EEPro100 problems in SMP on 2.4.5 ?

Hi. While doing some file tranfers to our new server (a Compaq Proliant
8way XEON 500 with 4gb ram and an EEPro100 NIC), the box socked solid (no oops, no response via network, no response via console). The other hardware in the system was a Compaq Smart Array 9SMART2 driver). It's running Slackware 7.1. The other system was a dual P3 450 running Redhat 7.1 (Linux velocity.kuro5hin.org 2.4.2-2smp #1 SMP Sun Apr 8 20:21:34 EDT 2001 i686 unknown) w/ 3c59x NIC. The Redhat machine experienced no problems.
In Uni processor mode, the system is totally stable. But only using 1/8th
of its power :-/ We had to roll back to 2.2.19 with a bigmem patch, but we'd like to have a stable 2.4 kernel to use (since it's so much better SMP wise, throughput wise, etc).
--
www.kuro5hin.org -- technology and culture, from the trenches.

Some things to try:

1: Include `magic sysrq' support in the kernel and use ALT-SYSRQ-T and S
when it has locked up. If you get some traces then please feed them
into `ksymoops -m System.map' and report back.

2: If the above doesn't work, add `nmi_watchdog=1' to the kernel boot
options. That may catch the lockup.

3: Replace the NIC with another eepro100. If the problem goes away then
chuck the old one.

4: Replace the NIC with one of a different type (ie: swap with the other
machine). If that fixes it we look at the ethernet driver. Otherwise we look at, umm, the rest of the kernel.

-

I have a lock problem on a dual P3 1GHz w/1GB RAM setup and 2.4.5, but it doesn't seem to be NIC related although I have two EEPro100's in it.

I get lockups while doing large disk reads that use larges amounts of memory (MySQL's myisamchk on a 600MB MyISAM table, for instance). The SCSI subsystem is an Adaptec Ultra160 card w/Ultra160 drives.

Since the system locks even when doing a fsck on one of the badly damaged drives (from the hard lock) during bootup, I am pretty sure this is an isolated problem.

In SMP mode, when the system does manage to boot, it will load up MySQL and run fine for about 10 minutes. Afterwards, the system hardlocks. In another instance, shutting down MySQL right after starting it also caused the system to hardlock in SMP mode. I thought it might have something to do with the bounce buffers, so I disabled the 4GB bigmem area, but that didn't change the situation.

The chipset is a VIA 686B Southbridge and VIA 694DP Northbridge, though I don't know whether this would affect the hard locks on the SCSI subsystem (SysRQ unresponsive).

It's rock solid with a uniprocessor kernel compiled with the exact same configuration as the SMP kernel (minus the SMP switch, of course (APIC+IO-APIC+4GBhighmem enabled. I've tried various kernels from 2.4.3 to 2.4.5-ac21).

Have you checked to see if it's the SCSI subsystem that is causing your locks?

--
Vibol Hou
KhmerConnection, http://khmer.cc
"Stay Connected."

-----Original Message-----
From: [email protected] [mailto:[email protected]]On Behalf Of Dylan Griffiths
Sent: Friday, June 29, 2001 8:42 PM
To: Linux kernel
Subject: EEPro100 problems in SMP on 2.4.5 ?

Hi. While doing some file tranfers to our new server (a Compaq Proliant
8way XEON 500 with 4gb ram and an EEPro100 NIC), the box socked solid (no oops, no response via network, no response via console). The other hardware in the system was a Compaq Smart Array 9SMART2 driver). It's running Slackware 7.1. The other system was a dual P3 450 running Redhat 7.1 (Linux velocity.kuro5hin.org 2.4.2-2smp #1 SMP Sun Apr 8 20:21:34 EDT 2001 i686 unknown) w/ 3c59x NIC. The Redhat machine experienced no problems.
In Uni processor mode, the system is totally stable. But only using 1/8th
of its power :-/ We had to roll back to 2.2.19 with a bigmem patch, but we'd like to have a stable 2.4 kernel to use (since it's so much better SMP wise, throughput wise, etc).
--
www.kuro5hin.org -- technology and culture, from the trenches.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

That was locked as well, AFAIK.

I'd love to do some of this, but since the box is now being shipped to a colo facility in New York, I don't really have a choice in the matter.

Hopefully someone here doing SMP + EEPro100 can see if they can reproduce the issue (2.4.5 kernel).

--
www.kuro5hin.org -- technology and culture, from the trenches.

I've had issues with the Intel cards, as well.

What revision of the card is it?

Have you tried the drivers available from Intel, to see if they do a better job? In my case, they didn't.

I've also had reports, for a linux-2.2.x kernel, that sometimes its guesswork as to whether stock kernel eepro100, the intel e100 driver, or Don Becker's eepro100 will work on the beasts.

HTH.