cpufreq/schedutil: Only bind threads if needed
Subsystems (2)
PATCH 1/1 cpufreq/schedutil: Only bind threads if needed
Remove the unconditional binding of sugov kthreads to the affected CPUs if the cpufreq driver indicates that updates can happen from any CPU. This allows userspace to set affinities to either save power (waking up bigger CPUs on HMP can be expensive) or increasing performance (by letting the utilized CPUs run without preemption of the sugov kthread).
Without this patch the behavior of sugov threads will basically be a boot-time dice roll on which CPU of the PD has to handle all the cpufreq updates. With the recent decreases of update filtering these two basic problems become more and more apparent:
1. The wake_cpu might be idle and we are waking it up from another CPU just for the cpufreq update. Apart from wasting power, the exit latency of it's idle state might be longer than the sugov threads running time, essentially delaying the cpufreq update unnecessarily.
2. We are preempting either the requesting or another busy CPU of the PD, while the update could be done from a CPU that we deem less important and pay the price of an IPI and two context-switches.
The change is essentially not setting PF_NO_SETAFFINITY on dvfs_possible_from_any_cpu, no behavior change if userspace doesn't touch affinities.
--- kernel/sched/cpufreq_schedutil.c | 6 +++++- kernel/sched/syscalls.c | 3 +++ 2 files changed, 8 insertions(+), 1 deletion(-)
kernel/sched/cpufreq_schedutil.c +5 -1
@@ -683,7 +683,11 @@ static int sugov_kthread_create(struct sugov_policy *sg_policy) } sg_policy->thread = thread; -kthread_bind_mask (thread, policy->related_cpus); +if (policy->dvfs_possible_from_any_cpu) +set_cpus_allowed_ptr (thread, policy->related_cpus); +else +kthread_bind_mask (thread, policy->related_cpus); +init_irq_work (&sg_policy->irq_work, sugov_irq_work);mutex_init (&sg_policy->work_lock);
kernel/sched/syscalls.c +3 -0
@@ -1159,6 +1159,9 @@ int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)if (!task_has_dl_policy (p) || !dl_bandwidth_enabled ())return 0 ; +if (dl_entity_is_special (&p->dl)) +return 0 ; +/* * Since bandwidth control happens on root_domain basis, * if admission test is enabled, we only admit -deadline
Signature · 1 line
2.34.1
Rafael J. Wysocki
maintainerQuoted 19 lines · Christian Loehle
Remove the unconditional binding of sugov kthreads to the affected CPUs if the cpufreq driver indicates that updates can happen from any CPU. This allows userspace to set affinities to either save power (waking up bigger CPUs on HMP can be expensive) or increasing performance (by letting the utilized CPUs run without preemption of the sugov kthread).
Without this patch the behavior of sugov threads will basically be a boot-time dice roll on which CPU of the PD has to handle all the cpufreq updates. With the recent decreases of update filtering these two basic problems become more and more apparent:
1. Thewake_cpumight be idle and we are waking it up from another CPU just for the cpufreq update. Apart from wasting power, the exit latency of it's idle state might be longer than the sugov threads running time, essentially delaying the cpufreq update unnecessarily.
2. We are preempting either the requesting or another busy CPU of the PD, while the update could be done from a CPU that we deem less important and pay the price of an IPI and two context-switches.
Quoted 3 lines
The change is essentially not setting
PF_NO_SETAFFINITYondvfs_possible_from_any_cpu, no behavior change if userspace doesn't touch affinities.
I'd like to hear from Viresh on this.
Quoted 28 lines
Signed-off-by Christian Loehle <[email protected]>--- kernel/sched/cpufreq_schedutil.c | 6 +++++- kernel/sched/syscalls.c | 3 +++ 2 files changed, 8 insertions(+), 1 deletion(-)kernel/sched/cpufreq_schedutil.c +5
-1@@ -683,7 +683,11 @@ static int sugov_kthread_create(struct sugov_policy *sg_policy) } sg_policy->thread = thread; -kthread_bind_mask (thread, policy->related_cpus); +if (policy->dvfs_possible_from_any_cpu) +set_cpus_allowed_ptr (thread, policy->related_cpus); +else +kthread_bind_mask (thread, policy->related_cpus); +init_irq_work (&sg_policy->irq_work, sugov_irq_work);mutex_init (&sg_policy->work_lock);kernel/sched/syscalls.c +0
-0
kernel/sched/syscalls.c int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask) · line 1164
Quoted 6 lines
kernel/sched/syscalls.c +3
-0@@ -1159,6 +1159,9 @@ int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)if (!task_has_dl_policy (p) || !dl_bandwidth_enabled ())return 0 ; +if (dl_entity_is_special (&p->dl)) +return 0 ; +
Care to explain this particular piece?
Quoted 4 lines
kernel/sched/syscalls.c +0
-1/* * Since bandwidth control happens on root_domain basis, * if admission test is enabled, we only admit -deadline --
Quoted 58 lines · Rafael J. Wysocki
Quoted 19 lines · Christian Loehle
Remove the unconditional binding of sugov kthreads to the affected CPUs if the cpufreq driver indicates that updates can happen from any CPU. This allows userspace to set affinities to either save power (waking up bigger CPUs on HMP can be expensive) or increasing performance (by letting the utilized CPUs run without preemption of the sugov kthread).
Without this patch the behavior of sugov threads will basically be a boot-time dice roll on which CPU of the PD has to handle all the cpufreq updates. With the recent decreases of update filtering these two basic problems become more and more apparent:
1. Thewake_cpumight be idle and we are waking it up from another CPU just for the cpufreq update. Apart from wasting power, the exit latency of it's idle state might be longer than the sugov threads running time, essentially delaying the cpufreq update unnecessarily.
2. We are preempting either the requesting or another busy CPU of the PD, while the update could be done from a CPU that we deem less important and pay the price of an IPI and two context-switches.Quoted 3 lines
The change is essentially not setting
PF_NO_SETAFFINITYondvfs_possible_from_any_cpu, no behavior change if userspace doesn't touch affinities.I'd like to hear from Viresh on this.
Quoted 31 lines
Signed-off-by Christian Loehle <[email protected]>--- kernel/sched/cpufreq_schedutil.c | 6 +++++- kernel/sched/syscalls.c | 3 +++ 2 files changed, 8 insertions(+), 1 deletion(-)kernel/sched/cpufreq_schedutil.c +5
-1@@ -683,7 +683,11 @@ static int sugov_kthread_create(struct sugov_policy *sg_policy) } sg_policy->thread = thread; -kthread_bind_mask (thread, policy->related_cpus); +if (policy->dvfs_possible_from_any_cpu) +set_cpus_allowed_ptr (thread, policy->related_cpus); +else +kthread_bind_mask (thread, policy->related_cpus); +init_irq_work (&sg_policy->irq_work, sugov_irq_work);mutex_init (&sg_policy->work_lock);kernel/sched/syscalls.c +0
-0@@ -1159,6 +1159,9 @@ int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)if (!task_has_dl_policy (p) || !dl_bandwidth_enabled ())return 0 ;
Quoted 5 lines
kernel/sched/syscalls.c int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask) · line 1164
Quoted 3 lines
kernel/sched/syscalls.c +3
-0+if (dl_entity_is_special (&p->dl)) +return 0 ; +Care to explain this particular piece?
Looks suspicious but the truncated comment below explains it:
/*
* Since bandwidth control happens on root_domain basis,
* if admission test is enabled, we only admit -deadline
* tasks allowed to run on all the CPUs in the task's
* root_domain.
*/
So that would only allow setting it to all CPUs for the relevant platforms unfortunately.
That should be fine though since the sugov task is pretty much a dummy in terms of bandwidth / admission control internally, so no harm done to not enforce this when userspace wants to set affinities.
...Unless Juri disagrees.
Quoted 5 lines
Quoted 5 lines
kernel/sched/syscalls.c +0
-1/* * Since bandwidth control happens on root_domain basis, * if admission test is enabled, we only admit -deadline --
Juri Lelli
maintainerHi,
Quoted 2 lines · Christian Loehle
Quoted 1 line · Rafael J. Wysocki
On Thu, Sep 12, 2024 at 3:53 PM Christian Loehle
...
Quoted 24 lines
Quoted 8 lines
Quoted 8 lines
kernel/sched/syscalls.c +0
-0@@ -1159,6 +1159,9 @@ int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)if (!task_has_dl_policy (p) || !dl_bandwidth_enabled ())return 0 ;Quoted 5 lines
kernel/sched/syscalls.c int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask) · line 1164
Quoted 3 lines
kernel/sched/syscalls.c +3
-0+if (dl_entity_is_special (&p->dl)) +return 0 ; +Care to explain this particular piece?
Looks suspicious but the truncated comment below explains it:
/*
* Since bandwidth control happens onroot_domainbasis,
* if admission test is enabled, we only admit -deadline
* tasks allowed to run on all the CPUs in the task's
*root_domain.
*/
So that would only allow setting it to all CPUs for the relevant platforms unfortunately.
Quoted 5 lines
That should be fine though since the sugov task is pretty much a dummy in terms of bandwidth / admission control internally, so no harm done to not enforce this when userspace wants to set affinities.
...Unless Juri disagrees.
Nope, I agree. :)
Wonder if we should put a comment along the lines of what you said above right above the new conditions (so that people will not need to wonder about it in the future). But not a strict requirement for me.
Thanks! (and apologies for the delay in replying)
Juri
Quoted 38 lines · Juri Lelli
Hi,
Quoted 2 lines · Christian Loehle
Quoted 1 line · Rafael J. Wysocki
On Thu, Sep 12, 2024 at 3:53 PM Christian Loehle
...
Quoted 24 lines
Quoted 8 lines
Quoted 8 lines
kernel/sched/syscalls.c +0
-0@@ -1159,6 +1159,9 @@ int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)if (!task_has_dl_policy (p) || !dl_bandwidth_enabled ())return 0 ;Quoted 5 lines
kernel/sched/syscalls.c int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask) · line 1164
Quoted 3 lines
kernel/sched/syscalls.c +3
-0+if (dl_entity_is_special (&p->dl)) +return 0 ; +Care to explain this particular piece?
Looks suspicious but the truncated comment below explains it:
/*
* Since bandwidth control happens onroot_domainbasis,
* if admission test is enabled, we only admit -deadline
* tasks allowed to run on all the CPUs in the task's
*root_domain.
*/
So that would only allow setting it to all CPUs for the relevant platforms unfortunately.Quoted 5 lines
That should be fine though since the sugov task is pretty much a dummy in terms of bandwidth / admission control internally, so no harm done to not enforce this when userspace wants to set affinities.
...Unless Juri disagrees.
Quoted 8 lines
Nope, I agree. :)
Wonder if we should put a comment along the lines of what you said above right above the new conditions (so that people will not need to wonder about it in the future). But not a strict requirement for me.
Thanks! (and apologies for the delay in replying)
Juri
Thank you Juri for taking a look. I agree with the comment, will do a v2.
Regards,
Christian
Quoted 21 lines · Rafael J. Wysocki
Quoted 19 lines · Christian Loehle
Remove the unconditional binding of sugov kthreads to the affected CPUs if the cpufreq driver indicates that updates can happen from any CPU. This allows userspace to set affinities to either save power (waking up bigger CPUs on HMP can be expensive) or increasing performance (by letting the utilized CPUs run without preemption of the sugov kthread).
Without this patch the behavior of sugov threads will basically be a boot-time dice roll on which CPU of the PD has to handle all the cpufreq updates. With the recent decreases of update filtering these two basic problems become more and more apparent:
1. Thewake_cpumight be idle and we are waking it up from another CPU just for the cpufreq update. Apart from wasting power, the exit latency of it's idle state might be longer than the sugov threads running time, essentially delaying the cpufreq update unnecessarily.
2. We are preempting either the requesting or another busy CPU of the PD, while the update could be done from a CPU that we deem less important and pay the price of an IPI and two context-switches.
Quoted 5 lines
Quoted 3 lines
The change is essentially not setting
PF_NO_SETAFFINITYondvfs_possible_from_any_cpu, no behavior change if userspace doesn't touch affinities.I'd like to hear from Viresh on this.
Looks good to me.
Signature · 1 line
viresh
Quoted 18 lines · Christian Loehle
Remove the unconditional binding of sugov kthreads to the affected CPUs if the cpufreq driver indicates that updates can happen from any CPU. This allows userspace to set affinities to either save power (waking up bigger CPUs on HMP can be expensive) or increasing performance (by letting the utilized CPUs run without preemption of the sugov kthread).
Without this patch the behavior of sugov threads will basically be a boot-time dice roll on which CPU of the PD has to handle all the cpufreq updates. With the recent decreases of update filtering these two basic problems become more and more apparent:
1. Thewake_cpumight be idle and we are waking it up from another CPU just for the cpufreq update. Apart from wasting power, the exit latency of it's idle state might be longer than the sugov threads running time, essentially delaying the cpufreq update unnecessarily.
2. We are preempting either the requesting or another busy CPU of the PD, while the update could be done from a CPU that we deem less important and pay the price of an IPI and two context-switches.
Quoted 5 lines
The change is essentially not setting
PF_NO_SETAFFINITYondvfs_possible_from_any_cpu, no behavior change if userspace doesn't touch affinities.Signed-off-by Christian Loehle <[email protected]>
I'll add some numbers to illustrate, although the example might not be particularly realistic.
The classic fio workload will trigger cpufreq update very often, so I used that on the Pixel6, affinity set to CPU7 (bitmask 80)(Big PD is [6,7]).
Without this patch we have either all sugov enqueues on CPU6 or CPU7, depending on where the first CPU frequency request (since boot) was issued from (the deadline select_task_rq is rather simple, so it will just wake_cpu if that is still valid, which here it always is).
I set different affinities for the sugov:6 worker and annotate IOPS (throughput) and power (mW average), the test is for 30s each. cpumask IOPS
80 7477
888.3 mW {'CPU-Big': 742.5668631414397, 'CPU-Mid': 11.919500003712095, 'CPU-Little': 133.79554163073317}
40 7378
942.8 mW {'CPU-Big': 797.4037245094225, 'CPU-Mid': 12.440698878099667, 'CPU-Little': 132.98899390286172}
f 7469
873.7 mW {'CPU-Big': 718.2933574826287, 'CPU-Mid': 11.89176754939742, 'CPU-Little': 143.55634276873246}
2 7501
872.8 mW {'CPU-Big': 718.6036614909397, 'CPU-Mid': 11.711731623773632, 'CPU-Little': 142.4830173663132}
1 7392
859.5 mW {'CPU-Big': 704.926078017567, 'CPU-Mid': 12.196892652625284, 'CPU-Little': 142.4231989686622}Throughput is somewhat comparable for all, anyway the frequency only bounces from capacity 1024 to 512 because of the instability of iowait boost.
For 40 (CPU6) we see significantly more power usage, as the sugov will prevent CPU6 from power down, as it gets woken up by the sugov worker often.
f,2 and 1 affinities have slightly higher power coming from the littles (as expected), but significantly less power from the bigs, since they don't have the double context-switch on preemption and the actual sugov CPU cycles.
1 has by far the least power, as it's always in WFI anyway, it handles the IO interrupts from fio on CPU7.
Granted that is a somewhat worst-case scenario just to illustrate the problem.
I also have a patch for preferring wake_cpu = smp_processor_id() for the sugov worker, which is somewhat adjacent to this, the numbers above make a case for that even without touching affinities. I thought this might be less controversial for now.