From: "Paul E. McKenney" <paulmck@kernel.org>
To: Z qiang <qiang.zhang1211@gmail.com>
Cc: frederic@kernel.org, neeraj.upadhyay@kernel.org,
joel@joelfernandes.org, rcu@vger.kernel.org,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH] rcu-tasks: Remove unnecessary lazy_jiffies in call_rcu_tasks_generic_timer()
Date: Sun, 17 Mar 2024 19:48:54 -0700 [thread overview]
Message-ID: <9247666f-8557-4419-a4aa-42193c386086@paulmck-laptop> (raw)
In-Reply-To: <CALm+0cVywbBNfzZBjcxJNvpeixkBrfPZu_4qSnOXQvqYTbjMyg@mail.gmail.com>
On Mon, Mar 18, 2024 at 10:35:44AM +0800, Z qiang wrote:
> > On Sun, Mar 17, 2024 at 10:31:22PM +0800, Z qiang wrote:
> > > > On Wed, Mar 13, 2024 at 12:49:50PM +0800, Z qiang wrote:
> > > > > > On Tue, Mar 12, 2024 at 07:35:24PM +0800, Zqiang wrote:
> > > > > > > The rcu_tasks_percpu structure's->lazy_timer is queued only when
> > > > > > > the rcu_tasks structure's->lazy_jiffies is not equal to zero in
> > > > > > > call_rcu_tasks_generic(), if the lazy_timer callback is invoked,
> > > > > > > that means the lazy_jiffes is not equal to zero, this commit
> > > > > > > therefore remove lazy_jiffies check in call_rcu_tasks_generic_timer().
> > > > > > >
> > > > > > > Signed-off-by: Zqiang <qiang.zhang1211@gmail.com>
> > > > > > > ---
> > > > > > > kernel/rcu/tasks.h | 2 +-
> > > > > > > 1 file changed, 1 insertion(+), 1 deletion(-)
> > > > > > >
> > > > > > > diff --git a/kernel/rcu/tasks.h b/kernel/rcu/tasks.h
> > > > > > > index b1254cf3c210..439e0b9a2656 100644
> > > > > > > --- a/kernel/rcu/tasks.h
> > > > > > > +++ b/kernel/rcu/tasks.h
> > > > > > > @@ -299,7 +299,7 @@ static void call_rcu_tasks_generic_timer(struct timer_list *tlp)
> > > > > > >
> > > > > > > rtp = rtpcp->rtpp;
> > > > > > > raw_spin_lock_irqsave_rcu_node(rtpcp, flags);
> > > > > > > - if (!rcu_segcblist_empty(&rtpcp->cblist) && rtp->lazy_jiffies) {
> > > > > > > + if (!rcu_segcblist_empty(&rtpcp->cblist)) {
> > > > > >
> > > > > > Good eyes!
> > > > > >
> > > > > > But did you test with something like a WARN_ON_ONCE(rtp->lazy_jiffies)?
> > > > >
> > > > > Hi, Paul
> > > > >
> > > > > + if (!rcu_segcblist_empty(&rtpcp->cblist) &&
> > > > > !WARN_ON_ONCE(!rtp->lazy_jiffies))
> > > > >
> > > > > I've done tests like this:
> > > > >
> > > > > 1. runqemu nographic kvm slirp qemuparams="-smp 4 -m 2048M -drive
> > > > > file=$PWD/share.img,if=virtio"
> > > > > bootparams="rcupdate.rcu_tasks_trace_lazy_ms=0" -d
> > > > >
> > > > > 2. insmod torture.ko
> > > > > insmod rcutorture.ko torture_type=tasks-tracing fwd_progress=4
> > > > >
> > > > > 3. bpftrace -e 't:timer:timer_expire_entry /args->function ==
> > > > > kaddr("call_rcu_tasks_generic_timer")/
> > > > > {
> > > > > printf("comm:%s,cpu:%d,stack:%s,func:%s\n", comm, cpu, kstack,
> > > > > ksym(args->function)); }'
> > > > >
> > > > > The call_rcu_tasks_generic_timer() has never been executed.
> > > >
> > > > Very good!
> > > >
> > > > Then if we get a couple of acks or reviews from the others acknowledging
> > > > that if they ever make ->lazy_jiffies be changeable at runtime, they
> > > > will remember to do something to adjust this logic appropriately, I will
> > > > take it. ;-)
> > > >
> > >
> > > Hi, Paul
> > >
> > > Would it be better to modify it as follows? set needwake not
> > > depend on lazy_jiffies, if ->lazy_jiffies be changed at runtime,
> > > and set it to zero, will miss the chance to wake up gp kthreads.
> > >
> > > diff --git a/kernel/rcu/tasks.h b/kernel/rcu/tasks.h
> > > index e7ac9138a4fd..aea2b71af36c 100644
> > > --- a/kernel/rcu/tasks.h
> > > +++ b/kernel/rcu/tasks.h
> > > @@ -299,11 +299,12 @@ static void call_rcu_tasks_generic_timer(struct
> > > timer_list *tlp)
> > >
> > > rtp = rtpcp->rtpp;
> > > raw_spin_lock_irqsave_rcu_node(rtpcp, flags);
> > > - if (!rcu_segcblist_empty(&rtpcp->cblist) && rtp->lazy_jiffies) {
> > > + if (!rcu_segcblist_empty(&rtpcp->cblist)) {
> > > if (!rtpcp->urgent_gp)
> > > rtpcp->urgent_gp = 1;
> > > needwake = true;
> > > - mod_timer(&rtpcp->lazy_timer, rcu_tasks_lazy_time(rtp));
> > > + if (rtp->lazy_jiffies)
> > > + mod_timer(&rtpcp->lazy_timer, rcu_tasks_lazy_time(rtp));
> > > }
> > > raw_spin_unlock_irqrestore_rcu_node(rtpcp, flags);
> > > if (needwake)
> >
> > Interesting approach!
> >
> > But how does that avoid defeating laziness by doing the wakeup early?
>
> Hello, Paul
>
> Does this mean that if cblist is empty, we will miss the opportunity to
> enqueue the timer again? If so, we can move mod_timer() outside
> the if condition.
> or I didn't understand your question?
Never mind! I was getting confused, and forgetting that this code has
already seen a timeout.
Let's see what others think.
Thanx, Paul
> Thanks
> Zqiang
>
>
> >
> > Thanx, Paul
> >
> > > Thanks
> > > Zqiang
> > >
> > >
> > > > Thanx, Paul
> > > >
> > > > > Thanks
> > > > > Zqiang
> > > > >
> > > > >
> > > > > >
> > > > > > Thanx, Paul
> > > > > >
> > > > > > > if (!rtpcp->urgent_gp)
> > > > > > > rtpcp->urgent_gp = 1;
> > > > > > > needwake = true;
> > > > > > > --
> > > > > > > 2.17.1
> > > > > > >
prev parent reply other threads:[~2024-03-18 2:48 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-03-12 11:35 [PATCH] rcu-tasks: Remove unnecessary lazy_jiffies in call_rcu_tasks_generic_timer() Zqiang
2024-03-12 22:05 ` Paul E. McKenney
2024-03-13 4:49 ` Z qiang
2024-03-17 7:38 ` Paul E. McKenney
2024-03-17 14:31 ` Z qiang
2024-03-17 20:28 ` Paul E. McKenney
2024-03-18 2:35 ` Z qiang
2024-03-18 2:48 ` Paul E. McKenney [this message]
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=9247666f-8557-4419-a4aa-42193c386086@paulmck-laptop \
--to=paulmck@kernel.org \
--cc=frederic@kernel.org \
--cc=joel@joelfernandes.org \
--cc=linux-kernel@vger.kernel.org \
--cc=neeraj.upadhyay@kernel.org \
--cc=qiang.zhang1211@gmail.com \
--cc=rcu@vger.kernel.org \
/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 read-only IMAP folder(s) and NNTP newsgroup(s).