Re: [PATCH v5 2/4] rcu: Reduce synchronize_rcu() latency

From: Frederic Weisbecker <frederic@kernel.org>
To: Z qiang <qiang.zhang1211@gmail.com>
Cc: "Uladzislau Rezki (Sony)" <urezki@gmail.com>,
	"Paul E . McKenney" <paulmck@kernel.org>,
	RCU <rcu@vger.kernel.org>,
	Neeraj upadhyay <Neeraj.Upadhyay@amd.com>,
	Boqun Feng <boqun.feng@gmail.com>,
	Hillf Danton <hdanton@sina.com>,
	Joel Fernandes <joel@joelfernandes.org>,
	LKML <linux-kernel@vger.kernel.org>,
	Oleksiy Avramchenko <oleksiy.avramchenko@sony.com>
Subject: Re: [PATCH v5 2/4] rcu: Reduce synchronize_rcu() latency
Date: Tue, 27 Feb 2024 17:16:27 +0100	[thread overview]
Message-ID: <Zd4K28_GkRQBn9Ik@localhost.localdomain> (raw)
In-Reply-To: <Zd3zptTLlUDCg1L7@localhost.localdomain>

Le Tue, Feb 27, 2024 at 03:37:26PM +0100, Frederic Weisbecker a écrit :
> Le Tue, Feb 27, 2024 at 02:39:55PM +0800, Z qiang a écrit :
> > > Can the following race happen?
> > >
> > > CPU 0                                                   CPU 1
> > > -----                                                   -----
> > >
> > > // wait_tail == HEAD1
> > > rcu_sr_normal_gp_cleanup() {
> > >     // has passed SR_MAX_USERS_WAKE_FROM_GP
> > >     wait_tail->next = next;
> > >     // done_tail = HEAD1
> > >     smp_store_release(&rcu_state.srs_done_tail, wait_tail);
> > >     queue_work() {
> > >         test_and_set_bit(WORK_STRUCT_PENDING_BIT, work_data_bits(work)
> > >         __queue_work()
> > >     }
> > > }
> > >
> > >                                                       set_work_pool_and_clear_pending()
> > >                                                       rcu_sr_normal_gp_cleanup_work() {
> > > // new GP, wait_tail == HEAD2
> > > rcu_sr_normal_gp_cleanup() {
> > >     // executes all completion, but stop at HEAD1
> > >     wait_tail->next = HEAD1;
> > >     // done_tail = HEAD2
> > >     smp_store_release(&rcu_state.srs_done_tail, wait_tail);
> > >     queue_work() {
> > >         test_and_set_bit(WORK_STRUCT_PENDING_BIT, work_data_bits(work)
> > >         __queue_work()
> > >     }
> > > }
> > >                                                           // done = HEAD2
> > >                                                           done = smp_load_acquire(&rcu_state.srs_done_tail);
> > >                                                           // head = HEAD1
> > >                                                           head = done->next;
> > >                                                           done->next = NULL;
> > >                                                           llist_for_each_safe() {
> > >                                                               // completes all callbacks, release HEAD1
> > >                                                           }
> > >                                                       }
> > >                                                       // Process second queue
> > >                                                       set_work_pool_and_clear_pending()
> > >                                                       rcu_sr_normal_gp_cleanup_work() {
> > >                                                           // done = HEAD2
> > >                                                           done = smp_load_acquire(&rcu_state.srs_done_tail);
> > >
> > > // new GP, wait_tail == HEAD3
> > > rcu_sr_normal_gp_cleanup() {
> > >     // Finds HEAD2 with ->next == NULL at the end
> > >     rcu_sr_put_wait_head(HEAD2)
> > 
> > It seems that we should move rcu_sr_put_wait_head() from
> > rcu_sr_normal_gp_cleanup() to
> > rcu_sr_normal_gp_cleanup_work(), if find wait_head->next == NULL, invoke
> > rcu_sr_put_wait_head() to release wait_head.
> 
> Well, rcu_sr_normal_gp_cleanup_work() already put all the wait heads
> that are _after_ srs_done_tail. But it can't put the srs_done_tail itself
> without introducing even worse races...
> 

(I forgot to mention this race actually concerns the last patch (4/4))