On Tue, 16 Oct 2018, Oliver Neukum wrote: > On Mo, 2018-10-15 at 09:50 -0400, Alan Stern wrote: > > > > It seems that a better approach would be to have an ioctl which would: > > > > fail if there are any active user URBs; > > > > drop the usbfs PM reference so the device can suspend; > > > > block interruptibly until the device resumes; > > Thus we would require user space to have a thread for each device > it wants to allow suspend for. True. I suppose instead of waiting on the usbfs device file, we could use select() on some sysfs file (something like power/runtime_status). Unlike the usbfs device file, keeping a sysfs file open won't interfere with runtime PM. And it might be useful for other user programs to have a way of being notified when a device (not necessarily USB!) goes into or out of suspend. > [..] > > What difference does it make if the URBs are killed by the user instead > > of the kernel? > > We stay within the limits of the timing (as well as we can) and the > case of a failed suspend is much easier to handle. > Else we have an unlimited time between cessation of IO and going > into suspend. The time could be limited be a timeout. In any case, we must recognize that userspace device drivers will never be as powerful as kernel drivers. The difficulty of handling suspend and resume is one example of this. Alan Stern
