I found that the initial corpus allocation of fork mode has certain defects.
I designed a new initial corpus allocation strategy based on size grouping.
This method can give more energy to the small seeds in the corpus and
increase the throughput of the test.
Fuzzbench data (glibfuzzer is -fork_corpus_groups=1):
https://www.fuzzbench.com/reports/experimental/2021-08-05-parallel/index.html
Reviewed By: morehouse
Differential Revision: https://reviews.llvm.org/D105084
Extend the existing single-pass algorithm for `Merger::Merge` with an algorithm that gives better results. This new implementation can be used with a new **set_cover_merge=1** flag.
This greedy set cover implementation gives a substantially smaller final corpus (40%-80% less testcases) while preserving the same features/coverage. At the same time, the execution time penalty is not that significant (+50% for ~1M corpus files and far less for smaller corpora). These results were obtained by comparing several targets with varying size corpora.
Change `Merger::CrashResistantMergeInternalStep` to collect all features from each file and not just unique ones. This is needed for the set cover algorithm to work correctly. The implementation of the algorithm in `Merger::SetCoverMerge` uses a bitvector to store features that are covered by a file while performing the pass. Collisions while indexing the bitvector are ignored similarly to the fuzzer.
Reviewed By: morehouse
Differential Revision: https://reviews.llvm.org/D105284
This change introduces libMutagen/libclang_rt.mutagen.a as a subset of libFuzzer/libclang_rt.fuzzer.a. This library contains only the fuzzing strategies used by libFuzzer to produce new test inputs from provided inputs, dictionaries, and SanitizerCoverage feedback.
Most of this change is simply moving sections of code to one side or the other of the library boundary. The only meaningful new code is:
* The Mutagen.h interface and its implementation in Mutagen.cpp.
* The following methods in MutagenDispatcher.cpp:
* UseCmp
* UseMemmem
* SetCustomMutator
* SetCustomCrossOver
* LateInitialize (similar to the MutationDispatcher's original constructor)
* Mutate_AddWordFromTORC (uses callbacks instead of accessing TPC directly)
* StartMutationSequence
* MutationSequence
* DictionaryEntrySequence
* RecommendDictionary
* RecommendDictionaryEntry
* FuzzerMutate.cpp (which now justs sets callbacks and handles printing)
* MutagenUnittest.cpp (which adds tests of Mutagen.h)
A note on performance: This change was tested with a 100 passes of test/fuzzer/LargeTest.cpp with 1000 runs per pass, both with and without the change. The running time distribution was qualitatively similar both with and without the change, and the average difference was within 30 microseconds (2.240 ms/run vs 2.212 ms/run, respectively). Both times were much higher than observed with the fully optimized system clang (~0.38 ms/run), most likely due to the combination of CMake "dev mode" settings (e.g. CMAKE_BUILD_TYPE="Debug", LLVM_ENABLE_LTO=OFF, etc.). The difference between the two versions built similarly seems to be "in the noise" and suggests no meaningful performance degradation.
Reviewed By: morehouse
Differential Revision: https://reviews.llvm.org/D102447
This change introduces libMutagen/libclang_rt.mutagen.a as a subset of libFuzzer/libclang_rt.fuzzer.a. This library contains only the fuzzing strategies used by libFuzzer to produce new test inputs from provided inputs, dictionaries, and SanitizerCoverage feedback.
Most of this change is simply moving sections of code to one side or the other of the library boundary. The only meaningful new code is:
* The Mutagen.h interface and its implementation in Mutagen.cpp.
* The following methods in MutagenDispatcher.cpp:
* UseCmp
* UseMemmem
* SetCustomMutator
* SetCustomCrossOver
* LateInitialize (similar to the MutationDispatcher's original constructor)
* Mutate_AddWordFromTORC (uses callbacks instead of accessing TPC directly)
* StartMutationSequence
* MutationSequence
* DictionaryEntrySequence
* RecommendDictionary
* RecommendDictionaryEntry
* FuzzerMutate.cpp (which now justs sets callbacks and handles printing)
* MutagenUnittest.cpp (which adds tests of Mutagen.h)
A note on performance: This change was tested with a 100 passes of test/fuzzer/LargeTest.cpp with 1000 runs per pass, both with and without the change. The running time distribution was qualitatively similar both with and without the change, and the average difference was within 30 microseconds (2.240 ms/run vs 2.212 ms/run, respectively). Both times were much higher than observed with the fully optimized system clang (~0.38 ms/run), most likely due to the combination of CMake "dev mode" settings (e.g. CMAKE_BUILD_TYPE="Debug", LLVM_ENABLE_LTO=OFF, etc.). The difference between the two versions built similarly seems to be "in the noise" and suggests no meaningful performance degradation.
Reviewed By: morehouse
Differential Revision: https://reviews.llvm.org/D102447
Attempting to build a standalone libFuzzer in Fuchsia's default toolchain for the purpose of cross-compiling the unit tests revealed a number of not-quite-proper type conversions. Fuchsia's toolchain include `-std=c++17` and `-Werror`, among others, leading to many errors like `-Wshorten-64-to-32`, `-Wimplicit-float-conversion`, etc.
Most of these have been addressed by simply making the conversion explicit with a `static_cast`. These typically fell into one of two categories: 1) conversions between types where high precision isn't critical, e.g. the "energy" calculations for `InputInfo`, and 2) conversions where the values will never reach the bits being truncated, e.g. `DftTimeInSeconds` is not going to exceed 136 years.
The major exception to this is the number of features: there are several places that treat features as `size_t`, and others as `uint32_t`. This change makes the decision to cap the features at 32 bits. The maximum value of a feature as produced by `TracePC::CollectFeatures` is roughly:
(NumPCsInPCTables + ValueBitMap::kMapSizeInBits + ExtraCountersBegin() - ExtraCountersEnd() + log2(SIZE_MAX)) * 8
It's conceivable for extremely large targets and/or extra counters that this limit could be reached. This shouldn't break fuzzing, but it will cause certain features to collide and lower the fuzzers overall precision. To address this, this change adds a warning to TracePC::PrintModuleInfo about excessive feature size if it is detected, and recommends refactoring the fuzzer into several smaller ones.
Reviewed By: morehouse
Differential Revision: https://reviews.llvm.org/D97992
Adds a new option, `handle_winexcept` to try to intercept uncaught
Visual C++ exceptions on Windows. On Linux, such exceptions are handled
implicitly by `std::terminate()` raising `SIBABRT`. This option brings the
Windows behavior in line with Linux.
Unfortunately this exception code is intentionally undocumented, however
has remained stable for the last decade. More information can be found
here: https://devblogs.microsoft.com/oldnewthing/20100730-00/?p=13273
Reviewed By: morehouse, metzman
Differential Revision: https://reviews.llvm.org/D89755
-print_full_coverage=1 produces a detailed branch coverage dump when run on a single file.
Uses same infrastructure as -print_coverage flag, but prints all branches (regardless of coverage status) in an easy-to-parse format.
Usage: For internal use with machine learning fuzzing models which require detailed coverage information on seed files to generate mutations.
Differential Revision: https://reviews.llvm.org/D85928
This patch adds a new command-line option -mutation_graph_file=FILE for
debugging purposes, which traces how corpus inputs evolve during a fuzzing
run. For each new input that is added to the corpus, a new vertex corresponding
to the added input, as well as a new edge that connects its base input to itself
are written to the given file. Each vertex is labeled with the filename of the
input, and each edge is labeled with the mutation sequence that led to the input
w.r.t. its base input.
The format of the mutation graph file is the dot file format. Once prepended and
appended with "graph {" and "}", respectively, the graph becomes a valid dot
file and can be visualized.
Differential Revision: https://reviews.llvm.org/D86560
This patch scales the energy computed by the Entropic schedule based on the
execution time of each input. The input execution time is compared with the
average execution time of inputs in the corpus, and, based on the amount by
which they differ, the energy is scaled from 0.1x (for inputs executing slow) to
3x (for inputs executing fast). Note that the exact scaling criteria and formula
is borrowed from AFL.
On FuzzBench, this gives a sizeable throughput increase, which in turn leads to
more coverage on several benchmarks. For details, see the following report.
https://storage.googleapis.com/fuzzer-test-suite-public/exectime-report/index.html
Differential Revision: https://reviews.llvm.org/D86092
This patch adds an option "cross_over_uniform_dist", which, if 1, considers all
inputs in the corpus for the crossover input selection. More specifically, this
patch uses a uniform distribution of all inputs in the corpus for the CrossOver
input selection. Note that input selection for mutation is still fully
determined by the scheduling policy (i.e., vanilla or Entropic); the uniform
distribution only applies to the secondary input selection, only for the
crossover mutation of the base input chosen by the scheduling policy. This way
the corpus inputs that have useful fragments in them, even though they are
deprioritized by the scheduling policy, have chances of getting mixed with other
inputs that are prioritized and selected as base input for mutation.
Differential Revision: https://reviews.llvm.org/D86954
This patch adds an option "keep_seed" to keep all initial seed inputs in the
corpus. Previously, only the initial seed inputs that find new coverage were
added to the corpus, and all the other initial inputs were discarded. We
observed in some circumstances that useful initial seed inputs are discarded as
they find no new coverage, even though they contain useful fragments in them
(e.g., SQLITE3 FuzzBench benchmark). This newly added option provides a way to
keeping seed inputs in the corpus for those circumstances. With this patch, and
with -keep_seed=1, all initial seed inputs are kept in the corpus regardless of
whether they find new coverage or not. Further, these seed inputs are not
replaced with smaller inputs even if -reduce_inputs=1.
Differential Revision: https://reviews.llvm.org/D86577
Currently, libFuzzer will exit with an error message if a non-existent
directory is provided for any of the appropriate arguments. For cases
where libFuzzer is used in a specialized embedded environment, it would
be much easier to have libFuzzer create the directories for the user.
This patch accommodates for this scenario by allowing the user to provide
the argument `-create_missing_dirs=1` which makes libFuzzer attempt to
create the `artifact_prefix`, `exact_artifact_path`,
`features_dir` and/or corpus directory if they don't already exist rather
than throw an error and exit.
Split off from D84808 as requested [here](https://reviews.llvm.org/D84808#2208546).
Reviewed By: morehouse
Differential Revision: https://reviews.llvm.org/D86733
Currently, libFuzzer will exit with an error message if a non-existent
directory is provided for any of the appropriate arguments. For cases
where libFuzzer is used in a specialized embedded environment, it would
be much easier to have libFuzzer create the directories for the user.
This patch accommodates for this scenario by allowing the user to provide
the argument `-create_missing_dirs=1` which makes libFuzzer attempt to
create the `artifact_prefix`, `exact_artifact_path`,
`features_dir` and/or corpus directory if they don't already exist rather
than throw an error and exit.
Split off from D84808 as requested [here](https://reviews.llvm.org/D84808#2208546).
Reviewed By: morehouse
Differential Revision: https://reviews.llvm.org/D86733
Currently, libFuzzer will exit with an error message if a non-existent
corpus directory is provided. However, if a user provides a non-existent
directory for the `artifact_prefix`, `exact_artifact_path`, or
`features_dir`, libFuzzer will continue execution but silently fail to
write artifacts/features.
To improve the user experience, this PR adds validation for the existence of
all user supplied directories before executing the main fuzzing loop. If they
don't exist, libFuzzer will exit with an error message.
Patch By: dgg5503
Reviewed By: morehouse
Differential Revision: https://reviews.llvm.org/D84808
When one tries to minimize timeouts using -minimize_crash=1,
minimization immediately fails. The following sequence of events is
responsible for this:
[parent] SIGALRM occurs
[parent] read() returns -EINTR (or -ERESTARTSYS according to strace)
[parent] fgets() returns NULL
[parent] ExecuteCommand() closes child's stdout and returns
[child ] SIGALRM occurs
[child ] AlarmCallback() attempts to write "ALARM: ..." to stdout
[child ] Dies with SIGPIPE without calling DumpCurrentUnit()
[parent] Does not see -exact_artifact_path and exits
When minimizing, the timer in parent is not necessary, so fix by not
setting it in this case.
Reviewed By: morehouse
Differential Revision: https://reviews.llvm.org/D85359
This adds a new extern "C" function that serves the same purpose. This removes the need for external users to depend on internal headers in order to use this feature. It also standardizes the interface in a way that other fuzzing engines will be able to match.
Patch By: IanPudney
Reviewed By: kcc
Differential Revision: https://reviews.llvm.org/D84561
The incorrect symbol will cause linking failures for 32-bit targets:
clang_rt.fuzzer-i386.lib(FuzzerDriver.obj) : error LNK2001: unresolved external symbol __libfuzzer_is_present
Verified no longer fails to link with this change for 32-bit and still succeeds for 64-bit MSVC.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D83594
Summary: This patch separates platform related macros in lib/fuzzer/FuzzerDefs.h into lib/fuzzer/FuzzerPlatform.h, and use FuzzerPlatform.h where necessary. This separation helps when compiling libFuzzer's interceptor module (under review); an unnecessary include of standard headers (such as string.h) may produce conflicts/ambiguation with the interceptor's declarations/definitions of library functions, which complicates interceptor implementation.
Reviewers: morehouse, hctim
Reviewed By: morehouse
Subscribers: krytarowski, #sanitizers
Tags: #sanitizers
Differential Revision: https://reviews.llvm.org/D83805
Summary:
This is collaboration between Marcel Boehme @ Monash, Australia and Valentin Manès plus Sang Kil Cha @ KAIST, South Korea.
We have made a few modifications to boost LibFuzzer performance by changing how weights are assigned to the seeds in the corpus. Essentially, seeds that reveal more "information" about globally rare features are assigned a higher weight. Our results on the Fuzzer Test Suite seem quite promising. In terms of bug finding, our Entropic patch usually finds the same errors much faster and in more runs. In terms of coverage, our version Entropic achieves the same coverage in less than half the time for the majority of subjects. For the lack of space, we shared more detailed performance results directly with @kcc. We'll publish the preprint with all the technical details as soon as it is accepted. Happy to share if you drop us an email.
There should be plenty of opportunities to optimise further. For instance, while Entropic achieves the same coverage in less than half the time, Entropic has a much lower #execs per second. We ran the perf-tool and found a few performance bottlenecks.
Thanks for open-sourcing LibFuzzer (and the entire LLVM Compiler Infrastructure)! This has been such a tremendous help to my research.
Patch By: Marcel Boehme
Reviewers: kcc, metzman, morehouse, Dor1s, vitalybuka
Reviewed By: kcc
Subscribers: dgg5503, Valentin, llvm-commits, kcc
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D73776
For CleanseCrashInput, discards stdout output anyway since it is not used.
These changes are to defend against aggressive PID recycle on windows to reduce the chance of contention on files.
Using pipe instead of file also workaround the problem that when the
process is spawned by llvm-lit, the aborted process keeps a handle to the
output file such that the output file can not be removed. This will
cause random test failures.
https://devblogs.microsoft.com/oldnewthing/20110107-00/?p=11803
Reviewers: kcc, vitalybuka
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D73329
Summary:
there is an ongoing work on interchangeable custom mutators
(https://github.com/google/clusterfuzz/pull/1333/files#r367706283)
and having some sort of signalling from libFuzzer that it has loaded
a custom mutator would be helpful.
The initial idea was to make the mutator to print something, but given
the anticipated variety of different mutators, it does not seem possible
to make all of them print the same message to signal their execution.
Reviewers: kcc, metzman
Reviewed By: metzman
Subscribers: #sanitizers, llvm-commits
Tags: #sanitizers, #llvm
Differential Revision: https://reviews.llvm.org/D73136
Summary:
Add basic support for emscripten.
This enables libFuzzer to build (using build.sh) for emscripten and fuzz
a target compiled with
-fsanitize-coverage=inline-8bit-counters.
Basic fuzzing and bug finding work with this commit.
RSS limit and timeouts will not work because they depend on system
functions that are not implemented/widely supported in emscripten.
Reviewers: kcc, vitalybuka, hctim
Reviewed By: hctim
Subscribers: #sanitizers, llvm-commits
Tags: #sanitizers, #llvm
Differential Revision: https://reviews.llvm.org/D71285
Summary:
Some custom mutators may not peform well when size restriction is
enforced by len_control. Because of that, it's safer to disable len_control
by default in such cases, but still allow users to enable it manually.
Bug example: https://bugs.chromium.org/p/chromium/issues/detail?id=919530.
Tested manually with LPM-based and regular fuzz targets.
Reviewers: kcc, vitalybuka, metzman
Reviewed By: kcc, metzman
Subscribers: delcypher, #sanitizers, llvm-commits
Tags: #llvm, #sanitizers
Differential Revision: https://reviews.llvm.org/D63334
llvm-svn: 363443
Summary:
Pass seed corpus list in a file to get around argument length limits on Windows.
This limit was preventing many uses of fork mode on Windows.
Reviewers: kcc, morehouse
Reviewed By: kcc
Subscribers: #sanitizers, llvm-commits
Tags: #sanitizers, #llvm
Differential Revision: https://reviews.llvm.org/D60980
llvm-svn: 359610
gtt1995
aristotelis