From: "Paul E. McKenney" <paulmck@linux.ibm.com>
To: Joel Fernandes <joel@joelfernandes.org>
Cc: rcu@vger.kernel.org, linux-kernel@vger.kernel.org,
mingo@kernel.org, jiangshanlai@gmail.com, dipankar@in.ibm.com,
akpm@linux-foundation.org, mathieu.desnoyers@efficios.com,
josh@joshtriplett.org, tglx@linutronix.de, peterz@infradead.org,
rostedt@goodmis.org, dhowells@redhat.com, edumazet@google.com,
fweisbec@gmail.com, oleg@redhat.com
Subject: Re: [PATCH RFC tip/core/rcu 02/14] rcu/nocb: Add bypass callback queueing
Date: Tue, 6 Aug 2019 17:35:01 -0700 [thread overview]
Message-ID: <20190807003501.GX28441@linux.ibm.com> (raw)
In-Reply-To: <20190807000313.GA161170@google.com>
On Tue, Aug 06, 2019 at 08:03:13PM -0400, Joel Fernandes wrote:
> On Fri, Aug 02, 2019 at 08:14:49AM -0700, Paul E. McKenney wrote:
> > Use of the rcu_data structure's segmented ->cblist for no-CBs CPUs
> > takes advantage of unrelated grace periods, thus reducing the memory
> > footprint in the face of floods of call_rcu() invocations. However,
> > the ->cblist field is a more-complex rcu_segcblist structure which must
> > be protected via locking. Even though there are only three entities
> > which can acquire this lock (the CPU invoking call_rcu(), the no-CBs
> > grace-period kthread, and the no-CBs callbacks kthread), the contention
> > on this lock is excessive under heavy stress.
> >
> > This commit therefore greatly reduces contention by provisioning
> > an rcu_cblist structure field named ->nocb_bypass within the
> > rcu_data structure. Each no-CBs CPU is permitted only a limited
> > number of enqueues onto the ->cblist per jiffy, controlled by a new
> > nocb_nobypass_lim_per_jiffy kernel boot parameter that defaults to
> > about 16 enqueues per millisecond (16 * 1000 / HZ). When that limit is
> > exceeded, the CPU instead enqueues onto the new ->nocb_bypass.
>
> Looks quite interesting. I am guessing the not-no-CB (regular) enqueues don't
> need to use the same technique because both enqueues / callback execution are
> happening on same CPU..
That is the theory! ;-)
> Still looking through patch but I understood the basic idea. Some nits below:
>
> [snip]
> > diff --git a/kernel/rcu/tree.h b/kernel/rcu/tree.h
> > index 2c3e9068671c..e4df86db8137 100644
> > --- a/kernel/rcu/tree.h
> > +++ b/kernel/rcu/tree.h
> > @@ -200,18 +200,26 @@ struct rcu_data {
> > atomic_t nocb_lock_contended; /* Contention experienced. */
> > int nocb_defer_wakeup; /* Defer wakeup of nocb_kthread. */
> > struct timer_list nocb_timer; /* Enforce finite deferral. */
> > + unsigned long nocb_gp_adv_time; /* Last call_rcu() CB adv (jiffies). */
> > +
> > + /* The following fields are used by call_rcu, hence own cacheline. */
> > + raw_spinlock_t nocb_bypass_lock ____cacheline_internodealigned_in_smp;
> > + struct rcu_cblist nocb_bypass; /* Lock-contention-bypass CB list. */
> > + unsigned long nocb_bypass_first; /* Time (jiffies) of first enqueue. */
> > + unsigned long nocb_nobypass_last; /* Last ->cblist enqueue (jiffies). */
> > + int nocb_nobypass_count; /* # ->cblist enqueues at ^^^ time. */
>
> Can these and below fields be ifdef'd out if !CONFIG_RCU_NOCB_CPU so as to
> keep the size of struct smaller for benefit of systems that don't use NOCB?
Please see below...
> > /* The following fields are used by GP kthread, hence own cacheline. */
> > raw_spinlock_t nocb_gp_lock ____cacheline_internodealigned_in_smp;
> > - bool nocb_gp_sleep;
> > - /* Is the nocb GP thread asleep? */
> > + struct timer_list nocb_bypass_timer; /* Force nocb_bypass flush. */
> > + bool nocb_gp_sleep; /* Is the nocb GP thread asleep? */
>
> And these too, I think.
>
>
> > struct swait_queue_head nocb_gp_wq; /* For nocb kthreads to sleep on. */
> > bool nocb_cb_sleep; /* Is the nocb CB thread asleep? */
> > struct task_struct *nocb_cb_kthread;
> > struct rcu_data *nocb_next_cb_rdp;
> > /* Next rcu_data in wakeup chain. */
> >
> > - /* The following fields are used by CB kthread, hence new cachline. */
> > + /* The following fields are used by CB kthread, hence new cacheline. */
> > struct rcu_data *nocb_gp_rdp ____cacheline_internodealigned_in_smp;
> > /* GP rdp takes GP-end wakeups. */
> > #endif /* #ifdef CONFIG_RCU_NOCB_CPU */
I believe that they in fact are all under CONFIG_RCU_NOCB_CPU.
> [snip]
> > +static void rcu_nocb_try_flush_bypass(struct rcu_data *rdp, unsigned long j)
> > +{
> > + rcu_lockdep_assert_cblist_protected(rdp);
> > + if (!rcu_segcblist_is_offloaded(&rdp->cblist) ||
> > + !rcu_nocb_bypass_trylock(rdp))
> > + return;
> > + WARN_ON_ONCE(!rcu_nocb_do_flush_bypass(rdp, NULL, j));
> > +}
> > +
> > +/*
> > + * See whether it is appropriate to use the ->nocb_bypass list in order
> > + * to control contention on ->nocb_lock. A limited number of direct
> > + * enqueues are permitted into ->cblist per jiffy. If ->nocb_bypass
> > + * is non-empty, further callbacks must be placed into ->nocb_bypass,
> > + * otherwise rcu_barrier() breaks. Use rcu_nocb_flush_bypass() to switch
> > + * back to direct use of ->cblist. However, ->nocb_bypass should not be
> > + * used if ->cblist is empty, because otherwise callbacks can be stranded
> > + * on ->nocb_bypass because we cannot count on the current CPU ever again
> > + * invoking call_rcu(). The general rule is that if ->nocb_bypass is
> > + * non-empty, the corresponding no-CBs grace-period kthread must not be
> > + * in an indefinite sleep state.
> > + *
> > + * Finally, it is not permitted to use the bypass during early boot,
> > + * as doing so would confuse the auto-initialization code. Besides
> > + * which, there is no point in worrying about lock contention while
> > + * there is only one CPU in operation.
> > + */
> > +static bool rcu_nocb_try_bypass(struct rcu_data *rdp, struct rcu_head *rhp,
> > + bool *was_alldone, unsigned long flags)
> > +{
> > + unsigned long c;
> > + unsigned long cur_gp_seq;
> > + unsigned long j = jiffies;
> > + long ncbs = rcu_cblist_n_cbs(&rdp->nocb_bypass);
> > +
> > + if (!rcu_segcblist_is_offloaded(&rdp->cblist)) {
> > + *was_alldone = !rcu_segcblist_pend_cbs(&rdp->cblist);
> > + return false; /* Not offloaded, no bypassing. */
> > + }
> > + lockdep_assert_irqs_disabled();
> > +
> > + // Don't use ->nocb_bypass during early boot.
>
> Very minor nit: comment style should be /* */
I thought that Linus said that "//" was now OK. Am I confused?
Thanx, Paul
next prev parent reply other threads:[~2019-08-07 0:35 UTC|newest]
Thread overview: 57+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-08-02 15:14 [PATCH tip/core/rcu 0/14] No-CBs bypass addition for v5.4 Paul E. McKenney
2019-08-02 15:14 ` [PATCH RFC tip/core/rcu 01/14] rcu/nocb: Atomic ->len field in rcu_segcblist structure Paul E. McKenney
2019-08-04 14:50 ` Peter Zijlstra
2019-08-04 14:52 ` Peter Zijlstra
2019-08-04 18:45 ` Paul E. McKenney
2019-08-04 18:42 ` Paul E. McKenney
2019-08-02 15:14 ` [PATCH RFC tip/core/rcu 02/14] rcu/nocb: Add bypass callback queueing Paul E. McKenney
2019-08-07 0:03 ` Joel Fernandes
2019-08-07 0:16 ` Joel Fernandes
2019-08-07 0:35 ` Paul E. McKenney [this message]
2019-08-07 0:40 ` Steven Rostedt
2019-08-07 1:17 ` Paul E. McKenney
2019-08-07 1:24 ` Steven Rostedt
2019-08-07 3:47 ` Paul E. McKenney
2019-08-02 15:14 ` [PATCH RFC tip/core/rcu 03/14] rcu/nocb: EXP Check use and usefulness of ->nocb_lock_contended Paul E. McKenney
2019-08-02 15:14 ` [PATCH RFC tip/core/rcu 04/14] rcu/nocb: Print no-CBs diagnostics when rcutorture writer unduly delayed Paul E. McKenney
2019-08-02 15:14 ` [PATCH RFC tip/core/rcu 05/14] rcu/nocb: Avoid synchronous wakeup in __call_rcu_nocb_wake() Paul E. McKenney
2019-08-02 15:14 ` [PATCH RFC tip/core/rcu 06/14] rcu/nocb: Advance CBs after merge in rcutree_migrate_callbacks() Paul E. McKenney
2019-08-02 15:14 ` [PATCH RFC tip/core/rcu 07/14] rcu/nocb: Reduce nocb_cb_wait() leaf rcu_node ->lock contention Paul E. McKenney
2019-08-02 15:14 ` [PATCH RFC tip/core/rcu 08/14] rcu/nocb: Reduce __call_rcu_nocb_wake() " Paul E. McKenney
2019-08-02 15:14 ` [PATCH RFC tip/core/rcu 09/14] rcu/nocb: Don't wake no-CBs GP kthread if timer posted under overload Paul E. McKenney
2019-08-02 15:14 ` [PATCH RFC tip/core/rcu 10/14] rcu: Allow rcu_do_batch() to dynamically adjust batch sizes Paul E. McKenney
2019-08-02 15:14 ` [PATCH RFC tip/core/rcu 11/14] EXP nohz: Add TICK_DEP_BIT_RCU Paul E. McKenney
2019-08-02 15:14 ` [PATCH RFC tip/core/rcu 12/14] rcu/nohz: Force on tick when invoking lots of callbacks Paul E. McKenney
2019-08-02 15:15 ` [PATCH RFC tip/core/rcu 13/14] rcutorture: Force on tick for readers and callback flooders Paul E. McKenney
2019-08-02 15:15 ` [PATCH RFC tip/core/rcu 14/14] rcu/nohz: Make multi_cpu_stop() enable tick on all online CPUs Paul E. McKenney
2019-08-04 14:43 ` Peter Zijlstra
2019-08-04 14:48 ` Peter Zijlstra
2019-08-04 18:41 ` Paul E. McKenney
2019-08-04 20:24 ` Paul E. McKenney
2019-08-05 4:19 ` Paul E. McKenney
2019-08-05 8:07 ` Peter Zijlstra
2019-08-05 14:47 ` Paul E. McKenney
2019-08-05 8:05 ` Peter Zijlstra
2019-08-05 14:54 ` Paul E. McKenney
2019-08-05 15:50 ` Peter Zijlstra
2019-08-05 17:48 ` Paul E. McKenney
2019-08-06 18:08 ` Paul E. McKenney
2019-08-07 21:41 ` Paul E. McKenney
2019-08-08 20:35 ` Paul E. McKenney
2019-08-08 21:30 ` Paul E. McKenney
2019-08-09 16:51 ` Paul E. McKenney
2019-08-09 18:07 ` Joel Fernandes
2019-08-09 18:39 ` Paul E. McKenney
2019-08-12 21:02 ` Frederic Weisbecker
2019-08-12 23:23 ` Paul E. McKenney
2019-08-13 1:33 ` Joel Fernandes
2019-08-13 12:30 ` Frederic Weisbecker
2019-08-13 14:48 ` Paul E. McKenney
2019-08-14 17:55 ` Joel Fernandes
2019-08-14 22:05 ` Paul E. McKenney
2019-08-15 15:07 ` Joel Fernandes
2019-08-15 17:23 ` Paul E. McKenney
2019-08-15 18:15 ` Joel Fernandes
2019-08-15 18:39 ` Paul E. McKenney
2019-08-15 19:42 ` Joel Fernandes
2019-08-13 21:06 ` Paul E. McKenney
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=20190807003501.GX28441@linux.ibm.com \
--to=paulmck@linux.ibm.com \
--cc=akpm@linux-foundation.org \
--cc=dhowells@redhat.com \
--cc=dipankar@in.ibm.com \
--cc=edumazet@google.com \
--cc=fweisbec@gmail.com \
--cc=jiangshanlai@gmail.com \
--cc=joel@joelfernandes.org \
--cc=josh@joshtriplett.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mathieu.desnoyers@efficios.com \
--cc=mingo@kernel.org \
--cc=oleg@redhat.com \
--cc=peterz@infradead.org \
--cc=rcu@vger.kernel.org \
--cc=rostedt@goodmis.org \
--cc=tglx@linutronix.de \
/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).