1 //===- yaml2minidump.cpp - Convert a YAML file to a minidump file ---------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8
9 #include "llvm/ObjectYAML/MinidumpYAML.h"
10 #include "llvm/ObjectYAML/yaml2obj.h"
11 #include "llvm/Support/ConvertUTF.h"
12 #include "llvm/Support/raw_ostream.h"
13 #include <optional>
14
15 using namespace llvm;
16 using namespace llvm::minidump;
17 using namespace llvm::MinidumpYAML;
18
19 namespace {
20 /// A helper class to manage the placement of various structures into the final
21 /// minidump binary. Space for objects can be allocated via various allocate***
22 /// methods, while the final minidump file is written by calling the writeTo
23 /// method. The plain versions of allocation functions take a reference to the
24 /// data which is to be written (and hence the data must be available until
25 /// writeTo is called), while the "New" versions allocate the data in an
26 /// allocator-managed buffer, which is available until the allocator object is
27 /// destroyed. For both kinds of functions, it is possible to modify the
28 /// data for which the space has been "allocated" until the final writeTo call.
29 /// This is useful for "linking" the allocated structures via their offsets.
30 class BlobAllocator {
31 public:
tell() const32 size_t tell() const { return NextOffset; }
33
allocateCallback(size_t Size,std::function<void (raw_ostream &)> Callback)34 size_t allocateCallback(size_t Size,
35 std::function<void(raw_ostream &)> Callback) {
36 size_t Offset = NextOffset;
37 NextOffset += Size;
38 Callbacks.push_back(std::move(Callback));
39 return Offset;
40 }
41
allocateBytes(ArrayRef<uint8_t> Data)42 size_t allocateBytes(ArrayRef<uint8_t> Data) {
43 return allocateCallback(
44 Data.size(), [Data](raw_ostream &OS) { OS << toStringRef(Data); });
45 }
46
allocateBytes(yaml::BinaryRef Data)47 size_t allocateBytes(yaml::BinaryRef Data) {
48 return allocateCallback(Data.binary_size(), [Data](raw_ostream &OS) {
49 Data.writeAsBinary(OS);
50 });
51 }
52
allocateArray(ArrayRef<T> Data)53 template <typename T> size_t allocateArray(ArrayRef<T> Data) {
54 return allocateBytes({reinterpret_cast<const uint8_t *>(Data.data()),
55 sizeof(T) * Data.size()});
56 }
57
58 template <typename T, typename RangeType>
59 std::pair<size_t, MutableArrayRef<T>>
60 allocateNewArray(const iterator_range<RangeType> &Range);
61
allocateObject(const T & Data)62 template <typename T> size_t allocateObject(const T &Data) {
63 return allocateArray(ArrayRef(Data));
64 }
65
66 template <typename T, typename... Types>
allocateNewObject(Types &&...Args)67 std::pair<size_t, T *> allocateNewObject(Types &&... Args) {
68 T *Object = new (Temporaries.Allocate<T>()) T(std::forward<Types>(Args)...);
69 return {allocateObject(*Object), Object};
70 }
71
72 size_t allocateString(StringRef Str);
73
74 void writeTo(raw_ostream &OS) const;
75
76 private:
77 size_t NextOffset = 0;
78
79 BumpPtrAllocator Temporaries;
80 std::vector<std::function<void(raw_ostream &)>> Callbacks;
81 };
82 } // namespace
83
84 template <typename T, typename RangeType>
85 std::pair<size_t, MutableArrayRef<T>>
allocateNewArray(const iterator_range<RangeType> & Range)86 BlobAllocator::allocateNewArray(const iterator_range<RangeType> &Range) {
87 size_t Num = std::distance(Range.begin(), Range.end());
88 MutableArrayRef<T> Array(Temporaries.Allocate<T>(Num), Num);
89 std::uninitialized_copy(Range.begin(), Range.end(), Array.begin());
90 return {allocateArray(Array), Array};
91 }
92
allocateString(StringRef Str)93 size_t BlobAllocator::allocateString(StringRef Str) {
94 SmallVector<UTF16, 32> WStr;
95 bool OK = convertUTF8ToUTF16String(Str, WStr);
96 assert(OK && "Invalid UTF8 in Str?");
97 (void)OK;
98
99 // The utf16 string is null-terminated, but the terminator is not counted in
100 // the string size.
101 WStr.push_back(0);
102 size_t Result =
103 allocateNewObject<support::ulittle32_t>(2 * (WStr.size() - 1)).first;
104 allocateNewArray<support::ulittle16_t>(make_range(WStr.begin(), WStr.end()));
105 return Result;
106 }
107
writeTo(raw_ostream & OS) const108 void BlobAllocator::writeTo(raw_ostream &OS) const {
109 size_t BeginOffset = OS.tell();
110 for (const auto &Callback : Callbacks)
111 Callback(OS);
112 assert(OS.tell() == BeginOffset + NextOffset &&
113 "Callbacks wrote an unexpected number of bytes.");
114 (void)BeginOffset;
115 }
116
layout(BlobAllocator & File,yaml::BinaryRef Data)117 static LocationDescriptor layout(BlobAllocator &File, yaml::BinaryRef Data) {
118 return {support::ulittle32_t(Data.binary_size()),
119 support::ulittle32_t(File.allocateBytes(Data))};
120 }
121
layout(BlobAllocator & File,MinidumpYAML::ExceptionStream & S)122 static size_t layout(BlobAllocator &File, MinidumpYAML::ExceptionStream &S) {
123 File.allocateObject(S.MDExceptionStream);
124
125 size_t DataEnd = File.tell();
126
127 // Lay out the thread context data, (which is not a part of the stream).
128 // TODO: This usually (always?) matches the thread context of the
129 // corresponding thread, and may overlap memory regions as well. We could
130 // add a level of indirection to the MinidumpYAML format (like an array of
131 // Blobs that the LocationDescriptors index into) to be able to distinguish
132 // the cases where location descriptions overlap vs happen to reference
133 // identical data.
134 S.MDExceptionStream.ThreadContext = layout(File, S.ThreadContext);
135
136 return DataEnd;
137 }
138
layout(BlobAllocator & File,MemoryListStream::entry_type & Range)139 static void layout(BlobAllocator &File, MemoryListStream::entry_type &Range) {
140 Range.Entry.Memory = layout(File, Range.Content);
141 }
142
layout(BlobAllocator & File,ModuleListStream::entry_type & M)143 static void layout(BlobAllocator &File, ModuleListStream::entry_type &M) {
144 M.Entry.ModuleNameRVA = File.allocateString(M.Name);
145
146 M.Entry.CvRecord = layout(File, M.CvRecord);
147 M.Entry.MiscRecord = layout(File, M.MiscRecord);
148 }
149
layout(BlobAllocator & File,ThreadListStream::entry_type & T)150 static void layout(BlobAllocator &File, ThreadListStream::entry_type &T) {
151 T.Entry.Stack.Memory = layout(File, T.Stack);
152 T.Entry.Context = layout(File, T.Context);
153 }
154
155 template <typename EntryT>
layout(BlobAllocator & File,MinidumpYAML::detail::ListStream<EntryT> & S)156 static size_t layout(BlobAllocator &File,
157 MinidumpYAML::detail::ListStream<EntryT> &S) {
158
159 File.allocateNewObject<support::ulittle32_t>(S.Entries.size());
160 for (auto &E : S.Entries)
161 File.allocateObject(E.Entry);
162
163 size_t DataEnd = File.tell();
164
165 // Lay out the auxiliary data, (which is not a part of the stream).
166 DataEnd = File.tell();
167 for (auto &E : S.Entries)
168 layout(File, E);
169
170 return DataEnd;
171 }
172
layout(BlobAllocator & File,Stream & S)173 static Directory layout(BlobAllocator &File, Stream &S) {
174 Directory Result;
175 Result.Type = S.Type;
176 Result.Location.RVA = File.tell();
177 std::optional<size_t> DataEnd;
178 switch (S.Kind) {
179 case Stream::StreamKind::Exception:
180 DataEnd = layout(File, cast<MinidumpYAML::ExceptionStream>(S));
181 break;
182 case Stream::StreamKind::MemoryInfoList: {
183 MemoryInfoListStream &InfoList = cast<MemoryInfoListStream>(S);
184 File.allocateNewObject<minidump::MemoryInfoListHeader>(
185 sizeof(minidump::MemoryInfoListHeader), sizeof(minidump::MemoryInfo),
186 InfoList.Infos.size());
187 File.allocateArray(ArrayRef(InfoList.Infos));
188 break;
189 }
190 case Stream::StreamKind::MemoryList:
191 DataEnd = layout(File, cast<MemoryListStream>(S));
192 break;
193 case Stream::StreamKind::ModuleList:
194 DataEnd = layout(File, cast<ModuleListStream>(S));
195 break;
196 case Stream::StreamKind::RawContent: {
197 RawContentStream &Raw = cast<RawContentStream>(S);
198 File.allocateCallback(Raw.Size, [&Raw](raw_ostream &OS) {
199 Raw.Content.writeAsBinary(OS);
200 assert(Raw.Content.binary_size() <= Raw.Size);
201 OS << std::string(Raw.Size - Raw.Content.binary_size(), '\0');
202 });
203 break;
204 }
205 case Stream::StreamKind::SystemInfo: {
206 SystemInfoStream &SystemInfo = cast<SystemInfoStream>(S);
207 File.allocateObject(SystemInfo.Info);
208 // The CSD string is not a part of the stream.
209 DataEnd = File.tell();
210 SystemInfo.Info.CSDVersionRVA = File.allocateString(SystemInfo.CSDVersion);
211 break;
212 }
213 case Stream::StreamKind::TextContent:
214 File.allocateArray(arrayRefFromStringRef(cast<TextContentStream>(S).Text));
215 break;
216 case Stream::StreamKind::ThreadList:
217 DataEnd = layout(File, cast<ThreadListStream>(S));
218 break;
219 }
220 // If DataEnd is not set, we assume everything we generated is a part of the
221 // stream.
222 Result.Location.DataSize =
223 DataEnd.value_or(File.tell()) - Result.Location.RVA;
224 return Result;
225 }
226
227 namespace llvm {
228 namespace yaml {
229
yaml2minidump(MinidumpYAML::Object & Obj,raw_ostream & Out,ErrorHandler)230 bool yaml2minidump(MinidumpYAML::Object &Obj, raw_ostream &Out,
231 ErrorHandler /*EH*/) {
232 BlobAllocator File;
233 File.allocateObject(Obj.Header);
234
235 std::vector<Directory> StreamDirectory(Obj.Streams.size());
236 Obj.Header.StreamDirectoryRVA = File.allocateArray(ArrayRef(StreamDirectory));
237 Obj.Header.NumberOfStreams = StreamDirectory.size();
238
239 for (const auto &[Index, Stream] : enumerate(Obj.Streams))
240 StreamDirectory[Index] = layout(File, *Stream);
241
242 File.writeTo(Out);
243 return true;
244 }
245
246 } // namespace yaml
247 } // namespace llvm
248