Messages in this thread | | | Date | Thu, 25 Aug 2011 00:40:12 +0200 | From | "Jan H. Schönherr" <> | Subject | Re: [PATCH v3 2/2] sched: Handle on_list ancestor in list_add_leaf_cfs_rq() |
| |
Am 24.08.2011 23:32, schrieb Paul Turner: >>> Now I don't really like the above because its hard to make the code go >>> away in the !FAIR_GROUP case, but maybe we can come up with something >>> for that. >> >> Hmmm... you might want to reconsider my original approach to solve this: >> http://lkml.org/lkml/2011/7/18/86 >> >> That might have been the cleanest one in this respect. >> >> Paul Turner did not like the introduced in-order removal, but the >> out-of-order removal is causing most problems. >> > > Sorry for the delayed reply -- I owe you some feedback on the updated > versions but have been buried with other work.
No problem.
> What I didn't like about the original approach was specifically the > positional dependence on enqueue/dequeue.
Maybe I misunderstood you, then.
If we can guarantee in-order removal of leaf_cfs_rqs, then there is no positional dependency. Any SE can be enqueued and dequeued anytime.
OTOH, the RCU splice variant has a positional dependence: calling enqueue_entity() outside of enqueue_task_fair() can go wrong easily as it depends on being called bottom-up and requires its caller to maintain state.
This is also partly true for the leaf_insertion_point variant: if a caller maintains state, then the pair enqueue_entity/enqueue_leaf_cfs_rq() also depends on being called bottom up.
> If we can't do the splicing > properly then I think we want something like: > https://lkml.org/lkml/2011/7/18/348 to avoid shooting ourselves in the > future later. > > See: https://lkml.org/lkml/2011/7/19/178 for why this should be cheap.
As far as I can tell, all three variants proposed so far work.
It is probably a matter of taste in the end. I'll happily help with whatever version tastes best. :)
Regards Jan
| |