deadlock

README

Online deadlock detection in go (golang).

Why

Deadlocks happen and are painful to debug.

What

go-deadlock provides (RW)Mutex drop-in replacements for sync.(RW)Mutex. It would not work if you create a spaghetti of channels. Mutexes only.

Installation

go get github.com/sasha-s/go-deadlock/...

Usage

import "github.com/sasha-s/go-deadlock"
var mu deadlock.Mutex
// Use normally, it works exactly like sync.Mutex does.
mu.Lock()

defer mu.Unlock()
// Or
var rw deadlock.RWMutex
rw.RLock()
defer rw.RUnlock()

Deadlocks

One of the most common sources of deadlocks is inconsistent lock ordering: say, you have two mutexes A and B, and in some goroutines you have

A.Lock() // defer A.Unlock() or similar.
...
B.Lock() // defer B.Unlock() or similar.

And in another goroutine the order of locks is reversed:

B.Lock() // defer B.Unlock() or similar.
...
A.Lock() // defer A.Unlock() or similar.

Another common sources of deadlocs is duplicate take a lock in a goroutine:

A.Rlock() or lock()

A.lock() or A.RLock()

This does not guarantee a deadlock (maybe the goroutines above can never be running at the same time), but it usually a design flaw at least.

go-deadlock can detect such cases (unless you cross goroutine boundary - say lock A, then spawn a goroutine, block until it is singals, and lock B inside of the goroutine), even if the deadlock itself happens very infrequently and is painful to reproduce!

Each time go-deadlock sees a lock attempt for lock B, it records the order A before B, for each lock that is currently being held in the same goroutine, and it prints (and exits the program by default) when it sees the locking order being violated.

In addition, if it sees that we are waiting on a lock for a long time (opts.DeadlockTimeout, 30 seconds by default), it reports a potential deadlock, also printing the stacktrace for a goroutine that is currently holding the lock we are desperately trying to grab.

Sample output

Inconsistent lock ordering:

POTENTIAL DEADLOCK: Inconsistent locking. saw this ordering in one goroutine:
happened before
inmem.go:623 bttest.(*server).ReadModifyWriteRow { r.mu.Lock() } <<<<<
inmem_test.go:118 bttest.TestConcurrentMutationsReadModifyAndGC.func4 { _, _ = s.ReadModifyWriteRow(ctx, rmw()) }

happened after
inmem.go:629 bttest.(*server).ReadModifyWriteRow { tbl.mu.RLock() } <<<<<
inmem_test.go:118 bttest.TestConcurrentMutationsReadModifyAndGC.func4 { _, _ = s.ReadModifyWriteRow(ctx, rmw()) }

in another goroutine: happened before
inmem.go:799 bttest.(*table).gc { t.mu.RLock() } <<<<<
inmem_test.go:125 bttest.TestConcurrentMutationsReadModifyAndGC.func5 { tbl.gc() }

happend after
inmem.go:814 bttest.(*table).gc { r.mu.Lock() } <<<<<
inmem_test.go:125 bttest.TestConcurrentMutationsReadModifyAndGC.func5 { tbl.gc() }

Waiting for a lock for a long time:

POTENTIAL DEADLOCK:
Previous place where the lock was grabbed
goroutine 240 lock 0xc820160440
inmem.go:799 bttest.(*table).gc { t.mu.RLock() } <<<<<
inmem_test.go:125 bttest.TestConcurrentMutationsReadModifyAndGC.func5 { tbl.gc() }

Have been trying to lock it again for more than 40ms
goroutine 68 lock 0xc820160440
inmem.go:785 bttest.(*table).mutableRow { t.mu.Lock() } <<<<<
inmem.go:428 bttest.(*server).MutateRow { r := tbl.mutableRow(string(req.RowKey)) }
inmem_test.go:111 bttest.TestConcurrentMutationsReadModifyAndGC.func3 { s.MutateRow(ctx, req) }


Here is what goroutine 240 doing now
goroutine 240 [select]:
github.com/sasha-s/go-deadlock.lock(0xc82028ca10, 0x5189e0, 0xc82013a9b0)
        /Users/sasha/go/src/github.com/sasha-s/go-deadlock/deadlock.go:163 +0x1640
github.com/sasha-s/go-deadlock.(*Mutex).Lock(0xc82013a9b0)
        /Users/sasha/go/src/github.com/sasha-s/go-deadlock/deadlock.go:54 +0x86
google.golang.org/cloud/bigtable/bttest.(*table).gc(0xc820160440)
        /Users/sasha/go/src/google.golang.org/cloud/bigtable/bttest/inmem.go:814 +0x28d
google.golang.org/cloud/bigtable/bttest.TestConcurrentMutationsReadModifyAndGC.func5(0xc82015c760, 0xc820160440)      /Users/sasha/go/src/google.golang.org/cloud/bigtable/bttest/inmem_test.go:125 +0x48
created by google.golang.org/cloud/bigtable/bttest.TestConcurrentMutationsReadModifyAndGC
        /Users/sasha/go/src/google.golang.org/cloud/bigtable/bttest/inmem_test.go:126 +0xb6f

Used in

cockroachdb: Potential deadlock between Gossip.SetStorage and Node.gossipStores

bigtable/bttest: A race between GC and row mutations

Need a mutex that works with net.context?

I have one.

Grabbing an RLock twice from the same goroutine

This is, surprisingly, not a good idea!

From RWMutex docs:

If a goroutine holds a RWMutex for reading and another goroutine might call Lock, no goroutine should expect to be able to acquire a read lock until the initial read lock is released. In particular, this prohibits recursive read locking. This is to ensure that the lock eventually becomes available; a blocked Lock call excludes new readers from acquiring the lock.

The following code will deadlock — run the example on playground or try it online with go-deadlock on wandbox:

package main

import (
	"fmt"
	"sync"
)

func main() {
	var mu sync.RWMutex

	chrlockTwice := make(chan struct{}) // Used to control rlockTwice
	rlockTwice := func() {
		mu.RLock()
		fmt.Println("first Rlock succeeded")
		<-chrlockTwice
		<-chrlockTwice
		fmt.Println("trying to Rlock again")
		mu.RLock()
		fmt.Println("second Rlock succeeded")
		mu.RUnlock()
		mu.RUnlock()
	}

	chLock := make(chan struct{}) // Used to contol lock
	lock := func() {
		<-chLock
		fmt.Println("about to Lock")
		mu.Lock()
		fmt.Println("Lock succeeded")
		mu.Unlock()
		<-chLock
	}

	control := func() {
		chrlockTwice <- struct{}{}
		chLock <- struct{}{}

		close(chrlockTwice)
		close(chLock)
	}

	go control()
	go lock()
	rlockTwice()
}

Documentation

Index

• Variables
• type Mutex
  • func (m *Mutex) Lock()
  • func (m *Mutex) Unlock()
• type RWMutex
  • func (m *RWMutex) Lock()
  • func (m *RWMutex) RLock()
  • func (m *RWMutex) RLocker() sync.Locker
  • func (m *RWMutex) RUnlock()
  • func (m *RWMutex) Unlock()

Variables

var Opts = struct {
	// Mutex/RWMutex would work exactly as their sync counterparts
	// -- almost no runtime penalty, no deadlock detection if Disable == true.
	Disable bool
	// Would disable lock order based deadlock detection if DisableLockOrderDetection == true.
	DisableLockOrderDetection bool
	// Waiting for a lock for longer than DeadlockTimeout is considered a deadlock.
	// Ignored is DeadlockTimeout <= 0.
	DeadlockTimeout time.Duration
	// OnPotentialDeadlock is called each time a potential deadlock is detected -- either based on
	// lock order or on lock wait time.
	OnPotentialDeadlock func()
	// Will keep MaxMapSize lock pairs (happens before // happens after) in the map.
	// The map resets once the threshold is reached.
	MaxMapSize int
	// Will dump stacktraces of all goroutines when inconsistent locking is detected.
	PrintAllCurrentGoroutines bool
	mu                        *sync.Mutex // Protects the LogBuf.
	// Will print deadlock info to log buffer.
	LogBuf io.Writer
}{
	DeadlockTimeout: time.Second * 30,
	OnPotentialDeadlock: func() {
		os.Exit(2)
	},
	MaxMapSize: 1024 * 64,

	LogBuf: os.Stderr,
	// contains filtered or unexported fields
}

Opts control how deadlock detection behaves. Options are supposed to be set once at a startup (say, when parsing flags).

Types

type Mutex

type Mutex struct {
	// contains filtered or unexported fields
}

A Mutex is a drop-in replacement for sync.Mutex. Performs deadlock detection unless disabled in Opts.

func (*Mutex) Lock

func (m *Mutex) Lock()

Lock locks the mutex. If the lock is already in use, the calling goroutine blocks until the mutex is available.

Unless deadlock detection is disabled, logs potential deadlocks to Opts.LogBuf, calling Opts.OnPotentialDeadlock on each occasion.

func (*Mutex) Unlock

func (m *Mutex) Unlock()

Unlock unlocks the mutex. It is a run-time error if m is not locked on entry to Unlock.

A locked Mutex is not associated with a particular goroutine. It is allowed for one goroutine to lock a Mutex and then arrange for another goroutine to unlock it.

type RWMutex

type RWMutex struct {
	// contains filtered or unexported fields
}

An RWMutex is a drop-in replacement for sync.RWMutex. Performs deadlock detection unless disabled in Opts.

func (*RWMutex) Lock

func (m *RWMutex) Lock()

Lock locks rw for writing. If the lock is already locked for reading or writing, Lock blocks until the lock is available. To ensure that the lock eventually becomes available, a blocked Lock call excludes new readers from acquiring the lock.

Unless deadlock detection is disabled, logs potential deadlocks to Opts.LogBuf, calling Opts.OnPotentialDeadlock on each occasion.

func (*RWMutex) RLock

func (m *RWMutex) RLock()

RLock locks the mutex for reading.

Unless deadlock detection is disabled, logs potential deadlocks to Opts.LogBuf, calling Opts.OnPotentialDeadlock on each occasion.

func (*RWMutex) RLocker

func (m *RWMutex) RLocker() sync.Locker

RLocker returns a Locker interface that implements the Lock and Unlock methods by calling RLock and RUnlock.

func (*RWMutex) RUnlock

func (m *RWMutex) RUnlock()

RUnlock undoes a single RLock call; it does not affect other simultaneous readers. It is a run-time error if rw is not locked for reading on entry to RUnlock.

func (*RWMutex) Unlock

func (m *RWMutex) Unlock()

Unlock unlocks the mutex for writing. It is a run-time error if rw is not locked for writing on entry to Unlock.

As with Mutexes, a locked RWMutex is not associated with a particular goroutine. One goroutine may RLock (Lock) an RWMutex and then arrange for another goroutine to RUnlock (Unlock) it.