| Message ID | 20240711130004.2157737-5-vschneid@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | sched/fair: Defer CFS throttle to user entry | expand |
On Thu, Jul 11, 2024 at 02:59:58PM +0200, Valentin Schneider wrote: > +void init_cfs_throttle_work(struct task_struct *p) > +{ > + /* Protect against double add, see throttle_cfs_rq() and throttle_cfs_rq_work() */ > + p->sched_throttle_work.next = &p->sched_throttle_work; > + init_task_work(&p->sched_throttle_work, throttle_cfs_rq_work); Yes, init_task_work() does not write .next, but can we please flip these two statements and avoid me having to double check that every time I seem them? :-) > +}
On 12/07/24 17:21, Peter Zijlstra wrote: > On Thu, Jul 11, 2024 at 02:59:58PM +0200, Valentin Schneider wrote: > >> +void init_cfs_throttle_work(struct task_struct *p) >> +{ >> + /* Protect against double add, see throttle_cfs_rq() and throttle_cfs_rq_work() */ >> + p->sched_throttle_work.next = &p->sched_throttle_work; >> + init_task_work(&p->sched_throttle_work, throttle_cfs_rq_work); > > Yes, init_task_work() does not write .next, but can we please flip these > two statements and avoid me having to double check that every time I > seem them? :-) > Easy enough :) >> +}
diff --git a/include/linux/sched.h b/include/linux/sched.h index 90691d99027e3..a4976eb5065fc 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -812,6 +812,7 @@ struct task_struct { #ifdef CONFIG_CGROUP_SCHED struct task_group *sched_task_group; + struct callback_head sched_throttle_work; #endif diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 6d35c48239be0..b811670d2c362 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -4329,6 +4329,10 @@ static void __sched_fork(unsigned long clone_flags, struct task_struct *p) p->se.cfs_rq = NULL; #endif +#ifdef CONFIG_CFS_BANDWIDTH + init_cfs_throttle_work(p); +#endif + #ifdef CONFIG_SCHEDSTATS /* Even if schedstat is disabled, there should not be garbage */ memset(&p->stats, 0, sizeof(p->stats)); diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 9057584ec06de..775547cdd3ce0 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -5759,6 +5759,18 @@ static int tg_throttle_down(struct task_group *tg, void *data) return 0; } +static void throttle_cfs_rq_work(struct callback_head *work) +{ + +} + +void init_cfs_throttle_work(struct task_struct *p) +{ + /* Protect against double add, see throttle_cfs_rq() and throttle_cfs_rq_work() */ + p->sched_throttle_work.next = &p->sched_throttle_work; + init_task_work(&p->sched_throttle_work, throttle_cfs_rq_work); +} + static bool throttle_cfs_rq(struct cfs_rq *cfs_rq) { struct rq *rq = rq_of(cfs_rq); diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index 4c36cc6803617..943bca8263ffe 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -2510,6 +2510,8 @@ extern void init_sched_dl_class(void); extern void init_sched_rt_class(void); extern void init_sched_fair_class(void); +extern void init_cfs_throttle_work(struct task_struct *p); + extern void reweight_task(struct task_struct *p, const struct load_weight *lw); extern void resched_curr(struct rq *rq);
Later commits will change CFS bandwidth control throttling from a per-cfs_rq basis to a per-task basis. Actual throttling of a task will happen in the return to user path, which will be implemented via a task_work callback. To ease reviewing, the infrastructure and helpers are added first, the actual behaviour will be implemented when switching to per-task throttling. Add a task_work node to struct task_struct, and have it initialised at sched_fork(). Signed-off-by: Valentin Schneider <vschneid@redhat.com> --- include/linux/sched.h | 1 + kernel/sched/core.c | 4 ++++ kernel/sched/fair.c | 12 ++++++++++++ kernel/sched/sched.h | 2 ++ 4 files changed, 19 insertions(+)