X-Git-Url: http://demsky.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FSupport%2FMemoryBuffer.cpp;h=5f4b7daae53d162affc9dd71d9cccaba140b4312;hb=ad591b341bf22bb4d3a7526d59d4ae6c414420ee;hp=542162d513b9f73c7b91be881b90416fb7a1973c;hpb=831338b320d9059b59564aa82f0703c3aff8c74f;p=oota-llvm.git diff --git a/lib/Support/MemoryBuffer.cpp b/lib/Support/MemoryBuffer.cpp index 542162d513b..5f4b7daae53 100644 --- a/lib/Support/MemoryBuffer.cpp +++ b/lib/Support/MemoryBuffer.cpp @@ -12,26 +12,27 @@ //===----------------------------------------------------------------------===// #include "llvm/Support/MemoryBuffer.h" -#include "llvm/ADT/OwningPtr.h" #include "llvm/ADT/SmallString.h" +#include "llvm/Config/config.h" +#include "llvm/Support/Errc.h" +#include "llvm/Support/Errno.h" +#include "llvm/Support/FileSystem.h" #include "llvm/Support/MathExtras.h" -#include "llvm/System/Errno.h" -#include "llvm/System/Path.h" -#include "llvm/System/Process.h" -#include "llvm/System/Program.h" +#include "llvm/Support/Path.h" +#include "llvm/Support/Process.h" +#include "llvm/Support/Program.h" #include +#include #include #include -#include +#include #include -#include +#include #if !defined(_MSC_VER) && !defined(__MINGW32__) #include -#include #else #include #endif -#include using namespace llvm; //===----------------------------------------------------------------------===// @@ -42,8 +43,10 @@ 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; } @@ -59,34 +62,45 @@ static void CopyStringRef(char *Memory, StringRef Data) { Memory[Data.size()] = 0; // Null terminate string. } -/// GetNamedBuffer - Allocates a new MemoryBuffer with Name copied after it. -template -static T* GetNamedBuffer(StringRef Buffer, StringRef Name) { - char *Mem = static_cast(operator new(sizeof(T) + Name.size() + 1)); - CopyStringRef(Mem + sizeof(T), Name); - return new (Mem) T(Buffer); +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 { public: - MemoryBufferMem(StringRef InputData) { - init(InputData.begin(), InputData.end()); + MemoryBufferMem(StringRef InputData, bool RequiresNullTerminator) { + init(InputData.begin(), InputData.end(), RequiresNullTerminator); } - virtual const char *getBufferIdentifier() const { + const char *getBufferIdentifier() const override { // The name is stored after the class itself. return reinterpret_cast(this + 1); } + + BufferKind getBufferKind() const 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! +/// that InputData must be a null terminated if RequiresNullTerminator is true! MemoryBuffer *MemoryBuffer::getMemBuffer(StringRef InputData, - StringRef BufferName) { - return GetNamedBuffer(InputData, BufferName); + StringRef BufferName, + bool RequiresNullTerminator) { + return new (NamedBufferAlloc(BufferName)) + MemoryBufferMem(InputData, RequiresNullTerminator); } /// getMemBufferCopy - Open the specified memory range as a MemoryBuffer, @@ -95,7 +109,7 @@ MemoryBuffer *MemoryBuffer::getMemBuffer(StringRef InputData, MemoryBuffer *MemoryBuffer::getMemBufferCopy(StringRef InputData, StringRef BufferName) { MemoryBuffer *Buf = getNewUninitMemBuffer(InputData.size(), BufferName); - if (!Buf) return 0; + if (!Buf) return nullptr; memcpy(const_cast(Buf->getBufferStart()), InputData.data(), InputData.size()); return Buf; @@ -109,12 +123,13 @@ MemoryBuffer *MemoryBuffer::getNewUninitMemBuffer(size_t Size, StringRef BufferName) { // Allocate space for the MemoryBuffer, the data and the name. It is important // that MemoryBuffer and data are aligned so PointerIntPair works with them. + // TODO: Is 16-byte alignment enough? We copy small object files with large + // alignment expectations into this buffer. size_t AlignedStringLen = - RoundUpToAlignment(sizeof(MemoryBufferMem) + BufferName.size() + 1, - sizeof(void*)); // TODO: Is sizeof(void*) enough? + RoundUpToAlignment(sizeof(MemoryBufferMem) + BufferName.size() + 1, 16); size_t RealLen = AlignedStringLen + Size + 1; char *Mem = static_cast(operator new(RealLen, std::nothrow)); - if (!Mem) return 0; + if (!Mem) return nullptr; // The name is stored after the class itself. CopyStringRef(Mem + sizeof(MemoryBufferMem), BufferName); @@ -123,7 +138,7 @@ MemoryBuffer *MemoryBuffer::getNewUninitMemBuffer(size_t Size, char *Buf = Mem + AlignedStringLen; Buf[Size] = 0; // Null terminate buffer. - return new (Mem) MemoryBufferMem(StringRef(Buf, Size)); + return new (Mem) MemoryBufferMem(StringRef(Buf, Size), true); } /// getNewMemBuffer - Allocate a new MemoryBuffer of the specified size that @@ -132,166 +147,265 @@ MemoryBuffer *MemoryBuffer::getNewUninitMemBuffer(size_t Size, /// the MemoryBuffer object. MemoryBuffer *MemoryBuffer::getNewMemBuffer(size_t Size, StringRef BufferName) { MemoryBuffer *SB = getNewUninitMemBuffer(Size, BufferName); - if (!SB) return 0; + if (!SB) return nullptr; memset(const_cast(SB->getBufferStart()), 0, Size); return SB; } - -/// getFileOrSTDIN - Open the specified file as a MemoryBuffer, or open stdin -/// 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) { +ErrorOr> +MemoryBuffer::getFileOrSTDIN(StringRef Filename, int64_t FileSize) { if (Filename == "-") - return getSTDIN(ErrStr); - return getFile(Filename, ErrStr, FileSize, FileInfo); + return getSTDIN(); + return getFile(Filename, FileSize); } -MemoryBuffer *MemoryBuffer::getFileOrSTDIN(const char *Filename, - std::string *ErrStr, - int64_t FileSize, - struct stat *FileInfo) { - if (strcmp(Filename, "-") == 0) - return getSTDIN(ErrStr); - return getFile(Filename, ErrStr, FileSize, FileInfo); -} //===----------------------------------------------------------------------===// // MemoryBuffer::getFile implementation. //===----------------------------------------------------------------------===// 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. -class MemoryBufferMMapFile : public MemoryBufferMem { -public: - MemoryBufferMMapFile(StringRef Buffer) - : MemoryBufferMem(Buffer) { } +/// \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 { + sys::fs::mapped_file_region MFR; + + static uint64_t getLegalMapOffset(uint64_t Offset) { + return Offset & ~(sys::fs::mapped_file_region::alignment() - 1); + } - ~MemoryBufferMMapFile() { - sys::Path::UnMapFilePages(getBufferStart(), getBufferSize()); + static uint64_t getLegalMapSize(uint64_t Len, uint64_t Offset) { + return Len + (Offset - getLegalMapOffset(Offset)); + } + + 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: - explicit FileCloser(int FD) : FD(FD) {} - ~FileCloser() { ::close(FD); } + MemoryBufferMMapFile(bool RequiresNullTerminator, int FD, uint64_t Len, + uint64_t Offset, std::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); + } + } + + const char *getBufferIdentifier() const override { + // The name is stored after the class itself. + return reinterpret_cast(this + 1); + } + + BufferKind getBufferKind() const override { + return MemoryBuffer_MMap; + } }; } -MemoryBuffer *MemoryBuffer::getFile(StringRef Filename, std::string *ErrStr, - int64_t FileSize, struct stat *FileInfo) { - SmallString<256> PathBuf(Filename.begin(), Filename.end()); - return MemoryBuffer::getFile(PathBuf.c_str(), ErrStr, FileSize, FileInfo); +static ErrorOr> +getMemoryBufferForStream(int FD, StringRef BufferName) { + 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 std::error_code(errno, std::generic_category()); + } + Buffer.set_size(Buffer.size() + ReadBytes); + } while (ReadBytes != 0); + + std::unique_ptr Ret( + MemoryBuffer::getMemBufferCopy(Buffer, BufferName)); + return std::move(Ret); } -MemoryBuffer *MemoryBuffer::getFile(const char *Filename, std::string *ErrStr, - int64_t FileSize, struct stat *FileInfo) { - int OpenFlags = O_RDONLY; -#ifdef O_BINARY - OpenFlags |= O_BINARY; // Open input file in binary mode on win32. -#endif - int FD = ::open(Filename, OpenFlags); - if (FD == -1) { - if (ErrStr) *ErrStr = sys::StrError(); - return 0; - } - FileCloser FC(FD); // Close FD on return. - +static ErrorOr> +getFileAux(const char *Filename, int64_t FileSize, bool RequiresNullTerminator, + bool IsVolatileSize); + +ErrorOr> +MemoryBuffer::getFile(Twine Filename, int64_t FileSize, + bool RequiresNullTerminator, bool IsVolatileSize) { + // Ensure the path is null terminated. + SmallString<256> PathBuf; + StringRef NullTerminatedName = Filename.toNullTerminatedStringRef(PathBuf); + return getFileAux(NullTerminatedName.data(), FileSize, RequiresNullTerminator, + IsVolatileSize); +} + +static ErrorOr> +getOpenFileImpl(int FD, const char *Filename, uint64_t FileSize, + uint64_t MapSize, int64_t Offset, bool RequiresNullTerminator, + bool IsVolatileSize); + +static ErrorOr> +getFileAux(const char *Filename, int64_t FileSize, bool RequiresNullTerminator, + bool IsVolatileSize) { + int FD; + std::error_code EC = sys::fs::openFileForRead(Filename, FD); + if (EC) + return EC; + + ErrorOr> Ret = + getOpenFileImpl(FD, Filename, FileSize, FileSize, 0, + RequiresNullTerminator, IsVolatileSize); + close(FD); + return Ret; +} + +static bool shouldUseMmap(int FD, + size_t FileSize, + size_t MapSize, + off_t Offset, + bool RequiresNullTerminator, + int PageSize, + bool IsVolatileSize) { + // mmap may leave the buffer without null terminator if the file size changed + // by the time the last page is mapped in, so avoid it if the file size is + // likely to change. + if (IsVolatileSize) + return false; + + // 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; + if (sys::fs::status(FD, Status)) + return false; + 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)) { - return GetNamedBuffer(StringRef(Pages, FileSize), - Filename); + + // 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; + + // 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 ErrorOr> +getOpenFileImpl(int FD, const char *Filename, uint64_t FileSize, + uint64_t MapSize, int64_t Offset, bool RequiresNullTerminator, + bool IsVolatileSize) { + 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; + std::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); + + FileSize = Status.getSize(); } + MapSize = FileSize; } - MemoryBuffer *Buf = MemoryBuffer::getNewUninitMemBuffer(FileSize, Filename); + if (shouldUseMmap(FD, FileSize, MapSize, Offset, RequiresNullTerminator, + PageSize, IsVolatileSize)) { + std::error_code EC; + std::unique_ptr Result( + new (NamedBufferAlloc(Filename)) + MemoryBufferMMapFile(RequiresNullTerminator, FD, MapSize, Offset, EC)); + if (!EC) + return std::move(Result); + } + + 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); + std::unique_ptr 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 std::error_code(errno, std::generic_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 std::error_code(errno, std::generic_category()); + } + if (NumRead == 0) { + memset(BufPtr, 0, BytesLeft); // zero-initialize rest of the buffer. + break; } BytesLeft -= NumRead; BufPtr += NumRead; } - return SB.take(); + return std::move(SB); } -//===----------------------------------------------------------------------===// -// MemoryBuffer::getSTDIN implementation. -//===----------------------------------------------------------------------===// +ErrorOr> +MemoryBuffer::getOpenFile(int FD, const char *Filename, uint64_t FileSize, + bool RequiresNullTerminator, bool IsVolatileSize) { + return getOpenFileImpl(FD, Filename, FileSize, FileSize, 0, + RequiresNullTerminator, IsVolatileSize); +} -MemoryBuffer *MemoryBuffer::getSTDIN(std::string *ErrStr) { +ErrorOr> +MemoryBuffer::getOpenFileSlice(int FD, const char *Filename, uint64_t MapSize, + int64_t Offset, bool IsVolatileSize) { + return getOpenFileImpl(FD, Filename, -1, MapSize, Offset, false, + IsVolatileSize); +} + +ErrorOr> MemoryBuffer::getSTDIN() { // 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(); - - 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(0, Buffer.end(), ChunkSize); - if (ReadBytes == -1) { - if (errno == EINTR) continue; - if (ErrStr) *ErrStr = sys::StrError(); - return 0; - } - Buffer.set_size(Buffer.size() + ReadBytes); - } while (ReadBytes != 0); + sys::ChangeStdinToBinary(); - return getMemBufferCopy(Buffer, ""); + return getMemoryBufferForStream(0, ""); }