kernel/sound/soc/sof/sof-audio.c, branch linux-6.2.y

ASoC: SOF: sof-audio: prepare_widgets: Check swidget for NULL on sink failure

2023-01-20T14:09:56Z

If the swidget is NULL we skip the preparing of the widget and jump to handle the sink path of the widget. If the prepare fails in this case we would undo the prepare but the swidget is NULL (we skipped the prepare for the widget). To avoid NULL pointer dereference in this case we must check swidget against NULL pointer once again. Fixes: 0ad84b11f2f8 ("ASoC: SOF: sof-audio: skip prepare/unprepare if swidget is NULL") Reported-by: kernel test robot Reported-by: Dan Carpenter Signed-off-by: Peter Ujfalusi Reviewed-by: AngeloGioacchino Del Regno Link: https://lore.kernel.org/r/20230120102125.30653-1-peter.ujfalusi@linux.intel.com Signed-off-by: Mark Brown

ASoC: SOF: sof-audio: Fixes for widget prepare and

2023-01-18T15:28:17Z

Merge series from Peter Ujfalusi : This series contains one fix (first patch) followed by a nice to have safety belts in case we get a widget from topology which is not handled by SOF and will not have corresponding swidget associated with.

ASoC: SOF: keep prepare/unprepare widgets in sink path

2023-01-18T12:08:32Z

The existing code return when a widget doesn't need to prepare/unprepare. This will prevent widgets in the sink path from being prepared/unprepared. Cc: # 6.1 Link: https://github.com/thesofproject/linux/issues/4021 Signed-off-by: Bard Liao Reviewed-by: Ranjani Sridharan Reviewed-by: Pierre-Louis Bossart Reviewed-by: Rander Wang Signed-off-by: Peter Ujfalusi Link: https://lore.kernel.org/r/20230118101255.29139-4-peter.ujfalusi@linux.intel.com Signed-off-by: Mark Brown

ASoC: SOF: sof-audio: skip prepare/unprepare if swidget is NULL

2023-01-18T12:08:32Z

Skip preparing/unpreparing widgets if the swidget pointer is NULL. This will be true in the case of virtual widgets in topology that were added for reusing the legacy HDA machine driver with SOF. Fixes: 9862dcf70245 ("ASoC: SOF: don't unprepare widget used other pipelines") Cc: # 6.1 Signed-off-by: Ranjani Sridharan Signed-off-by: Peter Ujfalusi Reviewed-by: Rander Wang Reviewed-by: AngeloGioacchino Del Regno Tested-by: AngeloGioacchino Del Regno Signed-off-by: Peter Ujfalusi Link: https://lore.kernel.org/r/20230118101255.29139-3-peter.ujfalusi@linux.intel.com Signed-off-by: Mark Brown

ASoC: SOF: sof-audio: unprepare when swidget->use_count > 0

2023-01-18T12:08:31Z

We should unprepare the widget if its use_count = 1. Fixes: 9862dcf70245 ("ASoC: SOF: don't unprepare widget used other pipelines") Cc: # 6.1 Signed-off-by: Bard Liao Reviewed-by: Ranjani Sridharan Reviewed-by: Pierre-Louis Bossart Reviewed-by: Rander Wang Signed-off-by: Peter Ujfalusi Link: https://lore.kernel.org/r/20230118101255.29139-2-peter.ujfalusi@linux.intel.com Signed-off-by: Mark Brown

ASoC: SOF: sof-audio: start with the right widget type

2023-01-17T15:17:30Z

If there is a connection between a playback stream and a capture stream, all widgets that are connected to the playback stream and the capture stream will be in the list. So, we have to start with the exactly right widget type. snd_soc_dapm_aif_out is for capture stream and a playback stream should start with a snd_soc_dapm_aif_in widget. Contrarily, snd_soc_dapm_dai_in is for playback stream, and a capture stream should start with a snd_soc_dapm_dai_out widget. Signed-off-by: Bard Liao Reviewed-by: Ranjani Sridharan Reviewed-by: Kai Vehmanen Reviewed-by: Péter Ujfalusi Reviewed-by: Pierre-Louis Bossart Signed-off-by: Peter Ujfalusi Link: https://lore.kernel.org/r/20230117123534.2075-1-peter.ujfalusi@linux.intel.com Signed-off-by: Mark Brown

ASoC: SOF: introduce new DEBUG_NOCODEC mode

2022-10-28T12:05:01Z

The existing NOCODEC mode enforces a build-time mutual exclusion with the HDaudio link support, mostly to avoid any dependency on the snd_hdac library and references to HDAudio codec/i915 stuff. This is very useful to track dependencies and test a minimal configuration, but very painful for developers and CI: a recompilation and reinstall of the kernel modules is required. This patch suggests an alternate middle ground where the selection of the machine driver and all codec-related actions are bypassed at run-time, contingent on a kernel module parameter being set. For example setting BIT(10) with 'options snd_sof sof_debug=0x401' is enough to switch from an HDaudio card to a nocodec one. This new DEBUG_NOCODEC mode is not suitable for distributions and end-users. It's not even recommended on all platforms, i.e. the NOCODEC mode is known not to work on specific devices where the BIOS did not configure support for I2S/DMIC interfaces. The usual development devices such as Chromebooks, Up boards and Intel RVP are the only recommended platforms where this mode can be supported. Note that the dynamic switch between HDaudio and nocodec may not always possible depending on hardware layout, pin-mux options, and BIOS settings. The audio subsustems on Intel platforms has to support 4 types of interfaces and pin-mux can be complicated. Reviewers might ask: why didn't we do this earlier? The main reason is that all the codec-related configurations were not cleanly separated out in the sof/intel directory. With all the cleanups done recently, adding this opt-in behavior is relatively straightforward. Tested on UpExtreme (WHL) and UpExtreme i11 (TGL). Signed-off-by: Pierre-Louis Bossart Reviewed-by: Rander Wang Reviewed-by: Bard Liao Reviewed-by: Péter Ujfalusi Reviewed-by: Ranjani Sridharan Link: https://lore.kernel.org/r/20221027193540.259520-22-pierre-louis.bossart@linux.intel.com Signed-off-by: Mark Brown

ASoC: SOF: don't unprepare widget used other pipelines

2022-09-20T18:07:34Z

If multiple pipeline are mixed into one, we can't unprepare the widget used by other pipelines. This patch checks use_count to address this case. Reviewed-by: Bard Liao Reviewed-by: Ranjani Sridharan Signed-off-by: Rander Wang Signed-off-by: Pierre-Louis Bossart Link: https://lore.kernel.org/r/20220920150107.2090695-3-pierre-louis.bossart@linux.intel.com Signed-off-by: Mark Brown

ASoC: SOF: clear prepare state when widget is unprepared

2022-09-20T18:07:33Z

Playback can't work after the first try sometimes. The reason is that some widgets don't have ipc_unprepare ops and driver will jump to sink_prepare so miss to set prepare state to false. Next time these widgets will not be prepared and it will result to error with different format of audio file since the last setting is not applicable. This patch makes sure that widget prepare state will be cleared to false when it is unprepared. Reviewed-by: Bard Liao Reviewed-by: Ranjani Sridharan Signed-off-by: Rander Wang Signed-off-by: Pierre-Louis Bossart Link: https://lore.kernel.org/r/20220920150107.2090695-2-pierre-louis.bossart@linux.intel.com Signed-off-by: Mark Brown

ASoC: SOF: add widget setup/free tracing

2022-09-19T14:44:03Z

Enables tracking of use_count during widget setup and free routines. Useful for debugging unbalanced use_counts during suspend/resume. Reviewed-by: Péter Ujfalusi Signed-off-by: Noah Klayman Signed-off-by: Bard Liao Signed-off-by: Pierre-Louis Bossart Link: https://lore.kernel.org/r/20220919122108.43764-2-pierre-louis.bossart@linux.intel.com Signed-off-by: Mark Brown