1 | //===- FuzzerMerge.h - merging corpa ----------------------------*- C++ -* ===// |
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 | // Merging Corpora. |
9 | // |
10 | // The task: |
11 | // Take the existing corpus (possibly empty) and merge new inputs into |
12 | // it so that only inputs with new coverage ('features') are added. |
13 | // The process should tolerate the crashes, OOMs, leaks, etc. |
14 | // |
15 | // Algorithm: |
16 | // The outer process collects the set of files and writes their names |
17 | // into a temporary "control" file, then repeatedly launches the inner |
18 | // process until all inputs are processed. |
19 | // The outer process does not actually execute the target code. |
20 | // |
21 | // The inner process reads the control file and sees a) list of all the inputs |
22 | // and b) the last processed input. Then it starts processing the inputs one |
23 | // by one. Before processing every input it writes one line to control file: |
24 | // STARTED INPUT_ID INPUT_SIZE |
25 | // After processing an input it writes the following lines: |
26 | // FT INPUT_ID Feature1 Feature2 Feature3 ... |
27 | // COV INPUT_ID Coverage1 Coverage2 Coverage3 ... |
28 | // If a crash happens while processing an input the last line in the control |
29 | // file will be "STARTED INPUT_ID" and so the next process will know |
30 | // where to resume. |
31 | // |
32 | // Once all inputs are processed by the inner process(es) the outer process |
33 | // reads the control files and does the merge based entirely on the contents |
34 | // of control file. |
35 | // It uses a single pass greedy algorithm choosing first the smallest inputs |
36 | // within the same size the inputs that have more new features. |
37 | // |
38 | //===----------------------------------------------------------------------===// |
39 | |
40 | #ifndef LLVM_FUZZER_MERGE_H |
41 | #define LLVM_FUZZER_MERGE_H |
42 | |
43 | #include "FuzzerDefs.h" |
44 | #include "FuzzerIO.h" |
45 | |
46 | #include <istream> |
47 | #include <ostream> |
48 | #include <set> |
49 | #include <vector> |
50 | |
51 | namespace fuzzer { |
52 | |
53 | struct MergeFileInfo { |
54 | std::string Name; |
55 | size_t Size = 0; |
56 | std::vector<uint32_t> Features, Cov; |
57 | }; |
58 | |
59 | struct Merger { |
60 | std::vector<MergeFileInfo> Files; |
61 | size_t NumFilesInFirstCorpus = 0; |
62 | size_t FirstNotProcessedFile = 0; |
63 | std::string LastFailure; |
64 | |
65 | bool Parse(std::istream &IS, bool ParseCoverage); |
66 | bool Parse(const std::string &Str, bool ParseCoverage); |
67 | void ParseOrExit(std::istream &IS, bool ParseCoverage); |
68 | size_t Merge(const std::set<uint32_t> &InitialFeatures, |
69 | std::set<uint32_t> *NewFeatures, |
70 | const std::set<uint32_t> &InitialCov, std::set<uint32_t> *NewCov, |
71 | std::vector<std::string> *NewFiles); |
72 | size_t SetCoverMerge(const std::set<uint32_t> &InitialFeatures, |
73 | std::set<uint32_t> *NewFeatures, |
74 | const std::set<uint32_t> &InitialCov, |
75 | std::set<uint32_t> *NewCov, |
76 | std::vector<std::string> *NewFiles); |
77 | size_t ApproximateMemoryConsumption() const; |
78 | std::set<uint32_t> AllFeatures() const; |
79 | }; |
80 | |
81 | void CrashResistantMerge(const std::vector<std::string> &Args, |
82 | const std::vector<SizedFile> &OldCorpus, |
83 | const std::vector<SizedFile> &NewCorpus, |
84 | std::vector<std::string> *NewFiles, |
85 | const std::set<uint32_t> &InitialFeatures, |
86 | std::set<uint32_t> *NewFeatures, |
87 | const std::set<uint32_t> &InitialCov, |
88 | std::set<uint32_t> *NewCov, const std::string &CFPath, |
89 | bool Verbose, bool IsSetCoverMerge); |
90 | |
91 | } // namespace fuzzer |
92 | |
93 | #endif // LLVM_FUZZER_MERGE_H |
94 | |