Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Thu, 3 Nov 2022 09:34:08 -0700
From:   "Paul E. McKenney" <paulmck@kernel.org>
To:     Uladzislau Rezki <urezki@gmail.com>
Cc:     Joel Fernandes <joel@joelfernandes.org>, rcu@vger.kernel.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH RFC] rcu/kfree: Do not request RCU when not needed
Message-ID: <20221103163408.GU5600@paulmck-ThinkPad-P17-Gen-1>
Reply-To: paulmck@kernel.org
References: <20221102184911.GP5600@paulmck-ThinkPad-P17-Gen-1>
 <755B5ED1-653D-4E57-B114-77CDE10A9033@joelfernandes.org>
 <20221102202813.GR5600@paulmck-ThinkPad-P17-Gen-1>
 <CAEXW_YQ+SxBoNUkPHhC3O0DJNQtZomN_4GPtvaWuDs5sSU4FAw@mail.gmail.com>
 <20221102223516.GT5600@paulmck-ThinkPad-P17-Gen-1>
 <Y2O3w3d3qmTg6VAP@pc638.lan>
 <Y2O8k2U+ACr1N6Fe@pc638.lan>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <Y2O8k2U+ACr1N6Fe@pc638.lan>
Precedence: bulk

On Thu, Nov 03, 2022 at 02:05:23PM +0100, Uladzislau Rezki wrote:
> On Thu, Nov 03, 2022 at 01:44:51PM +0100, Uladzislau Rezki wrote:
> > > 
> > > > Though I am thinking, workqueue context is normally used to invoke
> > > > code that can block, and would the issue you mentioned affect those as
> > > > well, or affect RCU when those non-RCU work items block. So for
> > > > example, when other things in the system that can queue things on the
> > > > system_wq and block.  (I might be throwing darts in the dark).
> > > > 
> > > > To be safe, we can implement your suggestion which is basically a form
> > > > of my initial patch.
> > > > 
> > > > Should we add Tejun to the thread?
> > > 
> > > Let's get organized first, but that would be a good thing.  Or I could
> > > reach out to Tejun internally.
> > > 
> > > For but one thing to get organized about, maybe kfree_rcu() should be
> > > using a workqueue with the WQ_MEM_RECLAIM flag set.
> > > 
> > It can be as an option to consider. Because such workqueue has some
> > special priority for better handling of memory releasing. I can have
> > a look at it closer to see how kvfree_rcu() works if it goes with WQ_MEM_RECLAIM.
> > 
> An extra note. It would work well with posted patch because we can
> directly queue the reclaim work to the WQ_MEM_RECLAIM queue.
> 
> As for now RCU-core kthreads like: rcugp, rcuop use "regular"
> queue. I think system_wq one.

Agreed, it is the expedited grace period that uses WQ_MEM_RECLAIM
workqueues.

							Thanx, Paul