1//===--- HeaderMap.cpp - A file that acts like dir of symlinks ------------===//
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// This file implements the HeaderMap interface.
10//
11//===----------------------------------------------------------------------===//
12
13#include "clang/Lex/HeaderMap.h"
14#include "clang/Lex/HeaderMapTypes.h"
15#include "clang/Basic/CharInfo.h"
16#include "clang/Basic/FileManager.h"
17#include "llvm/ADT/SmallString.h"
18#include "llvm/Support/Compiler.h"
19#include "llvm/Support/DataTypes.h"
20#include "llvm/Support/MathExtras.h"
21#include "llvm/Support/MemoryBuffer.h"
22#include "llvm/Support/SwapByteOrder.h"
23#include "llvm/Support/Debug.h"
24#include <cstring>
25#include <memory>
26#include <optional>
27using namespace clang;
28
29/// HashHMapKey - This is the 'well known' hash function required by the file
30/// format, used to look up keys in the hash table.  The hash table uses simple
31/// linear probing based on this function.
32static inline unsigned HashHMapKey(StringRef Str) {
33  unsigned Result = 0;
34  const char *S = Str.begin(), *End = Str.end();
35
36  for (; S != End; S++)
37    Result += toLowercase(*S) * 13;
38  return Result;
39}
40
41
42
43//===----------------------------------------------------------------------===//
44// Verification and Construction
45//===----------------------------------------------------------------------===//
46
47/// HeaderMap::Create - This attempts to load the specified file as a header
48/// map.  If it doesn't look like a HeaderMap, it gives up and returns null.
49/// If it looks like a HeaderMap but is obviously corrupted, it puts a reason
50/// into the string error argument and returns null.
51std::unique_ptr<HeaderMap> HeaderMap::Create(const FileEntry *FE,
52                                             FileManager &FM) {
53  // If the file is too small to be a header map, ignore it.
54  unsigned FileSize = FE->getSize();
55  if (FileSize <= sizeof(HMapHeader)) return nullptr;
56
57  auto FileBuffer = FM.getBufferForFile(FE);
58  if (!FileBuffer || !*FileBuffer)
59    return nullptr;
60  bool NeedsByteSwap;
61  if (!checkHeader(**FileBuffer, NeedsByteSwap))
62    return nullptr;
63  return std::unique_ptr<HeaderMap>(new HeaderMap(std::move(*FileBuffer), NeedsByteSwap));
64}
65
66bool HeaderMapImpl::checkHeader(const llvm::MemoryBuffer &File,
67                                bool &NeedsByteSwap) {
68  if (File.getBufferSize() <= sizeof(HMapHeader))
69    return false;
70  const char *FileStart = File.getBufferStart();
71
72  // We know the file is at least as big as the header, check it now.
73  const HMapHeader *Header = reinterpret_cast<const HMapHeader*>(FileStart);
74
75  // Sniff it to see if it's a headermap by checking the magic number and
76  // version.
77  if (Header->Magic == HMAP_HeaderMagicNumber &&
78      Header->Version == HMAP_HeaderVersion)
79    NeedsByteSwap = false;
80  else if (Header->Magic == llvm::ByteSwap_32(HMAP_HeaderMagicNumber) &&
81           Header->Version == llvm::ByteSwap_16(HMAP_HeaderVersion))
82    NeedsByteSwap = true;  // Mixed endianness headermap.
83  else
84    return false;  // Not a header map.
85
86  if (Header->Reserved != 0)
87    return false;
88
89  // Check the number of buckets.  It should be a power of two, and there
90  // should be enough space in the file for all of them.
91  uint32_t NumBuckets = NeedsByteSwap
92                            ? llvm::sys::getSwappedBytes(Header->NumBuckets)
93                            : Header->NumBuckets;
94  if (!llvm::isPowerOf2_32(NumBuckets))
95    return false;
96  if (File.getBufferSize() <
97      sizeof(HMapHeader) + sizeof(HMapBucket) * NumBuckets)
98    return false;
99
100  // Okay, everything looks good.
101  return true;
102}
103
104//===----------------------------------------------------------------------===//
105//  Utility Methods
106//===----------------------------------------------------------------------===//
107
108
109/// getFileName - Return the filename of the headermap.
110StringRef HeaderMapImpl::getFileName() const {
111  return FileBuffer->getBufferIdentifier();
112}
113
114unsigned HeaderMapImpl::getEndianAdjustedWord(unsigned X) const {
115  if (!NeedsBSwap) return X;
116  return llvm::ByteSwap_32(X);
117}
118
119/// getHeader - Return a reference to the file header, in unbyte-swapped form.
120/// This method cannot fail.
121const HMapHeader &HeaderMapImpl::getHeader() const {
122  // We know the file is at least as big as the header.  Return it.
123  return *reinterpret_cast<const HMapHeader*>(FileBuffer->getBufferStart());
124}
125
126/// getBucket - Return the specified hash table bucket from the header map,
127/// bswap'ing its fields as appropriate.  If the bucket number is not valid,
128/// this return a bucket with an empty key (0).
129HMapBucket HeaderMapImpl::getBucket(unsigned BucketNo) const {
130  assert(FileBuffer->getBufferSize() >=
131             sizeof(HMapHeader) + sizeof(HMapBucket) * BucketNo &&
132         "Expected bucket to be in range");
133
134  HMapBucket Result;
135  Result.Key = HMAP_EmptyBucketKey;
136
137  const HMapBucket *BucketArray =
138    reinterpret_cast<const HMapBucket*>(FileBuffer->getBufferStart() +
139                                        sizeof(HMapHeader));
140  const HMapBucket *BucketPtr = BucketArray+BucketNo;
141
142  // Load the values, bswapping as needed.
143  Result.Key    = getEndianAdjustedWord(BucketPtr->Key);
144  Result.Prefix = getEndianAdjustedWord(BucketPtr->Prefix);
145  Result.Suffix = getEndianAdjustedWord(BucketPtr->Suffix);
146  return Result;
147}
148
149std::optional<StringRef> HeaderMapImpl::getString(unsigned StrTabIdx) const {
150  // Add the start of the string table to the idx.
151  StrTabIdx += getEndianAdjustedWord(getHeader().StringsOffset);
152
153  // Check for invalid index.
154  if (StrTabIdx >= FileBuffer->getBufferSize())
155    return std::nullopt;
156
157  const char *Data = FileBuffer->getBufferStart() + StrTabIdx;
158  unsigned MaxLen = FileBuffer->getBufferSize() - StrTabIdx;
159  unsigned Len = strnlen(Data, MaxLen);
160
161  // Check whether the buffer is null-terminated.
162  if (Len == MaxLen && Data[Len - 1])
163    return std::nullopt;
164
165  return StringRef(Data, Len);
166}
167
168//===----------------------------------------------------------------------===//
169// The Main Drivers
170//===----------------------------------------------------------------------===//
171
172/// dump - Print the contents of this headermap to stderr.
173LLVM_DUMP_METHOD void HeaderMapImpl::dump() const {
174  const HMapHeader &Hdr = getHeader();
175  unsigned NumBuckets = getEndianAdjustedWord(Hdr.NumBuckets);
176
177  llvm::dbgs() << "Header Map " << getFileName() << ":\n  " << NumBuckets
178               << ", " << getEndianAdjustedWord(Hdr.NumEntries) << "\n";
179
180  auto getStringOrInvalid = [this](unsigned Id) -> StringRef {
181    if (std::optional<StringRef> S = getString(Id))
182      return *S;
183    return "<invalid>";
184  };
185
186  for (unsigned i = 0; i != NumBuckets; ++i) {
187    HMapBucket B = getBucket(i);
188    if (B.Key == HMAP_EmptyBucketKey) continue;
189
190    StringRef Key = getStringOrInvalid(B.Key);
191    StringRef Prefix = getStringOrInvalid(B.Prefix);
192    StringRef Suffix = getStringOrInvalid(B.Suffix);
193    llvm::dbgs() << "  " << i << ". " << Key << " -> '" << Prefix << "' '"
194                 << Suffix << "'\n";
195  }
196}
197
198StringRef HeaderMapImpl::lookupFilename(StringRef Filename,
199                                        SmallVectorImpl<char> &DestPath) const {
200  const HMapHeader &Hdr = getHeader();
201  unsigned NumBuckets = getEndianAdjustedWord(Hdr.NumBuckets);
202
203  // Don't probe infinitely.  This should be checked before constructing.
204  assert(llvm::isPowerOf2_32(NumBuckets) && "Expected power of 2");
205
206  // Linearly probe the hash table.
207  for (unsigned Bucket = HashHMapKey(Filename);; ++Bucket) {
208    HMapBucket B = getBucket(Bucket & (NumBuckets-1));
209    if (B.Key == HMAP_EmptyBucketKey) return StringRef(); // Hash miss.
210
211    // See if the key matches.  If not, probe on.
212    std::optional<StringRef> Key = getString(B.Key);
213    if (LLVM_UNLIKELY(!Key))
214      continue;
215    if (!Filename.equals_insensitive(*Key))
216      continue;
217
218    // If so, we have a match in the hash table.  Construct the destination
219    // path.
220    std::optional<StringRef> Prefix = getString(B.Prefix);
221    std::optional<StringRef> Suffix = getString(B.Suffix);
222
223    DestPath.clear();
224    if (LLVM_LIKELY(Prefix && Suffix)) {
225      DestPath.append(Prefix->begin(), Prefix->end());
226      DestPath.append(Suffix->begin(), Suffix->end());
227    }
228    return StringRef(DestPath.begin(), DestPath.size());
229  }
230}
231
232StringRef HeaderMapImpl::reverseLookupFilename(StringRef DestPath) const {
233  if (!ReverseMap.empty())
234    return ReverseMap.lookup(DestPath);
235
236  const HMapHeader &Hdr = getHeader();
237  unsigned NumBuckets = getEndianAdjustedWord(Hdr.NumBuckets);
238  StringRef RetKey;
239  for (unsigned i = 0; i != NumBuckets; ++i) {
240    HMapBucket B = getBucket(i);
241    if (B.Key == HMAP_EmptyBucketKey)
242      continue;
243
244    std::optional<StringRef> Key = getString(B.Key);
245    std::optional<StringRef> Prefix = getString(B.Prefix);
246    std::optional<StringRef> Suffix = getString(B.Suffix);
247    if (LLVM_LIKELY(Key && Prefix && Suffix)) {
248      SmallVector<char, 1024> Buf;
249      Buf.append(Prefix->begin(), Prefix->end());
250      Buf.append(Suffix->begin(), Suffix->end());
251      StringRef Value(Buf.begin(), Buf.size());
252      ReverseMap[Value] = *Key;
253
254      if (DestPath == Value)
255        RetKey = *Key;
256    }
257  }
258  return RetKey;
259}
260