From b57b2073acf85e985f513a12d8aae725f8942689 Mon Sep 17 00:00:00 2001 From: van Hauser Date: Wed, 3 Jul 2019 12:05:58 +0200 Subject: LAF_... -> AFL_LLVM_LAF_... --- docs/ChangeLog | 2 ++ 1 file changed, 2 insertions(+) (limited to 'docs/ChangeLog') diff --git a/docs/ChangeLog b/docs/ChangeLog index 0d730118..b758b211 100644 --- a/docs/ChangeLog +++ b/docs/ChangeLog @@ -30,6 +30,8 @@ Version ++2.52d (tbd): - added a -s seed switch to allow afl run with a fixed initial seed that is not updated. this is good for performance and path discovery tests as the random numbers are deterministic then + - llvm_mode LAF_... env variables can now be specified as AFL_LLVM_LAF_... + that is longer but in line with other llvm specific env vars - ... your idea or patch? -- cgit 1.4.1 From 7f6aaa53147afd4feb549214f49d0f5f69e4af6c Mon Sep 17 00:00:00 2001 From: van Hauser Date: Fri, 5 Jul 2019 11:28:08 +0200 Subject: final touches --- docs/ChangeLog | 2 + docs/env_variables.txt | 7 +++- llvm_mode/Makefile | 6 +++ llvm_mode/README.neverzero | 22 +++++++++++ llvm_mode/afl-llvm-pass.so.cc | 91 ++++++++++++++++++------------------------- 5 files changed, 72 insertions(+), 56 deletions(-) create mode 100644 llvm_mode/README.neverzero (limited to 'docs/ChangeLog') diff --git a/docs/ChangeLog b/docs/ChangeLog index 73c69196..a533de05 100644 --- a/docs/ChangeLog +++ b/docs/ChangeLog @@ -17,6 +17,8 @@ sending a mail to . Version ++2.52d (tbd): ----------------------------- + - added never zero counters for afl-gcc and optional (because of an + optimization issue in llvm < 9) for llvm_mode (AFL_LLVM_NEVER_ZERO=1) - added whitelist support for llvm_mode via AFL_LLVM_WHITELIST to allow only to instrument what is actually interesting. Gives more speed and less map pollution (originally by choller@mozilla) diff --git a/docs/env_variables.txt b/docs/env_variables.txt index 725dc82e..f8c6c86a 100644 --- a/docs/env_variables.txt +++ b/docs/env_variables.txt @@ -108,8 +108,11 @@ Then there are a few specific features that are only available in llvm_mode: OTHER ===== - - Setting AFL_NZERO_COUNTS=1 during compilation will use counters - that skip zero on overflow. + - Setting export AFL_LLVM_NOT_ZERO=1 during compilation will use counters + that skip zero on overflow. This is the default for llvm >= 9, + however for llvm versions below that this will increase an unnecessary + slowdown due a performance issue that is only fixed in llvm 9+. + This feature increases path discovery by a little bit. Note that AFL_INST_RATIO will behave a bit differently than for afl-gcc, because functions are *not* instrumented unconditionally - so low values diff --git a/llvm_mode/Makefile b/llvm_mode/Makefile index 6b277536..76de10c0 100644 --- a/llvm_mode/Makefile +++ b/llvm_mode/Makefile @@ -25,11 +25,17 @@ VERSION = $(shell grep '^\#define VERSION ' ../config.h | cut -d '"' -f2) LLVM_CONFIG ?= llvm-config #LLVM_OK = $(shell $(LLVM_CONFIG) --version | egrep -q '^[5-6]' && echo 0 || echo 1 ) LLVM_UNSUPPORTED = $(shell $(LLVM_CONFIG) --version | egrep -q '^9|3.0' && echo 1 || echo 0 ) +LLVM_MAJOR = ($shell $(LLVM_CONFIG) --version | sed 's/\..*//') ifeq "$(LLVM_UNSUPPORTED)" "1" $(warn llvm_mode only supports versions 3.8.0 up to 8.x ) endif +# this is not visible yet: +ifeq "$(LLVM_MAJOR)" "9" + $(info llvm_mode deteted llvm 9, enabling neverZero implementation) +endif + CFLAGS ?= -O3 -funroll-loops CFLAGS += -Wall -D_FORTIFY_SOURCE=2 -g -Wno-pointer-sign \ -DAFL_PATH=\"$(HELPER_PATH)\" -DBIN_PATH=\"$(BIN_PATH)\" \ diff --git a/llvm_mode/README.neverzero b/llvm_mode/README.neverzero new file mode 100644 index 00000000..ef873acb --- /dev/null +++ b/llvm_mode/README.neverzero @@ -0,0 +1,22 @@ +Usage +===== + +In larger, complex or reiterative programs the map that collects the edge pairs +can easily fill up and wrap. +This is not that much of an issue - unless by chance it wraps just to a 0 +when the program execution ends. +In this case afl-fuzz is not able to see that the pair has been accessed and +will ignore it. + +NeverZero prevents this behaviour. If a counter wraps, it jumps over the 0 +directly to a 1. This improves path discovery (by a very little amount) +at a very little cost (one instruction per edge). + +This is implemented in afl-gcc, however for llvm_mode this is optional if +the llvm version is below 9 - as there is a perfomance bug that is only fixed +in version 9 and onwards. + +If you want to enable this for llvm < 9 then set + +export AFL_LLVM_NOT_ZERO=1 + diff --git a/llvm_mode/afl-llvm-pass.so.cc b/llvm_mode/afl-llvm-pass.so.cc index 176692e3..cfeff968 100644 --- a/llvm_mode/afl-llvm-pass.so.cc +++ b/llvm_mode/afl-llvm-pass.so.cc @@ -118,7 +118,9 @@ bool AFLCoverage::runOnModule(Module &M) { } - char* neverZero_counters_str = getenv("AFL_NZERO_COUNTS"); +#if LLVM_VERSION_MAJOR < 9 + char* neverZero_counters_str = getenv("AFL_LLVM_NOT_ZERO"); +#endif /* Get globals for the SHM region and the previous location. Note that __afl_prev_loc is thread-local. */ @@ -236,75 +238,56 @@ bool AFLCoverage::runOnModule(Module &M) { LoadInst *Counter = IRB.CreateLoad(MapPtrIdx); Counter->setMetadata(M.getMDKindID("nosanitize"), MDNode::get(C, None)); - Value *Incr; - // if (neverZero_counters_str == NULL || neverZero_counters_str[0] != '4') - Incr = IRB.CreateAdd(Counter, ConstantInt::get(Int8Ty, 1)); - - if (neverZero_counters_str != NULL) { - /* hexcoder: Realize a counter that skips zero during overflow. - * Once this counter reaches its maximum value, it next increments to 1 - * - * Instead of - * Counter + 1 -> Counter - * we inject now this - * Counter + 1 -> {Counter, OverflowFlag} - * Counter + OverflowFlag -> Counter - */ - - // Solution #1 - creates - //mov ecx,edx - //add cl,0x1 - //adc dl,0x1 -/* - if (neverZero_counters_str[0] == '1') { - CallInst *AddOv = IRB.CreateBinaryIntrinsic(Intrinsic::uadd_with_overflow, Counter, ConstantInt::get(Int8Ty, 1)); - AddOv->setMetadata(M.getMDKindID("nosanitize"), MDNode::get(C, None)); - Value *SumWithOverflowBit = AddOv; - Incr = IRB.CreateAdd(IRB.CreateExtractValue(SumWithOverflowBit, 0), // sum - IRB.CreateZExt( // convert from one bit type to 8 bits type - IRB.CreateExtractValue(SumWithOverflowBit, 1), // overflow - Int8Ty)); - - // Solution #2 - creates the same code as #1 + Value *Incr = IRB.CreateAdd(Counter, ConstantInt::get(Int8Ty, 1)); + +#if LLVM_VERSION_MAJOR < 9 + if (neverZero_counters_str != NULL) { // with llvm 9 we make this the default as the bug in llvm is then fixed +#endif + /* hexcoder: Realize a counter that skips zero during overflow. + * Once this counter reaches its maximum value, it next increments to 1 + * + * Instead of + * Counter + 1 -> Counter + * we inject now this + * Counter + 1 -> {Counter, OverflowFlag} + * Counter + OverflowFlag -> Counter + */ +/* // we keep the old solutions just in case + // Solution #1 + if (neverZero_counters_str[0] == '1') { + CallInst *AddOv = IRB.CreateBinaryIntrinsic(Intrinsic::uadd_with_overflow, Counter, ConstantInt::get(Int8Ty, 1)); + AddOv->setMetadata(M.getMDKindID("nosanitize"), MDNode::get(C, None)); + Value *SumWithOverflowBit = AddOv; + Incr = IRB.CreateAdd(IRB.CreateExtractValue(SumWithOverflowBit, 0), // sum + IRB.CreateZExt( // convert from one bit type to 8 bits type + IRB.CreateExtractValue(SumWithOverflowBit, 1), // overflow + Int8Ty)); + // Solution #2 } else if (neverZero_counters_str[0] == '2') { - auto cf = IRB.CreateICmpULT(Incr, ConstantInt::get(Int8Ty, 1)); - Incr = IRB.CreateAdd(Incr, cf); - - // Solution #3 - creates - //add cl,0x1 - //cmp cl,0x1 - //adc cl,0x0 - } else if (neverZero_counters_str[0] == '3') { - auto cf = IRB.CreateICmpEQ(Incr, ConstantInt::get(Int8Ty, 0)); - Incr = IRB.CreateAdd(Incr, cf); - // Solution #4 - creates - // cmp dl, $0xff - // sete cl - // add dl,cl - // add dl,0x01 - } else if (neverZero_counters_str[0] == '4') { auto cf = IRB.CreateICmpEQ(Counter, ConstantInt::get(Int8Ty, 255)); Value *HowMuch = IRB.CreateAdd(ConstantInt::get(Int8Ty, 1), cf); Incr = IRB.CreateAdd(Counter, HowMuch); - - } else if (neverZero_counters_str[0] == '5') { + // Solution #3 + } else if (neverZero_counters_str[0] == '3') { */ + // this is the solution we choose because llvm9 should do the right thing here auto cf = IRB.CreateICmpEQ(Incr, ConstantInt::get(Int8Ty, 0)); auto carry = IRB.CreateZExt(cf, Int8Ty); Incr = IRB.CreateAdd(Incr, carry); /* - } else if (neverZero_counters_str[0] == '6') { + // Solution #4 + } else if (neverZero_counters_str[0] == '4') { auto cf = IRB.CreateICmpULT(Incr, ConstantInt::get(Int8Ty, 1)); auto carry = IRB.CreateZExt(cf, Int8Ty); Incr = IRB.CreateAdd(Incr, carry); - - // no other implementations yet - } else { + } else { fprintf(stderr, "Error: unknown value for AFL_NZERO_COUNTS: %s (valid is 1-4)\n", neverZero_counters_str); exit(-1); - } + } */ +#if LLVM_VERSION_MAJOR < 9 } +#endif IRB.CreateStore(Incr, MapPtrIdx)->setMetadata(M.getMDKindID("nosanitize"), MDNode::get(C, None)); -- cgit 1.4.1 From 71e22d9263fea7709c424ede2259332f8685498d Mon Sep 17 00:00:00 2001 From: van Hauser Date: Mon, 8 Jul 2019 11:39:06 +0200 Subject: updated docs --- docs/ChangeLog | 1 + docs/PATCHES | 1 + 2 files changed, 2 insertions(+) (limited to 'docs/ChangeLog') diff --git a/docs/ChangeLog b/docs/ChangeLog index b8d0d7ac..b4aec9ec 100644 --- a/docs/ChangeLog +++ b/docs/ChangeLog @@ -17,6 +17,7 @@ sending a mail to . Version ++2.52d (tbd): ----------------------------- + - added MOpt (github.com/puppet-meteor/MOpt-AFL) mode - added never zero counters for afl-gcc and optional (because of an optimization issue in llvm < 9) for llvm_mode (AFL_LLVM_NEVER_ZERO=1) - more cpu power for afl-system-config diff --git a/docs/PATCHES b/docs/PATCHES index cb050218..06da053e 100644 --- a/docs/PATCHES +++ b/docs/PATCHES @@ -17,6 +17,7 @@ afl-qemu-optimize-entrypoint.diff by mh(at)mh-sec(dot)de afl-qemu-speed.diff by abiondo on github afl-qemu-optimize-map.diff by mh(at)mh-sec(dot)de ++ MOpt (github.com/puppet-meteor/MOpt-AFL) was imported + AFLfast additions (github.com/mboehme/aflfast) were incorporated. + Qemu 3.1 upgrade with enhancement patches (github.com/andreafioraldi/afl) + Python mutator modules support (github.com/choller/afl) -- cgit 1.4.1 From eddfddccb2eb385afd076006e01fdfebacad9bec Mon Sep 17 00:00:00 2001 From: van Hauser Date: Fri, 12 Jul 2019 18:17:32 +0200 Subject: -E option and docu update --- afl-fuzz.c | 26 ++++++++-- docs/ChangeLog | 3 ++ docs/binaryonly_fuzzing.txt | 115 ++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 139 insertions(+), 5 deletions(-) create mode 100644 docs/binaryonly_fuzzing.txt (limited to 'docs/ChangeLog') diff --git a/afl-fuzz.c b/afl-fuzz.c index 484a351c..1e27cb34 100644 --- a/afl-fuzz.c +++ b/afl-fuzz.c @@ -88,6 +88,8 @@ u64 total_puppet_find = 0; u64 temp_puppet_find = 0; u64 most_time_key = 0; u64 most_time_puppet = 0; +u64 most_execs_key = 0; +u64 most_execs = 0; u64 old_hit_count = 0; int SPLICE_CYCLES_puppet; int limit_time_sig = 0; @@ -11344,18 +11346,21 @@ static void usage(u8* argv0) { " -n - fuzz without instrumentation (dumb mode)\n" " -x dir - optional fuzzer dictionary (see README)\n\n" + "Testing settings:\n" + " -s seed - use a fixed seed for the RNG\n" + " -V seconds - fuzz for a maximum total time of seconds then terminate\n" + " -E execs - fuzz for a maximum number of total executions then terminate\n\n" + "Other stuff:\n" " -T text - text banner to show on the screen\n" " -M / -S id - distributed mode (see parallel_fuzzing.txt)\n" " -C - crash exploration mode (the peruvian rabbit thing)\n" - " -V seconds - fuzz for a maximum total time of seconds then terminate\n" - " -s seed - use a fixed seed for the rng - important to testing\n" " -e ext - File extension for the temporarily generated test case\n\n" #ifdef USE_PYTHON "Compiled with Python 2.7 module support, see docs/python_mutators.txt\n" #endif - "For additional tips, please consult %s/README.\n\n", + "For additional tips, please consult %s/README\n\n", argv0, EXEC_TIMEOUT, MEM_LIMIT, doc_path); @@ -12050,7 +12055,7 @@ int main(int argc, char** argv) { gettimeofday(&tv, &tz); init_seed = tv.tv_sec ^ tv.tv_usec ^ getpid(); - while ((opt = getopt(argc, argv, "+i:o:f:m:t:T:dnCB:S:M:x:Qe:p:s:V:L:")) > 0) + while ((opt = getopt(argc, argv, "+i:o:f:m:t:T:dnCB:S:M:x:Qe:p:s:V:E:L:")) > 0) switch (opt) { @@ -12251,13 +12256,20 @@ int main(int argc, char** argv) { break; - case 'V': { + case 'V': { most_time_key = 1; if (sscanf(optarg, "%llu", &most_time_puppet) < 1 || optarg[0] == '-') FATAL("Bad syntax used for -V"); } break; + case 'E': { + most_execs_key = 1; + if (sscanf(optarg, "%llu", &most_execs) < 1 || optarg[0] == '-') + FATAL("Bad syntax used for -E"); + } + break; + case 'L': { /* MOpt mode */ if (limit_time_sig) FATAL("Multiple -L options not supported"); @@ -12583,6 +12595,10 @@ int main(int argc, char** argv) { if (most_time_puppet * 1000 < cur_ms_lv - start_time) break; } + if (most_execs_key == 1) { + if (most_execs >= total_execs) + break; + } } if (queue_cur) show_stats(); diff --git a/docs/ChangeLog b/docs/ChangeLog index b4aec9ec..dca674a2 100644 --- a/docs/ChangeLog +++ b/docs/ChangeLog @@ -20,6 +20,7 @@ Version ++2.52d (tbd): - added MOpt (github.com/puppet-meteor/MOpt-AFL) mode - added never zero counters for afl-gcc and optional (because of an optimization issue in llvm < 9) for llvm_mode (AFL_LLVM_NEVER_ZERO=1) + - added a new doc about binary only fuzzing: docs/binaryonly_fuzzing.txt - more cpu power for afl-system-config - added forkserver patch to afl-tmin, makes it much faster (originally from github.com/nccgroup/TriforceAFL) @@ -30,6 +31,8 @@ Version ++2.52d (tbd): see docs/python_mutators.txt (originally by choller@mozilla) - added AFL_CAL_FAST for slow applications and AFL_DEBUG_CHILD_OUTPUT for debugging + - added -V time and -E execs option to better comparison runs, runs afl-fuzz + for a specific time/executions. - added a -s seed switch to allow afl run with a fixed initial seed that is not updated. this is good for performance and path discovery tests as the random numbers are deterministic then diff --git a/docs/binaryonly_fuzzing.txt b/docs/binaryonly_fuzzing.txt new file mode 100644 index 00000000..ed654e2a --- /dev/null +++ b/docs/binaryonly_fuzzing.txt @@ -0,0 +1,115 @@ + +Fuzzing binary-only programs with afl++ +======================================= + +afl++, libfuzzer and others are great if you have the source code, and +it allows for very fast and coverage guided fuzzing. + +However, if there is only the binary program and not source code available, +then standard afl++ (dumb mode) is not effective. + +The following is a description of how these can be fuzzed with afl++ + +!!!!! +DTLR: try DYNINST with afl-dyninst. If it produces too many crashes then + use afl -Q qemu_mode. +!!!!! + + +QEMU +---- +Qemu is the "native" solution to the program. +It is available in the ./qemu_mode/ directory and once compiled it can +be accessed by the afl-fuzz -Q command line option. +The speed decrease is at about 50% +It the easiest to use alternative and even works for cross-platform binaries. + +As it is included in afl++ this needs no URL. + + +DYNINST +------- +Dyninst is a binary instrumentation framework similar to Pintool and Dynamorio +(see far below). Howver whereas Pintool and Dynamorio work at runtime, dyninst +instruments the target at load time, and then let it run. +This is great for some things, e.g. fuzzing, and not so effective for others, +e.g. malware analysis. + +So what we can do with dyninst is taking every basic block, and put afl's +instrumention code in there - and then save the binary. +Afterwards we can just fuzz the newly saved target binary with afl-fuzz. +Sounds great? It is. The issue though - this is a non-trivial problem to +insert instructions, which changes addresses in the process space and that +everything still works afterwards. Hence more often than not binaries +crash when they are run. + +The speed decrease is about 25-35% + +So if dyninst works, its the best option available. Otherwise it just doesn't +work well. + +https://github.com/vanhauser-thc/afl-dyninst + + +INTEL-PT +-------- +The big issue with Intel's PT is the small buffer size and the complex +encoding of the debug information collected through PT. +This makes the decoding very CPU intensive, hence slow and using up twice +the CPU resources. So to fairly compare Intel PT based afl fuzzers with +native afl or afl qemu we need to calculate in the higher CPU resources used. +As a result, the overall speed decrease is about 85-90% + +there are two afl intel-pt implementations: + +1. https://github.com/junxzm1990/afl-pt + => this needs Ubuntu 14.04.05 without any updates and the 4.4 kernel. + +2. https://github.com/hunter-ht-2018/ptfuzzer + => this needs a 4.14 or 4.15 kernel. the "nopti" kernel boot option must + be used + + +CORESIGHT +--------- + +Coresight is the ARM answer to Intel's PT. +There is no implementation so far which handle coresight and getting +it working on an ARM Linux is very difficult due custom kernel building +on embedded systems is difficult. And finding one that has coresight in +the ARM chip is difficult too. +My guess is that it is slower than Qemu, but faster than Intel PT. +If anyone finds any coresight implemention for afl please ping me: +vh@thc.org + + +PIN & DYNAMORIO +--------------- + +Pintool and Dynamorio are dynamic instrumentation engines, and they can be +used for getting basic block information at runtime. +Pintool is only available for Intel x32/x64 on Linux, Mac OS and Windows +whereas Dynamorio is additionally available for ARM and AARCH64. +Dynamorio is also 10x faster than Pintool. + +The big issue with Dynamorio (and therefore Pintool too) is speed. +Dynamorio has a speed decrease of 98-99% +Pintool has a speed decrease of 99.5% + +Hence Dynamorio is the option to go for if everything fails, and Pintool +only if Dynamorio fails too. + +Dynamorio solutions: + https://github.com/vanhauser-thc/afl-dynamorio + https://github.com/mxmssh/drAFL + https://github.com/googleprojectzero/winafl/ <= very good but windows only + +Pintool solutions: + https://github.com/vanhauser-thc/afl-pin + https://github.com/mothran/aflpin + https://github.com/spinpx/afl_pin_mode <= only old Pintool version supported + + +That's it! +News, corrections, updates? +Email vh@thc.org -- cgit 1.4.1