[Sound-open-firmware] [PATCH] FIX: Add dma_stop when receive trigger stop cmd in host

[Pan, Xiuli]

On 5/3/2018 15:09, Liam Girdwood wrote:
> On Thu, 2018-05-03 at 09:10 +0800, zhigangw wrote:
>> On 2018年05月02日 21:09, Pan, Xiuli wrote:
>>>
>>> On 5/2/2018 18:11, Liam Girdwood wrote:
>>>> On Wed, 2018-05-02 at 09:56 +0000, Jie, Yang wrote:
>>>>>> -----Original Message-----
>>>>>> From: sound-open-firmware-bounces at alsa-project.org [mailto:sound-open-
>>>>>> firmware-bounces at alsa-project.org] On Behalf Of Wu Zhigang
>>>>>> Sent: Wednesday, May 2, 2018 5:34 PM
>>>>>> To: sound-open-firmware at alsa-project.org
>>>>>> Cc: Wu Zhigang <zhigang.wu at linux.intel.com>; Wu, Zhigang
>>>>>> <zhigang.wu at intel.com>
>>>>>> Subject: [Sound-open-firmware] [PATCH] FIX: Add dma_stop when receive
>>>>>> trigger
>>>>>> stop cmd in host
>>>>>>
>>>>>> when ALSA lib detect xrun issue, it will send stop and start
>>>>>> command to DSP
>>>>>> without reset. it will hit error in hds_dma_start, if we did not
>>>>>> add the
>>>>>> dma_stop in
>>>>> Small typo. Should be 'hda_dma_start'.
>>>>>
>>>>> Others looks good to me.
>>>>>
>>>>>
>>>> Can someone confirm on BYT too.
>>> BYT did not need this handler. The host dma for BYT is also GP-DMA and
>>> need to assign the DMA every host copy. When we trigger the host comp
>>> stop, there will no more host DMA.
>>> This problem is only shown on the APL/CNL platforms.
>>>
>>> BTW, this bug is hit by the ALSA pcm lib xrun recovery. When ALSA
>>> trace the dma pointer position, it will wrongly though the DSP is in
>>> xrun and try to recover it by STOP/START pattern.
>>> Will the DSP be recovered from this kind of STOP/START?
>>>
>>> Thanks
>>> Xiuli
>> Actually, It should be added. Because we cannot confirm the ALSA lib
>> will not happen XRUN.
>> In this situation, we have to recovered from this case.
>> In the real environment, no one can confirm NO XRUN in ALSA.
>> we have to make sure our DSP is rebust.
>>
>> after fix this, there is still a xrun issue in the DSP.
>> thanks
> Ok, does this mean I should apply or is there another fix coming ?
Test with BYT and rt5651.
no critical regression happen.
Could not reproduce to the same ASLA xrun recover stop/start


Thanks
Xiuli
>
> Thanks
>
> Liam
>