[llvm] [clang] [ASAN] For Asan instrumented global, emit two symbols, one with actual size and other with instrumented size. (PR #70166)

[via cfe-commits]

https://github.com/skc7 updated https://github.com/llvm/llvm-project/pull/70166

>From 81dd1858512ce23154d5fd798edf8ec7c0e4c70d Mon Sep 17 00:00:00 2001
From: skc7 <Krishna.Sankisa at amd.com>
Date: Wed, 25 Oct 2023 10:46:10 +0530
Subject: [PATCH] [ASAN] For Asan instrumented globals, emit two symbols, with
 actual size and instrumented size.

---
 clang/test/CodeGen/asan_globals_symbols.cpp   | 15 ++++++++++
 .../asan_globals_symbols_ir_attribute.cpp     | 15 ++++++++++
 llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp    | 28 ++++++++++++++++++-
 .../Instrumentation/AddressSanitizer.cpp      |  3 ++
 .../AddressSanitizer/debug-info-global-var.ll |  2 +-
 5 files changed, 61 insertions(+), 2 deletions(-)
 create mode 100644 clang/test/CodeGen/asan_globals_symbols.cpp
 create mode 100644 clang/test/CodeGen/asan_globals_symbols_ir_attribute.cpp

diff --git a/clang/test/CodeGen/asan_globals_symbols.cpp b/clang/test/CodeGen/asan_globals_symbols.cpp
new file mode 100644
index 00000000000000..d53afb2433b171
--- /dev/null
+++ b/clang/test/CodeGen/asan_globals_symbols.cpp
@@ -0,0 +1,15 @@
+// RUN: %clang_cc1 -S -x c++ -std=c++11 -triple x86_64-linux \
+// RUN:   -fsanitize=address -o %t.out %s
+// RUN: FileCheck %s --input-file=%t.out --check-prefix=CHECK-A
+
+// CHECK-A: myGlobal:
+// CHECK-A: .size   myGlobal, 4
+// CHECK-A: myGlobal__sanitized_padded_global:
+// CHECK-A  .size   myGlobal__sanitized_padded_global, 32
+
+int myGlobal;
+
+int main() {
+    myGlobal = 0;
+    return 0;
+}
diff --git a/clang/test/CodeGen/asan_globals_symbols_ir_attribute.cpp b/clang/test/CodeGen/asan_globals_symbols_ir_attribute.cpp
new file mode 100644
index 00000000000000..f8ba5eb737696c
--- /dev/null
+++ b/clang/test/CodeGen/asan_globals_symbols_ir_attribute.cpp
@@ -0,0 +1,15 @@
+// RUN: %clang_cc1 %s -triple x86_64-unknown-linux-gnu -fsanitize=address -emit-llvm -o - | FileCheck -check-prefix=CHECK-ASAN %s
+
+// CHECK-ASAN: @myGlobal1 = global { i32, [28 x i8] } zeroinitializer, align 32 #[[ATTR0:[0-9]+]]
+// CHECK-ASAN: @myGlobal2 = global i32 0, no_sanitize_address, align 4
+// CHECK-NOT: #[[ATTR1:[0-9]+]]
+// CHECK-ASAN: attributes #[[ATTR0]] = { sanitized_padded_global }
+
+int myGlobal1;
+int __attribute__((no_sanitize("address"))) myGlobal2;
+
+int main() {
+    myGlobal1 = 0;
+    myGlobal2 = 0;
+    return 0;
+}
diff --git a/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp b/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
index 7df1c82bf357f6..cf1bdde6fc3425 100644
--- a/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
@@ -801,6 +801,19 @@ void AsmPrinter::emitGlobalVariable(const GlobalVariable *GV) {
   // sections and expected to be contiguous (e.g. ObjC metadata).
   const Align Alignment = getGVAlignment(GV, DL);
 
+  // Identify globals with "SanitizedPaddedGlobal" attribute and extract
+  // the actual global variable size.
+  uint64_t ActualSize = 0;
+  if (GV->hasAttribute(Attribute::SanitizedPaddedGlobal)) {
+    StructType *ST = dyn_cast<StructType>(GV->getValueType());
+    if (ST && ST->getNumElements() == 2) {
+      auto *ET0 = ST->getElementType(0);
+      if (ET0 && isa<ArrayType>(ST->getElementType(1))) {
+        ActualSize = DL.getTypeAllocSize(ET0);
+      }
+    }
+  }
+
   for (const HandlerInfo &HI : Handlers) {
     NamedRegionTimer T(HI.TimerName, HI.TimerDescription,
                        HI.TimerGroupName, HI.TimerGroupDescription,
@@ -911,6 +924,18 @@ void AsmPrinter::emitGlobalVariable(const GlobalVariable *GV) {
 
   MCSymbol *EmittedInitSym = GVSym;
 
+  if (GV->hasAttribute(Attribute::SanitizedPaddedGlobal)) {
+    OutStreamer->switchSection(TheSection);
+    emitLinkage(GV, EmittedInitSym);
+    OutStreamer->emitLabel(EmittedInitSym);
+    if (MAI->hasDotTypeDotSizeDirective())
+      OutStreamer->emitELFSize(EmittedInitSym,
+                               MCConstantExpr::create(ActualSize, OutContext));
+    EmittedInitSym = OutContext.getOrCreateSymbol(
+        GVSym->getName() + Twine("__sanitized_padded_global"));
+    emitVisibility(EmittedInitSym, GV->getVisibility(), !GV->isDeclaration());
+  }
+
   OutStreamer->switchSection(TheSection);
 
   emitLinkage(GV, EmittedInitSym);
@@ -918,7 +943,8 @@ void AsmPrinter::emitGlobalVariable(const GlobalVariable *GV) {
 
   OutStreamer->emitLabel(EmittedInitSym);
   MCSymbol *LocalAlias = getSymbolPreferLocal(*GV);
-  if (LocalAlias != EmittedInitSym)
+  if ((LocalAlias != EmittedInitSym) &&
+      !GV->hasAttribute(Attribute::SanitizedPaddedGlobal))
     OutStreamer->emitLabel(LocalAlias);
 
   emitGlobalConstant(GV->getParent()->getDataLayout(), GV->getInitializer());
diff --git a/llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp b/llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp
index caab98c732eeeb..a0dbc5224b3ebc 100644
--- a/llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp
+++ b/llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp
@@ -2496,6 +2496,9 @@ void ModuleAddressSanitizer::instrumentGlobals(IRBuilder<> &IRB, Module &M,
     // zero so we can copy the metadata over as is.
     NewGlobal->copyMetadata(G, 0);
 
+    // Attach "SanitizedPaddedGlobal" attribute to the new global.
+    NewGlobal->addAttribute(Attribute::SanitizedPaddedGlobal);
+
     Value *Indices2[2];
     Indices2[0] = IRB.getInt32(0);
     Indices2[1] = IRB.getInt32(0);
diff --git a/llvm/test/Instrumentation/AddressSanitizer/debug-info-global-var.ll b/llvm/test/Instrumentation/AddressSanitizer/debug-info-global-var.ll
index 0b516e0174d6d1..2815c1f04bff15 100644
--- a/llvm/test/Instrumentation/AddressSanitizer/debug-info-global-var.ll
+++ b/llvm/test/Instrumentation/AddressSanitizer/debug-info-global-var.ll
@@ -2,7 +2,7 @@
 source_filename = "version.c"
 target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.12.0"
-; CHECK: @version = constant { [5 x i8], [27 x i8] } {{.*}}, !dbg ![[GV:.*]]
+; CHECK: @version = constant { [5 x i8], [27 x i8] } {{.*}}, !dbg ![[GV:.*]] {{.*}}
 
 @version = constant [5 x i8] c"4.00\00", align 1, !dbg !0