[v3,2/3] media: uvcvideo: Do not halt the device after disconnect

Message ID 20231121-guenter-mini-v3-2-d8a5eae2312b@chromium.org
State New
Headers
Series uvcvideo: Attempt N to land UVC race conditions fixes |

Commit Message

Ricardo Ribalda Nov. 21, 2023, 7:53 p.m. UTC
  usb drivers should not call to any usb_() function after the
.disconnect() callback has been triggered.

If the camera is streaming, the uvc driver will call usb_set_interface or
usb_clear_halt once the device is being released. Let's fix this issue.

This is probably not the only driver affected with this kind of bug, but
until there is a better way to do it in the core this is the way to
solve this issue.

When/if a different mechanism is implemented in the core to solve the
lifetime of devices we will adopt it in uvc.

Trace:
[ 1065.389723] drivers/media/usb/uvc/uvc_driver.c:2248 uvc_disconnect enter
[ 1065.390160] drivers/media/usb/uvc/uvc_driver.c:2264 uvc_disconnect exit
[ 1065.433956] drivers/media/usb/uvc/uvc_v4l2.c:659 uvc_v4l2_release enter
[ 1065.433973] drivers/media/usb/uvc/uvc_video.c:2274 uvc_video_stop_streaming enter
[ 1065.434560] drivers/media/usb/uvc/uvc_video.c:2285 uvc_video_stop_streaming exit
[ 1065.435154] drivers/media/usb/uvc/uvc_v4l2.c:680 uvc_v4l2_release exit
[ 1065.435188] drivers/media/usb/uvc/uvc_driver.c:2248 uvc_disconnect enter

Signed-off-by: Ricardo Ribalda <ribalda@chromium.org>
---
 drivers/media/usb/uvc/uvc_driver.c |  2 ++
 drivers/media/usb/uvc/uvc_video.c  | 45 ++++++++++++++++++++++++--------------
 drivers/media/usb/uvc/uvcvideo.h   |  2 ++
 3 files changed, 32 insertions(+), 17 deletions(-)
  

Comments

Sergey Senozhatsky Nov. 22, 2023, 7:47 a.m. UTC | #1
On (23/11/21 19:53), Ricardo Ribalda wrote:
> diff --git a/drivers/media/usb/uvc/uvc_driver.c b/drivers/media/usb/uvc/uvc_driver.c
> index 08fcd2ffa727..413c32867617 100644
> --- a/drivers/media/usb/uvc/uvc_driver.c
> +++ b/drivers/media/usb/uvc/uvc_driver.c
> @@ -2257,6 +2257,8 @@ static void uvc_disconnect(struct usb_interface *intf)
>  		return;
>  
>  	uvc_unregister_video(dev);
> +	/* Barrier needed to synchronize with uvc_video_stop_streaming(). */
> +	smp_store_release(&dev->disconnected, true);
>  	kref_put(&dev->ref, uvc_delete);
>  }

[..]

> +void uvc_video_stop_streaming(struct uvc_streaming *stream)
> +{
> +	uvc_video_stop_transfer(stream, 1);
> +
> +	/*
> +	 * Barrier needed to synchronize with uvc_disconnect().
> +	 * We cannot call usb_* functions on a disconnected USB device.
> +	 */
> +	if (!smp_load_acquire(&stream->dev->disconnected))
> +		uvc_video_halt(stream);
> +
>  	uvc_video_clock_cleanup(stream);
>  }

Can the following happen?

CPU0                                            CPU1
 uvc_disconnect()
						uvc_video_stop_streaming()
 usb_set_intfdata()
 uvc_unregister_video()

						if (!smp_load(&dev->disconnected))
							uvc_video_halt()

 smp_store_release(&dev->disconnected, true);

 kref_put(&dev->ref, uvc_delete);
  
Sergey Senozhatsky Nov. 22, 2023, 8:01 a.m. UTC | #2
On (23/11/22 16:47), Sergey Senozhatsky wrote:
> Can the following happen?

Consider the following case (when CPU1 experienced a delay, a preemption
or anything):

> CPU0                                            CPU1
>  uvc_disconnect()
> 						uvc_video_stop_streaming()
>  usb_set_intfdata()
>  uvc_unregister_video()
> 
> 						if (!smp_load(&dev->disconnected))
> 
>  smp_store_release(&dev->disconnected, true);
> 
>  kref_put(&dev->ref, uvc_delete);

> 							uvc_video_halt()

That uvc_video_halt() cannot be legal, right?
  
Ricardo Ribalda Nov. 22, 2023, 9:55 a.m. UTC | #3
Hi Sergey

On Wed, 22 Nov 2023 at 09:01, Sergey Senozhatsky
<senozhatsky@chromium.org> wrote:
>
> On (23/11/22 16:47), Sergey Senozhatsky wrote:
> > Can the following happen?
>
> Consider the following case (when CPU1 experienced a delay, a preemption
> or anything):
>
> > CPU0                                            CPU1
> >  uvc_disconnect()
> >                                               uvc_video_stop_streaming()
> >  usb_set_intfdata()
> >  uvc_unregister_video()
> >
> >                                               if (!smp_load(&dev->disconnected))
> >
> >  smp_store_release(&dev->disconnected, true);
> >
> >  kref_put(&dev->ref, uvc_delete);
>
> >                                                       uvc_video_halt()
>
> That uvc_video_halt() cannot be legal, right?

This patch only takes care of calls to uvc_video_stop_streaming()
after .disconnect.

Guenter's patch from this series should take care of the concurrent
calls. I will resend making it explicit.

Thanks!
  
Sakari Ailus Nov. 22, 2023, 10:25 a.m. UTC | #4
Hi Ricardo,

Thank you for the patch.

On Tue, Nov 21, 2023 at 07:53:49PM +0000, Ricardo Ribalda wrote:
> usb drivers should not call to any usb_() function after the
> .disconnect() callback has been triggered.
> 
> If the camera is streaming, the uvc driver will call usb_set_interface or
> usb_clear_halt once the device is being released. Let's fix this issue.
> 
> This is probably not the only driver affected with this kind of bug, but
> until there is a better way to do it in the core this is the way to
> solve this issue.
> 
> When/if a different mechanism is implemented in the core to solve the
> lifetime of devices we will adopt it in uvc.
> 
> Trace:
> [ 1065.389723] drivers/media/usb/uvc/uvc_driver.c:2248 uvc_disconnect enter
> [ 1065.390160] drivers/media/usb/uvc/uvc_driver.c:2264 uvc_disconnect exit
> [ 1065.433956] drivers/media/usb/uvc/uvc_v4l2.c:659 uvc_v4l2_release enter
> [ 1065.433973] drivers/media/usb/uvc/uvc_video.c:2274 uvc_video_stop_streaming enter
> [ 1065.434560] drivers/media/usb/uvc/uvc_video.c:2285 uvc_video_stop_streaming exit
> [ 1065.435154] drivers/media/usb/uvc/uvc_v4l2.c:680 uvc_v4l2_release exit
> [ 1065.435188] drivers/media/usb/uvc/uvc_driver.c:2248 uvc_disconnect enter
> 
> Signed-off-by: Ricardo Ribalda <ribalda@chromium.org>
> ---
>  drivers/media/usb/uvc/uvc_driver.c |  2 ++
>  drivers/media/usb/uvc/uvc_video.c  | 45 ++++++++++++++++++++++++--------------
>  drivers/media/usb/uvc/uvcvideo.h   |  2 ++
>  3 files changed, 32 insertions(+), 17 deletions(-)
> 
> diff --git a/drivers/media/usb/uvc/uvc_driver.c b/drivers/media/usb/uvc/uvc_driver.c
> index 08fcd2ffa727..413c32867617 100644
> --- a/drivers/media/usb/uvc/uvc_driver.c
> +++ b/drivers/media/usb/uvc/uvc_driver.c
> @@ -2257,6 +2257,8 @@ static void uvc_disconnect(struct usb_interface *intf)
>  		return;
>  
>  	uvc_unregister_video(dev);
> +	/* Barrier needed to synchronize with uvc_video_stop_streaming(). */
> +	smp_store_release(&dev->disconnected, true);
>  	kref_put(&dev->ref, uvc_delete);
>  }
>  
> diff --git a/drivers/media/usb/uvc/uvc_video.c b/drivers/media/usb/uvc/uvc_video.c
> index 28dde08ec6c5..032b44e45b22 100644
> --- a/drivers/media/usb/uvc/uvc_video.c
> +++ b/drivers/media/usb/uvc/uvc_video.c
> @@ -2243,28 +2243,39 @@ int uvc_video_start_streaming(struct uvc_streaming *stream)
>  	return ret;
>  }
>  
> -void uvc_video_stop_streaming(struct uvc_streaming *stream)
> +static void uvc_video_halt(struct uvc_streaming *stream)
>  {
> -	uvc_video_stop_transfer(stream, 1);
> +	unsigned int epnum;
> +	unsigned int pipe;
> +	unsigned int dir;
>  
>  	if (stream->intf->num_altsetting > 1) {

Doesn't this imply the device is using isochronous mode?

>  		usb_set_interface(stream->dev->udev, stream->intfnum, 0);
> -	} else {
> -		/*
> -		 * UVC doesn't specify how to inform a bulk-based device
> -		 * when the video stream is stopped. Windows sends a
> -		 * CLEAR_FEATURE(HALT) request to the video streaming
> -		 * bulk endpoint, mimic the same behaviour.
> -		 */
> -		unsigned int epnum = stream->header.bEndpointAddress
> -				   & USB_ENDPOINT_NUMBER_MASK;
> -		unsigned int dir = stream->header.bEndpointAddress
> -				 & USB_ENDPOINT_DIR_MASK;
> -		unsigned int pipe;
> -
> -		pipe = usb_sndbulkpipe(stream->dev->udev, epnum) | dir;
> -		usb_clear_halt(stream->dev->udev, pipe);
> +		return;
>  	}
>  
> +	/*
> +	 * UVC doesn't specify how to inform a bulk-based device

Then this comment doesn't look right. What about the code? This isn't
mentioned in the commit message either.

> +	 * when the video stream is stopped. Windows sends a
> +	 * CLEAR_FEATURE(HALT) request to the video streaming
> +	 * bulk endpoint, mimic the same behaviour.
> +	 */
> +	epnum = stream->header.bEndpointAddress & USB_ENDPOINT_NUMBER_MASK;
> +	dir = stream->header.bEndpointAddress & USB_ENDPOINT_DIR_MASK;
> +	pipe = usb_sndbulkpipe(stream->dev->udev, epnum) | dir;
> +	usb_clear_halt(stream->dev->udev, pipe);
> +}
> +
> +void uvc_video_stop_streaming(struct uvc_streaming *stream)
> +{
> +	uvc_video_stop_transfer(stream, 1);
> +
> +	/*
> +	 * Barrier needed to synchronize with uvc_disconnect().
> +	 * We cannot call usb_* functions on a disconnected USB device.
> +	 */
> +	if (!smp_load_acquire(&stream->dev->disconnected))
> +		uvc_video_halt(stream);
> +
>  	uvc_video_clock_cleanup(stream);
>  }
> diff --git a/drivers/media/usb/uvc/uvcvideo.h b/drivers/media/usb/uvc/uvcvideo.h
> index 6fb0a78b1b00..4318ce8e31db 100644
> --- a/drivers/media/usb/uvc/uvcvideo.h
> +++ b/drivers/media/usb/uvc/uvcvideo.h
> @@ -559,6 +559,8 @@ struct uvc_device {
>  	unsigned int users;
>  	atomic_t nmappings;
>  
> +	bool disconnected;
> +
>  	/* Video control interface */
>  #ifdef CONFIG_MEDIA_CONTROLLER
>  	struct media_device mdev;
>
  
Ricardo Ribalda Nov. 22, 2023, 10:32 a.m. UTC | #5
Hi Sakari

On Wed, 22 Nov 2023 at 11:25, Sakari Ailus <sakari.ailus@iki.fi> wrote:
>
> Hi Ricardo,
>
> Thank you for the patch.
>
> On Tue, Nov 21, 2023 at 07:53:49PM +0000, Ricardo Ribalda wrote:
> > usb drivers should not call to any usb_() function after the
> > .disconnect() callback has been triggered.
> >
> > If the camera is streaming, the uvc driver will call usb_set_interface or
> > usb_clear_halt once the device is being released. Let's fix this issue.
> >
> > This is probably not the only driver affected with this kind of bug, but
> > until there is a better way to do it in the core this is the way to
> > solve this issue.
> >
> > When/if a different mechanism is implemented in the core to solve the
> > lifetime of devices we will adopt it in uvc.
> >
> > Trace:
> > [ 1065.389723] drivers/media/usb/uvc/uvc_driver.c:2248 uvc_disconnect enter
> > [ 1065.390160] drivers/media/usb/uvc/uvc_driver.c:2264 uvc_disconnect exit
> > [ 1065.433956] drivers/media/usb/uvc/uvc_v4l2.c:659 uvc_v4l2_release enter
> > [ 1065.433973] drivers/media/usb/uvc/uvc_video.c:2274 uvc_video_stop_streaming enter
> > [ 1065.434560] drivers/media/usb/uvc/uvc_video.c:2285 uvc_video_stop_streaming exit
> > [ 1065.435154] drivers/media/usb/uvc/uvc_v4l2.c:680 uvc_v4l2_release exit
> > [ 1065.435188] drivers/media/usb/uvc/uvc_driver.c:2248 uvc_disconnect enter
> >
> > Signed-off-by: Ricardo Ribalda <ribalda@chromium.org>
> > ---
> >  drivers/media/usb/uvc/uvc_driver.c |  2 ++
> >  drivers/media/usb/uvc/uvc_video.c  | 45 ++++++++++++++++++++++++--------------
> >  drivers/media/usb/uvc/uvcvideo.h   |  2 ++
> >  3 files changed, 32 insertions(+), 17 deletions(-)
> >
> > diff --git a/drivers/media/usb/uvc/uvc_driver.c b/drivers/media/usb/uvc/uvc_driver.c
> > index 08fcd2ffa727..413c32867617 100644
> > --- a/drivers/media/usb/uvc/uvc_driver.c
> > +++ b/drivers/media/usb/uvc/uvc_driver.c
> > @@ -2257,6 +2257,8 @@ static void uvc_disconnect(struct usb_interface *intf)
> >               return;
> >
> >       uvc_unregister_video(dev);
> > +     /* Barrier needed to synchronize with uvc_video_stop_streaming(). */
> > +     smp_store_release(&dev->disconnected, true);
> >       kref_put(&dev->ref, uvc_delete);
> >  }
> >
> > diff --git a/drivers/media/usb/uvc/uvc_video.c b/drivers/media/usb/uvc/uvc_video.c
> > index 28dde08ec6c5..032b44e45b22 100644
> > --- a/drivers/media/usb/uvc/uvc_video.c
> > +++ b/drivers/media/usb/uvc/uvc_video.c
> > @@ -2243,28 +2243,39 @@ int uvc_video_start_streaming(struct uvc_streaming *stream)
> >       return ret;
> >  }
> >
> > -void uvc_video_stop_streaming(struct uvc_streaming *stream)
> > +static void uvc_video_halt(struct uvc_streaming *stream)
> >  {
> > -     uvc_video_stop_transfer(stream, 1);
> > +     unsigned int epnum;
> > +     unsigned int pipe;
> > +     unsigned int dir;
> >
> >       if (stream->intf->num_altsetting > 1) {
>
> Doesn't this imply the device is using isochronous mode?

I haven't changed the behaviour for halt, it is just that git diff is
being a bit too creative here:

Basically it is doing:

void video_halt() {
   if (is_isoc()) {
     usb_set_interface(stream->dev->udev, stream->intfnum, 0);
     return;
   }
   usb_clear_halt();
}

instead of the old:

void video_halt() {
   if (is_isoc()) {
     usb_set_interface(stream->dev->udev, stream->intfnum, 0);
   }  else {
      usb_clear_halt();
   }
}

Thanks!
>
> >               usb_set_interface(stream->dev->udev, stream->intfnum, 0);
> > -     } else {
> > -             /*
> > -              * UVC doesn't specify how to inform a bulk-based device
> > -              * when the video stream is stopped. Windows sends a
> > -              * CLEAR_FEATURE(HALT) request to the video streaming
> > -              * bulk endpoint, mimic the same behaviour.
> > -              */
> > -             unsigned int epnum = stream->header.bEndpointAddress
> > -                                & USB_ENDPOINT_NUMBER_MASK;
> > -             unsigned int dir = stream->header.bEndpointAddress
> > -                              & USB_ENDPOINT_DIR_MASK;
> > -             unsigned int pipe;
> > -
> > -             pipe = usb_sndbulkpipe(stream->dev->udev, epnum) | dir;
> > -             usb_clear_halt(stream->dev->udev, pipe);
> > +             return;
> >       }
> >
> > +     /*
> > +      * UVC doesn't specify how to inform a bulk-based device
>
> Then this comment doesn't look right. What about the code? This isn't
> mentioned in the commit message either.
>
> > +      * when the video stream is stopped. Windows sends a
> > +      * CLEAR_FEATURE(HALT) request to the video streaming
> > +      * bulk endpoint, mimic the same behaviour.
> > +      */
> > +     epnum = stream->header.bEndpointAddress & USB_ENDPOINT_NUMBER_MASK;
> > +     dir = stream->header.bEndpointAddress & USB_ENDPOINT_DIR_MASK;
> > +     pipe = usb_sndbulkpipe(stream->dev->udev, epnum) | dir;
> > +     usb_clear_halt(stream->dev->udev, pipe);
> > +}
> > +
> > +void uvc_video_stop_streaming(struct uvc_streaming *stream)
> > +{
> > +     uvc_video_stop_transfer(stream, 1);
> > +
> > +     /*
> > +      * Barrier needed to synchronize with uvc_disconnect().
> > +      * We cannot call usb_* functions on a disconnected USB device.
> > +      */
> > +     if (!smp_load_acquire(&stream->dev->disconnected))
> > +             uvc_video_halt(stream);
> > +
> >       uvc_video_clock_cleanup(stream);
> >  }
> > diff --git a/drivers/media/usb/uvc/uvcvideo.h b/drivers/media/usb/uvc/uvcvideo.h
> > index 6fb0a78b1b00..4318ce8e31db 100644
> > --- a/drivers/media/usb/uvc/uvcvideo.h
> > +++ b/drivers/media/usb/uvc/uvcvideo.h
> > @@ -559,6 +559,8 @@ struct uvc_device {
> >       unsigned int users;
> >       atomic_t nmappings;
> >
> > +     bool disconnected;
> > +
> >       /* Video control interface */
> >  #ifdef CONFIG_MEDIA_CONTROLLER
> >       struct media_device mdev;
> >
>
> --
> Kind regards,
>
> Sakari Ailus
  
Sakari Ailus Nov. 22, 2023, 11:03 a.m. UTC | #6
Hi Ricardo,

On Wed, Nov 22, 2023 at 11:32:16AM +0100, Ricardo Ribalda wrote:
> Hi Sakari
> 
> On Wed, 22 Nov 2023 at 11:25, Sakari Ailus <sakari.ailus@iki.fi> wrote:
> >
> > Hi Ricardo,
> >
> > Thank you for the patch.
> >
> > On Tue, Nov 21, 2023 at 07:53:49PM +0000, Ricardo Ribalda wrote:
> > > usb drivers should not call to any usb_() function after the
> > > .disconnect() callback has been triggered.
> > >
> > > If the camera is streaming, the uvc driver will call usb_set_interface or
> > > usb_clear_halt once the device is being released. Let's fix this issue.
> > >
> > > This is probably not the only driver affected with this kind of bug, but
> > > until there is a better way to do it in the core this is the way to
> > > solve this issue.
> > >
> > > When/if a different mechanism is implemented in the core to solve the
> > > lifetime of devices we will adopt it in uvc.
> > >
> > > Trace:
> > > [ 1065.389723] drivers/media/usb/uvc/uvc_driver.c:2248 uvc_disconnect enter
> > > [ 1065.390160] drivers/media/usb/uvc/uvc_driver.c:2264 uvc_disconnect exit
> > > [ 1065.433956] drivers/media/usb/uvc/uvc_v4l2.c:659 uvc_v4l2_release enter
> > > [ 1065.433973] drivers/media/usb/uvc/uvc_video.c:2274 uvc_video_stop_streaming enter
> > > [ 1065.434560] drivers/media/usb/uvc/uvc_video.c:2285 uvc_video_stop_streaming exit
> > > [ 1065.435154] drivers/media/usb/uvc/uvc_v4l2.c:680 uvc_v4l2_release exit
> > > [ 1065.435188] drivers/media/usb/uvc/uvc_driver.c:2248 uvc_disconnect enter
> > >
> > > Signed-off-by: Ricardo Ribalda <ribalda@chromium.org>
> > > ---
> > >  drivers/media/usb/uvc/uvc_driver.c |  2 ++
> > >  drivers/media/usb/uvc/uvc_video.c  | 45 ++++++++++++++++++++++++--------------
> > >  drivers/media/usb/uvc/uvcvideo.h   |  2 ++
> > >  3 files changed, 32 insertions(+), 17 deletions(-)
> > >
> > > diff --git a/drivers/media/usb/uvc/uvc_driver.c b/drivers/media/usb/uvc/uvc_driver.c
> > > index 08fcd2ffa727..413c32867617 100644
> > > --- a/drivers/media/usb/uvc/uvc_driver.c
> > > +++ b/drivers/media/usb/uvc/uvc_driver.c
> > > @@ -2257,6 +2257,8 @@ static void uvc_disconnect(struct usb_interface *intf)
> > >               return;
> > >
> > >       uvc_unregister_video(dev);
> > > +     /* Barrier needed to synchronize with uvc_video_stop_streaming(). */
> > > +     smp_store_release(&dev->disconnected, true);
> > >       kref_put(&dev->ref, uvc_delete);
> > >  }
> > >
> > > diff --git a/drivers/media/usb/uvc/uvc_video.c b/drivers/media/usb/uvc/uvc_video.c
> > > index 28dde08ec6c5..032b44e45b22 100644
> > > --- a/drivers/media/usb/uvc/uvc_video.c
> > > +++ b/drivers/media/usb/uvc/uvc_video.c
> > > @@ -2243,28 +2243,39 @@ int uvc_video_start_streaming(struct uvc_streaming *stream)
> > >       return ret;
> > >  }
> > >
> > > -void uvc_video_stop_streaming(struct uvc_streaming *stream)
> > > +static void uvc_video_halt(struct uvc_streaming *stream)
> > >  {
> > > -     uvc_video_stop_transfer(stream, 1);
> > > +     unsigned int epnum;
> > > +     unsigned int pipe;
> > > +     unsigned int dir;
> > >
> > >       if (stream->intf->num_altsetting > 1) {
> >
> > Doesn't this imply the device is using isochronous mode?
> 
> I haven't changed the behaviour for halt, it is just that git diff is
> being a bit too creative here:
> 
> Basically it is doing:
> 
> void video_halt() {
>    if (is_isoc()) {
>      usb_set_interface(stream->dev->udev, stream->intfnum, 0);
>      return;
>    }
>    usb_clear_halt();
> }
> 
> instead of the old:
> 
> void video_halt() {
>    if (is_isoc()) {
>      usb_set_interface(stream->dev->udev, stream->intfnum, 0);
>    }  else {
>       usb_clear_halt();
>    }
> }
> 
> Thanks!

Oops. I missed the removal of the else branch altogether while reading the
patch.

Shouldn't you also ensure the disconnect callback won't return until the
streaming has been stopped here?

> >
> > >               usb_set_interface(stream->dev->udev, stream->intfnum, 0);
> > > -     } else {
> > > -             /*
> > > -              * UVC doesn't specify how to inform a bulk-based device
> > > -              * when the video stream is stopped. Windows sends a
> > > -              * CLEAR_FEATURE(HALT) request to the video streaming
> > > -              * bulk endpoint, mimic the same behaviour.
> > > -              */
> > > -             unsigned int epnum = stream->header.bEndpointAddress
> > > -                                & USB_ENDPOINT_NUMBER_MASK;
> > > -             unsigned int dir = stream->header.bEndpointAddress
> > > -                              & USB_ENDPOINT_DIR_MASK;
> > > -             unsigned int pipe;
> > > -
> > > -             pipe = usb_sndbulkpipe(stream->dev->udev, epnum) | dir;
> > > -             usb_clear_halt(stream->dev->udev, pipe);
> > > +             return;
> > >       }
> > >
> > > +     /*
> > > +      * UVC doesn't specify how to inform a bulk-based device
> >
> > Then this comment doesn't look right. What about the code? This isn't
> > mentioned in the commit message either.
> >
> > > +      * when the video stream is stopped. Windows sends a
> > > +      * CLEAR_FEATURE(HALT) request to the video streaming
> > > +      * bulk endpoint, mimic the same behaviour.
> > > +      */
> > > +     epnum = stream->header.bEndpointAddress & USB_ENDPOINT_NUMBER_MASK;
> > > +     dir = stream->header.bEndpointAddress & USB_ENDPOINT_DIR_MASK;
> > > +     pipe = usb_sndbulkpipe(stream->dev->udev, epnum) | dir;
> > > +     usb_clear_halt(stream->dev->udev, pipe);
> > > +}
> > > +
> > > +void uvc_video_stop_streaming(struct uvc_streaming *stream)
> > > +{
> > > +     uvc_video_stop_transfer(stream, 1);
> > > +
> > > +     /*
> > > +      * Barrier needed to synchronize with uvc_disconnect().
> > > +      * We cannot call usb_* functions on a disconnected USB device.
> > > +      */
> > > +     if (!smp_load_acquire(&stream->dev->disconnected))
> > > +             uvc_video_halt(stream);
> > > +
> > >       uvc_video_clock_cleanup(stream);
> > >  }
> > > diff --git a/drivers/media/usb/uvc/uvcvideo.h b/drivers/media/usb/uvc/uvcvideo.h
> > > index 6fb0a78b1b00..4318ce8e31db 100644
> > > --- a/drivers/media/usb/uvc/uvcvideo.h
> > > +++ b/drivers/media/usb/uvc/uvcvideo.h
> > > @@ -559,6 +559,8 @@ struct uvc_device {
> > >       unsigned int users;
> > >       atomic_t nmappings;
> > >
> > > +     bool disconnected;
> > > +
> > >       /* Video control interface */
> > >  #ifdef CONFIG_MEDIA_CONTROLLER
> > >       struct media_device mdev;
  
Ricardo Ribalda Nov. 22, 2023, 11:35 a.m. UTC | #7
Hi

On Wed, 22 Nov 2023 at 12:04, Sakari Ailus <sakari.ailus@iki.fi> wrote:
>
> Hi Ricardo,
>
> On Wed, Nov 22, 2023 at 11:32:16AM +0100, Ricardo Ribalda wrote:
> > Hi Sakari
> >
> > On Wed, 22 Nov 2023 at 11:25, Sakari Ailus <sakari.ailus@iki.fi> wrote:
> > >
> > > Hi Ricardo,
> > >
> > > Thank you for the patch.
> > >
> > > On Tue, Nov 21, 2023 at 07:53:49PM +0000, Ricardo Ribalda wrote:
> > > > usb drivers should not call to any usb_() function after the
> > > > .disconnect() callback has been triggered.
> > > >
> > > > If the camera is streaming, the uvc driver will call usb_set_interface or
> > > > usb_clear_halt once the device is being released. Let's fix this issue.
> > > >
> > > > This is probably not the only driver affected with this kind of bug, but
> > > > until there is a better way to do it in the core this is the way to
> > > > solve this issue.
> > > >
> > > > When/if a different mechanism is implemented in the core to solve the
> > > > lifetime of devices we will adopt it in uvc.
> > > >
> > > > Trace:
> > > > [ 1065.389723] drivers/media/usb/uvc/uvc_driver.c:2248 uvc_disconnect enter
> > > > [ 1065.390160] drivers/media/usb/uvc/uvc_driver.c:2264 uvc_disconnect exit
> > > > [ 1065.433956] drivers/media/usb/uvc/uvc_v4l2.c:659 uvc_v4l2_release enter
> > > > [ 1065.433973] drivers/media/usb/uvc/uvc_video.c:2274 uvc_video_stop_streaming enter
> > > > [ 1065.434560] drivers/media/usb/uvc/uvc_video.c:2285 uvc_video_stop_streaming exit
> > > > [ 1065.435154] drivers/media/usb/uvc/uvc_v4l2.c:680 uvc_v4l2_release exit
> > > > [ 1065.435188] drivers/media/usb/uvc/uvc_driver.c:2248 uvc_disconnect enter
> > > >
> > > > Signed-off-by: Ricardo Ribalda <ribalda@chromium.org>
> > > > ---
> > > >  drivers/media/usb/uvc/uvc_driver.c |  2 ++
> > > >  drivers/media/usb/uvc/uvc_video.c  | 45 ++++++++++++++++++++++++--------------
> > > >  drivers/media/usb/uvc/uvcvideo.h   |  2 ++
> > > >  3 files changed, 32 insertions(+), 17 deletions(-)
> > > >
> > > > diff --git a/drivers/media/usb/uvc/uvc_driver.c b/drivers/media/usb/uvc/uvc_driver.c
> > > > index 08fcd2ffa727..413c32867617 100644
> > > > --- a/drivers/media/usb/uvc/uvc_driver.c
> > > > +++ b/drivers/media/usb/uvc/uvc_driver.c
> > > > @@ -2257,6 +2257,8 @@ static void uvc_disconnect(struct usb_interface *intf)
> > > >               return;
> > > >
> > > >       uvc_unregister_video(dev);
> > > > +     /* Barrier needed to synchronize with uvc_video_stop_streaming(). */
> > > > +     smp_store_release(&dev->disconnected, true);
> > > >       kref_put(&dev->ref, uvc_delete);
> > > >  }
> > > >
> > > > diff --git a/drivers/media/usb/uvc/uvc_video.c b/drivers/media/usb/uvc/uvc_video.c
> > > > index 28dde08ec6c5..032b44e45b22 100644
> > > > --- a/drivers/media/usb/uvc/uvc_video.c
> > > > +++ b/drivers/media/usb/uvc/uvc_video.c
> > > > @@ -2243,28 +2243,39 @@ int uvc_video_start_streaming(struct uvc_streaming *stream)
> > > >       return ret;
> > > >  }
> > > >
> > > > -void uvc_video_stop_streaming(struct uvc_streaming *stream)
> > > > +static void uvc_video_halt(struct uvc_streaming *stream)
> > > >  {
> > > > -     uvc_video_stop_transfer(stream, 1);
> > > > +     unsigned int epnum;
> > > > +     unsigned int pipe;
> > > > +     unsigned int dir;
> > > >
> > > >       if (stream->intf->num_altsetting > 1) {
> > >
> > > Doesn't this imply the device is using isochronous mode?
> >
> > I haven't changed the behaviour for halt, it is just that git diff is
> > being a bit too creative here:
> >
> > Basically it is doing:
> >
> > void video_halt() {
> >    if (is_isoc()) {
> >      usb_set_interface(stream->dev->udev, stream->intfnum, 0);
> >      return;
> >    }
> >    usb_clear_halt();
> > }
> >
> > instead of the old:
> >
> > void video_halt() {
> >    if (is_isoc()) {
> >      usb_set_interface(stream->dev->udev, stream->intfnum, 0);
> >    }  else {
> >       usb_clear_halt();
> >    }
> > }
> >
> > Thanks!
>
> Oops. I missed the removal of the else branch altogether while reading the
> patch.
>
> Shouldn't you also ensure the disconnect callback won't return until the
> streaming has been stopped here?

This patch is just for calls after .disconnect. It will not protect
for concurrent calls.

I have sent a v4 making this explicit. We still need:

media: uvcvideo: Lock video streams and queues while unregistering

or similar.

Thanks!

>
> > >
> > > >               usb_set_interface(stream->dev->udev, stream->intfnum, 0);
> > > > -     } else {
> > > > -             /*
> > > > -              * UVC doesn't specify how to inform a bulk-based device
> > > > -              * when the video stream is stopped. Windows sends a
> > > > -              * CLEAR_FEATURE(HALT) request to the video streaming
> > > > -              * bulk endpoint, mimic the same behaviour.
> > > > -              */
> > > > -             unsigned int epnum = stream->header.bEndpointAddress
> > > > -                                & USB_ENDPOINT_NUMBER_MASK;
> > > > -             unsigned int dir = stream->header.bEndpointAddress
> > > > -                              & USB_ENDPOINT_DIR_MASK;
> > > > -             unsigned int pipe;
> > > > -
> > > > -             pipe = usb_sndbulkpipe(stream->dev->udev, epnum) | dir;
> > > > -             usb_clear_halt(stream->dev->udev, pipe);
> > > > +             return;
> > > >       }
> > > >
> > > > +     /*
> > > > +      * UVC doesn't specify how to inform a bulk-based device
> > >
> > > Then this comment doesn't look right. What about the code? This isn't
> > > mentioned in the commit message either.
> > >
> > > > +      * when the video stream is stopped. Windows sends a
> > > > +      * CLEAR_FEATURE(HALT) request to the video streaming
> > > > +      * bulk endpoint, mimic the same behaviour.
> > > > +      */
> > > > +     epnum = stream->header.bEndpointAddress & USB_ENDPOINT_NUMBER_MASK;
> > > > +     dir = stream->header.bEndpointAddress & USB_ENDPOINT_DIR_MASK;
> > > > +     pipe = usb_sndbulkpipe(stream->dev->udev, epnum) | dir;
> > > > +     usb_clear_halt(stream->dev->udev, pipe);
> > > > +}
> > > > +
> > > > +void uvc_video_stop_streaming(struct uvc_streaming *stream)
> > > > +{
> > > > +     uvc_video_stop_transfer(stream, 1);
> > > > +
> > > > +     /*
> > > > +      * Barrier needed to synchronize with uvc_disconnect().
> > > > +      * We cannot call usb_* functions on a disconnected USB device.
> > > > +      */
> > > > +     if (!smp_load_acquire(&stream->dev->disconnected))
> > > > +             uvc_video_halt(stream);
> > > > +
> > > >       uvc_video_clock_cleanup(stream);
> > > >  }
> > > > diff --git a/drivers/media/usb/uvc/uvcvideo.h b/drivers/media/usb/uvc/uvcvideo.h
> > > > index 6fb0a78b1b00..4318ce8e31db 100644
> > > > --- a/drivers/media/usb/uvc/uvcvideo.h
> > > > +++ b/drivers/media/usb/uvc/uvcvideo.h
> > > > @@ -559,6 +559,8 @@ struct uvc_device {
> > > >       unsigned int users;
> > > >       atomic_t nmappings;
> > > >
> > > > +     bool disconnected;
> > > > +
> > > >       /* Video control interface */
> > > >  #ifdef CONFIG_MEDIA_CONTROLLER
> > > >       struct media_device mdev;
>
> --
> Kind regards,
>
> Sakari Ailus
  

Patch

diff --git a/drivers/media/usb/uvc/uvc_driver.c b/drivers/media/usb/uvc/uvc_driver.c
index 08fcd2ffa727..413c32867617 100644
--- a/drivers/media/usb/uvc/uvc_driver.c
+++ b/drivers/media/usb/uvc/uvc_driver.c
@@ -2257,6 +2257,8 @@  static void uvc_disconnect(struct usb_interface *intf)
 		return;
 
 	uvc_unregister_video(dev);
+	/* Barrier needed to synchronize with uvc_video_stop_streaming(). */
+	smp_store_release(&dev->disconnected, true);
 	kref_put(&dev->ref, uvc_delete);
 }
 
diff --git a/drivers/media/usb/uvc/uvc_video.c b/drivers/media/usb/uvc/uvc_video.c
index 28dde08ec6c5..032b44e45b22 100644
--- a/drivers/media/usb/uvc/uvc_video.c
+++ b/drivers/media/usb/uvc/uvc_video.c
@@ -2243,28 +2243,39 @@  int uvc_video_start_streaming(struct uvc_streaming *stream)
 	return ret;
 }
 
-void uvc_video_stop_streaming(struct uvc_streaming *stream)
+static void uvc_video_halt(struct uvc_streaming *stream)
 {
-	uvc_video_stop_transfer(stream, 1);
+	unsigned int epnum;
+	unsigned int pipe;
+	unsigned int dir;
 
 	if (stream->intf->num_altsetting > 1) {
 		usb_set_interface(stream->dev->udev, stream->intfnum, 0);
-	} else {
-		/*
-		 * UVC doesn't specify how to inform a bulk-based device
-		 * when the video stream is stopped. Windows sends a
-		 * CLEAR_FEATURE(HALT) request to the video streaming
-		 * bulk endpoint, mimic the same behaviour.
-		 */
-		unsigned int epnum = stream->header.bEndpointAddress
-				   & USB_ENDPOINT_NUMBER_MASK;
-		unsigned int dir = stream->header.bEndpointAddress
-				 & USB_ENDPOINT_DIR_MASK;
-		unsigned int pipe;
-
-		pipe = usb_sndbulkpipe(stream->dev->udev, epnum) | dir;
-		usb_clear_halt(stream->dev->udev, pipe);
+		return;
 	}
 
+	/*
+	 * UVC doesn't specify how to inform a bulk-based device
+	 * when the video stream is stopped. Windows sends a
+	 * CLEAR_FEATURE(HALT) request to the video streaming
+	 * bulk endpoint, mimic the same behaviour.
+	 */
+	epnum = stream->header.bEndpointAddress & USB_ENDPOINT_NUMBER_MASK;
+	dir = stream->header.bEndpointAddress & USB_ENDPOINT_DIR_MASK;
+	pipe = usb_sndbulkpipe(stream->dev->udev, epnum) | dir;
+	usb_clear_halt(stream->dev->udev, pipe);
+}
+
+void uvc_video_stop_streaming(struct uvc_streaming *stream)
+{
+	uvc_video_stop_transfer(stream, 1);
+
+	/*
+	 * Barrier needed to synchronize with uvc_disconnect().
+	 * We cannot call usb_* functions on a disconnected USB device.
+	 */
+	if (!smp_load_acquire(&stream->dev->disconnected))
+		uvc_video_halt(stream);
+
 	uvc_video_clock_cleanup(stream);
 }
diff --git a/drivers/media/usb/uvc/uvcvideo.h b/drivers/media/usb/uvc/uvcvideo.h
index 6fb0a78b1b00..4318ce8e31db 100644
--- a/drivers/media/usb/uvc/uvcvideo.h
+++ b/drivers/media/usb/uvc/uvcvideo.h
@@ -559,6 +559,8 @@  struct uvc_device {
 	unsigned int users;
 	atomic_t nmappings;
 
+	bool disconnected;
+
 	/* Video control interface */
 #ifdef CONFIG_MEDIA_CONTROLLER
 	struct media_device mdev;