netdev - Re: [PATCH] l2tp: synchronise u64 stats writer callsites

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] [day] [month] [year] [list]

Message-ID: <20120621161955.GA29812@jackdaw>
Date:	Thu, 21 Jun 2012 17:19:56 +0100
From:	Tom Parkin <tparkin@...alix.com>
To:	David Laight <David.Laight@...LAB.COM>
Cc:	netdev@...r.kernel.org, James Chapman <jchapman@...alix.com>
Subject: Re: [PATCH] l2tp: synchronise u64 stats writer callsites

On Thu, Jun 21, 2012 at 04:18:12PM +0100, David Laight wrote:
> The purpose of the u64_stats_update_begin/end is to
> perform lockless writes of the stats.
> If you need to lock them (because multiple threads can
> be writing to stats covered by the same 'syncp' at the
> same time) then the reader might as well use the same lock.
> 
> Otherwise split the 'syncp' such that only one update
> can be happening (for each sync).

Thanks David.

I think the best approach is probably to attempt to partition the l2tp
statistics such that we can be sure of single-threaded writer access
for each dataset, which can then be covered by a 'syncp'.

If that turns out not to be possible, I suppose the fallback position
is to do away with the u64_stats_update* calls and just use a spinlock
instead.

I'll look at implementing the former and put a new patch together.

Tom
-- 
Tom Parkin
Katalix Systems Ltd
http://www.katalix.com
Catalysts for your Embedded Linux software development

Download attachment "signature.asc" of type "application/pgp-signature" (491 bytes)