On Mon, 09 Dec 2019 16:58:02 +0100, Takashi Iwai wrote:
On Mon, 09 Dec 2019 16:16:56 +0100, Takashi Sakamoto wrote:
Hi,
I'm not opposed to this patchset. The direction is preferable.
However changes in this patchset brings bugs in error path of each implementation for pcm.hw_params.
On Mon, Dec 09, 2019 at 10:48:41AM +0100, Takashi Iwai wrote:
diff --git a/sound/firewire/bebob/bebob_pcm.c b/sound/firewire/bebob/bebob_pcm.c index d4edd06d32cf..2ee91ff6891e 100644 --- a/sound/firewire/bebob/bebob_pcm.c +++ b/sound/firewire/bebob/bebob_pcm.c @@ -214,10 +214,6 @@ static int pcm_hw_params(struct snd_pcm_substream *substream, struct snd_bebob *bebob = substream->private_data; int err;
- err = snd_pcm_lib_malloc_pages(substream, params_buffer_bytes(hw_params));
- if (err < 0)
return err;- if (substream->runtime->status->state == SNDRV_PCM_STATE_OPEN) { unsigned int rate = params_rate(hw_params); unsigned int frames_per_period = params_period_size(hw_params);
@@ -231,7 +227,7 @@ static int pcm_hw_params(struct snd_pcm_substream *substream, mutex_unlock(&bebob->mutex); }
- return err;
- return 0;
}
This should return the err variable to report failure of isochronous resources reservation.
However I found that ALSA oxfw/fireface drivers returns 0 even if it fails. This is bug since v5.3 kernel.
Would you please apply below two patches into your three, then modify your patch for all firewire drivers to return error correctly?
Yes, that sounds like a correct fix. Thank you for spotting out.
So, please go ahead, submit these two patches, then I'll update the fix containing your correction above and submit as v2 (maybe only for this, as the whole series is too big to re-submit).
Never mind, the patches there were already good enough, so I applied these two now. Will resend v2 fix for this patch.
thanks,
Takashi