1//===- Minidump.cpp - Minidump object file implementation -----------------===//
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/Object/Minidump.h"
10#include "llvm/Object/Error.h"
11#include "llvm/Support/ConvertUTF.h"
12
13using namespace llvm;
14using namespace llvm::object;
15using namespace llvm::minidump;
16
17std::optional<ArrayRef<uint8_t>>
18MinidumpFile::getRawStream(minidump::StreamType Type) const {
19 auto It = StreamMap.find(Val: Type);
20 if (It != StreamMap.end())
21 return getRawStream(Stream: Streams[It->second]);
22 return std::nullopt;
23}
24
25Expected<std::string> MinidumpFile::getString(size_t Offset) const {
26 // Minidump strings consist of a 32-bit length field, which gives the size of
27 // the string in *bytes*. This is followed by the actual string encoded in
28 // UTF16.
29 auto ExpectedSize =
30 getDataSliceAs<support::ulittle32_t>(Data: getData(), Offset, Count: 1);
31 if (!ExpectedSize)
32 return ExpectedSize.takeError();
33 size_t Size = (*ExpectedSize)[0];
34 if (Size % 2 != 0)
35 return createError(Str: "String size not even");
36 Size /= 2;
37 if (Size == 0)
38 return "";
39
40 Offset += sizeof(support::ulittle32_t);
41 auto ExpectedData =
42 getDataSliceAs<support::ulittle16_t>(Data: getData(), Offset, Count: Size);
43 if (!ExpectedData)
44 return ExpectedData.takeError();
45
46 SmallVector<UTF16, 32> WStr(Size);
47 copy(Range&: *ExpectedData, Out: WStr.begin());
48
49 std::string Result;
50 if (!convertUTF16ToUTF8String(Src: WStr, Out&: Result))
51 return createError(Str: "String decoding failed");
52
53 return Result;
54}
55
56Expected<iterator_range<MinidumpFile::MemoryInfoIterator>>
57MinidumpFile::getMemoryInfoList() const {
58 std::optional<ArrayRef<uint8_t>> Stream =
59 getRawStream(Type: StreamType::MemoryInfoList);
60 if (!Stream)
61 return createError(Str: "No such stream");
62 auto ExpectedHeader =
63 getDataSliceAs<minidump::MemoryInfoListHeader>(Data: *Stream, Offset: 0, Count: 1);
64 if (!ExpectedHeader)
65 return ExpectedHeader.takeError();
66 const minidump::MemoryInfoListHeader &H = ExpectedHeader.get()[0];
67 Expected<ArrayRef<uint8_t>> Data =
68 getDataSlice(Data: *Stream, Offset: H.SizeOfHeader, Size: H.SizeOfEntry * H.NumberOfEntries);
69 if (!Data)
70 return Data.takeError();
71 return make_range(x: MemoryInfoIterator(*Data, H.SizeOfEntry),
72 y: MemoryInfoIterator({}, H.SizeOfEntry));
73}
74
75template <typename T>
76Expected<ArrayRef<T>> MinidumpFile::getListStream(StreamType Type) const {
77 std::optional<ArrayRef<uint8_t>> Stream = getRawStream(Type);
78 if (!Stream)
79 return createError(Str: "No such stream");
80 auto ExpectedSize = getDataSliceAs<support::ulittle32_t>(Data: *Stream, Offset: 0, Count: 1);
81 if (!ExpectedSize)
82 return ExpectedSize.takeError();
83
84 size_t ListSize = ExpectedSize.get()[0];
85
86 size_t ListOffset = 4;
87 // Some producers insert additional padding bytes to align the list to an
88 // 8-byte boundary. Check for that by comparing the list size with the overall
89 // stream size.
90 if (ListOffset + sizeof(T) * ListSize < Stream->size())
91 ListOffset = 8;
92
93 return getDataSliceAs<T>(*Stream, ListOffset, ListSize);
94}
95template Expected<ArrayRef<Module>>
96 MinidumpFile::getListStream(StreamType) const;
97template Expected<ArrayRef<Thread>>
98 MinidumpFile::getListStream(StreamType) const;
99template Expected<ArrayRef<MemoryDescriptor>>
100 MinidumpFile::getListStream(StreamType) const;
101
102Expected<ArrayRef<uint8_t>>
103MinidumpFile::getDataSlice(ArrayRef<uint8_t> Data, size_t Offset, size_t Size) {
104 // Check for overflow.
105 if (Offset + Size < Offset || Offset + Size < Size ||
106 Offset + Size > Data.size())
107 return createEOFError();
108 return Data.slice(N: Offset, M: Size);
109}
110
111Expected<std::unique_ptr<MinidumpFile>>
112MinidumpFile::create(MemoryBufferRef Source) {
113 ArrayRef<uint8_t> Data = arrayRefFromStringRef(Input: Source.getBuffer());
114 auto ExpectedHeader = getDataSliceAs<minidump::Header>(Data, Offset: 0, Count: 1);
115 if (!ExpectedHeader)
116 return ExpectedHeader.takeError();
117
118 const minidump::Header &Hdr = (*ExpectedHeader)[0];
119 if (Hdr.Signature != Header::MagicSignature)
120 return createError(Str: "Invalid signature");
121 if ((Hdr.Version & 0xffff) != Header::MagicVersion)
122 return createError(Str: "Invalid version");
123
124 auto ExpectedStreams = getDataSliceAs<Directory>(Data, Offset: Hdr.StreamDirectoryRVA,
125 Count: Hdr.NumberOfStreams);
126 if (!ExpectedStreams)
127 return ExpectedStreams.takeError();
128
129 DenseMap<StreamType, std::size_t> StreamMap;
130 for (const auto &StreamDescriptor : llvm::enumerate(First&: *ExpectedStreams)) {
131 StreamType Type = StreamDescriptor.value().Type;
132 const LocationDescriptor &Loc = StreamDescriptor.value().Location;
133
134 Expected<ArrayRef<uint8_t>> Stream =
135 getDataSlice(Data, Offset: Loc.RVA, Size: Loc.DataSize);
136 if (!Stream)
137 return Stream.takeError();
138
139 if (Type == StreamType::Unused && Loc.DataSize == 0) {
140 // Ignore dummy streams. This is technically ill-formed, but a number of
141 // existing minidumps seem to contain such streams.
142 continue;
143 }
144
145 if (Type == DenseMapInfo<StreamType>::getEmptyKey() ||
146 Type == DenseMapInfo<StreamType>::getTombstoneKey())
147 return createError(Str: "Cannot handle one of the minidump streams");
148
149 // Update the directory map, checking for duplicate stream types.
150 if (!StreamMap.try_emplace(Key: Type, Args: StreamDescriptor.index()).second)
151 return createError(Str: "Duplicate stream type");
152 }
153
154 return std::unique_ptr<MinidumpFile>(
155 new MinidumpFile(Source, Hdr, *ExpectedStreams, std::move(StreamMap)));
156}
157