1 | //===- FuzzerMutate.cpp - Mutate a test input -----------------------------===// |
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 | // Mutate a test input. |
9 | //===----------------------------------------------------------------------===// |
10 | |
11 | #include "FuzzerDefs.h" |
12 | #include "FuzzerExtFunctions.h" |
13 | #include "FuzzerIO.h" |
14 | #include "FuzzerMutate.h" |
15 | #include "FuzzerOptions.h" |
16 | #include "FuzzerTracePC.h" |
17 | |
18 | namespace fuzzer { |
19 | |
20 | const size_t Dictionary::kMaxDictSize; |
21 | static const size_t kMaxMutationsToPrint = 10; |
22 | |
23 | static void PrintASCII(const Word &W, const char *PrintAfter) { |
24 | PrintASCII(Data: W.data(), Size: W.size(), PrintAfter); |
25 | } |
26 | |
27 | MutationDispatcher::MutationDispatcher(Random &Rand, |
28 | const FuzzingOptions &Options) |
29 | : Rand(Rand), Options(Options) { |
30 | DefaultMutators.insert( |
31 | position: DefaultMutators.begin(), |
32 | l: { |
33 | {.Fn: &MutationDispatcher::Mutate_EraseBytes, .Name: "EraseBytes" }, |
34 | {.Fn: &MutationDispatcher::Mutate_InsertByte, .Name: "InsertByte" }, |
35 | {.Fn: &MutationDispatcher::Mutate_InsertRepeatedBytes, |
36 | .Name: "InsertRepeatedBytes" }, |
37 | {.Fn: &MutationDispatcher::Mutate_ChangeByte, .Name: "ChangeByte" }, |
38 | {.Fn: &MutationDispatcher::Mutate_ChangeBit, .Name: "ChangeBit" }, |
39 | {.Fn: &MutationDispatcher::Mutate_ShuffleBytes, .Name: "ShuffleBytes" }, |
40 | {.Fn: &MutationDispatcher::Mutate_ChangeASCIIInteger, .Name: "ChangeASCIIInt" }, |
41 | {.Fn: &MutationDispatcher::Mutate_ChangeBinaryInteger, .Name: "ChangeBinInt" }, |
42 | {.Fn: &MutationDispatcher::Mutate_CopyPart, .Name: "CopyPart" }, |
43 | {.Fn: &MutationDispatcher::Mutate_CrossOver, .Name: "CrossOver" }, |
44 | {.Fn: &MutationDispatcher::Mutate_AddWordFromManualDictionary, |
45 | .Name: "ManualDict" }, |
46 | {.Fn: &MutationDispatcher::Mutate_AddWordFromPersistentAutoDictionary, |
47 | .Name: "PersAutoDict" }, |
48 | }); |
49 | if(Options.UseCmp) |
50 | DefaultMutators.push_back( |
51 | x: {.Fn: &MutationDispatcher::Mutate_AddWordFromTORC, .Name: "CMP" }); |
52 | |
53 | if (EF->LLVMFuzzerCustomMutator) |
54 | Mutators.push_back(x: {.Fn: &MutationDispatcher::Mutate_Custom, .Name: "Custom" }); |
55 | else |
56 | Mutators = DefaultMutators; |
57 | |
58 | if (EF->LLVMFuzzerCustomCrossOver) |
59 | Mutators.push_back( |
60 | x: {.Fn: &MutationDispatcher::Mutate_CustomCrossOver, .Name: "CustomCrossOver" }); |
61 | } |
62 | |
63 | static char RandCh(Random &Rand) { |
64 | if (Rand.RandBool()) |
65 | return static_cast<char>(Rand(256)); |
66 | const char Special[] = "!*'();:@&=+$,/?%#[]012Az-`~.\xff\x00" ; |
67 | return Special[Rand(sizeof(Special) - 1)]; |
68 | } |
69 | |
70 | size_t MutationDispatcher::Mutate_Custom(uint8_t *Data, size_t Size, |
71 | size_t MaxSize) { |
72 | if (EF->__msan_unpoison) |
73 | EF->__msan_unpoison(Data, Size); |
74 | if (EF->__msan_unpoison_param) |
75 | EF->__msan_unpoison_param(4); |
76 | return EF->LLVMFuzzerCustomMutator(Data, Size, MaxSize, |
77 | Rand.Rand<unsigned int>()); |
78 | } |
79 | |
80 | size_t MutationDispatcher::Mutate_CustomCrossOver(uint8_t *Data, size_t Size, |
81 | size_t MaxSize) { |
82 | if (Size == 0) |
83 | return 0; |
84 | if (!CrossOverWith) return 0; |
85 | const Unit &Other = *CrossOverWith; |
86 | if (Other.empty()) |
87 | return 0; |
88 | CustomCrossOverInPlaceHere.resize(new_size: MaxSize); |
89 | auto &U = CustomCrossOverInPlaceHere; |
90 | |
91 | if (EF->__msan_unpoison) { |
92 | EF->__msan_unpoison(Data, Size); |
93 | EF->__msan_unpoison(Other.data(), Other.size()); |
94 | EF->__msan_unpoison(U.data(), U.size()); |
95 | } |
96 | if (EF->__msan_unpoison_param) |
97 | EF->__msan_unpoison_param(7); |
98 | size_t NewSize = EF->LLVMFuzzerCustomCrossOver( |
99 | Data, Size, Other.data(), Other.size(), U.data(), U.size(), |
100 | Rand.Rand<unsigned int>()); |
101 | |
102 | if (!NewSize) |
103 | return 0; |
104 | assert(NewSize <= MaxSize && "CustomCrossOver returned overisized unit" ); |
105 | memcpy(dest: Data, src: U.data(), n: NewSize); |
106 | return NewSize; |
107 | } |
108 | |
109 | size_t MutationDispatcher::Mutate_ShuffleBytes(uint8_t *Data, size_t Size, |
110 | size_t MaxSize) { |
111 | if (Size > MaxSize || Size == 0) return 0; |
112 | size_t ShuffleAmount = |
113 | Rand(std::min(a: Size, b: (size_t)8)) + 1; // [1,8] and <= Size. |
114 | size_t ShuffleStart = Rand(Size - ShuffleAmount); |
115 | assert(ShuffleStart + ShuffleAmount <= Size); |
116 | std::shuffle(first: Data + ShuffleStart, last: Data + ShuffleStart + ShuffleAmount, g&: Rand); |
117 | return Size; |
118 | } |
119 | |
120 | size_t MutationDispatcher::Mutate_EraseBytes(uint8_t *Data, size_t Size, |
121 | size_t MaxSize) { |
122 | if (Size <= 1) return 0; |
123 | size_t N = Rand(Size / 2) + 1; |
124 | assert(N < Size); |
125 | size_t Idx = Rand(Size - N + 1); |
126 | // Erase Data[Idx:Idx+N]. |
127 | memmove(dest: Data + Idx, src: Data + Idx + N, n: Size - Idx - N); |
128 | // Printf("Erase: %zd %zd => %zd; Idx %zd\n", N, Size, Size - N, Idx); |
129 | return Size - N; |
130 | } |
131 | |
132 | size_t MutationDispatcher::Mutate_InsertByte(uint8_t *Data, size_t Size, |
133 | size_t MaxSize) { |
134 | if (Size >= MaxSize) return 0; |
135 | size_t Idx = Rand(Size + 1); |
136 | // Insert new value at Data[Idx]. |
137 | memmove(dest: Data + Idx + 1, src: Data + Idx, n: Size - Idx); |
138 | Data[Idx] = RandCh(Rand); |
139 | return Size + 1; |
140 | } |
141 | |
142 | size_t MutationDispatcher::Mutate_InsertRepeatedBytes(uint8_t *Data, |
143 | size_t Size, |
144 | size_t MaxSize) { |
145 | const size_t kMinBytesToInsert = 3; |
146 | if (Size + kMinBytesToInsert >= MaxSize) return 0; |
147 | size_t MaxBytesToInsert = std::min(a: MaxSize - Size, b: (size_t)128); |
148 | size_t N = Rand(MaxBytesToInsert - kMinBytesToInsert + 1) + kMinBytesToInsert; |
149 | assert(Size + N <= MaxSize && N); |
150 | size_t Idx = Rand(Size + 1); |
151 | // Insert new values at Data[Idx]. |
152 | memmove(dest: Data + Idx + N, src: Data + Idx, n: Size - Idx); |
153 | // Give preference to 0x00 and 0xff. |
154 | uint8_t Byte = static_cast<uint8_t>( |
155 | Rand.RandBool() ? Rand(256) : (Rand.RandBool() ? 0 : 255)); |
156 | for (size_t i = 0; i < N; i++) |
157 | Data[Idx + i] = Byte; |
158 | return Size + N; |
159 | } |
160 | |
161 | size_t MutationDispatcher::Mutate_ChangeByte(uint8_t *Data, size_t Size, |
162 | size_t MaxSize) { |
163 | if (Size > MaxSize) return 0; |
164 | size_t Idx = Rand(Size); |
165 | Data[Idx] = RandCh(Rand); |
166 | return Size; |
167 | } |
168 | |
169 | size_t MutationDispatcher::Mutate_ChangeBit(uint8_t *Data, size_t Size, |
170 | size_t MaxSize) { |
171 | if (Size > MaxSize) return 0; |
172 | size_t Idx = Rand(Size); |
173 | Data[Idx] ^= 1 << Rand(8); |
174 | return Size; |
175 | } |
176 | |
177 | size_t MutationDispatcher::Mutate_AddWordFromManualDictionary(uint8_t *Data, |
178 | size_t Size, |
179 | size_t MaxSize) { |
180 | return AddWordFromDictionary(D&: ManualDictionary, Data, Size, MaxSize); |
181 | } |
182 | |
183 | size_t MutationDispatcher::ApplyDictionaryEntry(uint8_t *Data, size_t Size, |
184 | size_t MaxSize, |
185 | DictionaryEntry &DE) { |
186 | const Word &W = DE.GetW(); |
187 | bool UsePositionHint = DE.HasPositionHint() && |
188 | DE.GetPositionHint() + W.size() < Size && |
189 | Rand.RandBool(); |
190 | if (Rand.RandBool()) { // Insert W. |
191 | if (Size + W.size() > MaxSize) return 0; |
192 | size_t Idx = UsePositionHint ? DE.GetPositionHint() : Rand(Size + 1); |
193 | memmove(dest: Data + Idx + W.size(), src: Data + Idx, n: Size - Idx); |
194 | memcpy(dest: Data + Idx, src: W.data(), n: W.size()); |
195 | Size += W.size(); |
196 | } else { // Overwrite some bytes with W. |
197 | if (W.size() > Size) return 0; |
198 | size_t Idx = |
199 | UsePositionHint ? DE.GetPositionHint() : Rand(Size + 1 - W.size()); |
200 | memcpy(dest: Data + Idx, src: W.data(), n: W.size()); |
201 | } |
202 | return Size; |
203 | } |
204 | |
205 | // Somewhere in the past we have observed a comparison instructions |
206 | // with arguments Arg1 Arg2. This function tries to guess a dictionary |
207 | // entry that will satisfy that comparison. |
208 | // It first tries to find one of the arguments (possibly swapped) in the |
209 | // input and if it succeeds it creates a DE with a position hint. |
210 | // Otherwise it creates a DE with one of the arguments w/o a position hint. |
211 | DictionaryEntry MutationDispatcher::MakeDictionaryEntryFromCMP( |
212 | const void *Arg1, const void *Arg2, |
213 | const void *Arg1Mutation, const void *Arg2Mutation, |
214 | size_t ArgSize, const uint8_t *Data, |
215 | size_t Size) { |
216 | bool HandleFirst = Rand.RandBool(); |
217 | const void *ExistingBytes, *DesiredBytes; |
218 | Word W; |
219 | const uint8_t *End = Data + Size; |
220 | for (int Arg = 0; Arg < 2; Arg++) { |
221 | ExistingBytes = HandleFirst ? Arg1 : Arg2; |
222 | DesiredBytes = HandleFirst ? Arg2Mutation : Arg1Mutation; |
223 | HandleFirst = !HandleFirst; |
224 | W.Set(B: reinterpret_cast<const uint8_t*>(DesiredBytes), S: ArgSize); |
225 | const size_t kMaxNumPositions = 8; |
226 | size_t Positions[kMaxNumPositions]; |
227 | size_t NumPositions = 0; |
228 | for (const uint8_t *Cur = Data; |
229 | Cur < End && NumPositions < kMaxNumPositions; Cur++) { |
230 | Cur = |
231 | (const uint8_t *)SearchMemory(haystack: Cur, haystacklen: End - Cur, needle: ExistingBytes, needlelen: ArgSize); |
232 | if (!Cur) break; |
233 | Positions[NumPositions++] = Cur - Data; |
234 | } |
235 | if (!NumPositions) continue; |
236 | return DictionaryEntry(W, Positions[Rand(NumPositions)]); |
237 | } |
238 | DictionaryEntry DE(W); |
239 | return DE; |
240 | } |
241 | |
242 | |
243 | template <class T> |
244 | DictionaryEntry MutationDispatcher::MakeDictionaryEntryFromCMP( |
245 | T Arg1, T Arg2, const uint8_t *Data, size_t Size) { |
246 | if (Rand.RandBool()) Arg1 = Bswap(Arg1); |
247 | if (Rand.RandBool()) Arg2 = Bswap(Arg2); |
248 | T Arg1Mutation = static_cast<T>(Arg1 + Rand(-1, 1)); |
249 | T Arg2Mutation = static_cast<T>(Arg2 + Rand(-1, 1)); |
250 | return MakeDictionaryEntryFromCMP(&Arg1, &Arg2, &Arg1Mutation, &Arg2Mutation, |
251 | sizeof(Arg1), Data, Size); |
252 | } |
253 | |
254 | DictionaryEntry MutationDispatcher::MakeDictionaryEntryFromCMP( |
255 | const Word &Arg1, const Word &Arg2, const uint8_t *Data, size_t Size) { |
256 | return MakeDictionaryEntryFromCMP(Arg1: Arg1.data(), Arg2: Arg2.data(), Arg1Mutation: Arg1.data(), |
257 | Arg2Mutation: Arg2.data(), ArgSize: Arg1.size(), Data, Size); |
258 | } |
259 | |
260 | size_t MutationDispatcher::Mutate_AddWordFromTORC( |
261 | uint8_t *Data, size_t Size, size_t MaxSize) { |
262 | Word W; |
263 | DictionaryEntry DE; |
264 | switch (Rand(4)) { |
265 | case 0: { |
266 | auto X = TPC.TORC8.Get(I: Rand.Rand<size_t>()); |
267 | DE = MakeDictionaryEntryFromCMP(Arg1: X.A, Arg2: X.B, Data, Size); |
268 | } break; |
269 | case 1: { |
270 | auto X = TPC.TORC4.Get(I: Rand.Rand<size_t>()); |
271 | if ((X.A >> 16) == 0 && (X.B >> 16) == 0 && Rand.RandBool()) |
272 | DE = MakeDictionaryEntryFromCMP(Arg1: (uint16_t)X.A, Arg2: (uint16_t)X.B, Data, Size); |
273 | else |
274 | DE = MakeDictionaryEntryFromCMP(Arg1: X.A, Arg2: X.B, Data, Size); |
275 | } break; |
276 | case 2: { |
277 | auto X = TPC.TORCW.Get(I: Rand.Rand<size_t>()); |
278 | DE = MakeDictionaryEntryFromCMP(Arg1: X.A, Arg2: X.B, Data, Size); |
279 | } break; |
280 | case 3: if (Options.UseMemmem) { |
281 | auto X = TPC.MMT.Get(Idx: Rand.Rand<size_t>()); |
282 | DE = DictionaryEntry(X); |
283 | } break; |
284 | default: |
285 | assert(0); |
286 | } |
287 | if (!DE.GetW().size()) return 0; |
288 | Size = ApplyDictionaryEntry(Data, Size, MaxSize, DE); |
289 | if (!Size) return 0; |
290 | DictionaryEntry &DERef = |
291 | CmpDictionaryEntriesDeque[CmpDictionaryEntriesDequeIdx++ % |
292 | kCmpDictionaryEntriesDequeSize]; |
293 | DERef = DE; |
294 | CurrentDictionaryEntrySequence.push_back(x: &DERef); |
295 | return Size; |
296 | } |
297 | |
298 | size_t MutationDispatcher::Mutate_AddWordFromPersistentAutoDictionary( |
299 | uint8_t *Data, size_t Size, size_t MaxSize) { |
300 | return AddWordFromDictionary(D&: PersistentAutoDictionary, Data, Size, MaxSize); |
301 | } |
302 | |
303 | size_t MutationDispatcher::AddWordFromDictionary(Dictionary &D, uint8_t *Data, |
304 | size_t Size, size_t MaxSize) { |
305 | if (Size > MaxSize) return 0; |
306 | if (D.empty()) return 0; |
307 | DictionaryEntry &DE = D[Rand(D.size())]; |
308 | Size = ApplyDictionaryEntry(Data, Size, MaxSize, DE); |
309 | if (!Size) return 0; |
310 | DE.IncUseCount(); |
311 | CurrentDictionaryEntrySequence.push_back(x: &DE); |
312 | return Size; |
313 | } |
314 | |
315 | // Overwrites part of To[0,ToSize) with a part of From[0,FromSize). |
316 | // Returns ToSize. |
317 | size_t MutationDispatcher::CopyPartOf(const uint8_t *From, size_t FromSize, |
318 | uint8_t *To, size_t ToSize) { |
319 | // Copy From[FromBeg, FromBeg + CopySize) into To[ToBeg, ToBeg + CopySize). |
320 | size_t ToBeg = Rand(ToSize); |
321 | size_t CopySize = Rand(ToSize - ToBeg) + 1; |
322 | assert(ToBeg + CopySize <= ToSize); |
323 | CopySize = std::min(a: CopySize, b: FromSize); |
324 | size_t FromBeg = Rand(FromSize - CopySize + 1); |
325 | assert(FromBeg + CopySize <= FromSize); |
326 | memmove(dest: To + ToBeg, src: From + FromBeg, n: CopySize); |
327 | return ToSize; |
328 | } |
329 | |
330 | // Inserts part of From[0,ToSize) into To. |
331 | // Returns new size of To on success or 0 on failure. |
332 | size_t MutationDispatcher::InsertPartOf(const uint8_t *From, size_t FromSize, |
333 | uint8_t *To, size_t ToSize, |
334 | size_t MaxToSize) { |
335 | if (ToSize >= MaxToSize) return 0; |
336 | size_t AvailableSpace = MaxToSize - ToSize; |
337 | size_t MaxCopySize = std::min(a: AvailableSpace, b: FromSize); |
338 | size_t CopySize = Rand(MaxCopySize) + 1; |
339 | size_t FromBeg = Rand(FromSize - CopySize + 1); |
340 | assert(FromBeg + CopySize <= FromSize); |
341 | size_t ToInsertPos = Rand(ToSize + 1); |
342 | assert(ToInsertPos + CopySize <= MaxToSize); |
343 | size_t TailSize = ToSize - ToInsertPos; |
344 | if (To == From) { |
345 | MutateInPlaceHere.resize(new_size: MaxToSize); |
346 | memcpy(dest: MutateInPlaceHere.data(), src: From + FromBeg, n: CopySize); |
347 | memmove(dest: To + ToInsertPos + CopySize, src: To + ToInsertPos, n: TailSize); |
348 | memmove(dest: To + ToInsertPos, src: MutateInPlaceHere.data(), n: CopySize); |
349 | } else { |
350 | memmove(dest: To + ToInsertPos + CopySize, src: To + ToInsertPos, n: TailSize); |
351 | memmove(dest: To + ToInsertPos, src: From + FromBeg, n: CopySize); |
352 | } |
353 | return ToSize + CopySize; |
354 | } |
355 | |
356 | size_t MutationDispatcher::Mutate_CopyPart(uint8_t *Data, size_t Size, |
357 | size_t MaxSize) { |
358 | if (Size > MaxSize || Size == 0) return 0; |
359 | // If Size == MaxSize, `InsertPartOf(...)` will |
360 | // fail so there's no point using it in this case. |
361 | if (Size == MaxSize || Rand.RandBool()) |
362 | return CopyPartOf(From: Data, FromSize: Size, To: Data, ToSize: Size); |
363 | else |
364 | return InsertPartOf(From: Data, FromSize: Size, To: Data, ToSize: Size, MaxToSize: MaxSize); |
365 | } |
366 | |
367 | size_t MutationDispatcher::Mutate_ChangeASCIIInteger(uint8_t *Data, size_t Size, |
368 | size_t MaxSize) { |
369 | if (Size > MaxSize) return 0; |
370 | size_t B = Rand(Size); |
371 | while (B < Size && !isdigit(Data[B])) B++; |
372 | if (B == Size) return 0; |
373 | size_t E = B; |
374 | while (E < Size && isdigit(Data[E])) E++; |
375 | assert(B < E); |
376 | // now we have digits in [B, E). |
377 | // strtol and friends don't accept non-zero-teminated data, parse it manually. |
378 | uint64_t Val = Data[B] - '0'; |
379 | for (size_t i = B + 1; i < E; i++) |
380 | Val = Val * 10 + Data[i] - '0'; |
381 | |
382 | // Mutate the integer value. |
383 | switch(Rand(5)) { |
384 | case 0: Val++; break; |
385 | case 1: Val--; break; |
386 | case 2: Val /= 2; break; |
387 | case 3: Val *= 2; break; |
388 | case 4: Val = Rand(Val * Val); break; |
389 | default: assert(0); |
390 | } |
391 | // Just replace the bytes with the new ones, don't bother moving bytes. |
392 | for (size_t i = B; i < E; i++) { |
393 | size_t Idx = E + B - i - 1; |
394 | assert(Idx >= B && Idx < E); |
395 | Data[Idx] = (Val % 10) + '0'; |
396 | Val /= 10; |
397 | } |
398 | return Size; |
399 | } |
400 | |
401 | template<class T> |
402 | size_t ChangeBinaryInteger(uint8_t *Data, size_t Size, Random &Rand) { |
403 | if (Size < sizeof(T)) return 0; |
404 | size_t Off = Rand(Size - sizeof(T) + 1); |
405 | assert(Off + sizeof(T) <= Size); |
406 | T Val; |
407 | if (Off < 64 && !Rand(4)) { |
408 | Val = static_cast<T>(Size); |
409 | if (Rand.RandBool()) |
410 | Val = Bswap(Val); |
411 | } else { |
412 | memcpy(&Val, Data + Off, sizeof(Val)); |
413 | T Add = static_cast<T>(Rand(21)); |
414 | Add -= 10; |
415 | if (Rand.RandBool()) |
416 | Val = Bswap(T(Bswap(Val) + Add)); // Add assuming different endiannes. |
417 | else |
418 | Val = Val + Add; // Add assuming current endiannes. |
419 | if (Add == 0 || Rand.RandBool()) // Maybe negate. |
420 | Val = -Val; |
421 | } |
422 | memcpy(Data + Off, &Val, sizeof(Val)); |
423 | return Size; |
424 | } |
425 | |
426 | size_t MutationDispatcher::Mutate_ChangeBinaryInteger(uint8_t *Data, |
427 | size_t Size, |
428 | size_t MaxSize) { |
429 | if (Size > MaxSize) return 0; |
430 | switch (Rand(4)) { |
431 | case 3: return ChangeBinaryInteger<uint64_t>(Data, Size, Rand); |
432 | case 2: return ChangeBinaryInteger<uint32_t>(Data, Size, Rand); |
433 | case 1: return ChangeBinaryInteger<uint16_t>(Data, Size, Rand); |
434 | case 0: return ChangeBinaryInteger<uint8_t>(Data, Size, Rand); |
435 | default: assert(0); |
436 | } |
437 | return 0; |
438 | } |
439 | |
440 | size_t MutationDispatcher::Mutate_CrossOver(uint8_t *Data, size_t Size, |
441 | size_t MaxSize) { |
442 | if (Size > MaxSize) return 0; |
443 | if (Size == 0) return 0; |
444 | if (!CrossOverWith) return 0; |
445 | const Unit &O = *CrossOverWith; |
446 | if (O.empty()) return 0; |
447 | size_t NewSize = 0; |
448 | switch(Rand(3)) { |
449 | case 0: |
450 | MutateInPlaceHere.resize(new_size: MaxSize); |
451 | NewSize = CrossOver(Data1: Data, Size1: Size, Data2: O.data(), Size2: O.size(), |
452 | Out: MutateInPlaceHere.data(), MaxOutSize: MaxSize); |
453 | memcpy(dest: Data, src: MutateInPlaceHere.data(), n: NewSize); |
454 | break; |
455 | case 1: |
456 | NewSize = InsertPartOf(From: O.data(), FromSize: O.size(), To: Data, ToSize: Size, MaxToSize: MaxSize); |
457 | if (!NewSize) |
458 | NewSize = CopyPartOf(From: O.data(), FromSize: O.size(), To: Data, ToSize: Size); |
459 | break; |
460 | case 2: |
461 | NewSize = CopyPartOf(From: O.data(), FromSize: O.size(), To: Data, ToSize: Size); |
462 | break; |
463 | default: assert(0); |
464 | } |
465 | assert(NewSize > 0 && "CrossOver returned empty unit" ); |
466 | assert(NewSize <= MaxSize && "CrossOver returned overisized unit" ); |
467 | return NewSize; |
468 | } |
469 | |
470 | void MutationDispatcher::StartMutationSequence() { |
471 | CurrentMutatorSequence.clear(); |
472 | CurrentDictionaryEntrySequence.clear(); |
473 | } |
474 | |
475 | // Copy successful dictionary entries to PersistentAutoDictionary. |
476 | void MutationDispatcher::RecordSuccessfulMutationSequence() { |
477 | for (auto DE : CurrentDictionaryEntrySequence) { |
478 | // PersistentAutoDictionary.AddWithSuccessCountOne(DE); |
479 | DE->IncSuccessCount(); |
480 | assert(DE->GetW().size()); |
481 | // Linear search is fine here as this happens seldom. |
482 | if (!PersistentAutoDictionary.ContainsWord(W: DE->GetW())) |
483 | PersistentAutoDictionary.push_back(DE: *DE); |
484 | } |
485 | } |
486 | |
487 | void MutationDispatcher::PrintRecommendedDictionary() { |
488 | std::vector<DictionaryEntry> V; |
489 | for (auto &DE : PersistentAutoDictionary) |
490 | if (!ManualDictionary.ContainsWord(W: DE.GetW())) |
491 | V.push_back(x: DE); |
492 | if (V.empty()) return; |
493 | Printf(Fmt: "###### Recommended dictionary. ######\n" ); |
494 | for (auto &DE: V) { |
495 | assert(DE.GetW().size()); |
496 | Printf(Fmt: "\"" ); |
497 | PrintASCII(W: DE.GetW(), PrintAfter: "\"" ); |
498 | Printf(Fmt: " # Uses: %zd\n" , DE.GetUseCount()); |
499 | } |
500 | Printf(Fmt: "###### End of recommended dictionary. ######\n" ); |
501 | } |
502 | |
503 | void MutationDispatcher::PrintMutationSequence(bool Verbose) { |
504 | Printf(Fmt: "MS: %zd " , CurrentMutatorSequence.size()); |
505 | size_t EntriesToPrint = |
506 | Verbose ? CurrentMutatorSequence.size() |
507 | : std::min(a: kMaxMutationsToPrint, b: CurrentMutatorSequence.size()); |
508 | for (size_t i = 0; i < EntriesToPrint; i++) |
509 | Printf(Fmt: "%s-" , CurrentMutatorSequence[i].Name); |
510 | if (!CurrentDictionaryEntrySequence.empty()) { |
511 | Printf(Fmt: " DE: " ); |
512 | EntriesToPrint = Verbose ? CurrentDictionaryEntrySequence.size() |
513 | : std::min(a: kMaxMutationsToPrint, |
514 | b: CurrentDictionaryEntrySequence.size()); |
515 | for (size_t i = 0; i < EntriesToPrint; i++) { |
516 | Printf(Fmt: "\"" ); |
517 | PrintASCII(W: CurrentDictionaryEntrySequence[i]->GetW(), PrintAfter: "\"-" ); |
518 | } |
519 | } |
520 | } |
521 | |
522 | std::string MutationDispatcher::MutationSequence() { |
523 | std::string MS; |
524 | for (const auto &M : CurrentMutatorSequence) { |
525 | MS += M.Name; |
526 | MS += "-" ; |
527 | } |
528 | return MS; |
529 | } |
530 | |
531 | size_t MutationDispatcher::Mutate(uint8_t *Data, size_t Size, size_t MaxSize) { |
532 | return MutateImpl(Data, Size, MaxSize, Mutators); |
533 | } |
534 | |
535 | size_t MutationDispatcher::DefaultMutate(uint8_t *Data, size_t Size, |
536 | size_t MaxSize) { |
537 | return MutateImpl(Data, Size, MaxSize, Mutators&: DefaultMutators); |
538 | } |
539 | |
540 | // Mutates Data in place, returns new size. |
541 | size_t MutationDispatcher::MutateImpl(uint8_t *Data, size_t Size, |
542 | size_t MaxSize, |
543 | std::vector<Mutator> &Mutators) { |
544 | assert(MaxSize > 0); |
545 | // Some mutations may fail (e.g. can't insert more bytes if Size == MaxSize), |
546 | // in which case they will return 0. |
547 | // Try several times before returning un-mutated data. |
548 | for (int Iter = 0; Iter < 100; Iter++) { |
549 | auto M = Mutators[Rand(Mutators.size())]; |
550 | size_t NewSize = (this->*(M.Fn))(Data, Size, MaxSize); |
551 | if (NewSize && NewSize <= MaxSize) { |
552 | if (Options.OnlyASCII) |
553 | ToASCII(Data, Size: NewSize); |
554 | CurrentMutatorSequence.push_back(x: M); |
555 | return NewSize; |
556 | } |
557 | } |
558 | *Data = ' '; |
559 | return 1; // Fallback, should not happen frequently. |
560 | } |
561 | |
562 | // Mask represents the set of Data bytes that are worth mutating. |
563 | size_t MutationDispatcher::MutateWithMask(uint8_t *Data, size_t Size, |
564 | size_t MaxSize, |
565 | const std::vector<uint8_t> &Mask) { |
566 | size_t MaskedSize = std::min(a: Size, b: Mask.size()); |
567 | // * Copy the worthy bytes into a temporary array T |
568 | // * Mutate T |
569 | // * Copy T back. |
570 | // This is totally unoptimized. |
571 | auto &T = MutateWithMaskTemp; |
572 | if (T.size() < Size) |
573 | T.resize(new_size: Size); |
574 | size_t OneBits = 0; |
575 | for (size_t I = 0; I < MaskedSize; I++) |
576 | if (Mask[I]) |
577 | T[OneBits++] = Data[I]; |
578 | |
579 | if (!OneBits) return 0; |
580 | assert(!T.empty()); |
581 | size_t NewSize = Mutate(Data: T.data(), Size: OneBits, MaxSize: OneBits); |
582 | assert(NewSize <= OneBits); |
583 | (void)NewSize; |
584 | // Even if NewSize < OneBits we still use all OneBits bytes. |
585 | for (size_t I = 0, J = 0; I < MaskedSize; I++) |
586 | if (Mask[I]) |
587 | Data[I] = T[J++]; |
588 | return Size; |
589 | } |
590 | |
591 | void MutationDispatcher::AddWordToManualDictionary(const Word &W) { |
592 | ManualDictionary.push_back( |
593 | DE: {W, std::numeric_limits<size_t>::max()}); |
594 | } |
595 | |
596 | } // namespace fuzzer |
597 | |