Viresh,
On Wed, 20 Nov 2013, Viresh Kumar wrote:
Migration of timers from idle cores to non-idle ones for power saving is very well working and really saves a lot of power for us. What's currently not working is the migration of running timers Or timers which re-arms themselves.
There are complications with migrating timers which schedules themselves again from their handler. del_timer_sync() can't detect that the timer's handler yet has not finished.
Because you try to violate the semantics of the existing code. There is a reason why we enforce that running timers must be requeued on the base they are running on.
Now you try to duct tape it into submission. That's not going to fly.
The timer wheel code is not designed to allow that and it has lots of other short comings and historic burdens. I'm not going to accept more duct tape and fragile hackery into that code.
I'm already working on a complete replacement infrastructure, which avoids all the short comings of the current timer implementation including this one.
It's going to be a massive effort to convert everything over to the new infrastructure so we can kill the timer wheel, but that's less scary and less risky than trying to add more fragility to the existing code.
Thanks,
tglx