[clang] Support ASan on WASI

[kateinoigakukun]

I'm working on porting ASan to Wasm/WASI targets, and this is the first part of the change sets. I'll post runtime changes separately.

This change makes -fsanitize=address available for WASI target by replicating what we do for Emscripten because they share the same memory model.

[llvmbot]

@llvm/pr-subscribers-backend-webassembly
@llvm/pr-subscribers-compiler-rt-sanitizer
@llvm/pr-subscribers-clang
@llvm/pr-subscribers-llvm-transforms

@llvm/pr-subscribers-clang-driver

Author: Yuta Saito (kateinoigakukun)

Changes

I'm working on porting ASan to Wasm/WASI targets, and this is the first part of the change sets. I'll post runtime changes separately.

This change makes -fsanitize=address available for WASI target by replicating what we do for Emscripten because they share the same memory model.

---

Full diff: https://github.com/llvm/llvm-project/pull/139014.diff

2 Files Affected:

• (modified) clang/lib/Driver/ToolChains/WebAssembly.cpp (+6-1)
• (modified) llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp (+4-4)

diff --git a/clang/lib/Driver/ToolChains/WebAssembly.cpp b/clang/lib/Driver/ToolChains/WebAssembly.cpp
index cd12f2ae5a6de..9fcc33728c1ad 100644
--- a/clang/lib/Driver/ToolChains/WebAssembly.cpp
+++ b/clang/lib/Driver/ToolChains/WebAssembly.cpp
@@ -545,8 +545,13 @@ void WebAssembly::AddCXXStdlibLibArgs(const llvm::opt::ArgList &Args,
 SanitizerMask WebAssembly::getSupportedSanitizers() const {
   SanitizerMask Res = ToolChain::getSupportedSanitizers();
   if (getTriple().isOSEmscripten()) {
-    Res |= SanitizerKind::Vptr | SanitizerKind::Leak | SanitizerKind::Address;
+    Res |= SanitizerKind::Vptr | SanitizerKind::Leak;
   }
+
+  if (getTriple().isOSEmscripten() || getTriple().isOSWASI()) {
+    Res |= SanitizerKind::Address;
+  }
+
   // -fsanitize=function places two words before the function label, which are
   // -unsupported.
   Res &= ~SanitizerKind::Function;
diff --git a/llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp b/llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp
index c1dba77c3532b..840a5e3f31dfd 100644
--- a/llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp
+++ b/llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp
@@ -118,7 +118,7 @@ static const uint64_t kNetBSD_ShadowOffset64 = 1ULL << 46;
 static const uint64_t kNetBSDKasan_ShadowOffset64 = 0xdfff900000000000;
 static const uint64_t kPS_ShadowOffset64 = 1ULL << 40;
 static const uint64_t kWindowsShadowOffset32 = 3ULL << 28;
-static const uint64_t kEmscriptenShadowOffset = 0;
+static const uint64_t kWebAssemblyShadowOffset = 0;
 
 // The shadow memory space is dynamically allocated.
 static const uint64_t kWindowsShadowOffset64 = kDynamicShadowSentinel;
@@ -499,9 +499,9 @@ static ShadowMapping getShadowMapping(const Triple &TargetTriple, int LongSize,
   bool IsRISCV64 = TargetTriple.getArch() == Triple::riscv64;
   bool IsWindows = TargetTriple.isOSWindows();
   bool IsFuchsia = TargetTriple.isOSFuchsia();
-  bool IsEmscripten = TargetTriple.isOSEmscripten();
   bool IsAMDGPU = TargetTriple.isAMDGPU();
   bool IsHaiku = TargetTriple.isOSHaiku();
+  bool IsWasm = TargetTriple.isWasm();
 
   ShadowMapping Mapping;
 
@@ -525,8 +525,8 @@ static ShadowMapping getShadowMapping(const Triple &TargetTriple, int LongSize,
       Mapping.Offset = kDynamicShadowSentinel;
     else if (IsWindows)
       Mapping.Offset = kWindowsShadowOffset32;
-    else if (IsEmscripten)
-      Mapping.Offset = kEmscriptenShadowOffset;
+    else if (IsWasm)
+      Mapping.Offset = kWebAssemblyShadowOffset;
     else
       Mapping.Offset = kDefaultShadowOffset32;
   } else {  // LongSize == 64

[dschuff]

This looks good to me, @sbc100 and @sunfishcode might also be interested (presumably the sanitizer runtime will end up alongside wasi-libc in the wasi SDK?)

[kateinoigakukun]

presumably the sanitizer runtime will end up alongside wasi-libc in the wasi SDK?

@dschuff Yes, it will be placed alongside the other clang_rt libraries in WASI SDK and Swift SDK for Wasm/WASI