Re: [PATCH RFC v2] usb: gadget: uvc: check endpoint enabled state in uvcg_video_enable()

On Wed, Aug 12, 2026 at 3:28 PM syzbot <[email protected]> wrote:
>
> A NULL pointer dereference can occur in uvc_video_prep_requests() if
> userspace issues a VIDIOC_STREAMON ioctl before the USB host has fully
> configured the streaming endpoint.
>
> When userspace calls VIDIOC_STREAMON, the driver handles it in
> uvc_v4l2_streamon(), which in turn calls uvcg_video_enable(). Currently,
> uvcg_video_enable() only checks if the endpoint pointer is allocated
> (video->ep == NULL), but it fails to check if the endpoint is actually
> enabled (video->ep->enabled).
>
> The endpoint descriptor (video->ep->desc) is only assigned when the USB
> host explicitly selects the streaming alternate setting via a SET_INTERFACE
> control request. If userspace prematurely calls VIDIOC_STREAMON before the
> host has selected the streaming alternate setting, video->ep->desc will
> still be NULL.
>
> Because uvcg_video_enable() does not verify the endpoint's enabled state,
> it proceeds to call uvc_video_alloc_requests() and then
> uvc_video_prep_requests(). Inside uvc_video_prep_requests(), the code
> unconditionally dereferences video->ep->desc, resulting in a KASAN
> null-ptr-deref / general protection fault:
>
> Oops: general protection fault, probably for non-canonical address
> 0xdffffc0000000000: 0000 [#1] SMP KASAN NOPTI
> KASAN: null-ptr-deref in range [0x0000000000000000-0x0000000000000007]
> RIP: 0010:usb_endpoint_xfer_isoc include/uapi/linux/usb/ch9.h:571 [inline]
> RIP: 0010:uvc_video_prep_requests
> drivers/usb/gadget/function/uvc_video.c:508 [inline]
> RIP: 0010:uvc_video_alloc_requests
> drivers/usb/gadget/function/uvc_video.c:559 [inline]
> RIP: 0010:uvcg_video_enable+0x142/0xe70
> drivers/usb/gadget/function/uvc_video.c:784
> Call Trace:
>  <TASK>
>  uvc_v4l2_streamon+0x7e/0x110 drivers/usb/gadget/function/uvc_v4l2.c:531
>  __video_do_ioctl+0x8af/0xc70 drivers/media/v4l2-core/v4l2-ioctl.c:3133
>  video_usercopy+0x860/0x1430 drivers/media/v4l2-core/v4l2-ioctl.c:3475
>  v4l2_ioctl+0x18d/0x1e0 drivers/media/v4l2-core/v4l2-dev.c:366
>  vfs_ioctl fs/ioctl.c:51 [inline]
>  __do_sys_ioctl fs/ioctl.c:597 [inline]
>  __se_sys_ioctl+0xfc/0x170 fs/ioctl.c:583
>
> To fix this, update uvcg_video_enable() to also verify that the endpoint is
> enabled (!video->ep->enabled). This ensures that if userspace prematurely
> calls VIDIOC_STREAMON, the driver will safely reject the request with
> -ENODEV instead of crashing the kernel.
>
> Fixes: 48dbe731171e ("usb: gadget: uvc: set req_size and n_requests based on the frame interval")
> Assisted-by: Gemini:gemini-3.5-flash Gemini:gemini-3.1-pro-preview syzbot
> Reported-by: [email protected]
> Closes: https://syzkaller.appspot.com/bug?extid=44835f0858f11e3e923c
> Link: https://syzkaller.appspot.com/ai_job?id=e3902952-fe28-4ca4-a910-c485a9927925
> To: "Greg Kroah-Hartman" <[email protected]>
> To: <[email protected]>
> To: "Michael Grzeschik" <[email protected]>
> Cc: "Frank Li" <[email protected]>
> Cc: "Jimmy Hu" <[email protected]>
> Cc: "Hans Verkuil" <[email protected]>
> Cc: "Kees Cook" <[email protected]>
> Cc: "Laurent Pinchart" <[email protected]>
> Cc: <[email protected]>
> Cc: "Muhammad Bilal" <[email protected]>
> Cc: "Junzhong Pan" <[email protected]>
> Cc: "Xu Yang" <[email protected]>
>
> ---
> v2:
> - Check that uvc->state is UVC_STATE_CONNECTED in uvc_v4l2_streamon() before allowing streaming.
> - Update uvc->state to UVC_STATE_STREAMING before calling uvcg_video_enable(), and revert it on failure.
> - Check video->ep->desc in addition to video->ep->enabled in uvcg_video_enable().
> - Add safety checks for video->ep, video->ep->enabled, and video->ep->desc in uvc_video_alloc_requests().
>
> v1:
> https://lore.kernel.org/all/[email protected]/T/
> ---
> diff --git a/drivers/usb/gadget/function/uvc_v4l2.c b/drivers/usb/gadget/function/uvc_v4l2.c
> index dfa0521a2..a80663c9a 100644
> --- a/drivers/usb/gadget/function/uvc_v4l2.c
> +++ b/drivers/usb/gadget/function/uvc_v4l2.c
> @@ -527,17 +527,24 @@ uvc_v4l2_streamon(struct file *file, void *fh, enum v4l2_buf_type type)
>         if (type != video->queue.queue.type)
>                 return -EINVAL;
>
> +       if (uvc->state != UVC_STATE_CONNECTED)
> +               return -ENODEV;
> +
> +       uvc->state = UVC_STATE_STREAMING;
> +
>         /* Enable UVC video. */
>         ret = uvcg_video_enable(video);
> -       if (ret < 0)
> +       if (ret < 0) {
> +               if (uvc->state == UVC_STATE_STREAMING)
> +                       uvc->state = UVC_STATE_CONNECTED;
>                 return ret;
> +       }
>
>         /*
>          * Complete the alternate setting selection setup phase now that
>          * userspace is ready to provide video frames.
>          */
>         uvc_function_setup_continue(uvc, 0);
> -       uvc->state = UVC_STATE_STREAMING;

Instead of all that changes, would just this suffice (at the right place)?

        if (uvc->state != UVC_STATE_CONNECTED)
            return -ENODEV;

>
>         return 0;
>  }
> diff --git a/drivers/usb/gadget/function/uvc_video.c b/drivers/usb/gadget/function/uvc_video.c
> index 2f9700b3f..bd67a9d80 100644
> --- a/drivers/usb/gadget/function/uvc_video.c
> +++ b/drivers/usb/gadget/function/uvc_video.c
> @@ -551,6 +551,9 @@ uvc_video_alloc_requests(struct uvc_video *video)
>         unsigned int i;
>         int ret = -ENOMEM;
>
> +       if (video->ep == NULL || !video->ep->enabled || !video->ep->desc)
> +               return -ENODEV;
> +
>         /*
>          * calculate in uvc_video_prep_requests
>          * - video->uvc_num_requests
> @@ -764,7 +767,7 @@ int uvcg_video_enable(struct uvc_video *video)
>  {
>         int ret;
>
> -       if (video->ep == NULL) {
> +       if (video->ep == NULL || !video->ep->enabled || !video->ep->desc) {
>                 uvcg_info(&video->uvc->func,
>                           "Video enable failed, device is uninitialized.\n");
>                 return -ENODEV;
>
>
> base-commit: db2ddb87143519e20a95aa36c60b36107b736a58
> --
> This is an AI-generated patch subject to moderation.
> Reply with '#syz upstream' to Sign-off the patch as a human author
> and send it to the upstream kernel mailing lists.
> Reply with '#syz reject' to reject it ('#syz unreject' to undo).
>
> See https://goo.gle/syzbot-ai-patches for information about AI-generated patches.
> You can comment on the patch as usual, syzbot will try to address
> the comments and send a new version of the patch if necessary.
> syzbot engineers can be reached at [email protected].