Re: [PATCH] sched: skip autogroup when looking for all rt sched groups

From: Cheng Xu
Date: Thu Jun 23 2011 - 07:21:38 EST


On 06/22/2011 01:22 PM, Yong Zhang wrote:
> From: Yong Zhang <yong.zhang0@xxxxxxxxx>
> Subject: [PATCH] sched: skip autogroup when looking for all rt sched groups
>
> Since [commit ec514c48: sched: Fix rt_rq runtime leakage bug],
> 'cat /proc/sched_debug' will print data of root_task_group.rt_rq
> multi times, this is due to autogroup has no its own rt group,
> instead rt group of autogroup is linked to root_task_group.
>
> So skip it when we are looking for all rt sched groups, and it
> will also save some noop operation against root_task_group when
> __disable_runtime()/__enable_runtime().
>
> Signed-off-by: Yong Zhang <yong.zhang0@xxxxxxxxx>
> Cc: Ingo Molnar <mingo@xxxxxxx>
> Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> Cc: Mike Galbraith <efault@xxxxxx>
> ---
>
> Since webmail may mangle the patch, attach it too.
>
> kernel/sched_autogroup.h | 1 +
> kernel/sched_rt.c | 29 ++++++++++++++++++++++++-----
> 2 files changed, 25 insertions(+), 5 deletions(-)
>
> diff --git a/kernel/sched_autogroup.h b/kernel/sched_autogroup.h
> index 0557705..c2f0e72 100644
> --- a/kernel/sched_autogroup.h
> +++ b/kernel/sched_autogroup.h
> @@ -13,6 +13,7 @@ struct autogroup {
> int nice;
> };
>
> +static inline bool task_group_is_autogroup(struct task_group *tg);
> static inline struct task_group *
> autogroup_task_group(struct task_struct *p, struct task_group *tg);
>
> diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c
> index 10d0182..8edf487 100644
> --- a/kernel/sched_rt.c
> +++ b/kernel/sched_rt.c
> @@ -185,11 +185,30 @@ static inline u64 sched_rt_period(struct rt_rq *rt_rq)
>
> typedef struct task_group *rt_rq_iter_t;
>
> -#define for_each_rt_rq(rt_rq, iter, rq) \
> - for (iter = list_entry_rcu(task_groups.next, typeof(*iter), list); \
> - (&iter->list != &task_groups) && \
> - (rt_rq = iter->rt_rq[cpu_of(rq)]); \
> - iter = list_entry_rcu(iter->list.next, typeof(*iter), list))
> +/* autogroup is not related to rt group, skip it */
> +#define find_first_valid_task_group(iter) \
> + ({ \
> + list_for_each_entry_rcu(iter, &task_groups, list) \
> + if (!task_group_is_autogroup(iter)) \
> + break; \
> + iter; \
> + })
> +
> +#define find_next_valid_task_group(iter) \
> + ({ \
> + do { \
> + iter = list_entry_rcu(iter->list.next, \
> + typeof(*iter), list); \
> + } while (&iter->list != &task_groups && \
> + task_group_is_autogroup(iter)); \
> + iter; \
> + })
> +
> +#define for_each_rt_rq(rt_rq, iter, rq) \
> + for (find_first_valid_task_group(iter); \
> + (&iter->list != &task_groups) && \
> + (rt_rq = iter->rt_rq[cpu_of(rq)]); \
> + find_next_valid_task_group(iter))
>
> static inline void list_add_leaf_rt_rq(struct rt_rq *rt_rq)
> {

This looks like it works, but seems we might be able to do that with less code...

static inline struct task_group *next_task_group(struct task_group *tg)
{
do {
tg = list_entry_rcu(tg->list.next,
typeof(struct task_group), list);
} while ((&tg->list != &task_groups) && (task_group_is_autogroup(tg)));
if (&tg->list == &task_groups)
tg = NULL;
return tg;
}

#define for_each_rt_rq(rt_rq, iter, rq) \
for (iter = container_of(&task_groups, typeof(*iter), list); \
(iter = next_task_group(iter)) && \
(rt_rq = iter->rt_rq[cpu_of(rq)]);)


Thanks,
Cheng
--
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/