On 31.08.21 11:13, Johan Hovold wrote: > On Mon, Aug 30, 2021 at 10:46:13AM -0400, Alan Stern wrote: >> Perhaps the stable backport could be delayed for a while (say, one >> release cycle). > That might work. > >> Do you have alternative suggestions? I don't think we want special >> interruptible versions of usb_control_msg() and usb_bulk_msg() just for >> use by usbfs. > usbfs could carry a temporary local implementation as the documentation > for usb_start_wait_urb() currently suggests. I assume we can't limit the > usbfs timeouts. Upon further considerations forcing user space to handle signals also breaks the API, albeit in a more subtle manner. I'd suggest that we use wait_event_killable_timeout(). And do it the way Alan initially disliked, that is code a version for use by usbfs. Thus we'd avoid the issue of having an unkillable process, but we do not impose a need to handle signals. Regards Oliver
