[PATCH][Natty SRU] UBUNTU: SAUCE: uvcvideo: add SetInterface(0) in .reset_resume handler

[Andy Whitcroft]

On Tue, Jul 26, 2011 at 11:39:56AM +0800, Ming Lei wrote:
> From 2f485a24a5732b980f2d1c7097469f07cb848c70 Mon Sep 17 00:00:00 2001
> From: Ming Lei <ming.lei at canonical.com>
> Date: Sat, 16 Jul 2011 11:29:45 +0800
> Subject: [PATCH][Natty SRU] UBUNTU: SAUCE: uvcvideo: add SetInterface(0) in .reset_resume handler
> 
> As commented in uvc_video_init,
> 
> 	/* Alternate setting 0 should be the default, yet the XBox Live Vision
> 	 * Cam (and possibly other devices) crash or otherwise misbehave if
> 	 * they don't receive a SET_INTERFACE request before any other video
> 	 * control request.
> 	 */
> 
> so it does make sense to add the SetInterface(0) in .reset_resume
> handler so that this kind of devices can work well if they are reseted
> during resume from system or runtime suspend.
> 
> We have found, without the patch, Microdia camera(0c45:6437) can't send
> stream data any longer after it is reseted during resume from
> system suspend.
> 
> SRU Justification:
> 
> Impact:
> 	- without the patch, the Microdia camera(0c45:6437) can't
> 	work after system resume
> 
> Fix:
>         - After applying the patch, the camera can work well after
> 	system resume
> 
> BugLink: http://bugs.launchpad.net/bugs/733509
> 
> upstream discusstion:  http://marc.info/?t=131037773100002&r=1&w=2
> 
> Signed-off-by: Ming Lei <ming.lei at canonical.com>
> Cc: Jeremy Kerr <jeremy.kerr at canonical.com>
> ---
> BTW: The patch will enter 3.1-rc later, not sure how Oneiric will merge
> the patch. 
> ---
>  drivers/media/video/uvc/uvc_driver.c |   14 +++++++++++++-
>  1 files changed, 13 insertions(+), 1 deletions(-)
> 
> diff --git a/drivers/media/video/uvc/uvc_driver.c b/drivers/media/video/uvc/uvc_driver.c
> index b6eae48..41c6d1a 100644
> --- a/drivers/media/video/uvc/uvc_driver.c
> +++ b/drivers/media/video/uvc/uvc_driver.c
> @@ -1959,8 +1959,20 @@ static int __uvc_resume(struct usb_interface *intf, int reset)
>  	}
>  
>  	list_for_each_entry(stream, &dev->streams, list) {
> -		if (stream->intf == intf)
> +		if (stream->intf == intf) {
> +			/*
> +			 * After usb bus reset, some devices may
> +			 * misbehave if SetInterface(0) is not done, for
> +			 * example, Microdia camera(0c45:6437) will stop
> +			 * sending streaming data. Looks like XBox Live
> +			 * Vision Cam needs it too, as commented in
> +			 * uvc_video_init.
> +			 */
> +			if (reset)
> +				usb_set_interface(stream->dev->udev,
> +					stream->intfnum, 0);
>  			return uvc_video_resume(stream);
> +		}
>  	}
>  
>  	uvc_trace(UVC_TRACE_SUSPEND, "Resume: video streaming USB interface "

That is going to need some careful testing as it is applied to all
devices.  It appears to be sane however and do what it claims and
upstream seems happy with the new behaviour.  As I say this will need
careful monitoring in testing.

I am unsure if this might make it to Oneiric via stable, but for our
purposes it might make sense to apply this there and see what fallout we
get before committing it for Natty.

Overally I guess:

Acked-by: Andy Whitcroft <apw at canonical.com>

-apw