On Thu, Mar 21, 2019 at 05:20:17PM -0400, Julien Desfossez wrote: > On further investigation, we could see that the contention is mostly in the > way rq locks are taken. With this patchset, we lock the whole core if > cpu.tag is set for at least one cgroup. Due to this, __schedule() is more or > less serialized for the core and that attributes to the performance loss > that we are seeing. We also saw that newidle_balance() takes considerably > long time in load_balance() due to the rq spinlock contention. Do you think > it would help if the core-wide locking was only performed when absolutely > needed ?
Something like that could be done, but then you end up with 2 locks, something which I was hoping to avoid. Basically you keep rq->lock as it exists today, but add something like rq->core->core_lock, you then have to take that second lock (nested under rq->lock) for every scheduling action involving a tagged task. It makes things complicatd though; because now my head hurts thikning about pick_next_task(). (this can obviously do away with the whole rq->lock wrappery) Also, completely untested.. --- --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -146,6 +146,8 @@ void sched_core_enqueue(struct rq *rq, s if (!p->core_cookie) return; + raw_spin_lock(&rq->core->core_lock); + node = &rq->core_tree.rb_node; parent = *node; @@ -161,6 +163,8 @@ void sched_core_enqueue(struct rq *rq, s rb_link_node(&p->core_node, parent, node); rb_insert_color(&p->core_node, &rq->core_tree); + + raw_spin_unlock(&rq->core->core_lock); } void sched_core_dequeue(struct rq *rq, struct task_struct *p) @@ -170,7 +174,9 @@ void sched_core_dequeue(struct rq *rq, s if (!p->core_cookie) return; + raw_spin_lock(&rq->core->core_lock); rb_erase(&p->core_node, &rq->core_tree); + raw_spin_unlock(&rq->core->core_lock); } /* @@ -181,6 +187,8 @@ struct task_struct *sched_core_find(stru struct rb_node *node = rq->core_tree.rb_node; struct task_struct *node_task, *match; + lockdep_assert_held(&rq->core->core_lock); + /* * The idle task always matches any cookie! */ @@ -206,6 +214,8 @@ struct task_struct *sched_core_next(stru { struct rb_node *node = &p->core_node; + lockdep_assert_held(&rq->core->core_lock); + node = rb_next(node); if (!node) return NULL; @@ -3685,6 +3695,8 @@ pick_next_task(struct rq *rq, struct tas * If there were no {en,de}queues since we picked (IOW, the task * pointers are all still valid), and we haven't scheduled the last * pick yet, do so now. + * + * XXX probably OK without ->core_lock */ if (rq->core->core_pick_seq == rq->core->core_task_seq && rq->core->core_pick_seq != rq->core_sched_seq) { @@ -3710,6 +3722,20 @@ pick_next_task(struct rq *rq, struct tas if (!rq->nr_running) newidle_balance(rq, rf); + if (!rq->core->core_cookie) { + for_each_class(class) { + next = pick_task(rq, class, NULL); + if (next) + break; + } + + if (!next->core_cookie) { + set_next_task(rq, next); + return next; + } + } + + raw_spin_lock(&rq->core->core_lock); cpu = cpu_of(rq); smt_mask = cpu_smt_mask(cpu); @@ -3849,6 +3875,7 @@ next_class:; trace_printk("picked: %s/%d %lx\n", next->comm, next->pid, next->core_cookie); done: + raw_spin_unlock(&rq->core->core_lock); set_next_task(rq, next); return next; } --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -966,6 +966,7 @@ struct rq { struct rb_root core_tree; /* shared state */ + raw_spinlock_t core_lock; unsigned int core_task_seq; unsigned int core_pick_seq; unsigned long core_cookie; @@ -1007,9 +1008,6 @@ static inline bool sched_core_enabled(st static inline raw_spinlock_t *rq_lockp(struct rq *rq) { - if (sched_core_enabled(rq)) - return &rq->core->__lock; - return &rq->__lock; }