On Thu, Sep 02, 2010 at 11:12:59AM +0200, Clemens Ladisch wrote:
Takashi Iwai wrote:
I myself find it's OK to apply since it's pretty local change. Clemens, what was your objection at that time?
No objection per se, I wanted to find out what descriptor field is changed by the quirk, and where the crashes come from exactly. The latter is still unclear, but it seems a quirk like this is required.
I've committed the patch now.
Also, the Interface Class protocol for the device defaults to Vendor Specific Class, but the driver only recognizes Class 1 or 2. The device is really Class 1, so we have to force it to reflect this. There are 4 places that we need to set this."
Daniel, this seems to be a regression introduced with the UAC2 changes. The old driver accepted vendor-specific interfaces, if some quirk forced it to look at them.
Hmm, I don't follow. What's the problem, and why should that be caused by the UAC2 changes? Sorry, but I seem to have missed that thread.
Thanks, Daniel