Subject: Re: kern/20039 - ppp and mbuf leak
To: Simon Burge <>
From: Michael Kukat <>
List: netbsd-users
Date: 04/19/2003 13:35:18
Hi !

On Sat, 19 Apr 2003, Simon Burge wrote:
> > is there any ongoing effort in kern/20039, or has someone even solved the
> > problem? I seem to have the same problem here, but it seems not to be
> > connection-related, it also seems to depend on data transfer. Yesterday, i had
> > 4 machine freezes in about 2 GB download. Every 400-600 MB, the machine
> > freezes. I didn't find kern/20039 that time, so i didn't watch the mbuf count,
> > but after kernel panic messages about something like no memory left, this might
> > really be the same problem.
> The "fix" I've used is to switch to in-kernel pppoe (details mentioned
> elsewhere in this thread).
> Also, when I tested -current a month or three ago, it _didn't_ have the
> mbuf leak problem with rp-pppoe that existed on the 1.6 branch.  Are the
> problems you're seeing on -current or 1.6?

It was 1.6 distribution, so this could match the facts you are seeting. But
after some testing, i also switched to in-kernel pppoe now, and everything
works fine. I left my fingers off the last time because the statement in
pppoe(4), that just the local outgoing connections are clamped, not the
forwarded connections. But due to this can be set up in ipnat, everything works
fine now.

BTW... a really big performance gain. For now, i had 75% idle when downloading
large files with about 80 kb/s, now i have 95% idle :) (Pentium-233MMX)

Martin, thanx for pushing me in the rights direction to have a closer look at
in-kernel pppoe. Works very fine now.


--    Home Powered by: (Net|Open|Free)BSD IRIX NonStop-UX
Solaris AIX HP-UX Tru64 MUNIX Ultrix VMS SINIX Dolphin_Unix OpenStep MacOS A/UX