Re: [PATCH v9 26/33] sched/deadline: Start a period for a task that enters the class blocked
From: Mathieu Desnoyers
Date: Fri Nov 20 2026 - 14:40:50 EST
> Can gating the new period on dl_task_pi_blocked() in switched_to_dl() and
> dl_boost_blocked() still leave a sleeping task with an uninitialized or
> stale p->dl.deadline?
Your first case: yes. v10 will fix it in a new patch, "sched/deadline:
Start a period for a task requeued onto an rt_mutex".
task_blocks_on_rt_mutex() will start a new period for a proxy-locked
task whose deadline is 0 or has passed. The waiter then takes the new
deadline. A new case, requeue_pi_new_period, will test the deadline of
0. It fails without the fix.
Your second case: yes. v10 will compare the waiter with the deadline
that the owner would run on. "sched/deadline: Boost an owner that cannot
run on its own reservation" will do that. A deadline that has passed, 0
included, then counts as now + dl_deadline.
Your third case is how a boost works. A new donor does not give the
owner its deadline. It changes the parameters of the owner's next
refills. A queued owner keeps its deadline the same way.
> Does checking !dl_se->dl_deadline in dl_boost_blocked() miss a task that
> left SCHED_DEADLINE before its 0-lag time?
Yes, by reading. v7.4-rc2 sorts a deadline waiter blocked past its
deadline the same way. A new period at the boost would fix only that
moment: the deadline of any waiter can pass while it stays blocked.
v10 will leave this out, and list it as a known limit in its cover
letter.
Thanks,
Mathieu
--
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com