By reading golang src pipe.go to figure out how pipe works, I ran into these two write() & read() functions. What confuses me is that if reader calls read() func and hold l.lock and then waiting for data, how does writer call write() func and acquire l.lock to write data in?
func (p *pipe) write(b []byte) (n int, err error) {
// pipe uses nil to mean not available
if b == nil {
b = zero[:]
}
// One writer at a time.
p.wl.Lock()
defer p.wl.Unlock()
p.l.Lock()
defer p.l.Unlock()
if p.werr != nil {
err = ErrClosedPipe
return
}
p.data = b
p.rwait.Signal()
for {
if p.data == nil {
break
}
if p.rerr != nil {
err = p.rerr
break
}
if p.werr != nil {
err = ErrClosedPipe
break
}
p.wwait.Wait()
}
n = len(b) - len(p.data)
p.data = nil // in case of rerr or werr
return
}
and read:
func (p *pipe) read(b []byte) (n int, err error) {
// One reader at a time.
p.rl.Lock()
defer p.rl.Unlock()
p.l.Lock()
defer p.l.Unlock()
for {
if p.rerr != nil {
return 0, ErrClosedPipe
}
if p.data != nil {
break
}
if p.werr != nil {
return 0, p.werr
}
p.rwait.Wait()
}
n = copy(b, p.data)
p.data = p.data[n:]
if len(p.data) == 0 {
p.data = nil
p.wwait.Signal()
}
return
}
The mutex p.l is used in the read and write sync.Cond conditions, which will lock and unlock it as necessary.
Calling Wait on the condition unlocks its lock, and waits for a corresponding Signal call. You can see that the pipe uses p.wwait and p.rwait to coordinate the readers and writers within the Read and Write methods.