Re: [PATCH 2/2] sched: update runqueue clock before migrations away

From: Chris Redpath
Date: Wed Dec 18 2013 - 05:13:26 EST


On 17/12/13 18:03, bsegall@xxxxxxxxxx wrote:
__synchronize_entity_decay will decay load_avg_contrib in order to
figure out how much to remove from old_cfs_rq->blocked_load.
update_entity_load_avg will update the underlying runnable_avg_sum/period that
is used to update load_avg_contrib.

(Normally we update runnable_avg_sum, which updates load_avg_contrib via
__update_entity_load_avg_contrib. Here we go in the reverse direction
because we don't hold the right rq locks at the right times.)


Thanks Ben, got it now. The only question remaining for me to figure out is if I need to include the missed tick time in the contrib decay or not - I definitely need to include it in the negative decay count we send through a migration. I'll go and check the places we use the removed load and update blocked load again.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/