linux-kernel - Re: [PATCH] uvc-gadget: Fix Set Interface (alternate setting) response behaviour

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <3028836.5DEGZBmlnn@avalon>
Date:   Mon, 06 Mar 2017 22:50:14 +0200
From:   Laurent Pinchart <laurent.pinchart@...asonboard.com>
To:     Roger Quadros <rogerq@...com>
Cc:     balbi@...nel.org, b-liu@...com, nsekhar@...com,
        linux-usb@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] uvc-gadget: Fix Set Interface (alternate setting) response behaviour

Hi Roger,

Thank you for the patch.

On Friday 03 Mar 2017 13:17:15 Roger Quadros wrote:
> On alternate setting change, webcam gadget sends us a UVC_EVENT_STREAMON
> or UVC_EVENT_STREAMOFF event. It expects delayed status response on
> STREAMON event only but doesn't expect us to send that response over USB.
> It sends the delayed response when we issue the VIDIOC_STREAMON ioctl.
>
> So we must not send UVCIOC_SEND_RESPONSE ioctl in these cases that too
> with invalid response length.

The commit message only explains why we should not call UVCIOC_SEND_RESPONSE 
in response to a STREAMON event, but not why we shouldn't either in response 
to a STREAMOFF event. The patch is correct changing both, but I propose 
wording the above two paragraphs as follows.

"uvc-gadget: Do not send Set Interface (alternate setting) response twice

On alternate setting change, the webcam gadget sends us a UVC_EVENT_STREAMON 
or UVC_EVENT_STREAMOFF event. In the first case, the driver will issue a 
delayed status response automatically when we call the VIDIOC_STREAMON ioctl. 
In the second case, the driver sends the status response immediately. We must 
thus not send the status response manually with UVCIOC_SEND_RESPONSE in any of 
those cases."

If you're fine with that I'll change the message when applying, there's no 
need to resend the patch.

> Without this, the ISO streaming doesn't work if host application
> (e.g. luvcview) is closed and restarted.
> On dwc3 gadget controller it was resulting in Buffer Expiry error on
> the ISO endpoint.
> 
> Signed-off-by: Roger Quadros <rogerq@...com>
> ---
>  uvc-gadget.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/uvc-gadget.c b/uvc-gadget.c
> index 9ef315c..4d59ab8 100644
> --- a/uvc-gadget.c
> +++ b/uvc-gadget.c
> @@ -597,12 +597,12 @@ uvc_events_process(struct uvc_device *dev)
>  	case UVC_EVENT_STREAMON:
>  		uvc_video_reqbufs(dev, 4);
>  		uvc_video_stream(dev, 1);
> -		break;
> +		return;
> 
>  	case UVC_EVENT_STREAMOFF:
>  		uvc_video_stream(dev, 0);
>  		uvc_video_reqbufs(dev, 0);
> -		break;
> +		return;
>  	}
> 
>  	ioctl(dev->fd, UVCIOC_SEND_RESPONSE, &resp);

-- 
Regards,

Laurent Pinchart