Re: [PATCH 3/3] ALSA: usb-audio: Properly match with audio interface class

[Takashi Iwai <tiwai@suse.de>]

On Sat, 22 Aug 2020 17:27:35 +0200,
Alexander Tsoy wrote:
> 
> В Пн, 17/08/2020 в 10:21 +0200, Takashi Iwai пишет:
> > There are a few entries in the quirk table that set the device ID
> > with
> > USB_DEVICE() macro while having an extra bInterfaceClass field.  But
> > bInterfaceClass field is never checked unless the proper match_flags
> > is set, so those may match incorrectly with all interfaces.
> > 
> > Introduce another macro to match with the vid/pid pair and the audio
> > class interface, and apply it to such entries, so that they can match
> > properly.
> > 
> > Signed-off-by: Takashi Iwai <tiwai@suse.de>
> > ---
> >  sound/usb/quirks-table.h | 55 +++++++++++++++++---------------------
> > ----------
> >  1 file changed, 19 insertions(+), 36 deletions(-)
> > 
> > diff --git a/sound/usb/quirks-table.h b/sound/usb/quirks-table.h
> > index 988bb9d00192..7a80ef31bbe4 100644
> > --- a/sound/usb/quirks-table.h
> > +++ b/sound/usb/quirks-table.h
> > @@ -35,6 +35,14 @@
> >  	.bInterfaceClass = USB_CLASS_AUDIO, \
> >  	.bInterfaceSubClass = USB_SUBCLASS_AUDIOCONTROL
> >  
> > +/* Another standard entry matching with vid/pid and the audio class
> > */
> > +#define USB_AUDIO_CLASS(vend, prod) \
> > +	.match_flags = USB_DEVICE_ID_MATCH_DEVICE | \
> > +		       USB_DEVICE_ID_MATCH_INT_CLASS, \
> > +	.idVendor = vend, \
> > +	.idProduct = prod, \
> > +	.bInterfaceClass = USB_CLASS_AUDIO
> > +
> >  /* FTDI devices */
> >  {
> >  	USB_DEVICE(0x0403, 0xb8d8),
> > @@ -68,34 +76,14 @@
> >  	}
> >  },
> >  
> > -{
> > -	/* E-Mu 0202 USB */
> > -	.match_flags = USB_DEVICE_ID_MATCH_DEVICE,
> > -	.idVendor = 0x041e,
> > -	.idProduct = 0x3f02,
> > -	.bInterfaceClass = USB_CLASS_AUDIO,
> > -},
> > -{
> > -	/* E-Mu 0404 USB */
> > -	.match_flags = USB_DEVICE_ID_MATCH_DEVICE,
> > -	.idVendor = 0x041e,
> > -	.idProduct = 0x3f04,
> > -	.bInterfaceClass = USB_CLASS_AUDIO,
> > -},
> > -{
> > -	/* E-Mu Tracker Pre */
> > -	.match_flags = USB_DEVICE_ID_MATCH_DEVICE,
> > -	.idVendor = 0x041e,
> > -	.idProduct = 0x3f0a,
> > -	.bInterfaceClass = USB_CLASS_AUDIO,
> > -},
> > -{
> > -	/* E-Mu 0204 USB */
> > -	.match_flags = USB_DEVICE_ID_MATCH_DEVICE,
> > -	.idVendor = 0x041e,
> > -	.idProduct = 0x3f19,
> > -	.bInterfaceClass = USB_CLASS_AUDIO,
> > -},
> > +/* E-Mu 0202 USB */
> > +{ USB_AUDIO_CLASS(0x041e, 0x3f02) },
> > +/* E-Mu 0404 USB */
> > +{ USB_AUDIO_CLASS(0x041e, 0x3f04) },
> > +/* E-Mu Tracker Pre */
> > +{ USB_AUDIO_CLASS(0x041e, 0x3f0a) },
> > +/* E-Mu 0204 USB */
> > +{ USB_AUDIO_CLASS(0x041e, 0x3f19) },
> >  
> >  /*
> >   * HP Wireless Audio
> > @@ -2751,10 +2739,7 @@ YAMAHA_DEVICE(0x7010, "UB99"),
> >  },
> >  
> >  /* KeithMcMillen Stringport */
> > -{
> > -	USB_DEVICE(0x1f38, 0x0001),
> > -	.bInterfaceClass = USB_CLASS_AUDIO,
> > -},
> > +{ USB_AUDIO_CLASS(0x1f38, 0x0001) },
> >  
> >  /* Miditech devices */
> >  {
> > @@ -2977,10 +2962,7 @@ AU0828_DEVICE(0x2040, 0x7270, "Hauppauge",
> > "HVR-950Q"),
> >  },
> >  {
> >  	/* Tascam US122 MKII - playback-only support */
> > -	.match_flags = USB_DEVICE_ID_MATCH_DEVICE,
> > -	.idVendor = 0x0644,
> > -	.idProduct = 0x8021,
> > -	.bInterfaceClass = USB_CLASS_AUDIO,
> > +	USB_AUDIO_CLASS(0x0644, 0x8021),
> >  	.driver_info = (unsigned long) &(const struct
> > snd_usb_audio_quirk) {
> >  		.vendor_name = "TASCAM",
> >  		.product_name = "US122 MKII",
> > @@ -3612,3 +3594,4 @@ AU0828_DEVICE(0x2040, 0x7270, "Hauppauge",
> > "HVR-950Q"),
> >  
> >  #undef USB_DEVICE_VENDOR_SPEC
> >  #undef USB_AUDIO_DEVICE
> > +#undef USB_AUDIO_CLASS
> 
> I don't know anything about KeithMcMillen Stringport, but all other
> devices (US122 MKII and E-mu) seems to have Vendor-specific Class.

OK, then it must be with USB_DEVICE_VENDOR_SPEC macro instead although
they had .bInterfaceClass = USB_CLASS_AUDIO line wrongly.
Could you submit the correction patch, or shall I fix in my side?

> And
> since the driver should already match all compliant devices, the only
> potentional use for USB_AUDIO_CLASS macro is for devices
> with bInterfaceClass = USB_CLASS_AUDIO and with some invalid
> bInterfaceSubClass.

The problem is that without the class match in addition to the device
id match, it would match too eagerly, e.g. if a USB device defines
multiple different classes with the same device id like HID and
video.  So we should restrict the matching as much as possible.


thanks,

Takashi