On Fri, 11 Dec 2009, Ondrej Zary wrote: > On Thursday 10 December 2009, Alan Stern wrote: > > On Thu, 10 Dec 2009, Ondrej Zary wrote: > > > Narrowed down the code to this 170-line module which reproduces my > > > "rmmod usbtouchscreen" problem. Loading this module causes EHCI to fail. > > > Looks like it fails after calling usb_kill_urb(). Can a buggy device > > > cause this? > > > > I don't think so. But I am getting an idea of what the problem might > > be. Try applying the patch below and see if it makes any difference > > (make sure that CONFIG_USB_DEBUG is enabled). > > Thank you very much for the patch. It seems to finally solve this long > thread! Great! > I'm running 2.6.32 now (with usbtouchscreen copied from -next) so one hunk > had to be applied manually. With patch applied and CONFIG_USB_DEBUG enabled, > usbtouchscreen stopped working - the urb failed to submit. Return value of > usb_submit_urb() was not checked (and it wasn't my code this time :) Adding > a check revealed that the urb fails to submit with -EPIPE because of > endpoint vs. pipe mismatch. > > Fixing that (see the patch below) solved the rmmod problem! > --- linux-2.6.32/drivers/input/touchscreen/usbtouchscreen.5 2009-12-11 12:50:35.000000000 +0100 > +++ linux-2.6.32/drivers/input/touchscreen/usbtouchscreen.c 2009-12-11 20:15:51.000000000 +0100 > @@ -1375,10 +1375,16 @@ > input_set_abs_params(input_dev, ABS_PRESSURE, type->min_press, > type->max_press, 0, 0); > > - usb_fill_int_urb(usbtouch->irq, udev, > + if (usb_endpoint_type(endpoint) == USB_ENDPOINT_XFER_INT) > + usb_fill_int_urb(usbtouch->irq, udev, > usb_rcvintpipe(udev, endpoint->bEndpointAddress), > usbtouch->data, type->rept_size, > usbtouch_irq, usbtouch, endpoint->bInterval); > + else > + usb_fill_bulk_urb(usbtouch->irq, udev, > + usb_rcvbulkpipe(udev, endpoint->bEndpointAddress), > + usbtouch->data, type->rept_size, > + usbtouch_irq, usbtouch); > > usbtouch->irq->dev = udev; > usbtouch->irq->transfer_dma = usbtouch->data_dma; Okay. I'll submit my patch. Since it's not a bug fix, it won't go into the current release or the stable trees, but it should appear in 2.6.33. You should submit your fix to the appropriate maintainer. This was a tricky problem; I'm glad we finally found the underlying cause. Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
