X-Git-Url: http://plrg.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FSupport%2FMemoryBuffer.cpp;h=53437c858b6f708010b0d8e16eb080cab344797e;hb=5a1a1856a4dfa1335d937437fade5c0bbab06560;hp=4f135ead182c106cf9a05228bb03f769e6736910;hpb=d48c1f9b105c1c09e3914d9a08ad90f6a0893c9a;p=oota-llvm.git diff --git a/lib/Support/MemoryBuffer.cpp b/lib/Support/MemoryBuffer.cpp index 4f135ead182..53437c858b6 100644 --- a/lib/Support/MemoryBuffer.cpp +++ b/lib/Support/MemoryBuffer.cpp @@ -14,119 +14,148 @@ #include "llvm/Support/MemoryBuffer.h" #include "llvm/ADT/OwningPtr.h" #include "llvm/ADT/SmallString.h" -#include "llvm/System/Errno.h" -#include "llvm/System/Path.h" -#include "llvm/System/Process.h" -#include "llvm/System/Program.h" +#include "llvm/Config/config.h" +#include "llvm/Support/Errno.h" +#include "llvm/Support/FileSystem.h" +#include "llvm/Support/MathExtras.h" +#include "llvm/Support/Path.h" +#include "llvm/Support/Process.h" +#include "llvm/Support/Program.h" +#include "llvm/Support/system_error.h" #include +#include #include #include -#include -#include +#include #include +#include #if !defined(_MSC_VER) && !defined(__MINGW32__) #include -#include #else #include +// Simplistic definitinos of these macros for use in getOpenFile. +#ifndef S_ISREG +#define S_ISREG(x) (1) +#endif +#ifndef S_ISBLK +#define S_ISBLK(x) (0) +#endif #endif -#include using namespace llvm; //===----------------------------------------------------------------------===// // MemoryBuffer implementation itself. //===----------------------------------------------------------------------===// -MemoryBuffer::~MemoryBuffer() { - if (MustDeleteBuffer) - free((void*)BufferStart); -} - -/// initCopyOf - Initialize this source buffer with a copy of the specified -/// memory range. We make the copy so that we can null terminate it -/// successfully. -void MemoryBuffer::initCopyOf(const char *BufStart, const char *BufEnd) { - size_t Size = BufEnd-BufStart; - BufferStart = (char *)malloc(Size+1); - BufferEnd = BufferStart+Size; - memcpy(const_cast(BufferStart), BufStart, Size); - *const_cast(BufferEnd) = 0; // Null terminate buffer. - MustDeleteBuffer = true; -} +MemoryBuffer::~MemoryBuffer() { } /// init - Initialize this MemoryBuffer as a reference to externally allocated /// memory, memory that we know is already null terminated. -void MemoryBuffer::init(const char *BufStart, const char *BufEnd) { - assert(BufEnd[0] == 0 && "Buffer is not null terminated!"); +void MemoryBuffer::init(const char *BufStart, const char *BufEnd, + bool RequiresNullTerminator) { + assert((!RequiresNullTerminator || BufEnd[0] == 0) && + "Buffer is not null terminated!"); BufferStart = BufStart; BufferEnd = BufEnd; - MustDeleteBuffer = false; } //===----------------------------------------------------------------------===// // MemoryBufferMem implementation. //===----------------------------------------------------------------------===// +/// CopyStringRef - Copies contents of a StringRef into a block of memory and +/// null-terminates it. +static void CopyStringRef(char *Memory, StringRef Data) { + memcpy(Memory, Data.data(), Data.size()); + Memory[Data.size()] = 0; // Null terminate string. +} + namespace { +struct NamedBufferAlloc { + StringRef Name; + NamedBufferAlloc(StringRef Name) : Name(Name) {} +}; +} + +void *operator new(size_t N, const NamedBufferAlloc &Alloc) { + char *Mem = static_cast(operator new(N + Alloc.Name.size() + 1)); + CopyStringRef(Mem + N, Alloc.Name); + return Mem; +} + +namespace { +/// MemoryBufferMem - Named MemoryBuffer pointing to a block of memory. class MemoryBufferMem : public MemoryBuffer { - std::string FileID; public: - MemoryBufferMem(const char *Start, const char *End, StringRef FID, - bool Copy = false) - : FileID(FID) { - if (!Copy) - init(Start, End); - else - initCopyOf(Start, End); + MemoryBufferMem(StringRef InputData, bool RequiresNullTerminator) { + init(InputData.begin(), InputData.end(), RequiresNullTerminator); } - - virtual const char *getBufferIdentifier() const { - return FileID.c_str(); + + virtual const char *getBufferIdentifier() const LLVM_OVERRIDE { + // The name is stored after the class itself. + return reinterpret_cast(this + 1); + } + + virtual BufferKind getBufferKind() const LLVM_OVERRIDE { + return MemoryBuffer_Malloc; } }; } /// getMemBuffer - Open the specified memory range as a MemoryBuffer. Note -/// that EndPtr[0] must be a null byte and be accessible! -MemoryBuffer *MemoryBuffer::getMemBuffer(const char *StartPtr, - const char *EndPtr, - const char *BufferName) { - return new MemoryBufferMem(StartPtr, EndPtr, BufferName); +/// that InputData must be a null terminated if RequiresNullTerminator is true! +MemoryBuffer *MemoryBuffer::getMemBuffer(StringRef InputData, + StringRef BufferName, + bool RequiresNullTerminator) { + return new (NamedBufferAlloc(BufferName)) + MemoryBufferMem(InputData, RequiresNullTerminator); } /// getMemBufferCopy - Open the specified memory range as a MemoryBuffer, /// copying the contents and taking ownership of it. This has no requirements /// on EndPtr[0]. -MemoryBuffer *MemoryBuffer::getMemBufferCopy(const char *StartPtr, - const char *EndPtr, - const char *BufferName) { - return new MemoryBufferMem(StartPtr, EndPtr, BufferName, true); +MemoryBuffer *MemoryBuffer::getMemBufferCopy(StringRef InputData, + StringRef BufferName) { + MemoryBuffer *Buf = getNewUninitMemBuffer(InputData.size(), BufferName); + if (!Buf) return 0; + memcpy(const_cast(Buf->getBufferStart()), InputData.data(), + InputData.size()); + return Buf; } /// getNewUninitMemBuffer - Allocate a new MemoryBuffer of the specified size -/// that is completely initialized to zeros. Note that the caller should -/// initialize the memory allocated by this method. The memory is owned by -/// the MemoryBuffer object. +/// that is not initialized. Note that the caller should initialize the +/// memory allocated by this method. The memory is owned by the MemoryBuffer +/// object. MemoryBuffer *MemoryBuffer::getNewUninitMemBuffer(size_t Size, StringRef BufferName) { - char *Buf = (char *)malloc(Size+1); - if (!Buf) return 0; - Buf[Size] = 0; - MemoryBufferMem *SB = new MemoryBufferMem(Buf, Buf+Size, BufferName); - // The memory for this buffer is owned by the MemoryBuffer. - SB->MustDeleteBuffer = true; - return SB; + // Allocate space for the MemoryBuffer, the data and the name. It is important + // that MemoryBuffer and data are aligned so PointerIntPair works with them. + size_t AlignedStringLen = + RoundUpToAlignment(sizeof(MemoryBufferMem) + BufferName.size() + 1, + sizeof(void*)); // TODO: Is sizeof(void*) enough? + size_t RealLen = AlignedStringLen + Size + 1; + char *Mem = static_cast(operator new(RealLen, std::nothrow)); + if (!Mem) return 0; + + // The name is stored after the class itself. + CopyStringRef(Mem + sizeof(MemoryBufferMem), BufferName); + + // The buffer begins after the name and must be aligned. + char *Buf = Mem + AlignedStringLen; + Buf[Size] = 0; // Null terminate buffer. + + return new (Mem) MemoryBufferMem(StringRef(Buf, Size), true); } /// getNewMemBuffer - Allocate a new MemoryBuffer of the specified size that /// is completely initialized to zeros. Note that the caller should /// initialize the memory allocated by this method. The memory is owned by /// the MemoryBuffer object. -MemoryBuffer *MemoryBuffer::getNewMemBuffer(size_t Size, - const char *BufferName) { +MemoryBuffer *MemoryBuffer::getNewMemBuffer(size_t Size, StringRef BufferName) { MemoryBuffer *SB = getNewUninitMemBuffer(Size, BufferName); if (!SB) return 0; - memset(const_cast(SB->getBufferStart()), 0, Size+1); + memset(const_cast(SB->getBufferStart()), 0, Size); return SB; } @@ -135,13 +164,12 @@ MemoryBuffer *MemoryBuffer::getNewMemBuffer(size_t Size, /// if the Filename is "-". If an error occurs, this returns null and fills /// in *ErrStr with a reason. If stdin is empty, this API (unlike getSTDIN) /// returns an empty buffer. -MemoryBuffer *MemoryBuffer::getFileOrSTDIN(StringRef Filename, - std::string *ErrStr, - int64_t FileSize, - struct stat *FileInfo) { +error_code MemoryBuffer::getFileOrSTDIN(StringRef Filename, + OwningPtr &result, + int64_t FileSize) { if (Filename == "-") - return getSTDIN(); - return getFile(Filename, ErrStr, FileSize, FileInfo); + return getSTDIN(result); + return getFile(Filename, result, FileSize); } //===----------------------------------------------------------------------===// @@ -149,142 +177,252 @@ MemoryBuffer *MemoryBuffer::getFileOrSTDIN(StringRef Filename, //===----------------------------------------------------------------------===// namespace { -/// MemoryBufferMMapFile - This represents a file that was mapped in with the -/// sys::Path::MapInFilePages method. When destroyed, it calls the -/// sys::Path::UnMapFilePages method. +/// \brief Memory maps a file descriptor using sys::fs::mapped_file_region. +/// +/// This handles converting the offset into a legal offset on the platform. class MemoryBufferMMapFile : public MemoryBuffer { - std::string Filename; -public: - MemoryBufferMMapFile(StringRef filename, const char *Pages, uint64_t Size) - : Filename(filename) { - init(Pages, Pages+Size); + sys::fs::mapped_file_region MFR; + + static uint64_t getLegalMapOffset(uint64_t Offset) { + return Offset & ~(sys::fs::mapped_file_region::alignment() - 1); } - - virtual const char *getBufferIdentifier() const { - return Filename.c_str(); + + static uint64_t getLegalMapSize(uint64_t Len, uint64_t Offset) { + return Len + (Offset - getLegalMapOffset(Offset)); } - - ~MemoryBufferMMapFile() { - sys::Path::UnMapFilePages(getBufferStart(), getBufferSize()); + + const char *getStart(uint64_t Len, uint64_t Offset) { + return MFR.const_data() + (Offset - getLegalMapOffset(Offset)); } -}; -/// FileCloser - RAII object to make sure an FD gets closed properly. -class FileCloser { - int FD; public: - FileCloser(int FD) : FD(FD) {} - ~FileCloser() { ::close(FD); } + MemoryBufferMMapFile(bool RequiresNullTerminator, int FD, uint64_t Len, + uint64_t Offset, error_code EC) + : MFR(FD, false, sys::fs::mapped_file_region::readonly, + getLegalMapSize(Len, Offset), getLegalMapOffset(Offset), EC) { + if (!EC) { + const char *Start = getStart(Len, Offset); + init(Start, Start + Len, RequiresNullTerminator); + } + } + + virtual const char *getBufferIdentifier() const LLVM_OVERRIDE { + // The name is stored after the class itself. + return reinterpret_cast(this + 1); + } + + virtual BufferKind getBufferKind() const LLVM_OVERRIDE { + return MemoryBuffer_MMap; + } }; } -MemoryBuffer *MemoryBuffer::getFile(StringRef Filename, std::string *ErrStr, - int64_t FileSize, struct stat *FileInfo) { - int OpenFlags = 0; -#ifdef O_BINARY - OpenFlags |= O_BINARY; // Open input file in binary mode on win32. -#endif +static error_code getMemoryBufferForStream(int FD, + StringRef BufferName, + OwningPtr &result) { + const ssize_t ChunkSize = 4096*4; + SmallString Buffer; + ssize_t ReadBytes; + // Read into Buffer until we hit EOF. + do { + Buffer.reserve(Buffer.size() + ChunkSize); + ReadBytes = read(FD, Buffer.end(), ChunkSize); + if (ReadBytes == -1) { + if (errno == EINTR) continue; + return error_code(errno, posix_category()); + } + Buffer.set_size(Buffer.size() + ReadBytes); + } while (ReadBytes != 0); + + result.reset(MemoryBuffer::getMemBufferCopy(Buffer, BufferName)); + return error_code::success(); +} + +error_code MemoryBuffer::getFile(StringRef Filename, + OwningPtr &result, + int64_t FileSize, + bool RequiresNullTerminator) { + // Ensure the path is null terminated. SmallString<256> PathBuf(Filename.begin(), Filename.end()); - int FD = ::open(PathBuf.c_str(), O_RDONLY|OpenFlags); - if (FD == -1) { - if (ErrStr) *ErrStr = sys::StrError(); - return 0; - } - FileCloser FC(FD); // Close FD on return. - + return MemoryBuffer::getFile(PathBuf.c_str(), result, FileSize, + RequiresNullTerminator); +} + +static error_code getOpenFileImpl(int FD, const char *Filename, + OwningPtr &Result, + uint64_t FileSize, uint64_t MapSize, + int64_t Offset, bool RequiresNullTerminator); + +error_code MemoryBuffer::getFile(const char *Filename, + OwningPtr &result, + int64_t FileSize, + bool RequiresNullTerminator) { + int FD; + error_code EC = sys::fs::openFileForRead(Filename, FD); + if (EC) + return EC; + + error_code ret = getOpenFileImpl(FD, Filename, result, FileSize, FileSize, 0, + RequiresNullTerminator); + close(FD); + return ret; +} + +static bool shouldUseMmap(int FD, + size_t FileSize, + size_t MapSize, + off_t Offset, + bool RequiresNullTerminator, + int PageSize) { + // We don't use mmap for small files because this can severely fragment our + // address space. + if (MapSize < 4 * 4096 || MapSize < (unsigned)PageSize) + return false; + + if (!RequiresNullTerminator) + return true; + + // If we don't know the file size, use fstat to find out. fstat on an open // file descriptor is cheaper than stat on a random path. - if (FileSize == -1 || FileInfo) { - struct stat MyFileInfo; - struct stat *FileInfoPtr = FileInfo? FileInfo : &MyFileInfo; - - // TODO: This should use fstat64 when available. - if (fstat(FD, FileInfoPtr) == -1) { - if (ErrStr) *ErrStr = sys::StrError(); - return 0; - } - FileSize = FileInfoPtr->st_size; + // FIXME: this chunk of code is duplicated, but it avoids a fstat when + // RequiresNullTerminator = false and MapSize != -1. + if (FileSize == size_t(-1)) { + sys::fs::file_status Status; + error_code EC = sys::fs::status(FD, Status); + if (EC) + return EC; + FileSize = Status.getSize(); } - - - // If the file is large, try to use mmap to read it in. We don't use mmap - // for small files, because this can severely fragment our address space. Also - // don't try to map files that are exactly a multiple of the system page size, - // as the file would not have the required null terminator. - // - // FIXME: Can we just mmap an extra page in the latter case? - if (FileSize >= 4096*4 && - (FileSize & (sys::Process::GetPageSize()-1)) != 0) { - if (const char *Pages = sys::Path::MapInFilePages(FD, FileSize)) { - // Close the file descriptor, now that the whole file is in memory. - return new MemoryBufferMMapFile(Filename, Pages, FileSize); + + // If we need a null terminator and the end of the map is inside the file, + // we cannot use mmap. + size_t End = Offset + MapSize; + assert(End <= FileSize); + if (End != FileSize) + return false; + +#if defined(_WIN32) || defined(__CYGWIN__) + // Don't peek the next page if file is multiple of *physical* pagesize(4k) + // but is not multiple of AllocationGranularity(64k), + // when a null terminator is required. + // FIXME: It's not good to hardcode 4096 here. dwPageSize shows 4096. + if ((FileSize & (4096 - 1)) == 0) + return false; +#endif + + // Don't try to map files that are exactly a multiple of the system page size + // if we need a null terminator. + if ((FileSize & (PageSize -1)) == 0) + return false; + + return true; +} + +static error_code getOpenFileImpl(int FD, const char *Filename, + OwningPtr &result, + uint64_t FileSize, uint64_t MapSize, + int64_t Offset, bool RequiresNullTerminator) { + static int PageSize = sys::process::get_self()->page_size(); + + // Default is to map the full file. + if (MapSize == uint64_t(-1)) { + // If we don't know the file size, use fstat to find out. fstat on an open + // file descriptor is cheaper than stat on a random path. + if (FileSize == uint64_t(-1)) { + sys::fs::file_status Status; + error_code EC = sys::fs::status(FD, Status); + if (EC) + return EC; + + // If this not a file or a block device (e.g. it's a named pipe + // or character device), we can't trust the size. Create the memory + // buffer by copying off the stream. + sys::fs::file_type Type = Status.type(); + if (Type != sys::fs::file_type::regular_file && + Type != sys::fs::file_type::block_file) + return getMemoryBufferForStream(FD, Filename, result); + + FileSize = Status.getSize(); } + MapSize = FileSize; + } + + if (shouldUseMmap(FD, FileSize, MapSize, Offset, RequiresNullTerminator, + PageSize)) { + error_code EC; + result.reset(new (NamedBufferAlloc(Filename)) MemoryBufferMMapFile( + RequiresNullTerminator, FD, MapSize, Offset, EC)); + if (!EC) + return error_code::success(); } - MemoryBuffer *Buf = MemoryBuffer::getNewUninitMemBuffer(FileSize, Filename); + MemoryBuffer *Buf = MemoryBuffer::getNewUninitMemBuffer(MapSize, Filename); if (!Buf) { - // Failed to create a buffer. - if (ErrStr) *ErrStr = "could not allocate buffer"; - return 0; + // Failed to create a buffer. The only way it can fail is if + // new(std::nothrow) returns 0. + return make_error_code(errc::not_enough_memory); } OwningPtr SB(Buf); char *BufPtr = const_cast(SB->getBufferStart()); - size_t BytesLeft = FileSize; + size_t BytesLeft = MapSize; +#ifndef HAVE_PREAD + if (lseek(FD, Offset, SEEK_SET) == -1) + return error_code(errno, posix_category()); +#endif + while (BytesLeft) { +#ifdef HAVE_PREAD + ssize_t NumRead = ::pread(FD, BufPtr, BytesLeft, MapSize-BytesLeft+Offset); +#else ssize_t NumRead = ::read(FD, BufPtr, BytesLeft); +#endif if (NumRead == -1) { if (errno == EINTR) continue; // Error while reading. - if (ErrStr) *ErrStr = sys::StrError(); - return 0; - } else if (NumRead == 0) { - // We hit EOF early, truncate and terminate buffer. - Buf->BufferEnd = BufPtr; - *BufPtr = 0; - return SB.take(); + return error_code(errno, posix_category()); + } + if (NumRead == 0) { + assert(0 && "We got inaccurate FileSize value or fstat reported an " + "invalid file size."); + *BufPtr = '\0'; // null-terminate at the actual size. + break; } BytesLeft -= NumRead; BufPtr += NumRead; } - return SB.take(); + result.swap(SB); + return error_code::success(); } -//===----------------------------------------------------------------------===// -// MemoryBuffer::getSTDIN implementation. -//===----------------------------------------------------------------------===// - -namespace { -class STDINBufferFile : public MemoryBuffer { -public: - virtual const char *getBufferIdentifier() const { - return ""; - } -}; +error_code MemoryBuffer::getOpenFile(int FD, const char *Filename, + OwningPtr &Result, + uint64_t FileSize, + bool RequiresNullTerminator) { + return getOpenFileImpl(FD, Filename, Result, FileSize, FileSize, 0, + RequiresNullTerminator); } -MemoryBuffer *MemoryBuffer::getSTDIN() { - char Buffer[4096*4]; +error_code MemoryBuffer::getOpenFileSlice(int FD, const char *Filename, + OwningPtr &Result, + uint64_t MapSize, int64_t Offset) { + return getOpenFileImpl(FD, Filename, Result, -1, MapSize, Offset, false); +} - std::vector FileData; +//===----------------------------------------------------------------------===// +// MemoryBuffer::getSTDIN implementation. +//===----------------------------------------------------------------------===// +error_code MemoryBuffer::getSTDIN(OwningPtr &result) { // Read in all of the data from stdin, we cannot mmap stdin. // // FIXME: That isn't necessarily true, we should try to mmap stdin and // fallback if it fails. - sys::Program::ChangeStdinToBinary(); - size_t ReadBytes; - do { - ReadBytes = fread(Buffer, sizeof(char), sizeof(Buffer), stdin); - FileData.insert(FileData.end(), Buffer, Buffer+ReadBytes); - } while (ReadBytes == sizeof(Buffer)); - - FileData.push_back(0); // &FileData[Size] is invalid. So is &*FileData.end(). - size_t Size = FileData.size(); - MemoryBuffer *B = new STDINBufferFile(); - B->initCopyOf(&FileData[0], &FileData[Size-1]); - return B; + sys::ChangeStdinToBinary(); + + return getMemoryBufferForStream(0, "", result); }