Re: [PATCH] ALSA: usb-audio: validate full match when resolving quirk aliases

[Takashi Iwai <tiwai@suse.de>]

On Sat, 14 Mar 2026 04:34:46 +0100,
Cássio Gabriel wrote:
> 
> get_alias_quirk() resolves a quirk for an aliased USB ID by scanning
> usb_audio_ids[], but it currently checks only the vendor/product pair.
> 
> This can make an aliased ID pick the first entry with a matching
> vid:pid even when that entry also depends on interface descriptor
> fields that do not match the actual device or interface.
> 
> Fix it by re-checking each aliased candidate with usb_match_one_id()
> against the real interface before returning the quirk.
> 
> Signed-off-by: Cássio Gabriel <cassiogabrielcontato@gmail.com>
> ---
>  sound/usb/card.c | 21 ++++++++++++++++-----
>  1 file changed, 16 insertions(+), 5 deletions(-)
> 
> diff --git a/sound/usb/card.c b/sound/usb/card.c
> index 270dad84d825..ff4418017763 100644
> --- a/sound/usb/card.c
> +++ b/sound/usb/card.c
> @@ -870,17 +870,28 @@ static void find_last_interface(struct snd_usb_audio *chip)
>  
>  /* look for the corresponding quirk */
>  static const struct snd_usb_audio_quirk *
> -get_alias_quirk(struct usb_device *dev, unsigned int id)
> +get_alias_quirk(struct usb_interface *intf, unsigned int id)
>  {
>  	const struct usb_device_id *p;
> +	struct usb_device_id id_alias;
>  
>  	for (p = usb_audio_ids; p->match_flags; p++) {
> -		/* FIXME: this checks only vendor:product pair in the list */
>  		if ((p->match_flags & USB_DEVICE_ID_MATCH_DEVICE) ==
>  		    USB_DEVICE_ID_MATCH_DEVICE &&
>  		    p->idVendor == USB_ID_VENDOR(id) &&
> -		    p->idProduct == USB_ID_PRODUCT(id))
> -			return (const struct snd_usb_audio_quirk *)p->driver_info;
> +		    p->idProduct == USB_ID_PRODUCT(id)) {
> +			/*
> +			 * Re-check the aliased entry against the actual
> +			 * interface descriptors instead of matching only
> +			 * the vendor/product pair.
> +			 */
> +			id_alias = *p;
> +			id_alias.idVendor = USB_ID_VENDOR(id);
> +			id_alias.idProduct = USB_ID_PRODUCT(id);
> +
> +			if (usb_match_one_id(intf, &id_alias))

Hmm, is this really a correct logic?

In this case, USB_ID_VENDOR(id) and USB_ID_PRODUCT(id) are very same
as p->idVendor and p->idProduct, hence id_alias and *p are identical,
i.e. you're basically comparing against *p.

I suppose you wanted to substitute the original device vendor/product
IDs there instead?


thanks,

Takashi