Do not include throttled time as steal time
PATCH 1/1 Do not include throttled time as steal time
When the linux kernel is running as the guest OS and is configured for bandwidth control and steal time reporting, it can be confusing to users to see the throttled time show up in the steal time stats. The user will think they are not getting the cpu resources they have been configured.
--- kernel/sched_fair.c | 4 ++-- kernel/sched_stats.h | 7 ++++++- 2 files changed, 8 insertions(+), 3 deletions(-)
kernel/sched_fair.c +2 -2
@@ -707,7 +707,7 @@ account_entity_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se)#ifdef CONFIG_FAIR_GROUP_SCHED /* we need this in update_cfs_load and load-balance functions below */ -static inline int throttled_hierarchy (struct cfs_rq *cfs_rq); +inline int throttled_hierarchy (struct cfs_rq *cfs_rq);# ifdef CONFIG_SMP static void update_cfs_rq_load_contribution (struct cfs_rq *cfs_rq,int global_update)
@@ -1420,7 +1420,7 @@ static inline int cfs_rq_throttled(struct cfs_rq *cfs_rq) }/* check whether cfs_rq, or any parent, is throttled */ -static inline int throttled_hierarchy (struct cfs_rq *cfs_rq) +inline int throttled_hierarchy (struct cfs_rq *cfs_rq) {return cfs_rq->throttle_count; }
kernel/sched_stats.h +6 -1
@@ -213,14 +213,19 @@ static inline void sched_info_queued(struct task_struct *t)* sched_info_queued() to mark that it has now again started waiting on * the runqueue. */ +extern inline int throttled_hierarchy (struct cfs_rq *cfs_rq);static inline void sched_info_depart (struct task_struct *t) { +struct task_group *tg =task_group (t); +struct cfs_rq *cfs_rq;unsigned long long delta =task_rq (t)->clock - t->sched_info.last_arrival; + cfs_rq = tg->cfs_rq[smp_processor_id ()];rq_sched_info_depart (task_rq (t), delta); -if (t->state ==TASK_RUNNING ) + +if (t->state ==TASK_RUNNING && !throttled_hierarchy (cfs_rq))sched_info_queued (t); }
On Mon, 2011-12-05 at 09:08 -0600, Mike wolf wrote:
Hi Mike, couple of problems with this:
1) You failed to CC the appropriate maintainers for the piece of code you're trying to have changed. When in doubt see the MAINTAINERS file ;-)
2) You failed to CC the people who wrote the feature you're having a problem with.
Quoted 5 lines
When the linux kernel is running as the guest OS and is configured for bandwidth control and steal time reporting, it can be confusing to users to see the throttled time show up in the steal time stats. The user will think they are not getting the cpu resources they have been configured.
Supposedly this is a BAD (tm) thing :-)
Quoted 5 lines
Signed-off-by Mike Wolf <[email protected]>--- kernel/sched_fair.c | 4 ++-- kernel/sched_stats.h | 7 ++++++- 2 files changed, 8 insertions(+), 3 deletions(-)
3) You blink you loose, those files don't exist anymore. Patches are best provided against the development tree of the particular subsystem you're working against.
In this particular case tip/master is your target.
Quoted 5 lines
static inline void sched_info_depart(struct task_struct *t) {kernel/sched_stats.h +2
-0+struct task_group *tg =task_group (t); +struct cfs_rq *cfs_rq; unsigned long long delta = task_rq(t)->clock -
kernel/sched_stats.h static inline void sched_info_queued(struct task_struct *t) · line 230
Quoted 10 lines
kernel/sched_stats.h +3
-1t->sched_info.last_arrival; + cfs_rq = tg->cfs_rq[smp_processor_id ()];rq_sched_info_depart (task_rq (t), delta); -if (t->state ==TASK_RUNNING ) + +if (t->state ==TASK_RUNNING && !throttled_hierarchy (cfs_rq))sched_info_queued (t); }
4) so there's a lot more steal time crap all over the scheduler, you failed to explain why only this particular bit is important enough to change.
Quoted 7 lines · Peter Zijlstra
On Mon, 2011-12-05 at 09:08 -0600, Mike wolf wrote:
Hi Mike, couple of problems with this:
1) You failed to CC the appropriate maintainers for the piece of code you're trying to have changed. When in doubt see the
MAINTAINERSfile ;-)
yes will do
Quoted 2 lines
2) You failed to CC the people who wrote the feature you're having a problem with.
ok, I will add Glauber and others when the patch is respun/resubmitted
Quoted 28 lines
Quoted 5 lines
When the linux kernel is running as the guest OS and is configured for bandwidth control and steal time reporting, it can be confusing to users to see the throttled time show up in the steal time stats. The user will think they are not getting the cpu resources they have been configured.
Supposedly this is a BAD (tm) thing :-)
Quoted 5 lines
Signed-off-by Mike Wolf<[email protected]>--- kernel/sched_fair.c | 4 ++-- kernel/sched_stats.h | 7 ++++++- 2 files changed, 8 insertions(+), 3 deletions(-)3) You blink you loose, those files don't exist anymore. Patches are best provided against the development tree of the particular subsystem you're working against.
In this particular case tip/master is your target.
Quoted 10 lines
static inline void sched_info_depart(struct task_struct *t) {kernel/sched_stats.h +3
-0+struct task_group *tg =task_group (t); +struct cfs_rq *cfs_rq;unsigned long long delta =task_rq (t)->clock - t->sched_info.last_arrival; + cfs_rq = tg->cfs_rq[smp_processor_id ()];rq_sched_info_depart (task_rq (t), delta);
Quoted 8 lines
kernel/sched_stats.h static inline void sched_info_queued(struct task_struct *t) · line 230
Quoted 5 lines
kernel/sched_stats.h +2
-1-if (t->state ==TASK_RUNNING ) + +if (t->state ==TASK_RUNNING && !throttled_hierarchy (cfs_rq))sched_info_queued (t); }4) so there's a lot more steal time crap all over the scheduler, you failed to explain why only this particular bit is important enough to change.
I will make sure to explain why you would want the patch functionality better when I resubmit.