15.09.2014 16:14, Takashi Iwai wrote:
> At Mon, 15 Sep 2014 16:03:57 +0600,
> Alexander E. Patrakov wrote:
>>
>> 15.09.2014 14:49, Takashi Iwai пишет:
>>> At Sun, 14 Sep 2014 00:30:18 +0600,
>>> Alexander E. Patrakov wrote:
>>>>
>>>> Such negative returns are possible during an underrun if xrun detection
>>>> is disabled.
>>>>
>>>> So, don't store the result in an unsigned variable (where it will
>>>> overflow), and postpone the trigger in such case, too.
>>>>
>>>> Signed-off-by: Alexander E. Patrakov <patrakov@gmail.com>
>>>> ---
>>>>
>>>> The patch is only compile-tested and the second hunk may well be wrong.
>>>>
>>>> There are also similar issues in pcm_share.c, but, as I don't completely
>>>> understand the code there and cannot test that plugin at all due to
>>>> unrelated crashes, there will be no patch from me.
>>>
>>> In general, hw_avail must not be negative before starting the stream.
>>> If it is, then it means most likely the driver problem, so we should
>>> return error immediately instead.
>>
>> Thanks for the review. Would -EBADFD be a correct error code, or do you
>> want something different? or maybe even an assert?
>
> I'd take either EPIPE or EBADFD.
> An assert would be an overkill, IMO.

I have sent the fix to the list, but nobody reacted. Resending as an 
attachment to this message.

-- 
Alexander E. Patrakov