ALSA: hda: Fix page fault in snd_hda_codec_shutdown()
authorCezary Rojewski <cezary.rojewski@intel.com>
Wed, 6 Jul 2022 12:02:27 +0000 (14:02 +0200)
committerTakashi Iwai <tiwai@suse.de>
Sat, 9 Jul 2022 16:40:33 +0000 (18:40 +0200)
commit980b3a8790b402e959a6d773b38b771019682be1
treeb7b740a62850a5834c721888e61e98d445b8605a
parentebe043a3dfcade2756a41a3956365e1bfe4198cf
ALSA: hda: Fix page fault in snd_hda_codec_shutdown()

If early probe of HDAudio bus driver fails e.g.: due to missing
firmware file, snd_hda_codec_shutdown() ends in manipulating
uninitialized codec->pcm_list_head causing page fault.

Iinitialization of HDAudio codec in ASoC is split in two:
- snd_hda_codec_device_init()
- snd_hda_codec_device_new()

snd_hda_codec_device_init() is called during probe_codecs() by HDAudio
bus driver while snd_hda_codec_device_new() is called by
codec-component's ->probe(). The second call will not happen until all
components required by related sound card are present within the ASoC
framework. With firmware failing to load during the PCI's deferred
initialization i.e.: probe_work(), no platform components are ever
registered. HDAudio codec enumeration is done at that point though, so
the codec components became registered to ASoC framework, calling
snd_hda_codec_device_init() in the process.

Now, during platform reboot snd_hda_codec_shutdown() is called for every
codec found on the HDAudio bus causing oops if any of them has not
completed both of their initialization steps. Relocating field
initialization fixes the issue.

Signed-off-by: Cezary Rojewski <cezary.rojewski@intel.com>
Link: https://lore.kernel.org/r/20220706120230.427296-7-cezary.rojewski@intel.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
sound/pci/hda/hda_codec.c