1 ========================================================
2 LibFuzzer -- a library for coverage-guided fuzz testing.
3 ========================================================
11 This library is intended primarily for in-process coverage-guided fuzz testing
12 (fuzzing) of other libraries. The typical workflow looks like this:
14 * Build the Fuzzer library as a static archive (or just a set of .o files).
15 Note that the Fuzzer contains the main() function.
16 Preferably do *not* use sanitizers while building the Fuzzer.
17 * Build the library you are going to test with
18 `-fsanitize-coverage={bb,edge}[,indirect-calls,8bit-counters]`
19 and one of the sanitizers. We recommend to build the library in several
20 different modes (e.g. asan, msan, lsan, ubsan, etc) and even using different
21 optimizations options (e.g. -O0, -O1, -O2) to diversify testing.
22 * Build a test driver using the same options as the library.
23 The test driver is a C/C++ file containing interesting calls to the library
24 inside a single function ``extern "C" void LLVMFuzzerTestOneInput(const uint8_t *Data, size_t Size);``
25 * Link the Fuzzer, the library and the driver together into an executable
26 using the same sanitizer options as for the library.
27 * Collect the initial corpus of inputs for the
28 fuzzer (a directory with test inputs, one file per input).
29 The better your inputs are the faster you will find something interesting.
30 Also try to keep your inputs small, otherwise the Fuzzer will run too slow.
31 By default, the Fuzzer limits the size of every input to 64 bytes
32 (use ``-max_len=N`` to override).
33 * Run the fuzzer with the test corpus. As new interesting test cases are
34 discovered they will be added to the corpus. If a bug is discovered by
35 the sanitizer (asan, etc) it will be reported as usual and the reproducer
36 will be written to disk.
37 Each Fuzzer process is single-threaded (unless the library starts its own
38 threads). You can run the Fuzzer on the same corpus in multiple processes
42 The Fuzzer is similar in concept to AFL_,
43 but uses in-process Fuzzing, which is more fragile, more restrictive, but
44 potentially much faster as it has no overhead for process start-up.
45 It uses LLVM's SanitizerCoverage_ instrumentation to get in-process
48 The code resides in the LLVM repository, requires the fresh Clang compiler to build
49 and is used to fuzz various parts of LLVM,
50 but the Fuzzer itself does not (and should not) depend on any
51 part of LLVM and can be used for other projects w/o requiring the rest of LLVM.
55 The most important flags are::
57 seed 0 Random seed. If 0, seed is generated.
58 runs -1 Number of individual test runs (-1 for infinite runs).
59 max_len 64 Maximum length of the test input.
60 cross_over 1 If 1, cross over inputs.
61 mutate_depth 5 Apply this number of consecutive mutations to each input.
62 timeout 1200 Timeout in seconds (if positive). If one unit runs more than this number of seconds the process will abort.
64 save_minimized_corpus 0 If 1, the minimized corpus is saved into the first input directory
65 jobs 0 Number of jobs to run. If jobs >= 1 we spawn this number of jobs in separate worker processes with stdout/stderr redirected to fuzz-JOB.log.
66 workers 0 Number of simultaneous worker processes to run the jobs. If zero, "min(jobs,NumberOfCpuCores()/2)" is used.
67 sync_command 0 Execute an external command "<sync_command> <test_corpus>" to synchronize the test corpus.
68 sync_timeout 600 Minimum timeout between syncs.
69 use_traces 0 Experimental: use instruction traces
70 only_ascii 0 If 1, generate only ASCII (isprint+isspace) inputs.
73 For the full list of flags run the fuzzer binary with ``-help=1``.
81 A simple function that does something interesting if it receives the input "HI!"::
83 cat << EOF >> test_fuzzer.cc
84 extern "C" void LLVMFuzzerTestOneInput(const unsigned char *data, unsigned long size) {
85 if (size > 0 && data[0] == 'H')
86 if (size > 1 && data[1] == 'I')
87 if (size > 2 && data[2] == '!')
91 # Get lib/Fuzzer. Assuming that you already have fresh clang in PATH.
92 svn co http://llvm.org/svn/llvm-project/llvm/trunk/lib/Fuzzer
93 # Build lib/Fuzzer files.
94 clang -c -g -O2 -std=c++11 Fuzzer/*.cpp -IFuzzer
95 # Build test_fuzzer.cc with asan and link against lib/Fuzzer.
96 clang++ -fsanitize=address -fsanitize-coverage=edge test_fuzzer.cc Fuzzer*.o
97 # Run the fuzzer with no corpus.
100 You should get ``Illegal instruction (core dumped)`` pretty quickly.
105 Here we show how to use lib/Fuzzer on something real, yet simple: pcre2_::
107 COV_FLAGS=" -fsanitize-coverage=edge,indirect-calls,8bit-counters"
109 svn co svn://vcs.exim.org/pcre2/code/trunk pcre
110 # Get lib/Fuzzer. Assuming that you already have fresh clang in PATH.
111 svn co http://llvm.org/svn/llvm-project/llvm/trunk/lib/Fuzzer
112 # Build PCRE2 with AddressSanitizer and coverage.
113 (cd pcre; ./autogen.sh; CC="clang -fsanitize=address $COV_FLAGS" ./configure --prefix=`pwd`/../inst && make -j && make install)
114 # Build lib/Fuzzer files.
115 clang -c -g -O2 -std=c++11 Fuzzer/*.cpp -IFuzzer
116 # Build the actual function that does something interesting with PCRE2.
117 cat << EOF > pcre_fuzzer.cc
119 #include "pcre2posix.h"
120 extern "C" void LLVMFuzzerTestOneInput(const unsigned char *data, size_t size) {
121 if (size < 1) return;
122 char *str = new char[size+1];
123 memcpy(str, data, size);
126 if (0 == regcomp(&preg, str, 0)) {
127 regexec(&preg, str, 0, 0, 0);
133 clang++ -g -fsanitize=address $COV_FLAGS -c -std=c++11 -I inst/include/ pcre_fuzzer.cc
135 clang++ -g -fsanitize=address -Wl,--whole-archive inst/lib/*.a -Wl,-no-whole-archive Fuzzer*.o pcre_fuzzer.o -o pcre_fuzzer
137 This will give you a binary of the fuzzer, called ``pcre_fuzzer``.
138 Now, create a directory that will hold the test corpus::
142 For simple input languages like regular expressions this is all you need.
143 For more complicated inputs populate the directory with some input samples.
144 Now run the fuzzer with the corpus dir as the only parameter::
146 ./pcre_fuzzer ./CORPUS
148 You will see output like this::
151 #0 READ cov 0 bits 0 units 1 exec/s 0
152 #1 pulse cov 3 bits 0 units 1 exec/s 0
153 #1 INITED cov 3 bits 0 units 1 exec/s 0
154 #2 pulse cov 208 bits 0 units 1 exec/s 0
155 #2 NEW cov 208 bits 0 units 2 exec/s 0 L: 64
156 #3 NEW cov 217 bits 0 units 3 exec/s 0 L: 63
157 #4 pulse cov 217 bits 0 units 3 exec/s 0
159 * The ``Seed:`` line shows you the current random seed (you can change it with ``-seed=N`` flag).
160 * The ``READ`` line shows you how many input files were read (since you passed an empty dir there were inputs, but one dummy input was synthesised).
161 * The ``INITED`` line shows you that how many inputs will be fuzzed.
162 * The ``NEW`` lines appear with the fuzzer finds a new interesting input, which is saved to the CORPUS dir. If multiple corpus dirs are given, the first one is used.
163 * The ``pulse`` lines appear periodically to show the current status.
165 Now, interrupt the fuzzer and run it again the same way. You will see::
168 #0 READ cov 0 bits 0 units 564 exec/s 0
169 #1 pulse cov 502 bits 0 units 564 exec/s 0
171 #512 pulse cov 2933 bits 0 units 564 exec/s 512
172 #564 INITED cov 2991 bits 0 units 344 exec/s 564
173 #1024 pulse cov 2991 bits 0 units 344 exec/s 1024
174 #1455 NEW cov 2995 bits 0 units 345 exec/s 1455 L: 49
176 This time you were running the fuzzer with a non-empty input corpus (564 items).
177 As the first step, the fuzzer minimized the set to produce 344 interesting items (the ``INITED`` line)
179 It is quite convenient to store test corpuses in git.
180 As an example, here is a git repository with test inputs for the above PCRE2 fuzzer::
182 git clone https://github.com/kcc/fuzzing-with-sanitizers.git
183 ./pcre_fuzzer ./fuzzing-with-sanitizers/pcre2/C1/
185 You may run ``N`` independent fuzzer jobs in parallel on ``M`` CPUs::
187 N=100; M=4; ./pcre_fuzzer ./CORPUS -jobs=$N -workers=$M
189 By default (``-reload=1``) the fuzzer processes will periodically scan the CORPUS directory
190 and reload any new tests. This way the test inputs found by one process will be picked up
193 If ``-workers=$M`` is not supplied, ``min($N,NumberOfCpuCore/2)`` will be used.
197 Remember Heartbleed_?
198 As it was recently `shown <https://blog.hboeck.de/archives/868-How-Heartbleed-couldve-been-found.html>`_,
199 fuzzing with AddressSanitizer can find Heartbleed. Indeed, here are the step-by-step instructions
200 to find Heartbleed with LibFuzzer::
202 wget https://www.openssl.org/source/openssl-1.0.1f.tar.gz
203 tar xf openssl-1.0.1f.tar.gz
204 COV_FLAGS="-fsanitize-coverage=edge,indirect-calls" # -fsanitize-coverage=8bit-counters
205 (cd openssl-1.0.1f/ && ./config &&
206 make -j 32 CC="clang -g -fsanitize=address $COV_FLAGS")
207 # Get and build LibFuzzer
208 svn co http://llvm.org/svn/llvm-project/llvm/trunk/lib/Fuzzer
209 clang -c -g -O2 -std=c++11 Fuzzer/*.cpp -IFuzzer
210 # Get examples of key/pem files.
211 git clone https://github.com/hannob/selftls
212 cp selftls/server* . -v
213 cat << EOF > handshake-fuzz.cc
214 #include <openssl/ssl.h>
215 #include <openssl/err.h>
220 SSL_load_error_strings();
221 ERR_load_BIO_strings();
222 OpenSSL_add_all_algorithms();
223 assert (sctx = SSL_CTX_new(TLSv1_method()));
224 assert (SSL_CTX_use_certificate_file(sctx, "server.pem", SSL_FILETYPE_PEM));
225 assert (SSL_CTX_use_PrivateKey_file(sctx, "server.key", SSL_FILETYPE_PEM));
228 extern "C" void LLVMFuzzerTestOneInput(unsigned char *Data, size_t Size) {
229 static int unused = Init();
230 SSL *server = SSL_new(sctx);
231 BIO *sinbio = BIO_new(BIO_s_mem());
232 BIO *soutbio = BIO_new(BIO_s_mem());
233 SSL_set_bio(server, sinbio, soutbio);
234 SSL_set_accept_state(server);
235 BIO_write(sinbio, Data, Size);
236 SSL_do_handshake(server);
241 clang++ -g handshake-fuzz.cc -fsanitize=address \
242 openssl-1.0.1f/libssl.a openssl-1.0.1f/libcrypto.a Fuzzer*.o
243 # Run 20 independent fuzzer jobs.
244 ./a.out -jobs=20 -workers=20
248 #1048576 pulse cov 3424 bits 0 units 9 exec/s 24385
249 =================================================================
250 ==17488==ERROR: AddressSanitizer: heap-buffer-overflow on address 0x629000004748 at pc 0x00000048c979 bp 0x7fffe3e864f0 sp 0x7fffe3e85ca8
251 READ of size 60731 at 0x629000004748 thread T0
252 #0 0x48c978 in __asan_memcpy
253 #1 0x4db504 in tls1_process_heartbeat openssl-1.0.1f/ssl/t1_lib.c:2586:3
254 #2 0x580be3 in ssl3_read_bytes openssl-1.0.1f/ssl/s3_pkt.c:1092:4
259 Data-flow-guided fuzzing
260 ------------------------
263 With an additional compiler flag ``-fsanitize-coverage=trace-cmp`` (see SanitizerCoverageTraceDataFlow_)
264 and extra run-time flag ``-use_traces=1`` the fuzzer will try to apply *data-flow-guided fuzzing*.
265 That is, the fuzzer will record the inputs to comparison instructions, switch statements,
266 and several libc functions (``memcmp``, ``strcmp``, ``strncmp``, etc).
267 It will later use those recorded inputs during mutations.
269 This mode can be combined with DataFlowSanitizer_ to achieve better sensitivity.
273 LibFuzzer can be used in parallel with AFL_ on the same test corpus.
274 Both fuzzers expect the test corpus to reside in a directory, one file per input.
275 You can run both fuzzers on the same corpus in parallel::
277 ./afl-fuzz -i testcase_dir -o findings_dir /path/to/program -r @@
278 ./llvm-fuzz testcase_dir findings_dir # Will write new tests to testcase_dir
280 Periodically restart both fuzzers so that they can use each other's findings.
282 How good is my fuzzer?
283 ----------------------
285 Once you implement your target function ``LLVMFuzzerTestOneInput`` and fuzz it to death,
286 you will want to know whether the function or the corpus can be improved further.
287 One easy to use metric is, of course, code coverage.
288 You can get the coverage for your corpus like this::
290 ASAN_OPTIONS=coverage_pcs=1 ./fuzzer CORPUS_DIR -runs=0
292 This will run all the tests in the CORPUS_DIR but will not generate any new tests
293 and dump covered PCs to disk before exiting.
294 Then you can subtract the set of covered PCs from the set of all instrumented PCs in the binary,
295 see SanitizerCoverage_ for details.
297 User-supplied mutators
298 ----------------------
300 LibFuzzer allows to use custom (user-supplied) mutators,
301 see FuzzerInterface.h_
303 Fuzzing components of LLVM
304 ==========================
308 The inputs are random pieces of C++-like text.
310 Build (make sure to use fresh clang as the host compiler)::
312 cmake -GNinja -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DLLVM_USE_SANITIZER=Address -DLLVM_USE_SANITIZE_COVERAGE=YES -DCMAKE_BUILD_TYPE=Release /path/to/llvm
313 ninja clang-format-fuzzer
315 ./bin/clang-format-fuzzer CORPUS_DIR
317 Optionally build other kinds of binaries (asan+Debug, msan, ubsan, etc).
319 Tracking bug: https://llvm.org/bugs/show_bug.cgi?id=23052
324 The behavior is very similar to ``clang-format-fuzzer``.
326 Tracking bug: https://llvm.org/bugs/show_bug.cgi?id=23057
331 Tracking bug: https://llvm.org/bugs/show_bug.cgi?id=24639
336 We have a buildbot that runs the above fuzzers for LLVM components
337 24/7/365 at http://lab.llvm.org:8011/builders/sanitizer-x86_64-linux-fuzzer .
339 Pre-fuzzed test inputs in git
340 -----------------------------
342 The buildbot occumulates large test corpuses over time.
343 The corpuses are stored in git on github and can be used like this::
345 git clone https://github.com/kcc/fuzzing-with-sanitizers.git
346 bin/clang-format-fuzzer fuzzing-with-sanitizers/llvm/clang-format/C1
347 bin/clang-fuzzer fuzzing-with-sanitizers/llvm/clang/C1/
348 bin/llvm-as-fuzzer fuzzing-with-sanitizers/llvm/llvm-as/C1 -only_ascii=1
352 =========================
354 Q. Why Fuzzer does not use any of the LLVM support?
355 ---------------------------------------------------
357 There are two reasons.
359 First, we want this library to be used outside of the LLVM w/o users having to
360 build the rest of LLVM. This may sound unconvincing for many LLVM folks,
361 but in practice the need for building the whole LLVM frightens many potential
362 users -- and we want more users to use this code.
364 Second, there is a subtle technical reason not to rely on the rest of LLVM, or
365 any other large body of code (maybe not even STL). When coverage instrumentation
366 is enabled, it will also instrument the LLVM support code which will blow up the
367 coverage set of the process (since the fuzzer is in-process). In other words, by
368 using more external dependencies we will slow down the fuzzer while the main
369 reason for it to exist is extreme speed.
371 Q. What about Windows then? The Fuzzer contains code that does not build on Windows.
372 ------------------------------------------------------------------------------------
374 The sanitizer coverage support does not work on Windows either as of 01/2015.
375 Once it's there, we'll need to re-implement OS-specific parts (I/O, signals).
377 Q. When this Fuzzer is not a good solution for a problem?
378 ---------------------------------------------------------
380 * If the test inputs are validated by the target library and the validator
381 asserts/crashes on invalid inputs, the in-process fuzzer is not applicable
382 (we could use fork() w/o exec, but it comes with extra overhead).
383 * Bugs in the target library may accumulate w/o being detected. E.g. a memory
384 corruption that goes undetected at first and then leads to a crash while
385 testing another input. This is why it is highly recommended to run this
386 in-process fuzzer with all sanitizers to detect most bugs on the spot.
387 * It is harder to protect the in-process fuzzer from excessive memory
388 consumption and infinite loops in the target library (still possible).
389 * The target library should not have significant global state that is not
390 reset between the runs.
391 * Many interesting target libs are not designed in a way that supports
392 the in-process fuzzer interface (e.g. require a file path instead of a
394 * If a single test run takes a considerable fraction of a second (or
395 more) the speed benefit from the in-process fuzzer is negligible.
396 * If the target library runs persistent threads (that outlive
397 execution of one test) the fuzzing results will be unreliable.
399 Q. So, what exactly this Fuzzer is good for?
400 --------------------------------------------
402 This Fuzzer might be a good choice for testing libraries that have relatively
403 small inputs, each input takes < 1ms to run, and the library code is not expected
404 to crash on invalid inputs.
405 Examples: regular expression matchers, text or binary format parsers.
409 * GLIBC: https://sourceware.org/glibc/wiki/FuzzingLibc
413 * http://git.musl-libc.org/cgit/musl/commit/?id=39dfd58417ef642307d90306e1c7e50aaec5a35c
414 * http://www.openwall.com/lists/oss-security/2015/03/30/3
416 * pugixml: https://github.com/zeux/pugixml/issues/39
418 * PCRE: Search for "LLVM fuzzer" in http://vcs.pcre.org/pcre2/code/trunk/ChangeLog?view=markup
420 * ICU: http://bugs.icu-project.org/trac/ticket/11838
424 * Clang: https://llvm.org/bugs/show_bug.cgi?id=23057
426 * Clang-format: https://llvm.org/bugs/show_bug.cgi?id=23052
428 * libc++: https://llvm.org/bugs/show_bug.cgi?id=24411
430 * llvm-as: https://llvm.org/bugs/show_bug.cgi?id=24639
434 .. _pcre2: http://www.pcre.org/
436 .. _AFL: http://lcamtuf.coredump.cx/afl/
438 .. _SanitizerCoverage: http://clang.llvm.org/docs/SanitizerCoverage.html
439 .. _SanitizerCoverageTraceDataFlow: http://clang.llvm.org/docs/SanitizerCoverage.html#tracing-data-flow
440 .. _DataFlowSanitizer: http://clang.llvm.org/docs/DataFlowSanitizer.html
442 .. _Heartbleed: http://en.wikipedia.org/wiki/Heartbleed
444 .. _FuzzerInterface.h: https://github.com/llvm-mirror/llvm/blob/master/lib/Fuzzer/FuzzerInterface.h