Merge branch 'main' into fix/fixit-unknown-attributes

Author: a-tarasyuk

.github/new-prs-labeler.yml

@@ -554,6 +554,12 @@ flang:fir-hlfir:
 flang:codegen:
   - flang/**/CodeGen/**
 
+llvm:codegen:
+  - llvm/lib/CodeGen/*
+  - llvm/lib/CodeGen/MIRParser/*
+  - llvm/lib/CodeGen/LiveDebugValues/*
+  - llvm/lib/CodeGen/AsmPrinter/*
+
 llvm:globalisel:
   - llvm/**/GlobalISel/**
   - llvm/utils/TableGen/GlobalISel*

.github/workflows/containers/github-action-ci-windows/Dockerfile

@@ -85,7 +85,7 @@ RUN powershell -Command \
 RUN git config --system core.longpaths true & \
     git config --global core.autocrlf false
 
-ARG RUNNER_VERSION=2.324.0
+ARG RUNNER_VERSION=2.325.0
 ENV RUNNER_VERSION=$RUNNER_VERSION
 
 RUN powershell -Command \

.github/workflows/containers/github-action-ci/Dockerfile

@@ -86,7 +86,7 @@ WORKDIR /home/gha
 
 FROM ci-container as ci-container-agent
 
-ENV GITHUB_RUNNER_VERSION=2.324.0
+ENV GITHUB_RUNNER_VERSION=2.325.0
 
 RUN mkdir actions-runner && \
     cd actions-runner && \

bolt/include/bolt/Core/BinaryFunction.h

@@ -2348,6 +2348,7 @@ class BinaryFunction {
       releaseCFG();
       CurrentState = State::Emitted;
     }
+    clearList(Relocations);
   }
 
   /// Process LSDA information for the function.

bolt/lib/Core/BinaryContext.cpp

@@ -1032,10 +1032,8 @@ void BinaryContext::adjustCodePadding() {
 
     if (!hasValidCodePadding(BF)) {
       if (HasRelocations) {
-        if (opts::Verbosity >= 1) {
-          this->outs() << "BOLT-INFO: function " << BF
-                       << " has invalid padding. Ignoring the function.\n";
-        }
+        this->errs() << "BOLT-WARNING: function " << BF
+                     << " has invalid padding. Ignoring the function\n";
         BF.setIgnored();
       } else {
         BF.setMaxSize(BF.getSize());

bolt/lib/Core/BinaryFunction.cpp

@@ -1526,8 +1526,6 @@ Error BinaryFunction::disassemble() {
   if (uint64_t Offset = getFirstInstructionOffset())
     Labels[Offset] = BC.Ctx->createNamedTempSymbol();
 
-  clearList(Relocations);
-
   if (!IsSimple) {
     clearList(Instructions);
     return createNonFatalBOLTError("");
@@ -3244,16 +3242,26 @@ void BinaryFunction::setTrapOnEntry() {
 }
 
 void BinaryFunction::setIgnored() {
+  IsIgnored = true;
+
   if (opts::processAllFunctions()) {
     // We can accept ignored functions before they've been disassembled.
-    // In that case, they would still get disassembled and emited, but not
+    // In that case, they would still get disassembled and emitted, but not
     // optimized.
-    assert(CurrentState == State::Empty &&
-           "cannot ignore non-empty functions in current mode");
-    IsIgnored = true;
+    if (CurrentState != State::Empty) {
+      BC.errs() << "BOLT-ERROR: cannot ignore non-empty function " << *this
+                << " in current mode\n";
+      exit(1);
+    }
     return;
   }
 
+  IsSimple = false;
+  LLVM_DEBUG(dbgs() << "Ignoring " << getPrintName() << '\n');
+
+  if (CurrentState == State::Empty)
+    return;
+
   clearDisasmState();
 
   // Clear CFG state too.
@@ -3273,9 +3281,11 @@ void BinaryFunction::setIgnored() {
 
   CurrentState = State::Empty;
 
-  IsIgnored = true;
-  IsSimple = false;
-  LLVM_DEBUG(dbgs() << "Ignoring " << getPrintName() << '\n');
+  // Fix external references in the original function body.
+  if (BC.HasRelocations) {
+    LLVM_DEBUG(dbgs() << "Scanning refs in " << *this << '\n');
+    scanExternalRefs();
+  }
 }
 
 void BinaryFunction::duplicateConstantIslands() {
@@ -3764,7 +3774,6 @@ void BinaryFunction::postProcessBranches() {
 
 MCSymbol *BinaryFunction::addEntryPointAtOffset(uint64_t Offset) {
   assert(Offset && "cannot add primary entry point");
-  assert(CurrentState == State::Empty || CurrentState == State::Disassembled);
 
   const uint64_t EntryPointAddress = getAddress() + Offset;
   MCSymbol *LocalSymbol = getOrCreateLocalLabel(EntryPointAddress);
@@ -3773,6 +3782,8 @@ MCSymbol *BinaryFunction::addEntryPointAtOffset(uint64_t Offset) {
   if (EntrySymbol)
     return EntrySymbol;
 
+  assert(CurrentState == State::Empty || CurrentState == State::Disassembled);
+
   if (BinaryData *EntryBD = BC.getBinaryDataAtAddress(EntryPointAddress)) {
     EntrySymbol = EntryBD->getSymbol();
   } else {

bolt/lib/Passes/ADRRelaxationPass.cpp

@@ -81,17 +81,15 @@ void ADRRelaxationPass::runOnFunction(BinaryFunction &BF) {
         It = BB.eraseInstruction(std::prev(It));
       } else if (std::next(It) != BB.end() && BC.MIB->isNoop(*std::next(It))) {
         BB.eraseInstruction(std::next(It));
-      } else if (!opts::StrictMode && !BF.isSimple()) {
+      } else if (!BF.isSimple()) {
         // If the function is not simple, it may contain a jump table undetected
         // by us. This jump table may use an offset from the branch instruction
         // to land in the desired place. If we add new instructions, we
         // invalidate this offset, so we have to rely on linker-inserted NOP to
         // replace it with ADRP, and abort if it is not present.
         auto L = BC.scopeLock();
-        BC.errs() << formatv(
-            "BOLT-ERROR: Cannot relax adr in non-simple function "
-            "{0}. Use --strict option to override\n",
-            BF.getOneName());
+        BC.errs() << "BOLT-ERROR: cannot relax ADR in non-simple function "
+                  << BF << '\n';
         PassFailed = true;
         return;
       }

bolt/lib/Profile/DataAggregator.cpp

@@ -1220,7 +1220,7 @@ std::error_code DataAggregator::parseAggregatedLBREntry() {
   // Storage for parsed fields.
   StringRef EventName;
   std::optional<Location> Addr[3];
-  int64_t Counters[2];
+  int64_t Counters[2] = {0};
 
   while (Type == INVALID || Type == EVENT_NAME) {
     while (checkAndConsumeFS()) {

bolt/lib/Rewrite/RewriteInstance.cpp

@@ -3446,6 +3446,7 @@ void RewriteInstance::disassembleFunctions() {
         BC->outs() << "BOLT-INFO: could not disassemble function " << Function
                    << ". Will ignore.\n";
       // Forcefully ignore the function.
+      Function.scanExternalRefs();
       Function.setIgnored();
     });
 

bolt/test/AArch64/adr-relaxation-fail.s

@@ -0,0 +1,33 @@
+## Check that llvm-bolt generates a proper error message when ADR instruction
+## cannot be relaxed.
+
+# RUN: llvm-mc -filetype=obj -triple aarch64-unknown-unknown %s -o %t.o
+# RUN: %clang %cflags %t.o -o %t.exe -Wl,-q -static
+# RUN: not llvm-bolt %t.exe -o %t.bolt 2>&1 | FileCheck %s
+# RUN: not llvm-bolt %t.exe -o %t.bolt --strict 2>&1 | FileCheck %s
+
+# CHECK: BOLT-ERROR: cannot relax ADR in non-simple function _start
+
+## The function contains unknown control flow and llvm-bolt fails to recover
+## CFG. As BOLT has to preserve the function layout, the ADR instruction cannot
+## be relaxed into ADRP+ADD.
+  .text
+  .globl _start
+  .type _start, %function
+_start:
+  .cfi_startproc
+  adr x1, foo
+  adr x2, .L1
+.L1:
+  br x0
+  ret  x0
+  .cfi_endproc
+  .size _start, .-_start
+
+  .globl foo
+  .type foo, %function
+foo:
+  .cfi_startproc
+  ret  x0
+  .cfi_endproc
+  .size foo, .-foo