Re: [patch 2/2] sched: charge unaccounted run-time on entity re-weight

[Paul Turner <pjt@google.com>]

On 12/15/10 19:35, Paul Turner wrote:
> Hum -- forgot to refresh mbx file, slightly cleaner version (we can
> still charge unaccounted time against our queuing cfs_rq).
>

... Obviously this is supposed to be in lieu of the first patch.

I'm tired  :(

> - Paul
>
> -----
>
> sched: move periodic share updates to entity_tick()
>
> Long running entities that do not block (dequeue) require periodic
> updates to
> maintain accurate share values. (Note: group entities with several
> threads are
> quite likely to be non-blocking in many circumstances).
>
> By virtue of being long-running however, we will see entity ticks
> (otherwise
> the required update occurs in dequeue/put and we are done). Thus we can
> move
> the detection (and associated work) for these updates into the periodic
> path.
>
> This restores the 'atomicity' of update_curr() with respect to accounting.
>
> Signed-off-by: Paul Turner <pjt@google.com>
>
> ---
> kernel/sched_fair.c | 21 +++++++++++++++++----
> 1 file changed, 17 insertions(+), 4 deletions(-)
>
> Index: tip3/kernel/sched_fair.c
> ===================================================================
> --- tip3.orig/kernel/sched_fair.c
> +++ tip3/kernel/sched_fair.c
> @@ -564,11 +564,8 @@ __update_curr(struct cfs_rq *cfs_rq, str
>
> #if defined CONFIG_SMP && defined CONFIG_FAIR_GROUP_SCHED
> cfs_rq->load_unacc_exec_time += delta_exec;
> - if (cfs_rq->load_unacc_exec_time > sysctl_sched_shares_window) {
> - update_cfs_load(cfs_rq, 0);
> - update_cfs_shares(cfs_rq, 0);
> - }
> #endif
> }
>
> static void update_curr(struct cfs_rq *cfs_rq)
> @@ -809,6 +806,15 @@ static void update_cfs_shares(struct cfs
>
> reweight_entity(cfs_rq_of(se), se, shares);
> }
> +
> +static void update_cfs_rq_shares_tick(struct cfs_rq *cfs_rq)
> +{
> + /* rate limit updates by the averaging window */
> + if (cfs_rq->load_unacc_exec_time > sysctl_sched_shares_window) {
> + update_cfs_load(cfs_rq, 0);
> + update_cfs_shares(cfs_rq, 0);
> + }
> +}
> #else /* CONFIG_FAIR_GROUP_SCHED */
> static void update_cfs_load(struct cfs_rq *cfs_rq, int global_update)
> {
> @@ -1133,6 +1139,13 @@ entity_tick(struct cfs_rq *cfs_rq, struc
> */
> update_curr(cfs_rq);
>
> +#if defined CONFIG_SMP && defined CONFIG_FAIR_GROUP_SCHED
> + /*
> + * Update share accounting for long-running entities.
> + */
> + update_cfs_rq_shares_tick(cfs_rq);
> +#endif
> +
> #ifdef CONFIG_SCHED_HRTICK
> /*
> * queued ticks are scheduled to match the slice, so don't bother