On Wednesday 12 November 2003 17:47, Alan Stern wrote: > On Wed, 12 Nov 2003, Duncan Sands wrote: > > I'm getting various oopses in 2.6 due to actconfig being NULL > > unexpectedly in devio.c, namely in these releaseintf and findintif: > > > > The first oops occurs always at system shutdown. The second oops occurs > > sometimes at system startup. I can "fix" the first one by replacing > > down(&dev->serialize); > > with > > lock_kernel(); > > and > > up(&dev->serialize); > > with > > unlock_kernel(); > > in releasintf. The logic here is that although usbdev_release takes the > > BKL, it can lose it while sleeping in the down(&dev->serialize) in > > releaseintf, which is not what is intended I guess (and since it > > eliminates the first oops, there must be some truth to this theory). > > However things are more complicated because > > usb_driver_release_interface(&usbdevfs_driver, iface) can also sleep, > > which might also cause problems. I haven't analysed this yet. Maybe > > this change also fixed the second oops, but since that one only occurs > > about one time in four, it is hard to be sure. > > > > I will of course look into this further, but I thought I would mention it > > now in case anyone has any ideas. > > Usbfs is full of locking errors. In particular, there's no excuse for it > holding the BKL. > > In this case, the simplest solution is simply to insert a test for > actconfig != NULL before it gets dereferenced. To make certain that > actconfig won't change out from under you, be sure to hold dev->serialize > during that test. > > The real problem is that usbfs is trying to unbind a driver after usbcore > has already done so on its behalf. A better fix would be to repair > driver_disconnect() and the USBDEVFS_DISCONNECT case in proc_ioctl() > to make them use the > > test_and_clear_bit(intf, &ps->ifclaimed) > > test as they should.
Hi Alan, looking through devio.c I see many problems. A basic problem is that disconnect methods are now per-interface, but usbfs is still working on the per-device model. For example, in driver_disconnect the whole of ps->ifclaimed is reset to zero, not just the bit for the interface in question. The same goes for the use of ps->dev = NULL to indicate disconnection. That just doesn't mean anything any more - the device hasn't been disconnected, just one interface. It seems to me that this is all easy to fix: (1) take a reference to the usb device in usbdev_open (drop it in usbdev_release). (2) don't set ps->dev to NULL in driver_disconnect (3) just clear the right bit, and use destroy_async_on_interface in driver_disconnect Et voila as they say around here. Duncan. PS: I don't see what this has to do with my oopsen though :) ------------------------------------------------------- This SF.Net email sponsored by: ApacheCon 2003, 16-19 November in Las Vegas. Learn firsthand the latest developments in Apache, PHP, Perl, XML, Java, MySQL, WebDAV, and more! http://www.apachecon.com/ _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
