xref: /openbsd-src/gnu/llvm/compiler-rt/lib/sanitizer_common/symbolizer/scripts/build_symbolizer.sh (revision 810390e339a5425391477d5d41c78d7cab2424ac)
1*810390e3Srobert#!/usr/bin/env bash
23cab2bb3Spatrick#
33cab2bb3Spatrick# Run as: CLANG=bin/clang ZLIB_SRC=src/zlib \
43cab2bb3Spatrick#             build_symbolizer.sh runtime_build/lib/clang/4.0.0/lib/linux/
53cab2bb3Spatrick# zlib can be downloaded from http://www.zlib.net.
63cab2bb3Spatrick#
73cab2bb3Spatrick# Script compiles self-contained object file with symbolization code and injects
83cab2bb3Spatrick# it into the given set of runtime libraries. Script updates only libraries
93cab2bb3Spatrick# which has unresolved __sanitizer_symbolize_* symbols and matches architecture.
103cab2bb3Spatrick# Object file is be compiled from LLVM sources with dependencies like libc++ and
113cab2bb3Spatrick# zlib. Then it internalizes symbols in the file, so that it can be linked
123cab2bb3Spatrick# into arbitrary programs, avoiding conflicts with the program own symbols and
133cab2bb3Spatrick# avoiding dependencies on any program symbols. The only acceptable dependencies
143cab2bb3Spatrick# are libc and __sanitizer::internal_* from sanitizer runtime.
153cab2bb3Spatrick#
163cab2bb3Spatrick# Symbols exported by the object file will be used by Sanitizer runtime
173cab2bb3Spatrick# libraries to symbolize code/data in-process.
183cab2bb3Spatrick#
193cab2bb3Spatrick# The script will modify the output directory which is given as the first
203cab2bb3Spatrick# argument to the script.
213cab2bb3Spatrick#
223cab2bb3Spatrick# FIXME: We should really be using a simpler approach to building this object
233cab2bb3Spatrick# file, and it should be available as a regular cmake rule. Conceptually, we
243cab2bb3Spatrick# want to be doing "ld -r" followed by "objcopy -G" to create a relocatable
253cab2bb3Spatrick# object file with only our entry points exposed. However, this does not work at
263cab2bb3Spatrick# present, see PR30750.
273cab2bb3Spatrick
283cab2bb3Spatrickset -x
293cab2bb3Spatrickset -e
303cab2bb3Spatrickset -u
313cab2bb3Spatrick
323cab2bb3SpatrickSCRIPT_DIR=$(cd "$(dirname "$0")" && pwd)
333cab2bb3SpatrickSRC_DIR=$(readlink -f $SCRIPT_DIR/..)
343cab2bb3SpatrickTARGE_DIR=$(readlink -f $1)
353cab2bb3SpatrickCOMPILER_RT_SRC=$(readlink -f ${SCRIPT_DIR}/../../../..)
363cab2bb3SpatrickLLVM_SRC=${LLVM_SRC:-${COMPILER_RT_SRC}/../llvm}
373cab2bb3SpatrickLLVM_SRC=$(readlink -f $LLVM_SRC)
383cab2bb3Spatrick
393cab2bb3Spatrickif [[ "$ZLIB_SRC" == ""  ||
403cab2bb3Spatrick      ! -x "${ZLIB_SRC}/configure" ||
413cab2bb3Spatrick      ! -f "${ZLIB_SRC}/zlib.h" ]]; then
423cab2bb3Spatrick  echo "Missing or incomplete ZLIB_SRC"
433cab2bb3Spatrick  exit 1
443cab2bb3Spatrickfi
453cab2bb3SpatrickZLIB_SRC=$(readlink -f $ZLIB_SRC)
463cab2bb3Spatrick
473cab2bb3SpatrickCLANG="${CLANG:-`which clang`}"
483cab2bb3SpatrickCLANG_DIR=$(readlink -f $(dirname "$CLANG"))
493cab2bb3Spatrick
503cab2bb3SpatrickBUILD_DIR=$(readlink -f ./symbolizer)
513cab2bb3Spatrickmkdir -p $BUILD_DIR
523cab2bb3Spatrickcd $BUILD_DIR
533cab2bb3Spatrick
543cab2bb3SpatrickCC=$CLANG_DIR/clang
553cab2bb3SpatrickCXX=$CLANG_DIR/clang++
563cab2bb3SpatrickTBLGEN=$CLANG_DIR/llvm-tblgen
573cab2bb3SpatrickOPT=$CLANG_DIR/opt
58*810390e3SrobertAR=$CLANG_DIR/llvm-ar
59*810390e3SrobertLINK=$CLANG_DIR/llvm-link
603cab2bb3Spatrick
613cab2bb3Spatrickfor F in $CC $CXX $TBLGEN $LINK $OPT $AR; do
623cab2bb3Spatrick  if [[ ! -x "$F" ]]; then
633cab2bb3Spatrick    echo "Missing $F"
643cab2bb3Spatrick     exit 1
653cab2bb3Spatrick  fi
663cab2bb3Spatrickdone
673cab2bb3Spatrick
683cab2bb3SpatrickZLIB_BUILD=${BUILD_DIR}/zlib
693cab2bb3SpatrickLIBCXX_BUILD=${BUILD_DIR}/libcxx
703cab2bb3SpatrickLLVM_BUILD=${BUILD_DIR}/llvm
713cab2bb3SpatrickSYMBOLIZER_BUILD=${BUILD_DIR}/symbolizer
723cab2bb3Spatrick
733cab2bb3SpatrickFLAGS=${FLAGS:-}
74*810390e3SrobertTARGET_TRIPLE=$($CC -print-target-triple $FLAGS)
75*810390e3Srobertif [[ "$FLAGS" =~ "-m32" ]] ; then
76*810390e3Srobert  # Avoid new wrappers.
77*810390e3Srobert  FLAGS+=" -U_FILE_OFFSET_BITS"
78*810390e3Srobertfi
79*810390e3SrobertFLAGS+=" -fPIC -flto -Oz -g0 -DNDEBUG -target $TARGET_TRIPLE -Wno-unused-command-line-argument"
80*810390e3SrobertLINKFLAGS="-fuse-ld=lld -target $TARGET_TRIPLE"
813cab2bb3Spatrick
823cab2bb3Spatrick# Build zlib.
833cab2bb3Spatrickmkdir -p ${ZLIB_BUILD}
843cab2bb3Spatrickcd ${ZLIB_BUILD}
853cab2bb3Spatrickcp -r ${ZLIB_SRC}/* .
86*810390e3SrobertAR="${AR}" CC="${CC}" CFLAGS="$FLAGS -Wno-deprecated-non-prototype" RANLIB=/bin/true ./configure --static
87*810390e3Srobertmake -j libz.a
883cab2bb3Spatrick
893cab2bb3Spatrick# Build and install libcxxabi and libcxx.
903cab2bb3Spatrickif [[ ! -d ${LIBCXX_BUILD} ]]; then
913cab2bb3Spatrick  mkdir -p ${LIBCXX_BUILD}
923cab2bb3Spatrick  cd ${LIBCXX_BUILD}
93d89ec533Spatrick  LIBCXX_FLAGS="${FLAGS} -Wno-macro-redefined"
943cab2bb3Spatrick  cmake -GNinja \
95*810390e3Srobert    -DLLVM_ENABLE_RUNTIMES="libcxx;libcxxabi" \
963cab2bb3Spatrick    -DCMAKE_BUILD_TYPE=Release \
973cab2bb3Spatrick    -DCMAKE_C_COMPILER=$CC \
983cab2bb3Spatrick    -DCMAKE_CXX_COMPILER=$CXX \
993cab2bb3Spatrick    -DCMAKE_C_FLAGS_RELEASE="${LIBCXX_FLAGS}" \
1003cab2bb3Spatrick    -DCMAKE_CXX_FLAGS_RELEASE="${LIBCXX_FLAGS}" \
1013cab2bb3Spatrick    -DLIBCXXABI_ENABLE_ASSERTIONS=OFF \
1023cab2bb3Spatrick    -DLIBCXXABI_ENABLE_EXCEPTIONS=OFF \
1033cab2bb3Spatrick    -DLIBCXX_ENABLE_ASSERTIONS=OFF \
1043cab2bb3Spatrick    -DLIBCXX_ENABLE_EXCEPTIONS=OFF \
1053cab2bb3Spatrick    -DLIBCXX_ENABLE_RTTI=OFF \
106*810390e3Srobert    -DCMAKE_SHARED_LINKER_FLAGS="$LINKFLAGS" \
107*810390e3Srobert  $LLVM_SRC/../runtimes
1083cab2bb3Spatrickfi
1093cab2bb3Spatrickcd ${LIBCXX_BUILD}
1103cab2bb3Spatrickninja cxx cxxabi
1113cab2bb3Spatrick
1123cab2bb3SpatrickFLAGS="${FLAGS} -fno-rtti -fno-exceptions"
113*810390e3SrobertLLVM_CFLAGS="${FLAGS} -Wno-global-constructors"
114*810390e3SrobertLLVM_CXXFLAGS="${LLVM_CFLAGS} -nostdinc++ -I${ZLIB_BUILD} -isystem ${LIBCXX_BUILD}/include -isystem ${LIBCXX_BUILD}/include/c++/v1"
1153cab2bb3Spatrick
1163cab2bb3Spatrick# Build LLVM.
1173cab2bb3Spatrickif [[ ! -d ${LLVM_BUILD} ]]; then
1183cab2bb3Spatrick  mkdir -p ${LLVM_BUILD}
1193cab2bb3Spatrick  cd ${LLVM_BUILD}
1203cab2bb3Spatrick  cmake -GNinja \
1213cab2bb3Spatrick    -DCMAKE_BUILD_TYPE=Release \
1223cab2bb3Spatrick    -DCMAKE_C_COMPILER=$CC \
1233cab2bb3Spatrick    -DCMAKE_CXX_COMPILER=$CXX \
124*810390e3Srobert    -DCMAKE_C_FLAGS="${LLVM_CFLAGS}" \
125*810390e3Srobert    -DCMAKE_CXX_FLAGS="${LLVM_CXXFLAGS}" \
126*810390e3Srobert    -DCMAKE_EXE_LINKER_FLAGS="$LINKFLAGS -stdlib=libc++ -L${LIBCXX_BUILD}/lib" \
1273cab2bb3Spatrick    -DLLVM_TABLEGEN=$TBLGEN \
1283cab2bb3Spatrick    -DLLVM_ENABLE_ZLIB=ON \
1293cab2bb3Spatrick    -DLLVM_ENABLE_TERMINFO=OFF \
1303cab2bb3Spatrick    -DLLVM_ENABLE_THREADS=OFF \
1313cab2bb3Spatrick  $LLVM_SRC
1323cab2bb3Spatrickfi
1333cab2bb3Spatrickcd ${LLVM_BUILD}
134*810390e3Srobertninja LLVMSymbolize LLVMObject LLVMBinaryFormat LLVMDebugInfoDWARF LLVMSupport LLVMDebugInfoPDB LLVMDebuginfod LLVMMC LLVMDemangle LLVMTextAPI LLVMTargetParser
1353cab2bb3Spatrick
1363cab2bb3Spatrickcd ${BUILD_DIR}
1373cab2bb3Spatrickrm -rf ${SYMBOLIZER_BUILD}
1383cab2bb3Spatrickmkdir ${SYMBOLIZER_BUILD}
1393cab2bb3Spatrickcd ${SYMBOLIZER_BUILD}
1403cab2bb3Spatrick
1413cab2bb3Spatrickecho "Compiling..."
142*810390e3SrobertSYMBOLIZER_FLAGS="$LLVM_CXXFLAGS -I${LLVM_SRC}/include -I${LLVM_BUILD}/include -std=c++17"
1433cab2bb3Spatrick$CXX $SYMBOLIZER_FLAGS ${SRC_DIR}/sanitizer_symbolize.cpp ${SRC_DIR}/sanitizer_wrappers.cpp -c
1443cab2bb3Spatrick$AR rc symbolizer.a sanitizer_symbolize.o sanitizer_wrappers.o
1453cab2bb3Spatrick
146*810390e3SrobertSYMBOLIZER_API_LIST=__sanitizer_symbolize_code
147*810390e3SrobertSYMBOLIZER_API_LIST+=,__sanitizer_symbolize_data
148*810390e3SrobertSYMBOLIZER_API_LIST+=,__sanitizer_symbolize_flush
149*810390e3SrobertSYMBOLIZER_API_LIST+=,__sanitizer_symbolize_demangle
150*810390e3SrobertSYMBOLIZER_API_LIST+=,__sanitizer_symbolize_set_demangle
151*810390e3SrobertSYMBOLIZER_API_LIST+=,__sanitizer_symbolize_set_inline_frames
152*810390e3Srobert
153*810390e3SrobertLIBCXX_ARCHIVE_DIR=$(dirname $(find $LIBCXX_BUILD -name libc++.a | head -n1))
1543cab2bb3Spatrick
1553cab2bb3Spatrick# Merge all the object files together and copy the resulting library back.
156*810390e3Srobert$LINK $LIBCXX_ARCHIVE_DIR/libc++.a \
157*810390e3Srobert      $LIBCXX_ARCHIVE_DIR/libc++abi.a \
1583cab2bb3Spatrick      $LLVM_BUILD/lib/libLLVMSymbolize.a \
1593cab2bb3Spatrick      $LLVM_BUILD/lib/libLLVMObject.a \
1603cab2bb3Spatrick      $LLVM_BUILD/lib/libLLVMBinaryFormat.a \
1613cab2bb3Spatrick      $LLVM_BUILD/lib/libLLVMDebugInfoDWARF.a \
1623cab2bb3Spatrick      $LLVM_BUILD/lib/libLLVMSupport.a \
1633cab2bb3Spatrick      $LLVM_BUILD/lib/libLLVMDebugInfoPDB.a \
1641f9cb04fSpatrick      $LLVM_BUILD/lib/libLLVMDebugInfoMSF.a \
1651f9cb04fSpatrick      $LLVM_BUILD/lib/libLLVMDebugInfoCodeView.a \
166*810390e3Srobert      $LLVM_BUILD/lib/libLLVMDebuginfod.a \
1673cab2bb3Spatrick      $LLVM_BUILD/lib/libLLVMDemangle.a \
1683cab2bb3Spatrick      $LLVM_BUILD/lib/libLLVMMC.a \
1693cab2bb3Spatrick      $LLVM_BUILD/lib/libLLVMTextAPI.a \
170*810390e3Srobert      $LLVM_BUILD/lib/libLLVMTargetParser.a \
1713cab2bb3Spatrick      $ZLIB_BUILD/libz.a \
1723cab2bb3Spatrick      symbolizer.a \
173*810390e3Srobert      -ignore-non-bitcode -o all.bc
1743cab2bb3Spatrick
1753cab2bb3Spatrickecho "Optimizing..."
176*810390e3Srobert$OPT -passes=internalize -internalize-public-api-list=${SYMBOLIZER_API_LIST} all.bc -o opt.bc
1773cab2bb3Spatrick$CC $FLAGS -fno-lto -c opt.bc -o symbolizer.o
1783cab2bb3Spatrick
1793cab2bb3Spatrickecho "Checking undefined symbols..."
1803cab2bb3Spatricknm -f posix -g symbolizer.o | cut -f 1,2 -d \  | LC_COLLATE=C sort -u > undefined.new
1813cab2bb3Spatrick(diff -u $SCRIPT_DIR/global_symbols.txt undefined.new | grep -E "^\+[^+]") && \
1823cab2bb3Spatrick  (echo "Failed: unexpected symbols"; exit 1)
1833cab2bb3Spatrick
1843cab2bb3Spatrickarch() {
1853cab2bb3Spatrick  objdump -f $1 | grep -m1 -Po "(?<=file format ).*$"
1863cab2bb3Spatrick}
1873cab2bb3Spatrick
1883cab2bb3SpatrickSYMBOLIZER_FORMAT=$(arch symbolizer.o)
1893cab2bb3Spatrickecho "Injecting $SYMBOLIZER_FORMAT symbolizer..."
1903cab2bb3Spatrickfor A in $TARGE_DIR/libclang_rt.*san*.a; do
1913cab2bb3Spatrick  A_FORMAT=$(arch $A)
1923cab2bb3Spatrick  if [[ "$A_FORMAT" != "$SYMBOLIZER_FORMAT" ]] ; then
1933cab2bb3Spatrick    continue
1943cab2bb3Spatrick  fi
1953cab2bb3Spatrick  (nm -u $A 2>/dev/null | grep -E "__sanitizer_symbolize_code" >/dev/null) || continue
1963cab2bb3Spatrick  echo "$A"
1973cab2bb3Spatrick  $AR rcs $A symbolizer.o
1983cab2bb3Spatrickdone
1993cab2bb3Spatrick
2003cab2bb3Spatrickecho "Success!"
201