[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <aOTjSla1Yr3kz7op@jlelli-thinkpadt14gen4.remote.csb>
Date: Tue, 7 Oct 2025 11:54:18 +0200
From: Juri Lelli <juri.lelli@...hat.com>
To: Peter Zijlstra <peterz@...radead.org>
Cc: tj@...nel.org, linux-kernel@...r.kernel.org, mingo@...nel.org,
vincent.guittot@...aro.org, dietmar.eggemann@....com,
rostedt@...dmis.org, bsegall@...gle.com, mgorman@...e.de,
vschneid@...hat.com, longman@...hat.com, hannes@...xchg.org,
mkoutny@...e.com, void@...ifault.com, arighi@...dia.com,
changwoo@...lia.com, cgroups@...r.kernel.org,
sched-ext@...ts.linux.dev, liuwenfang@...or.com, tglx@...utronix.de
Subject: Re: [PATCH 10/12] sched: Add locking comments to sched_class methods
Hi Peter,
On 06/10/25 12:44, Peter Zijlstra wrote:
> 'Document' the locking context the various sched_class methods are
> called under.
>
> Signed-off-by: Peter Zijlstra (Intel) <peterz@...radead.org>
> ---
...
> --- a/kernel/sched/sched.h
> +++ b/kernel/sched/sched.h
> @@ -2345,8 +2345,7 @@ extern const u32 sched_prio_to_wmult[40
> /*
> * {de,en}queue flags:
> *
> - * DEQUEUE_SLEEP - task is no longer runnable
> - * ENQUEUE_WAKEUP - task just became runnable
> + * SLEEP/WAKEUP - task is no-longer/just-became runnable
> *
> * SAVE/RESTORE - an otherwise spurious dequeue/enqueue, done to ensure tasks
> * are in a known state which allows modification. Such pairs
> @@ -2359,6 +2358,11 @@ extern const u32 sched_prio_to_wmult[40
> *
> * MIGRATION - p->on_rq == TASK_ON_RQ_MIGRATING (used for DEADLINE)
> *
> + * DELAYED - de/re-queue a sched_delayed task
> + *
> + * CLASS - going to update p->sched_class; makes sched_change call the
> + * various switch methods.
> + *
> * ENQUEUE_HEAD - place at front of runqueue (tail if not specified)
> * ENQUEUE_REPLENISH - CBS (replenish runtime and postpone deadline)
> * ENQUEUE_MIGRATED - the task was migrated during wakeup
Not for this patch, but I wondered if, while we are at it, we wanted to
complete documentation of these flags. My new AI friend is suggesting
the following, is it very much garbage? :)
Thanks,
Juri
---
From: Claude <claude-sonnet-4-5@...hropic.com>
Date: Mon, 7 Oct 2025 12:44:13 +0200
Subject: sched: Document remaining DEQUEUE/ENQUEUE flags
Complete the flag documentation by adding descriptions for the three
previously undocumented flags: DEQUEUE_SPECIAL, DEQUEUE_THROTTLE, and
ENQUEUE_INITIAL.
DEQUEUE_SPECIAL is used when dequeuing tasks in special states (stopped,
traced, parked, dead, or frozen) that don't use the normal wait-loop
pattern and must not use delayed dequeue.
DEQUEUE_THROTTLE is used when removing tasks from the runqueue due to
CFS bandwidth throttling, preventing delayed dequeue to ensure proper
throttling behavior.
ENQUEUE_INITIAL is used when enqueueing newly created tasks in
wake_up_new_task(), allowing the fair scheduler to give them preferential
initial placement (half vslice when PLACE_DEADLINE_INITIAL is enabled).
Signed-off-by: Claude <claude-sonnet-4-5@...hropic.com>
Not-so-sure-yet: Juri Lelli <juri.lelli@...hat.com>
---
kernel/sched/sched.h | 10 ++++++++++
1 file changed, 10 insertions(+)
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 4c222fa8f908..1a2b3c8d9e4f 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -2364,10 +2364,20 @@ extern const u32 sched_prio_to_wmult[40];
* CLASS - going to update p->sched_class; makes sched_change call the
* various switch methods.
*
+ * DEQUEUE_SPECIAL - task is in a special state (STOPPED, TRACED, PARKED,
+ * DEAD, FROZEN) that doesn't use the normal wait-loop;
+ * disables delayed dequeue.
+ *
+ * DEQUEUE_THROTTLE - dequeuing due to CFS bandwidth throttling; disables
+ * delayed dequeue to ensure proper throttling.
+ *
* ENQUEUE_HEAD - place at front of runqueue (tail if not specified)
* ENQUEUE_REPLENISH - CBS (replenish runtime and postpone deadline)
* ENQUEUE_MIGRATED - the task was migrated during wakeup
+ * ENQUEUE_INITIAL - enqueuing a newly created task in wake_up_new_task();
+ * fair scheduler may give preferential initial placement
+ * (e.g., half vslice with PLACE_DEADLINE_INITIAL).
* ENQUEUE_RQ_SELECTED - ->select_task_rq() was called
*
*/
Powered by blists - more mailing lists