On Sat, 21 Apr 2012, Oliver Neukum wrote:

quoted

On the other hand, usbnet could call usb_unlink_urb from within a 
tasklet.

No. That would introduce a race.
Basically whenever a completion handler resubmits itself, any timeout
has to avoid a race. Checking for a timeout and unlinking must be atomic,
just as the completion handler must make the resubmission and noting
the time atomic.
If I cannot do this with a lock, then a much more complex pattern is necessary.

I see.  I never really understood the problem before.

Although the kerneldoc doesn't actually say so, it should be safe to
assume that usb_unlink_urb calls the completion routine directly _only_
in cases where the unlink succeeded.  (We could add this to the 
kerneldoc.)

Therefore: If the URB completes with status other than -ECONNRESET then 
you can safely take the lock for resubmission.  If the URB completes 
with status == -ECONNRESET then you know it was unlinked, so you don't 
need to take the lock -- the race has already been lost.

Does that solve your problem?

Alan Stern