Hi Morimoto-san, Thank you for your review.
> > If so, Although I don't have MSIOF hardware, I tested this on FSI hardware. I added a .close() callback to fsi_soc_component and forced fsi_dai_startup() to fail to trigger rollback in __soc_pcm_open(). When fsi_pcm_open() returned 1 from snd_pcm_hw_constraint_integer(), the .close() callback was skipped during rollback. When it returned 0, the callback was called as expected. >we would like to have such comment to indicate such background, > because it is not intuitive for people. > Otherwise, someone (maybe me :) will return to the original code. > I'll add a comment to explain why the return value must be normalized to 0 in the affected drivers. For MSIOF, I noticed that snd_dmaengine_pcm_open() already calls snd_pcm_hw_constraint_integer() for SNDRV_PCM_HW_PARAM_PERIODS, so the call in msiof_open() seems redundant. Since the first call succeeds, the second call should return 0. Would removing the redundant call be acceptable? Best regards, Phuc
