Add standalone driver for fuzz targets

Useful for cases that don't need to use a fuzzing engine, but just run
the fuzz targets over some set of inputs, like for regression/CI tests.

Also added a POP3 fuzzer dictionary, seed corpus, and README with
examples.

CMakeLists.txt

@@ -481,6 +481,9 @@ message(
     "\n       debugging:  ${USE_PERFTOOLS_DEBUG}"
     "\njemalloc:          ${ENABLE_JEMALLOC}"
     "\n"
+    "\nFuzz Targets:      ${ZEEK_ENABLE_FUZZING}"
+    "\nFuzz Engine:       ${ZEEK_FUZZING_ENGINE}"
+    "\n"
     "\n================================================================\n"
 )
 

src/fuzzers/CMakeLists.txt

@@ -5,13 +5,17 @@ if ( NOT ZEEK_ENABLE_FUZZING )
     return()
 endif ()
 
+if ( NOT DEFINED ZEEK_FUZZING_ENGINE AND DEFINED ENV{LIB_FUZZING_ENGINE} )
+    set(ZEEK_FUZZING_ENGINE $ENV{LIB_FUZZING_ENGINE} CACHE INTERNAL "" FORCE)
+endif ()
+
 macro(ADD_FUZZ_TARGET _name)
-    set(_fuzz_target zeek_fuzzer_${_name})
+    set(_fuzz_target zeek-${_name}-fuzzer)
-    set(_fuzz_source ${_name}.cc)
+    set(_fuzz_source ${_name}-fuzzer.cc)
 
     add_executable(${_fuzz_target} ${_fuzz_source} ${ARGN}
                    $<TARGET_OBJECTS:zeek_objs>
-                   $<TARGET_OBJECTS:zeek_fuzzer_objs>
+                   $<TARGET_OBJECTS:zeek_fuzzer_common>
                    ${zeek_HEADERS}
                    ${bro_SUBDIR_LIBS}
                    ${bro_PLUGIN_LIBS})
@@ -19,12 +23,17 @@ macro(ADD_FUZZ_TARGET _name)
     target_link_libraries(${_fuzz_target} ${zeekdeps}
                           ${CMAKE_THREAD_LIBS_INIT} ${CMAKE_DL_LIBS})
 
-    # TODO: Use LIB_FUZZING_ENGINE env. var. if it exists
+    if ( DEFINED ZEEK_FUZZING_ENGINE )
-    set_target_properties(${_fuzz_target} PROPERTIES LINK_FLAGS
+        set_target_properties(${_fuzz_target} PROPERTIES LINK_FLAGS
-                          "-fsanitize=fuzzer")
+                              ${ZEEK_FUZZING_ENGINE})
+    else ()
+        target_link_libraries(${_fuzz_target}
+                              $<TARGET_OBJECTS:zeek_fuzzer_standalone>)
+    endif ()
 endmacro ()
 
 include_directories(BEFORE ${CMAKE_CURRENT_SOURCE_DIR})
-add_library(zeek_fuzzer_objs OBJECT FuzzBuffer.cc)
+add_library(zeek_fuzzer_common OBJECT FuzzBuffer.cc)
+add_library(zeek_fuzzer_standalone OBJECT standalone-driver.cc)
 
 add_fuzz_target(pop3)

src/fuzzers/README

@@ -0,0 +1,79 @@
+Fuzz Testing
+============
+
+This directory contains fuzzing targets for various Zeek components.  The
+primary way to use these directly would be with a fuzzing engine such as
+libFuzzer: https://llvm.org/docs/LibFuzzer.html
+
+Example Build: Initial Fuzzing and Seed Corpus
+----------------------------------------------
+
+First configure and build for fuzzing (with libFuzzer) and code coverage::
+
+    $ LIB_FUZZING_ENGINE="-fsanitize=fuzzer" CC=clang CXX=clang++ \
+      CFLAGS="-fprofile-instr-generate -fcoverage-mapping" \
+      CXXFLAGS="-fprofile-instr-generate -fcoverage-mapping" \
+      ./configure --build-type=RelWithDebInfo --build-dir=./build-fuzz-cov \
+      --sanitizers=fuzzer-no-link --enable-fuzzing
+
+    $ cd build-fuzz-cov && make -j $(nproc)
+
+Now start fuzzing to generate an initial corpus (this uses the POP3 fuzzer as
+an example)::
+
+    $ mkdir corpus && ./src/fuzzers/zeek-pop3-fuzzer corpus \
+      -dict=../src/fuzzers/pop3.dict -max_total_time=300 -fork=$(($(nproc) - 1))
+
+You can set options, like the runtime and parallelism level, to taste.  For
+other fuzz targets, you'd also want to use a different dictionary or omit
+entirely.
+
+To minimize the size of the corpus::
+
+    $ mkdir min-corpus && ./src/fuzzers/zeek-pop3-fuzzer -merge=1 min-corpus corpus
+
+To check the code coverage of the corpus::
+
+    $ ./src/fuzzers/zeek-pop3-fuzzer min-corpus/*
+
+    $ llvm-profdata merge -sparse default.profraw -o zeek.profdata && \
+      llvm-cov report ./src/fuzzers/zeek-pop3-fuzzer -instr-profile=zeek.profdata \
+        ../src/analyzer/protocol/pop3/
+
+If the code coverage isn't satisfying, there may be something wrong with
+the fuzzer, it may need a better dictionary, or it may need to fuzz for longer.
+
+The corpus can be added to revision control for use in regression testing and
+as seed for OSS-Fuzz (check first that the zip file is a size that's sane to
+commit)::
+
+    zip -j ../src/fuzzers/pop3-corpus.zip min-corpus/*
+
+Example Build: Run Standalone Fuzz Targets
+------------------------------------------
+
+Fuzz targets can still be run without a fuzzing engine driving them.  In
+standalone mode, they'll process all input files provided as arguments
+(e.g. useful for regression testing).
+
+First configure and build::
+
+    $ CC=clang CXX=clang++ \
+      ./configure --build-type=debug --build-dir=./build-fuzz-check \
+      --sanitizers=address,fuzzer-no-link --enable-fuzzing
+
+    $ cd build-fuzz-cov && make -j $(nproc)
+
+Get a set of inputs to process (we're using the POP3 fuzzer/corpus as example)::
+
+    $ mkdir corpus && ( cd corpus && unzip ../../src/fuzzers/pop3-corpus.zip )
+
+Now run the standalone fuzzer on the input corpus::
+
+    $ ./src/fuzzers/zeek-pop3-fuzzer corpus/*
+
+Note that you can also configure this build for coverage reports to verify the
+code coverage (see the CFLAGS/CXXFLAGS from the first "Initial Fuzzing"
+section).  There's also the following ASan option which may need to be used::
+
+    $ export ASAN_OPTIONS=detect_odr_violation=0

src/fuzzers/fuzzer-setup.h

@@ -16,6 +16,7 @@ extern "C" int LLVMFuzzerInitialize(int* argc, char*** argv)
 	options.script_options_to_set.emplace_back("Site::local_nets={10.0.0.0/8}");
 	options.script_options_to_set.emplace_back("Log::default_writer=Log::WRITER_NONE");
 	options.deterministic_mode = true;
+	options.ignore_checksums = true;
 	options.abort_on_scripting_errors = true;
 
 	if ( zeek::setup(*argc, *argv, &options).code )

src/fuzzers/pop3-fuzzer.cc

@@ -9,7 +9,7 @@
 #include "analyzer/protocol/tcp/TCP.h"
 
 #include "FuzzBuffer.h"
-#include "fuzz-setup.h"
+#include "fuzzer-setup.h"
 
 static constexpr auto ZEEK_FUZZ_ANALYZER = "pop3";
 

src/fuzzers/pop3.dict

@@ -0,0 +1,21 @@
+"\x01PKT"
+"OK"
+"ERR"
+"USER"
+"PASS"
+"APOP"
+"AUTH"
+"STAT"
+"LIST"
+"RETR"
+"DELE"
+"RSET"
+"NOOP"
+"LAST"
+"QUIT"
+"TOP"
+"CAPA"
+"UIDL"
+"STLS"
+"XSENDER"
+"END"

src/fuzzers/standalone-driver.cc

@@ -0,0 +1,47 @@
+#include <cstdio>
+#include <cstdint>
+#include <cassert>
+#include <memory>
+#include <chrono>
+
+extern "C" int LLVMFuzzerTestOneInput(const uint8_t* data, size_t size);
+extern "C" int LLVMFuzzerInitialize(int* argc, char*** argv);
+
+int main(int argc, char** argv)
+	{
+	using namespace std::chrono;
+	auto agg_start = high_resolution_clock::now();
+	auto num_inputs = argc - 1;
+	printf("Standalone fuzzer processing %d inputs\n", num_inputs);
+
+	LLVMFuzzerInitialize(&argc, &argv);
+
+	for ( auto i = 0; i < num_inputs; ++i )
+		{
+		auto input_file_name = argv[i + 1];
+		printf("  %s:", input_file_name);
+
+		auto f = fopen(input_file_name, "r");
+		assert(f);
+
+		fseek(f, 0, SEEK_END);
+		auto input_length = ftell(f);
+		fseek(f, 0, SEEK_SET);
+
+		auto input_buffer = std::make_unique<uint8_t[]>(input_length);
+		auto bytes_read = fread(input_buffer.get(), 1, input_length, f);
+		assert(bytes_read == input_length);
+
+		auto start = high_resolution_clock::now();
+		LLVMFuzzerTestOneInput(input_buffer.get(), input_length);
+		auto stop = high_resolution_clock::now();
+		auto dt = duration<double>(stop - start).count();
+
+		printf(" %6zu bytes, %f seconds\n", input_length, dt);
+		fclose(f);
+		}
+
+	auto agg_stop = high_resolution_clock::now();
+	auto agg_dt = duration<double>(agg_stop - agg_start).count();
+	printf("Processed %d inputs in %fs\n", num_inputs, agg_dt);
+}