[ILUG] Re: pppd bugs (Linux PPP kernel driver suckage) (fwd)

[Michael Conry]

just thought some of this might be of interest. This guy was getting
errors similar enough to my own while using pppd 2.3.9 on top of ssh to
set up a tunnel. pppd was choking at high transfer rates. he gave excerpts
from the syslog on the client side of the tunnel, and they looked like my
error messages (not sure how similar the circumstances really are though).

---------- Forwarded message ----------
Date: Mon, 29 May 2000 19:36:26 +0200 (MEST)
From: Florian-Daniel Otel <otel at ce.chalmers.se>
To: michael conry <michael.conry at ucd.ie>
Cc: linux-kernel at vger.rutgers.edu
Subject: Re: pppd bugs (Linux PPP kernel driver suckage)


> Hi Florian, you don't know me
> i saw a post you made at
>  http://www.uwsg.iu.edu/hypermail/linux/kernel/9910.2/0675.html
> about bad frames (bad fcs, then frame, count=55, or whatever).
> I just wondered if you sorted out the problem. I have the same issue at
> the moment, with the same symptoms.

Well, not directly, but i found a work-around. After speaking with
Paul  Mackerras (pppd maintainer) days worth of debugging [well, i
spare the rest of the pain] seems that Linux 2.2 kernel ppp driver
sucks. Plain and simple. Right now i'm using that same setup (PPP over
SSH running over a 100 Mbps Eth) but the server side of the connection
is now a _fast_ dual Ultra60/Solaris8 machine. The Solaris is running
same PPP version (pppd-2.3.11) as the client side (Linux, 2.2.15) but 
with the Solaris kernel drivers supplied w/ pppd-2.3.11 as modules. Why this
works like a heavenly bliss and using Linux as a server
sucks...Well...i don't care anymore :) 


In the plans was to try one of the  2.3 kernels, see if bundled PPP
driver is any better. Needless to say, it never happened. Also, i was
planning to try sparc-linux on the same hardware to really
pinpoint the PPP kernel driver as the source of troubles. (I've tried
sparc-linux on the server side  but  on a slower dual-SS20 and the
same suckage happened, but there you might blame hardw. performance).  

All in all, what i would do is try a 2.3. kernel / latest-greatest
PPP kernel driver on some fast machine (in case you really intend to
push it _hard_ ;) ) . Not much hope, but still, might worth a shot.
  
If you have any results/experience  with that, i will be glad to hear
from you.  

Hope this helps (somehow),

> TIA
> michael conry



Florian.


P.S: I Cc: this into linux-kernel noise, maybe someone will pick up
the issue  - despite the total lack of any analysis/anatomy of the
claimed  PPP kernel driver problem...>)