On Wed, 13 Jul 2011, Ming Lei wrote:

> Hi,
> 
> On Tue, Jul 12, 2011 at 11:44 PM, Alan Stern <stern@rowland.harvard.edu> wrote:
> > On Tue, 12 Jul 2011, Ming Lei wrote:
> >
> >> Hi Laurent,
> >>
> >> After resume from sleep,  all the ISO packets from camera are like below:
> >>
> >> ffff880122d9f400 3527230728 S Zi:1:004:1 -115:1:2568 32 -18:0:1600
> >> -18:1600:1600 -18:3200:1600 -18:4800:1600 -18:6400:1600 51200 <
> >> ffff880122d9d400 3527234708 C Zi:1:004:1 0:1:2600:0 32 0:0:12
> >> 0:1600:12 0:3200:12 0:4800:12 0:6400:12 51200 = 0c8c0000 0000fa7e
> >> 012f1b05 00000000 00000000 00000000 00000000 00000000
> >>
> >> All are headed with 0c8c0000, see attached usbmon captures.
> >
> > Maybe this device needs a USB_QUIRK_RESET_RESUME entry in quirks.c.
> 
> I will try it, but seems unbind&bind driver don't produce extra usb reset signal
> to the device.
> 
> Also, the problem didn't happen in runtime pm case, just happen in
> wakeup from system suspend case. uvcvideo has enabled auto suspend
> already at default.

Why should system suspend be different from runtime suspend?  Have you
compared usbmon traces for the two types of suspend?

Alan Stern