blob: bbfd2cd8cbab5e254246a2dbdee5fffa4cc81a23 [file] [log] [blame]
//===- yaml2minidump.cpp - Convert a YAML file to a minidump file ---------===//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
#include "llvm/ObjectYAML/MinidumpYAML.h"
#include "llvm/ObjectYAML/yaml2obj.h"
#include "llvm/Support/ConvertUTF.h"
#include "llvm/Support/raw_ostream.h"
using namespace llvm;
using namespace llvm::minidump;
using namespace llvm::MinidumpYAML;
namespace {
/// A helper class to manage the placement of various structures into the final
/// minidump binary. Space for objects can be allocated via various allocate***
/// methods, while the final minidump file is written by calling the writeTo
/// method. The plain versions of allocation functions take a reference to the
/// data which is to be written (and hence the data must be available until
/// writeTo is called), while the "New" versions allocate the data in an
/// allocator-managed buffer, which is available until the allocator object is
/// destroyed. For both kinds of functions, it is possible to modify the
/// data for which the space has been "allocated" until the final writeTo call.
/// This is useful for "linking" the allocated structures via their offsets.
class BlobAllocator {
size_t tell() const { return NextOffset; }
size_t allocateCallback(size_t Size,
std::function<void(raw_ostream &)> Callback) {
size_t Offset = NextOffset;
NextOffset += Size;
return Offset;
size_t allocateBytes(ArrayRef<uint8_t> Data) {
return allocateCallback(
Data.size(), [Data](raw_ostream &OS) { OS << toStringRef(Data); });
size_t allocateBytes(yaml::BinaryRef Data) {
return allocateCallback(Data.binary_size(), [Data](raw_ostream &OS) {
template <typename T> size_t allocateArray(ArrayRef<T> Data) {
return allocateBytes({reinterpret_cast<const uint8_t *>(,
sizeof(T) * Data.size()});
template <typename T, typename RangeType>
std::pair<size_t, MutableArrayRef<T>>
allocateNewArray(const iterator_range<RangeType> &Range);
template <typename T> size_t allocateObject(const T &Data) {
return allocateArray(makeArrayRef(Data));
template <typename T, typename... Types>
std::pair<size_t, T *> allocateNewObject(Types &&... Args) {
T *Object = new (Temporaries.Allocate<T>()) T(std::forward<Types>(Args)...);
return {allocateObject(*Object), Object};
size_t allocateString(StringRef Str);
void writeTo(raw_ostream &OS) const;
size_t NextOffset = 0;
BumpPtrAllocator Temporaries;
std::vector<std::function<void(raw_ostream &)>> Callbacks;
} // namespace
template <typename T, typename RangeType>
std::pair<size_t, MutableArrayRef<T>>
BlobAllocator::allocateNewArray(const iterator_range<RangeType> &Range) {
size_t Num = std::distance(Range.begin(), Range.end());
MutableArrayRef<T> Array(Temporaries.Allocate<T>(Num), Num);
std::uninitialized_copy(Range.begin(), Range.end(), Array.begin());
return {allocateArray(Array), Array};
size_t BlobAllocator::allocateString(StringRef Str) {
SmallVector<UTF16, 32> WStr;
bool OK = convertUTF8ToUTF16String(Str, WStr);
assert(OK && "Invalid UTF8 in Str?");
// The utf16 string is null-terminated, but the terminator is not counted in
// the string size.
size_t Result =
allocateNewObject<support::ulittle32_t>(2 * (WStr.size() - 1)).first;
allocateNewArray<support::ulittle16_t>(make_range(WStr.begin(), WStr.end()));
return Result;
void BlobAllocator::writeTo(raw_ostream &OS) const {
size_t BeginOffset = OS.tell();
for (const auto &Callback : Callbacks)
assert(OS.tell() == BeginOffset + NextOffset &&
"Callbacks wrote an unexpected number of bytes.");
static LocationDescriptor layout(BlobAllocator &File, yaml::BinaryRef Data) {
return {support::ulittle32_t(Data.binary_size()),
static size_t layout(BlobAllocator &File, MinidumpYAML::ExceptionStream &S) {
size_t DataEnd = File.tell();
// Lay out the thread context data, (which is not a part of the stream).
// TODO: This usually (always?) matches the thread context of the
// corresponding thread, and may overlap memory regions as well. We could
// add a level of indirection to the MinidumpYAML format (like an array of
// Blobs that the LocationDescriptors index into) to be able to distinguish
// the cases where location descriptions overlap vs happen to reference
// identical data.
S.MDExceptionStream.ThreadContext = layout(File, S.ThreadContext);
return DataEnd;
static void layout(BlobAllocator &File, MemoryListStream::entry_type &Range) {
Range.Entry.Memory = layout(File, Range.Content);
static void layout(BlobAllocator &File, ModuleListStream::entry_type &M) {
M.Entry.ModuleNameRVA = File.allocateString(M.Name);
M.Entry.CvRecord = layout(File, M.CvRecord);
M.Entry.MiscRecord = layout(File, M.MiscRecord);
static void layout(BlobAllocator &File, ThreadListStream::entry_type &T) {
T.Entry.Stack.Memory = layout(File, T.Stack);
T.Entry.Context = layout(File, T.Context);
template <typename EntryT>
static size_t layout(BlobAllocator &File,
MinidumpYAML::detail::ListStream<EntryT> &S) {
for (auto &E : S.Entries)
size_t DataEnd = File.tell();
// Lay out the auxiliary data, (which is not a part of the stream).
DataEnd = File.tell();
for (auto &E : S.Entries)
layout(File, E);
return DataEnd;
static Directory layout(BlobAllocator &File, Stream &S) {
Directory Result;
Result.Type = S.Type;
Result.Location.RVA = File.tell();
Optional<size_t> DataEnd;
switch (S.Kind) {
case Stream::StreamKind::Exception:
DataEnd = layout(File, cast<MinidumpYAML::ExceptionStream>(S));
case Stream::StreamKind::MemoryInfoList: {
MemoryInfoListStream &InfoList = cast<MemoryInfoListStream>(S);
sizeof(minidump::MemoryInfoListHeader), sizeof(minidump::MemoryInfo),
case Stream::StreamKind::MemoryList:
DataEnd = layout(File, cast<MemoryListStream>(S));
case Stream::StreamKind::ModuleList:
DataEnd = layout(File, cast<ModuleListStream>(S));
case Stream::StreamKind::RawContent: {
RawContentStream &Raw = cast<RawContentStream>(S);
File.allocateCallback(Raw.Size, [&Raw](raw_ostream &OS) {
assert(Raw.Content.binary_size() <= Raw.Size);
OS << std::string(Raw.Size - Raw.Content.binary_size(), '\0');
case Stream::StreamKind::SystemInfo: {
SystemInfoStream &SystemInfo = cast<SystemInfoStream>(S);
// The CSD string is not a part of the stream.
DataEnd = File.tell();
SystemInfo.Info.CSDVersionRVA = File.allocateString(SystemInfo.CSDVersion);
case Stream::StreamKind::TextContent:
case Stream::StreamKind::ThreadList:
DataEnd = layout(File, cast<ThreadListStream>(S));
// If DataEnd is not set, we assume everything we generated is a part of the
// stream.
Result.Location.DataSize =
DataEnd.getValueOr(File.tell()) - Result.Location.RVA;
return Result;
namespace llvm {
namespace yaml {
bool yaml2minidump(MinidumpYAML::Object &Obj, raw_ostream &Out,
ErrorHandler /*EH*/) {
BlobAllocator File;
std::vector<Directory> StreamDirectory(Obj.Streams.size());
Obj.Header.StreamDirectoryRVA =
Obj.Header.NumberOfStreams = StreamDirectory.size();
for (auto &Stream : enumerate(Obj.Streams))
StreamDirectory[Stream.index()] = layout(File, *Stream.value());
return true;
} // namespace yaml
} // namespace llvm