netdev - Re: PPPoE performance regression

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: <1339609400.14785.23.camel@shinybook.infradead.org>
Date:	Wed, 13 Jun 2012 18:43:20 +0100
From:	David Woodhouse <dwmw2@...radead.org>
To:	Benjamin LaHaise <bcrl@...ck.org>
Cc:	Nathan Williams <nathan@...verse.com.au>,
	Karl Hiramoto <karl@...amoto.org>,
	"David S. Miller" <davem@...emloft.net>, netdev@...r.kernel.org,
	Paul Mackerras <paulus@...ba.org>,
	John Crispin <blogic@...nwrt.org>
Subject: Re: PPPoE performance regression

On Wed, 2012-06-13 at 13:21 -0400, Benjamin LaHaise wrote:
> On the whole question of PPPoE over intermediate ethernet links to ADSL 
> modems, I think it would be possible to limit latency by implementing a 
> sliding window clocked using LCP ECHO requests.  Does this sound worthwhile 
> implementing?

Maybe, for someone who actually cares about those who are using separate
ADSL modems, and doesn't think they should just better hardware if they
care about the performance and queue management :)

>   What sort of queue depths are you looking at for the ATM devices
> you're working on? 

We're managing the queue to keep it short. On the PPPoATM side, we (now)
strictly limit the number of packets between the ppp_generic code and
the ATM device. It's basically one in-flight and one waiting to be
handed to the device in the TX done interrupt. PPP is designed to feed
new packets with low latency, after all.

The Solos ADSL device *used* to eat as many packets as it had internal
RAM to store them in, acknowledging the "transmit" as soon as it had
buffered them. I got Nathan to fix that some time ago, and the internal
queue is fairly short now.

-- 
dwmw2

Download attachment "smime.p7s" of type "application/x-pkcs7-signature" (6171 bytes)