lkml.org 
[lkml]   [2016]   [Mar]   [1]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: sound: uninterruptible hang in snd_seq_oss_writeq_sync
On Tue, Mar 1, 2016 at 4:01 PM, Dmitry Vyukov <dvyukov@google.com> wrote:
> On Tue, Mar 1, 2016 at 3:44 PM, Takashi Iwai <tiwai@suse.de> wrote:
>> On Tue, 01 Mar 2016 13:33:27 +0100,
>> Dmitry Vyukov wrote:
>>>
>>> Hello,
>>>
>>> The following program creates an unkillable process:
>> .....
>>> The hang stack is:
>>>
>>> [<ffffffff85309f77>] snd_seq_oss_writeq_sync+0x327/0x790
>>> sound/core/seq/oss/seq_oss_writeq.c:121
>>
>> This is
>> wait_event_interruptible_timeout(q->sync_sleep, ! q->sync_event_put, HZ);
>>
>> and this should return zero
>> if (signal_pending(current))
>> /* interrupted - return 0 to finish sync */
>> q->sync_event_put = 0;
>> if (! q->sync_event_put || q->sync_time >= time)
>> return 0;
>> return 1;
>>
>>> [<ffffffff852fa353>] snd_seq_oss_drain_write+0x113/0x160
>>
>> ... and this loop should break:
>> while (snd_seq_oss_writeq_sync(dp->writeq))
>> ;
>>
>> So, I see no obvious error in the code, so far.
>>
>> I'm running your test program now with 8 parallel runs, but I couldn't
>> reproduce it. Any other specifics?


Hummm.. for me this process hangs every time, even if I just run it
once from console.
I've now retested in on clean commit
fc77dbd34c5c99bce46d40a2491937c3bcbd10af (without KASAN/KCOV and with
release gcc) and also got the same hang.
I can only think of a different config. I've attached mine, please try with it.

Are signals delivered if we are already in process of dying?
When I kill -9 this process it does not react, so presumably wait is
not unblocked...
[unhandled content-type:application/octet-stream]
\
 
 \ /
  Last update: 2016-03-01 16:41    [W:0.116 / U:1.480 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site