code review 71030044: cmd/ld: clear unused ctxt before morestack

cmd/ld: clear unused ctxt before morestack

For non-closure functions, the context register is uninitialized
on entry and will not be used, but morestack saves it and then the
garbage collector treats it as live. This can be a source of memory
leaks if the context register points at otherwise dead memory.
Avoid this by introducing a parallel set of morestack functions
that clear the context register, and use those for the non-closure functions.

I hope this will help with some of the finalizer flakiness, but it probably won't.

Fixes issue 7244.

[rsc, 2014-03-04 04:57:55]

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

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

[rsc, 2014-03-04 05:27:48]

As expected, this does not fix the amd64 finalizer flakiness. Still worth
doing.

[dvyukov, 2014-03-04 05:59:06]

LGTM

please also test something with -cpuprofile

[rsc, 2014-03-04 06:00:52]

On Tue, Mar 4, 2014 at 12:59 AM, <dvyukov@google.com> wrote:

> LGTM
>
> please also test something with -cpuprofile
>

there are pprof tests in runtime/pprof during all.bash. is that enough?

[dvyukov, 2014-03-04 06:29:18]

On Tue, Mar 4, 2014 at 10:00 AM, Russ Cox <rsc@golang.org> wrote:
> On Tue, Mar 4, 2014 at 12:59 AM, <dvyukov@google.com> wrote:
>>
>> LGTM
>>
>> please also test something with -cpuprofile
>
>
> there are pprof tests in runtime/pprof during all.bash. is that enough?

I've discovered a bunch of pprof bugs only on perf dashboard builders
that run everything under cpuprofiler, the bugs were not uncovered by
std pprof tests.

[rsc, 2014-03-04 18:51:45]

I take it back. This change was not sufficient to remove the finalizer
flakiness but it is definitely necessary. In particular, neither
CL 71080043 nor this one works in isolation, but together they do seem to
deflake the finalizer test.

[rsc, 2014-03-04 18:53:11]

*** Submitted as https://code.google.com/p/go/source/detail?r=b57ca889129c ***

cmd/ld: clear unused ctxt before morestack

For non-closure functions, the context register is uninitialized
on entry and will not be used, but morestack saves it and then the
garbage collector treats it as live. This can be a source of memory
leaks if the context register points at otherwise dead memory.
Avoid this by introducing a parallel set of morestack functions
that clear the context register, and use those for the non-closure functions.

I hope this will help with some of the finalizer flakiness, but it probably
won't.

Fixes issue 7244.

LGTM=dvyukov
R=khr, dvyukov
CC=golang-codereviews
https://codereview.appspot.com/71030044

[gobot, 2014-03-04 19:01:04]

This CL appears to have broken the linux-arm-panda builder.