On Tue, 18 May 2021 at 14:54, Odin Ugedal <odin@xxxxxxx> wrote:
>
> This fixes an issue where fairness is decreased since cfs_rq's can
> end up not being decayed properly. For two sibling control groups with
> the same priority, this can often lead to a load ratio of 99/1 (!!).
>
> This happen because when a cfs_rq is throttled, all the descendant cfs_rq's
> will be removed from the leaf list. When they initial cfs_rq is
> unthrottled, it will currently only re add descendant cfs_rq's if they
> have one or more entities enqueued. This is not a perfect heuristic.
What would be the other condition in addition to the current one
:cfs_rq->nr_running >= 1 ?
We need to add a cfs_rq in the list if it still contributes to the
tg->load_avg and the split of the share. Can't we add a condition for
this instead of adding a new field ?
>
> This fix change this behavior to save what cfs_rq's was removed from the
> list, and readds them properly on unthrottle.
>
> Can often lead to sutiations like this for equally weighted control
> groups:
>
> $ ps u -C stress
> USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND
> root 10009 88.8 0.0 3676 100 pts/1 R+ 11:04 0:13 stress --cpu 1
> root 10023 3.0 0.0 3676 104 pts/1 R+ 11:04 0:00 stress --cpu 1
>
> Fixes: 31bc6aeaab1d ("sched/fair: Optimize update_blocked_averages()")
> Signed-off-by: Odin Ugedal <odin@xxxxxxx>
> ---
> kernel/sched/fair.c | 11 ++++++-----
> kernel/sched/sched.h | 1 +
> 2 files changed, 7 insertions(+), 5 deletions(-)
>
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index ceda53c2a87a..e7423d658389 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -376,7 +376,8 @@ static inline bool list_add_leaf_cfs_rq(struct cfs_rq *cfs_rq)
> return false;
> }
>
> -static inline void list_del_leaf_cfs_rq(struct cfs_rq *cfs_rq)
> +/* Returns 1 if cfs_rq was present in the list and removed */
> +static inline bool list_del_leaf_cfs_rq(struct cfs_rq *cfs_rq)
> {
> if (cfs_rq->on_list) {
> struct rq *rq = rq_of(cfs_rq);
> @@ -393,7 +394,9 @@ static inline void list_del_leaf_cfs_rq(struct cfs_rq *cfs_rq)
>
> list_del_rcu(&cfs_rq->leaf_cfs_rq_list);
> cfs_rq->on_list = 0;
> + return 1;
> }
> + return 0;
> }
>
> static inline void assert_list_leaf_cfs_rq(struct rq *rq)
> @@ -4742,9 +4745,7 @@ static int tg_unthrottle_up(struct task_group *tg, void *data)
> if (!cfs_rq->throttle_count) {
> cfs_rq->throttled_clock_task_time += rq_clock_task(rq) -
> cfs_rq->throttled_clock_task;
> -
> - /* Add cfs_rq with already running entity in the list */
> - if (cfs_rq->nr_running >= 1)
> + if (cfs_rq->insert_on_unthrottle)
> list_add_leaf_cfs_rq(cfs_rq);
> }
>
> @@ -4759,7 +4760,7 @@ static int tg_throttle_down(struct task_group *tg, void *data)
> /* group is entering throttled state, stop time */
> if (!cfs_rq->throttle_count) {
> cfs_rq->throttled_clock_task = rq_clock_task(rq);
> - list_del_leaf_cfs_rq(cfs_rq);
> + cfs_rq->insert_on_unthrottle = list_del_leaf_cfs_rq(cfs_rq);
> }
> cfs_rq->throttle_count++;
>
> diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
> index a189bec13729..12a707d99ee6 100644
> --- a/kernel/sched/sched.h
> +++ b/kernel/sched/sched.h
> @@ -602,6 +602,7 @@ struct cfs_rq {
> u64 throttled_clock_task_time;
> int throttled;
> int throttle_count;
> + int insert_on_unthrottle;
> struct list_head throttled_list;
> #endif /* CONFIG_CFS_BANDWIDTH */
> #endif /* CONFIG_FAIR_GROUP_SCHED */
> --
> 2.31.1
>
