[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20081007143346.GA6384@linux.vnet.ibm.com>
Date: Tue, 7 Oct 2008 07:33:46 -0700
From: "Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>
To: Christoph Lameter <cl@...ux-foundation.org>
Cc: Evgeniy Polyakov <johnpol@....mipt.ru>,
Corey Minyard <minyard@....org>,
David Miller <davem@...emloft.net>, dada1@...mosbay.com,
linux-kernel@...r.kernel.org, netdev@...r.kernel.org,
shemminger@...tta.com
Subject: Re: [PATCH 3/3] Convert the UDP hash lock to RCU
On Tue, Oct 07, 2008 at 09:16:13AM -0500, Christoph Lameter wrote:
> Evgeniy Polyakov wrote:
> > On Mon, Oct 06, 2008 at 06:08:09PM -0500, Corey Minyard (minyard@....org) wrote:
> >> Would using SLAB_DESTROY_BY_RCU be ok, or would that be too expensive?
> >
> > I tested skb destruction via RCU path, and got 2.5 times worse numbers
> > with small-packets-bulk-transfer workload.
>
> Was this with regular RCU freeing? This will cool down the cacheline before
> frees. You need SLAB_DESTROY_BY_RCU to keep the objects cache hot.
Indeed!
But care is required -- SLAB_DESTROY_BY_RCU permits objects to be freed
and reallocated while a reader holds a reference. The only guarantee is
that the -type- of the data structure will not change while a reader holds
a reference. With something like UDP, this might well be sufficient.
Just be careful! ;-)
Thanx, Paul
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists