code review 124670043: runtime: fix deadlock when gctrace

runtime: fix deadlock when gctrace
Calling ReadMemStats which does stoptheworld on m0 holding locks
was not a good idea.
Stoptheworld holding locks is a recipe for deadlocks (added check for this).
Stoptheworld on g0 may or may not work (added check for this as well).
As far as I understand scavenger will print incorrect numbers now,
as stack usage is not subtracted from heap. But it's better than deadlocking.

[dvyukov, 2014-08-20 11:22:33]

Hello golang-codereviews@googlegroups.com (cc: khr@golang.org, rlh@golang.org,
rsc@golang.org),

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

[dvyukov, 2014-08-20 18:28:56]

I will appreciate review of this one.

This currently causes hangs on perf builders. And we've scheduled perf dashboard
deployment with Andrew... it must be around midnight pacific time. If builders
will hang, it will hinder deployment.

[rsc, 2014-08-20 18:31:03]

R=khr

[khr, 2014-08-20 22:02:38]

LGTM.

https://codereview.appspot.com/124670043/diff/50005/src/pkg/runtime/proc.c
File src/pkg/runtime/proc.c (right):

https://codereview.appspot.com/124670043/diff/50005/src/pkg/runtime/proc.c#ne...
src/pkg/runtime/proc.c:506: // we just don't do it today.
I thought we did do this today - when malloc is called from C.

[dvyukov, 2014-08-21 07:46:35]

On 2014/08/20 22:02:38, khr wrote:
> LGTM.
> 
> https://codereview.appspot.com/124670043/diff/50005/src/pkg/runtime/proc.c
> File src/pkg/runtime/proc.c (right):
> 
>
https://codereview.appspot.com/124670043/diff/50005/src/pkg/runtime/proc.c#ne...
> src/pkg/runtime/proc.c:506: // we just don't do it today.
> I thought we did do this today - when malloc is called from C.

We've tried but it did not work for some reason. Now we just return from GC when
on g0.

[dvyukov, 2014-08-21 07:46:59]

*** Submitted as https://code.google.com/p/go/source/detail?r=b839c52cb382 ***

runtime: fix deadlock when gctrace
Calling ReadMemStats which does stoptheworld on m0 holding locks
was not a good idea.
Stoptheworld holding locks is a recipe for deadlocks (added check for this).
Stoptheworld on g0 may or may not work (added check for this as well).
As far as I understand scavenger will print incorrect numbers now,
as stack usage is not subtracted from heap. But it's better than deadlocking.

LGTM=khr
R=golang-codereviews, rsc, khr
CC=golang-codereviews, rlh
https://codereview.appspot.com/124670043