diff options
Diffstat (limited to 'instrumentation')
| -rw-r--r-- | instrumentation/README.llvm.md | 24 | ||||
| -rw-r--r-- | instrumentation/SanitizerCoverageLTO.so.cc | 32 | ||||
| -rw-r--r-- | instrumentation/SanitizerCoveragePCGUARD.so.cc | 2 | ||||
| -rw-r--r-- | instrumentation/afl-compiler-rt.o.c | 249 | ||||
| -rw-r--r-- | instrumentation/afl-llvm-common.cc | 2 | ||||
| -rw-r--r-- | instrumentation/afl-llvm-dict2file.so.cc | 2 |
6 files changed, 294 insertions, 17 deletions
diff --git a/instrumentation/README.llvm.md b/instrumentation/README.llvm.md index c0677474..126cf1a2 100644 --- a/instrumentation/README.llvm.md +++ b/instrumentation/README.llvm.md @@ -280,3 +280,27 @@ Please note that the default counter implementations are not thread safe! Support for thread safe counters in mode LLVM CLASSIC can be activated with setting `AFL_LLVM_THREADSAFE_INST=1`. + +## 8) Source code coverage through instrumentation + +Measuring source code coverage is a common task in fuzzing, but it is very +difficut to do in some situations (e.g. when using snapshot fuzzing). + +When using the `AFL_LLVM_INSTRUMENT=llvm-codecov` option, afl-cc will use +native trace-pc-guard instrumentation but additionally select options that +are required to utilize the instrumentation for source code coverage. + +In particular, it will switch the instrumentation to be per basic block +instead of instrumenting edges, disable all guard pruning and enable the +experimental pc-table support that allows the runtime to gather 100% of +instrumented basic blocks at start, including their locations. + +Note: You must compile AFL with the `CODE_COVERAGE=1` option to enable the +respective parts in the AFL compiler runtime. Support is currently only +implemented for Nyx, but can in theory also work without Nyx. + +Note: You might have to adjust `MAP_SIZE_POW2` in include/config.h to ensure +that your coverage map is large enough to hold all basic blocks of your +target program without any collisions. + +More documentation on how to utilize this with Nyx will follow. diff --git a/instrumentation/SanitizerCoverageLTO.so.cc b/instrumentation/SanitizerCoverageLTO.so.cc index e779bb79..b3b0d2cd 100644 --- a/instrumentation/SanitizerCoverageLTO.so.cc +++ b/instrumentation/SanitizerCoverageLTO.so.cc @@ -1,4 +1,4 @@ -/* SanitizeCoverage.cpp ported to afl++ LTO :-) */ +/* SanitizeCoverage.cpp ported to AFL++ LTO :-) */ #define AFL_LLVM_PASS @@ -20,6 +20,8 @@ #if LLVM_VERSION_MAJOR < 17 #include "llvm/ADT/Triple.h" #include "llvm/Analysis/EHPersonalities.h" +#else + #include "llvm/IR/EHPersonalities.h" #endif #include "llvm/Analysis/PostDominators.h" #include "llvm/Analysis/ValueTracking.h" @@ -236,7 +238,7 @@ class ModuleSanitizerCoverageLTO SanitizerCoverageOptions Options; - // afl++ START + // AFL++ START // const SpecialCaseList * Allowlist; // const SpecialCaseList * Blocklist; uint32_t autodictionary = 1; @@ -262,7 +264,7 @@ class ModuleSanitizerCoverageLTO Value *MapPtrFixed = NULL; std::ofstream dFile; size_t found = 0; - // afl++ END + // AFL++ END }; @@ -406,7 +408,7 @@ bool ModuleSanitizerCoverageLTO::instrumentModule( Int8Ty = IRB.getInt8Ty(); Int1Ty = IRB.getInt1Ty(); - /* afl++ START */ + /* AFL++ START */ char *ptr; LLVMContext &Ctx = M.getContext(); Ct = &Ctx; @@ -980,7 +982,7 @@ bool ModuleSanitizerCoverageLTO::instrumentModule( } - // afl++ END + // AFL++ END SanCovTracePCIndir = M.getOrInsertFunction(SanCovTracePCIndirName, VoidTy, IntptrTy); @@ -1004,7 +1006,7 @@ bool ModuleSanitizerCoverageLTO::instrumentModule( for (auto &F : M) instrumentFunction(F, DTCallback, PDTCallback); - // afl++ START + // AFL++ START if (dFile.is_open()) dFile.close(); if (!getenv("AFL_LLVM_LTO_SKIPINIT") && @@ -1158,7 +1160,7 @@ bool ModuleSanitizerCoverageLTO::instrumentModule( } - // afl++ END + // AFL++ END // We don't reference these arrays directly in any of our runtime functions, // so we need to prevent them from being dead stripped. @@ -1215,10 +1217,10 @@ static bool shouldInstrumentBlock(const Function &F, const BasicBlock *BB, // (catchswitch blocks). if (BB->getFirstInsertionPt() == BB->end()) return false; - // afl++ START + // AFL++ START if (!Options.NoPrune && &F.getEntryBlock() == BB && F.size() > 1) return false; - // afl++ END + // AFL++ END if (Options.NoPrune || &F.getEntryBlock() == BB) return true; @@ -1260,10 +1262,10 @@ void ModuleSanitizerCoverageLTO::instrumentFunction( // if (Blocklist && Blocklist->inSection("coverage", "fun", F.getName())) // return; - // afl++ START + // AFL++ START if (!F.size()) return; if (!isInInstrumentList(&F, FMNAME)) return; - // afl++ END + // AFL++ END if (Options.CoverageType >= SanitizerCoverageOptions::SCK_Edge) SplitAllCriticalEdges( @@ -1561,7 +1563,7 @@ bool ModuleSanitizerCoverageLTO::InjectCoverage( for (size_t i = 0, N = AllBlocks.size(); i < N; i++) { - // afl++ START + // AFL++ START if (BlockList.size()) { int skip = 0; @@ -1583,7 +1585,7 @@ bool ModuleSanitizerCoverageLTO::InjectCoverage( } - // afl++ END + // AFL++ END InjectCoverageAtBlock(F, *AllBlocks[i], i, IsLeafFunc); @@ -1649,7 +1651,7 @@ void ModuleSanitizerCoverageLTO::InjectCoverageAtBlock(Function &F, if (Options.TracePCGuard) { - // afl++ START + // AFL++ START ++afl_global_id; if (dFile.is_open()) { @@ -1713,7 +1715,7 @@ void ModuleSanitizerCoverageLTO::InjectCoverageAtBlock(Function &F, // done :) inst++; - // afl++ END + // AFL++ END /* XXXXXXXXXXXXXXXXXXX diff --git a/instrumentation/SanitizerCoveragePCGUARD.so.cc b/instrumentation/SanitizerCoveragePCGUARD.so.cc index 8be9e329..41c38283 100644 --- a/instrumentation/SanitizerCoveragePCGUARD.so.cc +++ b/instrumentation/SanitizerCoveragePCGUARD.so.cc @@ -16,6 +16,8 @@ #if LLVM_VERSION_MAJOR < 17 #include "llvm/ADT/Triple.h" #include "llvm/Analysis/EHPersonalities.h" +#else + #include "llvm/IR/EHPersonalities.h" #endif #include "llvm/Analysis/PostDominators.h" #include "llvm/IR/CFG.h" diff --git a/instrumentation/afl-compiler-rt.o.c b/instrumentation/afl-compiler-rt.o.c index 0912e52b..5372fae0 100644 --- a/instrumentation/afl-compiler-rt.o.c +++ b/instrumentation/afl-compiler-rt.o.c @@ -14,6 +14,16 @@ */ +#ifdef __AFL_CODE_COVERAGE + #ifndef _GNU_SOURCE + #define _GNU_SOURCE + #endif + #ifndef __USE_GNU + #define __USE_GNU + #endif + #include <dlfcn.h> +#endif + #ifdef __ANDROID__ #include "android-ashmem.h" #endif @@ -105,6 +115,44 @@ u32 __afl_dictionary_len; u64 __afl_map_addr; u32 __afl_first_final_loc; +#ifdef __AFL_CODE_COVERAGE +typedef struct afl_module_info_t afl_module_info_t; + +struct afl_module_info_t { + + // A unique id starting with 0 + u32 id; + + // Name and base address of the module + char *name; + uintptr_t base_address; + + // PC Guard start/stop + u32 start; + u32 stop; + + // PC Table begin/end + const uintptr_t *pcs_beg; + const uintptr_t *pcs_end; + + u8 mapped; + + afl_module_info_t *next; + +}; + +typedef struct { + + uintptr_t PC, PCFlags; + +} PCTableEntry; + +afl_module_info_t *__afl_module_info = NULL; + +u32 __afl_pcmap_size = 0; +uintptr_t *__afl_pcmap_ptr = NULL; +#endif // __AFL_CODE_COVERAGE + /* 1 if we are running in afl, and the forkserver was started, else 0 */ u32 __afl_connected = 0; @@ -499,6 +547,7 @@ static void __afl_map_shm(void) { u8 *map_env = (u8 *)getenv("AFL_MAP_SIZE"); if (!map_env || atoi((char *)map_env) < MAP_SIZE) { + fprintf(stderr, "FS_ERROR_MAP_SIZE\n"); send_forkserver_error(FS_ERROR_MAP_SIZE); _exit(1); @@ -678,6 +727,27 @@ static void __afl_map_shm(void) { } +#ifdef __AFL_CODE_COVERAGE + char *pcmap_id_str = getenv("__AFL_PCMAP_SHM_ID"); + + if (pcmap_id_str) { + + __afl_pcmap_size = __afl_map_size * sizeof(void *); + u32 shm_id = atoi(pcmap_id_str); + + __afl_pcmap_ptr = (uintptr_t *)shmat(shm_id, NULL, 0); + + if (__afl_debug) { + + fprintf(stderr, "DEBUG: Received %p via shmat for pcmap\n", + __afl_pcmap_ptr); + + } + + } + +#endif // __AFL_CODE_COVERAGE + } /* unmap SHM. */ @@ -686,6 +756,17 @@ static void __afl_unmap_shm(void) { if (!__afl_already_initialized_shm) return; +#ifdef __AFL_CODE_COVERAGE + if (__afl_pcmap_size) { + + shmdt((void *)__afl_pcmap_ptr); + __afl_pcmap_ptr = NULL; + __afl_pcmap_size = 0; + + } + +#endif // __AFL_CODE_COVERAGE + char *id_str = getenv(SHM_ENV_VAR); if (id_str) { @@ -1507,6 +1588,102 @@ void __sanitizer_cov_trace_pc_guard(uint32_t *guard) { } +#ifdef __AFL_CODE_COVERAGE +void __sanitizer_cov_pcs_init(const uintptr_t *pcs_beg, + const uintptr_t *pcs_end) { + + if (__afl_debug) { + + fprintf(stderr, "DEBUG: __sanitizer_cov_pcs_init called\n"); + + } + + // If for whatever reason, we cannot get dlinfo here, then pc_guard_init also + // couldn't get it and we'd end up attributing to the wrong module. + Dl_info dlinfo; + if (!dladdr(__builtin_return_address(0), &dlinfo)) { + + fprintf(stderr, + "WARNING: Ignoring __sanitizer_cov_pcs_init callback due to " + "missing module info\n"); + return; + + } + + afl_module_info_t *last_module_info = __afl_module_info; + while (last_module_info && last_module_info->next) { + + last_module_info = last_module_info->next; + + } + + if (!last_module_info) { + + fprintf(stderr, + "ERROR: __sanitizer_cov_pcs_init called with no module info?!\n"); + abort(); + + } + + last_module_info->pcs_beg = pcs_beg; + last_module_info->pcs_end = pcs_end; + + // Now update the pcmap. If this is the last module coming in, after all + // pre-loaded code, then this will also map all of our delayed previous + // modules. + + if (!__afl_pcmap_ptr) { return; } + + for (afl_module_info_t *mod_info = __afl_module_info; mod_info; + mod_info = mod_info->next) { + + if (mod_info->mapped) { continue; } + + PCTableEntry *start = (PCTableEntry *)(mod_info->pcs_beg); + PCTableEntry *end = (PCTableEntry *)(mod_info->pcs_end); + + u32 in_module_index = 0; + + while (start < end) { + + if (mod_info->start + in_module_index >= __afl_map_size) { + + fprintf(stderr, "ERROR: __sanitizer_cov_pcs_init out of bounds?!\n"); + abort(); + + } + + uintptr_t PC = start->PC; + + // This is what `GetPreviousInstructionPc` in sanitizer runtime does + // for x86/x86-64. Needs more work for ARM and other archs. + PC = PC - 1; + + // Calculate relative offset in module + PC = PC - mod_info->base_address; + + __afl_pcmap_ptr[mod_info->start + in_module_index] = PC; + + start++; + in_module_index++; + + } + + mod_info->mapped = 1; + + if (__afl_debug) { + + fprintf(stderr, "DEBUG: __sanitizer_cov_pcs_init initialized %u PCs\n", + in_module_index); + + } + + } + +} + +#endif // __AFL_CODE_COVERAGE + /* Init callback. Populates instrumentation IDs. Note that we're using ID of 0 as a special value to indicate non-instrumented bits. That may still touch the bitmap, but in a fairly harmless way. */ @@ -1538,6 +1715,62 @@ void __sanitizer_cov_trace_pc_guard_init(uint32_t *start, uint32_t *stop) { if (start == stop || *start) { return; } +#ifdef __AFL_CODE_COVERAGE + u32 *orig_start = start; + afl_module_info_t *mod_info = NULL; + + Dl_info dlinfo; + if (dladdr(__builtin_return_address(0), &dlinfo)) { + + if (__afl_already_initialized_forkserver) { + + fprintf(stderr, "[pcmap] Error: Module was not preloaded: %s\n", + dlinfo.dli_fname); + + } else { + + afl_module_info_t *last_module_info = __afl_module_info; + while (last_module_info && last_module_info->next) { + + last_module_info = last_module_info->next; + + } + + mod_info = malloc(sizeof(afl_module_info_t)); + + mod_info->id = last_module_info ? last_module_info->id + 1 : 0; + mod_info->name = strdup(dlinfo.dli_fname); + mod_info->base_address = (uintptr_t)dlinfo.dli_fbase; + mod_info->start = 0; + mod_info->stop = 0; + mod_info->pcs_beg = NULL; + mod_info->pcs_end = NULL; + mod_info->mapped = 0; + mod_info->next = NULL; + + if (last_module_info) { + + last_module_info->next = mod_info; + + } else { + + __afl_module_info = mod_info; + + } + + fprintf(stderr, "[pcmap] Module: %s Base Address: %p\n", dlinfo.dli_fname, + dlinfo.dli_fbase); + + } + + } else { + + fprintf(stderr, "[pcmap] dladdr call failed\n"); + + } + +#endif // __AFL_CODE_COVERAGE + x = getenv("AFL_INST_RATIO"); if (x) { @@ -1625,6 +1858,22 @@ void __sanitizer_cov_trace_pc_guard_init(uint32_t *start, uint32_t *stop) { } +#ifdef __AFL_CODE_COVERAGE + if (mod_info) { + + mod_info->start = *orig_start; + mod_info->stop = *(stop - 1); + if (__afl_debug) { + + fprintf(stderr, "DEBUG: [pcmap] Start Index: %u Stop Index: %u\n", + mod_info->start, mod_info->stop); + + } + + } + +#endif // __AFL_CODE_COVERAGE + if (__afl_debug) { fprintf(stderr, diff --git a/instrumentation/afl-llvm-common.cc b/instrumentation/afl-llvm-common.cc index 5d82aa25..7f17b02d 100644 --- a/instrumentation/afl-llvm-common.cc +++ b/instrumentation/afl-llvm-common.cc @@ -584,7 +584,7 @@ bool isInInstrumentList(llvm::Function *F, std::string Filename) { } // Calculate the number of average collisions that would occur if all -// location IDs would be assigned randomly (like normal afl/afl++). +// location IDs would be assigned randomly (like normal afl/AFL++). // This uses the "balls in bins" algorithm. unsigned long long int calculateCollisions(uint32_t edges) { diff --git a/instrumentation/afl-llvm-dict2file.so.cc b/instrumentation/afl-llvm-dict2file.so.cc index 97155cd6..8ee13010 100644 --- a/instrumentation/afl-llvm-dict2file.so.cc +++ b/instrumentation/afl-llvm-dict2file.so.cc @@ -746,7 +746,7 @@ static void registerAFLdict2filePass(const PassManagerBuilder &, } static RegisterPass<AFLdict2filePass> X("afl-dict2file", - "afl++ dict2file instrumentation pass", + "AFL++ dict2file instrumentation pass", false, false); static RegisterStandardPasses RegisterAFLdict2filePass( |