Re: [rc8] Regression: key repeats and skips in music during compilation

[Richard Jonsson <richie@coderworld.net>]

Richard Jonsson skrev:
> Frans Pop skrev:
>> On Monday 07 April 2008, Dhaval Giani wrote:
>>> On Mon, Apr 07, 2008 at 12:29:26PM +0200, Rafael J. Wysocki wrote:
>>>> On Monday, 7 of April 2008, Frans Pop wrote:
>>>>> On Monday 07 April 2008, Rafael J. Wysocki wrote:
>>>>>> On Sunday, 6 of April 2008, Rafael J. Wysocki wrote:
>>>>>>> On Friday, 4 of April 2008, Frans Pop wrote:
>>>>>>>> While compiling glibc I am getting frequent short skips in
>>>>>>>> music play anddddddddddddddddd I have now three times seen key
>>>>>>>> repeats (as in this sentence). When there is a skip in the
>>>>>>>> music, there is also a delay in characters being typed
>>>>>>>> appearing on the screen.
>>>>>>> Can you test with CONFIG_GROUP_SCHED unset, please?
>>>>>> I was able to reproduce this issue with -rc8 on one of my test
>>>>>> systems,
>>>>> In the mean time I have verified that the issue also existed in rc6,
>>>>> so it's not a recent regression.
>>>>>
>>>>>> but I'm not any more with the current Linus' tree and
>>>>>> CONFIG_GROUP_SCHED unset, FWIW.
>>>>> Hmm. Does that mean that you _can_ reproduce it with git HEAD and
>>>>> CONFIG_GROUP_SCHED set? That would mean it is still an issue in HEAD
>>>>> and group scheduling is the culprit.
>>>> Well, CONFIG_GROUP_SCHED is known to cause problems, AFAICS, and some
>>>> fixes are scheduled for the 2.6.26 time frame.
>>> I thought we had most of the issues ironed out.
>>
>> Note that I can also reproduce the issue with a kernel that does *not* 
>> have CONFIG_GROUP_SCHED set. I still get the music skips (maybe less 
>> frequently, but that's hard to measure), and I've have also seen key 
>> repeats once.
>>
>> Anything I can do to provide additional info on this?
>>
>> Cheers,
>> FJP
>>
>> # CONFIG_CGROUPS is not set
>> # CONFIG_GROUP_SCHED is not set
>> (config otherwise identical to my initial mail)
>> -- 
>> To unsubscribe from this list: send the line "unsubscribe 
>> linux-kernel" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>> Please read the FAQ at  http://www.tux.org/lkml/
> 
> Hi
> 
> I reported about this problem a while back [1], and have been busy 
> bisecting. I have verified the issue in v2.6.25-rc6 and I'm about 99% 
> certain that the issue is not present in v2.6.25-rc5-300-gb663c6f.
> 
> I've been running this kernel for 3 days now with no problems.
> 
> It's a bit worrying to me that it's more frequent in builds between rc6 
> and rc7 than vanilla rc6. Newer builds than rc7 not tested.
> 
> I've not noticed stuttering audio, but at one time I saw stuttering 
> video in a way I don't normally see. May be incidence.
> 
> Maybe relevant config:
> # CONFIG_CGROUPS is not set
> CONFIG_GROUP_SCHED=y
> CONFIG_FAIR_GROUP_SCHED=y
> # CONFIG_RT_GROUP_SCHED is not set
> # CONFIG_CGROUP_SCHED is not set
> 
> [1]: http://lkml.org/lkml/2008/3/29/96

Bisecting this bug is very inefficient, but I've managed to narrow it 
down a bit more.

The log so far:
git-bisect start
# good: [b663c6fd98c9cf586279db03cec3257c413efd00] nfsd: fix oops on 
access from high-numbered ports
git-bisect good b663c6fd98c9cf586279db03cec3257c413efd00
# bad: [a978b30af3bab0dd9af9350eeda25e76123fa28e] Linux 2.6.25-rc6
git-bisect bad a978b30af3bab0dd9af9350eeda25e76123fa28e
# bad: [afbf331ed1252c85753ac6790356c11e171f3d0b] Merge 
git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
git-bisect bad afbf331ed1252c85753ac6790356c11e171f3d0b
# good: [bde1709aaa98f5004ab1580842c422be18eb4bc3] firewire: fw-ohci: 
use dma_alloc_coherent for ar_buffer
git-bisect good bde1709aaa98f5004ab1580842c422be18eb4bc3

This leaves these commits according to 'git bisect visualize':
[NET]: Fix tbench regression in 2.6.25-rc1
sched: simplify sched_slice()
sched: fix fair sleepers
sched: fix overload performance: buddy wakeups
sched: fix calc_delta_mine()
sched: fix update_load_add()/sub()  <-- currently bisecting, seems to be 
fine
sched: min_vruntime fix
sched: fix race in schedule()
Merge branch 'for-linus' of 
git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6
firewire: fw-ohci: shut up false compiler warning on PPC32
[SCTP]: Fix local_addr deletions during list traversals.


Hope this helps / Richard