We kill rq->rd on the CPU_DOWN_PREPARE stage:
cpuset_cpu_inactive -> cpuset_update_active_cpus -> partition_sched_domains ->
-> cpu_attach_domain -> rq_attach_root -> set_rq_offline
This unthrottles all throttled cfs_rqs.
But the cpu is still able to call schedule() till
take_cpu_down->__cpu_disable()
is called from stop_machine.
This case the tasks from just unthrottled cfs_rqs are pickable
in a standard scheduler way, and they are picked by dying cpu.
The cfs_rqs becomes throttled again, and migrate_tasks()
in migration_call skips their tasks (one more unthrottle
in migrate_tasks()->CPU_DYING does not happen, because rq->rd
is already NULL).
Patch sets runtime_enabled to zero. This guarantees, the runtime
is not accounted, and the cfs_rqs won't exceed given
cfs_rq->runtime_remaining = 1, and tasks will be pickable
in migrate_tasks(). runtime_enabled is recalculated again
when rq becomes online again.
Ben Segall also noticed, we always enable runtime in
tg_set_cfs_bandwidth(). Actually, we should do that for online
cpus only. To prevent races with unthrottle_offline_cfs_rqs()
we take get_online_cpus() lock.
v2: Fix race with tg_set_cfs_bandwidth().
Move cfs_rq->runtime_enabled=0 above unthrottle_cfs_rq().
v3: {get,put}_online_cpus()
Signed-off-by: Kirill Tkhai <[email protected]>
CC: Konstantin Khorenko <[email protected]>
CC: Ben Segall <[email protected]>
CC: Paul Turner <[email protected]>
CC: Srikar Dronamraju <[email protected]>
CC: Mike Galbraith <[email protected]>
CC: Peter Zijlstra <[email protected]>
CC: Ingo Molnar <[email protected]>
---
kernel/sched/core.c | 8 +++++++-
kernel/sched/fair.c | 22 ++++++++++++++++++++++
2 files changed, 29 insertions(+), 1 deletion(-)
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index ceea8d0..d9c4a08 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -7807,6 +7807,11 @@ static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
if (period > max_cfs_quota_period)
return -EINVAL;
+ /*
+ * Prevent race between setting of cfs_rq->runtime_enabled and
+ * unthrottle_offline_cfs_rqs().
+ */
+ get_online_cpus();
mutex_lock(&cfs_constraints_mutex);
ret = __cfs_schedulable(tg, period, quota);
if (ret)
@@ -7832,7 +7837,7 @@ static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
}
raw_spin_unlock_irq(&cfs_b->lock);
- for_each_possible_cpu(i) {
+ for_each_online_cpu(i) {
struct cfs_rq *cfs_rq = tg->cfs_rq[i];
struct rq *rq = cfs_rq->rq;
@@ -7848,6 +7853,7 @@ static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
cfs_bandwidth_usage_dec();
out_unlock:
mutex_unlock(&cfs_constraints_mutex);
+ put_online_cpus();
return ret;
}
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 1f9c457..5616d23 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -3776,6 +3776,19 @@ static void destroy_cfs_bandwidth(struct cfs_bandwidth *cfs_b)
hrtimer_cancel(&cfs_b->slack_timer);
}
+static void __maybe_unused update_runtime_enabled(struct rq *rq)
+{
+ struct cfs_rq *cfs_rq;
+
+ for_each_leaf_cfs_rq(rq, cfs_rq) {
+ struct cfs_bandwidth *cfs_b = &cfs_rq->tg->cfs_bandwidth;
+
+ raw_spin_lock(&cfs_b->lock);
+ cfs_rq->runtime_enabled = cfs_b->quota != RUNTIME_INF;
+ raw_spin_unlock(&cfs_b->lock);
+ }
+}
+
static void __maybe_unused unthrottle_offline_cfs_rqs(struct rq *rq)
{
struct cfs_rq *cfs_rq;
@@ -3789,6 +3802,12 @@ static void __maybe_unused unthrottle_offline_cfs_rqs(struct rq *rq)
* there's some valid quota amount
*/
cfs_rq->runtime_remaining = 1;
+ /*
+ * Offline rq is schedulable till cpu is completely disabled
+ * in take_cpu_down(), so we prevent new cfs throttling here.
+ */
+ cfs_rq->runtime_enabled = 0;
+
if (cfs_rq_throttled(cfs_rq))
unthrottle_cfs_rq(cfs_rq);
}
@@ -3832,6 +3851,7 @@ static inline struct cfs_bandwidth *tg_cfs_bandwidth(struct task_group *tg)
return NULL;
}
static inline void destroy_cfs_bandwidth(struct cfs_bandwidth *cfs_b) {}
+static inline void update_runtime_enabled(struct rq *rq) {}
static inline void unthrottle_offline_cfs_rqs(struct rq *rq) {}
#endif /* CONFIG_CFS_BANDWIDTH */
@@ -7325,6 +7345,8 @@ void trigger_load_balance(struct rq *rq)
static void rq_online_fair(struct rq *rq)
{
update_sysctl();
+
+ update_runtime_enabled(rq);
}
static void rq_offline_fair(struct rq *rq)
Kirill Tkhai <[email protected]> writes:
> We kill rq->rd on the CPU_DOWN_PREPARE stage:
>
> cpuset_cpu_inactive -> cpuset_update_active_cpus -> partition_sched_domains ->
> -> cpu_attach_domain -> rq_attach_root -> set_rq_offline
>
> This unthrottles all throttled cfs_rqs.
>
> But the cpu is still able to call schedule() till
>
> take_cpu_down->__cpu_disable()
>
> is called from stop_machine.
>
> This case the tasks from just unthrottled cfs_rqs are pickable
> in a standard scheduler way, and they are picked by dying cpu.
> The cfs_rqs becomes throttled again, and migrate_tasks()
> in migration_call skips their tasks (one more unthrottle
> in migrate_tasks()->CPU_DYING does not happen, because rq->rd
> is already NULL).
>
> Patch sets runtime_enabled to zero. This guarantees, the runtime
> is not accounted, and the cfs_rqs won't exceed given
> cfs_rq->runtime_remaining = 1, and tasks will be pickable
> in migrate_tasks(). runtime_enabled is recalculated again
> when rq becomes online again.
>
> Ben Segall also noticed, we always enable runtime in
> tg_set_cfs_bandwidth(). Actually, we should do that for online
> cpus only. To prevent races with unthrottle_offline_cfs_rqs()
> we take get_online_cpus() lock.
>
> v2: Fix race with tg_set_cfs_bandwidth().
> Move cfs_rq->runtime_enabled=0 above unthrottle_cfs_rq().
> v3: {get,put}_online_cpus()
>
> Signed-off-by: Kirill Tkhai <[email protected]>
> CC: Konstantin Khorenko <[email protected]>
> CC: Ben Segall <[email protected]>
> CC: Paul Turner <[email protected]>
> CC: Srikar Dronamraju <[email protected]>
> CC: Mike Galbraith <[email protected]>
> CC: Peter Zijlstra <[email protected]>
> CC: Ingo Molnar <[email protected]>
Reviewed-by: Ben Segall <[email protected]>
Although #ifdefing rq->online might work too. It's possibly faster but a
bit more ugly to look at, I don't care which way it is done.
> ---
> kernel/sched/core.c | 8 +++++++-
> kernel/sched/fair.c | 22 ++++++++++++++++++++++
> 2 files changed, 29 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index ceea8d0..d9c4a08 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -7807,6 +7807,11 @@ static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
> if (period > max_cfs_quota_period)
> return -EINVAL;
>
> + /*
> + * Prevent race between setting of cfs_rq->runtime_enabled and
> + * unthrottle_offline_cfs_rqs().
> + */
> + get_online_cpus();
> mutex_lock(&cfs_constraints_mutex);
> ret = __cfs_schedulable(tg, period, quota);
> if (ret)
> @@ -7832,7 +7837,7 @@ static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
> }
> raw_spin_unlock_irq(&cfs_b->lock);
>
> - for_each_possible_cpu(i) {
> + for_each_online_cpu(i) {
> struct cfs_rq *cfs_rq = tg->cfs_rq[i];
> struct rq *rq = cfs_rq->rq;
>
> @@ -7848,6 +7853,7 @@ static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
> cfs_bandwidth_usage_dec();
> out_unlock:
> mutex_unlock(&cfs_constraints_mutex);
> + put_online_cpus();
>
> return ret;
> }
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index 1f9c457..5616d23 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -3776,6 +3776,19 @@ static void destroy_cfs_bandwidth(struct cfs_bandwidth *cfs_b)
> hrtimer_cancel(&cfs_b->slack_timer);
> }
>
> +static void __maybe_unused update_runtime_enabled(struct rq *rq)
> +{
> + struct cfs_rq *cfs_rq;
> +
> + for_each_leaf_cfs_rq(rq, cfs_rq) {
> + struct cfs_bandwidth *cfs_b = &cfs_rq->tg->cfs_bandwidth;
> +
> + raw_spin_lock(&cfs_b->lock);
> + cfs_rq->runtime_enabled = cfs_b->quota != RUNTIME_INF;
> + raw_spin_unlock(&cfs_b->lock);
> + }
> +}
> +
> static void __maybe_unused unthrottle_offline_cfs_rqs(struct rq *rq)
> {
> struct cfs_rq *cfs_rq;
> @@ -3789,6 +3802,12 @@ static void __maybe_unused unthrottle_offline_cfs_rqs(struct rq *rq)
> * there's some valid quota amount
> */
> cfs_rq->runtime_remaining = 1;
> + /*
> + * Offline rq is schedulable till cpu is completely disabled
> + * in take_cpu_down(), so we prevent new cfs throttling here.
> + */
> + cfs_rq->runtime_enabled = 0;
> +
> if (cfs_rq_throttled(cfs_rq))
> unthrottle_cfs_rq(cfs_rq);
> }
> @@ -3832,6 +3851,7 @@ static inline struct cfs_bandwidth *tg_cfs_bandwidth(struct task_group *tg)
> return NULL;
> }
> static inline void destroy_cfs_bandwidth(struct cfs_bandwidth *cfs_b) {}
> +static inline void update_runtime_enabled(struct rq *rq) {}
> static inline void unthrottle_offline_cfs_rqs(struct rq *rq) {}
>
> #endif /* CONFIG_CFS_BANDWIDTH */
> @@ -7325,6 +7345,8 @@ void trigger_load_balance(struct rq *rq)
> static void rq_online_fair(struct rq *rq)
> {
> update_sysctl();
> +
> + update_runtime_enabled(rq);
> }
>
> static void rq_offline_fair(struct rq *rq)
Commit-ID: 0e59bdaea75f12a7d7c03672f4ac22c0119a1bc0
Gitweb: http://git.kernel.org/tip/0e59bdaea75f12a7d7c03672f4ac22c0119a1bc0
Author: Kirill Tkhai <[email protected]>
AuthorDate: Wed, 25 Jun 2014 12:19:42 +0400
Committer: Ingo Molnar <[email protected]>
CommitDate: Sat, 5 Jul 2014 11:17:42 +0200
sched/fair: Disable runtime_enabled on dying rq
We kill rq->rd on the CPU_DOWN_PREPARE stage:
cpuset_cpu_inactive -> cpuset_update_active_cpus -> partition_sched_domains ->
-> cpu_attach_domain -> rq_attach_root -> set_rq_offline
This unthrottles all throttled cfs_rqs.
But the cpu is still able to call schedule() till
take_cpu_down->__cpu_disable()
is called from stop_machine.
This case the tasks from just unthrottled cfs_rqs are pickable
in a standard scheduler way, and they are picked by dying cpu.
The cfs_rqs becomes throttled again, and migrate_tasks()
in migration_call skips their tasks (one more unthrottle
in migrate_tasks()->CPU_DYING does not happen, because rq->rd
is already NULL).
Patch sets runtime_enabled to zero. This guarantees, the runtime
is not accounted, and the cfs_rqs won't exceed given
cfs_rq->runtime_remaining = 1, and tasks will be pickable
in migrate_tasks(). runtime_enabled is recalculated again
when rq becomes online again.
Ben Segall also noticed, we always enable runtime in
tg_set_cfs_bandwidth(). Actually, we should do that for online
cpus only. To prevent races with unthrottle_offline_cfs_rqs()
we take get_online_cpus() lock.
Reviewed-by: Ben Segall <[email protected]>
Reviewed-by: Srikar Dronamraju <[email protected]>
Signed-off-by: Kirill Tkhai <[email protected]>
CC: Konstantin Khorenko <[email protected]>
CC: Paul Turner <[email protected]>
CC: Mike Galbraith <[email protected]>
Cc: Linus Torvalds <[email protected]>
Signed-off-by: Peter Zijlstra <[email protected]>
Link: http://lkml.kernel.org/r/1403684382.3462.42.camel@tkhai
Signed-off-by: Ingo Molnar <[email protected]>
---
kernel/sched/core.c | 8 +++++++-
kernel/sched/fair.c | 22 ++++++++++++++++++++++
2 files changed, 29 insertions(+), 1 deletion(-)
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index e50234b..2dbc63d 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -7817,6 +7817,11 @@ static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
if (period > max_cfs_quota_period)
return -EINVAL;
+ /*
+ * Prevent race between setting of cfs_rq->runtime_enabled and
+ * unthrottle_offline_cfs_rqs().
+ */
+ get_online_cpus();
mutex_lock(&cfs_constraints_mutex);
ret = __cfs_schedulable(tg, period, quota);
if (ret)
@@ -7842,7 +7847,7 @@ static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
}
raw_spin_unlock_irq(&cfs_b->lock);
- for_each_possible_cpu(i) {
+ for_each_online_cpu(i) {
struct cfs_rq *cfs_rq = tg->cfs_rq[i];
struct rq *rq = cfs_rq->rq;
@@ -7858,6 +7863,7 @@ static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota)
cfs_bandwidth_usage_dec();
out_unlock:
mutex_unlock(&cfs_constraints_mutex);
+ put_online_cpus();
return ret;
}
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index a140c6a..923fe32 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -3798,6 +3798,19 @@ static void destroy_cfs_bandwidth(struct cfs_bandwidth *cfs_b)
hrtimer_cancel(&cfs_b->slack_timer);
}
+static void __maybe_unused update_runtime_enabled(struct rq *rq)
+{
+ struct cfs_rq *cfs_rq;
+
+ for_each_leaf_cfs_rq(rq, cfs_rq) {
+ struct cfs_bandwidth *cfs_b = &cfs_rq->tg->cfs_bandwidth;
+
+ raw_spin_lock(&cfs_b->lock);
+ cfs_rq->runtime_enabled = cfs_b->quota != RUNTIME_INF;
+ raw_spin_unlock(&cfs_b->lock);
+ }
+}
+
static void __maybe_unused unthrottle_offline_cfs_rqs(struct rq *rq)
{
struct cfs_rq *cfs_rq;
@@ -3811,6 +3824,12 @@ static void __maybe_unused unthrottle_offline_cfs_rqs(struct rq *rq)
* there's some valid quota amount
*/
cfs_rq->runtime_remaining = 1;
+ /*
+ * Offline rq is schedulable till cpu is completely disabled
+ * in take_cpu_down(), so we prevent new cfs throttling here.
+ */
+ cfs_rq->runtime_enabled = 0;
+
if (cfs_rq_throttled(cfs_rq))
unthrottle_cfs_rq(cfs_rq);
}
@@ -3854,6 +3873,7 @@ static inline struct cfs_bandwidth *tg_cfs_bandwidth(struct task_group *tg)
return NULL;
}
static inline void destroy_cfs_bandwidth(struct cfs_bandwidth *cfs_b) {}
+static inline void update_runtime_enabled(struct rq *rq) {}
static inline void unthrottle_offline_cfs_rqs(struct rq *rq) {}
#endif /* CONFIG_CFS_BANDWIDTH */
@@ -7362,6 +7382,8 @@ void trigger_load_balance(struct rq *rq)
static void rq_online_fair(struct rq *rq)
{
update_sysctl();
+
+ update_runtime_enabled(rq);
}
static void rq_offline_fair(struct rq *rq)