On Wed, 27 Jan 2010, Oliver Neukum wrote: > Am Mittwoch, 27. Januar 2010 19:16:07 schrieben Sie: > > > > That's basically right. A driver doesn't have to kill its own URBs, > > > > but it is considered good form to do so. > > > > > > Ehm, there's the small matter of disconnect by sysfs or usbfs. > > > Or module unload. > > > > The reason for calling the disconnect routine doesn't matter; the core > > first disables the interface's endpoints, meaning that pending URBs > > will fail with -ESHUTDOWN and new submissions will fail with -ENOENT. > > However URBs for endpoint 0 are unaffected -- the driver has to handle > > them on its own. > > And there's the point of drivers using usb_claim_interface(). > Basically you should just kill them all. If a driver uses usb_claim_interface() then it should also use usb_release_interface(), which will disable endpoints on the interface being released. However the endpoints will be enabled again before usb_release_interface() returns, so there is a possibility that more URBs could be submitted before the disconnect is finished. I agree, it's overall cleaner for a driver to make sure it leaves no URBs active when it is disconnected. 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
