Re: Kernel: snd-usb-audio: probe of 2-2:1.0 failed with error -5

Re: Kernel: snd-usb-audio: probe of 2-2:1.0 failed with error -5

[Linus Torvalds]

On Tue, Nov 29, 2011 at 4:52 AM, Nixbits Admin <admin@nixbits.org> wrote:
>
> While knowing that your time is precious and without wanting to making you
> spend more than what you may be able to spare for my problem; i would like
> ask you if you could take a look at this issue.
>
> https://forums.gentoo.org/viewtopic-p-6885908.html#6885908
>
> snd-usb-audio: probe of 2-2:1.0 failed with error -5

This is related to the line above it:

  Nov 29 07:21:21  kernel: skipping empty audio interface (v1)
  Nov 29 07:21:21  kernel: snd-usb-audio: probe of 2-3:1.0 failed with error -5

which in turn is because of a zero "bInCollection" value in the device
control header. I'm adding some sound people to the cc, in case
anybody has more clues about what's wrong with that particular device
- I don't personally have any real clue about usb audio.

As to your UHCI/EHCI confusion - it's actually a USB confusion. EHCI
was done as a add-on to UHCI (or OHCI), to the point where there is
both a UHCI *and* an EHCI controller involved (on the same silicon)
and they have this odd hand-over thing where one part (the UHCI part)
handles legacy devices, and then the new bolted-on EHCI part handles
the full USB-2.0 functionality.

So the UHCI/EHCI confusion is deep in the silicon itself. If you
attach a usb-1.1 device, it will be handed off to the UHCI part, if
you attach a 2.0 device it will be handled by the EHCI part of the
controller. Oh the wonders of crazy backwards compatibility hacks..

                                      Linus

Re: Kernel: snd-usb-audio: probe of 2-2:1.0 failed with error -5

[Daniel Mack]

On 11/29/2011 05:58 PM, Linus Torvalds wrote:
> On Tue, Nov 29, 2011 at 4:52 AM, Nixbits Admin <admin@nixbits.org> wrote:
>>
>> While knowing that your time is precious and without wanting to making you
>> spend more than what you may be able to spare for my problem; i would like
>> ask you if you could take a look at this issue.
>>
>> https://forums.gentoo.org/viewtopic-p-6885908.html#6885908
>>
>> snd-usb-audio: probe of 2-2:1.0 failed with error -5
> 
> This is related to the line above it:
> 
>   Nov 29 07:21:21  kernel: skipping empty audio interface (v1)
>   Nov 29 07:21:21  kernel: snd-usb-audio: probe of 2-3:1.0 failed with error -5
> 
> which in turn is because of a zero "bInCollection" value in the device
> control header. I'm adding some sound people to the cc, in case
> anybody has more clues about what's wrong with that particular device
> - I don't personally have any real clue about usb audio.

As the dump in the forum post doesn't decode any of the USB audio
specific headers, we'd need to see the output of "lsusb -v" with that
device connected.

If the driver bails out at this point, there's something wrong with the
descriptors, or the interface has no audio support at all (which would
be odd as it could have been omitted entirely in the first place).

We might have to add a quirk for the device, let's see.


Daniel

Re: Kernel: snd-usb-audio: probe of 2-2:1.0 failed with error -5

[Clemens Ladisch]

Daniel Mack wrote:
> On 11/29/2011 05:58 PM, Linus Torvalds wrote:
>> On Tue, Nov 29, 2011 at 4:52 AM, Nixbits Admin <admin@nixbits.org> wrote:
>>> https://forums.gentoo.org/viewtopic-p-6885908.html#6885908
>>
>>   Nov 29 07:21:21  kernel: skipping empty audio interface (v1)
>>   Nov 29 07:21:21  kernel: snd-usb-audio: probe of 2-3:1.0 failed with error -5
>
> If the driver bails out at this point, there's something wrong with the
> descriptors, or the interface has no audio support at all (which would
> be odd as it could have been omitted entirely in the first place).
>
> We might have to add a quirk for the device, let's see.

Actually, this error message could be prevented by removing the quirk
that is already there.

That quirk does nothing more than providing a nice user-visible name for
the device.  In doing so, it uses the USB_DEVICE macro, which makes the
USB core trying to attach it to all interfaces, including the video ones.
Because of all the crap descriptors out there, the audio driver trusts
the quirk, is not fazed by an interface not labeled as "audio", and
happily begins to parse the interface's descriptors.

The quirk should just be restricted to the audio interfaces (like all
the other Logitech quirks).


N.A., please try the patch below.


Regards,
Clemens


--- a/sound/usb/quirks-table.h
+++ b/sound/usb/quirks-table.h
@@ -157,7 +157,13 @@
 	.bInterfaceSubClass = USB_SUBCLASS_AUDIOCONTROL
 },
 {
-	USB_DEVICE(0x046d, 0x0990),
+	.match_flags = USB_DEVICE_ID_MATCH_DEVICE |
+		       USB_DEVICE_ID_MATCH_INT_CLASS |
+		       USB_DEVICE_ID_MATCH_INT_SUBCLASS,
+	.idVendor = 0x046d,
+	.idProduct = 0x0990,
+	.bInterfaceClass = USB_CLASS_AUDIO,
+	.bInterfaceSubClass = USB_SUBCLASS_AUDIOCONTROL
 	.driver_info = (unsigned long) & (const struct snd_usb_audio_quirk) {
 		.vendor_name = "Logitech, Inc.",
 		.product_name = "QuickCam Pro 9000",

Re: Kernel: snd-usb-audio: probe of 2-2:1.0 failed with error -5

[Linus Torvalds]

On Tue, Nov 29, 2011 at 11:09 AM, Clemens Ladisch <clemens@ladisch.de> wrote:
>
> N.A., please try the patch below.

I don't think that will compile. At a minimum, there's a missing comma
there, afaik.

                    Linus

Re: Kernel: snd-usb-audio: probe of 2-2:1.0 failed with error -5

[Clemens Ladisch]

Linus Torvalds wrote:
> On Tue, Nov 29, 2011 at 11:09 AM, Clemens Ladisch <clemens@ladisch.de> wrote:
>> N.A., please try the patch below.
>
> I don't think that will compile. At a minimum, there's a missing comma
> there, afaik.

Oops, indeed.  Now compile-tested:

--- a/sound/usb/quirks-table.h
+++ b/sound/usb/quirks-table.h
@@ -157,7 +157,13 @@
 	.bInterfaceSubClass = USB_SUBCLASS_AUDIOCONTROL
 },
 {
-	USB_DEVICE(0x046d, 0x0990),
+	.match_flags = USB_DEVICE_ID_MATCH_DEVICE |
+		       USB_DEVICE_ID_MATCH_INT_CLASS |
+		       USB_DEVICE_ID_MATCH_INT_SUBCLASS,
+	.idVendor = 0x046d,
+	.idProduct = 0x0990,
+	.bInterfaceClass = USB_CLASS_AUDIO,
+	.bInterfaceSubClass = USB_SUBCLASS_AUDIOCONTROL,
 	.driver_info = (unsigned long) & (const struct snd_usb_audio_quirk) {
 		.vendor_name = "Logitech, Inc.",
 		.product_name = "QuickCam Pro 9000",

Re: Kernel: snd-usb-audio: probe of 2-2:1.0 failed with error -5

[Nixbits Admin]

I have included the output of lsusb -v and updated the topic.

https://forums.gentoo.org/viewtopic-p-6886404.html#6886404


 >>    Nov 29 07:21:21  kernel: skipping empty audio interface (v1)
 >>    Nov 29 07:21:21  kernel: snd-usb-audio: probe of 2-3:1.0 failed 
with error -5

H.

On 11/29/2011 12:20 PM, Daniel Mack wrote:
> On 11/29/2011 05:58 PM, Linus Torvalds wrote:
>> On Tue, Nov 29, 2011 at 4:52 AM, Nixbits Admin<admin@nixbits.org>  wrote:
>>>
>>> While knowing that your time is precious and without wanting to making you
>>> spend more than what you may be able to spare for my problem; i would like
>>> ask you if you could take a look at this issue.
>>>
>>> https://forums.gentoo.org/viewtopic-p-6885908.html#6885908
>>>
>>> snd-usb-audio: probe of 2-2:1.0 failed with error -5
>>
>> This is related to the line above it:
>>
>>    Nov 29 07:21:21  kernel: skipping empty audio interface (v1)
>>    Nov 29 07:21:21  kernel: snd-usb-audio: probe of 2-3:1.0 failed with error -5
>>
>> which in turn is because of a zero "bInCollection" value in the device
>> control header. I'm adding some sound people to the cc, in case
>> anybody has more clues about what's wrong with that particular device
>> - I don't personally have any real clue about usb audio.
>
> As the dump in the forum post doesn't decode any of the USB audio
> specific headers, we'd need to see the output of "lsusb -v" with that
> device connected.
>
> If the driver bails out at this point, there's something wrong with the
> descriptors, or the interface has no audio support at all (which would
> be odd as it could have been omitted entirely in the first place).
>
> We might have to add a quirk for the device, let's see.
>
>
> Daniel
>