From: "Paul E. McKenney" <paulmck@kernel.org>
To: Jakub Kicinski <kuba@kernel.org>
Cc: Eric Dumazet <edumazet@google.com>, Yan Zhai <yan@cloudflare.com>,
netdev@vger.kernel.org, "David S. Miller" <davem@davemloft.net>,
Paolo Abeni <pabeni@redhat.com>, Jiri Pirko <jiri@resnulli.us>,
Simon Horman <horms@kernel.org>,
Daniel Borkmann <daniel@iogearbox.net>,
Lorenzo Bianconi <lorenzo@kernel.org>,
Coco Li <lixiaoyan@google.com>, Wei Wang <weiwan@google.com>,
Alexander Duyck <alexanderduyck@fb.com>,
Hannes Frederic Sowa <hannes@stressinduktion.org>,
linux-kernel@vger.kernel.org, rcu@vger.kernel.org,
bpf@vger.kernel.org, kernel-team@cloudflare.com
Subject: Re: [PATCH] net: raise RCU qs after each threaded NAPI poll
Date: Wed, 28 Feb 2024 07:15:42 -0800 [thread overview]
Message-ID: <9a0052f9-b022-42c9-a5da-1d6ca3b00885@paulmck-laptop> (raw)
In-Reply-To: <20240228064343.578a5363@kernel.org>
On Wed, Feb 28, 2024 at 06:43:43AM -0800, Jakub Kicinski wrote:
> On Tue, 27 Feb 2024 20:42:24 -0800 Paul E. McKenney wrote:
> > On Tue, Feb 27, 2024 at 07:10:01PM -0800, Jakub Kicinski wrote:
> > > On Tue, 27 Feb 2024 10:32:22 -0800 Paul E. McKenney wrote:
> > > > The theory is that PREEMPT_RCU kernels have preemption, and get their
> > > > quiescent states that way.
> > >
> > > But that doesn't work well enough?
> > >
> > > Assuming that's the case why don't we add it with the inverse ifdef
> > > condition next to the cond_resched() which follows a few lines down?
> > >
> > > skb_defer_free_flush(sd);
> > > +
> > > + if (!IS_ENABLED(CONFIG_PREEMPT_RT))
> > > + rcu_softirq_qs();
> > > +
> > > local_bh_enable();
> > >
> > > if (!repoll)
> > > break;
> > >
> > > cond_resched();
> > > }
> > >
> > > We won't repoll majority of the time.
> >
> > I am not completely clear on what you are proposing, but one complication
> > is that We need preemption disabled across calls to rcu_softirq_qs()
> > and we cannot have preemption disabled across calls to cond_resched().
>
> I was thinking of using rcu_all_qs(), like cond_resched() does.
> Not sure how it compares in terms of functionality and cost.
It is probably a bit cheaper, but it does nothing for Tasks RCU. And that
"_all" in the name is a holdover from when there were separate mechanisms
for bh, sched, and preempt, so maybe we should change that name.
> > Another complication is that although CONFIG_PREEMPT_RT kernels are
> > built with CONFIG_PREEMPT_RCU, the reverse is not always the case.
> > And if we are not repolling, don't we have a high probability of doing
> > a voluntary context when we reach napi_thread_wait() at the beginning
> > of that loop?
>
> Very much so, which is why adding the cost of rcu_softirq_qs()
> for every NAPI run feels like an overkill.
Would it be better to do the rcu_softirq_qs() only once every 1000 times
or some such? Or once every HZ jiffies?
Or is there a better way?
Thanx, Paul
next prev parent reply other threads:[~2024-02-28 15:15 UTC|newest]
Thread overview: 47+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-02-27 15:44 [PATCH] net: raise RCU qs after each threaded NAPI poll Yan Zhai
2024-02-27 16:44 ` Eric Dumazet
2024-02-27 18:32 ` Paul E. McKenney
2024-02-27 21:22 ` Yan Zhai
2024-02-27 22:58 ` Paul E. McKenney
2024-02-28 3:10 ` Jakub Kicinski
2024-02-28 4:42 ` Paul E. McKenney
2024-02-28 14:43 ` Jakub Kicinski
2024-02-28 15:15 ` Paul E. McKenney [this message]
2024-02-28 15:35 ` Jakub Kicinski
2024-02-28 15:57 ` Yan Zhai
2024-02-28 11:50 ` Toke Høiland-Jørgensen
2024-02-28 15:10 ` Paul E. McKenney
2024-02-28 15:48 ` Yan Zhai
2024-02-28 17:47 ` Paul E. McKenney
2024-02-28 15:37 ` Joel Fernandes
2024-02-28 16:37 ` Yan Zhai
2024-02-28 17:18 ` Paul E. McKenney
2024-02-28 20:14 ` Joel Fernandes
2024-02-28 21:13 ` Paul E. McKenney
2024-02-28 21:27 ` Joel Fernandes
2024-02-28 21:52 ` Paul E. McKenney
2024-02-28 22:10 ` Joel Fernandes
2024-02-28 22:19 ` Paul E. McKenney
2024-02-28 22:33 ` Steven Rostedt
2024-02-28 22:48 ` Alexei Starovoitov
2024-02-28 22:58 ` Paul E. McKenney
2024-02-29 14:21 ` Joel Fernandes
2024-02-29 16:57 ` Paul E. McKenney
2024-02-29 17:41 ` Joel Fernandes
2024-02-29 18:29 ` Paul E. McKenney
2024-03-02 2:24 ` Joel Fernandes
2024-03-03 0:25 ` Paul E. McKenney
2024-03-03 1:01 ` Joel Fernandes
2024-03-04 9:16 ` Joel Fernandes
2024-03-05 17:53 ` Paul E. McKenney
2024-03-05 19:57 ` Mark Rutland
2024-03-05 21:52 ` Paul E. McKenney
2024-03-06 16:56 ` Steven Rostedt
2024-03-07 16:57 ` Mark Rutland
2024-03-07 18:34 ` Mark Rutland
2024-03-07 18:52 ` Steven Rostedt
2024-03-07 18:58 ` Paul E. McKenney
2024-03-04 9:16 ` Joel Fernandes
2024-02-28 22:49 ` Paul E. McKenney
2024-02-27 21:17 ` Yan Zhai
2024-02-28 23:53 ` Yan Zhai
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=9a0052f9-b022-42c9-a5da-1d6ca3b00885@paulmck-laptop \
--to=paulmck@kernel.org \
--cc=alexanderduyck@fb.com \
--cc=bpf@vger.kernel.org \
--cc=daniel@iogearbox.net \
--cc=davem@davemloft.net \
--cc=edumazet@google.com \
--cc=hannes@stressinduktion.org \
--cc=horms@kernel.org \
--cc=jiri@resnulli.us \
--cc=kernel-team@cloudflare.com \
--cc=kuba@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=lixiaoyan@google.com \
--cc=lorenzo@kernel.org \
--cc=netdev@vger.kernel.org \
--cc=pabeni@redhat.com \
--cc=rcu@vger.kernel.org \
--cc=weiwan@google.com \
--cc=yan@cloudflare.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).