This repository has been archived by the owner on Jun 18, 2024. It is now read-only.
scx: Implement SCX_KICK_IDLE #138
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Byte-Lab
approved these changes
Feb 7, 2024
kernel/sched/ext.c
Outdated
| struct rq *rq = cpu_rq(cpu); | ||
| unsigned long flags; | ||
| /* | ||
| * We can skip idle kicking if @cpu is going to go through at least one |
| if (cpu_online(cpu) || cpu == this_cpu) { | ||
| if (cpumask_test_cpu(cpu, this_rq->scx.cpus_to_preempt) && | ||
| rq->curr->sched_class == &ext_sched_class) | ||
| if (cpu_online(cpu) || cpu == cpu_of(this_rq)) { |
There was a problem hiding this comment.
So even though this functionally isn't a change from what was there before, would you mind please adding a comment explaining that a CPU can be offlined before IRQs are disabled to explain how it's possible that the CPU is offline but cpu == cpu_of(this_rq)?
There was a problem hiding this comment.
It's just that CPU on/offlining is a multi-stage operation and cpu hotplug ops themselves depend on scheduling working. I'm not sure how much detail we want to explain here. I'll mention it briefly.
- There's no reason to clear all scx_rq->cpus_* masks each time kick_cpus_irq_workfn() is run especially given that other cpumasks won't be set if ->cpus_to_kick isn't. Selectively clear bits which are set instead. - Don't access rq->scx.pnt_seq if not needed. - A bit of restructuring for redability and to avoid scanning ->cpus_to_wait unnecessarily. Signed-off-by: Tejun Heo <[email protected]>
No functional changes intended. Signed-off-by: Tejun Heo <[email protected]>
Wanting to wake a CPU from idle is not an uncommon operaiton that a BPF scheduler might want to do. The only way to do it is scx_bpf_kick_cpu(); however, this can lead to unnecessary churns when the target CPU is not idle and there isn't a good way to synchronize against idle transitions to avoid spurious kicks. This patch implements SCX_KICK_IDLE which kicks the target CPU if it's idle. The flag doesn't guarantee that a CPU will never be kicked unless idle but it optimizes away spurious kicks in most cases. Signed-off-by: Tejun Heo <[email protected]>
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Implement SCX_KICK_IDLE so that schedulers don't have to worry about causing too much churn with attempts to wake idle CPUs.