rocksdb/utilities
Hui Xiao b16655a547 Add missing synchronization in TestFSWritableFile (#10544)
Summary:
**Context:**
ajkr's command revealed an existing TSAN data race between `TestFSWritableFile::Append` and `TestFSWritableFile::Sync` on `TestFSWritableFile::state_`

```
$ make clean && COMPILE_WITH_TSAN=1 make -j56 db_stress
$ python3 tools/db_crashtest.py blackbox --simple --duration=3600 --interval=10 --sync_fault_injection=1 --disable_wal=0 --max_key=10000 --checkpoint_one_in=1000
```

The race is due to concurrent access from [checkpoint's WAL sync](https://github.com/facebook/rocksdb/blob/7.4.fb/utilities/fault_injection_fs.cc#L324) and [db put's WAL write when ‘sync_fault_injection=1 ‘](https://github.com/facebook/rocksdb/blob/7.4.fb/utilities/fault_injection_fs.cc#L208) to the `state_` on the same WAL `TestFSWritableFile` under the missing synchronization.

```
WARNING: ThreadSanitizer: data race (pid=11275)
Write of size 8 at 0x7b480003d850 by thread T23 (mutexes: write M69230):
#0 rocksdb::TestFSWritableFile::Sync(rocksdb::IOOptions const&, rocksdb::IODebugContext*) internal_repo_rocksdb/repo/utilities/fault_injection_fs.cc:297 (db_stress+0x716004)
https://github.com/facebook/rocksdb/issues/1 rocksdb::(anonymous namespace)::CompositeWritableFileWrapper::Sync() internal_repo_rocksdb/repo/env/composite_env.cc:154 (db_stress+0x4dfa78)
https://github.com/facebook/rocksdb/issues/2 rocksdb::(anonymous namespace)::LegacyWritableFileWrapper::Sync(rocksdb::IOOptions const&, rocksdb::IODebugContext*) internal_repo_rocksdb/repo/env/env.cc:280 (db_stress+0x6dfd24)
https://github.com/facebook/rocksdb/issues/3 rocksdb::WritableFileWriter::SyncInternal(bool) internal_repo_rocksdb/repo/file/writable_file_writer.cc:460 (db_stress+0xa1b98c)
https://github.com/facebook/rocksdb/issues/4 rocksdb::WritableFileWriter::SyncWithoutFlush(bool) internal_repo_rocksdb/repo/file/writable_file_writer.cc:435 (db_stress+0xa1e441)
https://github.com/facebook/rocksdb/issues/5 rocksdb::DBImpl::SyncWAL() internal_repo_rocksdb/repo/db/db_impl/db_impl.cc:1385 (db_stress+0x529458)
https://github.com/facebook/rocksdb/issues/6 rocksdb::DBImpl::FlushWAL(bool) internal_repo_rocksdb/repo/db/db_impl/db_impl.cc:1339 (db_stress+0x54f82a)
https://github.com/facebook/rocksdb/issues/7 rocksdb::DBImpl::GetLiveFilesStorageInfo(rocksdb::LiveFilesStorageInfoOptions const&, std::vector<rocksdb::LiveFileStorageInfo, std::allocator<rocksdb::LiveFileStorageInfo> >*) internal_repo_rocksdb/repo/db/db_filesnapshot.cc:387 (db_stress+0x5c831d)
https://github.com/facebook/rocksdb/issues/8 rocksdb::CheckpointImpl::CreateCustomCheckpoint(std::function<rocksdb::Status (std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, rocksdb::FileType)>, std::function<rocksdb::Status (std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, unsigned long, rocksdb::FileType, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, rocksdb::Temperature)>, std::function<rocksdb::Status (std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, rocksdb::FileType)>, unsigned long*, unsigned long, bool) internal_repo_rocksdb/repo/utilities/checkpoint/checkpoint_impl.cc:214 (db_stress+0x4c0343)
https://github.com/facebook/rocksdb/issues/9 rocksdb::CheckpointImpl::CreateCheckpoint(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, unsigned long, unsigned long*) internal_repo_rocksdb/repo/utilities/checkpoint/checkpoint_impl.cc:123 (db_stress+0x4c237e)
https://github.com/facebook/rocksdb/issues/10 rocksdb::StressTest::TestCheckpoint(rocksdb::ThreadState*, std::vector<int, std::allocator<int> > const&, std::vector<long, std::allocator<long> > const&) internal_repo_rocksdb/repo/db_stress_tool/db_stress_test_base.cc:1699 (db_stress+0x328340)
https://github.com/facebook/rocksdb/issues/11 rocksdb::StressTest::OperateDb(rocksdb::ThreadState*) internal_repo_rocksdb/repo/db_stress_tool/db_stress_test_base.cc:825 (db_stress+0x33921f)
https://github.com/facebook/rocksdb/issues/12 rocksdb::ThreadBody(void*) internal_repo_rocksdb/repo/db_stress_tool/db_stress_driver.cc:33 (db_stress+0x354857)
https://github.com/facebook/rocksdb/issues/13 rocksdb::(anonymous namespace)::StartThreadWrapper(void*) internal_repo_rocksdb/repo/env/env_posix.cc:447 (db_stress+0x6eb2ad)

Previous read of size 8 at 0x7b480003d850 by thread T64 (mutexes: write M980798978697532600, write M253744503184415024, write M1262):
#0 memcpy <null> (db_stress+0xbc9696)
https://github.com/facebook/rocksdb/issues/1 operator= internal_repo_rocksdb/repo/utilities/fault_injection_fs.h:35 (db_stress+0x70d5f1)
https://github.com/facebook/rocksdb/issues/2 rocksdb::FaultInjectionTestFS::WritableFileAppended(rocksdb::FSFileState const&) internal_repo_rocksdb/repo/utilities/fault_injection_fs.cc:827 (db_stress+0x70d5f1)
https://github.com/facebook/rocksdb/issues/3 rocksdb::TestFSWritableFile::Append(rocksdb::Slice const&, rocksdb::IOOptions const&, rocksdb::IODebugContext*) internal_repo_rocksdb/repo/utilities/fault_injection_fs.cc:173 (db_stress+0x7143af)
https://github.com/facebook/rocksdb/issues/4 rocksdb::(anonymous namespace)::CompositeWritableFileWrapper::Append(rocksdb::Slice const&) internal_repo_rocksdb/repo/env/composite_env.cc:115 (db_stress+0x4de3ab)
https://github.com/facebook/rocksdb/issues/5 rocksdb::(anonymous namespace)::LegacyWritableFileWrapper::Append(rocksdb::Slice const&, rocksdb::IOOptions const&, rocksdb::IODebugContext*) internal_repo_rocksdb/repo/env/env.cc:248 (db_stress+0x6df44b)
https://github.com/facebook/rocksdb/issues/6 rocksdb::WritableFileWriter::WriteBuffered(char const*, unsigned long, rocksdb::Env::IOPriority) internal_repo_rocksdb/repo/file/writable_file_writer.cc:551 (db_stress+0xa1a953)
https://github.com/facebook/rocksdb/issues/7 rocksdb::WritableFileWriter::Flush(rocksdb::Env::IOPriority) internal_repo_rocksdb/repo/file/writable_file_writer.cc:327 (db_stress+0xa16ee8)
https://github.com/facebook/rocksdb/issues/8 rocksdb::log::Writer::AddRecord(rocksdb::Slice const&, rocksdb::Env::IOPriority) internal_repo_rocksdb/repo/db/log_writer.cc:147 (db_stress+0x7f121f)
https://github.com/facebook/rocksdb/issues/9 rocksdb::DBImpl::WriteToWAL(rocksdb::WriteBatch const&, rocksdb::log::Writer*, unsigned long*, unsigned long*, rocksdb::Env::IOPriority, rocksdb::DBImpl::LogFileNumberSize&) internal_repo_rocksdb/repo/db/db_impl/db_impl_write.cc:1285 (db_stress+0x695042)
https://github.com/facebook/rocksdb/issues/10 rocksdb::DBImpl::WriteToWAL(rocksdb::WriteThread::WriteGroup const&, rocksdb::log::Writer*, unsigned long*, bool, bool, unsigned long, rocksdb::DBImpl::LogFileNumberSize&) internal_repo_rocksdb/repo/db/db_impl/db_impl_write.cc:1328 (db_stress+0x6907e8)
https://github.com/facebook/rocksdb/issues/11 rocksdb::DBImpl::PipelinedWriteImpl(rocksdb::WriteOptions const&, rocksdb::WriteBatch*, rocksdb::WriteCallback*, unsigned long*, unsigned long, bool, unsigned long*) internal_repo_rocksdb/repo/db/db_impl/db_impl_write.cc:731 (db_stress+0x68e8a7)
https://github.com/facebook/rocksdb/issues/12 rocksdb::DBImpl::WriteImpl(rocksdb::WriteOptions const&, rocksdb::WriteBatch*, rocksdb::WriteCallback*, unsigned long*, unsigned long, bool, unsigned long*, unsigned long, rocksdb::PreReleaseCallback*, rocksdb::PostMemTableCallback*) internal_repo_rocksdb/repo/db/db_impl/db_impl_write.cc:283 (db_stress+0x688370)
https://github.com/facebook/rocksdb/issues/13 rocksdb::DBImpl::Write(rocksdb::WriteOptions const&, rocksdb::WriteBatch*) internal_repo_rocksdb/repo/db/db_impl/db_impl_write.cc:126 (db_stress+0x69a7b5)
https://github.com/facebook/rocksdb/issues/14 rocksdb::DB::Put(rocksdb::WriteOptions const&, rocksdb::ColumnFamilyHandle*, rocksdb::Slice const&, rocksdb::Slice const&, rocksdb::Slice const&) internal_repo_rocksdb/repo/db/db_impl/db_impl_write.cc:2247 (db_stress+0x698634)
https://github.com/facebook/rocksdb/issues/15 rocksdb::DBImpl::Put(rocksdb::WriteOptions const&, rocksdb::ColumnFamilyHandle*, rocksdb::Slice const&, rocksdb::Slice const&, rocksdb::Slice const&) internal_repo_rocksdb/repo/db/db_impl/db_impl_write.cc:37 (db_stress+0x699868)
https://github.com/facebook/rocksdb/issues/16 rocksdb::NonBatchedOpsStressTest::TestPut(rocksdb::ThreadState*, rocksdb::WriteOptions&, rocksdb::ReadOptions const&, std::vector<int, std::allocator<int> > const&, std::vector<long, std::allocator<long> > const&, char (&) [100], std::unique_ptr<rocksdb::MutexLock, std::default_delete<rocksdb::MutexLock> >&) internal_repo_rocksdb/repo/db_stress_tool/no_batched_ops_stress.cc:681 (db_stress+0x38d20c)
https://github.com/facebook/rocksdb/issues/17 rocksdb::StressTest::OperateDb(rocksdb::ThreadState*) internal_repo_rocksdb/repo/db_stress_tool/db_stress_test_base.cc:897 (db_stress+0x3399ec)
https://github.com/facebook/rocksdb/issues/18 rocksdb::ThreadBody(void*) internal_repo_rocksdb/repo/db_stress_tool/db_stress_driver.cc:33 (db_stress+0x354857)
https://github.com/facebook/rocksdb/issues/19 rocksdb::(anonymous namespace)::StartThreadWrapper(void*) internal_repo_rocksdb/repo/env/env_posix.cc:447 (db_stress+0x6eb2ad)

Location is heap block of size 352 at 0x7b480003d800 allocated by thread T23:
#0 operator new(unsigned long) <null> (db_stress+0xb685dc)
https://github.com/facebook/rocksdb/issues/1 rocksdb::FaultInjectionTestFS::NewWritableFile(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, rocksdb::FileOptions const&, std::unique_ptr<rocksdb::FSWritableFile, std::default_delete<rocksdb::FSWritableFile> >*, rocksdb::IODebugContext*) internal_repo_rocksdb/repo/utilities/fault_injection_fs.cc:506 (db_stress+0x711192)
https://github.com/facebook/rocksdb/issues/2 rocksdb::CompositeEnv::NewWritableFile(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unique_ptr<rocksdb::WritableFile, std::default_delete<rocksdb::WritableFile> >*, rocksdb::EnvOptions const&) internal_repo_rocksdb/repo/env/composite_env.cc:329 (db_stress+0x4d33fa)
https://github.com/facebook/rocksdb/issues/3 rocksdb::EnvWrapper::NewWritableFile(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unique_ptr<rocksdb::WritableFile, std::default_delete<rocksdb::WritableFile> >*, rocksdb::EnvOptions const&) internal_repo_rocksdb/repo/include/rocksdb/env.h:1425 (db_stress+0x300662)
...
```

**Summary:**
- Added the missing lock in functions mentioned above along with three other functions with a similar need in TestFSWritableFile
- Added clarification comment

Pull Request resolved: https://github.com/facebook/rocksdb/pull/10544

Test Plan: - Past the above race condition repro

Reviewed By: ajkr

Differential Revision: D38886634

Pulled By: hx235

fbshipit-source-id: 0571bae9615f35b16fbd8168204607e306b1b486
2022-08-22 15:50:22 -07:00
..
agg_merge Add Aggregation Merge Operator (#9780) 2022-04-15 23:24:05 -07:00
backup Migrate to docker for CI run (#10496) 2022-08-10 17:34:38 -07:00
blob_db Remove own ToString() (#9955) 2022-05-06 13:03:58 -07:00
cassandra Migrate to docker for CI run (#10496) 2022-08-10 17:34:38 -07:00
checkpoint Fix race condition with WAL tracking and `FlushWAL(true /* sync */)` (#10185) 2022-06-17 16:45:28 -07:00
compaction_filters Make MergeOperator+CompactionFilter/Factory into Customizable Classes (#8481) 2021-08-06 08:27:25 -07:00
convenience
leveldb_options
memory Remove own ToString() (#9955) 2022-05-06 13:03:58 -07:00
merge_operators Remove using namespace (#9369) 2022-01-12 09:31:12 -08:00
option_change_migration Fix some typos in comments and HISTORY.md (#9798) 2022-04-04 09:32:57 -07:00
options Remove own ToString() (#9955) 2022-05-06 13:03:58 -07:00
persistent_cache Remove code that only compiles for Visual Studio versions older than 2015 (#10065) 2022-05-26 16:55:08 -07:00
simulator_cache Have Cache use Status::MemoryLimit (#10262) 2022-07-06 14:41:46 -07:00
table_properties_collectors Remove own ToString() (#9955) 2022-05-06 13:03:58 -07:00
trace Add rate limiter priority to ReadOptions (#9424) 2022-02-16 23:18:14 -08:00
transactions Fix two extra headers (#10525) 2022-08-12 15:54:35 -07:00
ttl Migrate to docker for CI run (#10496) 2022-08-10 17:34:38 -07:00
write_batch_with_index Migrate to docker for CI run (#10496) 2022-08-10 17:34:38 -07:00
cache_dump_load.cc Introduce a mechanism to dump out blocks from block cache and re-insert to secondary cache (#8912) 2021-10-07 11:42:31 -07:00
cache_dump_load_impl.cc Derive cache keys from SST unique IDs (#10394) 2022-08-12 13:49:49 -07:00
cache_dump_load_impl.h Remove deprecated block-based filter (#10184) 2022-06-16 15:51:33 -07:00
compaction_filters.cc Restore Regex support for ObjectLibrary::Register, rename new APIs to allow old one to be deprecated in the future (#9362) 2022-01-11 06:33:48 -08:00
counted_fs.cc Fix serious FSDirectory use-after-Close bug (missing fsync) (#10460) 2022-08-02 10:54:32 -07:00
counted_fs.h Explicitly closing all directory file descriptors (#10049) 2022-06-01 18:03:34 -07:00
debug.cc Add helper function to get debug type name (#10243) 2022-07-15 14:42:00 -07:00
env_mirror.cc Fix clang13 build error (#9374) 2022-01-11 10:36:22 -08:00
env_mirror_test.cc
env_timed.cc Make FileSystem a Customizable Class (#8649) 2021-11-02 09:07:11 -07:00
env_timed.h Make FileSystem a Customizable Class (#8649) 2021-11-02 09:07:11 -07:00
env_timed_test.cc
fault_injection_env.cc Explicitly closing all directory file descriptors (#10049) 2022-06-01 18:03:34 -07:00
fault_injection_env.h Explicitly closing all directory file descriptors (#10049) 2022-06-01 18:03:34 -07:00
fault_injection_fs.cc Add missing synchronization in TestFSWritableFile (#10544) 2022-08-22 15:50:22 -07:00
fault_injection_fs.h Add missing synchronization in TestFSWritableFile (#10544) 2022-08-22 15:50:22 -07:00
fault_injection_secondary_cache.cc Fix the memory leak in db_stress tests that are caused by `FaultInjectionSecondaryCache` and add `CompressedSecondaryCache` into stress tests. (#10523) 2022-08-18 21:53:27 -07:00
fault_injection_secondary_cache.h Fix the memory leak in db_stress tests that are caused by `FaultInjectionSecondaryCache` and add `CompressedSecondaryCache` into stress tests. (#10523) 2022-08-18 21:53:27 -07:00
memory_allocators.h Make MemoryAllocator into a Customizable class (#8980) 2021-12-17 04:20:47 -08:00
merge_operators.cc Restore Regex support for ObjectLibrary::Register, rename new APIs to allow old one to be deprecated in the future (#9362) 2022-01-11 06:33:48 -08:00
merge_operators.h Make MergeOperator+CompactionFilter/Factory into Customizable Classes (#8481) 2021-08-06 08:27:25 -07:00
object_registry.cc Added GetFactoryCount/Names/Types to ObjectRegistry (#9358) 2022-05-16 09:44:43 -07:00
object_registry_test.cc Added GetFactoryCount/Names/Types to ObjectRegistry (#9358) 2022-05-16 09:44:43 -07:00
util_merge_operators_test.cc
wal_filter.cc Make WalFilter, SstPartitionerFactory, FileChecksumGenFactory, and TableProperties Customizable (#8638) 2021-09-28 05:32:02 -07:00