Re: [PATCH 1/2] ALSA: hda/intel: Power down GPU at error on azx_probe_continue()

[Takashi Iwai]

On Tue, 13 Oct 2026 15:31:50 +0200,
Takashi Iwai wrote:
> 
> Since the recent fix commit fbaccb1b19c7 ("ALSA: hda: intel: Don't
> free card in the probe work"), we don't release the card but just
> toggles the init_failed flag to stop the operation in the case of
> error at probing in azx_probe_continue().  This leaves, however,
> display_power() true, which may block powering down the GPU.
> 
> Put display_power(false) at the error path for avoiding the potential
> blockage.
> 
> Fixes: fbaccb1b19c7 ("ALSA: hda: intel: Don't free card in the probe work")
> Signed-off-by: Takashi Iwai <tiwai@xxxxxxx>
> ---
>  sound/hda/controllers/intel.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/sound/hda/controllers/intel.c b/sound/hda/controllers/intel.c
> index 6b80b59b8743..b67d569fc859 100644
> --- a/sound/hda/controllers/intel.c
> +++ b/sound/hda/controllers/intel.c
> @@ -2453,6 +2453,7 @@ static int azx_probe_continue(struct azx *chip, bool free_on_error)
>  	if (err < 0) {
>  		if (hda->vga_switcheroo_registered)
>  			dev_warn(&pci->dev, "GPU sound probed, but not operational: please add a quirk to driver_denylist\n");
> +		display_power(chip, false);
>  		if (free_on_error) {
>  			pci_set_drvdata(pci, NULL);
>  			snd_card_free(chip->card);

On the second thought, turning off the display power here alone would
be problematic; the rest hardware cleanup procedure will run anyway in
azx_free() and with the display_power() off, it may lead to something
unexpected.

We can power up again temporarily in azx_free(), too.  But I'm not
sure whether it's a good idea.  Going back to calling azx_free() can
still work, but then we'll hit the old issue again about the already
bound codecs (that aren't detached by azx_free() itself).

Hmm.  Maybe better to leave in the current state for now until someone
sees a real problem on an actual hardware.


Takashi