Kernel 2.2.3/4/5 Alpha > 1 network card crash

Matthew G. Marsh (mgm@paktronix.com)
Sun, 4 Apr 1999 15:34:36 -0500 (CDT)


Well - I have been fighting this one for a few weeks now and I (kind of)
have it narrowed down. First off:
Alpha Avanti (Alphastation 200/166) - I have 7 of these and the first
three have done the exact same thing - have not started on the other 4.

Problem Scenario - Repeatable every time:

1) System is installed from RedHat 5.2 Alpha. Runs fine.
2) Updated all RPMS for use with 2.2 kernels. Runs fine.
Also tried compiling net-tools 1.51 for 2.2 case
3) Compile new 2.0.35 kernel for 2 Ethernet cards
1st card is internal DEC chip card
2nd card - I have tried PCI PCNet32, 3COM 3c509, 3COM 3c905 PCI,
Intel EEpro 100 PCI, Intel Epro10 ISA.
Kernel 2.0.35 work fine with both ethernets running and routing
4) Compile 2.2.(3,4,5) with exact same type of setup
Make a new config (no reuseing 2.0 config)
Same network driver options
5) Kernel 2.2.(345) works fine WITH ONE ETHERNET
Even if the second ethernet driver is installed (no ifconfig'ed)

MACHINE CRASHES back to SRM console with no OOPS or LOGS as soon as the
second ethernet card is IFCONFIG'ed

IE: ifconfig eth0 works fine pings routes etc...
ifconfig eth1 and about 5 second after hitting enter
BOOM - machine is at SRM prompt.

If I allow the network init script to run then the machine hangs hard at
the ifconfig of the second ethernet interface - must power off.

This whole sequence is repeatable. Rebooting the machine to 2.0.3(56) and
both ethernets work great.

I have tried:

Alexey's IP commands to IP ADDRESS the cards instead of IFCONFIG
Using 255.255.255.255 netmasks to try and turn off the auto route
(which is what I am suspecting)
Using several different ISA & PCI cards
None of the above change the sequence at all.

I am suspecting the auto-route or similar code paths because sometimes
after the IFCONFIG of the eth1 you can immediately run ifconfig alone and
see that the card is addressed but running route -n or ip route in this
brief time frame has no routes through that card. It is almost as though
once the route (or some such ip path) code starts to see the eth1 then the
machine crashes.

Oh - I can also config JUST eth1 and it works fine - then when I add eth0
- BOOM. So it is definately tied to the "second" real interface. And LO is
running in all of these tests.

Any CLUES - QUESTIONS - COMMENTS ?

Please note I am not currently subscribed to Axp-list but am on
Linux-Kernel. Thanks!

--------------------------------------------------
Matthew G. Marsh, President
Paktronix Systems LLC
1506 North 59th Street
Omaha NE 68104
Phone: (402) 932-7250
Email: mgm@paktronix.com
WWW: http://www.paktronix.com
--------------------------------------------------

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/