[alsa-devel] usb 1-2: 1:1: cannot set freq 44100 to ep 0x81 and potential fix

[Takashi Iwai]

On Mon, 21 Aug 2017 10:43:39 +0200,
Joakim Tjernlund wrote:
> 
> On Sun, 2017-08-20 at 15:14 +0200, Joakim Tjernlund wrote:
> > I have two USB headsets (Logitech Inc Logitech Stereo H650e and
> > Jabra Jabra UC VOICE 550a) who both causes lots of:
> > 
> >     usb 1-2: 1:1: cannot set freq 44100 to ep 0x81
> > 
> > errors when using them for pidgin audio chats. The actual freq
> > varies though.
> > 
> > I found that adding a small delay here:
> >  diff -u sound/usb/clock.c.org  sound/usb/clock.c 
> > --- sound/usb/clock.c.org	2017-08-18 14:38:05.597977669 +0200
> > +++ sound/usb/clock.c	2017-08-18 14:20:14.653661513 +0200
> > @@ -304,6 +304,7 @@
> >  			iface, fmt->altsetting, rate, ep);
> >  		return err;
> >  	}
> > +        msleep(1);
> >  
> >  	/* Don't check the sample rate for devices which we know don't
> >  	 * support reading */
> > 
> > it seems one cannot do these settings too fast
> > Ant better idea for a fix? kernel 4.9.44
> > 
> >  Jocke
> > 
> > PS.
> >    When starting a chat with pidgin-sipe, I see the same sample rate
> > freq setting twice, not sure where that comes from.
> >  
> 
> I found snd_usb_ctl_msg_quirk() and if I add the same logic as
> for Zoom R16/24:
> +
> +       /* Logitech H650e needs a tiny delay here, otherwise requests like
> +        * get/set frequency return as failed despite actually succeeding.
> +        */
> +       if (chip->usb_id == USB_ID(0x046D, 0x0A46) &&
> +           (requesttype & USB_TYPE_MASK) == USB_TYPE_CLASS)
> +               mdelay(1);
> +
> +       /* Jambra 550a needs a tiny delay here, otherwise requests like
> +        * get/set frequency return as failed despite actually succeeding.
> +        */
> +       if (chip->usb_id == USB_ID(0x0B0E, 0x0349) &&
> +           (requesttype & USB_TYPE_MASK) == USB_TYPE_CLASS)
> +               mdelay(1);
> 
> it works again. Is the above an acceptable patch ?

Yes, looks OK.  Care to cook up the fix and submit?

After merging your fix, I'll clean up the quirk function to use
switch/case instead of multiple if.


thanks,

Takashi