linux-kernel - Re: [PATCH v2 12/14] locks: give the blocked

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]

Message-ID: <20130613152036.GC20666@fieldses.org>
Date:	Thu, 13 Jun 2013 11:20:36 -0400
From:	"J. Bruce Fields" <bfields@...ldses.org>
To:	Jeff Layton <jlayton@...hat.com>
Cc:	viro@...iv.linux.org.uk, matthew@....cx, dhowells@...hat.com,
	sage@...tank.com, smfrench@...il.com, swhiteho@...hat.com,
	Trond.Myklebust@...app.com, akpm@...ux-foundation.org,
	linux-kernel@...r.kernel.org, linux-afs@...ts.infradead.org,
	ceph-devel@...r.kernel.org, linux-cifs@...r.kernel.org,
	samba-technical@...ts.samba.org, cluster-devel@...hat.com,
	linux-nfs@...r.kernel.org, linux-fsdevel@...r.kernel.org,
	piastryyy@...il.com
Subject: Re: [PATCH v2 12/14] locks: give the blocked_hash its own spinlock

On Thu, Jun 13, 2013 at 11:18:44AM -0400, Jeff Layton wrote:
> On Thu, 13 Jun 2013 11:02:47 -0400
> "J. Bruce Fields" <bfields@...ldses.org> wrote:
> 
> > On Tue, Jun 11, 2013 at 07:09:06AM -0400, Jeff Layton wrote:
> > > There's no reason we have to protect the blocked_hash and file_lock_list
> > > with the same spinlock. With the tests I have, breaking it in two gives
> > > a barely measurable performance benefit, but it seems reasonable to make
> > > this locking as granular as possible.
> > 
> > Out of curiosity...  In the typical case when adding/removing a lock,
> > aren't both lists being modified in rapid succession?
> > 
> > I wonder if it would be better to instead stick with one lock and take
> > care to acquire it only once to cover both manipulations.
> > 
> > --b.
> > 
> 
> That's not really the case...
> 
> Typically, when doing a call into __posix_lock_file with FL_SLEEP set,
> we either end up blocking on the lock or acquiring it. In either case,
> we'll only end up taking one of the global spinlocks. The reason for
> this is that blocker is what dequeues a waiter from the blocked_hash
> before waking it up (in locks_wake_up_posix_blocks).
> 
> Also, while this patch description doesn't spell it out, we require a
> truly global lock for deadlock detection. In a later patch though, I
> convert the file_lock_lock to a per-cpu spinlock. So we really do need
> to separate the locks here in order to make the per-cpu file_lock_list
> worthwhile.

Oh, right, got it!

--b.
--
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/