Re: [PATCH] ALSA: hda: Fix SSID detection for HP Dragonfly Folio G3

[Takashi Iwai <tiwai@suse.de>]

On Thu, 14 May 2026 15:31:10 +0200,
Heiko Schmid wrote:
> 
> The HP Dragonfly Folio 13.5 inch G3 (PCI SSID 103c:8a05/8a06) BIOS does
> not program the PCI subsystem ID correctly when the SOF audio driver is
> used. This causes the codec fixup lookup to fail as the PCI SSID reads
> as 0x0000:0x0000 instead of the correct 103c:8a06.
> 
> Fix this by falling back to the codec SSID when the PCI SSID is zero,
> and add the device-specific quirk entry for the HP Dragonfly Folio G3.
> 
> Signed-off-by: Heiko Schmid <heiko@future-machines.org>

I think the idea is good, but we don't have to restrict to both PCI
vendor and device being 0; PCI device 0 is used for wildcard, and
can't work in anyway.

About the patch:

> ---
>  sound/hda/codecs/realtek/alc269.c |  2 ++
>  sound/hda/common/auto_parser.c    | 10 ++++++++++
>  2 files changed, 12 insertions(+)
> 
> diff --git a/sound/hda/codecs/realtek/alc269.c b/sound/hda/codecs/realtek/alc269.c
> index 22b2f67..494dd0a 100644
> --- a/sound/hda/codecs/realtek/alc269.c
> +++ b/sound/hda/codecs/realtek/alc269.c
> @@ -6990,6 +6990,8 @@ static const struct hda_quirk alc269_fixup_tbl[] = {
>  	SND_PCI_QUIRK(0x103c, 0x89ae, "HP EliteBook 650 G9", ALC236_FIXUP_HP_GPIO_LED),
>  	SND_PCI_QUIRK(0x103c, 0x89c0, "HP ZBook Power 15.6 G9", ALC245_FIXUP_CS35L41_SPI_2_HP_GPIO_LED),
>  	SND_PCI_QUIRK(0x103c, 0x8a06, "HP Dragonfly Folio 13.5 G3", ALC245_FIXUP_CS35L41_SPI_4_HP_GPIO_LED),
> +	SND_PCI_QUIRK(0x103c, 0x8a06, "HP Dragonfly Folio 13.5 G3", ALC245_FIXUP_CS35L41_SPI_4_HP_GPIO_LED),
> +	SND_PCI_QUIRK(0x103c, 0x8a06, "HP Dragonfly Folio 13.5 G3", ALC245_FIXUP_CS35L41_SPI_4_HP_GPIO_LED),

Those additions are simply superfluous, no?  Also...

> diff --git a/sound/hda/common/auto_parser.c b/sound/hda/common/auto_parser.c
> index 8923813..e876378 100644
> --- a/sound/hda/common/auto_parser.c
> +++ b/sound/hda/common/auto_parser.c
> @@ -1048,6 +1048,16 @@ void snd_hda_pick_fixup(struct hda_codec *codec,
>  	if (codec->bus->pci) {
>  		pci_vendor = codec->bus->pci->subsystem_vendor;
>  		pci_device = codec->bus->pci->subsystem_device;
> +		/* Fallback: use codec SSID if PCI SSID is zero (e.g. HP Dragonfly Folio G3 with SOF) */
> +		if (!pci_vendor && !pci_device) {
> +			pci_vendor = codec->core.subsystem_id >> 16;
> +			pci_device = codec->core.subsystem_id & 0xffff;
> +		}
> +		/* Fallback: use codec SSID if PCI SSID is zero (e.g. HP Dragonfly Folio G3 with SOF) */
> +		if (!pci_vendor && !pci_device) {
> +			pci_vendor = codec->core.subsystem_id >> 16;
> +			pci_device = codec->core.subsystem_id & 0xffff;
> +		}

... here checking twice?  The patch looks somehow odd-merged.

And, I think we can just skip the PCI matching when either vendor or
device is 0.  So, something like below instead.

Could you verify whether it works?


Takashi

diff --git a/sound/hda/common/auto_parser.c b/sound/hda/common/auto_parser.c
index 8923813ce424..5bc95d3116ff 100644
--- a/sound/hda/common/auto_parser.c
+++ b/sound/hda/common/auto_parser.c
@@ -1013,7 +1013,7 @@ void snd_hda_pick_fixup(struct hda_codec *codec,
 	const char *name = NULL;
 	const char *type = NULL;
 	unsigned int vendor, device;
-	u16 pci_vendor, pci_device;
+	u16 pci_vendor = 0, pci_device = 0;
 	u16 codec_vendor, codec_device;
 
 	if (codec->fixup_id != HDA_FIXUP_ID_NOT_SET)
@@ -1066,7 +1066,7 @@ void snd_hda_pick_fixup(struct hda_codec *codec,
 	/* match primarily with the PCI SSID */
 	for (q = quirk; q->subvendor || q->subdevice; q++) {
 		/* if the entry is specific to codec SSID, check with it */
-		if (!codec->bus->pci || q->match_codec_ssid) {
+		if (!pci_vendor || !pci_device || q->match_codec_ssid) {
 			if (hda_quirk_match(codec_vendor, codec_device, q)) {
 				type = "codec SSID";
 				goto found_device;