linux-kernel - Re: [PATCH v9 6/7] rcu/tree: segcblist: Remove redundant smp

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Mon, 9 Nov 2020 17:41:51 -0800
From:   "Paul E. McKenney" <paulmck@...nel.org>
To:     Joel Fernandes <joel@...lfernandes.org>
Cc:     linux-kernel@...r.kernel.org,
        Frederic Weisbecker <frederic@...nel.org>,
        Josh Triplett <josh@...htriplett.org>,
        Lai Jiangshan <jiangshanlai@...il.com>,
        Marco Elver <elver@...gle.com>,
        Mathieu Desnoyers <mathieu.desnoyers@...icios.com>,
        rcu@...r.kernel.org, Steven Rostedt <rostedt@...dmis.org>,
        "Uladzislau Rezki (Sony)" <urezki@...il.com>, fweisbec@...il.com,
        neeraj.iitr10@...il.com
Subject: Re: [PATCH v9 6/7] rcu/tree: segcblist: Remove redundant smp_mb()s

On Fri, Nov 06, 2020 at 07:26:04PM -0500, Joel Fernandes wrote:
> On Wed, Nov 04, 2020 at 07:57:13PM -0800, Paul E. McKenney wrote:
> > On Tue, Nov 03, 2020 at 09:26:02AM -0500, Joel Fernandes (Google) wrote:
> > > This memory barrier is not needed as rcu_segcblist_add_len() already
> > > includes a memory barrier *before* and *after* the length of the list is
> > > updated.
> > > 
> > > Same reasoning for rcu_segcblist_enqueue().
> > 
> > I suggest a commit log like the following:
> > 
> > ------------------------------------------------------------------------
> > 
> > The full memory barriers in rcu_segcblist_enqueue() and in rcu_do_batch()
> > are not needed because rcu_segcblist_add_len(), and thus also
> > rcu_segcblist_inc_len(), already includes a memory barrier *before*
> > and *after* the length of the list is updated.
> > 
> > This commit therefore removes these redundant smp_mb() invocations.
> > 
> > ------------------------------------------------------------------------
> > 
> > Other than that, looks good!  I could hand-apply it, but that
> > would just cause more churn with the addition of the call to
> > rcu_segcblist_inc_seglen().  So could you please update the commit log
> > when you repost, whether to the mailing list or from your git tree?
> 
> Done, I updated it in my tree. I will share the link to tree on IRC.

And I have pulled this in, thank you!

							Thanx, Paul

> thanks,
> 
>  - Joel
> 
> > 
> > 							Thanx, Paul
> > 
> > > Reviewed-by: Frederic Weisbecker <frederic@...nel.org>
> > > Signed-off-by: Joel Fernandes (Google) <joel@...lfernandes.org>
> > > ---
> > >  kernel/rcu/rcu_segcblist.c | 1 -
> > >  kernel/rcu/tree.c          | 1 -
> > >  2 files changed, 2 deletions(-)
> > > 
> > > diff --git a/kernel/rcu/rcu_segcblist.c b/kernel/rcu/rcu_segcblist.c
> > > index e9e72d72f7a6..d96272e8d604 100644
> > > --- a/kernel/rcu/rcu_segcblist.c
> > > +++ b/kernel/rcu/rcu_segcblist.c
> > > @@ -268,7 +268,6 @@ void rcu_segcblist_enqueue(struct rcu_segcblist *rsclp,
> > >  			   struct rcu_head *rhp)
> > >  {
> > >  	rcu_segcblist_inc_len(rsclp);
> > > -	smp_mb(); /* Ensure counts are updated before callback is enqueued. */
> > >  	rcu_segcblist_inc_seglen(rsclp, RCU_NEXT_TAIL);
> > >  	rhp->next = NULL;
> > >  	WRITE_ONCE(*rsclp->tails[RCU_NEXT_TAIL], rhp);
> > > diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> > > index f6c6653b3ec2..fb2a5ac4a59c 100644
> > > --- a/kernel/rcu/tree.c
> > > +++ b/kernel/rcu/tree.c
> > > @@ -2525,7 +2525,6 @@ static void rcu_do_batch(struct rcu_data *rdp)
> > >  
> > >  	/* Update counts and requeue any remaining callbacks. */
> > >  	rcu_segcblist_insert_done_cbs(&rdp->cblist, &rcl);
> > > -	smp_mb(); /* List handling before counting for rcu_barrier(). */
> > >  	rcu_segcblist_add_len(&rdp->cblist, -count);
> > >  
> > >  	/* Reinstate batch limit if we have worked down the excess. */
> > > -- 
> > > 2.29.1.341.ge80a0c044ae-goog
> > >