[compiler-rt] [TSan][compiler-rt] Defer symbolization of Reports to as late as possible (PR #151120)

Dan Blackwell via llvm-commits llvm-commits at lists.llvm.org
Tue Jul 29 03:32:49 PDT 2025


https://github.com/DanBlackwell created https://github.com/llvm/llvm-project/pull/151120

This is the refactoring portion of: https://github.com/llvm/llvm-project/pull/150994. My aim is for this change to replicate current behaviour - just with Symbolization done explicitly (and later than previously).

>From 4a6ba6a392a3b2edbe70a885d4c87e2fa5a1a914 Mon Sep 17 00:00:00 2001
From: Dan Blackwell <dan_blackwell at apple.com>
Date: Fri, 18 Jul 2025 14:47:21 +0100
Subject: [PATCH] [TSan][compiler-rt] Defer symbolization of Reports to as late
 as possible

---
 .../lib/tsan/rtl/tsan_interceptors_posix.cpp  |  1 +
 .../lib/tsan/rtl/tsan_interface_ann.cpp       |  1 +
 compiler-rt/lib/tsan/rtl/tsan_mman.cpp        |  1 +
 compiler-rt/lib/tsan/rtl/tsan_report.h        | 13 +++
 compiler-rt/lib/tsan/rtl/tsan_rtl.h           |  1 +
 compiler-rt/lib/tsan/rtl/tsan_rtl_mutex.cpp   |  9 +-
 compiler-rt/lib/tsan/rtl/tsan_rtl_report.cpp  | 98 ++++++++++++++++---
 compiler-rt/lib/tsan/rtl/tsan_rtl_thread.cpp  |  1 +
 8 files changed, 108 insertions(+), 17 deletions(-)

diff --git a/compiler-rt/lib/tsan/rtl/tsan_interceptors_posix.cpp b/compiler-rt/lib/tsan/rtl/tsan_interceptors_posix.cpp
index 795e05394d71a..5d05789aaccdd 100644
--- a/compiler-rt/lib/tsan/rtl/tsan_interceptors_posix.cpp
+++ b/compiler-rt/lib/tsan/rtl/tsan_interceptors_posix.cpp
@@ -2146,6 +2146,7 @@ static void ReportErrnoSpoiling(ThreadState *thr, uptr pc, int sig) {
   rep.SetSigNum(sig);
   if (!IsFiredSuppression(ctx, ReportTypeErrnoInSignal, stack)) {
     rep.AddStack(stack, true);
+    rep.SymbolizeStackElems();
     OutputReport(thr, rep);
   }
 }
diff --git a/compiler-rt/lib/tsan/rtl/tsan_interface_ann.cpp b/compiler-rt/lib/tsan/rtl/tsan_interface_ann.cpp
index befd6a369026d..d16fd4cfb00d9 100644
--- a/compiler-rt/lib/tsan/rtl/tsan_interface_ann.cpp
+++ b/compiler-rt/lib/tsan/rtl/tsan_interface_ann.cpp
@@ -446,6 +446,7 @@ static void ReportMutexHeldWrongContext(ThreadState *thr, uptr pc) {
   VarSizeStackTrace trace;
   ObtainCurrentStack(thr, pc, &trace);
   rep.AddStack(trace, true);
+  rep.SymbolizeStackElems();
   OutputReport(thr, rep);
 }
 
diff --git a/compiler-rt/lib/tsan/rtl/tsan_mman.cpp b/compiler-rt/lib/tsan/rtl/tsan_mman.cpp
index 0ea83fb3b5982..3a0a6a1a4564f 100644
--- a/compiler-rt/lib/tsan/rtl/tsan_mman.cpp
+++ b/compiler-rt/lib/tsan/rtl/tsan_mman.cpp
@@ -185,6 +185,7 @@ static void SignalUnsafeCall(ThreadState *thr, uptr pc) {
   ThreadRegistryLock l(&ctx->thread_registry);
   ScopedReport rep(ReportTypeSignalUnsafe);
   rep.AddStack(stack, true);
+  rep.SymbolizeStackElems();
   OutputReport(thr, rep);
 }
 
diff --git a/compiler-rt/lib/tsan/rtl/tsan_report.h b/compiler-rt/lib/tsan/rtl/tsan_report.h
index 8975540ddfda2..53bb21964dbba 100644
--- a/compiler-rt/lib/tsan/rtl/tsan_report.h
+++ b/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;
 };
 
@@ -89,15 +93,23 @@ 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;
@@ -105,6 +117,7 @@ class ReportDesc {
   Vector<ReportStack*> stacks;
   Vector<ReportMop*> mops;
   Vector<ReportLocation*> locs;
+  Vector<AddedLocationAddr> added_location_addrs;
   Vector<ReportMutex*> mutexes;
   Vector<ReportThread*> threads;
   Vector<Tid> unique_tids;
diff --git a/compiler-rt/lib/tsan/rtl/tsan_rtl.h b/compiler-rt/lib/tsan/rtl/tsan_rtl.h
index 46276f20831d0..0d9d3bbf580a3 100644
--- a/compiler-rt/lib/tsan/rtl/tsan_rtl.h
+++ b/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;
 
diff --git a/compiler-rt/lib/tsan/rtl/tsan_rtl_mutex.cpp b/compiler-rt/lib/tsan/rtl/tsan_rtl_mutex.cpp
index 2a8aa1915c9ae..0027589939906 100644
--- a/compiler-rt/lib/tsan/rtl/tsan_rtl_mutex.cpp
+++ b/compiler-rt/lib/tsan/rtl/tsan_rtl_mutex.cpp
@@ -62,6 +62,7 @@ static void ReportMutexMisuse(ThreadState *thr, uptr pc, ReportType typ,
   ObtainCurrentStack(thr, pc, &trace);
   rep.AddStack(trace, true);
   rep.AddLocation(addr, 1);
+  rep.SymbolizeStackElems();
   OutputReport(thr, rep);
 }
 
@@ -539,15 +540,18 @@ void ReportDeadlock(ThreadState *thr, uptr pc, DDReport *r) {
   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) {
-        rep.AddStack(StackDepotGet(stk), true);
+        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.
-        rep.AddStack(StackTrace(&dummy_pc, 1), true);
+        stack = StackTrace(&dummy_pc, 1);
       }
+      rep.AddStack(stack, true);
     }
   }
+  rep.SymbolizeStackElems();
   OutputReport(thr, rep);
 }
 
@@ -572,6 +576,7 @@ void ReportDestroyLocked(ThreadState *thr, uptr pc, uptr addr,
     return;
   rep.AddStack(trace, true);
   rep.AddLocation(addr, 1);
+  rep.SymbolizeStackElems();
   OutputReport(thr, rep);
 }
 
diff --git a/compiler-rt/lib/tsan/rtl/tsan_rtl_report.cpp b/compiler-rt/lib/tsan/rtl/tsan_rtl_report.cpp
index 0820bf1adee43..f0848919694fa 100644
--- a/compiler-rt/lib/tsan/rtl/tsan_rtl_report.cpp
+++ b/compiler-rt/lib/tsan/rtl/tsan_rtl_report.cpp
@@ -11,6 +11,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "sanitizer_common/sanitizer_common.h"
+#include "sanitizer_common/sanitizer_internal_defs.h"
 #include "sanitizer_common/sanitizer_libc.h"
 #include "sanitizer_common/sanitizer_placement_new.h"
 #include "sanitizer_common/sanitizer_stackdepot.h"
@@ -187,10 +188,8 @@ void ScopedReportBase::AddMemoryAccess(uptr addr, uptr external_tag, Shadow s,
   mop->size = size;
   mop->write = !(typ & kAccessRead);
   mop->atomic = typ & kAccessAtomic;
-  mop->stack = SymbolizeStack(stack);
   mop->external_tag = external_tag;
-  if (mop->stack)
-    mop->stack->suppressable = true;
+  mop->stack_trace = stack;
   for (uptr i = 0; i < mset->Size(); i++) {
     MutexSet::Desc d = mset->Get(i);
     int id = this->AddMutex(d.addr, d.stack_id);
@@ -199,6 +198,80 @@ void ScopedReportBase::AddMemoryAccess(uptr addr, uptr external_tag, Shadow s,
   }
 }
 
+void ScopedReportBase::SymbolizeStackElems() {
+  // symbolize memory ops
+  for (usize i = 0; i < rep_->mops.Size(); i++) {
+    ReportMop *mop = rep_->mops[i];
+    mop->stack = SymbolizeStack(mop->stack_trace);
+    if (mop->stack)
+      mop->stack->suppressable = true;
+  }
+
+  Vector<ReportLocation *> locs_tmp;
+
+  // Repopulate the locations in the order they were added - take care with
+  // the added locations
+  usize locs_idx = 0;
+  for (usize i = 0; i < rep_->added_location_addrs.Size(); i++) {
+    AddedLocationAddr *added_loc_addr = &rep_->added_location_addrs[i];
+
+    for (; locs_idx < added_loc_addr->locs_idx; locs_idx++) {
+      ReportLocation *loc = rep_->locs[locs_idx];
+      loc->stack = SymbolizeStackId(loc->stack_id);
+      locs_tmp.PushBack(loc);
+    }
+
+    if (ReportLocation *added_loc = SymbolizeData(added_loc_addr->addr)) {
+      added_loc->suppressable = true;
+      locs_tmp.PushBack(added_loc);
+    }
+  }
+
+  // Append any remaining locations
+  for (; locs_idx < rep_->locs.Size(); locs_idx++) {
+    ReportLocation *loc = rep_->locs[locs_idx];
+    loc->stack = SymbolizeStackId(loc->stack_id);
+    locs_tmp.PushBack(loc);
+  }
+
+  rep_->locs.Reset();
+  for (usize i = 0; i < locs_tmp.Size(); i++) rep_->locs.PushBack(locs_tmp[i]);
+
+  // symbolize locations
+  for (usize i = 0; i < rep_->locs.Size(); i++) {
+    ReportLocation *loc = rep_->locs[i];
+    loc->stack = SymbolizeStackId(loc->stack_id);
+  }
+
+  usize offset = 0;
+  for (usize i = 0; i < rep_->added_location_addrs.Size(); i++) {
+    struct AddedLocationAddr *added_loc = &rep_->added_location_addrs[i];
+    if (ReportLocation *loc = SymbolizeData(added_loc->addr)) {
+      offset++;
+      loc->suppressable = true;
+      rep_->locs[added_loc->locs_idx] = loc;
+    }
+  }
+
+  // Check that all locs are populated (no-op in release)
+  for (usize i = 0; i < rep_->locs.Size(); i++)
+    CHECK_NE(rep_->locs[i], nullptr);
+
+  // symbolize threads
+  for (usize i = 0; i < rep_->threads.Size(); i++) {
+    ReportThread *rt = rep_->threads[i];
+    rt->stack = SymbolizeStackId(rt->stack_id);
+    if (rt->stack)
+      rt->stack->suppressable = rt->suppressable;
+  }
+
+  // symbolize mutexes
+  for (usize i = 0; i < rep_->mutexes.Size(); i++) {
+    ReportMutex *rm = rep_->mutexes[i];
+    rm->stack = SymbolizeStackId(rm->stack_id);
+  }
+}
+
 void ScopedReportBase::AddUniqueTid(Tid unique_tid) {
   rep_->unique_tids.PushBack(unique_tid);
 }
@@ -216,10 +289,8 @@ void ScopedReportBase::AddThread(const ThreadContext *tctx, bool suppressable) {
   rt->name = internal_strdup(tctx->name);
   rt->parent_tid = tctx->parent_tid;
   rt->thread_type = tctx->thread_type;
-  rt->stack = 0;
-  rt->stack = SymbolizeStackId(tctx->creation_stack_id);
-  if (rt->stack)
-    rt->stack->suppressable = suppressable;
+  rt->stack_id = tctx->creation_stack_id;
+  rt->suppressable = suppressable;
 }
 
 #if !SANITIZER_GO
@@ -270,7 +341,7 @@ int ScopedReportBase::AddMutex(uptr addr, StackID creation_stack_id) {
   rep_->mutexes.PushBack(rm);
   rm->id = rep_->mutexes.Size() - 1;
   rm->addr = addr;
-  rm->stack = SymbolizeStackId(creation_stack_id);
+  rm->stack_id = creation_stack_id;
   return rm->id;
 }
 
@@ -288,7 +359,7 @@ void ScopedReportBase::AddLocation(uptr addr, uptr size) {
     loc->fd_closed = closed;
     loc->fd = fd;
     loc->tid = creat_tid;
-    loc->stack = SymbolizeStackId(creat_stack);
+    loc->stack_id = creat_stack;
     rep_->locs.PushBack(loc);
     AddThread(creat_tid);
     return;
@@ -310,7 +381,7 @@ void ScopedReportBase::AddLocation(uptr addr, uptr size) {
     loc->heap_chunk_size = b->siz;
     loc->external_tag = b->tag;
     loc->tid = b->tid;
-    loc->stack = SymbolizeStackId(b->stk);
+    loc->stack_id = b->stk;
     rep_->locs.PushBack(loc);
     AddThread(b->tid);
     return;
@@ -324,11 +395,7 @@ void ScopedReportBase::AddLocation(uptr addr, uptr size) {
     AddThread(tctx);
   }
 #endif
-  if (ReportLocation *loc = SymbolizeData(addr)) {
-    loc->suppressable = true;
-    rep_->locs.PushBack(loc);
-    return;
-  }
+  rep_->added_location_addrs.PushBack({addr, rep_->locs.Size()});
 }
 
 #if !SANITIZER_GO
@@ -819,6 +886,7 @@ void ReportRace(ThreadState *thr, RawShadow *shadow_mem, Shadow cur, Shadow old,
       s[1].epoch() <= thr->last_sleep_clock.Get(s[1].sid()))
     rep.AddSleep(thr->last_sleep_stack_id);
 #endif
+  rep.SymbolizeStackElems();
   OutputReport(thr, rep);
 }
 
diff --git a/compiler-rt/lib/tsan/rtl/tsan_rtl_thread.cpp b/compiler-rt/lib/tsan/rtl/tsan_rtl_thread.cpp
index c6a8fd2acb6a8..36ee7af1fd973 100644
--- a/compiler-rt/lib/tsan/rtl/tsan_rtl_thread.cpp
+++ b/compiler-rt/lib/tsan/rtl/tsan_rtl_thread.cpp
@@ -96,6 +96,7 @@ void ThreadFinalize(ThreadState *thr) {
     ScopedReport rep(ReportTypeThreadLeak);
     rep.AddThread(leaks[i].tctx, true);
     rep.SetCount(leaks[i].count);
+    rep.SymbolizeStackElems();
     OutputReport(thr, rep);
   }
 #endif



More information about the llvm-commits mailing list