[llvm] [BOLT] Add support for safe-icf (PR #116275)

[Alexander Yermolovich via llvm-commits]

================
@@ -341,6 +341,61 @@ typedef std::unordered_map<BinaryFunction *, std::vector<BinaryFunction *>,
 namespace llvm {
 namespace bolt {
 
+Error IdenticalCodeFolding::processDataRelocations(
+    BinaryContext &BC, const SectionRef &SecRefRelData) {
+  for (const RelocationRef &Rel : SecRefRelData.relocations()) {
+    symbol_iterator SymbolIter = Rel.getSymbol();
+    const ObjectFile *OwningObj = Rel.getObject();
+    assert(SymbolIter != OwningObj->symbol_end() &&
+           "relocation Symbol expected");
+    const SymbolRef &Symbol = *SymbolIter;
+    const uint64_t SymbolAddress = cantFail(Symbol.getAddress());
+    const ELFObjectFileBase *ELFObj = dyn_cast<ELFObjectFileBase>(OwningObj);
+    if (!ELFObj)
+      return createFatalBOLTError(
+          Twine("BOLT-ERROR: Only ELFObjectFileBase is supported"));
+    const int64_t Addend = getRelocationAddend(ELFObj, Rel);
+    BinaryFunction *BF = BC.getBinaryFunctionAtAddress(SymbolAddress + Addend);
+    if (!BF)
+      continue;
+    BF->setUnsafeICF();
+  }
+  return Error::success();
+}
+
+Error IdenticalCodeFolding::markFunctionsUnsafeToFold(BinaryContext &BC) {
+  ErrorOr<BinarySection &> SecRelData = BC.getUniqueSectionByName(".rela.data");
----------------
ayermolo wrote:

I mean nothing is impossible with all the stuff that ends up in prod binaries between various compilers and languages. It's just from some common patterns with c++ this was the case.
Changing code to process all the data sections, skipping ".plt", ".got.plt", ".eh_frame", ".gcc_except_table", ".data.rel.ro", ".fini_array", ".init_array"
Processing ".fini_array", ".init_array" seperatly to mark things in them as safe.
and running bolt on HHVM some stats:

BOLT-DEBUG: Relocation type: R_X86_64_64
BOLT-DEBUG: Relocation type: R_X86_64_PC32
BOLT-DEBUG: Section with PC: .rela.rodata
BOLT-DEBUG: Section with function references: .rela.data
BOLT-DEBUG: Section with function references: .rela.fini_array
BOLT-DEBUG: Section with function references: .rela.init_array
BOLT-DEBUG: Section with function references: .rela.rodata
BOLT-DEBUG: Section with function references: .rela.tdata

not marking function symbols from ".fini_array", ".init_array" as safe did not change size of what got folded.

Trying to figure out how PC32 gets generated in the .rela.rodata.

https://github.com/llvm/llvm-project/pull/116275