ALSA: usb-audio: fix regression for fixed stream quirk

Commit 8f898e9 removed the redundant
reads of bInterfaceProtocol from the descriptors, but introduced a
regression to devices with quirks of type QUIRK_AUDIO_FIXED_ENDPOINT,
since fp->protocol is not set in setup process.

As a consequence, audio streams would not get initialized, as the
following logs show:

[   48.923043] setting usb interface 3:1
[   48.923056] Creating new capture data endpoint torvalds#81
[   48.923484] 4:3:1: cannot set freq 48000 to ep 0x81

This patch sets fp->protocol in create_fixed_stream_quirk() and
resolves the regression.

Signed-off-by: Eldad Zack <[email protected]>
Signed-off-by: Takashi Iwai <[email protected]>

sound/usb/quirks.c

@@ -129,6 +129,7 @@ static int create_fixed_stream_quirk(struct snd_usb_audio *chip,
 {
 	struct audioformat *fp;
 	struct usb_host_interface *alts;
+	struct usb_interface_descriptor *altsd;
 	int stream, err;
 	unsigned *rate_table = NULL;
 
@@ -166,6 +167,9 @@ static int create_fixed_stream_quirk(struct snd_usb_audio *chip,
 		return -EINVAL;
 	}
 	alts = &iface->altsetting[fp->altset_idx];
+	altsd = get_iface_desc(alts);
+	fp->protocol = altsd->bInterfaceProtocol;
+
 	if (fp->datainterval == 0)
 		fp->datainterval = snd_usb_parse_datainterval(chip, alts);
 	if (fp->maxpacksize == 0)