On Tue, 4 Aug 2009, Oliver Neukum wrote: > Am Dienstag, 4. August 2009 00:04:57 schrieb Alan Stern: > > On Mon, 3 Aug 2009, Oliver Neukum wrote: > > > Am Montag, 3. August 2009 23:11:07 schrieb Alan Stern: > > > > The tasklet races with disconnection anyway. Why add a bunch of > > > > special code in this one spot? > > > > > > The alternative is to cache the pointer. I don't like making unnecessary > > > copies. It is safe with respect to disconnect as disconnect sets a flag. > > > Which solution would you suggest? > > > > Caching the pointer would make the routine behave just like it used to > > before the possibility of using an interrupt endpoint was added. That > > seems safest. > > But it would break the devices for which that possibility was added. I wasn't clear. Add a new field, acm->rx_endpoint_desc of type "pointer to struct usb_endpoint_descriptor", and during initialization assign acm->rx_endpoint_desc = epread; Then in acm_rx_tasklet replace each occurrence of &ep->desc with acm->rx_endpoint_desc and remove the assignment to ep. The ability to use either bulk or interrupt endpoints is thus retained. Not only will this fix the problem, it also gives a net shrinkage of code (a complicated assignment is replaced by a simple assignment). 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
