mirror of
https://github.com/facebook/rocksdb.git
synced 2024-11-25 22:44:05 +00:00
d010b02e86
Summary: In follow-up to https://github.com/facebook/rocksdb/issues/11922, fix a race in functions like CreateColumnFamily and SetDBOptions where the DB reports one option setting but a different one is left in effect. To fix, we can add an extra mutex around these rare operations. We don't want to hold the DB mutex during I/O or other slow things because of the many purposes it serves, but a mutex more limited to these cases should be fine. I believe this would fix a write-write race in https://github.com/facebook/rocksdb/issues/10079 but not the read-write race. Intended follow-up to this: * Should be able to remove write thread synchronization from DBImpl::WriteOptionsFile Pull Request resolved: https://github.com/facebook/rocksdb/pull/11929 Test Plan: Added two mini-stress style regression tests that fail with >1% probability before this change: DBOptionsTest::SetStatsDumpPeriodSecRace ColumnFamilyTest::CreateAndDropPeriodicRace I haven't reproduced such an inconsistency between in-memory options and on disk latest options, but this change at least improves safety and adds a test anyway: DBOptionsTest::SetStatsDumpPeriodSecRace Reviewed By: ajkr Differential Revision: D50024506 Pulled By: pdillinger fbshipit-source-id: 1e99a9ed4d96fdcf3ac5061ec6b3cee78aecdda4
127 lines
3.6 KiB
C++
127 lines
3.6 KiB
C++
// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
|
|
// This source code is licensed under both the GPLv2 (found in the
|
|
// COPYING file in the root directory) and Apache 2.0 License
|
|
// (found in the LICENSE.Apache file in the root directory).
|
|
|
|
#pragma once
|
|
|
|
#include "monitoring/statistics_impl.h"
|
|
#include "port/port.h"
|
|
#include "rocksdb/statistics.h"
|
|
#include "rocksdb/system_clock.h"
|
|
#include "rocksdb/thread_status.h"
|
|
#include "util/stop_watch.h"
|
|
|
|
namespace ROCKSDB_NAMESPACE {
|
|
class InstrumentedCondVar;
|
|
|
|
// A wrapper class for port::Mutex that provides additional layer
|
|
// for collecting stats and instrumentation.
|
|
class InstrumentedMutex {
|
|
public:
|
|
explicit InstrumentedMutex(bool adaptive = false)
|
|
: mutex_(adaptive), stats_(nullptr), clock_(nullptr), stats_code_(0) {}
|
|
|
|
explicit InstrumentedMutex(SystemClock* clock, bool adaptive = false)
|
|
: mutex_(adaptive), stats_(nullptr), clock_(clock), stats_code_(0) {}
|
|
|
|
InstrumentedMutex(Statistics* stats, SystemClock* clock, int stats_code,
|
|
bool adaptive = false)
|
|
: mutex_(adaptive),
|
|
stats_(stats),
|
|
clock_(clock),
|
|
stats_code_(stats_code) {}
|
|
|
|
#ifdef COERCE_CONTEXT_SWITCH
|
|
InstrumentedMutex(Statistics* stats, SystemClock* clock, int stats_code,
|
|
InstrumentedCondVar* bg_cv, bool adaptive = false)
|
|
: mutex_(adaptive),
|
|
stats_(stats),
|
|
clock_(clock),
|
|
stats_code_(stats_code),
|
|
bg_cv_(bg_cv) {}
|
|
#endif
|
|
|
|
void Lock();
|
|
|
|
void Unlock() { mutex_.Unlock(); }
|
|
|
|
void AssertHeld() const { mutex_.AssertHeld(); }
|
|
|
|
private:
|
|
void LockInternal();
|
|
friend class InstrumentedCondVar;
|
|
port::Mutex mutex_;
|
|
Statistics* stats_;
|
|
SystemClock* clock_;
|
|
int stats_code_;
|
|
#ifdef COERCE_CONTEXT_SWITCH
|
|
InstrumentedCondVar* bg_cv_ = nullptr;
|
|
#endif
|
|
};
|
|
|
|
class ALIGN_AS(CACHE_LINE_SIZE) CacheAlignedInstrumentedMutex
|
|
: public InstrumentedMutex {
|
|
using InstrumentedMutex::InstrumentedMutex;
|
|
};
|
|
static_assert(alignof(CacheAlignedInstrumentedMutex) != CACHE_LINE_SIZE ||
|
|
sizeof(CacheAlignedInstrumentedMutex) % CACHE_LINE_SIZE == 0);
|
|
|
|
// RAII wrapper for InstrumentedMutex
|
|
class InstrumentedMutexLock {
|
|
public:
|
|
explicit InstrumentedMutexLock(InstrumentedMutex* mutex) : mutex_(mutex) {
|
|
mutex_->Lock();
|
|
}
|
|
|
|
~InstrumentedMutexLock() { mutex_->Unlock(); }
|
|
|
|
private:
|
|
InstrumentedMutex* const mutex_;
|
|
InstrumentedMutexLock(const InstrumentedMutexLock&) = delete;
|
|
void operator=(const InstrumentedMutexLock&) = delete;
|
|
};
|
|
|
|
// RAII wrapper for temporary releasing InstrumentedMutex inside
|
|
// InstrumentedMutexLock
|
|
class InstrumentedMutexUnlock {
|
|
public:
|
|
explicit InstrumentedMutexUnlock(InstrumentedMutex* mutex) : mutex_(mutex) {
|
|
mutex_->Unlock();
|
|
}
|
|
|
|
~InstrumentedMutexUnlock() { mutex_->Lock(); }
|
|
|
|
private:
|
|
InstrumentedMutex* const mutex_;
|
|
InstrumentedMutexUnlock(const InstrumentedMutexUnlock&) = delete;
|
|
void operator=(const InstrumentedMutexUnlock&) = delete;
|
|
};
|
|
|
|
class InstrumentedCondVar {
|
|
public:
|
|
explicit InstrumentedCondVar(InstrumentedMutex* instrumented_mutex)
|
|
: cond_(&(instrumented_mutex->mutex_)),
|
|
stats_(instrumented_mutex->stats_),
|
|
clock_(instrumented_mutex->clock_),
|
|
stats_code_(instrumented_mutex->stats_code_) {}
|
|
|
|
void Wait();
|
|
|
|
bool TimedWait(uint64_t abs_time_us);
|
|
|
|
void Signal() { cond_.Signal(); }
|
|
|
|
void SignalAll() { cond_.SignalAll(); }
|
|
|
|
private:
|
|
void WaitInternal();
|
|
bool TimedWaitInternal(uint64_t abs_time_us);
|
|
port::CondVar cond_;
|
|
Statistics* stats_;
|
|
SystemClock* clock_;
|
|
int stats_code_;
|
|
};
|
|
|
|
} // namespace ROCKSDB_NAMESPACE
|