[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20250902111740.hwMmUu4T@linutronix.de>
Date: Tue, 2 Sep 2025 13:17:40 +0200
From: Sebastian Andrzej Siewior <bigeasy@...utronix.de>
To: Lai Jiangshan <jiangshanlai@...il.com>
Cc: linux-rt-devel@...ts.linux.dev, linux-kernel@...r.kernel.org,
Clark Williams <clrkwllms@...nel.org>,
Ingo Molnar <mingo@...hat.com>,
Peter Zijlstra <peterz@...radead.org>,
Steven Rostedt <rostedt@...dmis.org>, Tejun Heo <tj@...nel.org>,
Thomas Gleixner <tglx@...utronix.de>
Subject: Re: [PATCH v2 1/3] workqueue: Provide a handshake for canceling BH
workers
On 2025-09-02 18:12:10 [+0800], Lai Jiangshan wrote:
> Hello
Hi Lai,
> > --- a/kernel/workqueue.c
> > +++ b/kernel/workqueue.c
> > @@ -222,7 +222,9 @@ struct worker_pool {
> > struct workqueue_attrs *attrs; /* I: worker attributes */
> > struct hlist_node hash_node; /* PL: unbound_pool_hash node */
> > int refcnt; /* PL: refcnt for unbound pools */
> > -
> > +#ifdef CONFIG_PREEMPT_RT
> > + spinlock_t cb_lock; /* BH worker cancel lock */
> > +#endif
> > /*
>
> Is it possible to use rt_mutex_init_proxy_locked(), rt_mutex_proxy_unlock()
> and rt_mutex_wait_proxy_lock()?
>
> Or is it possible to add something like rt_spinlock_init_proxy_locked(),
> rt_spinlock_proxy_unlock() and rt_spinlock_wait_proxy_lock() which work
> the same as the rt_mutex's proxy lock primitives but for non-sleep context?
I don't think so. I think non-sleep context is the killer part. Those
are for PI and this works by assigning waiter's priority, going to sleep
until "it" is done. Now if you want non-sleep then you would have to
remain on the CPU and spin until the "work" is done. This spinning would
work if the other task is on a remote CPU. But if both are on the same
CPU then spinning is not working.
> I think they will work as an rt variant of struct completion and
> they can be used for __flush_work() for BH work for preempt_rt
> as the same way as wait_for_completion() is used for normal work.
I completely dislike the idea of spinning until completion for the BH
work. Yes, we have three tasklet users which are doing this as of today.
We have no users which require this for workqueue and you can schedule a
workqueue from interrupt context.
Why do we even what this? My idea was to submit this and then hide later
behind an "atomic" API which will hopefully remain unused. As explained
it the previous thread: From RT's point of view it requires to acquire
and drop the cb_lock each time and boost all items on that worker until
the "barrier" work item arrives. While in general the "one" item is
boosted if needed.
> Thanks
> Lai
Sebastian
Powered by blists - more mailing lists