removes section and other symbols in ELF (#1220)

* removes an extra llvm from opam's depext

We already have llvm-9 installed so there is no reason to install the
default llvm also.

* enables options in configure-omake

though careful many of the options will be just ignored, either by the
script or by oasis/omake intergration.

Use the new feature immediately in the build-and-test action to enable
tests and specify the llvm version (which is for now ignored, but see
the next commit).

* pulls the llvm configuration from setup in omake

This is much better than specifying them manually as we did before :)

Also, switches back to the static linking with llvm when omake is
used.

* disables the emission of section, file, and other symbols for ELF

The section symbols confuse our symbolization procedure so that at the
end we get symbols named as `.text` instead of their real names. To
prevent this, we are not emitting anymore relocation information about
symbols that are sections, files, objects, and other non-functions.

This commit fixes the nightly regression test. The underlying issue is
triggered only on modern LLVM (e.g., it is not observable on LLVM 6
and before, but is definitely triggered on 9, not sure about the in
between versions).

Author: ivg

.github/workflows/build-and-test.yml

@@ -36,7 +36,11 @@ jobs:
                opam install omake
 
          - name: Configure BAP
-           run: opam exec -- ./configure-omake
+           run: |
+               opam exec -- ./configure-omake \
+                 --enable-tests \
+                 --with-llvm-version=9 \
+                 --with-llvm-config=llvm-config-9
 
          - name: Build BAP
            run: opam exec -- make

configure-omake

@@ -1,6 +1,6 @@
 #!/bin/sh
 
-rm -f _oasis
+rm -f _oasis setup.log.om setup.data _oasis_setup.om
 mv oasis/benchmarks oasis/benchmarks.backup
 mv oasis/piqi-printers oasis/piqi-printers.backup
 mv oasis/common oasis/common.backup
@@ -17,10 +17,32 @@ sed -i 's/.*Build$.*//' _oasis
 sed -i 's/.*CCOpt:.*//' _oasis
 sed -i 's/.*CCLib:.*//' _oasis
 
+FST=true
+
+
+for i in "$@"; do
+  if $FST; then
+    set --
+    FST=false
+  fi
+
+  case $i in
+    --*=*)
+      ARG=${i%%=*}
+      VAL=${i##*=}
+      set -- "$@" "$ARG" "$VAL"
+      ;;
+    *)
+      set -- "$@" "$i"
+      ;;
+  esac
+done
+
+
 # call setup twice so that it can capture the AB deps
 for pass in `seq 2`; do
     oasis setup
     ocamlfind ocamlopt unix.cmxa setup.ml -o setup.exe
     rm setup.cmx setup.cmi setup.o
-    ./setup.exe -configure --prefix $(opam config var prefix) --enable-tests
+    ./setup.exe -configure --prefix $(opam config var prefix) "$@"
 done

lib/bap_llvm/OMakefile

@@ -3,10 +3,14 @@
 # for more documentation.
 
 CFLAGS += -O2
-CXXFLAGS += -O2 -std=c++11 -fPIC -I$(ROOT)/lib/bap_disasm -I/usr/lib/llvm-6.0/include -std=c++0x -fuse-ld=gold -Wl,--no-keep-files-mapped -Wl,--no-map-whole-files -fPIC -fvisibility-inlines-hidden -Werror=date-time -std=c++11 -Wall -W -Wno-unused-parameter -Wwrite-strings -Wcast-qual -Wno-missing-field-initializers -pedantic -Wno-long-long -Wno-maybe-uninitialized -Wdelete-non-virtual-dtor -Wno-comment -ffunction-sections -fdata-sections -O2 -DNDEBUG  -fno-exceptions -D_GNU_SOURCE -D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS -D__STDC_LIMIT_MACROS
-
-OCAMLMKLIBFLAGS += -L/usr/lib/llvm-6.0/lib -lLLVM-6.0 -dllpath /usr/lib/llvm-6.0/lib -lstdc++  -lz -lrt -ldl -ltinfo -lpthread -lm -lcurses
-
+CXXFLAGS += -O2 -std=c++11 -fPIC -I$(ROOT)/lib/bap_disasm
+
+# pull the llvm configuration
+CXXFLAGS += $(split $(oasis_llvm_cxxflags))
+OCAMLMKLIBFLAGS += \
+	$(split $(oasis_llvm_libs)) \
+	$(split $(oasis_llvm_ldflags)) \
+	-lcurses -lstdc++
 
 include _oasis_hier.om
 

lib/bap_llvm/llvm_elf_loader.hpp

@@ -202,12 +202,16 @@ void emit_relocations(const ELFObjectFile<T> &obj, ogre_doc &s) {
         for (auto rel : sec.relocations()) {
             auto sym = rel.getSymbol();
             if (sym != prim::end_symbols(obj)) {
-                uint64_t raddr = prim::relocation_offset(rel);
-                if (auto addr = prim::symbol_address(*sym))
-                    if (*addr) s.entry("llvm:relocation") << raddr << *addr;
-                if (auto name = prim::symbol_name(*sym))
-                    if (!name->empty())
-                        s.entry("llvm:name-reference") << raddr << *name;
+                auto typ = prim::symbol_type(*sym);
+                if (typ && (*typ == SymbolRef::ST_Function ||
+                            *typ == SymbolRef::ST_Unknown)) {
+                    uint64_t raddr = prim::relocation_offset(rel);
+                    if (auto addr = prim::symbol_address(*sym))
+                        if (*addr) s.entry("llvm:relocation") << raddr << *addr;
+                    if (auto name = prim::symbol_name(*sym))
+                        if (!name->empty())
+                            s.entry("llvm:name-reference") << raddr << *name;
+                }
             }
         }
     }

opam/opam

@@ -308,7 +308,6 @@ depexts: [
     "llvm-9-dev"
     "time"
     "clang"
-    "llvm"
     "m4"
     "dejagnu"
   ] {os-distribution = "ubuntu"}