Environment variables

June 30, 2026 · View on GitHub

This document discusses the environment variables used by AFL++ to expose various exotic functions that may be (rarely) useful for power users or for some types of custom fuzzing setups. For general information about AFL++, see README.md.

Note: Most tools will warn on any unknown AFL++ environment variables; for example, because of typos. If you want to disable this check, then set the AFL_IGNORE_UNKNOWN_ENVS environment variable.

1) Settings for all compilers

Starting with AFL++ 3.0, there is only one compiler: afl-cc.

To select the different instrumentation modes, use one of the following options:

  • Pass the --afl-MODE command-line option to the compiler. Only this option accepts further AFL-specific command-line options.

  • Use a symlink to afl-cc: afl-clang, afl-clang++, afl-clang-fast, afl-clang-fast++, afl-clang-lto, afl-clang-lto++, afl-g++, afl-g++-fast, afl-gcc, afl-gcc-fast. This option does not accept AFL-specific command-line options. Instead, use environment variables.

  • Use the AFL_CC_COMPILER environment variable with MODE. To select MODE, use one of the following values:

    • GCC_PLUGIN (afl-g*-fast)
    • LLVM (afl-clang-fast*)
    • LTO (afl-clang-lto*).

The compile-time tools do not accept AFL-specific command-line options. The --afl-MODE command line option is the only exception. The other options make fairly broad use of environment variables instead:

  • Some build/configure scripts break with AFL++ compilers. To be able to pass them, do:

          export CC=afl-cc
      export CXX=afl-c++
      export AFL_NOOPT=1
      ./configure --disable-shared --disabler-werror
      unset AFL_NOOPT
      make

  • Setting AFL_CC, and AFL_CXX lets you use alternate downstream compilation tools, rather than the default 'clang', or 'gcc' binaries in your $PATH.

  • Setting AFL_COMPILER_LAUNCHER lets you prepend a command to the compiler invocation. This is useful for tools like ccache that can speed up compilation by caching object files. For example, setting AFL_COMPILER_LAUNCHER=ccache will prepend ccache to all compiler calls.

  • Most AFL tools do not print any output if stdout/stderr are redirected. If you want to get the output into a file, then set the AFL_DEBUG environment variable. This is sadly necessary for various build processes which fail otherwise.

  • By default, the wrapper appends -O3 to optimize builds. Very rarely, this will cause problems in programs built with -Werror, because -O3 enables more thorough code analysis and can spew out additional warnings. To disable optimizations, set AFL_DONT_OPTIMIZE. However, if -O... and/or -fno-unroll-loops are set, these are not overridden.

  • The optimization level can also be set with AFL_OPT_LEVEL, e.g. AFL_OPT_LEVEL=z for -Oz, default is 3

  • Setting AFL_HARDEN automatically adds code hardening options when invoking the downstream compiler. This currently includes -D_FORTIFY_SOURCE=2 and -fstack-protector-all. The setting is useful for catching non-crashing memory bugs at the expense of a very slight (sub-5%) performance loss.

  • Setting AFL_INST_RATIO to a percentage between 0 and 100 controls the probability of instrumenting every branch. This is (very rarely) useful when dealing with exceptionally complex programs that saturate the output bitmap. Examples include ffmpeg, perl, and v8.

    (If this ever happens, afl-fuzz will warn you ahead of the time by displaying the "bitmap density" field in fiery red.)

    Setting AFL_INST_RATIO to 0 is a valid choice. This will instrument only the transitions between function entry points, but not individual branches.

    Note that this is an outdated variable. Only the GCC plugin instrumentation honors it.

  • Setting AFL_INPUT_PLACEHOLDER to a string allows you to use that string as a placeholder instead of "@@" in the target command line arguments. Use this when "@@" conflicts with the parameters of your program. For eg. AFL_INPUT_PLACEHOLDER=NEW_PLACEHOLDER afl-fuzz -i in -o out -- ./targetProgram NEW_PLACEHOLDER

  • AFL_NO_BUILTIN causes the compiler to generate code suitable for use with libtokencap.so (but perhaps running a bit slower than without the flag).

  • AFL_PATH can be used to point a directory that contains LLVM/GCC plugins for AFL++, AFL++'s runtime objects and QEMU/Frida support files.

  • Setting AFL_QUIET will prevent afl-as and afl-cc banners from being displayed during compilation, in case you find them distracting.

  • Setting AFL_USE_... automatically enables supported sanitizers - provided that your compiler supports it. Available are:

    • AFL_USE_ASAN=1 - activates the address sanitizer (memory corruption detection)
    • AFL_USE_CFISAN=1 - activates the Control Flow Integrity sanitizer (e.g. type confusion vulnerabilities)
    • AFL_USE_LSAN - activates the leak sanitizer. To perform a leak check within your program at a certain point (such as at the end of an __AFL_LOOP()), you can run the macro __AFL_LEAK_CHECK(); which will cause an abort if any memory is leaked (you can combine this with the __AFL_LSAN_OFF(); and __AFL_LSAN_ON(); macros to avoid checking for memory leaks from memory allocated between these two calls.
    • AFL_USE_MSAN=1 - activates the memory sanitizer (uninitialized memory)
    • AFL_USE_TSAN=1 - activates the thread sanitizer to find thread race conditions
    • AFL_USE_UBSAN=1 - activates the undefined behavior sanitizer . AFL_USE_RTSAN . activates the realtime sanitizer (realtime violations in deterministic run time constraints). (clang 20 minimum)

  • AFL_LLVM_ONLY_FSRV/AFL_GCC_ONLY_FSRV will inject forkserver but not pc instrumentation. Please note this is different compared to AFL_LLVM_DISABLE_INSTRUMENTATION, which will totally disable forkserver implementation. This env is pretty useful in two cases:

    • SAND. In this case, the binaries built in this way will serve as extra oracles. Check the corresponding documents for details.
    • Compatible with LibAFL ForkserverExecutor implementation and thus faster to repeatedly run, compared to simple CommandExecutor.

  • TMPDIR is used by afl-as for temporary files; if this variable is not set, the tool defaults to /tmp.

2) Settings for LLVM and LTO: afl-clang-fast / afl-clang-fast++ / afl-clang-lto / afl-clang-lto++

The native instrumentation helpers (instrumentation and gcc_plugin) accept a subset of the settings discussed in section 1, with the exception of:

  • AFL_AS, since this toolchain does not directly invoke GNU as.

  • AFL_INST_RATIO, as we use collision free instrumentation by default. Not all passes support this option though as it is an outdated feature.

  • LLVM modes support AFL_LLVM_DICT2FILE=/absolute/path/file.txt which will write all constant string comparisons to this file to be used later with afl-fuzz' -x option.

  • An option to AFL_LLVM_DICT2FILE is AFL_LLVM_DICT2FILE_NO_MAIN=1 which skill not parse main().

  • TMPDIR and AFL_KEEP_ASSEMBLY, since no temporary assembly files are created.

  • LLVM modes compiling C++ will normally set rpath in the binary if LLVM is not in a usual location (/usr or /lib). Setting AFL_LLVM_NO_RPATH=1 disables this behaviour in case it isn't desired. For example, the compiling toolchain might be in a custom location, but the target machine has LLVM runtime libs in the search path.

Then there are a few specific features that are only available in instrumentation mode:

Select the instrumentation mode

AFL_LLVM_INSTRUMENT - this configures the instrumentation mode.

Available options:

  • CLANG - outdated clang instrumentation
  • CTX - context sensitive instrumentation (LTO mode only, see README.lto.md)
  • CALLER - single caller context instrumentation (LTO mode only)
  • GCC - outdated gcc instrumentation
  • LTO - LTO instrumentation
  • NATIVE - clang's original pcguard based instrumentation
  • PCGUARD - our own pcguard based instrumentation (default)

Bug-pass oracles

Setting any AFL_LLVM_BUG* variable during compilation enables afl-llvm-bug-pass.so, which adds runtime oracles for arithmetic-bound and logical-OOB bugs that ASan does not catch (the CVE-2023-4863 / libwebp-Huffman class).

  • AFL_LLVM_BUG=1 — enable all bug-pass modes (SCALAR, BUDGET, SIZEFILL, SLACK, ALLOCSIZE, and ALLOCSIZE_DERIVE). The size-derive mode only emits entries when a CmpLog map is available (needs a CmpLog binary for main fuzzing, so -c 0, not only -c target.cmplog).

  • AFL_LLVM_BUG_SCALAR=1 — max-value-per-arithmetic-site coverage and per-loop iteration counts. Treats internal scalar growth as a fitness signal in addition to edge coverage. Useful for finding inputs that drive computed sizes/lengths to corner cases.

  • AFL_LLVM_BUG_BUDGET=1 — at every call site of the form ptr += func() (or ptr = gep(ptr, func())), enforce that during the call all writes via the pointer arg lie in [ptr, ptr + return_value). Catches functions that write beyond what they claim to consume. This mode needs to see the callee body in the same LLVM module to instrument its stores; use LTO for cross-translation-unit APIs or expect those calls to be missed.

    Interaction with AFL_LLVM_BUG_SIZEFILL: when both are enabled, BUDGET additionally matches the out-param shape void fill(buf, size_t *out_n); fill(p, &n); p += n; (iconv-style streaming-parser APIs). The size value used for the budget check is the loaded *out_n rather than a return value. The validation hand-off is via the same findOutSizeParam precondition machinery SIZEFILL uses on its own call sites — so the FP surface is paired with SIZEFILL's existing one. Enabling BUDGET alone leaves the original ptr = gep(ptr, return_value) shape as the only matched pattern.

  • AFL_LLVM_BUG_SIZEFILL=1 — at every call site to a function that has a "NULL-means-size-only" sentinel parameter, verify the returned size and actual writes both fit the caller's known buffer. Like BUDGET, the store-tracking part is intra-module unless the target is built with LTO.

    Interaction with AFL_LLVM_BUG_ALLOCSIZE: SIZEFILL alone derives the per-call write-cap from the caller-buffer signature size plus a 64 KiB unrelated-store slack window (so an adjacent unrelated store inside the callee won't trip a spurious abort). When ALLOCSIZE is also enabled at compile time AND the buffer was registered by the ALLOCSIZE runtime (heap/calloc/realloc/posix_memalign/mmap/custom- allocator), the cap is tightened to the buffer's exact remaining extent. Users hitting an FP from SIZEFILL alone — typically a parser writing to an adjacent allocation within the 64 KiB slack — can usually silence it by also building with AFL_LLVM_BUG_ALLOCSIZE=1. Recommended canonical combination for SIZEFILL precision: AFL_LLVM_BUG_SIZEFILL=1 AFL_LLVM_BUG_ALLOCSIZE=1.

  • AFL_LLVM_BUG_ALLOCSIZE — enable the AllocSizeOracle pass: every call to malloc, calloc, realloc, posix_memalign, and free is rewritten to a tracked variant that records (base, size, alloc_site_id) in a runtime shadow table. Every store whose pointer base traces back to a tracked allocation emits one runtime hook with three feedback channels:

    1. Headroom (max-rule) — __afl_bug_map[hash(site)] holds the highest "closeness to end" seen across the run, so the fuzzer rewards inputs that approach but don't yet exceed the buffer.
    2. Proximity bucket — a synthetic edge hash(site, log2(remaining)) joining the bug-map; turns "wrote into the last 8/16/32/… bytes of buffer X" into a discoverable coverage event.
    3. Soft-OOB tripwire — abort()s with a diagnostic when a store hits or exceeds the recorded end, before ASan's shadow check would (useful for custom allocators ASan doesn't poison). Cost is one shadow lookup + one subtract + 2–3 max-rule writes per qualifying store. Combine with AFL_LLVM_BUG_BUDGET=1 and AFL_LLVM_BUG_SIZEFILL=1 to layer the aborting oracles.

    Interaction with AFL_LLVM_BUG_SIZEFILL: when both modes are enabled, SIZEFILL uses the ALLOCSIZE shadow to derive an exact per-call write-cap (the buffer's remaining extent), eliminating SIZEFILL's 64 KiB unrelated-store slack window. This both tightens SIZEFILL's tripwire (catches OOB at the buffer end precisely) and suppresses FPs from unrelated nearby allocations. See the AFL_LLVM_BUG_SIZEFILL entry above for details. The shadow itself spans up to 4 × 16 GiB windows pinned to live allocation bases — under glibc ASLR you typically need 1 (heap-only) or 2 (heap + mmap arena). When the cap is reached the runtime prints a one-shot [afl-bug] ALLOCSIZE: shadow window cap reached warning and subsequent allocations outside existing windows are silently untracked; pin the target's allocator with MALLOC_ARENA_MAX=1 (glibc) to mitigate.

    Coexistence with ASAN: when the target also links the AddressSanitizer runtime, ALLOCSIZE and ALLOCSIZE_DERIVE are disabled at startup by the bug-pass runtime — ASAN already enforces byte-granular OOB and reserves part of the address space for its shadow, so running both at once produces conflicting verdicts. The runtime emits a one-shot stderr note [afl-bug] ASAN detected; ALLOCSIZE/DERIVE modes disabled to avoid double-instrumentation and continues with the remaining modes (SCALAR/BUDGET/SIZEFILL/SLACK) enabled. For ALLOCSIZE runs, build the target without AFL_USE_ASAN.

  • AFL_LLVM_BUG_ALLOCSIZE_STACK — controls stack-alloca tracking. Default on when AFL_LLVM_BUG_ALLOCSIZE=1 is set; opt out with AFL_LLVM_BUG_ALLOCSIZE_STACK=0 for targets where the overhead is problematic (deep recursion, hot helpers with many large stack locals). When enabled, the pass walks each instrumented function's entry-block allocas; allocas whose static size is a multiple of the 64-byte shadow granule AND ≥ 64 bytes are registered via __afl_alloc_register after the alloca (or after llvm.lifetime.start when present) and unregistered before every ReturnInst/ResumeInst (or before llvm.lifetime.end). The pass forces alignment 64 on registered allocas so the painted shadow granules cover exactly the alloca — adjacent stack slots cannot trigger granule-aliasing FPs. Per-function cap: 16 instrumented allocas (more emit a one-shot [afl-bug] ALLOCSIZE: stack cap reached note and are skipped). Stack buffers < 64 bytes are not covered — pair with -fsanitize=address when fine-grained stack-OOB matters.

  • AFL_LLVM_BUG_ALLOCSIZE_FUNCS — comma-separated list of custom allocator function names that the pass should treat as additional allocator entry points (e.g., WebPSafeMalloc,WebPSafeCalloc). The pass inserts a post-call __afl_alloc_register and uses the widest integer argument of the callee as the size. Pair with AFL_LLVM_BUG_ALLOCSIZE=1. Note: targets must export the named function (non-static / extern linkage) so LLVM's IPO does not strip the size argument; static helpers can be specialized away by -O3.

  • AFL_LLVM_BUG_ALLOCSIZE_FREE_FUNCS — comma-separated list of the matching custom-free function names for the allocators listed in AFL_LLVM_BUG_ALLOCSIZE_FUNCS. Each free is rewritten to __afl_track_free so the runtime can drop the registration. Use a parallel ordering (same index = matching pair) when both are sets.

  • AFL_LLVM_BUG_SLACK=1 — per-icmp instrumentation: at every signed/ unsigned integer comparison the pass records |op0 - op1| mapped MIN- style (inverse-bucket) onto the shared bug map. Smaller slack = tighter match = more interesting. Helps the fuzzer drive multiple validation predicates toward their tight edges simultaneously (the libwebp CVE-2023-4863 pattern). Cheap; can be combined with any other mode.

  • AFL_LLVM_BUG_SCALAR_SLICE=1 — restrict SCALAR's arithmetic-site instrumentation to BinaryOperators that flow into a memory-size sink (allocator size args, GEP indices, memcpy/memset lengths). Implies AFL_LLVM_BUG_SCALAR=1. Off by default; turning it on silences pure- compute accumulators (hash builders, non-memory counters) but cuts SCALAR map pollution on very large targets at the cost of pure- compute coverage signal.

  • AFL_LLVM_BUG_ALLOCSIZE_DERIVE — enables ALLOCSIZE and, at every __afl_alloc_unregister (free of a tracked allocation), logs (record.size, record.max_observed_off) into a CmpLog routine slot keyed by (alloc-site ID, log2(record.size)). It needs a CmpLog binary (or AFL_CMPLOG_DEBUG=1) for the cmp_map to be allocated. Pair with afl-fuzz -l 2z to confirm the feature is in use; the fuzzer's existing CmpLog RTN dictionary mining harvests the entries automatically. This mode is also included by AFL_LLVM_BUG=1.

    The log-size component of the slot key spreads entries from a hot allocator (called from inner loops with varied sizes) across multiple CmpLog slots — without it, the slot's CMP_MAP_RTN_H hit cap saturates fast and every later size from that site is dropped. A site that allocates one fixed size still maps to one slot.

The runtime keeps its own private max-value bug-map (MAP_SIZE_BUG, 16384 u32 slots), separate from the IJON map, and reports oracle violations to stderr followed by abort().

Recommended usage:

  • Use normal coverage binaries as the main campaign baseline.
  • Add SCALAR and/or SLACK when you want extra guidance toward arithmetic bounds, tight comparisons, and computed-size corner cases.
  • Use BUDGET, SIZEFILL, and ALLOCSIZE as oracle builds for bug-finding or confirmation runs; they intentionally abort on detected contract violations.
  • Use AFL_LLVM_BUG_ALLOCSIZE_DERIVE=1 (or AFL_LLVM_BUG=1 when you want every bug-pass mode) with afl-fuzz -l 2Z as a CmpLog assistant build for mining allocation-size values.
  • Prefer LTO for BUDGET and SIZEFILL campaigns when the checked APIs are split across translation units.

CMPLOG

Setting AFL_LLVM_CMPLOG=1 during compilation will tell afl-clang-fast to produce a CmpLog binary.

For afl-gcc-fast, set AFL_GCC_CMPLOG=1 instead.

For more information, see instrumentation/README.cmplog.md.

CTX

Setting AFL_LLVM_CTX or AFL_LLVM_INSTRUMENT=CTX activates context sensitive branch coverage in LTO mode (afl-clang-lto) - meaning that each edge is additionally combined with its caller.

For more information, see instrumentation/README.lto.md.

INSTRUMENT LIST (selectively instrument files and functions)

This feature allows selective instrumentation of the source.

Setting AFL_LLVM_ALLOWLIST or AFL_LLVM_DENYLIST with a file name and/or function will only instrument (or skip) those files that match the names listed in the specified file.

For more information, see instrumentation/README.instrument_list.md.

INJECTIONS

This feature is able to find simple injection vulnerabilities in insecure calls to mysql/mariadb/nosql/postgresql/ldap and XSS in libxml2.

  • Setting AFL_LLVM_INJECTIONS_ALL will enable all injection hooking

  • Setting AFL_LLVM_INJECTIONS_SQL will enable SQL injection hooking

  • Setting AFL_LLVM_INJECTIONS_LDAP will enable LDAP injection hooking

  • Setting AFL_LLVM_INJECTIONS_XSS will enable XSS injection hooking

LAF-INTEL

This great feature will split compares into series of single byte comparisons to allow afl-fuzz to find otherwise rather impossible paths. It is not restricted to Intel CPUs. ;-)

  • Setting AFL_LLVM_LAF_TRANSFORM_COMPARES will split string compare functions.

  • Setting AFL_LLVM_LAF_SPLIT_COMPARES will split all floating point and 64, 32 and 16 bit integer CMP instructions.

  • Setting AFL_LLVM_LAF_SPLIT_FLOATS will split floating points, needs AFL_LLVM_LAF_SPLIT_COMPARES to be set.

  • Setting AFL_LLVM_LAF_SPLIT_SWITCHES will split all switch constructs.

  • Setting AFL_LLVM_LAF_ALL sets all of the above.

For more information, see instrumentation/README.laf-intel.md.

LTO

This is a different way of instrumentation: first it compiles all code in LTO (link time optimization) and then performs an edge inserting instrumentation which is 100% collision free (collisions are a big issue in AFL and AFL-like instrumentations). This is performed by using afl-clang-lto/afl-clang-lto++ instead of afl-clang-fast, but is only built if LLVM 11 or newer is used.

AFL_LLVM_INSTRUMENT=CFG will use Control Flow Graph instrumentation. (Not recommended for afl-clang-fast, default for afl-clang-lto as there it is a different and better kind of instrumentation.)

None of the following options are necessary to be used and are rather for manual use (which only ever the author of this LTO implementation will use). These are used if several separated instrumentations are performed which are then later combined.

  • AFL_LLVM_LTO_CALLER activates collision free CALLER instrumentation. AFL_LLVM_LTO_CTX is a synonym; AFL_LLVM_CALLER / AFL_LLVM_CTX also enable it.
  • AFL_LLVM_LTO_CALLER_DEPTH sets the maximum number of single-caller functions to walk up the call chain through to find a useful context split point. Default 0. Synonyms (checked in this precedence order): AFL_LLVM_LTO_CTX_DEPTH, AFL_LLVM_CALLER_DEPTH, AFL_LLVM_CTX_DEPTH.
  • AFL_LLVM_DOCUMENT_IDS=file will document to a file which edge ID was given to which function. This helps to identify functions with variable bytes or which functions were touched by an input.
  • AFL_LLVM_LTO_DONTWRITEID prevents that the highest location ID written into the instrumentation is set in a global variable.
  • AFL_LLVM_LTO_STARTID sets the starting location ID for the instrumentation. This defaults to 1.
  • AFL_LLVM_MAP_ADDR sets the fixed map address to a different address than the default 0x10000. A value of 0 or empty sets the map address to be dynamic (the original AFL way, which is slower).
  • AFL_LLVM_LTO_SKIPINIT skips adding initialization code. Some global vars (e.g. the highest location ID) are not injected. Needed to instrument with WAFL. For more information, see instrumentation/README.lto.md.

PATH (LTO and PCGUARD)

Setting AFL_LLVM_PATH (or AFL_LLVM_LTO_PATH / AFL_LLVM_PATH_MODE) under afl-clang-lto or afl-clang-fast (PCGUARD) enables Ball-Larus per-function path coverage in addition to the default edge coverage. Loops are treated as a single iteration (back-edges stripped). Functions with more than 100,000 acyclic paths that cannot be reduced by collapsing multi-way branches are skipped with a warning. The LTO build additionally composes with AFL_LLVM_LTO_CALLER to track (call_site, path) tuples.

Levels:

  • =1 — relaxed: every "guard-only" basic block (only loads/casts/ GEPs/cmps/phis/freezes/allocas/plain arithmetic + a terminator — no calls/stores/atomics) collapses via max() instead of sum(). Short-circuit &&/|| and switches on a bare loaded value collapse to one decision. Smallest map. An empty value (AFL_LLVM_PATH=) is rejected — set explicitly to 1/2/3/0.
  • =2 — restricted: like =1 but only 2-successor guard-only BBs collapse; switches/indirectbr keep their full multiplying effect.
  • =3 — strict Ball-Larus: every IR-level acyclic path is a unique slot.

AFL_LLVM_PATH_MAX_PATHS=N overrides the default 100,000-path cap above which a function is skipped (N >= 2). Useful for tightening or relaxing the cutoff on a per-target basis.

See instrumentation/README.lto.md and instrumentation/README.llvm.md for details.

NOT_ZERO

  • Setting 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.

  • Setting AFL_LLVM_SKIP_NEVERZERO=1 will not implement the skip zero test. If the target performs only a few loops, then this will give a small performance boost.

Thread safe instrumentation counters (in all modes)

Setting AFL_LLVM_THREADSAFE_INST will inject code that implements thread safe counters. The overhead is a little bit higher compared to the older non-thread safe case. Note that this disables neverzero (see NOT_ZERO).

Deny exec* calls

Setting AFL_LLVM_DENY_EXEC=1 during compilation will cause the instrumented binary to abort when any exec* family function is called. This is useful to prevent coverage map corruption that can occur when a target calls exec* functions, as the exec'd process will inherit the instrumentation but may not be the intended fuzzing target. Only enable this if your target should never call exec functions during normal operation.

Abort on entry of non-instrumented functions

When you build with an AFL_LLVM_ALLOWLIST or AFL_LLVM_DENYLIST (see INSTRUMENT LIST above), additionally setting AFL_LLVM_ABORTLIST=1 inserts an abort() call at the entry of every function that the allow/deny list excluded from instrumentation. Reaching such a function then crashes the target. This is useful to catch test cases that escape the intended fuzzing scope. It only affects functions skipped because of the allow/deny list. Functions that run automatically rather than through the fuzzing entry point are left untouched, so they cannot crash the target before, around or after the forkserver: compiler/sanitizer internals, available_externally definitions, constructors and destructors (C++ ctors/dtors and __attribute__((constructor))/ ((destructor)) functions), ifunc resolvers (run by the dynamic loader during relocation), exit/teardown callbacks registered with atexit, at_quick_exit, __cxa_atexit, __cxa_thread_atexit[_impl] or pthread_key_create, the LLVMFuzzerInitialize one-time harness setup function, and anything those reach through direct calls. It has no effect (and warns) if no allow/deny list is in use.

3) Settings for GCC / GCC_PLUGIN modes

There are a few specific features that are only available in GCC and GCC_PLUGIN mode.

  • GCC mode only: Setting AFL_KEEP_ASSEMBLY prevents afl-as from deleting instrumented assembly files. Useful for troubleshooting problems or understanding how the tool works.

    To get them in a predictable place, try something like:

    mkdir assembly_here
TMPDIR=$PWD/assembly_here AFL_KEEP_ASSEMBLY=1 make clean all

  • GCC_PLUGIN mode only: Setting AFL_GCC_INSTRUMENT_FILE or AFL_GCC_ALLOWLIST with a filename will only instrument those files that match the names listed in this file (one filename per line).

    Setting AFL_GCC_DENYLIST or AFL_GCC_BLOCKLIST with a file name and/or function will only skip those files that match the names listed in the specified file. See instrumentation/README.instrument_list.md for more information.

    Setting AFL_GCC_DISABLE_VERSION_CHECK=1 will disable the GCC plugin version check if the target GCC plugin differs from the system-installed version, resolving issues caused by version mismatches between GCC and the plugin.

    Setting AFL_GCC_OUT_OF_LINE=1 will instruct afl-gcc-fast to instrument the code with calls to an injected subroutine instead of the much more efficient inline instrumentation.

    Setting AFL_GCC_SKIP_NEVERZERO=1 will not implement the skip zero test. If the target performs only a few loops, then this will give a small performance boost.

4) Runtime settings

The following environment variables are for a compiled AFL++ target.

  • Setting AFL_DUMP_MAP_SIZE when executing the target directly will dump the map size of the target and exit.

  • Setting AFL_OLD_FORKSERVER will use the old AFL vanilla forkserver. This makes only sense when you a) compile in a classic colliding coverage mode (e.g. with the gcc plugin, afl-gcc-fast) or if the map size of the target is below MAP_SIZE (65536 by default), AND b) you want to use this compiled AFL++ target with a different tool that expects vanilla AFL behaviour, e.g. symcc, symqemu, nautilus, etc. You would use this option together with the target fuzzing application.

  • Setting AFL_DISABLE_LLVM_INSTRUMENTATION will disable collecting instrumentation. (More of an internal option.)

5) Settings for afl-fuzz

The main fuzzer binary accepts several options that disable a couple of sanity checks or alter some of the more exotic semantics of the tool:

  • Setting AFL_AUTORESUME will resume a fuzz run (same as providing -i -) for an existing out folder, even if a different -i was provided. Without this setting, afl-fuzz will refuse execution for a long-fuzzed out dir.

  • Benchmarking only: AFL_BENCH_JUST_ONE causes the fuzzer to exit after processing the first queue entry; and AFL_BENCH_UNTIL_CRASH causes it to exit soon after the first crash is found.

  • Setting AFL_ALLOW_CORES will allow writing core files on crashes. Not recommended unless you have crashes that do not reproduce stand-alone.

  • AFL_CMPLOG_ONLY_NEW will only perform the expensive cmplog feature for newly found test cases and not for test cases that are loaded on startup (-i in). This is an important feature to set when resuming a fuzzing session.

  • AFL_IGNORE_SEED_PROBLEMS will skip over crashes and timeouts in the seeds instead of exiting.

  • Setting AFL_CRASH_EXITCODE sets the exit code AFL++ treats as crash. For example, if AFL_CRASH_EXITCODE='-1' is set, each input resulting in a -1 return code (i.e. exit(-1) got called), will be treated as if a crash had occurred. This may be beneficial if you look for higher-level faulty conditions in which your target still exits gracefully.

  • Setting AFL_CUSTOM_MUTATOR_LIBRARY to a shared library with afl_custom_fuzz() creates additional mutations through this library. If afl-fuzz is compiled with Python (which is autodetected during building afl-fuzz), setting AFL_PYTHON_MODULE to a Python module can also provide additional mutations. If AFL_CUSTOM_MUTATOR_ONLY is also set, all mutations will solely be performed with the custom mutator. This feature allows to configure custom mutators which can be very helpful, e.g., fuzzing XML or other highly flexible structured input. For details, see custom_mutators.md.

  • Setting AFL_CUSTOM_MUTATOR_LATE_SEND will call the afl_custom_fuzz_send() function after the target has been restarted. (This is needed for e.g. TCP services.)

  • Setting AFL_CYCLE_SCHEDULES will switch to a different schedule every time a cycle is finished.

  • Setting AFL_DEBUG_CHILD will not suppress the child output. This lets you see all output of the child, making setup issues obvious. For example, in an unicornafl harness, you might see python stacktraces. You may also see other logs that way, indicating why the forkserver won't start. Not pretty but good for debugging purposes. Note that AFL_DEBUG_CHILD_OUTPUT is deprecated.

  • Setting AFL_DISABLE_TRIM tells afl-fuzz not to trim test cases. This is usually a bad idea!

  • Setting AFL_DISABLE_REDUNDANT disables any queue items that are redundant. This can be useful with huge queues.

  • Setting AFL_KEEP_TIMEOUTS will keep longer running inputs if they reach new coverage

  • On the contrary, if you are not interested in any timeouts, you can set AFL_IGNORE_TIMEOUTS to get a bit of speed instead.

  • AFL_EXIT_ON_SEED_ISSUES will restore the vanilla afl-fuzz behavior which does not allow crashes or timeout seeds in the initial -i corpus.

  • AFL_CRASHING_SEEDS_AS_NEW_CRASH will treat crashing seeds as new crash. these crashes will be written to crashes folder as op:dry_run, and orig:<seed_file_name>.

  • AFL_EXIT_ON_TIME causes afl-fuzz to terminate if no new paths were found within a specified period of time (in seconds). May be convenient for some types of automated jobs.

  • AFL_EXIT_WHEN_DONE causes afl-fuzz to terminate when all existing paths have been fuzzed and there were no new finds for a while. This is basically when the fuzzing state says state: finished

  • Setting AFL_EXPAND_HAVOC_NOW will start in the extended havoc mode that includes costly mutations. afl-fuzz automatically enables this mode when deemed useful otherwise.

  • AFL_FAST_CAL keeps the calibration stage about 2.5x faster (albeit less precise), which can help when starting a session against a slow target. AFL_CAL_FAST works too.

  • Setting AFL_FORCE_UI will force painting the UI on the screen even if no valid terminal was detected (for virtual consoles).

  • Setting AFL_FORKSRV_INIT_TMOUT allows you to specify a different timeout to wait for the forkserver to spin up. The specified value is the new timeout, in milliseconds. The default is the -t value times FORK_WAIT_MULT from config.h (usually 10), so for a -t 100, the default would wait for 1000 milliseconds. The AFL_FORKSRV_INIT_TMOUT value does not get multiplied. It overwrites the initial timeout afl-fuzz waits for the target to come up with a constant time. Setting a different time here is useful if the target has a very slow startup time, for example, when doing full-system fuzzing or emulation, but you don't want the actual runs to wait too long for timeouts.

  • Setting AFL_HANG_TMOUT allows you to specify a different timeout for deciding if a particular test case is a "hang". The default is 1 second or the value of the -t parameter, whichever is larger. Dialing the value down can be useful if you are very concerned about slow inputs, or if you don't want AFL++ to spend too much time classifying that stuff and just rapidly put all timeouts in that bin.

  • If you are Jakub, you may need AFL_I_DONT_CARE_ABOUT_MISSING_CRASHES. Others need not apply, unless they also want to disable the /proc/sys/kernel/core_pattern check.

  • If afl-fuzz encounters an incorrect fuzzing setup during a fuzzing session (not at startup), it will terminate. If you do not want this, then you can set AFL_IGNORE_PROBLEMS. If you additionally want to also ignore coverage from late loaded libraries, you can set AFL_IGNORE_PROBLEMS_COVERAGE.

  • When running with multiple afl-fuzz or with -F, setting AFL_IMPORT_FIRST causes the fuzzer to import test cases from other instances before doing anything else. This makes the "own finds" counter in the UI more accurate.

  • When running with multiple afl-fuzz or with -F, setting AFL_FINAL_SYNC will cause the fuzzer to perform a final import of test cases when terminating. This is beneficial for -M main fuzzers to ensure it has all unique test cases and hence you only need to afl-cmin this single queue.

  • Setting AFL_INPUT_LEN_MIN and AFL_INPUT_LEN_MAX are an alternative to the afl-fuzz -g/-G command line option to control the minimum/maximum of fuzzing input generated.

  • AFL_KILL_SIGNAL: Set the signal ID to be delivered to child processes on timeout. Unless you implement your own targets or instrumentation, you likely don't have to set it. By default, on timeout and on exit, SIGKILL (AFL_KILL_SIGNAL=9) will be delivered to the child.

  • AFL_FORK_SERVER_KILL_SIGNAL: Set the signal ID to be delivered to the fork server when AFL++ is terminated. Unless you implement your fork server, you likely do not have to set it. By default, SIGTERM (AFL_FORK_SERVER_KILL_SIGNAL=15) will be delivered to the fork server. If only AFL_KILL_SIGNAL is provided, AFL_FORK_SERVER_KILL_SIGNAL will be set to same value as AFL_KILL_SIGNAL to provide backward compatibility. If AFL_FORK_SERVER_KILL_SIGNAL is also set, it takes precedence.

    NOTE: Uncatchable signals, such as SIGKILL, cause child processes of the fork server to be orphaned and leaves them in a zombie state.

  • AFL_MAP_SIZE sets the size of the shared map that afl-analyze, afl-fuzz, afl-showmap, and afl-tmin create to gather instrumentation data from the target. This must be equal or larger than the size the target was compiled with.

  • Setting AFL_MAX_DET_EXTRAS will change the threshold at what number of elements in the -x dictionary and LTO autodict (combined) the probabilistic mode will kick off. In probabilistic mode, not all dictionary entries will be used all of the time for fuzzing mutations to not slow down fuzzing. The default count is 200 elements. So for the 200 + 1st element, there is a 1 in 201 chance, that one of the dictionary entries will not be used directly.

  • Setting AFL_NO_AFFINITY disables attempts to bind to a specific CPU core on Linux systems. This slows things down, but lets you run more instances of afl-fuzz than would be prudent (if you really want to).

  • AFL_NO_ARITH causes AFL++ to skip most of the deterministic arithmetics. This can be useful to speed up the fuzzing of text-based file formats.

  • Setting AFL_NO_AUTODICT will not load an LTO generated auto dictionary that is compiled into the target.

  • Setting AFL_NO_COLOR or AFL_NO_COLOUR will omit control sequences for coloring console output when configured with USE_COLOR and not ALWAYS_COLORED.

  • The CPU widget shown at the bottom of the screen is fairly simplistic and may complain of high load prematurely, especially on systems with low core counts. To avoid the alarming red color for very high CPU usages, you can set AFL_NO_CPU_RED.

  • Setting AFL_NO_FORKSRV disables the forkserver optimization, reverting to fork + execve() call for every tested input. This is useful mostly when working with unruly libraries that create threads or do other crazy things when initializing (before the instrumentation has a chance to run).

    Note that this setting inhibits some of the user-friendly diagnostics normally done when starting up the forkserver and causes a pretty significant performance drop.

  • AFL_NO_SNAPSHOT will advise afl-fuzz not to use the snapshot feature if the snapshot lkm is loaded.

  • AFL_NO_FASTRESUME will not try to read or write a fast resume file.

  • AFL_FORCE_FASTRESUME on the other hand will load the fast resume file (if it exists) even if the target binary was changed. Note that the coverage map size must be exactly the same for this to work.

  • Setting AFL_NO_UI inhibits the UI altogether and just periodically prints some basic stats. This behavior is also automatically triggered when the output from afl-fuzz is redirected to a file or to a pipe.

  • Setting AFL_NO_STARTUP_CALIBRATION will skip the initial calibration of all starting seeds, and start fuzzing at once. Use with care, this degrades the fuzzing performance!

  • Setting AFL_NO_WARN_INSTABILITY will suppress instability warnings.

  • In QEMU mode (-Q) and FRIDA mode (-O), AFL_PATH will be searched for afl-qemu-trace and afl-frida-trace.so.

  • If you are using persistent mode (you should, see instrumentation/README.persistent_mode.md), some targets keep inherent state due which a detected crash test case does not crash the target again when the test case is given. To be able to still re-trigger these crashes, you can use the AFL_PERSISTENT_RECORD variable with a value of how many previous fuzz cases to keep prior a crash. If set to e.g., 10, then the 9 previous inputs are written to out/default/crashes as RECORD:000000,cnt:000000 to RECORD:000000,cnt:000008 and RECORD:000000,cnt:000009 being the crash case. NOTE: This option needs to be enabled in config.h first!

  • Note that AFL_POST_LIBRARY is deprecated, use AFL_CUSTOM_MUTATOR_LIBRARY instead.

  • Setting AFL_PRELOAD causes AFL++ to set LD_PRELOAD for the target binary without disrupting the afl-fuzz process itself. This is useful, among other things, for bootstrapping libdislocator.so.

  • AFL_SHA1_FILENAMES causes AFL++ to generate files named by the SHA1 hash of their contents, rather than use the standard id:000000,... names. Warning: this disables any syncing to any AFL instances!

  • AFL_SHUFFLE_QUEUE randomly reorders the input queue on startup. Requested by some users for unorthodox parallelized fuzzing setups, but not advisable otherwise.

  • When developing custom instrumentation on top of afl-fuzz, you can use AFL_SKIP_BIN_CHECK to inhibit the checks for non-instrumented binaries and shell scripts; and AFL_DUMB_FORKSRV in conjunction with the -n setting to instruct afl-fuzz to still follow the fork server protocol without expecting any instrumentation data in return. Note that this also turns off auto map size detection.

  • Setting AFL_SKIP_CPUFREQ skips the check for CPU scaling policy. This is useful if you can't change the defaults (e.g., no root access to the system) and are OK with some performance loss.

  • Setting AFL_STATSD enables StatsD metrics collection. By default, AFL++ will send these metrics over UDP to 127.0.0.1:8125. The host and port are configurable with AFL_STATSD_HOST and AFL_STATSD_PORT respectively. To enable tags (banner and afl_version), you should provide AFL_STATSD_TAGS_FLAVOR that matches your StatsD server (see AFL_STATSD_TAGS_FLAVOR).

  • Setting AFL_STATSD_TAGS_FLAVOR to one of dogstatsd, influxdb, librato, or signalfx allows you to add tags to your fuzzing instances. This is especially useful when running multiple instances (-M/-S for example). Applied tags are banner and afl_version. banner corresponds to the name of the fuzzer provided through -M/-S. afl_version corresponds to the currently running AFL++ version (e.g., ++3.0c). Default (empty/non present) will add no tags to the metrics. For more information, see rpc_statsd.md.

  • AFL_SYNC_TIME allows you to specify a different minimal time (in minutes) between fuzzing instances synchronization. Default sync time is 20 minutes, note that time is halved for -M main nodes.

  • AFL_NO_SYNC disables any syncing whatsoever and takes priority on all other syncing parameters.

  • Setting AFL_TARGET_ENV causes AFL++ to set extra environment variables for the target binary. Example: AFL_TARGET_ENV="VAR1=1 VAR2='a b c'" afl-fuzz ... . This exists mostly for things like LD_LIBRARY_PATH but it would theoretically allow fuzzing of AFL++ itself (with 'target' AFL++ using some AFL_ vars that would disrupt work of 'fuzzer' AFL++). Note that when using QEMU mode, the AFL_TARGET_ENV environment variables will apply to QEMU, as well as the target binary. Therefore, in this case, you might want to use QEMU's QEMU_SET_ENV environment variable (see QEMU's documentation because the format is different from AFL_TARGET_ENV) to apply the environment variables to the target and not QEMU.

  • AFL_TESTCACHE_SIZE allows you to override the size of #define TESTCASE_CACHE in config.h. Recommended values are 50-250MB - or more if your fuzzing finds a huge amount of paths for large inputs.

  • AFL_TMPDIR is used to write the .cur_input file to if it exists, and in the normal output directory otherwise. You would use this to point to a ramdisk/tmpfs. This increases the speed by a small value but also reduces the stress on SSDs.

  • Setting AFL_TRY_AFFINITY tries to attempt binding to a specific CPU core on Linux systems, but will not terminate if that fails.

  • By default on Linux, persistent mode uses shared memory and futexes for child synchronization. This reduces the overhead of communication between afl-fuzz and the persistent target child. Setting AFL_OLD_CHILD_SYNC restores the old file descriptor based persistent synchronization behavior.

  • The following environment variables are only needed if you implemented your own forkserver or persistent mode, or if __AFL_LOOP or __AFL_INIT are in a shared library and not the main binary:

    • AFL_DEFER_FORKSRV enforces a deferred forkserver even if none was detected in the target binary
    • AFL_PERSISTENT enforces persistent mode even if none was detected in the target binary

  • If you need an early forkserver in your target because of early constructors in your target, you can set AFL_EARLY_FORKSERVER. Note that this is not a compile time option but a runtime option :-)

  • Set AFL_PIZZA_MODE to 1 to enable the April 1st stats menu, set to -1 to disable although it is 1st of April. 0 is the default and means enable on the 1st of April automatically.

  • If you need a specific interval to update fuzzer_stats file, you can set AFL_FUZZER_STATS_UPDATE_INTERVAL to the interval in seconds you'd the file to be updated. Note that will not be exact and with slow targets it can take seconds until there is a slice for the time test.

  • When using AFL_PRELOAD with a preload that disable fork() calls in the target, the forkserver becomes unable to fork. To overcome this issue, the AFL_PRELOAD_DISCRIMINATE_FORKSERVER_PARENT permits to be able to check in the preloaded library if the environment variable AFL_FORKSERVER_PARENT is set, to be able to use vanilla fork() in the forkserver, and the placeholder in the target. Here is a POC :

    // AFL_PRELOAD_DISCRIMINATE_FORKSERVER_PARENT=1 afl-fuzz ...
pid_t fork(void)
{
    if (getenv("AFL_FORKSERVER_PARENT") == NULL)
        return 0; // We are in the target
    else
        return real_fork(); // We are in the forkserver
}

  • AFL_FORKSRV_UID allows you to specify the UID that should be used when running the fork server. When setting this variable, user should ensure afl-fuzz binary has enough privileges to modify the UID (e.g. CAP_SETUID capability in Linux system).

  • AFL_FORKSRV_GID allows you to specify the GID and the supplementary group IDs that should be used when running the fork server. When setting this variable, user should ensure afl-fuzz binary has enough privileges to modify the GIDs (e.g. CAP_SETGID capability in Linux system).

  • When both AFL_FORKSRV_UID and AFL_FORKSRV_GID are set, afl-fuzz binary and the fork server no longer share any IDs. Thus, afl-fuzz binary changes the group owner of the created files to ensure that the fork server can still access them. In such case, user should ensure afl-fuzz binary has enough privileges to modify the ownership of entities (e.g. CAP_CHOWN capability in Linux system).

  • Setting AFL_FRAMESHIFT_DISABLE will disable the frameshift analysis stage. Frameshift automatically discovers size/offset fields in structured inputs and keeps them consistent as mutations insert or delete bytes. Disabling it may be useful for targets that do not consume structured binary formats, or when you want to avoid the overhead of the analysis entirely.

  • AFL_FRAMESHIFT_MAX_OVERHEAD controls the maximum fraction of total fuzzing time that frameshift analysis is allowed to consume. The value is a float between 0.0 and 1.0 (default 0.10, i.e. 10%). If the cumulative time spent in frameshift analysis exceeds this fraction of the overall run time, new analyses are skipped until the ratio drops back under the limit.

  • Normally a README.txt is written to the crashes/ directory when a first crash is found. Setting AFL_NO_CRASH_README will prevent this. Useful when counting crashes based on a file count in that directory.

6) Settings for afl-qemu-trace

The QEMU wrapper used to instrument binary-only code supports several settings:

  • Setting AFL_QEMU_CUSTOM_BIN will cause afl-fuzz to skip prepending afl-qemu-trace to your command line. Use this if you wish to use a custom afl-qemu-trace or if you need to modify the afl-qemu-trace arguments.

  • Setting AFL_COMPCOV_LEVEL enables the CompareCoverage tracing of all cmp and sub in x86 and x86_64 and memory comparison functions (e.g., strcmp, memcmp, ...) when libcompcov is preloaded using AFL_PRELOAD. More info at qemu_mode/libcompcov/README.md.

    There are two levels at the moment, AFL_COMPCOV_LEVEL=1 that instruments only comparisons with immediate values / read-only memory and AFL_COMPCOV_LEVEL=2 that instruments all the comparisons. Level 2 is more accurate but may need a larger shared memory.

  • AFL_DEBUG will print the found entry point for the binary to stderr. Use this if you are unsure if the entry point might be wrong - but use it directly, e.g., afl-qemu-trace ./program.

  • AFL_ENTRYPOINT allows you to specify a specific entry point into the binary (this can be very good for the performance!). The entry point is specified as hex address, e.g., 0x4004110. Note that the address must be the address of a basic block.

  • Setting AFL_INST_LIBS causes the translator to also instrument the code inside any dynamically linked libraries (notably including glibc).

  • You can use AFL_QEMU_INST_RANGES=0xaaaa-0xbbbb,0xcccc-0xdddd to just instrument specific memory locations, e.g. a specific library. You may also provide the filename of the library. Excluding ranges takes priority over any included ranges or AFL_INST_LIBS. See qemu_mode/README.md#partial-instrumenation).

  • You can use AFL_QEMU_EXCLUDE_RANGES=0xaaaa-0xbbbb,0xcccc-0xdddd to NOT instrument specific memory locations, e.g. a specific library. Excluding ranges takes priority over any included ranges or AFL_INST_LIBS.

  • It is possible to set AFL_INST_RATIO to skip the instrumentation on some of the basic blocks, which can be useful when dealing with very complex binaries.

  • Setting AFL_QEMU_COMPCOV enables the CompareCoverage tracing of all cmp and sub in x86 and x86_64. This is an alias of AFL_COMPCOV_LEVEL=1 when AFL_COMPCOV_LEVEL is not specified.

  • With AFL_QEMU_FORCE_DFL, you force QEMU to ignore the registered signal handlers of the target.

  • When the target is i386/x86_64, you can specify the address of the function that has to be the body of the persistent loop using AFL_QEMU_PERSISTENT_ADDR=start addr.

  • With AFL_QEMU_PERSISTENT_GPR=1, QEMU will save the original value of general purpose registers and restore them in each persistent cycle.

  • Another modality to execute the persistent loop is to specify also the AFL_QEMU_PERSISTENT_RET=end addr environment variable. With this variable assigned, instead of patching the return address, the specified instruction is transformed to a jump towards start addr.

  • With AFL_QEMU_PERSISTENT_RETADDR_OFFSET, you can specify the offset from the stack pointer in which QEMU can find the return address when start addr is hit.

  • With AFL_USE_QASAN, you can enable QEMU AddressSanitizer for dynamically linked binaries.

  • Using AFL_QEMU_IJON=test.conf allows qemu to call the ijon function, which allows aflpp to additionally gain awareness of changes in key variables of the target program, which is an excellent supplement to coverage-only feedback.(see qemu_mode/README.md for more details).

  • The underlying QEMU binary will recognize any standard "user space emulation" variables (e.g., QEMU_STACK_SIZE), but there should be no reason to touch them.

7) Settings for afl-frida-trace

The FRIDA wrapper used to instrument binary-only code supports many of the same options as afl-qemu-trace, but also has a number of additional advanced options. These are listed in brief below (see frida_mode/README.md for more details). These settings are provided for compatibility with QEMU mode, the preferred way to configure FRIDA mode is through its scripting support.

  • AFL_FRIDA_DEBUG_MAPS - See AFL_QEMU_DEBUG_MAPS
  • AFL_FRIDA_DRIVER_NO_HOOK - See AFL_QEMU_DRIVER_NO_HOOK. When using the QEMU driver to provide a main loop for a user provided LLVMFuzzerTestOneInput, this option configures the driver to read input from stdin rather than using in-memory test cases.
  • AFL_FRIDA_EXCLUDE_RANGES - See AFL_QEMU_EXCLUDE_RANGES
  • AFL_FRIDA_INST_COVERAGE_FILE - File to write DynamoRio format coverage information (e.g., to be loaded within IDA lighthouse).
  • AFL_FRIDA_INST_DEBUG_FILE - File to write raw assembly of original blocks and their instrumented counterparts during block compilation.
  • AFL_FRIDA_INST_JIT - Enable the instrumentation of Just-In-Time compiled code. Code is considered to be JIT if the executable segment is not backed by a file.
  • AFL_FRIDA_INST_NO_DYNAMIC_LOAD - Don't instrument the code loaded late at runtime. Strictly limits instrumentation to what has been included.
  • AFL_FRIDA_INST_NO_OPTIMIZE - Don't use optimized inline assembly coverage instrumentation (the default where available). Required to use AFL_FRIDA_INST_TRACE.
  • AFL_FRIDA_INST_NO_BACKPATCH - Disable backpatching. At the end of executing each block, control will return to FRIDA to identify the next block to execute.
  • AFL_FRIDA_INST_NO_PREFETCH - Disable prefetching. By default, the child will report instrumented blocks back to the parent so that it can also instrument them and they be inherited by the next child on fork, implies AFL_FRIDA_INST_NO_PREFETCH_BACKPATCH.
  • AFL_FRIDA_INST_NO_PREFETCH_BACKPATCH - Disable prefetching of stalker backpatching information. By default, the child will report applied backpatches to the parent so that they can be applied and then be inherited by the next child on fork.
  • AFL_FRIDA_INST_RANGES - See AFL_QEMU_INST_RANGES
  • AFL_FRIDA_INST_SEED - Sets the initial seed for the hash function used to generate block (and hence edge) IDs. Setting this to a constant value may be useful for debugging purposes, e.g., investigating unstable edges.
  • AFL_FRIDA_INST_TRACE - Log to stdout the address of executed blocks, implies AFL_FRIDA_INST_NO_OPTIMIZE.
  • AFL_FRIDA_INST_TRACE_UNIQUE - As per AFL_FRIDA_INST_TRACE, but each edge is logged only once, requires AFL_FRIDA_INST_NO_OPTIMIZE.
  • AFL_FRIDA_INST_UNSTABLE_COVERAGE_FILE - File to write DynamoRio format coverage information for unstable edges (e.g., to be loaded within IDA lighthouse).
  • AFL_FRIDA_JS_SCRIPT - Set the script to be loaded by the FRIDA scripting engine. See frida_mode/Scripting.md for details.
  • AFL_FRIDA_OUTPUT_STDOUT - Redirect the standard output of the target application to the named file (supersedes the setting of AFL_DEBUG_CHILD)
  • AFL_FRIDA_OUTPUT_STDERR - Redirect the standard error of the target application to the named file (supersedes the setting of AFL_DEBUG_CHILD)
  • AFL_FRIDA_PERSISTENT_ADDR - See AFL_QEMU_PERSISTENT_ADDR
  • AFL_FRIDA_PERSISTENT_CNT - See AFL_QEMU_PERSISTENT_CNT
  • AFL_FRIDA_PERSISTENT_DEBUG - Insert a Breakpoint into the instrumented code at AFL_FRIDA_PERSISTENT_HOOK and AFL_FRIDA_PERSISTENT_RET to allow the user to detect issues in the persistent loop using a debugger.
  • AFL_FRIDA_PERSISTENT_HOOK - See AFL_QEMU_PERSISTENT_HOOK
  • AFL_FRIDA_PERSISTENT_RET - See AFL_QEMU_PERSISTENT_RET
  • AFL_FRIDA_SECCOMP_FILE - Write a log of any syscalls made by the target to the specified file.
  • AFL_FRIDA_STALKER_ADJACENT_BLOCKS - Configure the number of adjacent blocks to fetch when generating instrumented code. By fetching blocks in the same order they appear in the original program, rather than the order of execution should help reduce locality and adjacency. This includes allowing us to vector between adjacent blocks using a NOP slide rather than an immediate branch.
  • AFL_FRIDA_STALKER_IC_ENTRIES - Configure the number of inline cache entries stored along-side branch instructions which provide a cache to avoid having to call back into FRIDA to find the next block. Default is 32.
  • AFL_FRIDA_STATS_FILE - Write statistics information about the code being instrumented to the given file name. The statistics are written only for the child process when new block is instrumented (when the AFL_FRIDA_STATS_INTERVAL has expired). Note that just because a new path is found does not mean a new block needs to be compiled. It could be that the existing blocks instrumented have been executed in a different order.
  • AFL_FRIDA_STATS_INTERVAL - The maximum frequency to output statistics information. Stats will be written whenever they are updated if the given interval has elapsed since last time they were written.
  • AFL_FRIDA_TRACEABLE - Set the child process to be traceable by any process to aid debugging and overcome the restrictions imposed by YAMA. Supported on Linux only. Permits a non-root user to use gcore or similar to collect a core dump of the instrumented target. Note that in order to capture the core dump you must set a sufficient timeout (using -t) to avoid afl-fuzz killing the process whilst it is being dumped.

8) Settings for afl-cmin

The corpus minimization script offers very little customization:

  • AFL_ALLOW_TMP permits this and some other scripts to run in /tmp. This is a modest security risk on multi-user systems with rogue users, but should be safe on dedicated fuzzing boxes.

  • AFL_KEEP_TRACES makes the tool keep traces and other metadata used for minimization and normally deleted at exit. The files can be found in the <out_dir>/.traces/ directory.

  • Setting AFL_PATH offers a way to specify the location of afl-showmap and afl-qemu-trace (the latter only in -Q mode).

  • AFL_PRINT_FILENAMES prints each filename to stdout, as it gets processed. This can help when embedding afl-cmin or afl-showmap in other scripts.

9) Settings for afl-tmin

Virtually nothing to play with. Well, in QEMU mode (-Q), AFL_PATH will be searched for afl-qemu-trace. In addition to this, TMPDIR may be used if a temporary file can't be created in the current working directory.

You can specify AFL_TMIN_EXACT if you want afl-tmin to require execution paths to match when minimizing crashes. This will make minimization less useful, but may prevent the tool from "jumping" from one crashing condition to another in very buggy software. You probably want to combine it with the -e flag.

10) Settings for afl-analyze

You can set AFL_ANALYZE_HEX to get file offsets printed as hexadecimal instead of decimal.

11) Settings for libdislocator

The library honors these environment variables:

  • AFL_ALIGNED_ALLOC=1 will force the alignment of the allocation size to max_align_t to be compliant with the C standard.

  • AFL_LD_HARD_FAIL alters the behavior by calling abort() on excessive allocations, thus causing what AFL++ would perceive as a crash. Useful for programs that are supposed to maintain a specific memory footprint.

  • AFL_LD_LIMIT_MB caps the size of the maximum heap usage permitted by the library, in megabytes. The default value is 1 GB. Once this is exceeded, allocations will return NULL.

  • AFL_LD_NO_CALLOC_OVER inhibits abort() on calloc() overflows. Most of the common allocators check for that internally and return NULL, so it's a security risk only in more exotic setups.

  • AFL_LD_VERBOSE causes the library to output some diagnostic messages that may be useful for pinpointing the cause of any observed issues.

12) Settings for libtokencap

This library accepts AFL_TOKEN_FILE to indicate the location to which the discovered tokens should be written.

13) Third-party variables set by afl-fuzz & other tools

Several variables are not directly interpreted by afl-fuzz, but are set to optimal values if not already present in the environment:

  • By default, ASAN_OPTIONS are set to (among others):

    abort_on_error=1
detect_leaks=0
malloc_context_size=0
symbolize=0
allocator_may_return_null=1

    If you want to set your own options, be sure to include abort_on_error=1 - otherwise, the fuzzer will not be able to detect crashes in the tested app. Similarly, include symbolize=0, since without it, AFL++ may have difficulty telling crashes and hangs apart.

  • Similarly, the default LSAN_OPTIONS are set to:

    exit_code=23
fast_unwind_on_malloc=0
symbolize=0
print_suppressions=0

    Be sure to include the first ones for LSAN and MSAN when customizing anything, since some MSAN and LSAN versions don't call abort() on error, and we need a way to detect faults.

  • In the same vein, by default, MSAN_OPTIONS are set to:

    exit_code=86 (required for legacy reasons)
abort_on_error=1
symbolize=0
msan_track_origins=0
allocator_may_return_null=1

  • By default, LD_BIND_NOW is set to speed up fuzzing by forcing the linker to do all the work before the fork server kicks in. You can override this by setting LD_BIND_LAZY beforehand, but it is almost certainly pointless.