data race in BM_ThreadYieldSpinLockThrashing

[esigo]

Describe your environment Describe any aspect of your environment relevant to the problem,
g++ (Ubuntu 9.3.0-17ubuntu1~20.04) 9.3.0

Steps to reproduce
Describe exactly how to reproduce the error. Include a code sample if applicable.
bazel test --copt=-DENABLE_METRICS_PREVIEW --copt=-DENABLE_LOGS_PREVIEW --test_output=errors --cache_test_results=no --config=tsan //api/test/common:spinlock_benchmark_smoketest
What is the expected behavior?
no race
What did you expect to see?
no race
What is the actual behavior?
What did you see instead?
a data race
Additional context
Add any other context about the problem here.

[esigo]

call stack

WARNING: ThreadSanitizer: data race (pid=12)
  Read of size 8 at 0x7ffc1ed406a8 by thread T4:
    #0 (anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}::operator()() const <null> (spinlock_benchmark_smoketest+0x545e)
    #1 {lambda(std::atomic<bool>&)#2} std::__invoke_impl<void, (anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}>(std::__invoke_other, std::atomic<bool>&&) <null> (spinlock_benchmark_smoketest+0x8c35)
    #2 std::__invoke_result<{lambda(std::atomic<bool>&)#2}>::type std::__invoke<(anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}>({lambda(std::atomic<bool>&)#2}&&, ({lambda()#1}&&)...) <null> (spinlock_benchmark_smoketest+0x8a16)
    #3 void std::thread::_Invoker<std::tuple<(anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}> >::_M_invoke<0ul>(std::_Index_tuple<0ul>) <null> (spinlock_benchmark_smoketest+0x8862)
    #4 std::thread::_Invoker<std::tuple<(anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}> >::operator()() <null> (spinlock_benchmark_smoketest+0x872a)
    #5 std::thread::_State_impl<std::thread::_Invoker<std::tuple<(anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}> > >::_M_run() <null> (spinlock_benchmark_smoketest+0x85ea)
    #6 <null> <null> (libstdc++.so.6+0xd6de3)

  Previous write of size 8 at 0x7ffc1ed406a8 by thread T3:
    #0 (anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}::operator()() const <null> (spinlock_benchmark_smoketest+0x546d)
    #1 {lambda(std::atomic<bool>&)#2} std::__invoke_impl<void, (anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}>(std::__invoke_other, std::atomic<bool>&&) <null> (spinlock_benchmark_smoketest+0x8c35)
    #2 std::__invoke_result<{lambda(std::atomic<bool>&)#2}>::type std::__invoke<(anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}>({lambda(std::atomic<bool>&)#2}&&, ({lambda()#1}&&)...) <null> (spinlock_benchmark_smoketest+0x8a16)
    #3 void std::thread::_Invoker<std::tuple<(anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}> >::_M_invoke<0ul>(std::_Index_tuple<0ul>) <null> (spinlock_benchmark_smoketest+0x8862)
    #4 std::thread::_Invoker<std::tuple<(anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}> >::operator()() <null> (spinlock_benchmark_smoketest+0x872a)
    #5 std::thread::_State_impl<std::thread::_Invoker<std::tuple<(anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}> > >::_M_run() <null> (spinlock_benchmark_smoketest+0x85ea)
    #6 <null> <null> (libstdc++.so.6+0xd6de3)

  Location is stack of main thread.

  Location is global '<null>' at 0x000000000000 ([stack]+0x00000001f6a8)

  Thread T4 (tid=66, running) created by main thread at:
    #0 pthread_create ../../../../src/libsanitizer/tsan/tsan_interceptors_posix.cpp:962 (libtsan.so.0+0x5ea79)
    #1 std::thread::_M_start_thread(std::unique_ptr<std::thread::_State, std::default_delete<std::thread::_State> >, void (*)()) <null> (libstdc++.so.6+0xd70a8)
    #2 void __gnu_cxx::new_allocator<std::thread>::construct<std::thread, (anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}>({lambda(std::atomic<bool>&)#2}*, (anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}&&) <null> (spinlock_benchmark_smoketest+0x6f52)
    #3 void std::allocator_traits<std::allocator<std::thread> >::construct<std::thread, (anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}>(std::allocator<std::thread>&, {lambda(std::atomic<bool>&)#2}*, (anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}&&) <null> (spinlock_benchmark_smoketest+0x69b9)
    #4 void std::vector<std::thread, std::allocator<std::thread> >::emplace_back<(anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}>((anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}&&) <null> (spinlock_benchmark_smoketest+0x5cf6)
    #5 void (anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}) <null> (spinlock_benchmark_smoketest+0x57db)
    #6 (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&) <null> (spinlock_benchmark_smoketest+0x4367)
    #7 benchmark::internal::FunctionBenchmark::Run(benchmark::State&) <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x7520e)
    #8 benchmark::internal::BenchmarkInstance::Run(unsigned long, int, benchmark::internal::ThreadTimer*, benchmark::internal::ThreadManager*) const <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x70c3a)
    #9 benchmark::internal::(anonymous namespace)::RunInThread(benchmark::internal::BenchmarkInstance const*, unsigned long, int, benchmark::internal::ThreadManager*) <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x8a04d)
    #10 benchmark::internal::(anonymous namespace)::BenchmarkRunner::DoNIterations() <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x8ae76)
    #11 benchmark::internal::(anonymous namespace)::BenchmarkRunner::DoOneRepetition(long) <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x8ba89)
    #12 benchmark::internal::(anonymous namespace)::BenchmarkRunner::BenchmarkRunner(benchmark::internal::BenchmarkInstance const&, std::vector<benchmark::BenchmarkReporter::Run, std::allocator<benchmark::BenchmarkReporter::Run> >*) <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x8a90a)
    #13 benchmark::internal::RunBenchmark(benchmark::internal::BenchmarkInstance const&, std::vector<benchmark::BenchmarkReporter::Run, std::allocator<benchmark::BenchmarkReporter::Run> >*) <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x8c006)
    #14 benchmark::internal::(anonymous namespace)::RunBenchmarks(std::vector<benchmark::internal::BenchmarkInstance, std::allocator<benchmark::internal::BenchmarkInstance> > const&, benchmark::BenchmarkReporter*, benchmark::BenchmarkReporter*) <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x69f9b)
    #15 benchmark::RunSpecifiedBenchmarks(benchmark::BenchmarkReporter*, benchmark::BenchmarkReporter*) <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x6aa43)
    #16 benchmark::RunSpecifiedBenchmarks() <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x6a4ef)
    #17 main <null> (spinlock_benchmark_smoketest+0x43ec)

  Thread T3 (tid=65, running) created by main thread at:
    #0 pthread_create ../../../../src/libsanitizer/tsan/tsan_interceptors_posix.cpp:962 (libtsan.so.0+0x5ea79)
    #1 std::thread::_M_start_thread(std::unique_ptr<std::thread::_State, std::default_delete<std::thread::_State> >, void (*)()) <null> (libstdc++.so.6+0xd70a8)
    #2 void __gnu_cxx::new_allocator<std::thread>::construct<std::thread, (anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}>({lambda(std::atomic<bool>&)#2}*, (anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}&&) <null> (spinlock_benchmark_smoketest+0x6f52)
    #3 void std::allocator_traits<std::allocator<std::thread> >::construct<std::thread, (anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}>(std::allocator<std::thread>&, {lambda(std::atomic<bool>&)#2}*, (anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}&&) <null> (spinlock_benchmark_smoketest+0x69b9)
    #4 void std::vector<std::thread, std::allocator<std::thread> >::emplace_back<(anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}>((anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}&&) <null> (spinlock_benchmark_smoketest+0x5cf6)
    #5 void (anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}) <null> (spinlock_benchmark_smoketest+0x57db)
    #6 (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&) <null> (spinlock_benchmark_smoketest+0x4367)
    #7 benchmark::internal::FunctionBenchmark::Run(benchmark::State&) <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x7520e)
    #8 benchmark::internal::BenchmarkInstance::Run(unsigned long, int, benchmark::internal::ThreadTimer*, benchmark::internal::ThreadManager*) const <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x70c3a)
    #9 benchmark::internal::(anonymous namespace)::RunInThread(benchmark::internal::BenchmarkInstance const*, unsigned long, int, benchmark::internal::ThreadManager*) <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x8a04d)
    #10 benchmark::internal::(anonymous namespace)::BenchmarkRunner::DoNIterations() <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x8ae76)
    #11 benchmark::internal::(anonymous namespace)::BenchmarkRunner::DoOneRepetition(long) <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x8ba89)
    #12 benchmark::internal::(anonymous namespace)::BenchmarkRunner::BenchmarkRunner(benchmark::internal::BenchmarkInstance const&, std::vector<benchmark::BenchmarkReporter::Run, std::allocator<benchmark::BenchmarkReporter::Run> >*) <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x8a90a)
    #13 benchmark::internal::RunBenchmark(benchmark::internal::BenchmarkInstance const&, std::vector<benchmark::BenchmarkReporter::Run, std::allocator<benchmark::BenchmarkReporter::Run> >*) <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x8c006)
    #14 benchmark::internal::(anonymous namespace)::RunBenchmarks(std::vector<benchmark::internal::BenchmarkInstance, std::allocator<benchmark::internal::BenchmarkInstance> > const&, benchmark::BenchmarkReporter*, benchmark::BenchmarkReporter*) <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x69f9b)
    #15 benchmark::RunSpecifiedBenchmarks(benchmark::BenchmarkReporter*, benchmark::BenchmarkReporter*) <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x6aa43)
    #16 benchmark::RunSpecifiedBenchmarks() <null> (libexternal_Scom_Ugithub_Ugoogle_Ubenchmark_Slibbenchmark.so+0x6a4ef)
    #17 main <null> (spinlock_benchmark_smoketest+0x43ec)

SUMMARY: ThreadSanitizer: data race (/086043e746a94dd8b5c16310132ae034/execroot/io_opentelemetry_cpp/bazel-out/k8-tsan-fastbuild/bin/api/test/common/spinlock_benchmark_smoketest+0x545e) in (anonymous namespace)::SpinThrash<std::atomic<bool>, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2}>(benchmark::State&, std::atomic<bool>&, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#1}, (anonymous namespace)::BM_ThreadYieldSpinLockThrashing(benchmark::State&)::{lambda(std::atomic<bool>&)#2})::{lambda()#1}::operator()() const

[esigo]

Appeared in #1094 ci.
I could reproduce it locally.

[esigo]

happened again here.

[esigo]

The lock function lock(spinlock) used in the problematic test, will return after some limited amount of failed tries to atomically exchange the mutex with true. After that, the supposedly protected value will be modified with no protection.
To fix the issue, the lock function should return only and only if it actually acquired the lock. Or somehow let the caller know that the lock was not successful and modification of value should be aborted.

[ThomsonTan]

@esigo thanks for the explanation. For lock(spinlock), does it call lock(), which does never return if the lock is not required?

[esigo]

@ThomsonTan, no the lock with problem is the lambda. @lalitb had the idea that, BM_ThreadYieldSpinLockThrashing is probably meant to be compared with the actual spin lock implementation.