socket.go中SimpleQuery中的本地互斥锁

Question

I tried to read code of mgo, and the function SimpleQuery in socket.go confused me. In this function, there are two local mutex: https://github.com/go-mgo/mgo/blob/v2-unstable/socket.go

func (socket *mongoSocket) SimpleQuery(op *queryOp) (data []byte, err error) {
    var wait, change sync.Mutex
    var replyDone bool
    var replyData []byte
    var replyErr error
    wait.Lock()
    op.replyFunc = func(err error, reply *replyOp, docNum int, docData []byte) {
        change.Lock()
        if !replyDone {
            replyDone = true
            replyErr = err
            if err == nil {
                replyData = docData
            }
        }
        change.Unlock()
        wait.Unlock()
    }
    err = socket.Query(op)
    if err != nil {
        return nil, err
    }
    wait.Lock()
    change.Lock()
    data = replyData
    err = replyErr
    change.Unlock()
    return data, err
}

As my understanding, for each call, there will be a new local mutex, so, it's actually protecting anything, why they put the mutex here?

Answer 1

The mutexes are protecting data structures, they are both captured in the closure given to the replyFunc callback.

The wait mutex blocks until the callback returns, and the change mutex protects the replyData and replyErr values.

The commit that added those only says:

Make SimpleQuery race safe, irrespective of what the server does.

So this may be extra protection against some unintended or legacy behavior, which is why it's using multiple mutexes rather than a single primitive to synchronize the operation.