On Wed, Mar 09, 2022 at 05:32:19PM +0800, Jiasheng Jiang wrote:
> On Wed, Mar 09, 2022 at 01:58:04AM +0800, Dan Carpenter wrote:
> > smatch warnings:
> > sound/soc/soc-compress.c:628 snd_soc_new_compress() error: we previously assumed 'codec_dai' could be null (see line 571)
>
> > de2c6f98817fa5d Jiasheng Jiang 2021-10-15 @571 if (codec_dai) {
> >
> > Can codec_dai be NULL? This new code assumes so.
>
> > aeb6fa0f15c71a1 Peng Donglin 2017-08-16 @628 rtd->dai_link->stream_name, codec_dai->name, num);
> > ^^^^^^^^^^^^^^^^
> > Unchecked dereference
>
> Actually, if 'codec_dai' is NULL, the check in line 584 will fail and return -EINVAL.
>
> a1068045883ed4a Vinod Koul 2016-01-07 584 if (playback + capture != 1) {
> 141dfc9e3751f5f Charles Keepax 2018-01-26 585 dev_err(rtd->card->dev,
> 141dfc9e3751f5f Charles Keepax 2018-01-26 586 "Compress ASoC: Invalid direction for P %d, C %d\n",
> a1068045883ed4a Vinod Koul 2016-01-07 587 playback, capture);
> a1068045883ed4a Vinod Koul 2016-01-07 588 return -EINVAL;
> a1068045883ed4a Vinod Koul 2016-01-07 589 }
>
Yeah. That's true. Smatch tries to look at conditions and figure out
the implications but this one is too complicated.
But if codec_dai can really be NULL then it would be easier to
understand if we reversed the check on codec_dai and returned early.
if (!codec_dai) {
dev_err(rtd->card->dev, "Missing codec\n");
return -EINVAL;
}
/* check client and interface hw capabilities */
if (snd_soc_dai_stream_valid(codec_dai, SNDRV_PCM_STREAM_PLAYBACK) &&
snd_soc_dai_stream_valid(cpu_dai, SNDRV_PCM_STREAM_PLAYBACK))
playback = 1;
regards,
dan carpenter
