[960] in linux-net channel archive

home help back first fref pref prev next nref lref last post

Re: Kernel Panic with 1.2.13 - bug in tcp_rcv()?

daemon@ATHENA.MIT.EDU (Paul Gortmaker)
Tue Aug 22 23:23:24 1995

From: Paul Gortmaker <gpg109@rsphy1.anu.edu.au>
To: dhollis@gp.magick.net (Daniel Hollis)
Date: Tue, 22 Aug 1995 15:14:24 +1000 (EST)
Cc: linux-kernel@vger.rutgers.edu, linux-net@vger.rutgers.edu
In-Reply-To: <m0sjyrn-000xACC@gp.magick.net> from "Daniel Hollis" at Aug 19, 95 06:07:59 pm

> 
> Latest kernel panic under 1.2.13. Keyboard VC switching still worked but 
> otherwise Linux was dead. Nothing in the logs to indicate any kind of 
> problem. The panic didn't get syslogged, so I had to write down the 
> screen and type it in by hand. Hope someone appreciates it!
> 
> Hardware:
> 486DX4/100, 40mb RAM
> 3c509 0x360 IRQ 5

Dan, you have reported about 8 crashes in the last month or two
with this same machine (40MB DX4/100). You originally had random crashes 
with a wd8013 clone, and then as well with a ne2000 clone, so you blamed the 
8390 drivers, as seen below.

>Apparently, any 8390 based card is not going to be stable with Linux. So
>here's the big question:
>
>Which card is >>>the most stable<<< under Linux? With the 8390 drivers in
>the state they are, that disqualifies any 80x3 and NEx000 card. My guess is
>it's some 3COM 3509 or other card?

[...]

>We'll may be trying an SMC Ultra card if we can get our hands on one, and 
>see if that makes any difference. Will the SMC prove anything or should we 
>try some other card to rule out our hardware and definitively point the 
>finger at the Linux networking code? What do we have to do to satisfy the 
>kernel developers that it's not our hardware? What card will prove it?

So then you moved onto a 3c509 card (as seen above) and now you still get
random crashes. (You have already reported 2 for the 3c509 so far.) You are 
right in assuming that not all three cards can be broken, but you then 
overlook the possibility of a flakey motherboard, and hence blame linux.

I suggest that you use more conservative settings in the CMOS
setup (if possible) -- things like adding ISA i/o wait states, etc.
If this is not possible,(or it doesn't help) then I would recommend you 
grab another motherboard, it would probably be cheaper than the two extra 
ethernet cards that you have bought.

Until you do this, I don't think you will find anyone who is going to take
too many of your crash reports very seriously anymore.

Good luck with it.
Paul.

home help back first fref pref prev next nref lref last post