----- Original Message -----
From: "Lennert Buytenhek" <[EMAIL PROTECTED]>
To: "Oleg" <[EMAIL PROTECTED]>
Cc: "Bart De Schuymer" <[EMAIL PROTECTED]>;
<[EMAIL PROTECTED]>
Sent: Saturday, January 26, 2002 2:33 PM
Subject: Re: [Bridge] 2.4.17 kernel panic...


> Hi!


Hello,

> I still haven't been able to reproduce it :(  Do you still have the
ability
> to test some things?  If so, can you try the following (incremental)
> patches from http://bridge.sf.net/patchtracker-nf.html ?


My problem is solved, at least the biggest problem. I was going to wait
until I bought some new nics to test as well before I post this, but seems
like it's taking a while for me to buy them ;)
This will happen when bughunting ebtables on a sparc is finished.

My machine was overclocked (forgot it was) and the three nics were too much
I guess. It's now underclocked and I don't get any panics anymore with those
ping -s 20000 and above.
More details:
With overclocking I could crash the bridge by sending a ping -s 40000 from a
box on the one side to a box on the other side of the bridge. Now I can do a
ping -s 60000 with no packet losses like that. The bridge also crashed when
doing those pings from the bridge box itself.
But if I now do a ping -s 20000 from the bridge machine to another box I
still get packet losses. But, no kernel panics...
I did a ping -s 60000 and let it wait a while, this message showed up on the
console:
>From destination_box: Frag reassembly time exceeded.
ping -s 20000 gives 50 % packet loss, with the destination box giving the
above message for the "lost" packets.
Hmm, even ping -s 2000 gives this message (and so, packet loss) for some
echo requests.
The bridge box is running on 300 Mhz now (underclocked, it's a 500 Mhz box).
Maybe this is a performance problem?
The motherboard of this machine is a piece of sh*t (packard bell comp *-*).

> 00_brnf_post_routing_paranoia.diff
> 01_brnf_clean_up_protocol_checks.diff

Your 01 diff is vs a file called br_netfilter.c2

> They remove some potentially unsafe assumptions from the code.  If these
> patches fix it, could you determine which of the two actually did the
> trick?

cheers,
Bart

_______________________________________________
Bridge mailing list
[EMAIL PROTECTED]
http://www.math.leidenuniv.nl/mailman/listinfo/bridge

Reply via email to