Re: [PATCH 0/2] USB gadget: Handle endpoint requests at the function level

"Robin Callender" <robin_callender@xxxxxxxxxxx> · Tue, 18 Aug 2009 14:07:23 -0700

----- Original Message ----- 
From: "Alan Stern" <stern@xxxxxxxxxxxxxxxxxxx>
To: "Robin Callender" <robin_callender@xxxxxxxxxxx>
Cc: "Laurent Pinchart" <laurent.pinchart@xxxxxxxxxxxxxxxx>; "linux-usb" 
<linux-usb@xxxxxxxxxxxxxxx>; "David Brownell" <david-b@xxxxxxxxxxx>; "Bryan 
Wu" <cooloney@xxxxxxxxxx>; "Mike Frysinger" <vapier@xxxxxxxxxx>
Sent: Tuesday, August 18, 2009 1:31 PM
Subject: Re: [PATCH 0/2] USB gadget: Handle endpoint requests at the 
function level

On Tue, 18 Aug 2009, Robin Callender wrote:

I agree, there is no wrong part (in the parsing sense) to the 
configuration
that I can discern.
But the device-to-host interactions do not progress to the next expected
interaction.
The next expected interaction is where the Endpoint CS
Get_Cur(SampleFrequency) is queried.
This would be the first action conducted over the configured endpoints.
There is something about this new "speaker+microphone" configuation which
windows-xp doesn't like, and is not sending Get_Cur(SampFreq).

Okay.  But it probably doesn't have any connection with the way the
endpoints are used.  Maybe your version of Windows can't handle
composite audio devies (that's just a wild guess; I have no experience
with USB audio devices and even less experience using them with
Windows).

Well, I have several other hardware-based usb audio devices which have 
composite configurations and they work without problems with the same 
winsows-xp system.
I have attached a configuration for a Turtle Beach adapter which has a 
composite audio configuration.

> It contains three interfaces.  Interface 0 doesn't use any endpoints,
> interface 1 uses endpoint 1 OUT, and interface 2 uses endpoint 2 IN.
> So what makes you think there's any sharing going on?

I have attached a kern.log section with debug spew showing what I am 
taking
as the sharing of the endpoint (circa line 7).
     Aug 18 10:53:01 ultraman kernel: [ 1177.261318] using net2280, OUT 
ep-a
IN ep-a
Perhaps I am misunderstanding this spew but it looks like OUT and IN are
both using ep-a.

Yes, it does look that way.  Evidently there is a bug somewhere.  But
the host wouldn't be aware of the bug -- not until it tried to transfer
some data.

I agree, but I see the lack of a tranfer as a symptom, not the root-cause of 
the problem.

BTW,  is there some ftp site I can upload these usb traces and configuration 
files?
They tend to a little large for attaching to mail.

Alan Stern

Attachment:
turtle_beach.pdf

Description: Adobe PDF document