +} diff --git a/fs/aufs/wkq.h b/fs/aufs/wkq.h new file mode 100755 index 0000000..cc1bb25 --- /dev/null +++ b/fs/aufs/wkq.h @@ -0,0 +1,81 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: wkq.h,v 1.9 2007/05/14 03:39:10 sfjro Exp $ */ + +#ifndef __AUFS_WKQ_H__ +#define __AUFS_WKQ_H__ + +#ifdef __KERNEL__ + +#include +#include +#include + +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,18) +#define DECLARE_COMPLETION_ONSTACK(work) DECLARE_COMPLETION(work) +#endif + +/* ---------------------------------------------------------------------- */ + +/* internal workqueue named AUFS_WKQ_NAME */ +struct au_wkq { + struct workqueue_struct *q; + + /* accounting */ + atomic_t busy; + unsigned int max_busy; +} ;//__attribute__ ((aligned)); + +typedef void (*au_wkq_func_t)(void *args); + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20) +#define AuInitWkq(wk, func) INIT_WORK(wk, func) +#define AuWkqFunc(name, arg) void name(struct work_struct *arg) +#else +typedef void (*work_func_t)(void *arg); +#define AuInitWkq(wk, func) INIT_WORK(wk, func, wk) +#define AuWkqFunc(name, arg) void name(void *arg) +#endif + +extern struct au_wkq *au_wkq; + +void au_wkq_run(au_wkq_func_t func, void *args, int dlgt, int do_wait); +//void au_wkq_wait_nwtask(void); +int __init au_wkq_init(void); +void au_wkq_fin(void); + +/* ---------------------------------------------------------------------- */ + +static inline int is_au_wkq(struct task_struct *tsk) +{ + return (!tsk->mm && !strcmp(current->comm, AUFS_WKQ_NAME)); +} + +static inline void au_wkq_wait(au_wkq_func_t func, void *args, int dlgt) +{ + au_wkq_run(func, args, dlgt, /*do_wait*/1); +} + +static inline void au_wkq_nowait(au_wkq_func_t func, void *args, int dlgt) +{ + au_wkq_run(func, args, dlgt, /*do_wait*/0); +} + +#endif /* __KERNEL__ */ +#endif /* __AUFS_WKQ_H__ */ diff --git a/fs/aufs/xino.c b/fs/aufs/xino.c new file mode 100755 index 0000000..145491e --- /dev/null +++ b/fs/aufs/xino.c @@ -0,0 +1,644 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: xino.c,v 1.27 2007/05/14 03:39:10 sfjro Exp $ */ + +//#include +#include +#include +#include "aufs.h" + +static readf_t find_readf(struct file *h_file) +{ + const struct file_operations *fop = h_file->f_op; + + if (fop) { + if (fop->read) + return fop->read; + if (fop->aio_read) + return do_sync_read; + } + return ERR_PTR(-ENOSYS); +} + +static writef_t find_writef(struct file *h_file) +{ + const struct file_operations *fop = h_file->f_op; + + if (fop) { + if (fop->write) + return fop->write; + if (fop->aio_write) + return do_sync_write; + } + return ERR_PTR(-ENOSYS); +} + +/* ---------------------------------------------------------------------- */ + +static ssize_t xino_fread(readf_t func, struct file *file, void *buf, + size_t size, loff_t *pos) +{ + ssize_t err; + mm_segment_t oldfs; + + LKTRTrace("%.*s, sz %lu, *pos %Ld\n", + DLNPair(file->f_dentry), (unsigned long)size, *pos); + + oldfs = get_fs(); + set_fs(KERNEL_DS); + do { + err = func(file, (char __user*)buf, size, pos); + } while (err == -EAGAIN || err == -EINTR); + set_fs(oldfs); + +#if 0 + if (err > 0) + fsnotify_access(file->f_dentry); +#endif + + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +static ssize_t do_xino_fwrite(writef_t func, struct file *file, void *buf, + size_t size, loff_t *pos) +{ + ssize_t err; + mm_segment_t oldfs; + + lockdep_off(); + oldfs = get_fs(); + set_fs(KERNEL_DS); + do { + err = func(file, (const char __user*)buf, size, pos); + } while (err == -EAGAIN || err == -EINTR); + set_fs(oldfs); + lockdep_on(); + +#if 0 + if (err > 0) + fsnotify_modify(file->f_dentry); +#endif + + TraceErr(err); + return err; +} + +struct do_xino_fwrite_args { + ssize_t *errp; + writef_t func; + struct file *file; + void *buf; + size_t size; + loff_t *pos; +}; + +static void call_do_xino_fwrite(void *args) +{ + struct do_xino_fwrite_args *a = args; + *a->errp = do_xino_fwrite(a->func, a->file, a->buf, a->size, a->pos); +} + +static ssize_t xino_fwrite(writef_t func, struct file *file, void *buf, + size_t size, loff_t *pos) +{ + ssize_t err; + + LKTRTrace("%.*s, sz %lu, *pos %Ld\n", + DLNPair(file->f_dentry), (unsigned long)size, *pos); + + // signal block and no wkq? + /* + * it breaks RLIMIT_FSIZE and normal user's limit, + * users should care about quota and real 'filesystem full.' + */ + if (!is_au_wkq(current)) { + struct do_xino_fwrite_args args = { + .errp = &err, + .func = func, + .file = file, + .buf = buf, + .size = size, + .pos = pos + }; + au_wkq_wait(call_do_xino_fwrite, &args, /*dlgt*/0); + } else + err = do_xino_fwrite(func, file, buf, size, pos); + + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +/* + * write @ino to the xinofile for the specified branch{@sb, @bindex} + * at the position of @_ino. + * when @ino is zero, it is written to the xinofile and means no entry. + */ +int xino_write(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino, + struct xino *xino) +{ + struct aufs_branch *br; + loff_t pos; + ssize_t sz; + + LKTRTrace("b%d, hi%lu, i%lu\n", bindex, h_ino, xino->ino); + //DEBUG_ON(!xino->ino /* || !xino->h_gen */); + //WARN_ON(bindex == 0 && h_ino == 31); + + if (unlikely(!au_flag_test(sb, AuFlag_XINO))) + return 0; + + br = stobr(sb, bindex); + DEBUG_ON(!br || !br->br_xino); + pos = h_ino * sizeof(*xino); + sz = xino_fwrite(br->br_xino_write, br->br_xino, xino, sizeof(*xino), + &pos); + //if (LktrCond) sz = 1; + if (sz == sizeof(*xino)) + return 0; /* success */ + + IOErr("write failed (%ld)\n", (long)sz); + return -EIO; +} + +int xino_write0(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino) +{ + struct xino xino = { + .ino = 0 + }; + return xino_write(sb, bindex, h_ino, &xino); +} + +// why is not atomic_long_inc_return defined? +static DEFINE_SPINLOCK(alir_lock); +static long atomic_long_inc_return(atomic_long_t *a) +{ + long l; + + spin_lock(&alir_lock); + atomic_long_inc(a); + l = atomic_long_read(a); + spin_unlock(&alir_lock); + return l; +} + +ino_t xino_new_ino(struct super_block *sb) +{ + ino_t ino; + + TraceEnter(); + ino = atomic_long_inc_return(&stosi(sb)->si_xino); + BUILD_BUG_ON(AUFS_FIRST_INO < AUFS_ROOT_INO); + if (ino >= AUFS_ROOT_INO) + return ino; + else { + atomic_long_dec(&stosi(sb)->si_xino); + IOErr1("inode number overflow\n"); + return 0; + } +} + +/* + * read @ino from xinofile for the specified branch{@sb, @bindex} + * at the position of @h_ino. + * if @ino does not exist and @do_new is true, get new one. + */ +int xino_read(struct super_block *sb, aufs_bindex_t bindex, ino_t h_ino, + struct xino *xino) +{ + int err; + struct aufs_branch *br; + struct file *file; + loff_t pos; + ssize_t sz; + + LKTRTrace("b%d, hi%lu\n", bindex, h_ino); + + err = 0; + xino->ino = 0; + if (unlikely(!au_flag_test(sb, AuFlag_XINO))) + return 0; /* no ino */ + + br = stobr(sb, bindex); + file = br->br_xino; + DEBUG_ON(!file); + pos = h_ino * sizeof(*xino); + if (i_size_read(file->f_dentry->d_inode) < pos + sizeof(*xino)) + return 0; /* no ino */ + + sz = xino_fread(br->br_xino_read, file, xino, sizeof(*xino), &pos); + if (sz == sizeof(*xino)) + return 0; /* success */ + + err = sz; + if (unlikely(sz >= 0)) { + err = -EIO; + IOErr("xino read error (%ld)\n", (long)sz); + } + + TraceErr(err); + return err; +} + +/* ---------------------------------------------------------------------- */ + +struct file *xino_create(struct super_block *sb, char *fname, int silent, + struct dentry *parent) +{ + struct file *file; + int err; + struct dentry *hidden_parent; + struct inode *hidden_dir; + //const int udba = au_flag_test(sb, AuFlag_UDBA_INOTIFY); + + LKTRTrace("%s\n", fname); + //DEBUG_ON(!au_flag_test(sb, AuFlag_XINO)); + + // LSM may detect it + // use sio? + file = vfsub_filp_open(fname, O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE, + S_IRUGO | S_IWUGO); + //file = ERR_PTR(-1); + if (IS_ERR(file)) { + if (!silent) + Err("open %s(%ld)\n", fname, PTR_ERR(file)); + return file; + } +#if 0 + if (unlikely(udba && parent)) + au_direval_dec(parent); +#endif + + /* keep file count */ + hidden_parent = dget_parent(file->f_dentry); + hidden_dir = hidden_parent->d_inode; + hi_lock_parent(hidden_dir); + err = vfsub_unlink(hidden_dir, file->f_dentry, /*dlgt*/0); +#if 0 + if (unlikely(!err && udba && parent)) + au_direval_dec(parent); +#endif + i_unlock(hidden_dir); + dput(hidden_parent); + if (unlikely(err)) { + if (!silent) + Err("unlink %s(%d)\n", fname, err); + goto out; + } + if (sb != file->f_dentry->d_sb) + return file; /* success */ + + if (!silent) + Err("%s must be outside\n", fname); + err = -EINVAL; + + out: + fput(file); + file = ERR_PTR(err); + return file; +} + +/* + * find another branch who is on the same filesystem of the specified + * branch{@btgt}. search until @bend. + */ +static int is_sb_shared(struct super_block *sb, aufs_bindex_t btgt, + aufs_bindex_t bend) +{ + aufs_bindex_t bindex; + struct super_block *tgt_sb = sbr_sb(sb, btgt); + + for (bindex = 0; bindex <= bend; bindex++) + if (unlikely(btgt != bindex && tgt_sb == sbr_sb(sb, bindex))) + return bindex; + return -1; +} + +/* + * create a new xinofile at the same place/path as @base_file. + */ +static struct file *xino_create2(struct file *base_file) +{ + struct file *file; + int err; + struct dentry *base, *dentry, *parent; + struct inode *dir; + struct qstr *name; + struct lkup_args lkup = { + .nfsmnt = NULL, + .dlgt = 0 + }; + + base = base_file->f_dentry; + LKTRTrace("%.*s\n", DLNPair(base)); + parent = dget_parent(base); + dir = parent->d_inode; + IMustLock(dir); + + file = ERR_PTR(-EINVAL); + if (unlikely(au_is_nfs(parent->d_sb))) + goto out; + + // do not superio, nor NFS. + name = &base->d_name; + dentry = lkup_one(name->name, parent, name->len, &lkup); + //if (LktrCond) {dput(dentry); dentry = ERR_PTR(-1);} + if (IS_ERR(dentry)) { + file = (void*)dentry; + Err("%.*s lookup err %ld\n", LNPair(name), PTR_ERR(dentry)); + goto out; + } + err = vfsub_create(dir, dentry, S_IRUGO | S_IWUGO, NULL, /*dlgt*/0); + //if (LktrCond) {vfs_unlink(dir, dentry); err = -1;} + if (unlikely(err)) { + file = ERR_PTR(err); + Err("%.*s create err %d\n", LNPair(name), err); + goto out_dput; + } + file = dentry_open(dget(dentry), mntget(base_file->f_vfsmnt), + O_RDWR | O_CREAT | O_EXCL | O_LARGEFILE); + //if (LktrCond) {fput(file); file = ERR_PTR(-1);} + if (IS_ERR(file)) { + Err("%.*s open err %ld\n", LNPair(name), PTR_ERR(file)); + goto out_dput; + } + err = vfsub_unlink(dir, dentry, /*dlgt*/0); + //if (LktrCond) err = -1; + if (!err) + goto out_dput; /* success */ + + Err("%.*s unlink err %d\n", LNPair(name), err); + fput(file); + file = ERR_PTR(err); + + out_dput: + dput(dentry); + out: + dput(parent); + TraceErrPtr(file); + return file; +} + +/* + * initialize the xinofile for the specified branch{@sb, @bindex} + * at the place/path where @base_file indicates. + * test whether another branch is on the same filesystem or not, + * if @do_test is true. + */ +int xino_init(struct super_block *sb, aufs_bindex_t bindex, + struct file *base_file, int do_test) +{ + int err; + struct aufs_branch *br; + aufs_bindex_t bshared, bend; + struct file *file; + struct inode *inode, *hidden_inode; + struct xino xino; + + LKTRTrace("b%d, base_file %p, do_test %d\n", + bindex, base_file, do_test); + SiMustWriteLock(sb); + DEBUG_ON(!au_flag_test(sb, AuFlag_XINO)); + br = stobr(sb, bindex); + DEBUG_ON(br->br_xino); + + file = NULL; + bshared = -1; + bend = sbend(sb); + if (do_test) + bshared = is_sb_shared(sb, bindex, bend); + if (unlikely(bshared >= 0)) { + struct aufs_branch *shared_br = stobr(sb, bshared); + if (shared_br->br_xino) { + file = shared_br->br_xino; + get_file(file); + } + } + + if (!file) { + struct dentry *parent = dget_parent(base_file->f_dentry); + struct inode *dir = parent->d_inode; + + hi_lock_parent(dir); + file = xino_create2(base_file); + //if (LktrCond) {fput(file); file = ERR_PTR(-1);} + i_unlock(dir); + dput(parent); + err = PTR_ERR(file); + if (IS_ERR(file)) + goto out; + } + br->br_xino_read = find_readf(file); + err = PTR_ERR(br->br_xino_read); + if (IS_ERR(br->br_xino_read)) + goto out_put; + br->br_xino_write = find_writef(file); + err = PTR_ERR(br->br_xino_write); + if (IS_ERR(br->br_xino_write)) + goto out_put; + br->br_xino = file; + + inode = sb->s_root->d_inode; + hidden_inode = au_h_iptr_i(inode, bindex); + xino.ino = inode->i_ino; + //xino.h_gen = hidden_inode->i_generation; + //WARN_ON(xino.h_gen == AuXino_INVALID_HGEN); + err = xino_write(sb, bindex, hidden_inode->i_ino, &xino); + //if (LktrCond) err = -1; + if (!err) + return 0; /* success */ + + br->br_xino = NULL; + + out_put: + fput(file); + out: + TraceErr(err); + return err; +} + +/* + * set xino mount option. + */ +int xino_set(struct super_block *sb, struct opt_xino *xino, int remount) +{ + int err, sparse; + aufs_bindex_t bindex, bend; + struct aufs_branch *br; + struct dentry *parent; + struct qstr *name; + struct file *cur_xino; + struct inode *dir; + + LKTRTrace("%s\n", xino->path); + + err = 0; + name = &xino->file->f_dentry->d_name; + parent = dget_parent(xino->file->f_dentry); + dir = parent->d_inode; + cur_xino = stobr(sb, 0)->br_xino; + if (remount + && cur_xino + && cur_xino->f_dentry->d_parent == parent + && name->len == cur_xino->f_dentry->d_name.len + && !memcmp(name->name, cur_xino->f_dentry->d_name.name, name->len)) + goto out; + + au_flag_set(sb, AuFlag_XINO); + bend = sbend(sb); + for (bindex = bend; bindex >= 0; bindex--) { + br = stobr(sb, bindex); + if (unlikely(br->br_xino && file_count(br->br_xino) > 1)) { + fput(br->br_xino); + br->br_xino = NULL; + } + } + + for (bindex = 0; bindex <= bend; bindex++) { + struct file *file; + struct inode *inode; + + br = stobr(sb, bindex); + if (unlikely(!br->br_xino)) + continue; + + DEBUG_ON(file_count(br->br_xino) != 1); + hi_lock_parent(dir); + file = xino_create2(xino->file); + //if (LktrCond) {fput(file); file = ERR_PTR(-1);} + err = PTR_ERR(file); + if (IS_ERR(file)) { + i_unlock(dir); + break; + } + inode = br->br_xino->f_dentry->d_inode; + err = au_copy_file(file, br->br_xino, i_size_read(inode), sb, + &sparse); + //if (LktrCond) err = -1; + i_unlock(dir); + if (unlikely(err)) { + fput(file); + break; + } + fput(br->br_xino); + br->br_xino = file; + br->br_xino_read = find_readf(file); + DEBUG_ON(IS_ERR(br->br_xino_read)); + br->br_xino_write = find_writef(file); + DEBUG_ON(IS_ERR(br->br_xino_write)); + } + + for (bindex = 0; bindex <= bend; bindex++) + if (unlikely(!stobr(sb, bindex)->br_xino)) { + err = xino_init(sb, bindex, xino->file, /*do_test*/1); + //if (LktrCond) {fput(stobr(sb, bindex)->br_xino); + //stobr(sb, bindex)->br_xino = NULL; err = -1;} + if (!err) + continue; + IOErr("creating xino for branch %d(%d), " + "forcing noxino\n", bindex, err); + err = -EIO; + break; + } + out: + dput(parent); + if (!err) + au_flag_set(sb, AuFlag_XINO); + else + au_flag_clr(sb, AuFlag_XINO); + TraceErr(err); + return err; +} + +/* + * clear xino mount option + */ +int xino_clr(struct super_block *sb) +{ + aufs_bindex_t bindex, bend; + + TraceEnter(); + SiMustWriteLock(sb); + + bend = sbend(sb); + for (bindex = 0; bindex <= bend; bindex++) { + struct aufs_branch *br; + br = stobr(sb, bindex); + if (br->br_xino) { + fput(br->br_xino); + br->br_xino = NULL; + } + } + + //todo: need to make iunique() to return the larger inode number + + au_flag_clr(sb, AuFlag_XINO); + return 0; +} + +/* + * create a xinofile at the default place/path. + */ +struct file *xino_def(struct super_block *sb) +{ + struct file *file; + aufs_bindex_t bend, bindex, bwr; + char *page, *p; + + bend = sbend(sb); + bwr = -1; + for (bindex = 0; bindex <= bend; bindex++) + if (br_writable(sbr_perm(sb, bindex)) + && !au_is_nfs(au_h_dptr_i(sb->s_root, bindex)->d_sb)) { + bwr = bindex; + break; + } + + if (bwr != -1) { + // todo: rewrite with lkup_one() + file = ERR_PTR(-ENOMEM); + page = __getname(); + //if (LktrCond) {__putname(page); page = NULL;} + if (unlikely(!page)) + goto out; + p = d_path(au_h_dptr_i(sb->s_root, bwr), sbr_mnt(sb, bwr), page, + PATH_MAX - sizeof(AUFS_XINO_FNAME)); + //if (LktrCond) p = ERR_PTR(-1); + file = (void*)p; + if (p && !IS_ERR(p)) { + strcat(p, "/" AUFS_XINO_FNAME); + LKTRTrace("%s\n", p); + file = xino_create(sb, p, /*silent*/0, sb->s_root); + //if (LktrCond) {fput(file); file = ERR_PTR(-1);} + } + __putname(page); + } else { + file = xino_create(sb, AUFS_XINO_DEFPATH, /*silent*/0, + /*parent*/NULL); + //if (LktrCond) {fput(file); file = ERR_PTR(-1);} + } + + out: + TraceErrPtr(file); + return file; +} diff --git a/fs/squashfs/Makefile b/fs/squashfs/Makefile new file mode 100644 index 0000000..1bc7b06 --- /dev/null +++ b/fs/squashfs/Makefile @@ -0,0 +1,7 @@ +# +# Makefile for the linux squashfs routines. +# + +obj-$(CONFIG_SQUASHFS) += squashfs.o +squashfs-y += inode.o +squashfs-y += squashfs2_0.o diff --git a/fs/squashfs/inode.c b/fs/squashfs/inode.c new file mode 100644 index 0000000..895b699 --- /dev/null +++ b/fs/squashfs/inode.c @@ -0,0 +1,2329 @@ +/* + * Squashfs - a compressed read only filesystem for Linux + * + * Copyright (c) 2002, 2003, 2004, 2005, 2006, 2007 + * Phillip Lougher + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2, + * or (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + * + * inode.c + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "squashfs.h" + +static void vfs_read_inode(struct inode *i); +static struct dentry *squashfs_get_parent(struct dentry *child); +static int squashfs_read_inode(struct inode *i, squashfs_inode_t inode); +static int squashfs_statfs(struct dentry *, struct kstatfs *); +static int squashfs_symlink_readpage(struct file *file, struct page *page); +static long long read_blocklist(struct inode *inode, int index, + int readahead_blks, char *block_list, + unsigned short **block_p, unsigned int *bsize); +static int squashfs_readpage(struct file *file, struct page *page); +static int squashfs_readpage4K(struct file *file, struct page *page); +static int squashfs_readdir(struct file *, void *, filldir_t); +static struct dentry *squashfs_lookup(struct inode *, struct dentry *, + struct nameidata *); +static int squashfs_remount(struct super_block *s, int *flags, char *data); +static void squashfs_put_super(struct super_block *); +static int squashfs_get_sb(struct file_system_type *,int, const char *, void *, + struct vfsmount *); +static struct inode *squashfs_alloc_inode(struct super_block *sb); +static void squashfs_destroy_inode(struct inode *inode); +static int init_inodecache(void); +static void destroy_inodecache(void); + +static struct file_system_type squashfs_fs_type = { + .owner = THIS_MODULE, + .name = "squashfs", + .get_sb = squashfs_get_sb, + .kill_sb = kill_block_super, + .fs_flags = FS_REQUIRES_DEV +}; + +static const unsigned char squashfs_filetype_table[] = { + DT_UNKNOWN, DT_DIR, DT_REG, DT_LNK, DT_BLK, DT_CHR, DT_FIFO, DT_SOCK +}; + +static struct super_operations squashfs_super_ops = { + .alloc_inode = squashfs_alloc_inode, + .destroy_inode = squashfs_destroy_inode, + .statfs = squashfs_statfs, + .put_super = squashfs_put_super, + .remount_fs = squashfs_remount +}; + +static struct super_operations squashfs_export_super_ops = { + .alloc_inode = squashfs_alloc_inode, + .destroy_inode = squashfs_destroy_inode, + .statfs = squashfs_statfs, + .put_super = squashfs_put_super, + .read_inode = vfs_read_inode +}; + +static struct export_operations squashfs_export_ops = { + .get_parent = squashfs_get_parent +}; + +SQSH_EXTERN const struct address_space_operations squashfs_symlink_aops = { + .readpage = squashfs_symlink_readpage +}; + +SQSH_EXTERN const struct address_space_operations squashfs_aops = { + .readpage = squashfs_readpage +}; + +SQSH_EXTERN const struct address_space_operations squashfs_aops_4K = { + .readpage = squashfs_readpage4K +}; + +static const struct file_operations squashfs_dir_ops = { + .read = generic_read_dir, + .readdir = squashfs_readdir +}; + +SQSH_EXTERN struct inode_operations squashfs_dir_inode_ops = { + .lookup = squashfs_lookup +}; + + +static struct buffer_head *get_block_length(struct super_block *s, + int *cur_index, int *offset, int *c_byte) +{ + struct squashfs_sb_info *msblk = s->s_fs_info; + unsigned short temp; + struct buffer_head *bh; + + if (!(bh = sb_bread(s, *cur_index))) + goto out; + + if (msblk->devblksize - *offset == 1) { + if (msblk->swap) + ((unsigned char *) &temp)[1] = *((unsigned char *) + (bh->b_data + *offset)); + else + ((unsigned char *) &temp)[0] = *((unsigned char *) + (bh->b_data + *offset)); + brelse(bh); + if (!(bh = sb_bread(s, ++(*cur_index)))) + goto out; + if (msblk->swap) + ((unsigned char *) &temp)[0] = *((unsigned char *) + bh->b_data); + else + ((unsigned char *) &temp)[1] = *((unsigned char *) + bh->b_data); + *c_byte = temp; + *offset = 1; + } else { + if (msblk->swap) { + ((unsigned char *) &temp)[1] = *((unsigned char *) + (bh->b_data + *offset)); + ((unsigned char *) &temp)[0] = *((unsigned char *) + (bh->b_data + *offset + 1)); + } else { + ((unsigned char *) &temp)[0] = *((unsigned char *) + (bh->b_data + *offset)); + ((unsigned char *) &temp)[1] = *((unsigned char *) + (bh->b_data + *offset + 1)); + } + *c_byte = temp; + *offset += 2; + } + + if (SQUASHFS_CHECK_DATA(msblk->sblk.flags)) { + if (*offset == msblk->devblksize) { + brelse(bh); + if (!(bh = sb_bread(s, ++(*cur_index)))) + goto out; + *offset = 0; + } + if (*((unsigned char *) (bh->b_data + *offset)) != + SQUASHFS_MARKER_BYTE) { + ERROR("Metadata block marker corrupt @ %x\n", + *cur_index); + brelse(bh); + goto out; + } + (*offset)++; + } + return bh; + +out: + return NULL; +} + + +SQSH_EXTERN unsigned int squashfs_read_data(struct super_block *s, char *buffer, + long long index, unsigned int length, + long long *next_index, int srclength) +{ + struct squashfs_sb_info *msblk = s->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + struct buffer_head *bh[((SQUASHFS_FILE_MAX_SIZE - 1) >> + msblk->devblksize_log2) + 2]; + unsigned int offset = index & ((1 << msblk->devblksize_log2) - 1); + unsigned int cur_index = index >> msblk->devblksize_log2; + int bytes, avail_bytes, b = 0, k = 0; + unsigned int compressed; + unsigned int c_byte = length; + + if (c_byte) { + bytes = msblk->devblksize - offset; + compressed = SQUASHFS_COMPRESSED_BLOCK(c_byte); + c_byte = SQUASHFS_COMPRESSED_SIZE_BLOCK(c_byte); + + TRACE("Block @ 0x%llx, %scompressed size %d, src size %d\n", index, compressed + ? "" : "un", (unsigned int) c_byte, srclength); + + if (c_byte > srclength || index < 0 || (index + c_byte) > sblk->bytes_used) + goto read_failure; + + if (!(bh[0] = sb_getblk(s, cur_index))) + goto block_release; + + for (b = 1; bytes < c_byte; b++) { + if (!(bh[b] = sb_getblk(s, ++cur_index))) + goto block_release; + bytes += msblk->devblksize; + } + ll_rw_block(READ, b, bh); + } else { + if (index < 0 || (index + 2) > sblk->bytes_used) + goto read_failure; + + if (!(bh[0] = get_block_length(s, &cur_index, &offset, + &c_byte))) + goto read_failure; + + bytes = msblk->devblksize - offset; + compressed = SQUASHFS_COMPRESSED(c_byte); + c_byte = SQUASHFS_COMPRESSED_SIZE(c_byte); + + TRACE("Block @ 0x%llx, %scompressed size %d\n", index, compressed + ? "" : "un", (unsigned int) c_byte); + + if (c_byte > srclength || (index + c_byte) > sblk->bytes_used) + goto read_failure; + + for (b = 1; bytes < c_byte; b++) { + if (!(bh[b] = sb_getblk(s, ++cur_index))) + goto block_release; + bytes += msblk->devblksize; + } + ll_rw_block(READ, b - 1, bh + 1); + } + + if (compressed) { + int zlib_err = 0; + + /* + * uncompress block + */ + + mutex_lock(&msblk->read_data_mutex); + + msblk->stream.next_out = buffer; + msblk->stream.avail_out = srclength; + + for (bytes = 0; k < b; k++) { + avail_bytes = (c_byte - bytes) > (msblk->devblksize - offset) ? + msblk->devblksize - offset : + c_byte - bytes; + wait_on_buffer(bh[k]); + if (!buffer_uptodate(bh[k])) + goto release_mutex; + + msblk->stream.next_in = bh[k]->b_data + offset; + msblk->stream.avail_in = avail_bytes; + + if (k == 0) { + zlib_err = zlib_inflateInit(&msblk->stream); + if (zlib_err != Z_OK) { + ERROR("zlib_inflateInit returned unexpected result 0x%x, srclength %d\n", + zlib_err, srclength); + goto release_mutex; + } + + if (avail_bytes == 0) { + offset = 0; + brelse(bh[k]); + continue; + } + } + + zlib_err = zlib_inflate(&msblk->stream, Z_NO_FLUSH); + if (zlib_err != Z_OK && zlib_err != Z_STREAM_END) { + ERROR("zlib_inflate returned unexpected result 0x%x, srclength %d, avail_in %d, avail_out %d\n", + zlib_err, srclength, msblk->stream.avail_in, msblk->stream.avail_out); + goto release_mutex; + } + + bytes += avail_bytes; + offset = 0; + brelse(bh[k]); + } + + if (zlib_err != Z_STREAM_END) + goto release_mutex; + + zlib_err = zlib_inflateEnd(&msblk->stream); + if (zlib_err != Z_OK) { + ERROR("zlib_inflateEnd returned unexpected result 0x%x, srclength %d\n", + zlib_err, srclength); + goto release_mutex; + } + bytes = msblk->stream.total_out; + mutex_unlock(&msblk->read_data_mutex); + } else { + int i; + + for(i = 0; i < b; i++) { + wait_on_buffer(bh[i]); + if(!buffer_uptodate(bh[i])) + goto block_release; + } + + for (bytes = 0; k < b; k++) { + avail_bytes = (c_byte - bytes) > (msblk->devblksize - offset) ? + msblk->devblksize - offset : + c_byte - bytes; + memcpy(buffer + bytes, bh[k]->b_data + offset, avail_bytes); + bytes += avail_bytes; + offset = 0; + brelse(bh[k]); + } + } + + if (next_index) + *next_index = index + c_byte + (length ? 0 : + (SQUASHFS_CHECK_DATA(msblk->sblk.flags) + ? 3 : 2)); + return bytes; + +release_mutex: + mutex_unlock(&msblk->read_data_mutex); + +block_release: + for (; k < b; k++) + brelse(bh[k]); + +read_failure: + ERROR("sb_bread failed reading block 0x%x\n", cur_index); + return 0; +} + + +SQSH_EXTERN int squashfs_get_cached_block(struct super_block *s, char *buffer, + long long block, unsigned int offset, + int length, long long *next_block, + unsigned int *next_offset) +{ + struct squashfs_sb_info *msblk = s->s_fs_info; + int n, i, bytes, return_length = length; + long long next_index; + + TRACE("Entered squashfs_get_cached_block [%llx:%x]\n", block, offset); + + while ( 1 ) { + for (i = 0; i < SQUASHFS_CACHED_BLKS; i++) + if (msblk->block_cache[i].block == block) + break; + + mutex_lock(&msblk->block_cache_mutex); + + if (i == SQUASHFS_CACHED_BLKS) { + /* read inode header block */ + for (i = msblk->next_cache, n = SQUASHFS_CACHED_BLKS; + n ; n --, i = (i + 1) % + SQUASHFS_CACHED_BLKS) + if (msblk->block_cache[i].block != + SQUASHFS_USED_BLK) + break; + + if (n == 0) { + wait_queue_t wait; + + init_waitqueue_entry(&wait, current); + add_wait_queue(&msblk->waitq, &wait); + set_current_state(TASK_UNINTERRUPTIBLE); + mutex_unlock(&msblk->block_cache_mutex); + schedule(); + set_current_state(TASK_RUNNING); + remove_wait_queue(&msblk->waitq, &wait); + continue; + } + msblk->next_cache = (i + 1) % SQUASHFS_CACHED_BLKS; + + if (msblk->block_cache[i].block == + SQUASHFS_INVALID_BLK) { + if (!(msblk->block_cache[i].data = + kmalloc(SQUASHFS_METADATA_SIZE, + GFP_KERNEL))) { + ERROR("Failed to allocate cache" + "block\n"); + mutex_unlock(&msblk->block_cache_mutex); + goto out; + } + } + + msblk->block_cache[i].block = SQUASHFS_USED_BLK; + mutex_unlock(&msblk->block_cache_mutex); + + msblk->block_cache[i].length = squashfs_read_data(s, + msblk->block_cache[i].data, block, 0, &next_index, SQUASHFS_METADATA_SIZE); + if (msblk->block_cache[i].length == 0) { + ERROR("Unable to read cache block [%llx:%x]\n", + block, offset); + mutex_lock(&msblk->block_cache_mutex); + msblk->block_cache[i].block = SQUASHFS_INVALID_BLK; + kfree(msblk->block_cache[i].data); + wake_up(&msblk->waitq); + mutex_unlock(&msblk->block_cache_mutex); + goto out; + } + + mutex_lock(&msblk->block_cache_mutex); + wake_up(&msblk->waitq); + msblk->block_cache[i].block = block; + msblk->block_cache[i].next_index = next_index; + TRACE("Read cache block [%llx:%x]\n", block, offset); + } + + if (msblk->block_cache[i].block != block) { + mutex_unlock(&msblk->block_cache_mutex); + continue; + } + + bytes = msblk->block_cache[i].length - offset; + + if (bytes < 1) { + mutex_unlock(&msblk->block_cache_mutex); + goto out; + } else if (bytes >= length) { + if (buffer) + memcpy(buffer, msblk->block_cache[i].data + + offset, length); + if (msblk->block_cache[i].length - offset == length) { + *next_block = msblk->block_cache[i].next_index; + *next_offset = 0; + } else { + *next_block = block; + *next_offset = offset + length; + } + mutex_unlock(&msblk->block_cache_mutex); + goto finish; + } else { + if (buffer) { + memcpy(buffer, msblk->block_cache[i].data + + offset, bytes); + buffer += bytes; + } + block = msblk->block_cache[i].next_index; + mutex_unlock(&msblk->block_cache_mutex); + length -= bytes; + offset = 0; + } + } + +finish: + return return_length; +out: + return 0; +} + + +static int get_fragment_location(struct super_block *s, unsigned int fragment, + long long *fragment_start_block, + unsigned int *fragment_size) +{ + struct squashfs_sb_info *msblk = s->s_fs_info; + long long start_block = + msblk->fragment_index[SQUASHFS_FRAGMENT_INDEX(fragment)]; + int offset = SQUASHFS_FRAGMENT_INDEX_OFFSET(fragment); + struct squashfs_fragment_entry fragment_entry; + + if (msblk->swap) { + struct squashfs_fragment_entry sfragment_entry; + + if (!squashfs_get_cached_block(s, (char *) &sfragment_entry, + start_block, offset, + sizeof(sfragment_entry), &start_block, + &offset)) + goto out; + SQUASHFS_SWAP_FRAGMENT_ENTRY(&fragment_entry, &sfragment_entry); + } else + if (!squashfs_get_cached_block(s, (char *) &fragment_entry, + start_block, offset, + sizeof(fragment_entry), &start_block, + &offset)) + goto out; + + *fragment_start_block = fragment_entry.start_block; + *fragment_size = fragment_entry.size; + + return 1; + +out: + return 0; +} + + +SQSH_EXTERN void release_cached_fragment(struct squashfs_sb_info *msblk, struct + squashfs_fragment_cache *fragment) +{ + mutex_lock(&msblk->fragment_mutex); + fragment->locked --; + wake_up(&msblk->fragment_wait_queue); + mutex_unlock(&msblk->fragment_mutex); +} + + +SQSH_EXTERN struct squashfs_fragment_cache *get_cached_fragment(struct super_block + *s, long long start_block, + int length) +{ + int i, n; + struct squashfs_sb_info *msblk = s->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + + while ( 1 ) { + mutex_lock(&msblk->fragment_mutex); + + for (i = 0; i < SQUASHFS_CACHED_FRAGMENTS && + msblk->fragment[i].block != start_block; i++); + + if (i == SQUASHFS_CACHED_FRAGMENTS) { + for (i = msblk->next_fragment, n = + SQUASHFS_CACHED_FRAGMENTS; n && + msblk->fragment[i].locked; n--, i = (i + 1) % + SQUASHFS_CACHED_FRAGMENTS); + + if (n == 0) { + wait_queue_t wait; + + init_waitqueue_entry(&wait, current); + add_wait_queue(&msblk->fragment_wait_queue, + &wait); + set_current_state(TASK_UNINTERRUPTIBLE); + mutex_unlock(&msblk->fragment_mutex); + schedule(); + set_current_state(TASK_RUNNING); + remove_wait_queue(&msblk->fragment_wait_queue, + &wait); + continue; + } + msblk->next_fragment = (msblk->next_fragment + 1) % + SQUASHFS_CACHED_FRAGMENTS; + + if (msblk->fragment[i].data == NULL) + if (!(msblk->fragment[i].data = SQUASHFS_ALLOC + (SQUASHFS_FILE_MAX_SIZE))) { + ERROR("Failed to allocate fragment " + "cache block\n"); + mutex_unlock(&msblk->fragment_mutex); + goto out; + } + + msblk->fragment[i].block = SQUASHFS_INVALID_BLK; + msblk->fragment[i].locked = 1; + mutex_unlock(&msblk->fragment_mutex); + + if (!(msblk->fragment[i].length = squashfs_read_data(s, + msblk->fragment[i].data, + start_block, length, NULL, sblk->block_size))) { + ERROR("Unable to read fragment cache block " + "[%llx]\n", start_block); + msblk->fragment[i].locked = 0; + smp_mb(); + goto out; + } + + mutex_lock(&msblk->fragment_mutex); + msblk->fragment[i].block = start_block; + TRACE("New fragment %d, start block %lld, locked %d\n", + i, msblk->fragment[i].block, + msblk->fragment[i].locked); + mutex_unlock(&msblk->fragment_mutex); + break; + } + + msblk->fragment[i].locked++; + mutex_unlock(&msblk->fragment_mutex); + TRACE("Got fragment %d, start block %lld, locked %d\n", i, + msblk->fragment[i].block, + msblk->fragment[i].locked); + break; + } + + return &msblk->fragment[i]; + +out: + return NULL; +} + + +static void squashfs_new_inode(struct squashfs_sb_info *msblk, struct inode *i, + struct squashfs_base_inode_header *inodeb) +{ + i->i_ino = inodeb->inode_number; + i->i_mtime.tv_sec = inodeb->mtime; + i->i_atime.tv_sec = inodeb->mtime; + i->i_ctime.tv_sec = inodeb->mtime; + i->i_uid = msblk->uid[inodeb->uid]; + i->i_mode = inodeb->mode; + i->i_size = 0; + if (inodeb->guid == SQUASHFS_GUIDS) + i->i_gid = i->i_uid; + else + i->i_gid = msblk->guid[inodeb->guid]; +} + + +static squashfs_inode_t squashfs_inode_lookup(struct super_block *s, int ino) +{ + struct squashfs_sb_info *msblk = s->s_fs_info; + long long start = msblk->inode_lookup_table[SQUASHFS_LOOKUP_BLOCK(ino - 1)]; + int offset = SQUASHFS_LOOKUP_BLOCK_OFFSET(ino - 1); + squashfs_inode_t inode; + + TRACE("Entered squashfs_inode_lookup, inode_number = %d\n", ino); + + if (msblk->swap) { + squashfs_inode_t sinode; + + if (!squashfs_get_cached_block(s, (char *) &sinode, start, offset, + sizeof(sinode), &start, &offset)) + goto out; + SQUASHFS_SWAP_INODE_T((&inode), &sinode); + } else if (!squashfs_get_cached_block(s, (char *) &inode, start, offset, + sizeof(inode), &start, &offset)) + goto out; + + TRACE("squashfs_inode_lookup, inode = 0x%llx\n", inode); + + return inode; + +out: + return SQUASHFS_INVALID_BLK; +} + + +static void vfs_read_inode(struct inode *i) +{ + struct squashfs_sb_info *msblk = i->i_sb->s_fs_info; + squashfs_inode_t inode = squashfs_inode_lookup(i->i_sb, i->i_ino); + + TRACE("Entered vfs_read_inode\n"); + + if(inode != SQUASHFS_INVALID_BLK) + (msblk->read_inode)(i, inode); +} + + +static struct dentry *squashfs_get_parent(struct dentry *child) +{ + struct inode *i = child->d_inode; + struct inode *parent = iget(i->i_sb, SQUASHFS_I(i)->u.s2.parent_inode); + struct dentry *rv; + + TRACE("Entered squashfs_get_parent\n"); + + if(parent == NULL) { + rv = ERR_PTR(-EACCES); + goto out; + } + + rv = d_alloc_anon(parent); + if(rv == NULL) + rv = ERR_PTR(-ENOMEM); + +out: + return rv; +} + + +SQSH_EXTERN struct inode *squashfs_iget(struct super_block *s, squashfs_inode_t inode, unsigned int inode_number) +{ + struct squashfs_sb_info *msblk = s->s_fs_info; + struct inode *i = iget_locked(s, inode_number); + + TRACE("Entered squashfs_iget\n"); + + if(i && (i->i_state & I_NEW)) { + (msblk->read_inode)(i, inode); + unlock_new_inode(i); + } + + return i; +} + + +static int squashfs_read_inode(struct inode *i, squashfs_inode_t inode) +{ + struct super_block *s = i->i_sb; + struct squashfs_sb_info *msblk = s->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + long long block = SQUASHFS_INODE_BLK(inode) + + sblk->inode_table_start; + unsigned int offset = SQUASHFS_INODE_OFFSET(inode); + long long next_block; + unsigned int next_offset; + union squashfs_inode_header id, sid; + struct squashfs_base_inode_header *inodeb = &id.base, + *sinodeb = &sid.base; + + TRACE("Entered squashfs_read_inode\n"); + + if (msblk->swap) { + if (!squashfs_get_cached_block(s, (char *) sinodeb, block, + offset, sizeof(*sinodeb), &next_block, + &next_offset)) + goto failed_read; + SQUASHFS_SWAP_BASE_INODE_HEADER(inodeb, sinodeb, + sizeof(*sinodeb)); + } else + if (!squashfs_get_cached_block(s, (char *) inodeb, block, + offset, sizeof(*inodeb), &next_block, + &next_offset)) + goto failed_read; + + squashfs_new_inode(msblk, i, inodeb); + + switch(inodeb->inode_type) { + case SQUASHFS_FILE_TYPE: { + unsigned int frag_size; + long long frag_blk; + struct squashfs_reg_inode_header *inodep = &id.reg; + struct squashfs_reg_inode_header *sinodep = &sid.reg; + + if (msblk->swap) { + if (!squashfs_get_cached_block(s, (char *) + sinodep, block, offset, + sizeof(*sinodep), &next_block, + &next_offset)) + goto failed_read; + SQUASHFS_SWAP_REG_INODE_HEADER(inodep, sinodep); + } else + if (!squashfs_get_cached_block(s, (char *) + inodep, block, offset, + sizeof(*inodep), &next_block, + &next_offset)) + goto failed_read; + + frag_blk = SQUASHFS_INVALID_BLK; + if (inodep->fragment != SQUASHFS_INVALID_FRAG && + !get_fragment_location(s, + inodep->fragment, &frag_blk, &frag_size)) + goto failed_read; + + i->i_nlink = 1; + i->i_size = inodep->file_size; + i->i_fop = &generic_ro_fops; + i->i_mode |= S_IFREG; + i->i_blocks = ((i->i_size - 1) >> 9) + 1; + SQUASHFS_I(i)->u.s1.fragment_start_block = frag_blk; + SQUASHFS_I(i)->u.s1.fragment_size = frag_size; + SQUASHFS_I(i)->u.s1.fragment_offset = inodep->offset; + SQUASHFS_I(i)->start_block = inodep->start_block; + SQUASHFS_I(i)->u.s1.block_list_start = next_block; + SQUASHFS_I(i)->offset = next_offset; + if (sblk->block_size > 4096) + i->i_data.a_ops = &squashfs_aops; + else + i->i_data.a_ops = &squashfs_aops_4K; + + TRACE("File inode %x:%x, start_block %llx, " + "block_list_start %llx, offset %x\n", + SQUASHFS_INODE_BLK(inode), offset, + inodep->start_block, next_block, + next_offset); + break; + } + case SQUASHFS_LREG_TYPE: { + unsigned int frag_size; + long long frag_blk; + struct squashfs_lreg_inode_header *inodep = &id.lreg; + struct squashfs_lreg_inode_header *sinodep = &sid.lreg; + + if (msblk->swap) { + if (!squashfs_get_cached_block(s, (char *) + sinodep, block, offset, + sizeof(*sinodep), &next_block, + &next_offset)) + goto failed_read; + SQUASHFS_SWAP_LREG_INODE_HEADER(inodep, sinodep); + } else + if (!squashfs_get_cached_block(s, (char *) + inodep, block, offset, + sizeof(*inodep), &next_block, + &next_offset)) + goto failed_read; + + frag_blk = SQUASHFS_INVALID_BLK; + if (inodep->fragment != SQUASHFS_INVALID_FRAG && + !get_fragment_location(s, + inodep->fragment, &frag_blk, &frag_size)) + goto failed_read; + + i->i_nlink = inodep->nlink; + i->i_size = inodep->file_size; + i->i_fop = &generic_ro_fops; + i->i_mode |= S_IFREG; + i->i_blocks = ((i->i_size - 1) >> 9) + 1; + SQUASHFS_I(i)->u.s1.fragment_start_block = frag_blk; + SQUASHFS_I(i)->u.s1.fragment_size = frag_size; + SQUASHFS_I(i)->u.s1.fragment_offset = inodep->offset; + SQUASHFS_I(i)->start_block = inodep->start_block; + SQUASHFS_I(i)->u.s1.block_list_start = next_block; + SQUASHFS_I(i)->offset = next_offset; + if (sblk->block_size > 4096) + i->i_data.a_ops = &squashfs_aops; + else + i->i_data.a_ops = &squashfs_aops_4K; + + TRACE("File inode %x:%x, start_block %llx, " + "block_list_start %llx, offset %x\n", + SQUASHFS_INODE_BLK(inode), offset, + inodep->start_block, next_block, + next_offset); + break; + } + case SQUASHFS_DIR_TYPE: { + struct squashfs_dir_inode_header *inodep = &id.dir; + struct squashfs_dir_inode_header *sinodep = &sid.dir; + + if (msblk->swap) { + if (!squashfs_get_cached_block(s, (char *) + sinodep, block, offset, + sizeof(*sinodep), &next_block, + &next_offset)) + goto failed_read; + SQUASHFS_SWAP_DIR_INODE_HEADER(inodep, sinodep); + } else + if (!squashfs_get_cached_block(s, (char *) + inodep, block, offset, + sizeof(*inodep), &next_block, + &next_offset)) + goto failed_read; + + i->i_nlink = inodep->nlink; + i->i_size = inodep->file_size; + i->i_op = &squashfs_dir_inode_ops; + i->i_fop = &squashfs_dir_ops; + i->i_mode |= S_IFDIR; + SQUASHFS_I(i)->start_block = inodep->start_block; + SQUASHFS_I(i)->offset = inodep->offset; + SQUASHFS_I(i)->u.s2.directory_index_count = 0; + SQUASHFS_I(i)->u.s2.parent_inode = inodep->parent_inode; + + TRACE("Directory inode %x:%x, start_block %x, offset " + "%x\n", SQUASHFS_INODE_BLK(inode), + offset, inodep->start_block, + inodep->offset); + break; + } + case SQUASHFS_LDIR_TYPE: { + struct squashfs_ldir_inode_header *inodep = &id.ldir; + struct squashfs_ldir_inode_header *sinodep = &sid.ldir; + + if (msblk->swap) { + if (!squashfs_get_cached_block(s, (char *) + sinodep, block, offset, + sizeof(*sinodep), &next_block, + &next_offset)) + goto failed_read; + SQUASHFS_SWAP_LDIR_INODE_HEADER(inodep, + sinodep); + } else + if (!squashfs_get_cached_block(s, (char *) + inodep, block, offset, + sizeof(*inodep), &next_block, + &next_offset)) + goto failed_read; + + i->i_nlink = inodep->nlink; + i->i_size = inodep->file_size; + i->i_op = &squashfs_dir_inode_ops; + i->i_fop = &squashfs_dir_ops; + i->i_mode |= S_IFDIR; + SQUASHFS_I(i)->start_block = inodep->start_block; + SQUASHFS_I(i)->offset = inodep->offset; + SQUASHFS_I(i)->u.s2.directory_index_start = next_block; + SQUASHFS_I(i)->u.s2.directory_index_offset = + next_offset; + SQUASHFS_I(i)->u.s2.directory_index_count = + inodep->i_count; + SQUASHFS_I(i)->u.s2.parent_inode = inodep->parent_inode; + + TRACE("Long directory inode %x:%x, start_block %x, " + "offset %x\n", + SQUASHFS_INODE_BLK(inode), offset, + inodep->start_block, inodep->offset); + break; + } + case SQUASHFS_SYMLINK_TYPE: { + struct squashfs_symlink_inode_header *inodep = + &id.symlink; + struct squashfs_symlink_inode_header *sinodep = + &sid.symlink; + + if (msblk->swap) { + if (!squashfs_get_cached_block(s, (char *) + sinodep, block, offset, + sizeof(*sinodep), &next_block, + &next_offset)) + goto failed_read; + SQUASHFS_SWAP_SYMLINK_INODE_HEADER(inodep, + sinodep); + } else + if (!squashfs_get_cached_block(s, (char *) + inodep, block, offset, + sizeof(*inodep), &next_block, + &next_offset)) + goto failed_read; + + i->i_nlink = inodep->nlink; + i->i_size = inodep->symlink_size; + i->i_op = &page_symlink_inode_operations; + i->i_data.a_ops = &squashfs_symlink_aops; + i->i_mode |= S_IFLNK; + SQUASHFS_I(i)->start_block = next_block; + SQUASHFS_I(i)->offset = next_offset; + + TRACE("Symbolic link inode %x:%x, start_block %llx, " + "offset %x\n", + SQUASHFS_INODE_BLK(inode), offset, + next_block, next_offset); + break; + } + case SQUASHFS_BLKDEV_TYPE: + case SQUASHFS_CHRDEV_TYPE: { + struct squashfs_dev_inode_header *inodep = &id.dev; + struct squashfs_dev_inode_header *sinodep = &sid.dev; + + if (msblk->swap) { + if (!squashfs_get_cached_block(s, (char *) + sinodep, block, offset, + sizeof(*sinodep), &next_block, + &next_offset)) + goto failed_read; + SQUASHFS_SWAP_DEV_INODE_HEADER(inodep, sinodep); + } else + if (!squashfs_get_cached_block(s, (char *) + inodep, block, offset, + sizeof(*inodep), &next_block, + &next_offset)) + goto failed_read; + + i->i_nlink = inodep->nlink; + i->i_mode |= (inodeb->inode_type == + SQUASHFS_CHRDEV_TYPE) ? S_IFCHR : + S_IFBLK; + init_special_inode(i, i->i_mode, + old_decode_dev(inodep->rdev)); + + TRACE("Device inode %x:%x, rdev %x\n", + SQUASHFS_INODE_BLK(inode), offset, + inodep->rdev); + break; + } + case SQUASHFS_FIFO_TYPE: + case SQUASHFS_SOCKET_TYPE: { + struct squashfs_ipc_inode_header *inodep = &id.ipc; + struct squashfs_ipc_inode_header *sinodep = &sid.ipc; + + if (msblk->swap) { + if (!squashfs_get_cached_block(s, (char *) + sinodep, block, offset, + sizeof(*sinodep), &next_block, + &next_offset)) + goto failed_read; + SQUASHFS_SWAP_IPC_INODE_HEADER(inodep, sinodep); + } else + if (!squashfs_get_cached_block(s, (char *) + inodep, block, offset, + sizeof(*inodep), &next_block, + &next_offset)) + goto failed_read; + + i->i_nlink = inodep->nlink; + i->i_mode |= (inodeb->inode_type == SQUASHFS_FIFO_TYPE) + ? S_IFIFO : S_IFSOCK; + init_special_inode(i, i->i_mode, 0); + break; + } + default: + ERROR("Unknown inode type %d in squashfs_iget!\n", + inodeb->inode_type); + goto failed_read1; + } + + return 1; + +failed_read: + ERROR("Unable to read inode [%llx:%x]\n", block, offset); + +failed_read1: + make_bad_inode(i); + return 0; +} + + +static int read_inode_lookup_table(struct super_block *s) +{ + struct squashfs_sb_info *msblk = s->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + unsigned int length = SQUASHFS_LOOKUP_BLOCK_BYTES(sblk->inodes); + + TRACE("In read_inode_lookup_table, length %d\n", length); + + /* Allocate inode lookup table */ + if (!(msblk->inode_lookup_table = kmalloc(length, GFP_KERNEL))) { + ERROR("Failed to allocate inode lookup table\n"); + return 0; + } + + if (!squashfs_read_data(s, (char *) msblk->inode_lookup_table, + sblk->lookup_table_start, length | + SQUASHFS_COMPRESSED_BIT_BLOCK, NULL, length)) { + ERROR("unable to read inode lookup table\n"); + return 0; + } + + if (msblk->swap) { + int i; + long long block; + + for (i = 0; i < SQUASHFS_LOOKUP_BLOCKS(sblk->inodes); i++) { + SQUASHFS_SWAP_LOOKUP_BLOCKS((&block), + &msblk->inode_lookup_table[i], 1); + msblk->inode_lookup_table[i] = block; + } + } + + return 1; +} + + +static int read_fragment_index_table(struct super_block *s) +{ + struct squashfs_sb_info *msblk = s->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + unsigned int length = SQUASHFS_FRAGMENT_INDEX_BYTES(sblk->fragments); + + if(length == 0) + return 1; + + /* Allocate fragment index table */ + if (!(msblk->fragment_index = kmalloc(length, GFP_KERNEL))) { + ERROR("Failed to allocate fragment index table\n"); + return 0; + } + + if (!squashfs_read_data(s, (char *) msblk->fragment_index, + sblk->fragment_table_start, length | + SQUASHFS_COMPRESSED_BIT_BLOCK, NULL, length)) { + ERROR("unable to read fragment index table\n"); + return 0; + } + + if (msblk->swap) { + int i; + long long fragment; + + for (i = 0; i < SQUASHFS_FRAGMENT_INDEXES(sblk->fragments); i++) { + SQUASHFS_SWAP_FRAGMENT_INDEXES((&fragment), + &msblk->fragment_index[i], 1); + msblk->fragment_index[i] = fragment; + } + } + + return 1; +} + + +static int supported_squashfs_filesystem(struct squashfs_sb_info *msblk, int silent) +{ + struct squashfs_super_block *sblk = &msblk->sblk; + + msblk->read_inode = squashfs_read_inode; + msblk->read_blocklist = read_blocklist; + msblk->read_fragment_index_table = read_fragment_index_table; + + if (sblk->s_major == 1) { + if (!squashfs_1_0_supported(msblk)) { + SERROR("Major/Minor mismatch, Squashfs 1.0 filesystems " + "are unsupported\n"); + SERROR("Please recompile with " + "Squashfs 1.0 support enabled\n"); + return 0; + } + } else if (sblk->s_major == 2) { + if (!squashfs_2_0_supported(msblk)) { + SERROR("Major/Minor mismatch, Squashfs 2.0 filesystems " + "are unsupported\n"); + SERROR("Please recompile with " + "Squashfs 2.0 support enabled\n"); + return 0; + } + } else if(sblk->s_major != SQUASHFS_MAJOR || sblk->s_minor > + SQUASHFS_MINOR) { + SERROR("Major/Minor mismatch, trying to mount newer %d.%d " + "filesystem\n", sblk->s_major, sblk->s_minor); + SERROR("Please update your kernel\n"); + return 0; + } + + return 1; +} + + +static int squashfs_fill_super(struct super_block *s, void *data, int silent) +{ + struct squashfs_sb_info *msblk; + struct squashfs_super_block *sblk; + int i; + char b[BDEVNAME_SIZE]; + struct inode *root; + + TRACE("Entered squashfs_read_superblock\n"); + + if (!(s->s_fs_info = kmalloc(sizeof(struct squashfs_sb_info), + GFP_KERNEL))) { + ERROR("Failed to allocate superblock\n"); + goto failure; + } + memset(s->s_fs_info, 0, sizeof(struct squashfs_sb_info)); + msblk = s->s_fs_info; + if (!(msblk->stream.workspace = vmalloc(zlib_inflate_workspacesize()))) { + ERROR("Failed to allocate zlib workspace\n"); + goto failure; + } + sblk = &msblk->sblk; + + msblk->devblksize = sb_min_blocksize(s, BLOCK_SIZE); + msblk->devblksize_log2 = ffz(~msblk->devblksize); + + mutex_init(&msblk->read_data_mutex); + mutex_init(&msblk->read_page_mutex); + mutex_init(&msblk->block_cache_mutex); + mutex_init(&msblk->fragment_mutex); + mutex_init(&msblk->meta_index_mutex); + + init_waitqueue_head(&msblk->waitq); + init_waitqueue_head(&msblk->fragment_wait_queue); + + sblk->bytes_used = sizeof(struct squashfs_super_block); + if (!squashfs_read_data(s, (char *) sblk, SQUASHFS_START, + sizeof(struct squashfs_super_block) | + SQUASHFS_COMPRESSED_BIT_BLOCK, NULL, sizeof(struct squashfs_super_block))) { + SERROR("unable to read superblock\n"); + goto failed_mount; + } + + /* Check it is a SQUASHFS superblock */ + msblk->swap = 0; + if ((s->s_magic = sblk->s_magic) != SQUASHFS_MAGIC) { + if (sblk->s_magic == SQUASHFS_MAGIC_SWAP) { + struct squashfs_super_block ssblk; + + WARNING("Mounting a different endian SQUASHFS " + "filesystem on %s\n", bdevname(s->s_bdev, b)); + + SQUASHFS_SWAP_SUPER_BLOCK(&ssblk, sblk); + memcpy(sblk, &ssblk, sizeof(struct squashfs_super_block)); + msblk->swap = 1; + } else { + SERROR("Can't find a SQUASHFS superblock on %s\n", + bdevname(s->s_bdev, b)); + goto failed_mount; + } + } + + /* Check the MAJOR & MINOR versions */ + if(!supported_squashfs_filesystem(msblk, silent)) + goto failed_mount; + + /* Check the filesystem does not extend beyond the end of the + block device */ + if(sblk->bytes_used < 0 || sblk->bytes_used > i_size_read(s->s_bdev->bd_inode)) + goto failed_mount; + + /* Check the root inode for sanity */ + if (SQUASHFS_INODE_OFFSET(sblk->root_inode) > SQUASHFS_METADATA_SIZE) + goto failed_mount; + + TRACE("Found valid superblock on %s\n", bdevname(s->s_bdev, b)); + TRACE("Inodes are %scompressed\n", + SQUASHFS_UNCOMPRESSED_INODES + (sblk->flags) ? "un" : ""); + TRACE("Data is %scompressed\n", + SQUASHFS_UNCOMPRESSED_DATA(sblk->flags) + ? "un" : ""); + TRACE("Check data is %s present in the filesystem\n", + SQUASHFS_CHECK_DATA(sblk->flags) ? + "" : "not"); + TRACE("Filesystem size %lld bytes\n", sblk->bytes_used); + TRACE("Block size %d\n", sblk->block_size); + TRACE("Number of inodes %d\n", sblk->inodes); + if (sblk->s_major > 1) + TRACE("Number of fragments %d\n", sblk->fragments); + TRACE("Number of uids %d\n", sblk->no_uids); + TRACE("Number of gids %d\n", sblk->no_guids); + TRACE("sblk->inode_table_start %llx\n", sblk->inode_table_start); + TRACE("sblk->directory_table_start %llx\n", sblk->directory_table_start); + if (sblk->s_major > 1) + TRACE("sblk->fragment_table_start %llx\n", + sblk->fragment_table_start); + TRACE("sblk->uid_start %llx\n", sblk->uid_start); + + s->s_flags |= MS_RDONLY; + s->s_op = &squashfs_super_ops; + + /* Init inode_table block pointer array */ + if (!(msblk->block_cache = kmalloc(sizeof(struct squashfs_cache) * + SQUASHFS_CACHED_BLKS, GFP_KERNEL))) { + ERROR("Failed to allocate block cache\n"); + goto failed_mount; + } + + for (i = 0; i < SQUASHFS_CACHED_BLKS; i++) + msblk->block_cache[i].block = SQUASHFS_INVALID_BLK; + + msblk->next_cache = 0; + + /* Allocate read_page block */ + if (!(msblk->read_page = kmalloc(sblk->block_size, GFP_KERNEL))) { + ERROR("Failed to allocate read_page block\n"); + goto failed_mount; + } + + /* Allocate uid and gid tables */ + if (!(msblk->uid = kmalloc((sblk->no_uids + sblk->no_guids) * + sizeof(unsigned int), GFP_KERNEL))) { + ERROR("Failed to allocate uid/gid table\n"); + goto failed_mount; + } + msblk->guid = msblk->uid + sblk->no_uids; + + if (msblk->swap) { + unsigned int suid[sblk->no_uids + sblk->no_guids]; + + if (!squashfs_read_data(s, (char *) &suid, sblk->uid_start, + ((sblk->no_uids + sblk->no_guids) * + sizeof(unsigned int)) | + SQUASHFS_COMPRESSED_BIT_BLOCK, NULL, (sblk->no_uids + sblk->no_guids) * sizeof(unsigned int))) { + ERROR("unable to read uid/gid table\n"); + goto failed_mount; + } + + SQUASHFS_SWAP_DATA(msblk->uid, suid, (sblk->no_uids + + sblk->no_guids), (sizeof(unsigned int) * 8)); + } else + if (!squashfs_read_data(s, (char *) msblk->uid, sblk->uid_start, + ((sblk->no_uids + sblk->no_guids) * + sizeof(unsigned int)) | + SQUASHFS_COMPRESSED_BIT_BLOCK, NULL, (sblk->no_uids + sblk->no_guids) * sizeof(unsigned int))) { + ERROR("unable to read uid/gid table\n"); + goto failed_mount; + } + + + if (sblk->s_major == 1 && squashfs_1_0_supported(msblk)) + goto allocate_root; + + if (!(msblk->fragment = kmalloc(sizeof(struct squashfs_fragment_cache) * + SQUASHFS_CACHED_FRAGMENTS, GFP_KERNEL))) { + ERROR("Failed to allocate fragment block cache\n"); + goto failed_mount; + } + + for (i = 0; i < SQUASHFS_CACHED_FRAGMENTS; i++) { + msblk->fragment[i].locked = 0; + msblk->fragment[i].block = SQUASHFS_INVALID_BLK; + msblk->fragment[i].data = NULL; + } + + msblk->next_fragment = 0; + + /* Allocate and read fragment index table */ + if (msblk->read_fragment_index_table(s) == 0) + goto failed_mount; + + if(sblk->s_major < 3 || sblk->lookup_table_start == SQUASHFS_INVALID_BLK) + goto allocate_root; + + /* Allocate and read inode lookup table */ + if (read_inode_lookup_table(s) == 0) + goto failed_mount; + + s->s_op = &squashfs_export_super_ops; + s->s_export_op = &squashfs_export_ops; + +allocate_root: + root = new_inode(s); + if ((msblk->read_inode)(root, sblk->root_inode) == 0) + goto failed_mount; + insert_inode_hash(root); + + if ((s->s_root = d_alloc_root(root)) == NULL) { + ERROR("Root inode create failed\n"); + iput(root); + goto failed_mount; + } + + TRACE("Leaving squashfs_read_super\n"); + return 0; + +failed_mount: + kfree(msblk->inode_lookup_table); + kfree(msblk->fragment_index); + kfree(msblk->fragment); + kfree(msblk->uid); + kfree(msblk->read_page); + kfree(msblk->block_cache); + kfree(msblk->fragment_index_2); + vfree(msblk->stream.workspace); + kfree(s->s_fs_info); + s->s_fs_info = NULL; + return -EINVAL; + +failure: + return -ENOMEM; +} + + +static int squashfs_statfs(struct dentry *dentry, struct kstatfs *buf) +{ + struct squashfs_sb_info *msblk = dentry->d_sb->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + + TRACE("Entered squashfs_statfs\n"); + + buf->f_type = SQUASHFS_MAGIC; + buf->f_bsize = sblk->block_size; + buf->f_blocks = ((sblk->bytes_used - 1) >> sblk->block_log) + 1; + buf->f_bfree = buf->f_bavail = 0; + buf->f_files = sblk->inodes; + buf->f_ffree = 0; + buf->f_namelen = SQUASHFS_NAME_LEN; + + return 0; +} + + +static int squashfs_symlink_readpage(struct file *file, struct page *page) +{ + struct inode *inode = page->mapping->host; + int index = page->index << PAGE_CACHE_SHIFT, length, bytes; + long long block = SQUASHFS_I(inode)->start_block; + int offset = SQUASHFS_I(inode)->offset; + void *pageaddr = kmap(page); + + TRACE("Entered squashfs_symlink_readpage, page index %ld, start block " + "%llx, offset %x\n", page->index, + SQUASHFS_I(inode)->start_block, + SQUASHFS_I(inode)->offset); + + for (length = 0; length < index; length += bytes) { + if (!(bytes = squashfs_get_cached_block(inode->i_sb, NULL, + block, offset, PAGE_CACHE_SIZE, &block, + &offset))) { + ERROR("Unable to read symbolic link [%llx:%x]\n", block, + offset); + goto skip_read; + } + } + + if (length != index) { + ERROR("(squashfs_symlink_readpage) length != index\n"); + bytes = 0; + goto skip_read; + } + + bytes = (i_size_read(inode) - length) > PAGE_CACHE_SIZE ? PAGE_CACHE_SIZE : + i_size_read(inode) - length; + + if (!(bytes = squashfs_get_cached_block(inode->i_sb, pageaddr, block, + offset, bytes, &block, &offset))) + ERROR("Unable to read symbolic link [%llx:%x]\n", block, offset); + +skip_read: + memset(pageaddr + bytes, 0, PAGE_CACHE_SIZE - bytes); + kunmap(page); + flush_dcache_page(page); + SetPageUptodate(page); + unlock_page(page); + + return 0; +} + + +struct meta_index *locate_meta_index(struct inode *inode, int index, int offset) +{ + struct meta_index *meta = NULL; + struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info; + int i; + + mutex_lock(&msblk->meta_index_mutex); + + TRACE("locate_meta_index: index %d, offset %d\n", index, offset); + + if(msblk->meta_index == NULL) + goto not_allocated; + + for (i = 0; i < SQUASHFS_META_NUMBER; i ++) + if (msblk->meta_index[i].inode_number == inode->i_ino && + msblk->meta_index[i].offset >= offset && + msblk->meta_index[i].offset <= index && + msblk->meta_index[i].locked == 0) { + TRACE("locate_meta_index: entry %d, offset %d\n", i, + msblk->meta_index[i].offset); + meta = &msblk->meta_index[i]; + offset = meta->offset; + } + + if (meta) + meta->locked = 1; + +not_allocated: + mutex_unlock(&msblk->meta_index_mutex); + + return meta; +} + + +struct meta_index *empty_meta_index(struct inode *inode, int offset, int skip) +{ + struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info; + struct meta_index *meta = NULL; + int i; + + mutex_lock(&msblk->meta_index_mutex); + + TRACE("empty_meta_index: offset %d, skip %d\n", offset, skip); + + if(msblk->meta_index == NULL) { + if (!(msblk->meta_index = kmalloc(sizeof(struct meta_index) * + SQUASHFS_META_NUMBER, GFP_KERNEL))) { + ERROR("Failed to allocate meta_index\n"); + goto failed; + } + for(i = 0; i < SQUASHFS_META_NUMBER; i++) { + msblk->meta_index[i].inode_number = 0; + msblk->meta_index[i].locked = 0; + } + msblk->next_meta_index = 0; + } + + for(i = SQUASHFS_META_NUMBER; i && + msblk->meta_index[msblk->next_meta_index].locked; i --) + msblk->next_meta_index = (msblk->next_meta_index + 1) % + SQUASHFS_META_NUMBER; + + if(i == 0) { + TRACE("empty_meta_index: failed!\n"); + goto failed; + } + + TRACE("empty_meta_index: returned meta entry %d, %p\n", + msblk->next_meta_index, + &msblk->meta_index[msblk->next_meta_index]); + + meta = &msblk->meta_index[msblk->next_meta_index]; + msblk->next_meta_index = (msblk->next_meta_index + 1) % + SQUASHFS_META_NUMBER; + + meta->inode_number = inode->i_ino; + meta->offset = offset; + meta->skip = skip; + meta->entries = 0; + meta->locked = 1; + +failed: + mutex_unlock(&msblk->meta_index_mutex); + return meta; +} + + +void release_meta_index(struct inode *inode, struct meta_index *meta) +{ + meta->locked = 0; + smp_mb(); +} + + +static int read_block_index(struct super_block *s, int blocks, char *block_list, + long long *start_block, int *offset) +{ + struct squashfs_sb_info *msblk = s->s_fs_info; + unsigned int *block_listp; + int block = 0; + + if (msblk->swap) { + char sblock_list[blocks << 2]; + + if (!squashfs_get_cached_block(s, sblock_list, *start_block, + *offset, blocks << 2, start_block, offset)) { + ERROR("Unable to read block list [%llx:%x]\n", + *start_block, *offset); + goto failure; + } + SQUASHFS_SWAP_INTS(((unsigned int *)block_list), + ((unsigned int *)sblock_list), blocks); + } else + if (!squashfs_get_cached_block(s, block_list, *start_block, + *offset, blocks << 2, start_block, offset)) { + ERROR("Unable to read block list [%llx:%x]\n", + *start_block, *offset); + goto failure; + } + + for (block_listp = (unsigned int *) block_list; blocks; + block_listp++, blocks --) + block += SQUASHFS_COMPRESSED_SIZE_BLOCK(*block_listp); + + return block; + +failure: + return -1; +} + + +#define SIZE 256 + +static inline int calculate_skip(int blocks) { + int skip = (blocks - 1) / ((SQUASHFS_SLOTS * SQUASHFS_META_ENTRIES + 1) * SQUASHFS_META_INDEXES); + return skip >= 7 ? 7 : skip + 1; +} + + +static int get_meta_index(struct inode *inode, int index, + long long *index_block, int *index_offset, + long long *data_block, char *block_list) +{ + struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + int skip = calculate_skip(i_size_read(inode) >> sblk->block_log); + int offset = 0; + struct meta_index *meta; + struct meta_entry *meta_entry; + long long cur_index_block = SQUASHFS_I(inode)->u.s1.block_list_start; + int cur_offset = SQUASHFS_I(inode)->offset; + long long cur_data_block = SQUASHFS_I(inode)->start_block; + int i; + + index /= SQUASHFS_META_INDEXES * skip; + + while ( offset < index ) { + meta = locate_meta_index(inode, index, offset + 1); + + if (meta == NULL) { + if ((meta = empty_meta_index(inode, offset + 1, + skip)) == NULL) + goto all_done; + } else { + if(meta->entries == 0) + goto failed; + offset = index < meta->offset + meta->entries ? index : + meta->offset + meta->entries - 1; + meta_entry = &meta->meta_entry[offset - meta->offset]; + cur_index_block = meta_entry->index_block + sblk->inode_table_start; + cur_offset = meta_entry->offset; + cur_data_block = meta_entry->data_block; + TRACE("get_meta_index: offset %d, meta->offset %d, " + "meta->entries %d\n", offset, meta->offset, + meta->entries); + TRACE("get_meta_index: index_block 0x%llx, offset 0x%x" + " data_block 0x%llx\n", cur_index_block, + cur_offset, cur_data_block); + } + + for (i = meta->offset + meta->entries; i <= index && + i < meta->offset + SQUASHFS_META_ENTRIES; i++) { + int blocks = skip * SQUASHFS_META_INDEXES; + + while (blocks) { + int block = blocks > (SIZE >> 2) ? (SIZE >> 2) : + blocks; + int res = read_block_index(inode->i_sb, block, + block_list, &cur_index_block, + &cur_offset); + + if (res == -1) + goto failed; + + cur_data_block += res; + blocks -= block; + } + + meta_entry = &meta->meta_entry[i - meta->offset]; + meta_entry->index_block = cur_index_block - sblk->inode_table_start; + meta_entry->offset = cur_offset; + meta_entry->data_block = cur_data_block; + meta->entries ++; + offset ++; + } + + TRACE("get_meta_index: meta->offset %d, meta->entries %d\n", + meta->offset, meta->entries); + + release_meta_index(inode, meta); + } + +all_done: + *index_block = cur_index_block; + *index_offset = cur_offset; + *data_block = cur_data_block; + + return offset * SQUASHFS_META_INDEXES * skip; + +failed: + release_meta_index(inode, meta); + return -1; +} + + +static long long read_blocklist(struct inode *inode, int index, + int readahead_blks, char *block_list, + unsigned short **block_p, unsigned int *bsize) +{ + long long block_ptr; + int offset; + long long block; + int res = get_meta_index(inode, index, &block_ptr, &offset, &block, + block_list); + + TRACE("read_blocklist: res %d, index %d, block_ptr 0x%llx, offset" + " 0x%x, block 0x%llx\n", res, index, block_ptr, offset, + block); + + if(res == -1) + goto failure; + + index -= res; + + while ( index ) { + int blocks = index > (SIZE >> 2) ? (SIZE >> 2) : index; + int res = read_block_index(inode->i_sb, blocks, block_list, + &block_ptr, &offset); + if (res == -1) + goto failure; + block += res; + index -= blocks; + } + + if (read_block_index(inode->i_sb, 1, block_list, + &block_ptr, &offset) == -1) + goto failure; + *bsize = *((unsigned int *) block_list); + + return block; + +failure: + return 0; +} + + +static int squashfs_readpage(struct file *file, struct page *page) +{ + struct inode *inode = page->mapping->host; + struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + unsigned char *block_list; + long long block; + unsigned int bsize, i = 0, bytes = 0, byte_offset = 0; + int index = page->index >> (sblk->block_log - PAGE_CACHE_SHIFT); + void *pageaddr; + struct squashfs_fragment_cache *fragment = NULL; + char *data_ptr = msblk->read_page; + + int mask = (1 << (sblk->block_log - PAGE_CACHE_SHIFT)) - 1; + int start_index = page->index & ~mask; + int end_index = start_index | mask; + + TRACE("Entered squashfs_readpage, page index %lx, start block %llx\n", + page->index, + SQUASHFS_I(inode)->start_block); + + if (!(block_list = kmalloc(SIZE, GFP_KERNEL))) { + ERROR("Failed to allocate block_list\n"); + goto skip_read; + } + + if (page->index >= ((i_size_read(inode) + PAGE_CACHE_SIZE - 1) >> + PAGE_CACHE_SHIFT)) + goto skip_read; + + if (SQUASHFS_I(inode)->u.s1.fragment_start_block == SQUASHFS_INVALID_BLK + || index < (i_size_read(inode) >> + sblk->block_log)) { + if ((block = (msblk->read_blocklist)(inode, index, 1, + block_list, NULL, &bsize)) == 0) + goto skip_read; + + mutex_lock(&msblk->read_page_mutex); + + if (!(bytes = squashfs_read_data(inode->i_sb, msblk->read_page, + block, bsize, NULL, sblk->block_size))) { + ERROR("Unable to read page, block %llx, size %x\n", block, + bsize); + mutex_unlock(&msblk->read_page_mutex); + goto skip_read; + } + } else { + if ((fragment = get_cached_fragment(inode->i_sb, + SQUASHFS_I(inode)-> + u.s1.fragment_start_block, + SQUASHFS_I(inode)->u.s1.fragment_size)) + == NULL) { + ERROR("Unable to read page, block %llx, size %x\n", + SQUASHFS_I(inode)-> + u.s1.fragment_start_block, + (int) SQUASHFS_I(inode)-> + u.s1.fragment_size); + goto skip_read; + } + bytes = SQUASHFS_I(inode)->u.s1.fragment_offset + + (i_size_read(inode) & (sblk->block_size + - 1)); + byte_offset = SQUASHFS_I(inode)->u.s1.fragment_offset; + data_ptr = fragment->data; + } + + for (i = start_index; i <= end_index && byte_offset < bytes; + i++, byte_offset += PAGE_CACHE_SIZE) { + struct page *push_page; + int avail = (bytes - byte_offset) > PAGE_CACHE_SIZE ? + PAGE_CACHE_SIZE : bytes - byte_offset; + + TRACE("bytes %d, i %d, byte_offset %d, available_bytes %d\n", + bytes, i, byte_offset, avail); + + push_page = (i == page->index) ? page : + grab_cache_page_nowait(page->mapping, i); + + if (!push_page) + continue; + + if (PageUptodate(push_page)) + goto skip_page; + + pageaddr = kmap_atomic(push_page, KM_USER0); + memcpy(pageaddr, data_ptr + byte_offset, avail); + memset(pageaddr + avail, 0, PAGE_CACHE_SIZE - avail); + kunmap_atomic(pageaddr, KM_USER0); + flush_dcache_page(push_page); + SetPageUptodate(push_page); +skip_page: + unlock_page(push_page); + if(i != page->index) + page_cache_release(push_page); + } + + if (SQUASHFS_I(inode)->u.s1.fragment_start_block == SQUASHFS_INVALID_BLK + || index < (i_size_read(inode) >> + sblk->block_log)) + mutex_unlock(&msblk->read_page_mutex); + else + release_cached_fragment(msblk, fragment); + + kfree(block_list); + return 0; + +skip_read: + pageaddr = kmap_atomic(page, KM_USER0); + memset(pageaddr + bytes, 0, PAGE_CACHE_SIZE - bytes); + kunmap_atomic(pageaddr, KM_USER0); + flush_dcache_page(page); + SetPageUptodate(page); + unlock_page(page); + + kfree(block_list); + return 0; +} + + +static int squashfs_readpage4K(struct file *file, struct page *page) +{ + struct inode *inode = page->mapping->host; + struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + unsigned char *block_list; + long long block; + unsigned int bsize, bytes = 0; + void *pageaddr; + + TRACE("Entered squashfs_readpage4K, page index %lx, start block %llx\n", + page->index, + SQUASHFS_I(inode)->start_block); + + if (page->index >= ((i_size_read(inode) + PAGE_CACHE_SIZE - 1) >> + PAGE_CACHE_SHIFT)) { + block_list = NULL; + goto skip_read; + } + + if (!(block_list = kmalloc(SIZE, GFP_KERNEL))) { + ERROR("Failed to allocate block_list\n"); + goto skip_read; + } + + if (SQUASHFS_I(inode)->u.s1.fragment_start_block == SQUASHFS_INVALID_BLK + || page->index < (i_size_read(inode) >> + sblk->block_log)) { + block = (msblk->read_blocklist)(inode, page->index, 1, + block_list, NULL, &bsize); + if(block == 0) + goto skip_read; + + mutex_lock(&msblk->read_page_mutex); + bytes = squashfs_read_data(inode->i_sb, msblk->read_page, block, + bsize, NULL, sblk->block_size); + if (bytes) { + pageaddr = kmap_atomic(page, KM_USER0); + memcpy(pageaddr, msblk->read_page, bytes); + kunmap_atomic(pageaddr, KM_USER0); + } else + ERROR("Unable to read page, block %llx, size %x\n", + block, bsize); + mutex_unlock(&msblk->read_page_mutex); + } else { + struct squashfs_fragment_cache *fragment = + get_cached_fragment(inode->i_sb, + SQUASHFS_I(inode)-> + u.s1.fragment_start_block, + SQUASHFS_I(inode)-> u.s1.fragment_size); + if (fragment) { + bytes = i_size_read(inode) & (sblk->block_size - 1); + pageaddr = kmap_atomic(page, KM_USER0); + memcpy(pageaddr, fragment->data + SQUASHFS_I(inode)-> + u.s1.fragment_offset, bytes); + kunmap_atomic(pageaddr, KM_USER0); + release_cached_fragment(msblk, fragment); + } else + ERROR("Unable to read page, block %llx, size %x\n", + SQUASHFS_I(inode)-> + u.s1.fragment_start_block, (int) + SQUASHFS_I(inode)-> u.s1.fragment_size); + } + +skip_read: + pageaddr = kmap_atomic(page, KM_USER0); + memset(pageaddr + bytes, 0, PAGE_CACHE_SIZE - bytes); + kunmap_atomic(pageaddr, KM_USER0); + flush_dcache_page(page); + SetPageUptodate(page); + unlock_page(page); + + kfree(block_list); + return 0; +} + + +static int get_dir_index_using_offset(struct super_block *s, long long + *next_block, unsigned int *next_offset, + long long index_start, + unsigned int index_offset, int i_count, + long long f_pos) +{ + struct squashfs_sb_info *msblk = s->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + int i, length = 0; + struct squashfs_dir_index index; + + TRACE("Entered get_dir_index_using_offset, i_count %d, f_pos %d\n", + i_count, (unsigned int) f_pos); + + f_pos =- 3; + if (f_pos == 0) + goto finish; + + for (i = 0; i < i_count; i++) { + if (msblk->swap) { + struct squashfs_dir_index sindex; + squashfs_get_cached_block(s, (char *) &sindex, + index_start, index_offset, + sizeof(sindex), &index_start, + &index_offset); + SQUASHFS_SWAP_DIR_INDEX(&index, &sindex); + } else + squashfs_get_cached_block(s, (char *) &index, + index_start, index_offset, + sizeof(index), &index_start, + &index_offset); + + if (index.index > f_pos) + break; + + squashfs_get_cached_block(s, NULL, index_start, index_offset, + index.size + 1, &index_start, + &index_offset); + + length = index.index; + *next_block = index.start_block + sblk->directory_table_start; + } + + *next_offset = (length + *next_offset) % SQUASHFS_METADATA_SIZE; + +finish: + return length + 3; +} + + +static int get_dir_index_using_name(struct super_block *s, long long + *next_block, unsigned int *next_offset, + long long index_start, + unsigned int index_offset, int i_count, + const char *name, int size) +{ + struct squashfs_sb_info *msblk = s->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + int i, length = 0; + struct squashfs_dir_index *index; + char *str; + + TRACE("Entered get_dir_index_using_name, i_count %d\n", i_count); + + if (!(str = kmalloc(sizeof(struct squashfs_dir_index) + + (SQUASHFS_NAME_LEN + 1) * 2, GFP_KERNEL))) { + ERROR("Failed to allocate squashfs_dir_index\n"); + goto failure; + } + + index = (struct squashfs_dir_index *) (str + SQUASHFS_NAME_LEN + 1); + strncpy(str, name, size); + str[size] = '\0'; + + for (i = 0; i < i_count; i++) { + if (msblk->swap) { + struct squashfs_dir_index sindex; + squashfs_get_cached_block(s, (char *) &sindex, + index_start, index_offset, + sizeof(sindex), &index_start, + &index_offset); + SQUASHFS_SWAP_DIR_INDEX(index, &sindex); + } else + squashfs_get_cached_block(s, (char *) index, + index_start, index_offset, + sizeof(struct squashfs_dir_index), + &index_start, &index_offset); + + squashfs_get_cached_block(s, index->name, index_start, + index_offset, index->size + 1, + &index_start, &index_offset); + + index->name[index->size + 1] = '\0'; + + if (strcmp(index->name, str) > 0) + break; + + length = index->index; + *next_block = index->start_block + sblk->directory_table_start; + } + + *next_offset = (length + *next_offset) % SQUASHFS_METADATA_SIZE; + kfree(str); +failure: + return length + 3; +} + + +static int squashfs_readdir(struct file *file, void *dirent, filldir_t filldir) +{ + struct inode *i = file->f_dentry->d_inode; + struct squashfs_sb_info *msblk = i->i_sb->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + long long next_block = SQUASHFS_I(i)->start_block + + sblk->directory_table_start; + int next_offset = SQUASHFS_I(i)->offset, length = 0, + dir_count; + struct squashfs_dir_header dirh; + struct squashfs_dir_entry *dire; + + TRACE("Entered squashfs_readdir [%llx:%x]\n", next_block, next_offset); + + if (!(dire = kmalloc(sizeof(struct squashfs_dir_entry) + + SQUASHFS_NAME_LEN + 1, GFP_KERNEL))) { + ERROR("Failed to allocate squashfs_dir_entry\n"); + goto finish; + } + + while(file->f_pos < 3) { + char *name; + int size, i_ino; + + if(file->f_pos == 0) { + name = "."; + size = 1; + i_ino = i->i_ino; + } else { + name = ".."; + size = 2; + i_ino = SQUASHFS_I(i)->u.s2.parent_inode; + } + TRACE("Calling filldir(%x, %s, %d, %d, %d, %d)\n", + (unsigned int) dirent, name, size, (int) + file->f_pos, i_ino, + squashfs_filetype_table[1]); + + if (filldir(dirent, name, size, + file->f_pos, i_ino, + squashfs_filetype_table[1]) < 0) { + TRACE("Filldir returned less than 0\n"); + goto finish; + } + file->f_pos += size; + } + + length = get_dir_index_using_offset(i->i_sb, &next_block, &next_offset, + SQUASHFS_I(i)->u.s2.directory_index_start, + SQUASHFS_I(i)->u.s2.directory_index_offset, + SQUASHFS_I(i)->u.s2.directory_index_count, + file->f_pos); + + while (length < i_size_read(i)) { + /* read directory header */ + if (msblk->swap) { + struct squashfs_dir_header sdirh; + + if (!squashfs_get_cached_block(i->i_sb, (char *) &sdirh, + next_block, next_offset, sizeof(sdirh), + &next_block, &next_offset)) + goto failed_read; + + length += sizeof(sdirh); + SQUASHFS_SWAP_DIR_HEADER(&dirh, &sdirh); + } else { + if (!squashfs_get_cached_block(i->i_sb, (char *) &dirh, + next_block, next_offset, sizeof(dirh), + &next_block, &next_offset)) + goto failed_read; + + length += sizeof(dirh); + } + + dir_count = dirh.count + 1; + while (dir_count--) { + if (msblk->swap) { + struct squashfs_dir_entry sdire; + if (!squashfs_get_cached_block(i->i_sb, (char *) + &sdire, next_block, next_offset, + sizeof(sdire), &next_block, + &next_offset)) + goto failed_read; + + length += sizeof(sdire); + SQUASHFS_SWAP_DIR_ENTRY(dire, &sdire); + } else { + if (!squashfs_get_cached_block(i->i_sb, (char *) + dire, next_block, next_offset, + sizeof(*dire), &next_block, + &next_offset)) + goto failed_read; + + length += sizeof(*dire); + } + + if (!squashfs_get_cached_block(i->i_sb, dire->name, + next_block, next_offset, + dire->size + 1, &next_block, + &next_offset)) + goto failed_read; + + length += dire->size + 1; + + if (file->f_pos >= length) + continue; + + dire->name[dire->size + 1] = '\0'; + + TRACE("Calling filldir(%x, %s, %d, %d, %x:%x, %d, %d)\n", + (unsigned int) dirent, dire->name, + dire->size + 1, (int) file->f_pos, + dirh.start_block, dire->offset, + dirh.inode_number + dire->inode_number, + squashfs_filetype_table[dire->type]); + + if (filldir(dirent, dire->name, dire->size + 1, + file->f_pos, + dirh.inode_number + dire->inode_number, + squashfs_filetype_table[dire->type]) + < 0) { + TRACE("Filldir returned less than 0\n"); + goto finish; + } + file->f_pos = length; + } + } + +finish: + kfree(dire); + return 0; + +failed_read: + ERROR("Unable to read directory block [%llx:%x]\n", next_block, + next_offset); + kfree(dire); + return 0; +} + + +static struct dentry *squashfs_lookup(struct inode *i, struct dentry *dentry, + struct nameidata *nd) +{ + const unsigned char *name = dentry->d_name.name; + int len = dentry->d_name.len; + struct inode *inode = NULL; + struct squashfs_sb_info *msblk = i->i_sb->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + long long next_block = SQUASHFS_I(i)->start_block + + sblk->directory_table_start; + int next_offset = SQUASHFS_I(i)->offset, length = 0, + dir_count; + struct squashfs_dir_header dirh; + struct squashfs_dir_entry *dire; + + TRACE("Entered squashfs_lookup [%llx:%x]\n", next_block, next_offset); + + if (!(dire = kmalloc(sizeof(struct squashfs_dir_entry) + + SQUASHFS_NAME_LEN + 1, GFP_KERNEL))) { + ERROR("Failed to allocate squashfs_dir_entry\n"); + goto exit_lookup; + } + + if (len > SQUASHFS_NAME_LEN) + goto exit_lookup; + + length = get_dir_index_using_name(i->i_sb, &next_block, &next_offset, + SQUASHFS_I(i)->u.s2.directory_index_start, + SQUASHFS_I(i)->u.s2.directory_index_offset, + SQUASHFS_I(i)->u.s2.directory_index_count, name, + len); + + while (length < i_size_read(i)) { + /* read directory header */ + if (msblk->swap) { + struct squashfs_dir_header sdirh; + if (!squashfs_get_cached_block(i->i_sb, (char *) &sdirh, + next_block, next_offset, sizeof(sdirh), + &next_block, &next_offset)) + goto failed_read; + + length += sizeof(sdirh); + SQUASHFS_SWAP_DIR_HEADER(&dirh, &sdirh); + } else { + if (!squashfs_get_cached_block(i->i_sb, (char *) &dirh, + next_block, next_offset, sizeof(dirh), + &next_block, &next_offset)) + goto failed_read; + + length += sizeof(dirh); + } + + dir_count = dirh.count + 1; + while (dir_count--) { + if (msblk->swap) { + struct squashfs_dir_entry sdire; + if (!squashfs_get_cached_block(i->i_sb, (char *) + &sdire, next_block,next_offset, + sizeof(sdire), &next_block, + &next_offset)) + goto failed_read; + + length += sizeof(sdire); + SQUASHFS_SWAP_DIR_ENTRY(dire, &sdire); + } else { + if (!squashfs_get_cached_block(i->i_sb, (char *) + dire, next_block,next_offset, + sizeof(*dire), &next_block, + &next_offset)) + goto failed_read; + + length += sizeof(*dire); + } + + if (!squashfs_get_cached_block(i->i_sb, dire->name, + next_block, next_offset, dire->size + 1, + &next_block, &next_offset)) + goto failed_read; + + length += dire->size + 1; + + if (name[0] < dire->name[0]) + goto exit_lookup; + + if ((len == dire->size + 1) && !strncmp(name, dire->name, len)) { + squashfs_inode_t ino = SQUASHFS_MKINODE(dirh.start_block, + dire->offset); + + TRACE("calling squashfs_iget for directory " + "entry %s, inode %x:%x, %d\n", name, + dirh.start_block, dire->offset, + dirh.inode_number + dire->inode_number); + + inode = squashfs_iget(i->i_sb, ino, dirh.inode_number + dire->inode_number); + + goto exit_lookup; + } + } + } + +exit_lookup: + kfree(dire); + if (inode) + return d_splice_alias(inode, dentry); + d_add(dentry, inode); + return ERR_PTR(0); + +failed_read: + ERROR("Unable to read directory block [%llx:%x]\n", next_block, + next_offset); + goto exit_lookup; +} + + +static int squashfs_remount(struct super_block *s, int *flags, char *data) +{ + *flags |= MS_RDONLY; + return 0; +} + + +static void squashfs_put_super(struct super_block *s) +{ + int i; + + if (s->s_fs_info) { + struct squashfs_sb_info *sbi = s->s_fs_info; + if (sbi->block_cache) + for (i = 0; i < SQUASHFS_CACHED_BLKS; i++) + if (sbi->block_cache[i].block != + SQUASHFS_INVALID_BLK) + kfree(sbi->block_cache[i].data); + if (sbi->fragment) + for (i = 0; i < SQUASHFS_CACHED_FRAGMENTS; i++) + SQUASHFS_FREE(sbi->fragment[i].data); + kfree(sbi->fragment); + kfree(sbi->block_cache); + kfree(sbi->read_page); + kfree(sbi->uid); + kfree(sbi->fragment_index); + kfree(sbi->fragment_index_2); + kfree(sbi->meta_index); + vfree(sbi->stream.workspace); + kfree(s->s_fs_info); + s->s_fs_info = NULL; + } +} + + +static int squashfs_get_sb(struct file_system_type *fs_type, int flags, + const char *dev_name, void *data, + struct vfsmount *mnt) +{ + return get_sb_bdev(fs_type, flags, dev_name, data, squashfs_fill_super, + mnt); +} + + +static int __init init_squashfs_fs(void) +{ + int err = init_inodecache(); + if (err) + goto out; + + printk(KERN_INFO "squashfs: version 3.2-r2 (2007/01/15) " + "Phillip Lougher\n"); + + if ((err = register_filesystem(&squashfs_fs_type))) + destroy_inodecache(); + +out: + return err; +} + + +static void __exit exit_squashfs_fs(void) +{ + unregister_filesystem(&squashfs_fs_type); + destroy_inodecache(); +} + + +static struct kmem_cache * squashfs_inode_cachep; + + +static struct inode *squashfs_alloc_inode(struct super_block *sb) +{ + struct squashfs_inode_info *ei; + ei = kmem_cache_alloc(squashfs_inode_cachep, GFP_KERNEL); + if (!ei) + return NULL; + return &ei->vfs_inode; +} + + +static void squashfs_destroy_inode(struct inode *inode) +{ + kmem_cache_free(squashfs_inode_cachep, SQUASHFS_I(inode)); +} + + +static void init_once(void * foo, struct kmem_cache * cachep, unsigned long flags) +{ + struct squashfs_inode_info *ei = foo; + + if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) == + SLAB_CTOR_CONSTRUCTOR) + inode_init_once(&ei->vfs_inode); +} + + +static int __init init_inodecache(void) +{ + squashfs_inode_cachep = kmem_cache_create("squashfs_inode_cache", + sizeof(struct squashfs_inode_info), + 0, SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT, + init_once, NULL); + if (squashfs_inode_cachep == NULL) + return -ENOMEM; + return 0; +} + + +static void destroy_inodecache(void) +{ + kmem_cache_destroy(squashfs_inode_cachep); +} + + +module_init(init_squashfs_fs); +module_exit(exit_squashfs_fs); +MODULE_DESCRIPTION("squashfs 3.2-r2, a compressed read-only filesystem"); +MODULE_AUTHOR("Phillip Lougher "); +MODULE_LICENSE("GPL"); diff --git a/fs/squashfs/squashfs.h b/fs/squashfs/squashfs.h new file mode 100644 index 0000000..6f863f0 --- /dev/null +++ b/fs/squashfs/squashfs.h @@ -0,0 +1,87 @@ +/* + * Squashfs - a compressed read only filesystem for Linux + * + * Copyright (c) 2002, 2003, 2004, 2005, 2006, 2007 + * Phillip Lougher + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2, + * or (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + * + * squashfs.h + */ + +#ifdef CONFIG_SQUASHFS_1_0_COMPATIBILITY +#undef CONFIG_SQUASHFS_1_0_COMPATIBILITY +#endif + +#ifdef SQUASHFS_TRACE +#define TRACE(s, args...) printk(KERN_NOTICE "SQUASHFS: "s, ## args) +#else +#define TRACE(s, args...) {} +#endif + +#define ERROR(s, args...) printk(KERN_ERR "SQUASHFS error: "s, ## args) + +#define SERROR(s, args...) do { \ + if (!silent) \ + printk(KERN_ERR "SQUASHFS error: "s, ## args);\ + } while(0) + +#define WARNING(s, args...) printk(KERN_WARNING "SQUASHFS: "s, ## args) + +static inline struct squashfs_inode_info *SQUASHFS_I(struct inode *inode) +{ + return list_entry(inode, struct squashfs_inode_info, vfs_inode); +} + +#if defined(CONFIG_SQUASHFS_1_0_COMPATIBILITY ) || defined(CONFIG_SQUASHFS_2_0_COMPATIBILITY) +#define SQSH_EXTERN +extern unsigned int squashfs_read_data(struct super_block *s, char *buffer, + long long index, unsigned int length, + long long *next_index, int srclength); +extern int squashfs_get_cached_block(struct super_block *s, char *buffer, + long long block, unsigned int offset, + int length, long long *next_block, + unsigned int *next_offset); +extern void release_cached_fragment(struct squashfs_sb_info *msblk, struct + squashfs_fragment_cache *fragment); +extern struct squashfs_fragment_cache *get_cached_fragment(struct super_block + *s, long long start_block, + int length); +extern struct inode *squashfs_iget(struct super_block *s, squashfs_inode_t inode, unsigned int inode_number); +extern const struct address_space_operations squashfs_symlink_aops; +extern const struct address_space_operations squashfs_aops; +extern const struct address_space_operations squashfs_aops_4K; +extern struct inode_operations squashfs_dir_inode_ops; +#else +#define SQSH_EXTERN static +#endif + +#ifdef CONFIG_SQUASHFS_1_0_COMPATIBILITY +extern int squashfs_1_0_supported(struct squashfs_sb_info *msblk); +#else +static inline int squashfs_1_0_supported(struct squashfs_sb_info *msblk) +{ + return 0; +} +#endif + +#ifdef CONFIG_SQUASHFS_2_0_COMPATIBILITY +extern int squashfs_2_0_supported(struct squashfs_sb_info *msblk); +#else +static inline int squashfs_2_0_supported(struct squashfs_sb_info *msblk) +{ + return 0; +} +#endif diff --git a/fs/squashfs/squashfs2_0.c b/fs/squashfs/squashfs2_0.c new file mode 100644 index 0000000..d8d9d55 --- /dev/null +++ b/fs/squashfs/squashfs2_0.c @@ -0,0 +1,742 @@ +/* + * Squashfs - a compressed read only filesystem for Linux + * + * Copyright (c) 2002, 2003, 2004, 2005, 2006, 2007 + * Phillip Lougher + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2, + * or (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + * + * squashfs2_0.c + */ + +#include +#include +#include +#include +#include +#include + +#include "squashfs.h" +static int squashfs_readdir_2(struct file *file, void *dirent, filldir_t filldir); +static struct dentry *squashfs_lookup_2(struct inode *, struct dentry *, + struct nameidata *); + +static struct file_operations squashfs_dir_ops_2 = { + .read = generic_read_dir, + .readdir = squashfs_readdir_2 +}; + +static struct inode_operations squashfs_dir_inode_ops_2 = { + .lookup = squashfs_lookup_2 +}; + +static unsigned char squashfs_filetype_table[] = { + DT_UNKNOWN, DT_DIR, DT_REG, DT_LNK, DT_BLK, DT_CHR, DT_FIFO, DT_SOCK +}; + +static int read_fragment_index_table_2(struct super_block *s) +{ + struct squashfs_sb_info *msblk = s->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + + if (!(msblk->fragment_index_2 = kmalloc(SQUASHFS_FRAGMENT_INDEX_BYTES_2 + (sblk->fragments), GFP_KERNEL))) { + ERROR("Failed to allocate uid/gid table\n"); + return 0; + } + + if (SQUASHFS_FRAGMENT_INDEX_BYTES_2(sblk->fragments) && + !squashfs_read_data(s, (char *) + msblk->fragment_index_2, + sblk->fragment_table_start, + SQUASHFS_FRAGMENT_INDEX_BYTES_2 + (sblk->fragments) | + SQUASHFS_COMPRESSED_BIT_BLOCK, NULL, SQUASHFS_FRAGMENT_INDEX_BYTES_2(sblk->fragments))) { + ERROR("unable to read fragment index table\n"); + return 0; + } + + if (msblk->swap) { + int i; + unsigned int fragment; + + for (i = 0; i < SQUASHFS_FRAGMENT_INDEXES_2(sblk->fragments); + i++) { + SQUASHFS_SWAP_FRAGMENT_INDEXES_2((&fragment), + &msblk->fragment_index_2[i], 1); + msblk->fragment_index_2[i] = fragment; + } + } + + return 1; +} + + +static int get_fragment_location_2(struct super_block *s, unsigned int fragment, + long long *fragment_start_block, + unsigned int *fragment_size) +{ + struct squashfs_sb_info *msblk = s->s_fs_info; + long long start_block = + msblk->fragment_index_2[SQUASHFS_FRAGMENT_INDEX_2(fragment)]; + int offset = SQUASHFS_FRAGMENT_INDEX_OFFSET_2(fragment); + struct squashfs_fragment_entry_2 fragment_entry; + + if (msblk->swap) { + struct squashfs_fragment_entry_2 sfragment_entry; + + if (!squashfs_get_cached_block(s, (char *) &sfragment_entry, + start_block, offset, + sizeof(sfragment_entry), &start_block, + &offset)) + goto out; + SQUASHFS_SWAP_FRAGMENT_ENTRY_2(&fragment_entry, &sfragment_entry); + } else + if (!squashfs_get_cached_block(s, (char *) &fragment_entry, + start_block, offset, + sizeof(fragment_entry), &start_block, + &offset)) + goto out; + + *fragment_start_block = fragment_entry.start_block; + *fragment_size = fragment_entry.size; + + return 1; + +out: + return 0; +} + + +static void squashfs_new_inode(struct squashfs_sb_info *msblk, struct inode *i, + struct squashfs_base_inode_header_2 *inodeb, unsigned int ino) +{ + struct squashfs_super_block *sblk = &msblk->sblk; + + i->i_ino = ino; + i->i_mtime.tv_sec = sblk->mkfs_time; + i->i_atime.tv_sec = sblk->mkfs_time; + i->i_ctime.tv_sec = sblk->mkfs_time; + i->i_uid = msblk->uid[inodeb->uid]; + i->i_mode = inodeb->mode; + i->i_nlink = 1; + i->i_size = 0; + if (inodeb->guid == SQUASHFS_GUIDS) + i->i_gid = i->i_uid; + else + i->i_gid = msblk->guid[inodeb->guid]; +} + + +static int squashfs_read_inode_2(struct inode *i, squashfs_inode_t inode) +{ + struct super_block *s = i->i_sb; + struct squashfs_sb_info *msblk = s->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + unsigned int block = SQUASHFS_INODE_BLK(inode) + + sblk->inode_table_start; + unsigned int offset = SQUASHFS_INODE_OFFSET(inode); + unsigned int ino = i->i_ino; + long long next_block; + unsigned int next_offset; + union squashfs_inode_header_2 id, sid; + struct squashfs_base_inode_header_2 *inodeb = &id.base, + *sinodeb = &sid.base; + + TRACE("Entered squashfs_iget\n"); + + if (msblk->swap) { + if (!squashfs_get_cached_block(s, (char *) sinodeb, block, + offset, sizeof(*sinodeb), &next_block, + &next_offset)) + goto failed_read; + SQUASHFS_SWAP_BASE_INODE_HEADER_2(inodeb, sinodeb, + sizeof(*sinodeb)); + } else + if (!squashfs_get_cached_block(s, (char *) inodeb, block, + offset, sizeof(*inodeb), &next_block, + &next_offset)) + goto failed_read; + + squashfs_new_inode(msblk, i, inodeb, ino); + + switch(inodeb->inode_type) { + case SQUASHFS_FILE_TYPE: { + struct squashfs_reg_inode_header_2 *inodep = &id.reg; + struct squashfs_reg_inode_header_2 *sinodep = &sid.reg; + long long frag_blk; + unsigned int frag_size = 0; + + if (msblk->swap) { + if (!squashfs_get_cached_block(s, (char *) + sinodep, block, offset, + sizeof(*sinodep), &next_block, + &next_offset)) + goto failed_read; + SQUASHFS_SWAP_REG_INODE_HEADER_2(inodep, sinodep); + } else + if (!squashfs_get_cached_block(s, (char *) + inodep, block, offset, + sizeof(*inodep), &next_block, + &next_offset)) + goto failed_read; + + frag_blk = SQUASHFS_INVALID_BLK; + if (inodep->fragment != SQUASHFS_INVALID_FRAG && + !get_fragment_location_2(s, + inodep->fragment, &frag_blk, &frag_size)) + goto failed_read; + + i->i_size = inodep->file_size; + i->i_fop = &generic_ro_fops; + i->i_mode |= S_IFREG; + i->i_mtime.tv_sec = inodep->mtime; + i->i_atime.tv_sec = inodep->mtime; + i->i_ctime.tv_sec = inodep->mtime; + i->i_blocks = ((i->i_size - 1) >> 9) + 1; + SQUASHFS_I(i)->u.s1.fragment_start_block = frag_blk; + SQUASHFS_I(i)->u.s1.fragment_size = frag_size; + SQUASHFS_I(i)->u.s1.fragment_offset = inodep->offset; + SQUASHFS_I(i)->start_block = inodep->start_block; + SQUASHFS_I(i)->u.s1.block_list_start = next_block; + SQUASHFS_I(i)->offset = next_offset; + if (sblk->block_size > 4096) + i->i_data.a_ops = &squashfs_aops; + else + i->i_data.a_ops = &squashfs_aops_4K; + + TRACE("File inode %x:%x, start_block %x, " + "block_list_start %llx, offset %x\n", + SQUASHFS_INODE_BLK(inode), offset, + inodep->start_block, next_block, + next_offset); + break; + } + case SQUASHFS_DIR_TYPE: { + struct squashfs_dir_inode_header_2 *inodep = &id.dir; + struct squashfs_dir_inode_header_2 *sinodep = &sid.dir; + + if (msblk->swap) { + if (!squashfs_get_cached_block(s, (char *) + sinodep, block, offset, + sizeof(*sinodep), &next_block, + &next_offset)) + goto failed_read; + SQUASHFS_SWAP_DIR_INODE_HEADER_2(inodep, sinodep); + } else + if (!squashfs_get_cached_block(s, (char *) + inodep, block, offset, + sizeof(*inodep), &next_block, + &next_offset)) + goto failed_read; + + i->i_size = inodep->file_size; + i->i_op = &squashfs_dir_inode_ops_2; + i->i_fop = &squashfs_dir_ops_2; + i->i_mode |= S_IFDIR; + i->i_mtime.tv_sec = inodep->mtime; + i->i_atime.tv_sec = inodep->mtime; + i->i_ctime.tv_sec = inodep->mtime; + SQUASHFS_I(i)->start_block = inodep->start_block; + SQUASHFS_I(i)->offset = inodep->offset; + SQUASHFS_I(i)->u.s2.directory_index_count = 0; + SQUASHFS_I(i)->u.s2.parent_inode = 0; + + TRACE("Directory inode %x:%x, start_block %x, offset " + "%x\n", SQUASHFS_INODE_BLK(inode), + offset, inodep->start_block, + inodep->offset); + break; + } + case SQUASHFS_LDIR_TYPE: { + struct squashfs_ldir_inode_header_2 *inodep = &id.ldir; + struct squashfs_ldir_inode_header_2 *sinodep = &sid.ldir; + + if (msblk->swap) { + if (!squashfs_get_cached_block(s, (char *) + sinodep, block, offset, + sizeof(*sinodep), &next_block, + &next_offset)) + goto failed_read; + SQUASHFS_SWAP_LDIR_INODE_HEADER_2(inodep, + sinodep); + } else + if (!squashfs_get_cached_block(s, (char *) + inodep, block, offset, + sizeof(*inodep), &next_block, + &next_offset)) + goto failed_read; + + i->i_size = inodep->file_size; + i->i_op = &squashfs_dir_inode_ops_2; + i->i_fop = &squashfs_dir_ops_2; + i->i_mode |= S_IFDIR; + i->i_mtime.tv_sec = inodep->mtime; + i->i_atime.tv_sec = inodep->mtime; + i->i_ctime.tv_sec = inodep->mtime; + SQUASHFS_I(i)->start_block = inodep->start_block; + SQUASHFS_I(i)->offset = inodep->offset; + SQUASHFS_I(i)->u.s2.directory_index_start = next_block; + SQUASHFS_I(i)->u.s2.directory_index_offset = + next_offset; + SQUASHFS_I(i)->u.s2.directory_index_count = + inodep->i_count; + SQUASHFS_I(i)->u.s2.parent_inode = 0; + + TRACE("Long directory inode %x:%x, start_block %x, " + "offset %x\n", + SQUASHFS_INODE_BLK(inode), offset, + inodep->start_block, inodep->offset); + break; + } + case SQUASHFS_SYMLINK_TYPE: { + struct squashfs_symlink_inode_header_2 *inodep = + &id.symlink; + struct squashfs_symlink_inode_header_2 *sinodep = + &sid.symlink; + + if (msblk->swap) { + if (!squashfs_get_cached_block(s, (char *) + sinodep, block, offset, + sizeof(*sinodep), &next_block, + &next_offset)) + goto failed_read; + SQUASHFS_SWAP_SYMLINK_INODE_HEADER_2(inodep, + sinodep); + } else + if (!squashfs_get_cached_block(s, (char *) + inodep, block, offset, + sizeof(*inodep), &next_block, + &next_offset)) + goto failed_read; + + i->i_size = inodep->symlink_size; + i->i_op = &page_symlink_inode_operations; + i->i_data.a_ops = &squashfs_symlink_aops; + i->i_mode |= S_IFLNK; + SQUASHFS_I(i)->start_block = next_block; + SQUASHFS_I(i)->offset = next_offset; + + TRACE("Symbolic link inode %x:%x, start_block %llx, " + "offset %x\n", + SQUASHFS_INODE_BLK(inode), offset, + next_block, next_offset); + break; + } + case SQUASHFS_BLKDEV_TYPE: + case SQUASHFS_CHRDEV_TYPE: { + struct squashfs_dev_inode_header_2 *inodep = &id.dev; + struct squashfs_dev_inode_header_2 *sinodep = &sid.dev; + + if (msblk->swap) { + if (!squashfs_get_cached_block(s, (char *) + sinodep, block, offset, + sizeof(*sinodep), &next_block, + &next_offset)) + goto failed_read; + SQUASHFS_SWAP_DEV_INODE_HEADER_2(inodep, sinodep); + } else + if (!squashfs_get_cached_block(s, (char *) + inodep, block, offset, + sizeof(*inodep), &next_block, + &next_offset)) + goto failed_read; + + i->i_mode |= (inodeb->inode_type == + SQUASHFS_CHRDEV_TYPE) ? S_IFCHR : + S_IFBLK; + init_special_inode(i, i->i_mode, + old_decode_dev(inodep->rdev)); + + TRACE("Device inode %x:%x, rdev %x\n", + SQUASHFS_INODE_BLK(inode), offset, + inodep->rdev); + break; + } + case SQUASHFS_FIFO_TYPE: + case SQUASHFS_SOCKET_TYPE: { + + i->i_mode |= (inodeb->inode_type == SQUASHFS_FIFO_TYPE) + ? S_IFIFO : S_IFSOCK; + init_special_inode(i, i->i_mode, 0); + break; + } + default: + ERROR("Unknown inode type %d in squashfs_iget!\n", + inodeb->inode_type); + goto failed_read1; + } + + return 1; + +failed_read: + ERROR("Unable to read inode [%x:%x]\n", block, offset); + +failed_read1: + return 0; +} + + +static int get_dir_index_using_offset(struct super_block *s, long long + *next_block, unsigned int *next_offset, + long long index_start, + unsigned int index_offset, int i_count, + long long f_pos) +{ + struct squashfs_sb_info *msblk = s->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + int i, length = 0; + struct squashfs_dir_index_2 index; + + TRACE("Entered get_dir_index_using_offset, i_count %d, f_pos %d\n", + i_count, (unsigned int) f_pos); + + if (f_pos == 0) + goto finish; + + for (i = 0; i < i_count; i++) { + if (msblk->swap) { + struct squashfs_dir_index_2 sindex; + squashfs_get_cached_block(s, (char *) &sindex, + index_start, index_offset, + sizeof(sindex), &index_start, + &index_offset); + SQUASHFS_SWAP_DIR_INDEX_2(&index, &sindex); + } else + squashfs_get_cached_block(s, (char *) &index, + index_start, index_offset, + sizeof(index), &index_start, + &index_offset); + + if (index.index > f_pos) + break; + + squashfs_get_cached_block(s, NULL, index_start, index_offset, + index.size + 1, &index_start, + &index_offset); + + length = index.index; + *next_block = index.start_block + sblk->directory_table_start; + } + + *next_offset = (length + *next_offset) % SQUASHFS_METADATA_SIZE; + +finish: + return length; +} + + +static int get_dir_index_using_name(struct super_block *s, long long + *next_block, unsigned int *next_offset, + long long index_start, + unsigned int index_offset, int i_count, + const char *name, int size) +{ + struct squashfs_sb_info *msblk = s->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + int i, length = 0; + struct squashfs_dir_index_2 *index; + char *str; + + TRACE("Entered get_dir_index_using_name, i_count %d\n", i_count); + + if (!(str = kmalloc(sizeof(struct squashfs_dir_index) + + (SQUASHFS_NAME_LEN + 1) * 2, GFP_KERNEL))) { + ERROR("Failed to allocate squashfs_dir_index\n"); + goto failure; + } + + index = (struct squashfs_dir_index_2 *) (str + SQUASHFS_NAME_LEN + 1); + strncpy(str, name, size); + str[size] = '\0'; + + for (i = 0; i < i_count; i++) { + if (msblk->swap) { + struct squashfs_dir_index_2 sindex; + squashfs_get_cached_block(s, (char *) &sindex, + index_start, index_offset, + sizeof(sindex), &index_start, + &index_offset); + SQUASHFS_SWAP_DIR_INDEX_2(index, &sindex); + } else + squashfs_get_cached_block(s, (char *) index, + index_start, index_offset, + sizeof(struct squashfs_dir_index_2), + &index_start, &index_offset); + + squashfs_get_cached_block(s, index->name, index_start, + index_offset, index->size + 1, + &index_start, &index_offset); + + index->name[index->size + 1] = '\0'; + + if (strcmp(index->name, str) > 0) + break; + + length = index->index; + *next_block = index->start_block + sblk->directory_table_start; + } + + *next_offset = (length + *next_offset) % SQUASHFS_METADATA_SIZE; + kfree(str); +failure: + return length; +} + + +static int squashfs_readdir_2(struct file *file, void *dirent, filldir_t filldir) +{ + struct inode *i = file->f_dentry->d_inode; + struct squashfs_sb_info *msblk = i->i_sb->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + long long next_block = SQUASHFS_I(i)->start_block + + sblk->directory_table_start; + int next_offset = SQUASHFS_I(i)->offset, length = 0, + dir_count; + struct squashfs_dir_header_2 dirh; + struct squashfs_dir_entry_2 *dire; + + TRACE("Entered squashfs_readdir_2 [%llx:%x]\n", next_block, next_offset); + + if (!(dire = kmalloc(sizeof(struct squashfs_dir_entry) + + SQUASHFS_NAME_LEN + 1, GFP_KERNEL))) { + ERROR("Failed to allocate squashfs_dir_entry\n"); + goto finish; + } + + length = get_dir_index_using_offset(i->i_sb, &next_block, &next_offset, + SQUASHFS_I(i)->u.s2.directory_index_start, + SQUASHFS_I(i)->u.s2.directory_index_offset, + SQUASHFS_I(i)->u.s2.directory_index_count, + file->f_pos); + + while (length < i_size_read(i)) { + /* read directory header */ + if (msblk->swap) { + struct squashfs_dir_header_2 sdirh; + + if (!squashfs_get_cached_block(i->i_sb, (char *) &sdirh, + next_block, next_offset, sizeof(sdirh), + &next_block, &next_offset)) + goto failed_read; + + length += sizeof(sdirh); + SQUASHFS_SWAP_DIR_HEADER_2(&dirh, &sdirh); + } else { + if (!squashfs_get_cached_block(i->i_sb, (char *) &dirh, + next_block, next_offset, sizeof(dirh), + &next_block, &next_offset)) + goto failed_read; + + length += sizeof(dirh); + } + + dir_count = dirh.count + 1; + while (dir_count--) { + if (msblk->swap) { + struct squashfs_dir_entry_2 sdire; + if (!squashfs_get_cached_block(i->i_sb, (char *) + &sdire, next_block, next_offset, + sizeof(sdire), &next_block, + &next_offset)) + goto failed_read; + + length += sizeof(sdire); + SQUASHFS_SWAP_DIR_ENTRY_2(dire, &sdire); + } else { + if (!squashfs_get_cached_block(i->i_sb, (char *) + dire, next_block, next_offset, + sizeof(*dire), &next_block, + &next_offset)) + goto failed_read; + + length += sizeof(*dire); + } + + if (!squashfs_get_cached_block(i->i_sb, dire->name, + next_block, next_offset, + dire->size + 1, &next_block, + &next_offset)) + goto failed_read; + + length += dire->size + 1; + + if (file->f_pos >= length) + continue; + + dire->name[dire->size + 1] = '\0'; + + TRACE("Calling filldir(%x, %s, %d, %d, %x:%x, %d)\n", + (unsigned int) dirent, dire->name, + dire->size + 1, (int) file->f_pos, + dirh.start_block, dire->offset, + squashfs_filetype_table[dire->type]); + + if (filldir(dirent, dire->name, dire->size + 1, + file->f_pos, SQUASHFS_MK_VFS_INODE( + dirh.start_block, dire->offset), + squashfs_filetype_table[dire->type]) + < 0) { + TRACE("Filldir returned less than 0\n"); + goto finish; + } + file->f_pos = length; + } + } + +finish: + kfree(dire); + return 0; + +failed_read: + ERROR("Unable to read directory block [%llx:%x]\n", next_block, + next_offset); + kfree(dire); + return 0; +} + + +static struct dentry *squashfs_lookup_2(struct inode *i, struct dentry *dentry, + struct nameidata *nd) +{ + const unsigned char *name = dentry->d_name.name; + int len = dentry->d_name.len; + struct inode *inode = NULL; + struct squashfs_sb_info *msblk = i->i_sb->s_fs_info; + struct squashfs_super_block *sblk = &msblk->sblk; + long long next_block = SQUASHFS_I(i)->start_block + + sblk->directory_table_start; + int next_offset = SQUASHFS_I(i)->offset, length = 0, + dir_count; + struct squashfs_dir_header_2 dirh; + struct squashfs_dir_entry_2 *dire; + int sorted = sblk->s_major == 2 && sblk->s_minor >= 1; + + TRACE("Entered squashfs_lookup_2 [%llx:%x]\n", next_block, next_offset); + + if (!(dire = kmalloc(sizeof(struct squashfs_dir_entry) + + SQUASHFS_NAME_LEN + 1, GFP_KERNEL))) { + ERROR("Failed to allocate squashfs_dir_entry\n"); + goto exit_loop; + } + + if (len > SQUASHFS_NAME_LEN) + goto exit_loop; + + length = get_dir_index_using_name(i->i_sb, &next_block, &next_offset, + SQUASHFS_I(i)->u.s2.directory_index_start, + SQUASHFS_I(i)->u.s2.directory_index_offset, + SQUASHFS_I(i)->u.s2.directory_index_count, name, + len); + + while (length < i_size_read(i)) { + /* read directory header */ + if (msblk->swap) { + struct squashfs_dir_header_2 sdirh; + if (!squashfs_get_cached_block(i->i_sb, (char *) &sdirh, + next_block, next_offset, sizeof(sdirh), + &next_block, &next_offset)) + goto failed_read; + + length += sizeof(sdirh); + SQUASHFS_SWAP_DIR_HEADER_2(&dirh, &sdirh); + } else { + if (!squashfs_get_cached_block(i->i_sb, (char *) &dirh, + next_block, next_offset, sizeof(dirh), + &next_block, &next_offset)) + goto failed_read; + + length += sizeof(dirh); + } + + dir_count = dirh.count + 1; + while (dir_count--) { + if (msblk->swap) { + struct squashfs_dir_entry_2 sdire; + if (!squashfs_get_cached_block(i->i_sb, (char *) + &sdire, next_block,next_offset, + sizeof(sdire), &next_block, + &next_offset)) + goto failed_read; + + length += sizeof(sdire); + SQUASHFS_SWAP_DIR_ENTRY_2(dire, &sdire); + } else { + if (!squashfs_get_cached_block(i->i_sb, (char *) + dire, next_block,next_offset, + sizeof(*dire), &next_block, + &next_offset)) + goto failed_read; + + length += sizeof(*dire); + } + + if (!squashfs_get_cached_block(i->i_sb, dire->name, + next_block, next_offset, dire->size + 1, + &next_block, &next_offset)) + goto failed_read; + + length += dire->size + 1; + + if (sorted && name[0] < dire->name[0]) + goto exit_loop; + + if ((len == dire->size + 1) && !strncmp(name, + dire->name, len)) { + squashfs_inode_t ino = + SQUASHFS_MKINODE(dirh.start_block, + dire->offset); + unsigned int inode_number = SQUASHFS_MK_VFS_INODE(dirh.start_block, + dire->offset); + + TRACE("calling squashfs_iget for directory " + "entry %s, inode %x:%x, %lld\n", name, + dirh.start_block, dire->offset, ino); + + inode = squashfs_iget(i->i_sb, ino, inode_number); + + goto exit_loop; + } + } + } + +exit_loop: + kfree(dire); + d_add(dentry, inode); + return ERR_PTR(0); + +failed_read: + ERROR("Unable to read directory block [%llx:%x]\n", next_block, + next_offset); + goto exit_loop; +} + + +int squashfs_2_0_supported(struct squashfs_sb_info *msblk) +{ + struct squashfs_super_block *sblk = &msblk->sblk; + + msblk->read_inode = squashfs_read_inode_2; + msblk->read_fragment_index_table = read_fragment_index_table_2; + + sblk->bytes_used = sblk->bytes_used_2; + sblk->uid_start = sblk->uid_start_2; + sblk->guid_start = sblk->guid_start_2; + sblk->inode_table_start = sblk->inode_table_start_2; + sblk->directory_table_start = sblk->directory_table_start_2; + sblk->fragment_table_start = sblk->fragment_table_start_2; + + return 1; +} diff --git a/include/linux/aufs_type.h b/include/linux/aufs_type.h new file mode 100755 index 0000000..8b4629e --- /dev/null +++ b/include/linux/aufs_type.h @@ -0,0 +1,97 @@ +/* + * Copyright (C) 2005, 2006, 2007 Junjiro Okajima + * + * This program, aufs is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/* $Id: aufs_type.h,v 1.55 2007/05/14 03:40:57 sfjro Exp $ */ + +#include + +#ifndef __AUFS_TYPE_H__ +#define __AUFS_TYPE_H__ + +#define AUFS_VERSION "20070514" + +/* ---------------------------------------------------------------------- */ + +#ifdef CONFIG_AUFS_BRANCH_MAX_127 +typedef char aufs_bindex_t; +#define AUFS_BRANCH_MAX 127 +#else +typedef short aufs_bindex_t; +#ifdef CONFIG_AUFS_BRANCH_MAX_511 +#define AUFS_BRANCH_MAX 511 +#elif defined(CONFIG_AUFS_BRANCH_MAX_1023) +#define AUFS_BRANCH_MAX 1023 +#elif defined(CONFIG_AUFS_BRANCH_MAX_32767) +#define AUFS_BRANCH_MAX 32767 +#else +#error unknown CONFIG_AUFS_BRANCH_MAX value +#endif +#endif + +#define AUFS_NAME "aufs" +#define AUFS_FSTYPE AUFS_NAME + +#define AUFS_ROOT_INO 2 +#define AUFS_FIRST_INO 11 + +#define AUFS_WH_PFX ".wh." +#define AUFS_WH_PFX_LEN ((int)sizeof(AUFS_WH_PFX) - 1) +#define AUFS_XINO_FNAME "." AUFS_NAME ".xino" +#define AUFS_XINO_DEFPATH "/tmp/" AUFS_XINO_FNAME +#define AUFS_DIRWH_DEF 3 +#define AUFS_RDCACHE_DEF 10 /* seconds */ +#define AUFS_WKQ_NAME AUFS_NAME "d" +#define AUFS_NWKQ_DEF 4 + +#ifdef CONFIG_AUFS_COMPAT +#define AUFS_DIROPQ_NAME "__dir_opaque" +#else +#define AUFS_DIROPQ_NAME AUFS_WH_PFX ".opq" /* whiteouted doubly */ +#endif +#define AUFS_WH_DIROPQ AUFS_WH_PFX AUFS_DIROPQ_NAME + +/* will be whiteouted doubly */ +#define AUFS_WH_BASENAME AUFS_WH_PFX AUFS_NAME +#define AUFS_WH_PLINKDIR AUFS_WH_PFX "plink" + +/* ---------------------------------------------------------------------- */ + +/* ioctl */ +enum {AuCtlErr, AuCtlErr_Last}; +enum { + AuCtl_REFRESH, //AuCtl_REFRESHV, + //AuCtl_FLUSH_PLINK, + //AuCtl_CPUP, + AuCtl_CPDOWN, AuCtl_MVDOWN +}; + +struct aufs_ctl_cp { + int bsrc, bdst; + int err; +}; + +#define Type 'A' +#define AUFS_CTL_REFRESH _IO(Type, AuCtl_REFRESH) +//#define AUFS_CTL_REFRESHV _IO(Type, AuCtl_REFRESHV) +//#define AUFS_CTL_FLUSH_PLINK _IOR(Type, AuCtl_FLUSH_PLINK) +//#define AUFS_CTL_CPUP _IOWR(Type, AuCtl_CPUP, struct aufs_ctl_cp) +#define AUFS_CTL_CPDOWN _IOWR(Type, AuCtl_CPDOWN, struct aufs_ctl_cp) +#define AUFS_CTL_MVDOWN _IOWR(Type, AuCtl_MVDOWN, struct aufs_ctl_cp) +#undef Type + +#endif /* __AUFS_TYPE_H__ */ diff --git a/include/linux/squashfs_fs.h b/include/linux/squashfs_fs.h new file mode 100644 index 0000000..a9380ad --- /dev/null +++ b/include/linux/squashfs_fs.h @@ -0,0 +1,934 @@ +#ifndef SQUASHFS_FS +#define SQUASHFS_FS + +/* + * Squashfs + * + * Copyright (c) 2002, 2003, 2004, 2005, 2006, 2007 + * Phillip Lougher + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2, + * or (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + * + * squashfs_fs.h + */ + +#ifndef CONFIG_SQUASHFS_2_0_COMPATIBILITY +#define CONFIG_SQUASHFS_2_0_COMPATIBILITY +#endif + +#ifdef CONFIG_SQUASHFS_VMALLOC +#define SQUASHFS_ALLOC(a) vmalloc(a) +#define SQUASHFS_FREE(a) vfree(a) +#else +#define SQUASHFS_ALLOC(a) kmalloc(a, GFP_KERNEL) +#define SQUASHFS_FREE(a) kfree(a) +#endif +#define SQUASHFS_CACHED_FRAGMENTS CONFIG_SQUASHFS_FRAGMENT_CACHE_SIZE +#define SQUASHFS_MAJOR 3 +#define SQUASHFS_MINOR 0 +#define SQUASHFS_MAGIC 0x73717368 +#define SQUASHFS_MAGIC_SWAP 0x68737173 +#define SQUASHFS_START 0 + +/* size of metadata (inode and directory) blocks */ +#define SQUASHFS_METADATA_SIZE 8192 +#define SQUASHFS_METADATA_LOG 13 + +/* default size of data blocks */ +#define SQUASHFS_FILE_SIZE 65536 +#define SQUASHFS_FILE_LOG 16 + +#define SQUASHFS_FILE_MAX_SIZE 65536 + +/* Max number of uids and gids */ +#define SQUASHFS_UIDS 256 +#define SQUASHFS_GUIDS 255 + +/* Max length of filename (not 255) */ +#define SQUASHFS_NAME_LEN 256 + +#define SQUASHFS_INVALID ((long long) 0xffffffffffff) +#define SQUASHFS_INVALID_FRAG ((unsigned int) 0xffffffff) +#define SQUASHFS_INVALID_BLK ((long long) -1) +#define SQUASHFS_USED_BLK ((long long) -2) + +/* Filesystem flags */ +#define SQUASHFS_NOI 0 +#define SQUASHFS_NOD 1 +#define SQUASHFS_CHECK 2 +#define SQUASHFS_NOF 3 +#define SQUASHFS_NO_FRAG 4 +#define SQUASHFS_ALWAYS_FRAG 5 +#define SQUASHFS_DUPLICATE 6 +#define SQUASHFS_EXPORT 7 + +#define SQUASHFS_BIT(flag, bit) ((flag >> bit) & 1) + +#define SQUASHFS_UNCOMPRESSED_INODES(flags) SQUASHFS_BIT(flags, \ + SQUASHFS_NOI) + +#define SQUASHFS_UNCOMPRESSED_DATA(flags) SQUASHFS_BIT(flags, \ + SQUASHFS_NOD) + +#define SQUASHFS_UNCOMPRESSED_FRAGMENTS(flags) SQUASHFS_BIT(flags, \ + SQUASHFS_NOF) + +#define SQUASHFS_NO_FRAGMENTS(flags) SQUASHFS_BIT(flags, \ + SQUASHFS_NO_FRAG) + +#define SQUASHFS_ALWAYS_FRAGMENTS(flags) SQUASHFS_BIT(flags, \ + SQUASHFS_ALWAYS_FRAG) + +#define SQUASHFS_DUPLICATES(flags) SQUASHFS_BIT(flags, \ + SQUASHFS_DUPLICATE) + +#define SQUASHFS_EXPORTABLE(flags) SQUASHFS_BIT(flags, \ + SQUASHFS_EXPORT) + +#define SQUASHFS_CHECK_DATA(flags) SQUASHFS_BIT(flags, \ + SQUASHFS_CHECK) + +#define SQUASHFS_MKFLAGS(noi, nod, check_data, nof, no_frag, always_frag, \ + duplicate_checking, exortable) (noi | (nod << 1) | (check_data << 2) \ + | (nof << 3) | (no_frag << 4) | (always_frag << 5) | \ + (duplicate_checking << 6) | (exportable << 7)) + +/* Max number of types and file types */ +#define SQUASHFS_DIR_TYPE 1 +#define SQUASHFS_FILE_TYPE 2 +#define SQUASHFS_SYMLINK_TYPE 3 +#define SQUASHFS_BLKDEV_TYPE 4 +#define SQUASHFS_CHRDEV_TYPE 5 +#define SQUASHFS_FIFO_TYPE 6 +#define SQUASHFS_SOCKET_TYPE 7 +#define SQUASHFS_LDIR_TYPE 8 +#define SQUASHFS_LREG_TYPE 9 + +/* 1.0 filesystem type definitions */ +#define SQUASHFS_TYPES 5 +#define SQUASHFS_IPC_TYPE 0 + +/* Flag whether block is compressed or uncompressed, bit is set if block is + * uncompressed */ +#define SQUASHFS_COMPRESSED_BIT (1 << 15) + +#define SQUASHFS_COMPRESSED_SIZE(B) (((B) & ~SQUASHFS_COMPRESSED_BIT) ? \ + (B) & ~SQUASHFS_COMPRESSED_BIT : SQUASHFS_COMPRESSED_BIT) + +#define SQUASHFS_COMPRESSED(B) (!((B) & SQUASHFS_COMPRESSED_BIT)) + +#define SQUASHFS_COMPRESSED_BIT_BLOCK (1 << 24) + +#define SQUASHFS_COMPRESSED_SIZE_BLOCK(B) (((B) & \ + ~SQUASHFS_COMPRESSED_BIT_BLOCK) ? (B) & \ + ~SQUASHFS_COMPRESSED_BIT_BLOCK : SQUASHFS_COMPRESSED_BIT_BLOCK) + +#define SQUASHFS_COMPRESSED_BLOCK(B) (!((B) & SQUASHFS_COMPRESSED_BIT_BLOCK)) + +/* + * Inode number ops. Inodes consist of a compressed block number, and an + * uncompressed offset within that block + */ +#define SQUASHFS_INODE_BLK(a) ((unsigned int) ((a) >> 16)) + +#define SQUASHFS_INODE_OFFSET(a) ((unsigned int) ((a) & 0xffff)) + +#define SQUASHFS_MKINODE(A, B) ((squashfs_inode_t)(((squashfs_inode_t) (A)\ + << 16) + (B))) + +/* Compute 32 bit VFS inode number from squashfs inode number */ +#define SQUASHFS_MK_VFS_INODE(a, b) ((unsigned int) (((a) << 8) + \ + ((b) >> 2) + 1)) +/* XXX */ + +/* Translate between VFS mode and squashfs mode */ +#define SQUASHFS_MODE(a) ((a) & 0xfff) + +/* fragment and fragment table defines */ +#define SQUASHFS_FRAGMENT_BYTES(A) ((A) * sizeof(struct squashfs_fragment_entry)) + +#define SQUASHFS_FRAGMENT_INDEX(A) (SQUASHFS_FRAGMENT_BYTES(A) / \ + SQUASHFS_METADATA_SIZE) + +#define SQUASHFS_FRAGMENT_INDEX_OFFSET(A) (SQUASHFS_FRAGMENT_BYTES(A) % \ + SQUASHFS_METADATA_SIZE) + +#define SQUASHFS_FRAGMENT_INDEXES(A) ((SQUASHFS_FRAGMENT_BYTES(A) + \ + SQUASHFS_METADATA_SIZE - 1) / \ + SQUASHFS_METADATA_SIZE) + +#define SQUASHFS_FRAGMENT_INDEX_BYTES(A) (SQUASHFS_FRAGMENT_INDEXES(A) *\ + sizeof(long long)) + +/* inode lookup table defines */ +#define SQUASHFS_LOOKUP_BYTES(A) ((A) * sizeof(squashfs_inode_t)) + +#define SQUASHFS_LOOKUP_BLOCK(A) (SQUASHFS_LOOKUP_BYTES(A) / \ + SQUASHFS_METADATA_SIZE) + +#define SQUASHFS_LOOKUP_BLOCK_OFFSET(A) (SQUASHFS_LOOKUP_BYTES(A) % \ + SQUASHFS_METADATA_SIZE) + +#define SQUASHFS_LOOKUP_BLOCKS(A) ((SQUASHFS_LOOKUP_BYTES(A) + \ + SQUASHFS_METADATA_SIZE - 1) / \ + SQUASHFS_METADATA_SIZE) + +#define SQUASHFS_LOOKUP_BLOCK_BYTES(A) (SQUASHFS_LOOKUP_BLOCKS(A) *\ + sizeof(long long)) + +/* cached data constants for filesystem */ +#define SQUASHFS_CACHED_BLKS 8 + +#define SQUASHFS_MAX_FILE_SIZE_LOG 64 + +#define SQUASHFS_MAX_FILE_SIZE ((long long) 1 << \ + (SQUASHFS_MAX_FILE_SIZE_LOG - 2)) + +#define SQUASHFS_MARKER_BYTE 0xff + +/* meta index cache */ +#define SQUASHFS_META_INDEXES (SQUASHFS_METADATA_SIZE / sizeof(unsigned int)) +#define SQUASHFS_META_ENTRIES 31 +#define SQUASHFS_META_NUMBER 8 +#define SQUASHFS_SLOTS 4 + +struct meta_entry { + long long data_block; + unsigned int index_block; + unsigned short offset; + unsigned short pad; +}; + +struct meta_index { + unsigned int inode_number; + unsigned int offset; + unsigned short entries; + unsigned short skip; + unsigned short locked; + unsigned short pad; + struct meta_entry meta_entry[SQUASHFS_META_ENTRIES]; +}; + + +/* + * definitions for structures on disk + */ + +typedef long long squashfs_block_t; +typedef long long squashfs_inode_t; + +struct squashfs_super_block { + unsigned int s_magic; + unsigned int inodes; + unsigned int bytes_used_2; + unsigned int uid_start_2; + unsigned int guid_start_2; + unsigned int inode_table_start_2; + unsigned int directory_table_start_2; + unsigned int s_major:16; + unsigned int s_minor:16; + unsigned int block_size_1:16; + unsigned int block_log:16; + unsigned int flags:8; + unsigned int no_uids:8; + unsigned int no_guids:8; + unsigned int mkfs_time /* time of filesystem creation */; + squashfs_inode_t root_inode; + unsigned int block_size; + unsigned int fragments; + unsigned int fragment_table_start_2; + long long bytes_used; + long long uid_start; + long long guid_start; + long long inode_table_start; + long long directory_table_start; + long long fragment_table_start; + long long lookup_table_start; +} __attribute__ ((packed)); + +struct squashfs_dir_index { + unsigned int index; + unsigned int start_block; + unsigned char size; + unsigned char name[0]; +} __attribute__ ((packed)); + +#define SQUASHFS_BASE_INODE_HEADER \ + unsigned int inode_type:4; \ + unsigned int mode:12; \ + unsigned int uid:8; \ + unsigned int guid:8; \ + unsigned int mtime; \ + unsigned int inode_number; + +struct squashfs_base_inode_header { + SQUASHFS_BASE_INODE_HEADER; +} __attribute__ ((packed)); + +struct squashfs_ipc_inode_header { + SQUASHFS_BASE_INODE_HEADER; + unsigned int nlink; +} __attribute__ ((packed)); + +struct squashfs_dev_inode_header { + SQUASHFS_BASE_INODE_HEADER; + unsigned int nlink; + unsigned short rdev; +} __attribute__ ((packed)); + +struct squashfs_symlink_inode_header { + SQUASHFS_BASE_INODE_HEADER; + unsigned int nlink; + unsigned short symlink_size; + char symlink[0]; +} __attribute__ ((packed)); + +struct squashfs_reg_inode_header { + SQUASHFS_BASE_INODE_HEADER; + squashfs_block_t start_block; + unsigned int fragment; + unsigned int offset; + unsigned int file_size; + unsigned short block_list[0]; +} __attribute__ ((packed)); + +struct squashfs_lreg_inode_header { + SQUASHFS_BASE_INODE_HEADER; + unsigned int nlink; + squashfs_block_t start_block; + unsigned int fragment; + unsigned int offset; + long long file_size; + unsigned short block_list[0]; +} __attribute__ ((packed)); + +struct squashfs_dir_inode_header { + SQUASHFS_BASE_INODE_HEADER; + unsigned int nlink; + unsigned int file_size:19; + unsigned int offset:13; + unsigned int start_block; + unsigned int parent_inode; +} __attribute__ ((packed)); + +struct squashfs_ldir_inode_header { + SQUASHFS_BASE_INODE_HEADER; + unsigned int nlink; + unsigned int file_size:27; + unsigned int offset:13; + unsigned int start_block; + unsigned int i_count:16; + unsigned int parent_inode; + struct squashfs_dir_index index[0]; +} __attribute__ ((packed)); + +union squashfs_inode_header { + struct squashfs_base_inode_header base; + struct squashfs_dev_inode_header dev; + struct squashfs_symlink_inode_header symlink; + struct squashfs_reg_inode_header reg; + struct squashfs_lreg_inode_header lreg; + struct squashfs_dir_inode_header dir; + struct squashfs_ldir_inode_header ldir; + struct squashfs_ipc_inode_header ipc; +}; + +struct squashfs_dir_entry { + unsigned int offset:13; + unsigned int type:3; + unsigned int size:8; + int inode_number:16; + char name[0]; +} __attribute__ ((packed)); + +struct squashfs_dir_header { + unsigned int count:8; + unsigned int start_block; + unsigned int inode_number; +} __attribute__ ((packed)); + +struct squashfs_fragment_entry { + long long start_block; + unsigned int size; + unsigned int pending; +} __attribute__ ((packed)); + +extern int squashfs_uncompress_block(void *d, int dstlen, void *s, int srclen); +extern int squashfs_uncompress_init(void); +extern int squashfs_uncompress_exit(void); + +/* + * macros to convert each packed bitfield structure from little endian to big + * endian and vice versa. These are needed when creating or using a filesystem + * on a machine with different byte ordering to the target architecture. + * + */ + +#define SQUASHFS_SWAP_START \ + int bits;\ + int b_pos;\ + unsigned long long val;\ + unsigned char *s;\ + unsigned char *d; + +#define SQUASHFS_SWAP_SUPER_BLOCK(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_MEMSET(s, d, sizeof(struct squashfs_super_block));\ + SQUASHFS_SWAP((s)->s_magic, d, 0, 32);\ + SQUASHFS_SWAP((s)->inodes, d, 32, 32);\ + SQUASHFS_SWAP((s)->bytes_used_2, d, 64, 32);\ + SQUASHFS_SWAP((s)->uid_start_2, d, 96, 32);\ + SQUASHFS_SWAP((s)->guid_start_2, d, 128, 32);\ + SQUASHFS_SWAP((s)->inode_table_start_2, d, 160, 32);\ + SQUASHFS_SWAP((s)->directory_table_start_2, d, 192, 32);\ + SQUASHFS_SWAP((s)->s_major, d, 224, 16);\ + SQUASHFS_SWAP((s)->s_minor, d, 240, 16);\ + SQUASHFS_SWAP((s)->block_size_1, d, 256, 16);\ + SQUASHFS_SWAP((s)->block_log, d, 272, 16);\ + SQUASHFS_SWAP((s)->flags, d, 288, 8);\ + SQUASHFS_SWAP((s)->no_uids, d, 296, 8);\ + SQUASHFS_SWAP((s)->no_guids, d, 304, 8);\ + SQUASHFS_SWAP((s)->mkfs_time, d, 312, 32);\ + SQUASHFS_SWAP((s)->root_inode, d, 344, 64);\ + SQUASHFS_SWAP((s)->block_size, d, 408, 32);\ + SQUASHFS_SWAP((s)->fragments, d, 440, 32);\ + SQUASHFS_SWAP((s)->fragment_table_start_2, d, 472, 32);\ + SQUASHFS_SWAP((s)->bytes_used, d, 504, 64);\ + SQUASHFS_SWAP((s)->uid_start, d, 568, 64);\ + SQUASHFS_SWAP((s)->guid_start, d, 632, 64);\ + SQUASHFS_SWAP((s)->inode_table_start, d, 696, 64);\ + SQUASHFS_SWAP((s)->directory_table_start, d, 760, 64);\ + SQUASHFS_SWAP((s)->fragment_table_start, d, 824, 64);\ + SQUASHFS_SWAP((s)->lookup_table_start, d, 888, 64);\ +} + +#define SQUASHFS_SWAP_BASE_INODE_CORE(s, d, n)\ + SQUASHFS_MEMSET(s, d, n);\ + SQUASHFS_SWAP((s)->inode_type, d, 0, 4);\ + SQUASHFS_SWAP((s)->mode, d, 4, 12);\ + SQUASHFS_SWAP((s)->uid, d, 16, 8);\ + SQUASHFS_SWAP((s)->guid, d, 24, 8);\ + SQUASHFS_SWAP((s)->mtime, d, 32, 32);\ + SQUASHFS_SWAP((s)->inode_number, d, 64, 32); + +#define SQUASHFS_SWAP_BASE_INODE_HEADER(s, d, n) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE(s, d, n)\ +} + +#define SQUASHFS_SWAP_IPC_INODE_HEADER(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE(s, d, \ + sizeof(struct squashfs_ipc_inode_header))\ + SQUASHFS_SWAP((s)->nlink, d, 96, 32);\ +} + +#define SQUASHFS_SWAP_DEV_INODE_HEADER(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE(s, d, \ + sizeof(struct squashfs_dev_inode_header)); \ + SQUASHFS_SWAP((s)->nlink, d, 96, 32);\ + SQUASHFS_SWAP((s)->rdev, d, 128, 16);\ +} + +#define SQUASHFS_SWAP_SYMLINK_INODE_HEADER(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE(s, d, \ + sizeof(struct squashfs_symlink_inode_header));\ + SQUASHFS_SWAP((s)->nlink, d, 96, 32);\ + SQUASHFS_SWAP((s)->symlink_size, d, 128, 16);\ +} + +#define SQUASHFS_SWAP_REG_INODE_HEADER(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE(s, d, \ + sizeof(struct squashfs_reg_inode_header));\ + SQUASHFS_SWAP((s)->start_block, d, 96, 64);\ + SQUASHFS_SWAP((s)->fragment, d, 160, 32);\ + SQUASHFS_SWAP((s)->offset, d, 192, 32);\ + SQUASHFS_SWAP((s)->file_size, d, 224, 32);\ +} + +#define SQUASHFS_SWAP_LREG_INODE_HEADER(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE(s, d, \ + sizeof(struct squashfs_lreg_inode_header));\ + SQUASHFS_SWAP((s)->nlink, d, 96, 32);\ + SQUASHFS_SWAP((s)->start_block, d, 128, 64);\ + SQUASHFS_SWAP((s)->fragment, d, 192, 32);\ + SQUASHFS_SWAP((s)->offset, d, 224, 32);\ + SQUASHFS_SWAP((s)->file_size, d, 256, 64);\ +} + +#define SQUASHFS_SWAP_DIR_INODE_HEADER(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE(s, d, \ + sizeof(struct squashfs_dir_inode_header));\ + SQUASHFS_SWAP((s)->nlink, d, 96, 32);\ + SQUASHFS_SWAP((s)->file_size, d, 128, 19);\ + SQUASHFS_SWAP((s)->offset, d, 147, 13);\ + SQUASHFS_SWAP((s)->start_block, d, 160, 32);\ + SQUASHFS_SWAP((s)->parent_inode, d, 192, 32);\ +} + +#define SQUASHFS_SWAP_LDIR_INODE_HEADER(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE(s, d, \ + sizeof(struct squashfs_ldir_inode_header));\ + SQUASHFS_SWAP((s)->nlink, d, 96, 32);\ + SQUASHFS_SWAP((s)->file_size, d, 128, 27);\ + SQUASHFS_SWAP((s)->offset, d, 155, 13);\ + SQUASHFS_SWAP((s)->start_block, d, 168, 32);\ + SQUASHFS_SWAP((s)->i_count, d, 200, 16);\ + SQUASHFS_SWAP((s)->parent_inode, d, 216, 32);\ +} + +#define SQUASHFS_SWAP_DIR_INDEX(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_MEMSET(s, d, sizeof(struct squashfs_dir_index));\ + SQUASHFS_SWAP((s)->index, d, 0, 32);\ + SQUASHFS_SWAP((s)->start_block, d, 32, 32);\ + SQUASHFS_SWAP((s)->size, d, 64, 8);\ +} + +#define SQUASHFS_SWAP_DIR_HEADER(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_MEMSET(s, d, sizeof(struct squashfs_dir_header));\ + SQUASHFS_SWAP((s)->count, d, 0, 8);\ + SQUASHFS_SWAP((s)->start_block, d, 8, 32);\ + SQUASHFS_SWAP((s)->inode_number, d, 40, 32);\ +} + +#define SQUASHFS_SWAP_DIR_ENTRY(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_MEMSET(s, d, sizeof(struct squashfs_dir_entry));\ + SQUASHFS_SWAP((s)->offset, d, 0, 13);\ + SQUASHFS_SWAP((s)->type, d, 13, 3);\ + SQUASHFS_SWAP((s)->size, d, 16, 8);\ + SQUASHFS_SWAP((s)->inode_number, d, 24, 16);\ +} + +#define SQUASHFS_SWAP_FRAGMENT_ENTRY(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_MEMSET(s, d, sizeof(struct squashfs_fragment_entry));\ + SQUASHFS_SWAP((s)->start_block, d, 0, 64);\ + SQUASHFS_SWAP((s)->size, d, 64, 32);\ +} + +#define SQUASHFS_SWAP_INODE_T(s, d) SQUASHFS_SWAP_LONG_LONGS(s, d, 1) + +#define SQUASHFS_SWAP_SHORTS(s, d, n) {\ + int entry;\ + int bit_position;\ + SQUASHFS_SWAP_START\ + SQUASHFS_MEMSET(s, d, n * 2);\ + for(entry = 0, bit_position = 0; entry < n; entry++, bit_position += \ + 16)\ + SQUASHFS_SWAP(s[entry], d, bit_position, 16);\ +} + +#define SQUASHFS_SWAP_INTS(s, d, n) {\ + int entry;\ + int bit_position;\ + SQUASHFS_SWAP_START\ + SQUASHFS_MEMSET(s, d, n * 4);\ + for(entry = 0, bit_position = 0; entry < n; entry++, bit_position += \ + 32)\ + SQUASHFS_SWAP(s[entry], d, bit_position, 32);\ +} + +#define SQUASHFS_SWAP_LONG_LONGS(s, d, n) {\ + int entry;\ + int bit_position;\ + SQUASHFS_SWAP_START\ + SQUASHFS_MEMSET(s, d, n * 8);\ + for(entry = 0, bit_position = 0; entry < n; entry++, bit_position += \ + 64)\ + SQUASHFS_SWAP(s[entry], d, bit_position, 64);\ +} + +#define SQUASHFS_SWAP_DATA(s, d, n, bits) {\ + int entry;\ + int bit_position;\ + SQUASHFS_SWAP_START\ + SQUASHFS_MEMSET(s, d, n * bits / 8);\ + for(entry = 0, bit_position = 0; entry < n; entry++, bit_position += \ + bits)\ + SQUASHFS_SWAP(s[entry], d, bit_position, bits);\ +} + +#define SQUASHFS_SWAP_FRAGMENT_INDEXES(s, d, n) SQUASHFS_SWAP_LONG_LONGS(s, d, n) +#define SQUASHFS_SWAP_LOOKUP_BLOCKS(s, d, n) SQUASHFS_SWAP_LONG_LONGS(s, d, n) + +#ifdef CONFIG_SQUASHFS_1_0_COMPATIBILITY + +struct squashfs_base_inode_header_1 { + unsigned int inode_type:4; + unsigned int mode:12; /* protection */ + unsigned int uid:4; /* index into uid table */ + unsigned int guid:4; /* index into guid table */ +} __attribute__ ((packed)); + +struct squashfs_ipc_inode_header_1 { + unsigned int inode_type:4; + unsigned int mode:12; /* protection */ + unsigned int uid:4; /* index into uid table */ + unsigned int guid:4; /* index into guid table */ + unsigned int type:4; + unsigned int offset:4; +} __attribute__ ((packed)); + +struct squashfs_dev_inode_header_1 { + unsigned int inode_type:4; + unsigned int mode:12; /* protection */ + unsigned int uid:4; /* index into uid table */ + unsigned int guid:4; /* index into guid table */ + unsigned short rdev; +} __attribute__ ((packed)); + +struct squashfs_symlink_inode_header_1 { + unsigned int inode_type:4; + unsigned int mode:12; /* protection */ + unsigned int uid:4; /* index into uid table */ + unsigned int guid:4; /* index into guid table */ + unsigned short symlink_size; + char symlink[0]; +} __attribute__ ((packed)); + +struct squashfs_reg_inode_header_1 { + unsigned int inode_type:4; + unsigned int mode:12; /* protection */ + unsigned int uid:4; /* index into uid table */ + unsigned int guid:4; /* index into guid table */ + unsigned int mtime; + unsigned int start_block; + unsigned int file_size:32; + unsigned short block_list[0]; +} __attribute__ ((packed)); + +struct squashfs_dir_inode_header_1 { + unsigned int inode_type:4; + unsigned int mode:12; /* protection */ + unsigned int uid:4; /* index into uid table */ + unsigned int guid:4; /* index into guid table */ + unsigned int file_size:19; + unsigned int offset:13; + unsigned int mtime; + unsigned int start_block:24; +} __attribute__ ((packed)); + +#define SQUASHFS_SWAP_BASE_INODE_CORE_1(s, d, n) \ + SQUASHFS_MEMSET(s, d, n);\ + SQUASHFS_SWAP((s)->inode_type, d, 0, 4);\ + SQUASHFS_SWAP((s)->mode, d, 4, 12);\ + SQUASHFS_SWAP((s)->uid, d, 16, 4);\ + SQUASHFS_SWAP((s)->guid, d, 20, 4); + +#define SQUASHFS_SWAP_BASE_INODE_HEADER_1(s, d, n) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE_1(s, d, n)\ +} + +#define SQUASHFS_SWAP_IPC_INODE_HEADER_1(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE_1(s, d, \ + sizeof(struct squashfs_ipc_inode_header_1));\ + SQUASHFS_SWAP((s)->type, d, 24, 4);\ + SQUASHFS_SWAP((s)->offset, d, 28, 4);\ +} + +#define SQUASHFS_SWAP_DEV_INODE_HEADER_1(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE_1(s, d, \ + sizeof(struct squashfs_dev_inode_header_1));\ + SQUASHFS_SWAP((s)->rdev, d, 24, 16);\ +} + +#define SQUASHFS_SWAP_SYMLINK_INODE_HEADER_1(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE_1(s, d, \ + sizeof(struct squashfs_symlink_inode_header_1));\ + SQUASHFS_SWAP((s)->symlink_size, d, 24, 16);\ +} + +#define SQUASHFS_SWAP_REG_INODE_HEADER_1(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE_1(s, d, \ + sizeof(struct squashfs_reg_inode_header_1));\ + SQUASHFS_SWAP((s)->mtime, d, 24, 32);\ + SQUASHFS_SWAP((s)->start_block, d, 56, 32);\ + SQUASHFS_SWAP((s)->file_size, d, 88, 32);\ +} + +#define SQUASHFS_SWAP_DIR_INODE_HEADER_1(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE_1(s, d, \ + sizeof(struct squashfs_dir_inode_header_1));\ + SQUASHFS_SWAP((s)->file_size, d, 24, 19);\ + SQUASHFS_SWAP((s)->offset, d, 43, 13);\ + SQUASHFS_SWAP((s)->mtime, d, 56, 32);\ + SQUASHFS_SWAP((s)->start_block, d, 88, 24);\ +} + +#endif + +#ifdef CONFIG_SQUASHFS_2_0_COMPATIBILITY + +struct squashfs_dir_index_2 { + unsigned int index:27; + unsigned int start_block:29; + unsigned char size; + unsigned char name[0]; +} __attribute__ ((packed)); + +struct squashfs_base_inode_header_2 { + unsigned int inode_type:4; + unsigned int mode:12; /* protection */ + unsigned int uid:8; /* index into uid table */ + unsigned int guid:8; /* index into guid table */ +} __attribute__ ((packed)); + +struct squashfs_ipc_inode_header_2 { + unsigned int inode_type:4; + unsigned int mode:12; /* protection */ + unsigned int uid:8; /* index into uid table */ + unsigned int guid:8; /* index into guid table */ +} __attribute__ ((packed)); + +struct squashfs_dev_inode_header_2 { + unsigned int inode_type:4; + unsigned int mode:12; /* protection */ + unsigned int uid:8; /* index into uid table */ + unsigned int guid:8; /* index into guid table */ + unsigned short rdev; +} __attribute__ ((packed)); + +struct squashfs_symlink_inode_header_2 { + unsigned int inode_type:4; + unsigned int mode:12; /* protection */ + unsigned int uid:8; /* index into uid table */ + unsigned int guid:8; /* index into guid table */ + unsigned short symlink_size; + char symlink[0]; +} __attribute__ ((packed)); + +struct squashfs_reg_inode_header_2 { + unsigned int inode_type:4; + unsigned int mode:12; /* protection */ + unsigned int uid:8; /* index into uid table */ + unsigned int guid:8; /* index into guid table */ + unsigned int mtime; + unsigned int start_block; + unsigned int fragment; + unsigned int offset; + unsigned int file_size:32; + unsigned short block_list[0]; +} __attribute__ ((packed)); + +struct squashfs_dir_inode_header_2 { + unsigned int inode_type:4; + unsigned int mode:12; /* protection */ + unsigned int uid:8; /* index into uid table */ + unsigned int guid:8; /* index into guid table */ + unsigned int file_size:19; + unsigned int offset:13; + unsigned int mtime; + unsigned int start_block:24; +} __attribute__ ((packed)); + +struct squashfs_ldir_inode_header_2 { + unsigned int inode_type:4; + unsigned int mode:12; /* protection */ + unsigned int uid:8; /* index into uid table */ + unsigned int guid:8; /* index into guid table */ + unsigned int file_size:27; + unsigned int offset:13; + unsigned int mtime; + unsigned int start_block:24; + unsigned int i_count:16; + struct squashfs_dir_index_2 index[0]; +} __attribute__ ((packed)); + +union squashfs_inode_header_2 { + struct squashfs_base_inode_header_2 base; + struct squashfs_dev_inode_header_2 dev; + struct squashfs_symlink_inode_header_2 symlink; + struct squashfs_reg_inode_header_2 reg; + struct squashfs_dir_inode_header_2 dir; + struct squashfs_ldir_inode_header_2 ldir; + struct squashfs_ipc_inode_header_2 ipc; +}; + +struct squashfs_dir_header_2 { + unsigned int count:8; + unsigned int start_block:24; +} __attribute__ ((packed)); + +struct squashfs_dir_entry_2 { + unsigned int offset:13; + unsigned int type:3; + unsigned int size:8; + char name[0]; +} __attribute__ ((packed)); + +struct squashfs_fragment_entry_2 { + unsigned int start_block; + unsigned int size; +} __attribute__ ((packed)); + +#define SQUASHFS_SWAP_BASE_INODE_CORE_2(s, d, n)\ + SQUASHFS_MEMSET(s, d, n);\ + SQUASHFS_SWAP((s)->inode_type, d, 0, 4);\ + SQUASHFS_SWAP((s)->mode, d, 4, 12);\ + SQUASHFS_SWAP((s)->uid, d, 16, 8);\ + SQUASHFS_SWAP((s)->guid, d, 24, 8);\ + +#define SQUASHFS_SWAP_BASE_INODE_HEADER_2(s, d, n) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE_2(s, d, n)\ +} + +#define SQUASHFS_SWAP_IPC_INODE_HEADER_2(s, d) \ + SQUASHFS_SWAP_BASE_INODE_HEADER_2(s, d, sizeof(struct squashfs_ipc_inode_header_2)) + +#define SQUASHFS_SWAP_DEV_INODE_HEADER_2(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE_2(s, d, \ + sizeof(struct squashfs_dev_inode_header_2)); \ + SQUASHFS_SWAP((s)->rdev, d, 32, 16);\ +} + +#define SQUASHFS_SWAP_SYMLINK_INODE_HEADER_2(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE_2(s, d, \ + sizeof(struct squashfs_symlink_inode_header_2));\ + SQUASHFS_SWAP((s)->symlink_size, d, 32, 16);\ +} + +#define SQUASHFS_SWAP_REG_INODE_HEADER_2(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE_2(s, d, \ + sizeof(struct squashfs_reg_inode_header_2));\ + SQUASHFS_SWAP((s)->mtime, d, 32, 32);\ + SQUASHFS_SWAP((s)->start_block, d, 64, 32);\ + SQUASHFS_SWAP((s)->fragment, d, 96, 32);\ + SQUASHFS_SWAP((s)->offset, d, 128, 32);\ + SQUASHFS_SWAP((s)->file_size, d, 160, 32);\ +} + +#define SQUASHFS_SWAP_DIR_INODE_HEADER_2(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE_2(s, d, \ + sizeof(struct squashfs_dir_inode_header_2));\ + SQUASHFS_SWAP((s)->file_size, d, 32, 19);\ + SQUASHFS_SWAP((s)->offset, d, 51, 13);\ + SQUASHFS_SWAP((s)->mtime, d, 64, 32);\ + SQUASHFS_SWAP((s)->start_block, d, 96, 24);\ +} + +#define SQUASHFS_SWAP_LDIR_INODE_HEADER_2(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_SWAP_BASE_INODE_CORE_2(s, d, \ + sizeof(struct squashfs_ldir_inode_header_2));\ + SQUASHFS_SWAP((s)->file_size, d, 32, 27);\ + SQUASHFS_SWAP((s)->offset, d, 59, 13);\ + SQUASHFS_SWAP((s)->mtime, d, 72, 32);\ + SQUASHFS_SWAP((s)->start_block, d, 104, 24);\ + SQUASHFS_SWAP((s)->i_count, d, 128, 16);\ +} + +#define SQUASHFS_SWAP_DIR_INDEX_2(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_MEMSET(s, d, sizeof(struct squashfs_dir_index_2));\ + SQUASHFS_SWAP((s)->index, d, 0, 27);\ + SQUASHFS_SWAP((s)->start_block, d, 27, 29);\ + SQUASHFS_SWAP((s)->size, d, 56, 8);\ +} +#define SQUASHFS_SWAP_DIR_HEADER_2(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_MEMSET(s, d, sizeof(struct squashfs_dir_header_2));\ + SQUASHFS_SWAP((s)->count, d, 0, 8);\ + SQUASHFS_SWAP((s)->start_block, d, 8, 24);\ +} + +#define SQUASHFS_SWAP_DIR_ENTRY_2(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_MEMSET(s, d, sizeof(struct squashfs_dir_entry_2));\ + SQUASHFS_SWAP((s)->offset, d, 0, 13);\ + SQUASHFS_SWAP((s)->type, d, 13, 3);\ + SQUASHFS_SWAP((s)->size, d, 16, 8);\ +} + +#define SQUASHFS_SWAP_FRAGMENT_ENTRY_2(s, d) {\ + SQUASHFS_SWAP_START\ + SQUASHFS_MEMSET(s, d, sizeof(struct squashfs_fragment_entry_2));\ + SQUASHFS_SWAP((s)->start_block, d, 0, 32);\ + SQUASHFS_SWAP((s)->size, d, 32, 32);\ +} + +#define SQUASHFS_SWAP_FRAGMENT_INDEXES_2(s, d, n) SQUASHFS_SWAP_INTS(s, d, n) + +/* fragment and fragment table defines */ +#define SQUASHFS_FRAGMENT_BYTES_2(A) (A * sizeof(struct squashfs_fragment_entry_2)) + +#define SQUASHFS_FRAGMENT_INDEX_2(A) (SQUASHFS_FRAGMENT_BYTES_2(A) / \ + SQUASHFS_METADATA_SIZE) + +#define SQUASHFS_FRAGMENT_INDEX_OFFSET_2(A) (SQUASHFS_FRAGMENT_BYTES_2(A) % \ + SQUASHFS_METADATA_SIZE) + +#define SQUASHFS_FRAGMENT_INDEXES_2(A) ((SQUASHFS_FRAGMENT_BYTES_2(A) + \ + SQUASHFS_METADATA_SIZE - 1) / \ + SQUASHFS_METADATA_SIZE) + +#define SQUASHFS_FRAGMENT_INDEX_BYTES_2(A) (SQUASHFS_FRAGMENT_INDEXES_2(A) *\ + sizeof(int)) + +#endif + +#ifdef __KERNEL__ + +/* + * macros used to swap each structure entry, taking into account + * bitfields and different bitfield placing conventions on differing + * architectures + */ + +#include + +#ifdef __BIG_ENDIAN + /* convert from little endian to big endian */ +#define SQUASHFS_SWAP(value, p, pos, tbits) _SQUASHFS_SWAP(value, p, pos, \ + tbits, b_pos) +#else + /* convert from big endian to little endian */ +#define SQUASHFS_SWAP(value, p, pos, tbits) _SQUASHFS_SWAP(value, p, pos, \ + tbits, 64 - tbits - b_pos) +#endif + +#define _SQUASHFS_SWAP(value, p, pos, tbits, SHIFT) {\ + b_pos = pos % 8;\ + val = 0;\ + s = (unsigned char *)p + (pos / 8);\ + d = ((unsigned char *) &val) + 7;\ + for(bits = 0; bits < (tbits + b_pos); bits += 8) \ + *d-- = *s++;\ + value = (val >> (SHIFT))/* & ((1 << tbits) - 1)*/;\ +} + +#define SQUASHFS_MEMSET(s, d, n) memset(s, 0, n); + +#endif +#endif diff --git a/include/linux/squashfs_fs_i.h b/include/linux/squashfs_fs_i.h new file mode 100644 index 0000000..798891a --- /dev/null +++ b/include/linux/squashfs_fs_i.h @@ -0,0 +1,45 @@ +#ifndef SQUASHFS_FS_I +#define SQUASHFS_FS_I +/* + * Squashfs + * + * Copyright (c) 2002, 2003, 2004, 2005, 2006, 2007 + * Phillip Lougher + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2, + * or (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + * + * squashfs_fs_i.h + */ + +struct squashfs_inode_info { + long long start_block; + unsigned int offset; + union { + struct { + long long fragment_start_block; + unsigned int fragment_size; + unsigned int fragment_offset; + long long block_list_start; + } s1; + struct { + long long directory_index_start; + unsigned int directory_index_offset; + unsigned int directory_index_count; + unsigned int parent_inode; + } s2; + } u; + struct inode vfs_inode; +}; +#endif diff --git a/include/linux/squashfs_fs_sb.h b/include/linux/squashfs_fs_sb.h new file mode 100644 index 0000000..8f3bf99 --- /dev/null +++ b/include/linux/squashfs_fs_sb.h @@ -0,0 +1,74 @@ +#ifndef SQUASHFS_FS_SB +#define SQUASHFS_FS_SB +/* + * Squashfs + * + * Copyright (c) 2002, 2003, 2004, 2005, 2006, 2007 + * Phillip Lougher + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2, + * or (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + * + * squashfs_fs_sb.h + */ + +#include + +struct squashfs_cache { + long long block; + int length; + long long next_index; + char *data; +}; + +struct squashfs_fragment_cache { + long long block; + int length; + unsigned int locked; + char *data; +}; + +struct squashfs_sb_info { + struct squashfs_super_block sblk; + int devblksize; + int devblksize_log2; + int swap; + struct squashfs_cache *block_cache; + struct squashfs_fragment_cache *fragment; + int next_cache; + int next_fragment; + int next_meta_index; + unsigned int *uid; + unsigned int *guid; + long long *fragment_index; + unsigned int *fragment_index_2; + char *read_page; + struct mutex read_data_mutex; + struct mutex read_page_mutex; + struct mutex block_cache_mutex; + struct mutex fragment_mutex; + struct mutex meta_index_mutex; + wait_queue_head_t waitq; + wait_queue_head_t fragment_wait_queue; + struct meta_index *meta_index; + z_stream stream; + long long *inode_lookup_table; + int (*read_inode)(struct inode *i, squashfs_inode_t \ + inode); + long long (*read_blocklist)(struct inode *inode, int \ + index, int readahead_blks, char *block_list, \ + unsigned short **block_p, unsigned int *bsize); + int (*read_fragment_index_table)(struct super_block *s); +}; +#endif diff --git a/init/Kconfig b/init/Kconfig index b170aa1..bcfc3b4 100644 --- a/init/Kconfig +++ b/init/Kconfig @@ -244,23 +244,21 @@ config AUDITSYSCALL ensure that INOTIFY is configured. config IKCONFIG - tristate "Kernel .config support" + tristate "Kernel .miniconfig support" ---help--- - This option enables the complete Linux kernel ".config" file + This option enables the mini Linux kernel ".miniconfig" file contents to be saved in the kernel. It provides documentation of which kernel options are used in a running kernel or in an - on-disk kernel. This information can be extracted from the kernel - image file with the script scripts/extract-ikconfig and used as - input to rebuild the current kernel or to build another kernel. - It can also be extracted from a running kernel by reading - /proc/config.gz if enabled (below). + on-disk kernel. + It can be extracted from a running kernel by reading + /proc/miniconfig.gz if enabled (below). config IKCONFIG_PROC - bool "Enable access to .config through /proc/config.gz" + bool "Enable access to .miniconfig through /proc/miniconfig.gz" depends on IKCONFIG && PROC_FS ---help--- This option enables access to the kernel configuration file - through /proc/config.gz. + through /proc/miniconfig.gz. config CPUSETS bool "Cpuset support" diff --git a/init/LzmaDecode.c b/init/LzmaDecode.c new file mode 100644 index 0000000..21bf40b --- /dev/null +++ b/init/LzmaDecode.c @@ -0,0 +1,588 @@ +/* + LzmaDecode.c + LZMA Decoder (optimized for Speed version) + + LZMA SDK 4.22 Copyright (c) 1999-2005 Igor Pavlov (2005-06-10) + http://www.7-zip.org/ + + LZMA SDK is licensed under two licenses: + 1) GNU Lesser General Public License (GNU LGPL) + 2) Common Public License (CPL) + It means that you can select one of these two licenses and + follow rules of that license. + + SPECIAL EXCEPTION: + Igor Pavlov, as the author of this Code, expressly permits you to + statically or dynamically link your Code (or bind by name) to the + interfaces of this file without subjecting your linked Code to the + terms of the CPL or GNU LGPL. Any modifications or additions + to this file, however, are subject to the LGPL or CPL terms. +*/ + +#include "LzmaDecode.h" + +#ifndef Byte +#define Byte unsigned char +#endif + +#define kNumTopBits 24 +#define kTopValue ((UInt32)1 << kNumTopBits) + +#define kNumBitModelTotalBits 11 +#define kBitModelTotal (1 << kNumBitModelTotalBits) +#define kNumMoveBits 5 + +#define RC_READ_BYTE (*Buffer++) + +#define RC_INIT2 Code = 0; Range = 0xFFFFFFFF; \ + { int i; for(i = 0; i < 5; i++) { RC_TEST; Code = (Code << 8) | RC_READ_BYTE; }} + +#ifdef _LZMA_IN_CB + +#define RC_TEST { if (Buffer == BufferLim) \ + { SizeT size; int result = InCallback->Read(InCallback, &Buffer, &size); if (result != LZMA_RESULT_OK) return result; \ + BufferLim = Buffer + size; if (size == 0) return LZMA_RESULT_DATA_ERROR; }} + +#define RC_INIT Buffer = BufferLim = 0; RC_INIT2 + +#else + +#define RC_TEST { if (Buffer == BufferLim) return LZMA_RESULT_DATA_ERROR; } + +#define RC_INIT(buffer, bufferSize) Buffer = buffer; BufferLim = buffer + bufferSize; RC_INIT2 + +#endif + +#define RC_NORMALIZE if (Range < kTopValue) { RC_TEST; Range <<= 8; Code = (Code << 8) | RC_READ_BYTE; } + +#define IfBit0(p) RC_NORMALIZE; bound = (Range >> kNumBitModelTotalBits) * *(p); if (Code < bound) +#define UpdateBit0(p) Range = bound; *(p) += (kBitModelTotal - *(p)) >> kNumMoveBits; +#define UpdateBit1(p) Range -= bound; Code -= bound; *(p) -= (*(p)) >> kNumMoveBits; + +#define RC_GET_BIT2(p, mi, A0, A1) IfBit0(p) \ + { UpdateBit0(p); mi <<= 1; A0; } else \ + { UpdateBit1(p); mi = (mi + mi) + 1; A1; } + +#define RC_GET_BIT(p, mi) RC_GET_BIT2(p, mi, ; , ;) + +#define RangeDecoderBitTreeDecode(probs, numLevels, res) \ + { int i = numLevels; res = 1; \ + do { CProb *p = probs + res; RC_GET_BIT(p, res) } while(--i != 0); \ + res -= (1 << numLevels); } + + +#define kNumPosBitsMax 4 +#define kNumPosStatesMax (1 << kNumPosBitsMax) + +#define kLenNumLowBits 3 +#define kLenNumLowSymbols (1 << kLenNumLowBits) +#define kLenNumMidBits 3 +#define kLenNumMidSymbols (1 << kLenNumMidBits) +#define kLenNumHighBits 8 +#define kLenNumHighSymbols (1 << kLenNumHighBits) + +#define LenChoice 0 +#define LenChoice2 (LenChoice + 1) +#define LenLow (LenChoice2 + 1) +#define LenMid (LenLow + (kNumPosStatesMax << kLenNumLowBits)) +#define LenHigh (LenMid + (kNumPosStatesMax << kLenNumMidBits)) +#define kNumLenProbs (LenHigh + kLenNumHighSymbols) + + +#define kNumStates 12 +#define kNumLitStates 7 + +#define kStartPosModelIndex 4 +#define kEndPosModelIndex 14 +#define kNumFullDistances (1 << (kEndPosModelIndex >> 1)) + +#define kNumPosSlotBits 6 +#define kNumLenToPosStates 4 + +#define kNumAlignBits 4 +#define kAlignTableSize (1 << kNumAlignBits) + +#define kMatchMinLen 2 + +#define IsMatch 0 +#define IsRep (IsMatch + (kNumStates << kNumPosBitsMax)) +#define IsRepG0 (IsRep + kNumStates) +#define IsRepG1 (IsRepG0 + kNumStates) +#define IsRepG2 (IsRepG1 + kNumStates) +#define IsRep0Long (IsRepG2 + kNumStates) +#define PosSlot (IsRep0Long + (kNumStates << kNumPosBitsMax)) +#define SpecPos (PosSlot + (kNumLenToPosStates << kNumPosSlotBits)) +#define Align (SpecPos + kNumFullDistances - kEndPosModelIndex) +#define LenCoder (Align + kAlignTableSize) +#define RepLenCoder (LenCoder + kNumLenProbs) +#define Literal (RepLenCoder + kNumLenProbs) + +#if Literal != LZMA_BASE_SIZE +StopCompilingDueBUG +#endif + +int LzmaDecodeProperties(CLzmaProperties *propsRes, const unsigned char *propsData, int size) +{ + unsigned char prop0; + if (size < LZMA_PROPERTIES_SIZE) + return LZMA_RESULT_DATA_ERROR; + prop0 = propsData[0]; + if (prop0 >= (9 * 5 * 5)) + return LZMA_RESULT_DATA_ERROR; + { + for (propsRes->pb = 0; prop0 >= (9 * 5); propsRes->pb++, prop0 -= (9 * 5)); + for (propsRes->lp = 0; prop0 >= 9; propsRes->lp++, prop0 -= 9); + propsRes->lc = prop0; + /* + unsigned char remainder = (unsigned char)(prop0 / 9); + propsRes->lc = prop0 % 9; + propsRes->pb = remainder / 5; + propsRes->lp = remainder % 5; + */ + } + + #ifdef _LZMA_OUT_READ + { + int i; + propsRes->DictionarySize = 0; + for (i = 0; i < 4; i++) + propsRes->DictionarySize += (UInt32)(propsData[1 + i]) << (i * 8); + if (propsRes->DictionarySize == 0) + propsRes->DictionarySize = 1; + } + #endif + return LZMA_RESULT_OK; +} + +#define kLzmaStreamWasFinishedId (-1) + +int LzmaDecode(CLzmaDecoderState *vs, + #ifdef _LZMA_IN_CB + ILzmaInCallback *InCallback, + #else + const unsigned char *inStream, SizeT inSize, SizeT *inSizeProcessed, + #endif + unsigned char *outStream, SizeT outSize, SizeT *outSizeProcessed) +{ + CProb *p = vs->Probs; + SizeT nowPos = 0; + Byte previousByte = 0; + UInt32 posStateMask = (1 << (vs->Properties.pb)) - 1; + UInt32 literalPosMask = (1 << (vs->Properties.lp)) - 1; + int lc = vs->Properties.lc; + + #ifdef _LZMA_OUT_READ + + UInt32 Range = vs->Range; + UInt32 Code = vs->Code; + #ifdef _LZMA_IN_CB + const Byte *Buffer = vs->Buffer; + const Byte *BufferLim = vs->BufferLim; + #else + const Byte *Buffer = inStream; + const Byte *BufferLim = inStream + inSize; + #endif + int state = vs->State; + UInt32 rep0 = vs->Reps[0], rep1 = vs->Reps[1], rep2 = vs->Reps[2], rep3 = vs->Reps[3]; + int len = vs->RemainLen; + UInt32 globalPos = vs->GlobalPos; + UInt32 distanceLimit = vs->DistanceLimit; + + Byte *dictionary = vs->Dictionary; + UInt32 dictionarySize = vs->Properties.DictionarySize; + UInt32 dictionaryPos = vs->DictionaryPos; + + Byte tempDictionary[4]; + + #ifndef _LZMA_IN_CB + *inSizeProcessed = 0; + #endif + *outSizeProcessed = 0; + if (len == kLzmaStreamWasFinishedId) + return LZMA_RESULT_OK; + + if (dictionarySize == 0) + { + dictionary = tempDictionary; + dictionarySize = 1; + tempDictionary[0] = vs->TempDictionary[0]; + } + + if (len == kLzmaNeedInitId) + { + { + UInt32 numProbs = Literal + ((UInt32)LZMA_LIT_SIZE << (lc + vs->Properties.lp)); + UInt32 i; + for (i = 0; i < numProbs; i++) + p[i] = kBitModelTotal >> 1; + rep0 = rep1 = rep2 = rep3 = 1; + state = 0; + globalPos = 0; + distanceLimit = 0; + dictionaryPos = 0; + dictionary[dictionarySize - 1] = 0; + #ifdef _LZMA_IN_CB + RC_INIT; + #else + RC_INIT(inStream, inSize); + #endif + } + len = 0; + } + while(len != 0 && nowPos < outSize) + { + UInt32 pos = dictionaryPos - rep0; + if (pos >= dictionarySize) + pos += dictionarySize; + outStream[nowPos++] = dictionary[dictionaryPos] = dictionary[pos]; + if (++dictionaryPos == dictionarySize) + dictionaryPos = 0; + len--; + } + if (dictionaryPos == 0) + previousByte = dictionary[dictionarySize - 1]; + else + previousByte = dictionary[dictionaryPos - 1]; + + #else /* if !_LZMA_OUT_READ */ + + int state = 0; + UInt32 rep0 = 1, rep1 = 1, rep2 = 1, rep3 = 1; + int len = 0; + const Byte *Buffer; + const Byte *BufferLim; + UInt32 Range; + UInt32 Code; + + #ifndef _LZMA_IN_CB + *inSizeProcessed = 0; + #endif + *outSizeProcessed = 0; + + { + UInt32 i; + UInt32 numProbs = Literal + ((UInt32)LZMA_LIT_SIZE << (lc + vs->Properties.lp)); + for (i = 0; i < numProbs; i++) + p[i] = kBitModelTotal >> 1; + } + + #ifdef _LZMA_IN_CB + RC_INIT; + #else + RC_INIT(inStream, inSize); + #endif + + #endif /* _LZMA_OUT_READ */ + + while(nowPos < outSize) + { + CProb *prob; + UInt32 bound; + int posState = (int)( + (nowPos + #ifdef _LZMA_OUT_READ + + globalPos + #endif + ) + & posStateMask); + + prob = p + IsMatch + (state << kNumPosBitsMax) + posState; + IfBit0(prob) + { + int symbol = 1; + UpdateBit0(prob) + prob = p + Literal + (LZMA_LIT_SIZE * + ((( + (nowPos + #ifdef _LZMA_OUT_READ + + globalPos + #endif + ) + & literalPosMask) << lc) + (previousByte >> (8 - lc)))); + + if (state >= kNumLitStates) + { + int matchByte; + #ifdef _LZMA_OUT_READ + UInt32 pos = dictionaryPos - rep0; + if (pos >= dictionarySize) + pos += dictionarySize; + matchByte = dictionary[pos]; + #else + matchByte = outStream[nowPos - rep0]; + #endif + do + { + int bit; + CProb *probLit; + matchByte <<= 1; + bit = (matchByte & 0x100); + probLit = prob + 0x100 + bit + symbol; + RC_GET_BIT2(probLit, symbol, if (bit != 0) break, if (bit == 0) break) + } + while (symbol < 0x100); + } + while (symbol < 0x100) + { + CProb *probLit = prob + symbol; + RC_GET_BIT(probLit, symbol) + } + previousByte = (Byte)symbol; + + outStream[nowPos++] = previousByte; + #ifdef _LZMA_OUT_READ + if (distanceLimit < dictionarySize) + distanceLimit++; + + dictionary[dictionaryPos] = previousByte; + if (++dictionaryPos == dictionarySize) + dictionaryPos = 0; + #endif + if (state < 4) state = 0; + else if (state < 10) state -= 3; + else state -= 6; + } + else + { + UpdateBit1(prob); + prob = p + IsRep + state; + IfBit0(prob) + { + UpdateBit0(prob); + rep3 = rep2; + rep2 = rep1; + rep1 = rep0; + state = state < kNumLitStates ? 0 : 3; + prob = p + LenCoder; + } + else + { + UpdateBit1(prob); + prob = p + IsRepG0 + state; + IfBit0(prob) + { + UpdateBit0(prob); + prob = p + IsRep0Long + (state << kNumPosBitsMax) + posState; + IfBit0(prob) + { + #ifdef _LZMA_OUT_READ + UInt32 pos; + #endif + UpdateBit0(prob); + + #ifdef _LZMA_OUT_READ + if (distanceLimit == 0) + #else + if (nowPos == 0) + #endif + return LZMA_RESULT_DATA_ERROR; + + state = state < kNumLitStates ? 9 : 11; + #ifdef _LZMA_OUT_READ + pos = dictionaryPos - rep0; + if (pos >= dictionarySize) + pos += dictionarySize; + previousByte = dictionary[pos]; + dictionary[dictionaryPos] = previousByte; + if (++dictionaryPos == dictionarySize) + dictionaryPos = 0; + #else + previousByte = outStream[nowPos - rep0]; + #endif + outStream[nowPos++] = previousByte; + #ifdef _LZMA_OUT_READ + if (distanceLimit < dictionarySize) + distanceLimit++; + #endif + + continue; + } + else + { + UpdateBit1(prob); + } + } + else + { + UInt32 distance; + UpdateBit1(prob); + prob = p + IsRepG1 + state; + IfBit0(prob) + { + UpdateBit0(prob); + distance = rep1; + } + else + { + UpdateBit1(prob); + prob = p + IsRepG2 + state; + IfBit0(prob) + { + UpdateBit0(prob); + distance = rep2; + } + else + { + UpdateBit1(prob); + distance = rep3; + rep3 = rep2; + } + rep2 = rep1; + } + rep1 = rep0; + rep0 = distance; + } + state = state < kNumLitStates ? 8 : 11; + prob = p + RepLenCoder; + } + { + int numBits, offset; + CProb *probLen = prob + LenChoice; + IfBit0(probLen) + { + UpdateBit0(probLen); + probLen = prob + LenLow + (posState << kLenNumLowBits); + offset = 0; + numBits = kLenNumLowBits; + } + else + { + UpdateBit1(probLen); + probLen = prob + LenChoice2; + IfBit0(probLen) + { + UpdateBit0(probLen); + probLen = prob + LenMid + (posState << kLenNumMidBits); + offset = kLenNumLowSymbols; + numBits = kLenNumMidBits; + } + else + { + UpdateBit1(probLen); + probLen = prob + LenHigh; + offset = kLenNumLowSymbols + kLenNumMidSymbols; + numBits = kLenNumHighBits; + } + } + RangeDecoderBitTreeDecode(probLen, numBits, len); + len += offset; + } + + if (state < 4) + { + int posSlot; + state += kNumLitStates; + prob = p + PosSlot + + ((len < kNumLenToPosStates ? len : kNumLenToPosStates - 1) << + kNumPosSlotBits); + RangeDecoderBitTreeDecode(prob, kNumPosSlotBits, posSlot); + if (posSlot >= kStartPosModelIndex) + { + int numDirectBits = ((posSlot >> 1) - 1); + rep0 = (2 | ((UInt32)posSlot & 1)); + if (posSlot < kEndPosModelIndex) + { + rep0 <<= numDirectBits; + prob = p + SpecPos + rep0 - posSlot - 1; + } + else + { + numDirectBits -= kNumAlignBits; + do + { + RC_NORMALIZE + Range >>= 1; + rep0 <<= 1; + if (Code >= Range) + { + Code -= Range; + rep0 |= 1; + } + } + while (--numDirectBits != 0); + prob = p + Align; + rep0 <<= kNumAlignBits; + numDirectBits = kNumAlignBits; + } + { + int i = 1; + int mi = 1; + do + { + CProb *prob3 = prob + mi; + RC_GET_BIT2(prob3, mi, ; , rep0 |= i); + i <<= 1; + } + while(--numDirectBits != 0); + } + } + else + rep0 = posSlot; + if (++rep0 == (UInt32)(0)) + { + /* it's for stream version */ + len = kLzmaStreamWasFinishedId; + break; + } + } + + len += kMatchMinLen; + #ifdef _LZMA_OUT_READ + if (rep0 > distanceLimit) + #else + if (rep0 > nowPos) + #endif + return LZMA_RESULT_DATA_ERROR; + + #ifdef _LZMA_OUT_READ + if (dictionarySize - distanceLimit > (UInt32)len) + distanceLimit += len; + else + distanceLimit = dictionarySize; + #endif + + do + { + #ifdef _LZMA_OUT_READ + UInt32 pos = dictionaryPos - rep0; + if (pos >= dictionarySize) + pos += dictionarySize; + previousByte = dictionary[pos]; + dictionary[dictionaryPos] = previousByte; + if (++dictionaryPos == dictionarySize) + dictionaryPos = 0; + #else + previousByte = outStream[nowPos - rep0]; + #endif + len--; + outStream[nowPos++] = previousByte; + } + while(len != 0 && nowPos < outSize); + } + } + RC_NORMALIZE; + + #ifdef _LZMA_OUT_READ + vs->Range = Range; + vs->Code = Code; + vs->DictionaryPos = dictionaryPos; + vs->GlobalPos = globalPos + (UInt32)nowPos; + vs->DistanceLimit = distanceLimit; + vs->Reps[0] = rep0; + vs->Reps[1] = rep1; + vs->Reps[2] = rep2; + vs->Reps[3] = rep3; + vs->State = state; + vs->RemainLen = len; + vs->TempDictionary[0] = tempDictionary[0]; + #endif + + #ifdef _LZMA_IN_CB + vs->Buffer = Buffer; + vs->BufferLim = BufferLim; + #else + *inSizeProcessed = (SizeT)(Buffer - inStream); + #endif + *outSizeProcessed = nowPos; + return LZMA_RESULT_OK; +} diff --git a/init/LzmaDecode.h b/init/LzmaDecode.h new file mode 100644 index 0000000..213062a --- /dev/null +++ b/init/LzmaDecode.h @@ -0,0 +1,131 @@ +/* + LzmaDecode.h + LZMA Decoder interface + + LZMA SDK 4.21 Copyright (c) 1999-2005 Igor Pavlov (2005-06-08) + http://www.7-zip.org/ + + LZMA SDK is licensed under two licenses: + 1) GNU Lesser General Public License (GNU LGPL) + 2) Common Public License (CPL) + It means that you can select one of these two licenses and + follow rules of that license. + + SPECIAL EXCEPTION: + Igor Pavlov, as the author of this code, expressly permits you to + statically or dynamically link your code (or bind by name) to the + interfaces of this file without subjecting your linked code to the + terms of the CPL or GNU LGPL. Any modifications or additions + to this file, however, are subject to the LGPL or CPL terms. +*/ + +#ifndef __LZMADECODE_H +#define __LZMADECODE_H + +/* #define _LZMA_IN_CB */ +/* Use callback for input data */ + +/* #define _LZMA_OUT_READ */ +/* Use read function for output data */ + +/* #define _LZMA_PROB32 */ +/* It can increase speed on some 32-bit CPUs, + but memory usage will be doubled in that case */ + +/* #define _LZMA_LOC_OPT */ +/* Enable local speed optimizations inside code */ + +/* #define _LZMA_SYSTEM_SIZE_T */ +/* Use system's size_t. You can use it to enable 64-bit sizes supporting*/ + +#ifndef UInt32 +#ifdef _LZMA_UINT32_IS_ULONG +#define UInt32 unsigned long +#else +#define UInt32 unsigned int +#endif +#endif + +#ifndef SizeT +#ifdef _LZMA_SYSTEM_SIZE_T +#include +#define SizeT size_t +#else +#define SizeT UInt32 +#endif +#endif + +#ifdef _LZMA_PROB32 +#define CProb UInt32 +#else +#define CProb unsigned short +#endif + +#define LZMA_RESULT_OK 0 +#define LZMA_RESULT_DATA_ERROR 1 + +#ifdef _LZMA_IN_CB +typedef struct _ILzmaInCallback +{ + int (*Read)(void *object, const unsigned char **buffer, SizeT *bufferSize); +} ILzmaInCallback; +#endif + +#define LZMA_BASE_SIZE 1846 +#define LZMA_LIT_SIZE 768 + +#define LZMA_PROPERTIES_SIZE 5 + +typedef struct _CLzmaProperties +{ + int lc; + int lp; + int pb; + #ifdef _LZMA_OUT_READ + UInt32 DictionarySize; + #endif +}CLzmaProperties; + +int LzmaDecodeProperties(CLzmaProperties *propsRes, const unsigned char *propsData, int size); + +#define LzmaGetNumProbs(Properties) (LZMA_BASE_SIZE + (LZMA_LIT_SIZE << ((Properties)->lc + (Properties)->lp))) + +#define kLzmaNeedInitId (-2) + +typedef struct _CLzmaDecoderState +{ + CLzmaProperties Properties; + CProb *Probs; + + #ifdef _LZMA_IN_CB + const unsigned char *Buffer; + const unsigned char *BufferLim; + #endif + + #ifdef _LZMA_OUT_READ + unsigned char *Dictionary; + UInt32 Range; + UInt32 Code; + UInt32 DictionaryPos; + UInt32 GlobalPos; + UInt32 DistanceLimit; + UInt32 Reps[4]; + int State; + int RemainLen; + unsigned char TempDictionary[4]; + #endif +} CLzmaDecoderState; + +#ifdef _LZMA_OUT_READ +#define LzmaDecoderInit(vs) { (vs)->RemainLen = kLzmaNeedInitId; } +#endif + +int LzmaDecode(CLzmaDecoderState *vs, + #ifdef _LZMA_IN_CB + ILzmaInCallback *inCallback, + #else + const unsigned char *inStream, SizeT inSize, SizeT *inSizeProcessed, + #endif + unsigned char *outStream, SizeT outSize, SizeT *outSizeProcessed); + +#endif diff --git a/init/do_mounts_rd.c b/init/do_mounts_rd.c index ed652f4..5fd1ec5 100644 --- a/init/do_mounts_rd.c +++ b/init/do_mounts_rd.c @@ -5,7 +5,9 @@ #include #include #include +#include #include +#include #include #include "do_mounts.h" @@ -31,6 +33,9 @@ static int __init ramdisk_start_setup(char *str) __setup("ramdisk_start=", ramdisk_start_setup); static int __init crd_load(int in_fd, int out_fd); +#ifdef CONFIG_LZMA_INITRD +static int __init lzma_rd_load(int in_fd, int out_fd); +#endif /* * This routine tries to find a RAM disk image to load, and returns the @@ -39,6 +44,7 @@ static int __init crd_load(int in_fd, int out_fd); * numbers could not be found. * * We currently check for the following magic numbers: + * squashfs * minix * ext2 * romfs @@ -53,6 +59,7 @@ identify_ramdisk_image(int fd, int start_block) struct ext2_super_block *ext2sb; struct romfs_super_block *romfsb; struct cramfs_super *cramfsb; + struct squashfs_super_block *squashfsb; int nblocks = -1; unsigned char *buf; @@ -64,6 +71,7 @@ identify_ramdisk_image(int fd, int start_block) ext2sb = (struct ext2_super_block *) buf; romfsb = (struct romfs_super_block *) buf; cramfsb = (struct cramfs_super *) buf; + squashfsb = (struct squashfs_super_block *) buf; memset(buf, 0xe5, size); /* @@ -82,6 +90,17 @@ identify_ramdisk_image(int fd, int start_block) nblocks = 0; goto done; } + /* + * handle lzma compressed initrd, returns nblocks=1 as indication + */ + if( buf[0] < 9 * 5 * 5 && buf[9] == 0 && buf[10] == 0 && buf[11] == 0 + && buf[12] == 0 ) + { + printk( KERN_NOTICE "RAMDISK: LZMA image found at block %d\n", + start_block); + nblocks = 1; // just a convenient return flag + goto done; + } /* romfs is at block zero too */ if (romfsb->word0 == ROMSB_WORD0 && @@ -101,6 +120,18 @@ identify_ramdisk_image(int fd, int start_block) goto done; } + /* squashfs is at block zero too */ + if (squashfsb->s_magic == SQUASHFS_MAGIC) { + printk(KERN_NOTICE + "RAMDISK: squashfs filesystem found at block %d\n", + start_block); + if (squashfsb->s_major < 3) + nblocks = (squashfsb->bytes_used_2+BLOCK_SIZE-1)>>BLOCK_SIZE_BITS; + else + nblocks = (squashfsb->bytes_used+BLOCK_SIZE-1)>>BLOCK_SIZE_BITS; + goto done; + } + /* * Read block 1 to test for minix and ext2 superblock */ @@ -172,7 +203,22 @@ int __init rd_load_image(char *from) #endif goto done; } - +#ifdef CONFIG_LZMA_INITRD + /* + * handle lzma compressed image + */ + if ( nblocks == 1 ) + { + nblocks = 0; + if ( lzma_rd_load(in_fd, out_fd) == 0 ) + { + printk("\nLZMA initrd loaded successfully\n"); + goto successful_load; + } + printk(KERN_NOTICE "LZMA initrd is not in the correct format\n"); + goto done; + } +#endif /* * NOTE NOTE: nblocks is not actually blocks but * the number of kibibytes of data to load into a ramdisk. @@ -393,6 +439,134 @@ static void __init error(char *x) unzip_error = 1; } +#ifdef CONFIG_LZMA_INITRD +#define _LZMA_IN_CB +#define _LZMA_OUT_READ +#include "LzmaDecode.h" +#include "LzmaDecode.c" + +static int read_byte(void *object, const unsigned char **buffer, SizeT *bufferSize); + +/* + * Do the lzma decompression + */ +static int __init lzma_rd_load(int in_fd, int out_fd) +{ + unsigned int i; + CLzmaDecoderState state; + unsigned char* outputbuffer; + unsigned int uncompressedSize = 0; + unsigned char* p; + unsigned int kBlockSize = 0x10000; + unsigned int nowPos = 0; + unsigned int outsizeProcessed = 0; + int res; + ILzmaInCallback callback; + + insize = 0; /* valid bytes in inbuf */ + inptr = 0; /* index of next byte to be processed in inbuf */ + exit_code = 0; + crd_infd = in_fd; + inbuf = kmalloc(INBUFSIZ, GFP_KERNEL); + if (inbuf == 0) + { + printk(KERN_ERR "RAMDISK: Couldn't allocate lzma input buffer\n"); + return -1; + } + + callback.Read = read_byte; + + /* lzma args */ + i = get_byte(); + state.Properties.lc = i % 9, i = i / 9; + state.Properties.lp = i % 5, state.Properties.pb = i / 5; + + /* read dictionary size */ + p = (char*)&state.Properties.DictionarySize; + for (i = 0; i < 4; i++) + *p++ = get_byte(); + + /* get uncompressedSize */ + p= (char*)&uncompressedSize; + for (i = 0; i < 4; i++) + *p++ = get_byte(); + + /* skip big file */ + for (i = 0; i < 4; i++) + get_byte(); + + printk( KERN_NOTICE "RAMDISK: LZMA lc=%d,lp=%d,pb=%d,dictSize=%d,origSize=%d\n", + state.Properties.lc, state.Properties.lp, state.Properties.pb, state.Properties.DictionarySize, uncompressedSize); + outputbuffer = kmalloc(kBlockSize, GFP_KERNEL); + if (outputbuffer == 0) { + printk(KERN_ERR "RAMDISK: Couldn't allocate lzma output buffer\n"); + return -1; + } + + state.Probs = (CProb*)kmalloc( LzmaGetNumProbs(&state.Properties)*sizeof(CProb), GFP_KERNEL); + if ( state.Probs == 0) { + printk(KERN_ERR "RAMDISK: Couldn't allocate lzma workspace\n"); + return -1; + } + +#ifdef CONFIG_LZMA_INITRD_KMALLOC_ONLY + state.Dictionary = kmalloc( state.Properties.DictionarySize, GFP_KERNEL); +#else + state.Dictionary = vmalloc( state.Properties.DictionarySize); +#endif + if ( state.Dictionary == 0) { + printk(KERN_ERR "RAMDISK: Couldn't allocate lzma dictionary\n"); + return -1; + } + + printk( KERN_NOTICE "LZMA initrd by Ming-Ching Tiew " ); + + LzmaDecoderInit( &state ); + + for( nowPos =0; nowPos < uncompressedSize ; ) + { + UInt32 blockSize = uncompressedSize - nowPos; + if( blockSize > kBlockSize) + blockSize = kBlockSize; + res = LzmaDecode( &state, &callback, outputbuffer, blockSize, &outsizeProcessed); + if( res != 0 ) { + printk( KERN_ERR "RAMDISK: Lzma decode failure\n"); + return -1; + } + if( outsizeProcessed == 0 ) + { + uncompressedSize = nowPos; + printk( KERN_NOTICE "RAMDISK nowPos=%d, uncompressedSize=%d\n", + nowPos, uncompressedSize ); + break; + } + sys_write(out_fd, outputbuffer, outsizeProcessed ); + nowPos += outsizeProcessed; + printk( "."); + } + +#ifdef CONFIG_LZMA_INITRD_KMALLOC_ONLY + kfree(state.Dictionary); +#else + vfree(state.Dictionary); +#endif + kfree(inbuf); + kfree(outputbuffer); + kfree(state.Probs); + return 0; +} + +static int read_byte(void *object, const unsigned char **buffer, SizeT *bufferSize) +{ + static unsigned char val; + *bufferSize = 1; + val = get_byte(); + *buffer = &val; + return LZMA_RESULT_OK; +} + +#endif /*CONFIG_LZMA_INITRD*/ + static int __init crd_load(int in_fd, int out_fd) { int result; diff --git a/init/initramfs.c b/init/initramfs.c index 00eff7a..30d32a2 100644 --- a/init/initramfs.c +++ b/init/initramfs.c @@ -6,6 +6,7 @@ #include #include #include +#include static __initdata char *message; static void __init error(char *x) @@ -441,6 +442,118 @@ static void __init flush_window(void) outcnt = 0; } +#ifdef CONFIG_LZMA_INITRAM_FS +#define _LZMA_IN_CB +#define _LZMA_OUT_READ +#include "LzmaDecode.h" +#ifndef CONFIG_LZMA_INITRD + #include "LzmaDecode.c" +#endif +static int read_byte(void *object, const unsigned char **buffer, SizeT *bufferSize) +{ + static unsigned char val; + *bufferSize = 1; + val = get_byte(); + *buffer = &val; + return LZMA_RESULT_OK; +} + +static int __init lzma_unzip(void) +{ + unsigned int i; + CLzmaDecoderState state; + unsigned char* outputbuffer; + unsigned int uncompressedSize = 0; + unsigned char* p; + unsigned int kBlockSize = 0x10000; + unsigned int nowPos = 0; + unsigned int outsizeProcessed = 0; + int res; + ILzmaInCallback callback; + + callback.Read = read_byte; + + // lzma args + i = get_byte(); + state.Properties.lc = i % 9, i = i / 9; + state.Properties.lp = i % 5, state.Properties.pb = i / 5; + + // read dictionary size + p = (char*)&state.Properties.DictionarySize; + for (i = 0; i < 4; i++) + *p++ = get_byte(); + + // get uncompressedSize + p= (char*)&uncompressedSize; + for (i = 0; i < 4; i++) + *p++ = get_byte(); + + // skip big file + for (i = 0; i < 4; i++) + get_byte(); + + printk( KERN_NOTICE "initramfs: LZMA lc=%d,lp=%d,pb=%d,dictSize=%d,origSize=%d\n", + state.Properties.lc,state.Properties.lp,state.Properties.pb,state.Properties.DictionarySize, uncompressedSize); + outputbuffer = kmalloc(kBlockSize, GFP_KERNEL); + if (outputbuffer == 0) { + printk(KERN_ERR "initramfs: Couldn't allocate lzma output buffer\n"); + return -1; + } + + state.Probs = (CProb*) kmalloc( LzmaGetNumProbs(&state.Properties)*sizeof(CProb), GFP_KERNEL); + if ( state.Probs == 0) { + printk(KERN_ERR "initramfs: Couldn't allocate lzma workspace\n"); + return -1; + } + +#ifdef CONFIG_LZMA_INITRAM_FS_KMALLOC_ONLY + state.Dictionary = kmalloc( state.Properties.DictionarySize, GFP_KERNEL); +#else + state.Dictionary = vmalloc( state.Properties.DictionarySize); +#endif + if ( state.Dictionary == 0) { + printk(KERN_ERR "initramfs: Couldn't allocate lzma dictionary\n"); + return -1; + } + + printk( KERN_NOTICE "LZMA initramfs by Ming-Ching Tiew " ); + + LzmaDecoderInit( &state ); + + for( nowPos =0; nowPos < uncompressedSize ; ) + { + UInt32 blockSize = uncompressedSize - nowPos; + if( blockSize > kBlockSize) + blockSize = kBlockSize; + res = LzmaDecode( &state, &callback, outputbuffer, blockSize, &outsizeProcessed); + if( res != 0 ) { + panic( KERN_ERR "initramfs: Lzma decode failure\n"); + return -1; + } + if( outsizeProcessed == 0 ) + { + uncompressedSize = nowPos; + printk( KERN_NOTICE "initramfs: nowPos=%d, uncompressedSize=%d\n", + nowPos, uncompressedSize ); + break; + } + flush_buffer(outputbuffer, outsizeProcessed); + nowPos += outsizeProcessed; + printk( "."); + } + +#ifdef CONFIG_LZMA_INITRAM_FS_KMALLOC_ONLY + kfree(state.Dictionary); +#else + vfree(state.Dictionary); +#endif + kfree(outputbuffer); + kfree(state.Probs); + return 0; +} + +#endif /*CONFIG LZMA_INITRAM_FS*/ + static char * __init unpack_to_rootfs(char *buf, unsigned len, int check_only) { int written; @@ -475,12 +588,31 @@ static char * __init unpack_to_rootfs(char *buf, unsigned len, int check_only) inptr = 0; outcnt = 0; /* bytes in output buffer */ bytes_out = 0; - crc = (ulg)0xffffffffL; /* shift register contents */ - makecrc(); - gunzip(); - if (state != Reset) + if( inbuf[0] == 037 && ((inbuf[1] == 0213) || (inbuf[1] == 0236))) + { + printk( KERN_NOTICE "detected gzip initramfs\n"); + crc = (ulg)0xffffffffL; /* shift register contents */ + makecrc(); + gunzip(); + if (state != Reset) error("junk in gzipped archive"); - this_header = saved_offset + inptr; + } +#ifdef CONFIG_LZMA_INITRAM_FS + else if( inbuf[0] < 9 * 5 * 5 && buf[9] == 0 && buf[10] == 0 + && buf[11] == 0 && buf[12] == 0 ) + { + printk( KERN_NOTICE "detected lzma initramfs\n"); + lzma_unzip(); + } +#endif + else + { + // skip forward ? + crc = (ulg)0xffffffffL; /* shift register contents */ + makecrc(); + gunzip(); + } + this_header = saved_offset + inptr; buf += inptr; len -= inptr; } diff --git a/kernel/Makefile b/kernel/Makefile index ac6b27a..bd498a2 100644 --- a/kernel/Makefile +++ b/kernel/Makefile @@ -66,7 +66,7 @@ $(obj)/configs.o: $(obj)/config_data.h # config_data.h contains the same information as ikconfig.h but gzipped. # Info from config_data can be extracted from /proc/config* targets += config_data.gz -$(obj)/config_data.gz: .config FORCE +$(obj)/config_data.gz: .miniconfig FORCE $(call if_changed,gzip) quiet_cmd_ikconfiggz = IKCFG $@ diff --git a/kernel/configs.c b/kernel/configs.c index 8fa1fb2..c8407eb 100644 --- a/kernel/configs.c +++ b/kernel/configs.c @@ -88,7 +88,7 @@ static int __init ikconfig_init(void) struct proc_dir_entry *entry; /* create the current config file */ - entry = create_proc_entry("config.gz", S_IFREG | S_IRUGO, + entry = create_proc_entry("miniconfig.gz", S_IFREG | S_IRUGO, &proc_root); if (!entry) return -ENOMEM; @@ -104,7 +104,7 @@ static int __init ikconfig_init(void) static void __exit ikconfig_cleanup(void) { - remove_proc_entry("config.gz", &proc_root); + remove_proc_entry("miniconfig.gz", &proc_root); } module_init(ikconfig_init); diff --git a/kernel/time/clocksource.c b/kernel/time/clocksource.c index fe5c7db..a5150e6 100644 --- a/kernel/time/clocksource.c +++ b/kernel/time/clocksource.c @@ -85,8 +85,8 @@ static void clocksource_ratewd(struct clocksource *cs, int64_t delta) if (delta > -WATCHDOG_TRESHOLD && delta < WATCHDOG_TRESHOLD) return; - printk(KERN_WARNING "Clocksource %s unstable (delta = %Ld ns)\n", - cs->name, delta); +/* printk(KERN_WARNING "Clocksource %s unstable (delta = %Ld ns)\n", + cs->name, delta); */ cs->flags &= ~(CLOCK_SOURCE_VALID_FOR_HRES | CLOCK_SOURCE_WATCHDOG); clocksource_change_rating(cs, 0); cs->flags &= ~CLOCK_SOURCE_WATCHDOG; diff --git a/kernel/timer.c b/kernel/timer.c index dd6c2c1..3a8f485 100644 --- a/kernel/timer.c +++ b/kernel/timer.c @@ -916,8 +916,8 @@ static void change_clocksource(void) tick_clock_notify(); - printk(KERN_INFO "Time: %s clocksource has been installed.\n", - clock->name); +/* printk(KERN_INFO "Time: %s clocksource has been installed.\n", + clock->name); */ } #else static inline void change_clocksource(void) { } diff --git a/miniconfig.sh b/miniconfig.sh new file mode 100755 index 0000000..28e7433 --- /dev/null +++ b/miniconfig.sh @@ -0,0 +1,2 @@ +#!/bin/sh -f +make allnoconfig KCONFIG_ALLCONFIG=.miniconfig diff --git a/scripts/Makefile.lib b/scripts/Makefile.lib index fc498fe..e98172c 100644 --- a/scripts/Makefile.lib +++ b/scripts/Makefile.lib @@ -162,4 +162,9 @@ cmd_objcopy = $(OBJCOPY) $(OBJCOPYFLAGS) $(OBJCOPYFLAGS_$(@F)) $< $@ quiet_cmd_gzip = GZIP $@ cmd_gzip = gzip -f -9 < $< > $@ +# LZMA +# +quiet_cmd_lzma = LZMA $@ +cmd_lzma = lzma e $< $@ -lc7 -lp0 -pb0 2>/dev/null + diff --git a/scripts/gen_lzma_initramfs_list.sh b/scripts/gen_lzma_initramfs_list.sh new file mode 100644 index 0000000..be3ed6a --- /dev/null +++ b/scripts/gen_lzma_initramfs_list.sh @@ -0,0 +1,292 @@ +#!/bin/bash +# Copyright (C) Martin Schlemmer +# Copyright (c) 2006 Sam Ravnborg +# +# Released under the terms of the GNU GPL +# +# Generate a cpio packed initramfs. It uses gen_init_cpio to generate +# the cpio archive, and gzip to pack it. +# The script may also be used to generate the inputfile used for gen_init_cpio +# This script assumes that gen_init_cpio is located in usr/ directory + +# error out on errors +set -e + +usage() { +cat << EOF +Usage: +$0 [-o ] [-u ] [-g ] { -s | -d | } ... + -o Create lzma initramfs file named using + gen_init_cpio and lzma + -u User ID to map to user ID 0 (root). + is only meaningful if + is a directory. + -g Group ID to map to group ID 0 (root). + is only meaningful if + is a directory. + File list or directory for cpio archive. + If is a .cpio file it will be used + as direct input to initramfs. + -s Create lzma file with small dictionary size + -d Output the default cpio list. + +All options except -o and -l may be repeated and are interpreted +sequentially and immediately. -u and -g states are preserved across + options so an explicit "-u 0 -g 0" is required +to reset the root/group mapping. +EOF +} + +list_default_initramfs() { + # echo usr/kinit/kinit + : +} + +default_initramfs() { + cat <<-EOF >> ${output} + # This is a very simple, default initramfs + + dir /dev 0755 0 0 + nod /dev/console 0600 0 0 c 5 1 + dir /root 0700 0 0 + # file /kinit usr/kinit/kinit 0755 0 0 + # slink /init kinit 0755 0 0 + EOF +} + +filetype() { + local argv1="$1" + + # symlink test must come before file test + if [ -L "${argv1}" ]; then + echo "slink" + elif [ -f "${argv1}" ]; then + echo "file" + elif [ -d "${argv1}" ]; then + echo "dir" + elif [ -b "${argv1}" -o -c "${argv1}" ]; then + echo "nod" + elif [ -p "${argv1}" ]; then + echo "pipe" + elif [ -S "${argv1}" ]; then + echo "sock" + else + echo "invalid" + fi + return 0 +} + +list_print_mtime() { + : +} + +print_mtime() { + local my_mtime="0" + + if [ -e "$1" ]; then + my_mtime=$(find "$1" -printf "%T@\n" | sort -r | head -n 1) + fi + + echo "# Last modified: ${my_mtime}" >> ${output} + echo "" >> ${output} +} + +list_parse() { + echo "$1 \\" +} + +# for each file print a line in following format +# +# for links, devices etc the format differs. See gen_init_cpio for details +parse() { + local location="$1" + local name="${location/${srcdir}//}" + # change '//' into '/' + name="${name//\/\///}" + local mode="$2" + local uid="$3" + local gid="$4" + local ftype=$(filetype "${location}") + # remap uid/gid to 0 if necessary + [ "$uid" -eq "$root_uid" ] && uid=0 + [ "$gid" -eq "$root_gid" ] && gid=0 + local str="${mode} ${uid} ${gid}" + + [ "${ftype}" == "invalid" ] && return 0 + [ "${location}" == "${srcdir}" ] && return 0 + + case "${ftype}" in + "file") + str="${ftype} ${name} ${location} ${str}" + ;; + "nod") + local dev_type= + local maj=$(LC_ALL=C ls -l "${location}" | \ + gawk '{sub(/,/, "", $5); print $5}') + local min=$(LC_ALL=C ls -l "${location}" | \ + gawk '{print $6}') + + if [ -b "${location}" ]; then + dev_type="b" + else + dev_type="c" + fi + str="${ftype} ${name} ${str} ${dev_type} ${maj} ${min}" + ;; + "slink") + local target=$(LC_ALL=C ls -l "${location}" | \ + gawk '{print $11}') + str="${ftype} ${name} ${target} ${str}" + ;; + *) + str="${ftype} ${name} ${str}" + ;; + esac + + echo "${str}" >> ${output} + + return 0 +} + +unknown_option() { + printf "ERROR: unknown option \"$arg\"\n" >&2 + printf "If the filename validly begins with '-', " >&2 + printf "then it must be prefixed\n" >&2 + printf "by './' so that it won't be interpreted as an option." >&2 + printf "\n" >&2 + usage >&2 + exit 1 +} + +list_header() { + : +} + +header() { + printf "\n#####################\n# $1\n" >> ${output} +} + +# process one directory (incl sub-directories) +dir_filelist() { + ${dep_list}header "$1" + + srcdir=$(echo "$1" | sed -e 's://*:/:g') + dirlist=$(find "${srcdir}" -printf "%p %m %U %G\n" 2>/dev/null) + + # If $dirlist is only one line, then the directory is empty + if [ "$(echo "${dirlist}" | wc -l)" -gt 1 ]; then + ${dep_list}print_mtime "$1" + + echo "${dirlist}" | \ + while read x; do + ${dep_list}parse ${x} + done + fi +} + +# if only one file is specified and it is .cpio file then use it direct as fs +# if a directory is specified then add all files in given direcotry to fs +# if a regular file is specified assume it is in gen_initramfs format +input_file() { + source="$1" + if [ -f "$1" ]; then + ${dep_list}header "$1" + is_cpio="$(echo "$1" | sed 's/^.*\.cpio/cpio/')" + if [ $2 -eq 0 -a ${is_cpio} == "cpio" ]; then + cpio_file=$1 + [ ! -z ${dep_list} ] && echo "$1" + return 0 + fi + if [ -z ${dep_list} ]; then + print_mtime "$1" >> ${output} + cat "$1" >> ${output} + else + cat "$1" | while read type dir file perm ; do + if [ "$type" == "file" ]; then + echo "$file \\"; + fi + done + fi + elif [ -d "$1" ]; then + dir_filelist "$1" + else + echo " ${prog}: Cannot open '$1'" >&2 + exit 1 + fi +} + +prog=$0 +root_uid=0 +root_gid=0 +dep_list= +cpio_file= +cpio_list= +output="/dev/stdout" +output_file="" +opt="" + +arg="$1" +case "$arg" in + "-l") # files included in initramfs - used by kbuild + dep_list="list_" + echo "deps_initramfs := \\" + shift + ;; + "-o") # generate lzma-ed cpio image named $1 + shift + output_file="$1" + cpio_list="$(mktemp ${TMPDIR:-/tmp}/cpiolist.XXXXXX)" + output=${cpio_list} + shift + ;; +esac +while [ $# -gt 0 ]; do + arg="$1" + shift + case "$arg" in + "-u") # map $1 to uid=0 (root) + root_uid="$1" + shift + ;; + "-g") # map $1 to gid=0 (root) + root_gid="$1" + shift + ;; + "-s") + opt="-d16" + ;; + "-d") # display default initramfs list + default_list="$arg" + ${dep_list}default_initramfs + ;; + "-h") + usage + exit 0 + ;; + *) + case "$arg" in + "-"*) + unknown_option + ;; + *) # input file/dir - process it + input_file "$arg" "$#" + ;; + esac + ;; + esac +done + +# If output_file is set we will generate cpio archive and lzma it +# we are carefull to delete tmp files +if [ ! -z ${output_file} ]; then + if [ -z ${cpio_file} ]; then + cpio_tfile="$(mktemp ${TMPDIR:-/tmp}/cpiofile.XXXXXX)" + usr/gen_init_cpio ${cpio_list} > ${cpio_tfile} + else + cpio_tfile=${cpio_file} + fi + rm ${cpio_list} + lzma e ${cpio_tfile} ${output_file} ${opt} + [ -z ${cpio_file} ] && rm ${cpio_tfile} +fi +exit 0 diff --git a/shrinkconfig.sh b/shrinkconfig.sh new file mode 100755 index 0000000..e7a3df7 --- /dev/null +++ b/shrinkconfig.sh @@ -0,0 +1,79 @@ +#! /bin/bash + +# shrinkconfig copyright 2006 by Rob Landley +# Licensed under the GNU General Public License version 2. + +if [ $# -ne 1 ] +then + echo "Turns current .config into a miniconfig file." + echo "Usage: shrinkconfig mini.config" + exit 1 +fi + +if [ ! -f .config ] +then + echo "Need a .config file to shrink." + exit 1 +fi +LENGTH=$(wc -l < .config) + +OUTPUT="$1" +cp .config "$OUTPUT" +if [ $? -ne 0 ] +then + echo "Couldn't create $OUTPUT" + exit 1 +fi + +# If we get interrupted, clean up the mess + +KERNELOUTPUT="" + +function cleanup +{ + echo + echo "Interrupted." + [ ! -z "$KERNELOUTPUT" ] && rm -rf "$KERNELOUTPUT" + rm "$OUTPUT" + exit 1 +} + +trap cleanup HUP INT QUIT TERM + +# Since the "O=" argument to make doesn't work recursively, we need to jump +# through a few hoops to avoid overwriting the .config that we're shrinking. + +# If we're building out of tree, we'll have absolute paths to source and build +# directories in the Makefile. + +KERNELSRC=$(sed -n -e 's/KERNELSRC[^/]*:=[^/]*//p' Makefile) +[ -z "$KERNELSRC" ] && KERNELSRC=$(pwd) +KERNELOUTPUT=`pwd`/.config.minitemp + +mkdir -p "$KERNELOUTPUT" || exit 1 + +echo "Shrinking .config to $OUTPUT..." + +for I in $(seq 1 $LENGTH) +do + echo -n -e "\r"$I/$LENGTH lines $(wc -c < "$OUTPUT") bytes + + sed -n "${I}!p" "$OUTPUT" > "$KERNELOUTPUT"/.config.test + # Do a config with this file + make -C "$KERNELSRC" O="$KERNELOUTPUT" allnoconfig KCONFIG_ALLCONFIG="$KERNELOUTPUT"/.config.test > /dev/null + + # Compare. The date changes, so expect a small difference each time. + D=$(diff "$KERNELOUTPUT"/.config .config | wc -l) + if [ $D -eq 4 ] + then + mv "$KERNELOUTPUT"/.config.test "$OUTPUT" + LENGTH=$[$LENGTH-1] + else + I=$[$I + 1] + fi +done + +rm -rf "$KERNELOUTPUT" + +# One extra echo to preserve status line. +echo diff --git a/usr/Makefile b/usr/Makefile index 201f27f..8e1f6ea 100644 --- a/usr/Makefile +++ b/usr/Makefile @@ -19,6 +19,7 @@ $(obj)/initramfs_data.o: $(obj)/initramfs_data.cpio.gz FORCE hostprogs-y := gen_init_cpio initramfs := $(CONFIG_SHELL) $(srctree)/scripts/gen_initramfs_list.sh +lzma_initramfs := $(CONFIG_SHELL) $(srctree)/scripts/gen_lzma_initramfs_list.sh ramfs-input := $(if $(filter-out "",$(CONFIG_INITRAMFS_SOURCE)), \ $(shell echo $(CONFIG_INITRAMFS_SOURCE)),-d) ramfs-args := \ @@ -36,6 +37,14 @@ endif quiet_cmd_initfs = GEN $@ cmd_initfs = $(initramfs) -o $@ $(ramfs-args) $(ramfs-input) +ifdef CONFIG_LZMA_INITRAM_FS_SMALLMEM +quiet_cmd_lzma_initfs = LZRAMFS $@ + cmd_lzma_initfs = $(lzma_initramfs) -o $@ $(ramfs-args) -s $(ramfs-input) +else +quiet_cmd_lzma_initfs = LZRAMFS $@ + cmd_lzma_initfs = $(lzma_initramfs) -o $@ $(ramfs-args) $(ramfs-input) +endif + targets := initramfs_data.cpio.gz # do not try to update files included in initramfs $(deps_initramfs): ; @@ -48,5 +57,9 @@ $(deps_initramfs): klibcdirs # 4) arguments to gen_initramfs.sh changes $(obj)/initramfs_data.cpio.gz: $(obj)/gen_init_cpio $(deps_initramfs) klibcdirs $(Q)$(initramfs) -l $(ramfs-input) > $(obj)/.initramfs_data.cpio.gz.d +ifdef CONFIG_LZMA_INITRAM_FS + $(call if_changed,lzma_initfs) +else $(call if_changed,initfs) +endif