Skip to content

TSan warning about isFiberRunning reading a stale pointer #122

Description

@vadimskipin

TSan reports a problem in one of our tests:

WARNING: ThreadSanitizer: data race (pid=40521)
  Write of size 8 at 0xff12bac3ee88 by thread T430:
    #0 operator delete(void*, unsigned long, std::align_val_t) <null> (amber-storage-bench+0x104748) (BuildId: 724e2958dfce8c8c5b51a12a902811a05da08c68)
    #1 std::default_delete<silk::Fiber>::operator()(silk::Fiber*) const /usr/lib/gcc/aarch64-linux-gnu/13/../../../../include/c++/13/bits/unique_ptr.h:99:2 (amber-storage-bench+0x18a804) (BuildId: 724e2958dfce8c8c5b51a12a902811a05da08c68)
    #2 std::unique_ptr<silk::Fiber, std::default_delete<silk::Fiber>>::~unique_ptr() /usr/lib/gcc/aarch64-linux-gnu/13/../../../../include/c++/13/bits/unique_ptr.h:404:4 (amber-storage-bench+0x18a804)
    #3 __call_tls_dtors stdlib/cxa_thread_atexit_impl.c:159:7 (libc.so.6+0x3d7d4) (BuildId: 4a81130bc30ded618f1edc030ecc5439e066dc68)

  Previous atomic read of size 1 at 0xff12bac3ee88 by thread T438:
    #0 std::atomic<silk::FiberState>::load(std::memory_order) const /usr/lib/gcc/aarch64-linux-gnu/13/../../../../include/c++/13/atomic:297:2 (amber-storage-bench+0x184160) (BuildId: 724e2958dfce8c8c5b51a12a902811a05da08c68)
    #1 silk::FiberScheduler::isFiberRunning(silk::Fiber*) contrib/silk/src/fibers/fiber.cpp:1444:25 (amber-storage-bench+0x184160)
    #2 silk::FiberMutex::lockSlow(silk::FiberMutex::State, unsigned long*) contrib/silk/src/fibers/mutex.cpp:40:42 (amber-storage-bench+0x18d6fc) (BuildId: 724e2958dfce8c8c5b51a12a902811a05da08c68)
    #3 silk::FiberMutex::lock(unsigned long*) contrib/silk/include/silk/fibers/mutex.h:60:9 (amber-storage-bench+0x10f7b4) (BuildId: 724e2958dfce8c8c5b51a12a902811a05da08c68)
    #4 std::unique_lock<silk::FiberMutex>::lock() /usr/lib/gcc/aarch64-linux-gnu/13/../../../../include/c++/13/bits/unique_lock.h:141:17 (amber-storage-bench+0x10f7b4)
    #5 std::unique_lock<silk::FiberMutex>::unique_lock(silk::FiberMutex&) /usr/lib/gcc/aarch64-linux-gnu/13/../../../../include/c++/13/bits/unique_lock.h:71:2 (amber-storage-bench+0x10f7b4)
    #6 amber::LogWriter::flush() src/storage/log-writer.cpp:430:26 (amber-storage-bench+0x10f7b4)
    #7 amber::LogStoreImpl::runWriter(amber::LogWriter*) src/storage/log-store-worker.cpp:262:25 (amber-storage-bench+0x11f05c) (BuildId: 724e2958dfce8c8c5b51a12a902811a05da08c68)
    #8 amber::LogStoreImpl::writerFiberMain(amber::LogStoreImpl::WriterFiberParams*) src/storage/log-store.h:301:78 (amber-storage-bench+0x10cc2c) (BuildId: 724e2958dfce8c8c5b51a12a902811a05da08c68)
    #9 silk::Fiber::fiberContextMain(transfer_t) contrib/silk/src/fibers/fiber.cpp:469:21 (amber-storage-bench+0x17ffc4) (BuildId: 724e2958dfce8c8c5b51a12a902811a05da08c68)
    #10 make_fcontext contrib/silk/contrib/fcontext/asm/make_arm64_aapcs_elf_gas.S:94 (amber-storage-bench+0x1b1184) (BuildId: 724e2958dfce8c8c5b51a12a902811a05da08c68)

  Thread T430 (tid=40638, running) created by main thread at:
    #0 pthread_create <null> (amber-storage-bench+0x74dac) (BuildId: 724e2958dfce8c8c5b51a12a902811a05da08c68)
    #1 std::thread::_M_start_thread(std::unique_ptr<std::thread::_State, std::default_delete<std::thread::_State>>, void (*)()) <null> (libstdc++.so.6+0xe8b60) (BuildId: dd1ec2f8de311a1d2a43412b9ad99b3a5e8aa0fe)
    #2 benchmark::internal::(anonymous namespace)::ThreadRunnerDefault::RunThreads(std::function<void (int)> const&) contrib/benchmark/src/benchmark_runner.cc:202:18 (amber-storage-bench+0x15bdcc) (BuildId: 724e2958dfce8c8c5b51a12a902811a05da08c68)
    #3 benchmark::internal::BenchmarkRunner::DoNIterations() contrib/benchmark/src/benchmark_runner.cc:335:18 (amber-storage-bench+0x159870) (BuildId: 724e2958dfce8c8c5b51a12a902811a05da08c68)
    #4 benchmark::internal::BenchmarkRunner::DoOneRepetition() contrib/benchmark/src/benchmark_runner.cc:509:9 (amber-storage-bench+0x15b250) (BuildId: 724e2958dfce8c8c5b51a12a902811a05da08c68)
    #5 benchmark::internal::(anonymous namespace)::RunBenchmarks(std::vector<benchmark::internal::BenchmarkInstance, std::allocator<benchmark::internal::BenchmarkInstance>> const&, benchmark::BenchmarkReporter*, benchmark::BenchmarkReporter*) contrib/benchmark/src/benchmark.cc:484:14 (amber-storage-bench+0x122e10) (BuildId: 724e2958dfce8c8c5b51a12a902811a05da08c68)
    #6 benchmark::RunSpecifiedBenchmarks(benchmark::BenchmarkReporter*, benchmark::BenchmarkReporter*, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>) contrib/benchmark/src/benchmark.cc:674:5 (amber-storage-bench+0x122e10)
    #7 benchmark::RunSpecifiedBenchmarks() contrib/benchmark/src/benchmark.cc:587:10 (amber-storage-bench+0x121f5c) (BuildId: 724e2958dfce8c8c5b51a12a902811a05da08c68)
    #8 main src/storage/benchmarks/main.cpp:14:5 (amber-storage-bench+0x106214) (BuildId: 724e2958dfce8c8c5b51a12a902811a05da08c68)

This looks like proxy-fiber was deleted and isFiberRunning uses a stale pointer. This does not happen with regular fibers because they are allocated from a memory pool.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions