[ofw] [PATCH] complib/user: fix timer race conditions

[Fab Tillier]

Hefty, Sean wrote on Fri, 25 Jun 2010 at 08:12:27

>> Why the per-timer timer queue?  The only time you take advantage of it
>> is in cl_timer_destroy, but at that point the timer should no longer be
>> in the process of being queued.  You could just as easily call
>> __clear_timer_handle, and then do a blocking cancel of the timer.
> 
> DeleteTimerQueueEx cancels all outstanding timers and waits for them to
> complete.  This allows us to call DeleteTimerQueueTimer in a non-
> blocking fashion from start/trim and not worry that we stop tracking
> the timer.  Cleanup happens appropriately.
> 
> Calling DeleteTimerQueueTimer in a blocking fashion from destroy only
> guarantees that the last timer we created has been canceled.  However,
> previous timers may still be lingering.

Ok, that makes sense, looks good to me then. Does it work?

-Fab