ASoC: SOF: Intel: hda-dai: Drop call to hda_link_dma_cleanup in dai_s… #5049
Conversation
…uspend() The DAI ops and sequences have been reworked but the hda_dai_suspend() function kept the call to hda_link_dma_cleanup(). This causes a system lockup on suspend with paused streams. Drop the stale call to fix the system lockup. Note: that this patch will _not_ fix the suspend while pause, it is not working, but at least the system is not locked up hard. Signed-off-by: Peter Ujfalusi <peter.ujfalusi@linux.intel.com>
| cpu_dai); | ||
| if (ret < 0) | ||
| return ret; | ||
|
|
There was a problem hiding this comment.
This seems to contradict the comments line 599-602, and it's not clear why this causes a hard lockup?
There was a problem hiding this comment.
Yeah, something is not right for sure, but it prevents the lockup.
If we suspend with active audio we run through the pre_trigger/trigger/post_trigger sequence (and SUSPEND/STOP/PAUSE_PUSH is identical up to post_trigger) then we call hda_link_dma_cleanup().
With the suspend while paused we just have a longer delay between the triggers and the hda_link_dma_cleanup()
But, we have failing FE triggers, so we might fail to stop host side and this has some side effect on the link side if it is cleaned?
We cannot see this as the system is hard locked..
|
Likely just a side effect that it is fixing the lockup, closing. |
…uspend()
The DAI ops and sequences have been reworked but the hda_dai_suspend() function kept the call to hda_link_dma_cleanup().
This causes a system lockup on suspend with paused streams.
Drop the stale call to fix the system lockup.
Note: that this patch will not fix the suspend while pause, it is not working, but at least the system is not locked up hard.