Re: PPPoE performance regression

[David Woodhouse <dwmw2@infradead.org>]

[-- Attachment #1: Type: text/plain, Size: 2130 bytes --]

On Wed, 2012-06-13 at 12:31 -0400, Benjamin LaHaise wrote:
> I would contend that PPPoE over br2684 is not the common case.  The vast 
> majority of users in client mode are going to be using PPPoE over an 
> ethernet link to a DSL modem (or cable or wireless radios even).  Just look 
> at what DSL modems are available for users in computer stores / what ISPs 
> actually ship to their users.  Real ATM exposing devices are rare.

I'm looking at the class of device on which OpenWRT runs. Linux is *on*
the router with the ADSL port, not connected to it via Ethernet.

I can buy lots of these in the shop. Anything that's an ADSL *router*
rather than *modem* is likely to be running, or at least capable of
running, Linux.

Admittedly, if you have access to the native ADSL interface then you'd
do a lot better to run PPPoA — but I already fixed this issue for PPPoA.
There are people in some parts of the world who are using PPPoEoA and
putting up with the resulting MTU issues because the *ISP* doesn't
support proper PPPoA.

And even if it *were* rare... this is the case that *should* work best,
where we have complete control of the hardware. There's no excuse for
the behaviour that we currently see with PPPoE on BR2684.

> > On the ISP side if the skb ends up sitting on a receive queue of a user
> > socket, and nothing is servicing that socket, surely the transmits on
> > this channel weren't happening anyway?
> 
> True, but it's a design issue we've had to contend with elsewhere in the 
> various tunnelling protocols.
> 
> Don't get me wrong: I am very much in favour of intelligent queue 
> management, but this approach simply does not work for the vast majority 
> of PPPoE users, while it adds overhead that will negatively impact access 
> concentrators. 

I think that's largely true of BQL in general, isn't it? That's OK; it's
a config option. I suspect if I make this accounting of PPPoE / L2TP
packets conditional on BQL (or perhaps on a separate config option
PPP_BQL) that ought to address your concern about the cases where you
don't need it?

-- 
dwmw2

[-- Attachment #2: smime.p7s --]
[-- Type: application/x-pkcs7-signature, Size: 6171 bytes --]