On Sat, Apr 28, 2018 at 10:33:55AM +0200, Martin Pieuchot wrote:
> If USB transfers aren't completed in order, the stack cannot recover.
>
> So don't try to continue when this happen and make it clear what the
> problem is. Currently the kernel panic in many different ways or
> worse enters an infinite loop sometimes after that happened.
>
> ok?
OK
>
> Index: usbdi.c
> ===================================================================
> RCS file: /cvs/src/sys/dev/usb/usbdi.c,v
> retrieving revision 1.97
> diff -u -p -r1.97 usbdi.c
> --- usbdi.c 3 Feb 2018 13:37:37 -0000 1.97
> +++ usbdi.c 28 Apr 2018 08:18:18 -0000
> @@ -769,13 +769,11 @@ usb_transfer_complete(struct usbd_xfer *
>
> if (!pipe->repeat) {
> /* Remove request from queue. */
> + KASSERT(xfer == SIMPLEQ_FIRST(&pipe->queue));
> + SIMPLEQ_REMOVE_HEAD(&pipe->queue, next);
> #ifdef DIAGNOSTIC
> - if (xfer != SIMPLEQ_FIRST(&pipe->queue))
> - printf("usb_transfer_complete: bad dequeue %p != %p\n",
> - xfer, SIMPLEQ_FIRST(&pipe->queue));
> xfer->busy_free = XFER_FREE;
> #endif
> - SIMPLEQ_REMOVE_HEAD(&pipe->queue, next);
> }
> DPRINTFN(5,("usb_transfer_complete: repeat=%d new head=%p\n",
> pipe->repeat, SIMPLEQ_FIRST(&pipe->queue)));
