linux-kernel - Re: [PATCH 4/5] rcusync: introduce rcu_sync

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, 7 Oct 2013 12:49:00 +0200
From:	Peter Zijlstra <peterz@...radead.org>
To:	Oleg Nesterov <oleg@...hat.com>
Cc:	Paul McKenney <paulmck@...ux.vnet.ibm.com>,
	Mel Gorman <mgorman@...e.de>, Rik van Riel <riel@...hat.com>,
	Srikar Dronamraju <srikar@...ux.vnet.ibm.com>,
	Ingo Molnar <mingo@...nel.org>,
	Andrea Arcangeli <aarcange@...hat.com>,
	Johannes Weiner <hannes@...xchg.org>,
	Thomas Gleixner <tglx@...utronix.de>,
	Steven Rostedt <rostedt@...dmis.org>,
	Linus Torvalds <torvalds@...ux-foundation.org>,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH 4/5] rcusync: introduce rcu_sync_struct->exclusive mode

On Sun, Oct 06, 2013 at 03:22:40PM +0200, Oleg Nesterov wrote:
> On 10/04, Peter Zijlstra wrote:
> > That would yield something like so I suppose:
> >
> > void rcu_sync_enter(struct rcu_sync_struct *rss)
> > {
> > 	bool need_wait, need_sync;
> >
> > 	spin_lock_irq(&rss->rss_lock);
> > 	if (rss->exclusive && rss->gp_count) {
> > 		__wait_event_locked(rss->gp_wait, rss->gp_count);
>                                                   ^^^^^^^^^^^^^
> I guess you meant !rss->gp_count.

Uh yes, obviously :-)

> I am obviously biased, but imho the code looks worse this way.
> I like the current simple "need_wait" and "gp_count != 0" logic.

Yeah, I know.. but it doesn't add the extra variable and doesn't play
games with the completion implementation.

> And afaics this is racy,
> 
> Yes, but if rcu_sync_exit() does __wake_up_locked(), then
> autoremove_wake_function() makes waitqueue_active() == F. If the pending
> rcu_sync_func() takes ->rss_lock first we have a problem.

Ah indeed, it seems I got confused between DECLARE_WAITQUEUE and
DEFINE_WAIT; there's too damn many variants there :/

> Easy to fix, but needs more complications.
> 
> Or we can simply ignore the fact that rcu_sync_func() can race with
> wakeup. This can lead to unnecessary sched_sync() but this case is
> unlikely. IOW,
> 
> 	spin_lock_irq(&rss->rss_lock);
> 	if (rss->exclusive)
> 		wait_event_locked(rss->gp_wait, !rss->gp_count);
> 	need_wait = rss->gp_count++;
> 	need_sync = rss->gp_state == GP_IDLE;
> 	if (need_sync)
> 		rss->gp_state = GP_PENDING;
> 	spin_unlock_irq(&rss->lock);
> 
> But still I don't like the (imho) unnecessary complications. And the
> fact we can race with rcu_sync_func() even if this is very unlikely,
> this just doesn't look good.

OK.. I'll give up trying to wreck this stuff ;-)
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/