[hwasan] Remove lazy thread-initialisation

This was an experiment made possible by a non-standard feature of the
Android dynamic loader.

It required introducing a flag to tell the compiler which ABI was being
targeted.
This flag is no longer needed, since the generated code now works for
both ABI's.

We leave that flag untouched for backwards compatibility. This also
means that if we need to distinguish between targeted ABI's again
we can do that without disturbing any existing workflows.

We leave a comment in the source code and mention in the help text to
explain this for any confused person reading the code in the future.

Patch by Matthew Malcomson

Differential Revision: https://reviews.llvm.org/D69574

Author: David Spickett

clang/include/clang/Driver/Options.td

@@ -1032,10 +1032,16 @@ def fno_sanitize_address_use_odr_indicator
     : Flag<["-"], "fno-sanitize-address-use-odr-indicator">,
       Group<f_clang_Group>,
       HelpText<"Disable ODR indicator globals">;
+// Note: This flag was introduced when it was necessary to distinguish between
+//       ABI for correct codegen.  This is no longer needed, but the flag is
+//       not removed since targeting either ABI will behave the same.
+//       This way we cause no disturbance to existing scripts & code, and if we
+//       want to use this flag in the future we will cause no disturbance then
+//       either.
 def fsanitize_hwaddress_abi_EQ
     : Joined<["-"], "fsanitize-hwaddress-abi=">,
       Group<f_clang_Group>,
-      HelpText<"Select the HWAddressSanitizer ABI to target (interceptor or platform, default interceptor)">;
+      HelpText<"Select the HWAddressSanitizer ABI to target (interceptor or platform, default interceptor). This option is currently unused.">;
 def fsanitize_recover : Flag<["-"], "fsanitize-recover">, Group<f_clang_Group>;
 def fno_sanitize_recover : Flag<["-"], "fno-sanitize-recover">,
                            Flags<[CoreOption, DriverOption]>,

compiler-rt/lib/hwasan/hwasan_interceptors.cpp

@@ -202,23 +202,33 @@ INTERCEPTOR_ALIAS(__sanitizer_struct_mallinfo, mallinfo);
 INTERCEPTOR_ALIAS(int, mallopt, int cmd, int value);
 INTERCEPTOR_ALIAS(void, malloc_stats, void);
 #endif
-#endif // HWASAN_WITH_INTERCEPTORS
 
+struct ThreadStartArg {
+  thread_callback_t callback;
+  void *param;
+};
+
+static void *HwasanThreadStartFunc(void *arg) {
+  __hwasan_thread_enter();
+  ThreadStartArg A = *reinterpret_cast<ThreadStartArg*>(arg);
+  UnmapOrDie(arg, GetPageSizeCached());
+  return A.callback(A.param);
+}
 
-#if HWASAN_WITH_INTERCEPTORS && !defined(__aarch64__)
-INTERCEPTOR(int, pthread_create, void *th, void *attr,
-            void *(*callback)(void *), void *param) {
+INTERCEPTOR(int, pthread_create, void *th, void *attr, void *(*callback)(void*),
+            void * param) {
   ScopedTaggingDisabler disabler;
+  ThreadStartArg *A = reinterpret_cast<ThreadStartArg *> (MmapOrDie(
+      GetPageSizeCached(), "pthread_create"));
+  *A = {callback, param};
   int res = REAL(pthread_create)(UntagPtr(th), UntagPtr(attr),
-                                 callback, param);
+                                 &HwasanThreadStartFunc, A);
   return res;
 }
-#endif
 
-#if HWASAN_WITH_INTERCEPTORS
 DEFINE_REAL(int, vfork)
 DECLARE_EXTERN_INTERCEPTOR_AND_WRAPPER(int, vfork)
-#endif
+#endif // HWASAN_WITH_INTERCEPTORS
 
 #if HWASAN_WITH_INTERCEPTORS && defined(__aarch64__)
 // Get and/or change the set of blocked signals.
@@ -331,9 +341,7 @@ void InitializeInterceptors() {
 #if defined(__linux__)
   INTERCEPT_FUNCTION(vfork);
 #endif  // __linux__
-#if !defined(__aarch64__)
   INTERCEPT_FUNCTION(pthread_create);
-#endif  // __aarch64__
 #endif
 
   inited = 1;

compiler-rt/lib/hwasan/hwasan_linux.cpp

@@ -354,12 +354,7 @@ void AndroidTestTlsSlot() {}
 #endif
 
 Thread *GetCurrentThread() {
-  uptr *ThreadLong = GetCurrentThreadLongPtr();
-#if HWASAN_WITH_INTERCEPTORS
-  if (!*ThreadLong)
-    __hwasan_thread_enter();
-#endif
-  auto *R = (StackAllocationsRingBuffer *)ThreadLong;
+  auto *R = (StackAllocationsRingBuffer *)GetCurrentThreadLongPtr();
   return hwasanThreadList().GetThreadByBufferAddress((uptr)(R->Next()));
 }
 

llvm/lib/Transforms/Instrumentation/HWAddressSanitizer.cpp

@@ -284,7 +284,6 @@ class HWAddressSanitizer {
 
   FunctionCallee HwasanTagMemoryFunc;
   FunctionCallee HwasanGenerateTagFunc;
-  FunctionCallee HwasanThreadEnterFunc;
 
   Constant *ShadowGlobal;
 
@@ -473,9 +472,6 @@ void HWAddressSanitizer::initializeCallbacks(Module &M) {
 
   HWAsanHandleVfork =
       M.getOrInsertFunction("__hwasan_handle_vfork", IRB.getVoidTy(), IntptrTy);
-
-  HwasanThreadEnterFunc =
-      M.getOrInsertFunction("__hwasan_thread_enter", IRB.getVoidTy());
 }
 
 Value *HWAddressSanitizer::getDynamicShadowIfunc(IRBuilder<> &IRB) {
@@ -934,34 +930,13 @@ void HWAddressSanitizer::emitPrologue(IRBuilder<> &IRB, bool WithFrameRecord) {
   Value *SlotPtr = getHwasanThreadSlotPtr(IRB, IntptrTy);
   assert(SlotPtr);
 
-  Instruction *ThreadLong = IRB.CreateLoad(IntptrTy, SlotPtr);
-
-  Function *F = IRB.GetInsertBlock()->getParent();
-  if (F->getFnAttribute("hwasan-abi").getValueAsString() == "interceptor") {
-    Value *ThreadLongEqZero =
-        IRB.CreateICmpEQ(ThreadLong, ConstantInt::get(IntptrTy, 0));
-    auto *Br = cast<BranchInst>(SplitBlockAndInsertIfThen(
-        ThreadLongEqZero, cast<Instruction>(ThreadLongEqZero)->getNextNode(),
-        false, MDBuilder(*C).createBranchWeights(1, 100000)));
-
-    IRB.SetInsertPoint(Br);
-    // FIXME: This should call a new runtime function with a custom calling
-    // convention to avoid needing to spill all arguments here.
-    IRB.CreateCall(HwasanThreadEnterFunc);
-    LoadInst *ReloadThreadLong = IRB.CreateLoad(IntptrTy, SlotPtr);
-
-    IRB.SetInsertPoint(&*Br->getSuccessor(0)->begin());
-    PHINode *ThreadLongPhi = IRB.CreatePHI(IntptrTy, 2);
-    ThreadLongPhi->addIncoming(ThreadLong, ThreadLong->getParent());
-    ThreadLongPhi->addIncoming(ReloadThreadLong, ReloadThreadLong->getParent());
-    ThreadLong = ThreadLongPhi;
-  }
-
+  Value *ThreadLong = IRB.CreateLoad(IntptrTy, SlotPtr);
   // Extract the address field from ThreadLong. Unnecessary on AArch64 with TBI.
   Value *ThreadLongMaybeUntagged =
       TargetTriple.isAArch64() ? ThreadLong : untagPointer(IRB, ThreadLong);
 
   if (WithFrameRecord) {
+    Function *F = IRB.GetInsertBlock()->getParent();
     StackBaseTag = IRB.CreateAShr(ThreadLong, 3);
 
     // Prepare ring buffer data.