On 2/24/20 9:47 AM, Dmitry Osipenko wrote:
24.02.2020 15:01, Mark Brown пишет:
On Thu, Feb 20, 2020 at 11:49:55AM +0200, Kai Vehmanen wrote:
ASoC component open/close and snd_soc_component_module_get/put are called independently for each component-substream pair, so the logic added in commit dd03907bf129 ("ASoC: soc-pcm: call snd_soc_component_open/close() once") was not sufficient and led to PCM playback and module unload errors.
Implement handling of failures directly in soc_pcm_components_open(), so that any successfully opened components are closed upon error with other components. This allows to clean up error handling in soc_pcm_open() without adding more state tracking.
Do people have thoughts on this? I do like this approach but can't really test effectively myself.
I haven't tried to review this patch, but it works fine on NVIDIA Tegra:
Tested-by: Dmitry Osipenko digetx@gmail.com
LGTM - I believe Kai tested for Intel platforms.
Reviewed-by: Pierre-Louis Bossart pierre-louis.bossart@linux.intel.com
Note that the code could be simplified further with a for_each_rtd_components_rollback() macro, and we could also simplify the hw_free code below - the use of the 'last' pointer is not really necessary since we already have an index. That's for another cleanup though.
static int soc_pcm_components_hw_free(struct snd_pcm_substream *substream, struct snd_soc_component *last) { struct snd_soc_pcm_runtime *rtd = substream->private_data; struct snd_soc_component *component; int i, r, ret = 0;
for_each_rtd_components(rtd, i, component) { if (component == last) break;