Merge pull request #11303 from DanBlackwell/cherry-pick-tsan-deadlock-fix-apple

🍒[TSan] Fix deadlocks during TSan error reporting on Apple platforms

Author: DanBlackwell

compiler-rt/lib/tsan/rtl/tsan_interceptors_posix.cpp

@@ -22,6 +22,7 @@
 #include "sanitizer_common/sanitizer_internal_defs.h"
 #include "sanitizer_common/sanitizer_libc.h"
 #include "sanitizer_common/sanitizer_linux.h"
+#include "sanitizer_common/sanitizer_placement_new.h"
 #include "sanitizer_common/sanitizer_platform_interceptors.h"
 #include "sanitizer_common/sanitizer_platform_limits_netbsd.h"
 #include "sanitizer_common/sanitizer_platform_limits_posix.h"
@@ -2141,13 +2142,29 @@ static void ReportErrnoSpoiling(ThreadState *thr, uptr pc, int sig) {
   // StackTrace::GetNestInstructionPc(pc) is used because return address is
   // expected, OutputReport() will undo this.
   ObtainCurrentStack(thr, StackTrace::GetNextInstructionPc(pc), &stack);
-  ThreadRegistryLock l(&ctx->thread_registry);
-  ScopedReport rep(ReportTypeErrnoInSignal);
-  rep.SetSigNum(sig);
-  if (!IsFiredSuppression(ctx, ReportTypeErrnoInSignal, stack)) {
-    rep.AddStack(stack, true);
-    OutputReport(thr, rep);
+  // Use alloca, because malloc during signal handling deadlocks
+  ScopedReport *rep = (ScopedReport *)__builtin_alloca(sizeof(ScopedReport));
+  bool suppressed;
+  // Take a new scope as Apple platforms require the below locks released
+  // before symbolizing in order to avoid a deadlock
+  {
+    ThreadRegistryLock l(&ctx->thread_registry);
+    new (rep) ScopedReport(ReportTypeErrnoInSignal);
+    rep->SetSigNum(sig);
+    suppressed = IsFiredSuppression(ctx, ReportTypeErrnoInSignal, stack);
+    if (!suppressed)
+      rep->AddStack(stack, true);
+#if SANITIZER_APPLE
+  }  // Close this scope to release the locks before writing report
+#endif
+    if (!suppressed)
+      OutputReport(thr, *rep);
+
+    // Need to manually destroy this because we used placement new to allocate
+    rep->~ScopedReport();
+#if !SANITIZER_APPLE
   }
+#endif
 }
 
 static void CallUserSignalHandler(ThreadState *thr, bool sync, bool acquire,

compiler-rt/lib/tsan/rtl/tsan_interface_ann.cpp

@@ -437,16 +437,30 @@ void __tsan_mutex_post_divert(void *addr, unsigned flagz) {
 }
 
 static void ReportMutexHeldWrongContext(ThreadState *thr, uptr pc) {
-  ThreadRegistryLock l(&ctx->thread_registry);
-  ScopedReport rep(ReportTypeMutexHeldWrongContext);
-  for (uptr i = 0; i < thr->mset.Size(); ++i) {
-    MutexSet::Desc desc = thr->mset.Get(i);
-    rep.AddMutex(desc.addr, desc.stack_id);
+  // Use alloca, because malloc during signal handling deadlocks
+  ScopedReport *rep = (ScopedReport *)__builtin_alloca(sizeof(ScopedReport));
+  // Take a new scope as Apple platforms require the below locks released
+  // before symbolizing in order to avoid a deadlock
+  {
+    ThreadRegistryLock l(&ctx->thread_registry);
+    new (rep) ScopedReport(ReportTypeMutexHeldWrongContext);
+    for (uptr i = 0; i < thr->mset.Size(); ++i) {
+      MutexSet::Desc desc = thr->mset.Get(i);
+      rep->AddMutex(desc.addr, desc.stack_id);
+    }
+    VarSizeStackTrace trace;
+    ObtainCurrentStack(thr, pc, &trace);
+    rep->AddStack(trace, true);
+#if SANITIZER_APPLE
+  }  // Close this scope to release the locks
+#endif
+    OutputReport(thr, *rep);
+
+    // Need to manually destroy this because we used placement new to allocate
+    rep->~ScopedReport();
+#if !SANITIZER_APPLE
   }
-  VarSizeStackTrace trace;
-  ObtainCurrentStack(thr, pc, &trace);
-  rep.AddStack(trace, true);
-  OutputReport(thr, rep);
+#endif
 }
 
 INTERFACE_ATTRIBUTE

compiler-rt/lib/tsan/rtl/tsan_mman.cpp

@@ -182,10 +182,24 @@ static void SignalUnsafeCall(ThreadState *thr, uptr pc) {
   ObtainCurrentStack(thr, pc, &stack);
   if (IsFiredSuppression(ctx, ReportTypeSignalUnsafe, stack))
     return;
-  ThreadRegistryLock l(&ctx->thread_registry);
-  ScopedReport rep(ReportTypeSignalUnsafe);
-  rep.AddStack(stack, true);
-  OutputReport(thr, rep);
+  // Use alloca, because malloc during signal handling deadlocks
+  ScopedReport *rep = (ScopedReport *)__builtin_alloca(sizeof(ScopedReport));
+  // Take a new scope as Apple platforms require the below locks released
+  // before symbolizing in order to avoid a deadlock
+  {
+    ThreadRegistryLock l(&ctx->thread_registry);
+    new (rep) ScopedReport(ReportTypeSignalUnsafe);
+    rep->AddStack(stack, true);
+#if SANITIZER_APPLE
+  }  // Close this scope to release the locks
+#endif
+    OutputReport(thr, *rep);
+
+    // Need to manually destroy this because we used placement new to allocate
+    rep->~ScopedReport();
+#if !SANITIZER_APPLE
+  }
+#endif
 }
 
 

compiler-rt/lib/tsan/rtl/tsan_report.h

@@ -12,6 +12,8 @@
 #ifndef TSAN_REPORT_H
 #define TSAN_REPORT_H
 
+#include "sanitizer_common/sanitizer_internal_defs.h"
+#include "sanitizer_common/sanitizer_stacktrace.h"
 #include "sanitizer_common/sanitizer_symbolizer.h"
 #include "sanitizer_common/sanitizer_thread_registry.h"
 #include "sanitizer_common/sanitizer_vector.h"
@@ -56,6 +58,7 @@ struct ReportMop {
   bool atomic;
   uptr external_tag;
   Vector<ReportMopMutex> mset;
+  StackTrace stack_trace;
   ReportStack *stack;
 
   ReportMop();
@@ -79,6 +82,7 @@ struct ReportLocation {
   int fd = 0;
   bool fd_closed = false;
   bool suppressable = false;
+  StackID stack_id = 0;
   ReportStack *stack = nullptr;
 };
 
@@ -89,22 +93,31 @@ struct ReportThread {
   ThreadType thread_type;
   char *name;
   Tid parent_tid;
+  StackID stack_id;
   ReportStack *stack;
+  bool suppressable;
 };
 
 struct ReportMutex {
   int id;
   uptr addr;
+  StackID stack_id;
   ReportStack *stack;
 };
 
+struct AddedLocationAddr {
+  uptr addr;
+  usize locs_idx;
+};
+
 class ReportDesc {
  public:
   ReportType typ;
   uptr tag;
   Vector<ReportStack*> stacks;
   Vector<ReportMop*> mops;
   Vector<ReportLocation*> locs;
+  Vector<AddedLocationAddr> added_location_addrs;
   Vector<ReportMutex*> mutexes;
   Vector<ReportThread*> threads;
   Vector<Tid> unique_tids;

compiler-rt/lib/tsan/rtl/tsan_rtl.h

@@ -420,6 +420,7 @@ class ScopedReportBase {
   void AddSleep(StackID stack_id);
   void SetCount(int count);
   void SetSigNum(int sig);
+  void SymbolizeStackElems(void);
 
   const ReportDesc *GetReport() const;
 
@@ -498,7 +499,7 @@ void ForkChildAfter(ThreadState *thr, uptr pc, bool start_thread);
 
 void ReportRace(ThreadState *thr, RawShadow *shadow_mem, Shadow cur, Shadow old,
                 AccessType typ);
-bool OutputReport(ThreadState *thr, const ScopedReport &srep);
+bool OutputReport(ThreadState *thr, ScopedReport &srep);
 bool IsFiredSuppression(Context *ctx, ReportType type, StackTrace trace);
 bool IsExpectedReport(uptr addr, uptr size);
 

compiler-rt/lib/tsan/rtl/tsan_rtl_access.cpp

@@ -419,6 +419,11 @@ NOINLINE void TraceRestartMemoryAccess(ThreadState* thr, uptr pc, uptr addr,
 
 ALWAYS_INLINE USED void MemoryAccess(ThreadState* thr, uptr pc, uptr addr,
                                      uptr size, AccessType typ) {
+#if SANITIZER_APPLE && !SANITIZER_GO
+  // Swift symbolizer can be intercepted and deadlock without this
+  if (thr->in_symbolizer)
+    return;
+#endif
   RawShadow* shadow_mem = MemToShadow(addr);
   UNUSED char memBuf[4][64];
   DPrintf2("#%d: Access: %d@%d %p/%zd typ=0x%x {%s, %s, %s, %s}\n", thr->tid,

compiler-rt/lib/tsan/rtl/tsan_rtl_mutex.cpp

@@ -11,14 +11,15 @@
 //===----------------------------------------------------------------------===//
 
 #include <sanitizer_common/sanitizer_deadlock_detector_interface.h>
+#include <sanitizer_common/sanitizer_placement_new.h>
 #include <sanitizer_common/sanitizer_stackdepot.h>
 
-#include "tsan_rtl.h"
 #include "tsan_flags.h"
-#include "tsan_sync.h"
+#include "tsan_platform.h"
 #include "tsan_report.h"
+#include "tsan_rtl.h"
 #include "tsan_symbolize.h"
-#include "tsan_platform.h"
+#include "tsan_sync.h"
 
 namespace __tsan {
 
@@ -55,14 +56,28 @@ static void ReportMutexMisuse(ThreadState *thr, uptr pc, ReportType typ,
     return;
   if (!ShouldReport(thr, typ))
     return;
-  ThreadRegistryLock l(&ctx->thread_registry);
-  ScopedReport rep(typ);
-  rep.AddMutex(addr, creation_stack_id);
-  VarSizeStackTrace trace;
-  ObtainCurrentStack(thr, pc, &trace);
-  rep.AddStack(trace, true);
-  rep.AddLocation(addr, 1);
-  OutputReport(thr, rep);
+  // Use alloca, because malloc during signal handling deadlocks
+  ScopedReport *rep = (ScopedReport *)__builtin_alloca(sizeof(ScopedReport));
+  // Take a new scope as Apple platforms require the below locks released
+  // before symbolizing in order to avoid a deadlock
+  {
+    ThreadRegistryLock l(&ctx->thread_registry);
+    new (rep) ScopedReport(typ);
+    rep->AddMutex(addr, creation_stack_id);
+    VarSizeStackTrace trace;
+    ObtainCurrentStack(thr, pc, &trace);
+    rep->AddStack(trace, true);
+    rep->AddLocation(addr, 1);
+#if SANITIZER_APPLE
+  }  // Close this scope to release the locks
+#endif
+    OutputReport(thr, *rep);
+
+    // Need to manually destroy this because we used placement new to allocate
+    rep->~ScopedReport();
+#if !SANITIZER_APPLE
+  }
+#endif
 }
 
 static void RecordMutexLock(ThreadState *thr, uptr pc, uptr addr,
@@ -528,51 +543,81 @@ void AfterSleep(ThreadState *thr, uptr pc) {
 void ReportDeadlock(ThreadState *thr, uptr pc, DDReport *r) {
   if (r == 0 || !ShouldReport(thr, ReportTypeDeadlock))
     return;
-  ThreadRegistryLock l(&ctx->thread_registry);
-  ScopedReport rep(ReportTypeDeadlock);
-  for (int i = 0; i < r->n; i++) {
-    rep.AddMutex(r->loop[i].mtx_ctx0, r->loop[i].stk[0]);
-    rep.AddUniqueTid((int)r->loop[i].thr_ctx);
-    rep.AddThread((int)r->loop[i].thr_ctx);
-  }
-  uptr dummy_pc = 0x42;
-  for (int i = 0; i < r->n; i++) {
-    for (int j = 0; j < (flags()->second_deadlock_stack ? 2 : 1); j++) {
-      u32 stk = r->loop[i].stk[j];
-      if (stk && stk != kInvalidStackID) {
-        rep.AddStack(StackDepotGet(stk), true);
-      } else {
-        // Sometimes we fail to extract the stack trace (FIXME: investigate),
-        // but we should still produce some stack trace in the report.
-        rep.AddStack(StackTrace(&dummy_pc, 1), true);
+  // Use alloca, because malloc during signal handling deadlocks
+  ScopedReport *rep = (ScopedReport *)__builtin_alloca(sizeof(ScopedReport));
+  // Take a new scope as Apple platforms require the below locks released
+  // before symbolizing in order to avoid a deadlock
+  {
+    ThreadRegistryLock l(&ctx->thread_registry);
+    new (rep) ScopedReport(ReportTypeDeadlock);
+    for (int i = 0; i < r->n; i++) {
+      rep->AddMutex(r->loop[i].mtx_ctx0, r->loop[i].stk[0]);
+      rep->AddUniqueTid((int)r->loop[i].thr_ctx);
+      rep->AddThread((int)r->loop[i].thr_ctx);
+    }
+    uptr dummy_pc = 0x42;
+    for (int i = 0; i < r->n; i++) {
+      for (int j = 0; j < (flags()->second_deadlock_stack ? 2 : 1); j++) {
+        u32 stk = r->loop[i].stk[j];
+        StackTrace stack;
+        if (stk && stk != kInvalidStackID) {
+          stack = StackDepotGet(stk);
+        } else {
+          // Sometimes we fail to extract the stack trace (FIXME: investigate),
+          // but we should still produce some stack trace in the report.
+          stack = StackTrace(&dummy_pc, 1);
+        }
+        rep->AddStack(stack, true);
       }
     }
+#if SANITIZER_APPLE
+  }  // Close this scope to release the locks
+#endif
+    OutputReport(thr, *rep);
+
+    // Need to manually destroy this because we used placement new to allocate
+    rep->~ScopedReport();
+#if !SANITIZER_APPLE
   }
-  OutputReport(thr, rep);
+#endif
 }
 
 void ReportDestroyLocked(ThreadState *thr, uptr pc, uptr addr,
                          FastState last_lock, StackID creation_stack_id) {
-  // We need to lock the slot during RestoreStack because it protects
-  // the slot journal.
-  Lock slot_lock(&ctx->slots[static_cast<uptr>(last_lock.sid())].mtx);
-  ThreadRegistryLock l0(&ctx->thread_registry);
-  Lock slots_lock(&ctx->slot_mtx);
-  ScopedReport rep(ReportTypeMutexDestroyLocked);
-  rep.AddMutex(addr, creation_stack_id);
-  VarSizeStackTrace trace;
-  ObtainCurrentStack(thr, pc, &trace);
-  rep.AddStack(trace, true);
-
-  Tid tid;
-  DynamicMutexSet mset;
-  uptr tag;
-  if (!RestoreStack(EventType::kLock, last_lock.sid(), last_lock.epoch(), addr,
-                    0, kAccessWrite, &tid, &trace, mset, &tag))
-    return;
-  rep.AddStack(trace, true);
-  rep.AddLocation(addr, 1);
-  OutputReport(thr, rep);
+  // Use alloca, because malloc during signal handling deadlocks
+  ScopedReport *rep = (ScopedReport *)__builtin_alloca(sizeof(ScopedReport));
+  // Take a new scope as Apple platforms require the below locks released
+  // before symbolizing in order to avoid a deadlock
+  {
+    // We need to lock the slot during RestoreStack because it protects
+    // the slot journal.
+    Lock slot_lock(&ctx->slots[static_cast<uptr>(last_lock.sid())].mtx);
+    ThreadRegistryLock l0(&ctx->thread_registry);
+    Lock slots_lock(&ctx->slot_mtx);
+    new (rep) ScopedReport(ReportTypeMutexDestroyLocked);
+    rep->AddMutex(addr, creation_stack_id);
+    VarSizeStackTrace trace;
+    ObtainCurrentStack(thr, pc, &trace);
+    rep->AddStack(trace, true);
+
+    Tid tid;
+    DynamicMutexSet mset;
+    uptr tag;
+    if (!RestoreStack(EventType::kLock, last_lock.sid(), last_lock.epoch(),
+                      addr, 0, kAccessWrite, &tid, &trace, mset, &tag))
+      return;
+    rep->AddStack(trace, true);
+    rep->AddLocation(addr, 1);
+#if SANITIZER_APPLE
+  }  // Close this scope to release the locks
+#endif
+    OutputReport(thr, *rep);
+
+    // Need to manually destroy this because we used placement new to allocate
+    rep->~ScopedReport();
+#if !SANITIZER_APPLE
+  }
+#endif
 }
 
 }  // namespace __tsan