* Alan Stern | 2014-12-05 16:21:02 [-0500]: >On Fri, 5 Dec 2014, Sebastian Andrzej Siewior wrote: >> So instead, I hold the device struct in the HCD and the HCD struct on >> every USB-buf-alloc. That means after a disconnect we still have a >> refcount on usb_hcd and device and it will be cleaned "later" once the >> last USB-buffer is released. > >This is not a valid solution. Notice that your _hcd_buffer_free still >dereferences hcd->driver; that will not point to anything useful if you >rmmod the HCD. Hmm. You're right, that one is gone. >Also, you neglected to move the calls to hcd_buffer_destroy from >usb_remove_hcd to hcd_release. I add them, I didn't move them. >On the whole, it would be easier if the UVC driver could release its >coherent DMA buffers during the disconnect callback. If that's not >feasible we'll have to find some other solution. I had one patch doing that. Let me grab it out on Monday. >Alan Stern > Sebastian -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
