Messages in this thread Patch in this message | | | From | Juri Lelli <> | Subject | [RFD PATCH 5/5] sched/deadline: make bandwidth enforcement scale-invariant | Date | Fri, 24 Mar 2017 14:09:00 +0000 |
| |
Apply frequency and cpu scale-invariance correction factor to bandwidth enforcement (similar to what we already do to fair utilization tracking).
Each delta_exec gets scaled considering current frequency and maximum cpu capacity; which means that the reservation runtime parameter (that need to be specified profiling the task execution at max frequency on biggest capacity core) gets thus scaled accordingly.
Signed-off-by: Juri Lelli <juri.lelli@arm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: Luca Abeni <luca.abeni@santannapisa.it> Cc: Claudio Scordino <claudio@evidence.eu.com> --- kernel/sched/deadline.c | 27 +++++++++++++++++++++++---- kernel/sched/fair.c | 2 -- kernel/sched/sched.h | 2 ++ 3 files changed, 25 insertions(+), 6 deletions(-)
diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c index 853de524c6c6..7141d6f51ee0 100644 --- a/kernel/sched/deadline.c +++ b/kernel/sched/deadline.c @@ -940,7 +940,9 @@ static void update_curr_dl(struct rq *rq) { struct task_struct *curr = rq->curr; struct sched_dl_entity *dl_se = &curr->dl; - u64 delta_exec; + u64 delta_exec, scaled_delta_exec; + unsigned long scale_freq, scale_cpu; + int cpu = cpu_of(rq); if (!dl_task(curr) || !on_dl_rq(dl_se)) return; @@ -974,9 +976,26 @@ static void update_curr_dl(struct rq *rq) if (unlikely(dl_entity_is_special(dl_se))) return; - if (unlikely(dl_se->flags & SCHED_FLAG_RECLAIM)) - delta_exec = grub_reclaim(delta_exec, rq, curr->dl.dl_bw); - dl_se->runtime -= delta_exec; + /* + * XXX When clock frequency is controlled by the scheduler (via + * schedutil governor) we implement GRUB-PA: the spare reclaimed + * bandwidth is used to clock down frequency. + * + * However, what below seems to assume scheduler to always be in + * control of clock frequency; when running at a fixed frequency + * (e.g., performance or userspace governor), shouldn't we instead + * use the grub_reclaim mechanism below? + * + * if (unlikely(dl_se->flags & SCHED_FLAG_RECLAIM)) + * delta_exec = grub_reclaim(delta_exec, rq, curr->dl.dl_bw); + * dl_se->runtime -= delta_exec; + */ + scale_freq = arch_scale_freq_capacity(NULL, cpu); + scale_cpu = arch_scale_cpu_capacity(NULL, cpu); + + scaled_delta_exec = cap_scale(delta_exec, scale_freq); + scaled_delta_exec = cap_scale(scaled_delta_exec, scale_cpu); + dl_se->runtime -= scaled_delta_exec; throttle: if (dl_runtime_exceeded(dl_se) || dl_se->dl_yielded) { diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 2805bd7c8994..37f12d0a3bc4 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -2818,8 +2818,6 @@ static u32 __compute_runnable_contrib(u64 n) return contrib + runnable_avg_yN_sum[n]; } -#define cap_scale(v, s) ((v)*(s) >> SCHED_CAPACITY_SHIFT) - /* * We can represent the historical contribution to runnable average as the * coefficients of a geometric series. To do this we sub-divide our runnable diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index 7b5e81120813..81bd048ed181 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -155,6 +155,8 @@ static inline int task_has_dl_policy(struct task_struct *p) return dl_policy(p->policy); } +#define cap_scale(v, s) ((v)*(s) >> SCHED_CAPACITY_SHIFT) + static inline int dl_entity_is_special(struct sched_dl_entity *dl_se) { return dl_se->flags & SCHED_FLAG_SPECIAL; -- 2.10.0
| |