code review 12167043: runtime: do not park sysmon thread if any goroutines ar...

runtime: do not park sysmon thread if any goroutines are running
Sysmon thread parks if no goroutines are running (runtime.sched.npidle == runtime.gomaxprocs).
Currently it's unparked when a goroutine enters syscall, it was enough
to retake P's from blocking syscalls.
But it's not enough for reliable goroutine preemption. We need to ensure that
sysmon runs if any goroutines are running.

[dvyukov, 2013-07-31 14:18:52]

Hello golang-dev@googlegroups.com,

I'd like you to review this change to
https://dvyukov%40google.com@code.google.com/p/go/

[dvyukov, 2013-07-31 14:19:34]

This is the piece I forgot to move from my original preemption patch:
https://codereview.appspot.com/9136045/diff/28002/src/pkg/runtime/proc.c

[rsc, 2013-07-31 15:27:34]

LGTM

Is that really a non-atomic swap? Will it mean multiple notewakeup calls?
Does that matter?

[dvyukov, 2013-07-31 15:59:49]

On 2013/07/31 15:27:34, rsc wrote:
> LGTM
> 
> Is that really a non-atomic swap? Will it mean multiple notewakeup calls?
> Does that matter?

It's always under sched lock.

[dvyukov, 2013-07-31 16:00:06]

*** Submitted as https://code.google.com/p/go/source/detail?r=475e11851fc1 ***

runtime: do not park sysmon thread if any goroutines are running
Sysmon thread parks if no goroutines are running (runtime.sched.npidle ==
runtime.gomaxprocs).
Currently it's unparked when a goroutine enters syscall, it was enough
to retake P's from blocking syscalls.
But it's not enough for reliable goroutine preemption. We need to ensure that
sysmon runs if any goroutines are running.

R=golang-dev, rsc
CC=golang-dev
https://codereview.appspot.com/12167043