Update fuzzing tools and build scripts

2022-01-16 12:59:23 +01:00 · 2022-01-16 12:59:23 +01:00 · 97f8ae52c1
parent 21d2d51911
commit 97f8ae52c1
5 changed files with 108 additions and 40 deletions
--- a/configure.ac
+++ b/configure.ac
@ -73,6 +73,14 @@ AM_PATH_PYTHON([2.7])

 PKG_PROG_PKG_CONFIG

+AC_ARG_ENABLE([fuzzing],
+  [AS_HELP_STRING([--enable-fuzzing], [Turn on fuzzing build (for developers)])],
+  [enable_fuzzing=yes;
+   AC_SUBST([LIB_FUZZING_ENGINE])
+   AC_DEFINE([FUZZING], 1, [Define to 1 if this is a fuzzing build])
+  ], [enable_fuzzing=no; LIB_FUZZING_ENGINE=""])
+AM_CONDITIONAL([FUZZING], [test "$enable_fuzzing" = "yes"])
+
 AC_ARG_ENABLE([cfi],
  [AS_HELP_STRING([--enable-cfi], [Turn on clang's Control Flow Integrity (CFI)])],
  [
@ -90,6 +98,7 @@ AC_ARG_ENABLE([ubsan],
  [AS_HELP_STRING([--enable-ubsan], [Turn on Undefined Behavior Sanitizer (UBSan)])],
  [
    if test "$enable_ubsan" = yes; then
+      # Set basic UBSAN compiler flags. Add your own to CFLAGS.
      CFLAGS=$CFLAGS" -fsanitize=undefined -fno-sanitize-recover=undefined"
    fi
  ], [ enable_ubsan=no ])
@ -98,6 +107,7 @@ AC_ARG_ENABLE([asan],
  [AS_HELP_STRING([--enable-asan], [Turn on Address Sanitizer (ASan)])],
  [
    if test "$enable_asan" = yes; then
+      # Set basic ASAN compiler flags. Add your own to CFLAGS.
      CFLAGS=$CFLAGS" -fsanitize=address -fno-omit-frame-pointer"
      AX_CHECK_COMPILE_FLAG([-fsanitize-address-use-after-scope], [CFLAGS="$CFLAGS -fsanitize-address-use-after-scope"])
    fi
@ -409,4 +419,5 @@ AC_MSG_NOTICE([Summary of build options:
  Docs:              $enable_gtk_doc
  Man pages:         $enable_man
  Tests:             ${TESTS_INFO}
+  Fuzzing build:     $enable_fuzzing, $LIB_FUZZING_ENGINE
 ])
--- a/fuzz/Makefile.am
+++ b/fuzz/Makefile.am
@ -1,15 +1,18 @@
 AM_CFLAGS = $(WERROR_CFLAGS) $(WARN_CFLAGS)
-AM_CPPFLAGS = -I$(top_srcdir)/include -I$(srcdir) -DSRCDIR=\"$(abs_srcdir)\" -DTEST_RUN
+AM_CPPFLAGS = -I$(top_srcdir)/include -I$(srcdir) -DSRCDIR=\"$(abs_srcdir)\"
 #AM_LDFLAGS = -static
-AM_LDFLAGS = -no-install
 LDADD = ../src/libpsl.la $(LIBICU_LIBS) $(LIBIDN_LIBS) $(LIBIDN2_LIBS)

+if !FUZZING
+  MAIN = main.c fuzzer.h
+endif
+
 if WITH_LIBICU
  XLIBS = $(LIBICU_LIBS)
  XTYPE = _icu
-  libpsl_icu_fuzzer_SOURCES = libpsl_fuzzer.c main.c fuzzer.h
-  libpsl_icu_load_fuzzer_SOURCES = libpsl_load_fuzzer.c main.c fuzzer.h
-  libpsl_icu_load_dafsa_fuzzer_SOURCES = libpsl_load_dafsa_fuzzer.c main.c fuzzer.h
+  libpsl_icu_fuzzer_SOURCES = libpsl_fuzzer.c $(MAIN)
+  libpsl_icu_load_fuzzer_SOURCES = libpsl_load_fuzzer.c $(MAIN)
+  libpsl_icu_load_dafsa_fuzzer_SOURCES = libpsl_load_dafsa_fuzzer.c $(MAIN)
  PSL_TESTS = \
   libpsl_icu_fuzzer$(EXEEXT) \
   libpsl_icu_load_fuzzer$(EXEEXT) \
@ -18,9 +21,9 @@ else
 if WITH_LIBIDN2
  XLIBS = -lidn2 -lunistring
  XTYPE = _idn2
-  libpsl_idn2_fuzzer_SOURCES = libpsl_fuzzer.c main.c fuzzer.h
-  libpsl_idn2_load_fuzzer_SOURCES = libpsl_load_fuzzer.c main.c fuzzer.h
-  libpsl_idn2_load_dafsa_fuzzer_SOURCES = libpsl_load_dafsa_fuzzer.c main.c fuzzer.h
+  libpsl_idn2_fuzzer_SOURCES = libpsl_fuzzer.c $(MAIN)
+  libpsl_idn2_load_fuzzer_SOURCES = libpsl_load_fuzzer.c $(MAIN)
+  libpsl_idn2_load_dafsa_fuzzer_SOURCES = libpsl_load_dafsa_fuzzer.c $(MAIN)
  PSL_TESTS = \
   libpsl_idn2_fuzzer$(EXEEXT) \
   libpsl_idn2_load_fuzzer$(EXEEXT) \
@ -29,9 +32,9 @@ else
 if WITH_LIBIDN
  XLIBS = -lidn -lunistring
  XTYPE = _idn
-  libpsl_idn_fuzzer_SOURCES = libpsl_fuzzer.c main.c fuzzer.h
-  libpsl_idn_load_fuzzer_SOURCES = libpsl_load_fuzzer.c main.c fuzzer.h
-  libpsl_idn_load_dafsa_fuzzer_SOURCES = libpsl_load_dafsa_fuzzer.c main.c fuzzer.h
+  libpsl_idn_fuzzer_SOURCES = libpsl_fuzzer.c $(MAIN)
+  libpsl_idn_load_fuzzer_SOURCES = libpsl_load_fuzzer.c $(MAIN)
+  libpsl_idn_load_dafsa_fuzzer_SOURCES = libpsl_load_dafsa_fuzzer.c $(MAIN)
  PSL_TESTS = \
   libpsl_idn_fuzzer$(EXEEXT) \
   libpsl_idn_load_fuzzer$(EXEEXT) \
@ -39,9 +42,9 @@ if WITH_LIBIDN
 else
  XLIBS =
  XTYPE =
-  libpsl_fuzzer_SOURCES = libpsl_fuzzer.c main.c fuzzer.h
-  libpsl_load_fuzzer_SOURCES = libpsl_load_fuzzer.c main.c fuzzer.h
-  libpsl_load_dafsa_fuzzer_SOURCES = libpsl_load_dafsa_fuzzer.c main.c fuzzer.h
+  libpsl_fuzzer_SOURCES = libpsl_fuzzer.c $(MAIN)
+  libpsl_load_fuzzer_SOURCES = libpsl_load_fuzzer.c $(MAIN)
+  libpsl_load_dafsa_fuzzer_SOURCES = libpsl_load_dafsa_fuzzer.c $(MAIN)
  PSL_TESTS = \
   libpsl_fuzzer$(EXEEXT) \
   libpsl_load_fuzzer$(EXEEXT) \
@ -50,7 +53,16 @@ endif
 endif
 endif

+if FUZZING
+  bin_PROGRAMS = $(PSL_TESTS)
+  LDADD += $(LIB_FUZZING_ENGINE)
+  AM_LDFLAGS = -no-install
+else
+  AM_CPPFLAGS += -DTEST_RUN
+  AM_TESTS_ENVIRONMENT = export VALGRIND_TESTS"=@VALGRIND_TESTS@";
+  TESTS = $(PSL_TESTS)
  check_PROGRAMS = $(PSL_TESTS)
+endif

 EXTRA_DIST = meson.build

@ -61,7 +73,6 @@ dist-hook:
 	find . -name '*.repro' -exec cp -vr '{}' $(distdir) ';'

 TESTS_ENVIRONMENT = TESTS_VALGRIND="@VALGRIND_ENVIRONMENT@"
-TESTS = $(PSL_TESTS)

 clean-local:
 	rm -rf *.gc?? *.log lcov coverage.info *_fuzzer *.o
@ -69,7 +80,7 @@ clean-local:
 fuzz-coverage: $(PSL_TESTS)
 	find . -name '*_fuzzer' -exec ./coverage.sh '{}' ';'

-CXX ?= clang-5.0
+CXX ?= clang
 CXXFLAGS ?= $(CFLAGS)

 oss-fuzz:
--- a/fuzz/README.md
+++ b/fuzz/README.md
@ -12,24 +12,58 @@ regression testing with top dir 'make check' or 'make check-valgrind'.
 The ./configure runs below are for libidn2.
 To test libicu replace 'libidn2' with 'libicu', to test with
 libidn replace 'libidn2' by 'libidn'.
+To test without IDNA libraries replace `--enable-runtime=...` with `--disable-runtime`
+and replace `--enable-builtin=...` with `--disable-builtin`.


 # Running a fuzzer using clang

 Use the following commands on top dir:
 ```
-export CC=clang-5.0
-export CFLAGS="-O1 -fno-omit-frame-pointer -gline-tables-only -DFUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION -fsanitize=address -fsanitize-address-use-after-scope -fsanitize-coverage=trace-pc-guard,trace-cmp"
-./configure --enable-static --disable-gtk-doc --enable-runtime=libidn2 --enable-builtin=libidn2
-make clean
+export CC=clang
+export LIB_FUZZING_ENGINE="-lFuzzer -lstdc++"
+export UBSAN_OPTIONS=print_stacktrace=1
+export ASAN_SYMBOLIZER_PATH=/usr/bin/llvm-symbolizer
+
+# set flags for clang asan
+CFLAGS="-O1 -fno-omit-frame-pointer -gline-tables-only -DFUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION -fsanitize=address,leak,nonnull-attribute -fsanitize-address-use-after-scope -fsanitize-coverage=trace-pc-guard,trace-cmp"
+
+# set flags for clang ubsan
+CFLAGS="$CFLAGS -fsanitize=bool,array-bounds,float-divide-by-zero,function,integer-divide-by-zero,return,shift,signed-integer-overflow,vla-bound,vptr,undefined,alignment,null,enum,integer,builtin,float-divide-by-zero,function,object-size,returns-nonnull-attribute,unsigned-integer-overflow,unreachable -fsanitize=fuzzer-no-link"
+
+# unsigned-integer-overflow is not UB, so recover when we see it.
+CFLAGS="$CFLAGS -fno-sanitize-recover=all -fsanitize-recover=unsigned-integer-overflow"
+
+export CFLAGS
+
+./configure --enable-static --disable-gtk-doc --enable-fuzzing --enable-runtime=libidn2 --enable-builtin=libidn2
 make -j$(nproc)
 cd fuzz

-# build and run libpsl_fuzzer
+# run libpsl_fuzzer
 ./run-clang.sh libpsl_fuzzer
 ```


+If you see a crash, then a crash corpora is written that can be used for further
+investigation. E.g.
+```
+==2410==ERROR: AddressSanitizer: heap-use-after-free on address 0x602000004e90 at pc 0x00000049cf9c bp 0x7fffb5543f70 sp 0x7ff
+fb5543720
+...
+Test unit written to ./crash-adc83b19e793491b1c6ea0fd8b46cd9f32e592fc
+```
+
+To reproduce the crash:
+```
+./libpsl_fuzzer < ./crash-adc83b19e793491b1c6ea0fd8b46cd9f32e592fc
+```
+
+You can also copy/move that file into libpsl_fuzzer.repro/
+and re-build the project without fuzzing for a valgrind run, if you like that better.
+Just a `./configure` and a `make check-valgrind` should reproduce it.
+
+
 # Running a fuzzer using AFL

 Use the following commands on top dir:
@ -67,7 +101,7 @@ directory.

 # Clang CFI instrumentation
 ```
-CC=clang-5.0 CFLAGS="-B/usr/bin/gold -O0 -fsanitize=cfi -flto -fvisibility=default -fno-sanitize-trap=all" ./configure
+CC=clang CFLAGS="-B/usr/bin/gold -O0 -fsanitize=cfi -flto -fvisibility=default -fno-sanitize-trap=all" ./configure
 make clean
 make
 make check
--- a/fuzz/get_ossfuzz_corpora
+++ b/fuzz/get_ossfuzz_corpora
@ -1,15 +1,8 @@
 #!/bin/sh -eu

-# First step: In the top directory execute
-# export CC=clang-5.0
-# export CFLAGS="-O1 -fno-omit-frame-pointer -gline-tables-only -DFUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION -fsanitize=address -fsanitize-address-use-after-scope -fsanitize-coverage=trace-pc-guard,trace-cmp"
-# ./configure --enable-static --disable-gtk-doc --enable-runtime=libidn2 --enable-builtin=libidn2
-# make clean
-# make -j$(nproc) LIB_FUZZING_ENGINE="-lFuzzer"
-# cd fuzz
-# make -j$(nproc) check
+# As a first step see README.md and follow the steps under "Running a fuzzer using clang".

-# 
+# You might need 'gsutil' to download new corpora from the Google cloud:
 # Read the docs at https://github.com/google/oss-fuzz/blob/master/docs/corpora.md
 # then install 'google-cloud-sdk' and execute 'gcloud init'.
 # Now 'gsutil' should be ready to use.
@ -20,13 +13,20 @@ if test -z "$1"; then
  exit 1
 fi

+if ! grep -q FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION Makefile; then
+  echo "The fuzzers haven't been built for fuzzing (maybe for regression testing !?)"
+  echo "Please built regarding README.md and try again."
+  exit 1
+fi
+
 fuzzer=$1
 project=$(echo $1 | cut -d'_' -f1)

 # sync/copy the OSS-Fuzz corpora into the .new directory
-mkdir -p ${fuzzer}.new
-cp -p ${fuzzer}.in/* ${fuzzer}.new
-gsutil -m rsync gs://${project}-corpus.clusterfuzz-external.appspot.com/libFuzzer/${fuzzer} ${fuzzer}.new
+mkdir -p ${fuzzer}.in ${fuzzer}.new
+cp -fp ${fuzzer}.in/* ${fuzzer}.new 2>/dev/null || true
+gsutil cp $(gsutil ls gs://${project}-backup.clusterfuzz-external.appspot.com/corpus/libFuzzer/${fuzzer}|tail -n 1) ${fuzzer}.new/
+(cd ${fuzzer}.new && unzip -o *.zip && mv *.zip ..)

 # create fuzzer target
 BUILD_ONLY=1 ./run-clang.sh ${fuzzer}
@ -35,8 +35,9 @@ BUILD_ONLY=1 ./run-clang.sh ${fuzzer}
 ./${fuzzer} -merge=1 ${fuzzer}.in ${fuzzer}.new

 # now clear .new dir and put all corpora there
-rm -f ${fuzzer}.new/*
-mv ${fuzzer}.in/* ${fuzzer}.new
+rm -rf ${fuzzer}.new
+mv ${fuzzer}.in ${fuzzer}.new
+mkdir ${fuzzer}.in

 # now merge again (optimizes number of corpora)
 ./${fuzzer} -merge=1 ${fuzzer}.in ${fuzzer}.new
--- a/fuzz/run-clang.sh
+++ b/fuzz/run-clang.sh
@ -35,10 +35,21 @@ if test -z "$1"; then
 	exit 1
 fi

+if ! grep -q FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION Makefile; then
+  echo "The fuzzers haven't been built for fuzzing (maybe for regression testing !?)"
+  echo "Please built regarding README.md and try again."
+  exit 1
+fi
+
+# you'll need ~2GB free memory per worker !
 fuzzer=$1
-workers=$(($(nproc) - 1))
+workers=$(($(nproc) - 0))
 jobs=$workers

+if test -n "$BUILD_ONLY"; then
+  exit 0
+fi
+
 case $fuzzer in
  libpsl_idn2_*)
    cfile="libpsl_"$(echo $fuzzer|cut -d'_' -f3-)".c"
@ -54,11 +65,11 @@ case $fuzzer in
    XLIBS=
 esac

-clang-5.0 \
+clang \
 $CFLAGS -Og -g -I../include -I.. \
 ${cfile} -o ${fuzzer} \
 -Wl,-Bstatic ../src/.libs/libpsl.a -lFuzzer \
- -Wl,-Bdynamic $XLIBS -lclang-5.0 -lpthread -lm -lstdc++
+ -Wl,-Bdynamic $XLIBS -lpthread -lm -lstdc++

 if test -n "$BUILD_ONLY"; then
  exit 0