On Fri, Aug 17, 2012 at 03:05:15PM +0200, Oliver Neukum wrote: > On Friday 17 August 2012 15:01:16 Sebastian Andrzej Siewior wrote: > > > It is possible that you get -ENOMEM while a command is already being > > > executed. The worst case is probably that you got a command to the device > > > and you need to do autosense, but you get -ENOMEM. > > > > Looking at the code, I see that we have have anchors so were are good. Except > > for the cmd_urb. > > That means the endless re-try loops stops once you plug the cable. > > I am afraid this is not entirely true. Somebody may do a soft disconnect (sysfs) > between scheduling the work and executing it. In that case the work function > will happily submit URBs to an interface it is no longer bound to. And as the work > is shared we cannot simply shut it down in disconnect() I see. So the problem is that we have it in the uas_work_list and uas_disconnect() frees that URB which we re-submit. So the obvious thing would be something per-interface but do need this re-submit at all? > > Regards > Oliver > Sebastian -- 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
