| Message ID | 20231013151450.257891-2-vincent.guittot@linaro.org |
|---|---|
| State | Superseded |
| Headers | show |
| Series | Rework interface between scheduler and schedutil governor | expand |
* Vincent Guittot <vincent.guittot@linaro.org> wrote: > + > + /* The minimum utilization returns the highest level between: > + * - the computed DL bandwidth needed with the irq pressure which > + * steals time to the deadline task. > + * - The minimum bandwidth requirement for CFS. > + */ Nit: please use the standard multi-line Linux kernel comment style. > + /* The maximum hint is a soft bandwidth requirement which can be lower > + * than the actual utilization because of max uclamp requirments > */ Ditto. > +unsigned long sugov_effective_cpu_perf(int cpu, unsigned long actual, > + unsigned long min, > + unsigned long max) > +{ > + unsigned long target; > + struct rq *rq = cpu_rq(cpu); > + > + if (rt_rq_is_runnable(&rq->rt)) > + return max; > + > + /* Provide at least enough capacity for DL + irq */ > + target = min; s/ / / s/irq/IRQ/ Thanks, Ingo
On Fri, 13 Oct 2023 at 20:21, Ingo Molnar <mingo@kernel.org> wrote: > > > * Vincent Guittot <vincent.guittot@linaro.org> wrote: > > > + > > + /* The minimum utilization returns the highest level between: > > + * - the computed DL bandwidth needed with the irq pressure which > > + * steals time to the deadline task. > > + * - The minimum bandwidth requirement for CFS. > > + */ > > Nit: please use the standard multi-line Linux kernel comment style. Yes, I don't know how I ended up with such comment style. I will fix it and others below > > > + /* The maximum hint is a soft bandwidth requirement which can be lower > > + * than the actual utilization because of max uclamp requirments > > */ > > Ditto. > > > +unsigned long sugov_effective_cpu_perf(int cpu, unsigned long actual, > > + unsigned long min, > > + unsigned long max) > > +{ > > + unsigned long target; > > + struct rq *rq = cpu_rq(cpu); > > + > > + if (rt_rq_is_runnable(&rq->rt)) > > + return max; > > + > > + /* Provide at least enough capacity for DL + irq */ > > + target = min; > > s/ / / > s/irq/IRQ/ > > Thanks, > > Ingo
On 10/18/23 08:04, Beata Michalska wrote: > Hi Vincent, > > On Fri, Oct 13, 2023 at 05:14:49PM +0200, Vincent Guittot wrote: >> The current method to take into account uclamp hints when estimating the >> target frequency can end into situation where the selected target >> frequency is finally higher than uclamp hints whereas there are no real >> needs. Such cases mainly happen because we are currently mixing the >> traditional scheduler utilization signal with the uclamp performance >> hints. By adding these 2 metrics, we loose an important information when >> it comes to select the target frequency and we have to make some >> assumptions which can't fit all cases. [snip] >> >> diff --git a/include/linux/energy_model.h b/include/linux/energy_model.h >> index b9caa01dfac4..adec808b371a 100644 >> --- a/include/linux/energy_model.h >> +++ b/include/linux/energy_model.h >> @@ -243,7 +243,6 @@ static inline unsigned long em_cpu_energy(struct em_perf_domain *pd, >> scale_cpu = arch_scale_cpu_capacity(cpu); >> ps = &pd->table[pd->nr_perf_states - 1]; >> >> - max_util = map_util_perf(max_util); > Even though the effective_cpu_util does no longer include the headroom, it is > being applied by sugov further down the line (sugov_effective_cpu_perf). > Won't that bring back the original problem when freq selection within EM is > not align with the one performed by sugov ? It should be OK here to remove the above line. The map_util_perf() is done before this em_cpu_energy() call, in the new code in [snip] >> } >> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c >> index 922905194c0c..d4f7b2f49c44 100644 >> --- a/kernel/sched/fair.c >> +++ b/kernel/sched/fair.c >> @@ -7628,7 +7628,7 @@ static inline void eenv_pd_busy_time(struct energy_env *eenv, >> for_each_cpu(cpu, pd_cpus) { >> unsigned long util = cpu_util(cpu, p, -1, 0); >> >> - busy_time += effective_cpu_util(cpu, util, ENERGY_UTIL, NULL); >> + busy_time += effective_cpu_util(cpu, util, NULL, NULL); >> } >> >> eenv->pd_busy_time = min(eenv->pd_cap, busy_time); >> @@ -7651,7 +7651,7 @@ eenv_pd_max_util(struct energy_env *eenv, struct cpumask *pd_cpus, >> for_each_cpu(cpu, pd_cpus) { >> struct task_struct *tsk = (cpu == dst_cpu) ? p : NULL; >> unsigned long util = cpu_util(cpu, p, dst_cpu, 1); >> - unsigned long eff_util; >> + unsigned long eff_util, min, max; >> >> /* >> * Performance domain frequency: utilization clamping >> @@ -7660,7 +7660,23 @@ eenv_pd_max_util(struct energy_env *eenv, struct cpumask *pd_cpus, >> * NOTE: in case RT tasks are running, by default the >> * FREQUENCY_UTIL's utilization can be max OPP. >> */ >> - eff_util = effective_cpu_util(cpu, util, FREQUENCY_UTIL, tsk); >> + eff_util = effective_cpu_util(cpu, util, &min, &max); >> + >> + /* Task's uclamp can modify min and max value */ >> + if (tsk && uclamp_is_used()) { >> + min = max(min, uclamp_eff_value(p, UCLAMP_MIN)); >> + >> + /* >> + * If there is no active max uclamp constraint, >> + * directly use task's one otherwise keep max >> + */ >> + if (uclamp_rq_is_idle(cpu_rq(cpu))) >> + max = uclamp_eff_value(p, UCLAMP_MAX); >> + else >> + max = max(max, uclamp_eff_value(p, UCLAMP_MAX)); >> + } >> + >> + eff_util = sugov_effective_cpu_perf(cpu, eff_util, min, max); > This will include the headroom so won't it inflate the util here ? Yes, that's the goal. It will inflate when needed. Currently, the problem is that we always inflate (blindly) in the em_cpu_energy(). We don't know if the util value which is comming is from uclamp_max and the frequency should not be higher, because something want to clamp it. The other question would be: What if the PD has 4 CPUs, the max util found is 500 and is from uclamp_max, but there is onother util on some other CPU 490? That CPU is allowed and can have the +20% freq in voting. In current design we don't punish the whole domain in such scenario.
On Wed, 18 Oct 2023 at 09:05, Beata Michalska <beata.michalska@arm.com> wrote: > > Hi Vincent, > > On Fri, Oct 13, 2023 at 05:14:49PM +0200, Vincent Guittot wrote: > > The current method to take into account uclamp hints when estimating the > > target frequency can end into situation where the selected target > > frequency is finally higher than uclamp hints whereas there are no real > > needs. Such cases mainly happen because we are currently mixing the > > traditional scheduler utilization signal with the uclamp performance > > hints. By adding these 2 metrics, we loose an important information when > > it comes to select the target frequency and we have to make some > > assumptions which can't fit all cases. > > > > Rework the interface between the scheduler and schedutil governor in order > > to propagate all information down to the cpufreq governor. > /governor/driver ? no only schedutil governor uses the interface > > > > effective_cpu_util() interface changes and now returns the actual > > utilization of the CPU with 2 optional inputs: > > - The minimum performance for this CPU; typically the capacity to handle > > the deadline task and the interrupt pressure. But also uclamp_min > > request when available. > > - The maximum targeting performance for this CPU which reflects the > > maximum level that we would like to not exceed. By default it will be > > the CPU capacity but can be reduced because of some performance hints > > set with uclamp. The value can be lower than actual utilization and/or > > min performance level. > > > > A new sugov_effective_cpu_perf() interface is also available to compute > > the final performance level that is targeted for the CPU after applying > > some cpufreq headroom and taking into account all inputs. > > > > With these 2 functions, schedutil is now able to decide when it must go > > above uclamp hints. It now also have a generic way to get the min > > perfromance level. > > > > The dependency between energy model and cpufreq governor and its headroom > > policy doesn't exist anymore. > > > > eenv_pd_max_util asks schedutil for the targeted performance after > > applying the impact of the waking task. > > > > Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org> > > --- > > include/linux/energy_model.h | 1 - > > kernel/sched/core.c | 85 ++++++++++++-------------------- > > kernel/sched/cpufreq_schedutil.c | 43 ++++++++++++---- > > kernel/sched/fair.c | 22 +++++++-- > > kernel/sched/sched.h | 24 +++------ > > 5 files changed, 91 insertions(+), 84 deletions(-) > > > > diff --git a/include/linux/energy_model.h b/include/linux/energy_model.h > > index b9caa01dfac4..adec808b371a 100644 > > --- a/include/linux/energy_model.h > > +++ b/include/linux/energy_model.h > > @@ -243,7 +243,6 @@ static inline unsigned long em_cpu_energy(struct em_perf_domain *pd, > > scale_cpu = arch_scale_cpu_capacity(cpu); > > ps = &pd->table[pd->nr_perf_states - 1]; > > > > - max_util = map_util_perf(max_util); > Even though the effective_cpu_util does no longer include the headroom, it is > being applied by sugov further down the line (sugov_effective_cpu_perf). > Won't that bring back the original problem when freq selection within EM is > not align with the one performed by sugov ? They are both aligned. Here, the goal is to move the headroom computation and decision inside schedutil function instead of taking care to apply the same headroom policy everywhere in the scheduler > > max_util = min(max_util, allowed_cpu_cap); > > freq = map_util_freq(max_util, ps->frequency, scale_cpu); > > > > diff --git a/kernel/sched/core.c b/kernel/sched/core.c > > index a3f9cd52eec5..78228abd1219 100644 > > --- a/kernel/sched/core.c > > +++ b/kernel/sched/core.c > > @@ -7381,18 +7381,13 @@ int sched_core_idle_cpu(int cpu) > > * required to meet deadlines. > > */ > > unsigned long effective_cpu_util(int cpu, unsigned long util_cfs, > > - enum cpu_util_type type, > > - struct task_struct *p) > > + unsigned long *min, > > + unsigned long *max) > > { > > - unsigned long dl_util, util, irq, max; > > + unsigned long util, irq, scale; > > struct rq *rq = cpu_rq(cpu); > > > > - max = arch_scale_cpu_capacity(cpu); > > - > > - if (!uclamp_is_used() && > > - type == FREQUENCY_UTIL && rt_rq_is_runnable(&rq->rt)) { > > - return max; > > - } > > + scale = arch_scale_cpu_capacity(cpu); > > > > /* > > * Early check to see if IRQ/steal time saturates the CPU, can be > > @@ -7400,45 +7395,36 @@ unsigned long effective_cpu_util(int cpu, unsigned long util_cfs, > > * update_irq_load_avg(). > > */ > > irq = cpu_util_irq(rq); > > - if (unlikely(irq >= max)) > > - return max; > > + if (unlikely(irq >= scale)) { > > + if (min) > > + *min = scale; > > + if (max) > > + *max = scale; > > + return scale; > > + } > > + > > + /* The minimum utilization returns the highest level between: > > + * - the computed DL bandwidth needed with the irq pressure which > > + * steals time to the deadline task. > > + * - The minimum bandwidth requirement for CFS. > > + */ > > + if (min) > > + *min = max(irq + cpu_bw_dl(rq), uclamp_rq_get(rq, UCLAMP_MIN)); > > > > /* > > * Because the time spend on RT/DL tasks is visible as 'lost' time to > > * CFS tasks and we use the same metric to track the effective > > * utilization (PELT windows are synchronized) we can directly add them > > * to obtain the CPU's actual utilization. > > - * > > - * CFS and RT utilization can be boosted or capped, depending on > > - * utilization clamp constraints requested by currently RUNNABLE > > - * tasks. > > - * When there are no CFS RUNNABLE tasks, clamps are released and > > - * frequency will be gracefully reduced with the utilization decay. > > */ > > util = util_cfs + cpu_util_rt(rq); > > - if (type == FREQUENCY_UTIL) > > - util = uclamp_rq_util_with(rq, util, p); > > - > > - dl_util = cpu_util_dl(rq); > > - > > - /* > > - * For frequency selection we do not make cpu_util_dl() a permanent part > > - * of this sum because we want to use cpu_bw_dl() later on, but we need > > - * to check if the CFS+RT+DL sum is saturated (ie. no idle time) such > > - * that we select f_max when there is no idle time. > > - * > > - * NOTE: numerical errors or stop class might cause us to not quite hit > > - * saturation when we should -- something for later. > > - */ > > - if (util + dl_util >= max) > > - return max; > > + util += cpu_util_dl(rq); > > > > - /* > > - * OTOH, for energy computation we need the estimated running time, so > > - * include util_dl and ignore dl_bw. > > - */ > > - if (type == ENERGY_UTIL) > > - util += dl_util; > > + if (util >= scale) { > > + if (max) > > + *max = scale; > > + return scale; > > + } > > > > /* > > * There is still idle time; further improve the number by using the > > @@ -7449,28 +7435,21 @@ unsigned long effective_cpu_util(int cpu, unsigned long util_cfs, > > * U' = irq + --------- * U > > * max > > */ > > - util = scale_irq_capacity(util, irq, max); > > + util = scale_irq_capacity(util, irq, scale); > > util += irq; > > > > - /* > > - * Bandwidth required by DEADLINE must always be granted while, for > > - * FAIR and RT, we use blocked utilization of IDLE CPUs as a mechanism > > - * to gracefully reduce the frequency when no tasks show up for longer > > - * periods of time. > > - * > > - * Ideally we would like to set bw_dl as min/guaranteed freq and util + > > - * bw_dl as requested freq. However, cpufreq is not yet ready for such > > - * an interface. So, we only do the latter for now. > > + /* The maximum hint is a soft bandwidth requirement which can be lower > > + * than the actual utilization because of max uclamp requirments > > */ > > - if (type == FREQUENCY_UTIL) > > - util += cpu_bw_dl(rq); > > + if (max) > > + *max = min(scale, uclamp_rq_get(rq, UCLAMP_MAX)); > > > > - return min(max, util); > > + return min(scale, util); > > } > > > > unsigned long sched_cpu_util(int cpu) > > { > > - return effective_cpu_util(cpu, cpu_util_cfs(cpu), ENERGY_UTIL, NULL); > > + return effective_cpu_util(cpu, cpu_util_cfs(cpu), NULL, NULL); > > } > > #endif /* CONFIG_SMP */ > > > > diff --git a/kernel/sched/cpufreq_schedutil.c b/kernel/sched/cpufreq_schedutil.c > > index 458d359f5991..8cb323522b90 100644 > > --- a/kernel/sched/cpufreq_schedutil.c > > +++ b/kernel/sched/cpufreq_schedutil.c > > @@ -47,7 +47,7 @@ struct sugov_cpu { > > u64 last_update; > > > > unsigned long util; > > - unsigned long bw_dl; > > + unsigned long bw_min; > > > > /* The field below is for single-CPU policies only: */ > > #ifdef CONFIG_NO_HZ_COMMON > > @@ -143,7 +143,6 @@ static unsigned int get_next_freq(struct sugov_policy *sg_policy, > > unsigned int freq = arch_scale_freq_invariant() ? > > policy->cpuinfo.max_freq : policy->cur; > > > > - util = map_util_perf(util); > > freq = map_util_freq(util, freq, max); > > > > if (freq == sg_policy->cached_raw_freq && !sg_policy->need_freq_update) > > @@ -153,14 +152,38 @@ static unsigned int get_next_freq(struct sugov_policy *sg_policy, > > return cpufreq_driver_resolve_freq(policy, freq); > > } > > > > +unsigned long sugov_effective_cpu_perf(int cpu, unsigned long actual, > > + unsigned long min, > > + unsigned long max) > > +{ > > + unsigned long target; > > + struct rq *rq = cpu_rq(cpu); > > + > > + if (rt_rq_is_runnable(&rq->rt)) > > + return max; > > + > > + /* Provide at least enough capacity for DL + irq */ > > + target = min; > Minor, but it might be worth to mention UCLAMP_MIN here, otherwise the return > statement might get bit confusing. yes, i can add it > > + > > + actual = map_util_perf(actual); > > + /* Actually we don't need to target the max performance */ > > + if (actual < max) > > + max = actual; > > + > > + /* > > + * Ensure at least minimum performance while providing more compute > > + * capacity when possible. > > + */ > > + return max(target, max); > > +} > > + > > static void sugov_get_util(struct sugov_cpu *sg_cpu) > > { > > - unsigned long util = cpu_util_cfs_boost(sg_cpu->cpu); > > - struct rq *rq = cpu_rq(sg_cpu->cpu); > > + unsigned long min, max, util = cpu_util_cfs_boost(sg_cpu->cpu); > > > > - sg_cpu->bw_dl = cpu_bw_dl(rq); > > - sg_cpu->util = effective_cpu_util(sg_cpu->cpu, util, > > - FREQUENCY_UTIL, NULL); > > + util = effective_cpu_util(sg_cpu->cpu, util, &min, &max); > > + sg_cpu->bw_min = map_util_perf(min); > > + sg_cpu->util = sugov_effective_cpu_perf(sg_cpu->cpu, util, min, max); > > } > > > > /** > > @@ -306,7 +329,7 @@ static inline bool sugov_cpu_is_busy(struct sugov_cpu *sg_cpu) { return false; } > > */ > > static inline void ignore_dl_rate_limit(struct sugov_cpu *sg_cpu) > > { > > - if (cpu_bw_dl(cpu_rq(sg_cpu->cpu)) > sg_cpu->bw_dl) > > + if (cpu_bw_dl(cpu_rq(sg_cpu->cpu)) > sg_cpu->bw_min) > > sg_cpu->sg_policy->limits_changed = true; > Because the bw_min now includes the headroom, this might not catch potential > changes in DL util, not sure though if that could be an issue. this is ok, because the min is still higher than cpu_bw_dl() > > } > > > > @@ -407,8 +430,8 @@ static void sugov_update_single_perf(struct update_util_data *hook, u64 time, > > sugov_cpu_is_busy(sg_cpu) && sg_cpu->util < prev_util) > > sg_cpu->util = prev_util; > > > > - cpufreq_driver_adjust_perf(sg_cpu->cpu, map_util_perf(sg_cpu->bw_dl), > > - map_util_perf(sg_cpu->util), max_cap); > > + cpufreq_driver_adjust_perf(sg_cpu->cpu, sg_cpu->bw_min, > > + sg_cpu->util, max_cap); > > > > sg_cpu->sg_policy->last_freq_update_time = time; > > } > > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c > > index 922905194c0c..d4f7b2f49c44 100644 > > --- a/kernel/sched/fair.c > > +++ b/kernel/sched/fair.c > > @@ -7628,7 +7628,7 @@ static inline void eenv_pd_busy_time(struct energy_env *eenv, > > for_each_cpu(cpu, pd_cpus) { > > unsigned long util = cpu_util(cpu, p, -1, 0); > > > > - busy_time += effective_cpu_util(cpu, util, ENERGY_UTIL, NULL); > > + busy_time += effective_cpu_util(cpu, util, NULL, NULL); > > } > > > > eenv->pd_busy_time = min(eenv->pd_cap, busy_time); > > @@ -7651,7 +7651,7 @@ eenv_pd_max_util(struct energy_env *eenv, struct cpumask *pd_cpus, > > for_each_cpu(cpu, pd_cpus) { > > struct task_struct *tsk = (cpu == dst_cpu) ? p : NULL; > > unsigned long util = cpu_util(cpu, p, dst_cpu, 1); > > - unsigned long eff_util; > > + unsigned long eff_util, min, max; > > > > /* > > * Performance domain frequency: utilization clamping > > @@ -7660,7 +7660,23 @@ eenv_pd_max_util(struct energy_env *eenv, struct cpumask *pd_cpus, > > * NOTE: in case RT tasks are running, by default the > > * FREQUENCY_UTIL's utilization can be max OPP. > > */ > > - eff_util = effective_cpu_util(cpu, util, FREQUENCY_UTIL, tsk); > > + eff_util = effective_cpu_util(cpu, util, &min, &max); > > + > > + /* Task's uclamp can modify min and max value */ > > + if (tsk && uclamp_is_used()) { > > + min = max(min, uclamp_eff_value(p, UCLAMP_MIN)); > > + > > + /* > > + * If there is no active max uclamp constraint, > > + * directly use task's one otherwise keep max > > + */ > > + if (uclamp_rq_is_idle(cpu_rq(cpu))) > > + max = uclamp_eff_value(p, UCLAMP_MAX); > > + else > > + max = max(max, uclamp_eff_value(p, UCLAMP_MAX)); > > + } > > + > > + eff_util = sugov_effective_cpu_perf(cpu, eff_util, min, max); > This will include the headroom so won't it inflate the util here ? If you look into sugov_effective_cpu_perf(), the headroom is not applied on everything now and in particular not on the max. > > --- > BR > B. > > max_util = max(max_util, eff_util); > > } > > > > diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h > > index 65cad0e5729e..3873b4de7cfa 100644 > > --- a/kernel/sched/sched.h > > +++ b/kernel/sched/sched.h > > @@ -2962,24 +2962,14 @@ static inline void cpufreq_update_util(struct rq *rq, unsigned int flags) {} > > #endif > > > > #ifdef CONFIG_SMP > > -/** > > - * enum cpu_util_type - CPU utilization type > > - * @FREQUENCY_UTIL: Utilization used to select frequency > > - * @ENERGY_UTIL: Utilization used during energy calculation > > - * > > - * The utilization signals of all scheduling classes (CFS/RT/DL) and IRQ time > > - * need to be aggregated differently depending on the usage made of them. This > > - * enum is used within effective_cpu_util() to differentiate the types of > > - * utilization expected by the callers, and adjust the aggregation accordingly. > > - */ > > -enum cpu_util_type { > > - FREQUENCY_UTIL, > > - ENERGY_UTIL, > > -}; > > - > > unsigned long effective_cpu_util(int cpu, unsigned long util_cfs, > > - enum cpu_util_type type, > > - struct task_struct *p); > > + unsigned long *min, > > + unsigned long *max); > > + > > +unsigned long sugov_effective_cpu_perf(int cpu, unsigned long actual, > > + unsigned long min, > > + unsigned long max); > > + > > > > /* > > * Verify the fitness of task @p to run on @cpu taking into account the > > -- > > 2.34.1 > >
On 20/10/2023 15:58, Vincent Guittot wrote: > On Fri, 20 Oct 2023 at 11:48, Dietmar Eggemann <dietmar.eggemann@arm.com> wrote: >> >> On 13/10/2023 17:14, Vincent Guittot wrote: [...] >>> A new sugov_effective_cpu_perf() interface is also available to compute >>> the final performance level that is targeted for the CPU after applying >>> some cpufreq headroom and taking into account all inputs. >>> >>> With these 2 functions, schedutil is now able to decide when it must go >>> above uclamp hints. It now also have a generic way to get the min >>> perfromance level. >>> >>> The dependency between energy model and cpufreq governor and its headroom >>> policy doesn't exist anymore. >> >> But the dependency that both are doing the same thing still exists, right? > > For the energy model itself, it is now fully removed; only EAS still > has to estimate which perf level will be selected by schedutil but it > uses now a schedutil function without having to care about headroom > and cpufreq governor policy I see now. (1) replaces (2) so only schedutil and EAS, EM dependency is gone. compute_energy() max_util = eenv_pd_max_util() sugov_effective_cpu_perf() actual = map_util_perf(actual) (1) energy = em_cpu_energy(..., max_util, ...); max_util = map_util_perf(max_util) (2) [...] >>> unsigned long effective_cpu_util(int cpu, unsigned long util_cfs, >>> - enum cpu_util_type type, >>> - struct task_struct *p) >>> + unsigned long *min, >>> + unsigned long *max) >> >> FREQUENCY_UTIL relates to *min != NULL and *max != NULL >> >> ENERGY_UTIL relates to *min == NULL and *max == NULL >> >> so both must be either NULL or !NULL. >> >> Calling it with one equa NULL and the other with !NULL should be >> undefined, right? > > At now there is no user but one could consider only asking for min or > max. So I would not say undefined but unused OK. [...] >>> - * OTOH, for energy computation we need the estimated running time, so >>> - * include util_dl and ignore dl_bw. >>> - */ >>> - if (type == ENERGY_UTIL) >>> - util += dl_util; >>> + if (util >= scale) { >>> + if (max) >>> + *max = scale; >> >> But that means that ucamp_max cannot constrain a system in which the >> 'util > ucamp_max'. I guess that's related to you saying uclamp_min is a >> hard req and uclamp_max is a soft req. I don't think that's in sync with >> the rest of the uclamp_max implantation. > > That's a mistake, I made a shortcut here. I wanted to save the > scale_irq_capacity() step but forgot to update max 1st. > > Will fix it I see. [...] >> effective_cpu_util for FREQUENCY_UTIL (i.e. (*min != NULL && *max != >> NULL)) is slightly different. >> >> missing: >> >> if (!uclamp_is_used() && rt_rq_is_runnable(&rq->rt) >> return max >> >> probably moved into sugov_effective_cpu_perf() (which is only called >> for `FREQUENCY_UTIL`) ? > > yes, it's in sugov_effective_cpu_perf() OK. [...] >>> @@ -306,7 +329,7 @@ static inline bool sugov_cpu_is_busy(struct sugov_cpu *sg_cpu) { return false; } >>> */ >>> static inline void ignore_dl_rate_limit(struct sugov_cpu *sg_cpu) >>> { >>> - if (cpu_bw_dl(cpu_rq(sg_cpu->cpu)) > sg_cpu->bw_dl) >>> + if (cpu_bw_dl(cpu_rq(sg_cpu->cpu)) > sg_cpu->bw_min) >> >> bw_min is more than DL right? > > yes > > Interruptions are preempting DL so we should include them > And now that we can take into account uclamp_min, use it when > computing the min perf parameter of cpufreq_driver_adjust_perf() OK.
diff --git a/include/linux/energy_model.h b/include/linux/energy_model.h index b9caa01dfac4..adec808b371a 100644 --- a/include/linux/energy_model.h +++ b/include/linux/energy_model.h @@ -243,7 +243,6 @@ static inline unsigned long em_cpu_energy(struct em_perf_domain *pd, scale_cpu = arch_scale_cpu_capacity(cpu); ps = &pd->table[pd->nr_perf_states - 1]; - max_util = map_util_perf(max_util); max_util = min(max_util, allowed_cpu_cap); freq = map_util_freq(max_util, ps->frequency, scale_cpu); diff --git a/kernel/sched/core.c b/kernel/sched/core.c index a3f9cd52eec5..78228abd1219 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -7381,18 +7381,13 @@ int sched_core_idle_cpu(int cpu) * required to meet deadlines. */ unsigned long effective_cpu_util(int cpu, unsigned long util_cfs, - enum cpu_util_type type, - struct task_struct *p) + unsigned long *min, + unsigned long *max) { - unsigned long dl_util, util, irq, max; + unsigned long util, irq, scale; struct rq *rq = cpu_rq(cpu); - max = arch_scale_cpu_capacity(cpu); - - if (!uclamp_is_used() && - type == FREQUENCY_UTIL && rt_rq_is_runnable(&rq->rt)) { - return max; - } + scale = arch_scale_cpu_capacity(cpu); /* * Early check to see if IRQ/steal time saturates the CPU, can be @@ -7400,45 +7395,36 @@ unsigned long effective_cpu_util(int cpu, unsigned long util_cfs, * update_irq_load_avg(). */ irq = cpu_util_irq(rq); - if (unlikely(irq >= max)) - return max; + if (unlikely(irq >= scale)) { + if (min) + *min = scale; + if (max) + *max = scale; + return scale; + } + + /* The minimum utilization returns the highest level between: + * - the computed DL bandwidth needed with the irq pressure which + * steals time to the deadline task. + * - The minimum bandwidth requirement for CFS. + */ + if (min) + *min = max(irq + cpu_bw_dl(rq), uclamp_rq_get(rq, UCLAMP_MIN)); /* * Because the time spend on RT/DL tasks is visible as 'lost' time to * CFS tasks and we use the same metric to track the effective * utilization (PELT windows are synchronized) we can directly add them * to obtain the CPU's actual utilization. - * - * CFS and RT utilization can be boosted or capped, depending on - * utilization clamp constraints requested by currently RUNNABLE - * tasks. - * When there are no CFS RUNNABLE tasks, clamps are released and - * frequency will be gracefully reduced with the utilization decay. */ util = util_cfs + cpu_util_rt(rq); - if (type == FREQUENCY_UTIL) - util = uclamp_rq_util_with(rq, util, p); - - dl_util = cpu_util_dl(rq); - - /* - * For frequency selection we do not make cpu_util_dl() a permanent part - * of this sum because we want to use cpu_bw_dl() later on, but we need - * to check if the CFS+RT+DL sum is saturated (ie. no idle time) such - * that we select f_max when there is no idle time. - * - * NOTE: numerical errors or stop class might cause us to not quite hit - * saturation when we should -- something for later. - */ - if (util + dl_util >= max) - return max; + util += cpu_util_dl(rq); - /* - * OTOH, for energy computation we need the estimated running time, so - * include util_dl and ignore dl_bw. - */ - if (type == ENERGY_UTIL) - util += dl_util; + if (util >= scale) { + if (max) + *max = scale; + return scale; + } /* * There is still idle time; further improve the number by using the @@ -7449,28 +7435,21 @@ unsigned long effective_cpu_util(int cpu, unsigned long util_cfs, * U' = irq + --------- * U * max */ - util = scale_irq_capacity(util, irq, max); + util = scale_irq_capacity(util, irq, scale); util += irq; - /* - * Bandwidth required by DEADLINE must always be granted while, for - * FAIR and RT, we use blocked utilization of IDLE CPUs as a mechanism - * to gracefully reduce the frequency when no tasks show up for longer - * periods of time. - * - * Ideally we would like to set bw_dl as min/guaranteed freq and util + - * bw_dl as requested freq. However, cpufreq is not yet ready for such - * an interface. So, we only do the latter for now. + /* The maximum hint is a soft bandwidth requirement which can be lower + * than the actual utilization because of max uclamp requirments */ - if (type == FREQUENCY_UTIL) - util += cpu_bw_dl(rq); + if (max) + *max = min(scale, uclamp_rq_get(rq, UCLAMP_MAX)); - return min(max, util); + return min(scale, util); } unsigned long sched_cpu_util(int cpu) { - return effective_cpu_util(cpu, cpu_util_cfs(cpu), ENERGY_UTIL, NULL); + return effective_cpu_util(cpu, cpu_util_cfs(cpu), NULL, NULL); } #endif /* CONFIG_SMP */ diff --git a/kernel/sched/cpufreq_schedutil.c b/kernel/sched/cpufreq_schedutil.c index 458d359f5991..8cb323522b90 100644 --- a/kernel/sched/cpufreq_schedutil.c +++ b/kernel/sched/cpufreq_schedutil.c @@ -47,7 +47,7 @@ struct sugov_cpu { u64 last_update; unsigned long util; - unsigned long bw_dl; + unsigned long bw_min; /* The field below is for single-CPU policies only: */ #ifdef CONFIG_NO_HZ_COMMON @@ -143,7 +143,6 @@ static unsigned int get_next_freq(struct sugov_policy *sg_policy, unsigned int freq = arch_scale_freq_invariant() ? policy->cpuinfo.max_freq : policy->cur; - util = map_util_perf(util); freq = map_util_freq(util, freq, max); if (freq == sg_policy->cached_raw_freq && !sg_policy->need_freq_update) @@ -153,14 +152,38 @@ static unsigned int get_next_freq(struct sugov_policy *sg_policy, return cpufreq_driver_resolve_freq(policy, freq); } +unsigned long sugov_effective_cpu_perf(int cpu, unsigned long actual, + unsigned long min, + unsigned long max) +{ + unsigned long target; + struct rq *rq = cpu_rq(cpu); + + if (rt_rq_is_runnable(&rq->rt)) + return max; + + /* Provide at least enough capacity for DL + irq */ + target = min; + + actual = map_util_perf(actual); + /* Actually we don't need to target the max performance */ + if (actual < max) + max = actual; + + /* + * Ensure at least minimum performance while providing more compute + * capacity when possible. + */ + return max(target, max); +} + static void sugov_get_util(struct sugov_cpu *sg_cpu) { - unsigned long util = cpu_util_cfs_boost(sg_cpu->cpu); - struct rq *rq = cpu_rq(sg_cpu->cpu); + unsigned long min, max, util = cpu_util_cfs_boost(sg_cpu->cpu); - sg_cpu->bw_dl = cpu_bw_dl(rq); - sg_cpu->util = effective_cpu_util(sg_cpu->cpu, util, - FREQUENCY_UTIL, NULL); + util = effective_cpu_util(sg_cpu->cpu, util, &min, &max); + sg_cpu->bw_min = map_util_perf(min); + sg_cpu->util = sugov_effective_cpu_perf(sg_cpu->cpu, util, min, max); } /** @@ -306,7 +329,7 @@ static inline bool sugov_cpu_is_busy(struct sugov_cpu *sg_cpu) { return false; } */ static inline void ignore_dl_rate_limit(struct sugov_cpu *sg_cpu) { - if (cpu_bw_dl(cpu_rq(sg_cpu->cpu)) > sg_cpu->bw_dl) + if (cpu_bw_dl(cpu_rq(sg_cpu->cpu)) > sg_cpu->bw_min) sg_cpu->sg_policy->limits_changed = true; } @@ -407,8 +430,8 @@ static void sugov_update_single_perf(struct update_util_data *hook, u64 time, sugov_cpu_is_busy(sg_cpu) && sg_cpu->util < prev_util) sg_cpu->util = prev_util; - cpufreq_driver_adjust_perf(sg_cpu->cpu, map_util_perf(sg_cpu->bw_dl), - map_util_perf(sg_cpu->util), max_cap); + cpufreq_driver_adjust_perf(sg_cpu->cpu, sg_cpu->bw_min, + sg_cpu->util, max_cap); sg_cpu->sg_policy->last_freq_update_time = time; } diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 922905194c0c..d4f7b2f49c44 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -7628,7 +7628,7 @@ static inline void eenv_pd_busy_time(struct energy_env *eenv, for_each_cpu(cpu, pd_cpus) { unsigned long util = cpu_util(cpu, p, -1, 0); - busy_time += effective_cpu_util(cpu, util, ENERGY_UTIL, NULL); + busy_time += effective_cpu_util(cpu, util, NULL, NULL); } eenv->pd_busy_time = min(eenv->pd_cap, busy_time); @@ -7651,7 +7651,7 @@ eenv_pd_max_util(struct energy_env *eenv, struct cpumask *pd_cpus, for_each_cpu(cpu, pd_cpus) { struct task_struct *tsk = (cpu == dst_cpu) ? p : NULL; unsigned long util = cpu_util(cpu, p, dst_cpu, 1); - unsigned long eff_util; + unsigned long eff_util, min, max; /* * Performance domain frequency: utilization clamping @@ -7660,7 +7660,23 @@ eenv_pd_max_util(struct energy_env *eenv, struct cpumask *pd_cpus, * NOTE: in case RT tasks are running, by default the * FREQUENCY_UTIL's utilization can be max OPP. */ - eff_util = effective_cpu_util(cpu, util, FREQUENCY_UTIL, tsk); + eff_util = effective_cpu_util(cpu, util, &min, &max); + + /* Task's uclamp can modify min and max value */ + if (tsk && uclamp_is_used()) { + min = max(min, uclamp_eff_value(p, UCLAMP_MIN)); + + /* + * If there is no active max uclamp constraint, + * directly use task's one otherwise keep max + */ + if (uclamp_rq_is_idle(cpu_rq(cpu))) + max = uclamp_eff_value(p, UCLAMP_MAX); + else + max = max(max, uclamp_eff_value(p, UCLAMP_MAX)); + } + + eff_util = sugov_effective_cpu_perf(cpu, eff_util, min, max); max_util = max(max_util, eff_util); } diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index 65cad0e5729e..3873b4de7cfa 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -2962,24 +2962,14 @@ static inline void cpufreq_update_util(struct rq *rq, unsigned int flags) {} #endif #ifdef CONFIG_SMP -/** - * enum cpu_util_type - CPU utilization type - * @FREQUENCY_UTIL: Utilization used to select frequency - * @ENERGY_UTIL: Utilization used during energy calculation - * - * The utilization signals of all scheduling classes (CFS/RT/DL) and IRQ time - * need to be aggregated differently depending on the usage made of them. This - * enum is used within effective_cpu_util() to differentiate the types of - * utilization expected by the callers, and adjust the aggregation accordingly. - */ -enum cpu_util_type { - FREQUENCY_UTIL, - ENERGY_UTIL, -}; - unsigned long effective_cpu_util(int cpu, unsigned long util_cfs, - enum cpu_util_type type, - struct task_struct *p); + unsigned long *min, + unsigned long *max); + +unsigned long sugov_effective_cpu_perf(int cpu, unsigned long actual, + unsigned long min, + unsigned long max); + /* * Verify the fitness of task @p to run on @cpu taking into account the
The current method to take into account uclamp hints when estimating the target frequency can end into situation where the selected target frequency is finally higher than uclamp hints whereas there are no real needs. Such cases mainly happen because we are currently mixing the traditional scheduler utilization signal with the uclamp performance hints. By adding these 2 metrics, we loose an important information when it comes to select the target frequency and we have to make some assumptions which can't fit all cases. Rework the interface between the scheduler and schedutil governor in order to propagate all information down to the cpufreq governor. effective_cpu_util() interface changes and now returns the actual utilization of the CPU with 2 optional inputs: - The minimum performance for this CPU; typically the capacity to handle the deadline task and the interrupt pressure. But also uclamp_min request when available. - The maximum targeting performance for this CPU which reflects the maximum level that we would like to not exceed. By default it will be the CPU capacity but can be reduced because of some performance hints set with uclamp. The value can be lower than actual utilization and/or min performance level. A new sugov_effective_cpu_perf() interface is also available to compute the final performance level that is targeted for the CPU after applying some cpufreq headroom and taking into account all inputs. With these 2 functions, schedutil is now able to decide when it must go above uclamp hints. It now also have a generic way to get the min perfromance level. The dependency between energy model and cpufreq governor and its headroom policy doesn't exist anymore. eenv_pd_max_util asks schedutil for the targeted performance after applying the impact of the waking task. Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org> --- include/linux/energy_model.h | 1 - kernel/sched/core.c | 85 ++++++++++++-------------------- kernel/sched/cpufreq_schedutil.c | 43 ++++++++++++---- kernel/sched/fair.c | 22 +++++++-- kernel/sched/sched.h | 24 +++------ 5 files changed, 91 insertions(+), 84 deletions(-)