netdev - Re: [PATCH 2/4] e1000: add module option to set transmit descriptor size

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: <20080627152642.GG32408@gospo.rdu.redhat.com>
Date:	Fri, 27 Jun 2008 11:26:42 -0400
From:	Andy Gospodarek <andy@...yhouse.net>
To:	Arthur Jones <ajones@...erbed.com>
Cc:	Jeff Garzik <jeff@...zik.org>,
	Andy Gospodarek <andy@...yhouse.net>,
	"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
	"jeffrey.t.kirsher@...el.com" <jeffrey.t.kirsher@...el.com>,
	"jesse.brandeburg@...el.com" <jesse.brandeburg@...el.com>
Subject: Re: [PATCH 2/4] e1000: add module option to set transmit descriptor size

On Fri, Jun 27, 2008 at 07:55:22AM -0700, Arthur Jones wrote:
> Hi Jeff, ...
> 
> On Fri, Jun 27, 2008 at 07:21:22AM -0700, Arthur Jones wrote:
> > On Thu, Jun 26, 2008 at 10:10:37PM -0700, Jeff Garzik wrote:
> > > Andy Gospodarek wrote:
> > > > This patch added the TxDescPower parameter to the e1000 module.  This
> > > > parameter represents the size-order of each transmit descriptor.  The
> > > > valid size for descriptors would be 2^7 (128) - 2^12 (4096) bytes each.
> > > > As this value decreases one may want to consider increasing the
> > > > TxDescriptors value to maintain the same amount of frame memory.  This
> > > > patch does not change the defaults for any particular hardware model, so
> > > > this will not have an effect on existing users.
> > > [...]
> > > I don't see why a module parameter is needed.  In practice, nobody will
> > > ever think about this option, much less use it.
> 
> One thing that may not be immediately obvious is
> that setting the descriptor payload size low to avoid
> the chip bug has a very negative effect on performance.
> I don't want to lower the descriptor payload size
> unless I absolutely have to -- e.g. environments where
> link flap on chip reset causes major distruptions to the
> network...
> 

Arthur,

That is absolutely correct.  For those that want to tune, the
performance hit can be significant and I would never recommend that
someone tweaks these settings unless they have problems first.

-andy

--
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