Re: woody/linux 2.4/raid/athlon/almost a horror story
On Fri, Jun 29, 2001 at 11:58:45PM -0700, aphro@portal.aphroland.org wrote:
> what a horrible experience.
>
> firstly, i compiled my 2.4.5 kernel with "Athlon/Duron'
> support, and was promptly greeted with "Illegal Instruction"
> on e2fsck when i tried to boot the system. Kind of
> odd considering i have an athlon 1309mhz proc ..
1309 is odd as well, and a typo probably, too. :-)
It might be memory/bus/cache corruption problems. Linux 2.4 may just
be hammering your hardware harder than 2.2 did.
If you can clock your system busses down, stretch memory timings and
reduce the cpu speed, you might want to try that and see if you can boot
with "init=/bin/sh". That way, it will not attempt to fsck and mount
your filesystems readwriteable, just boot the kernel and give you a shell.
Type "reboot" or apply vulcan pinch to exit.
Alternatively, prepare a memtest86 bootdisk and boot from it. You may
need to leave it running for hours and hours. If you are daring,
you may try to overclock your busses, memory and cpu just a little
and run memtest again. Too bad that memtest cannot generate heavy io
on peripheral devices while running its tests. Heavy device activity
can severely affect power supply quality, reflecting in turn on the
reliability of the mainboard chipset and busses.
> anyways i booted back to 2.2.19(good kernel..good
> kernel..) and recompiled for PPRO instead of athlon.
> this time it got farther, and begun to scan my
That sounds a lot like the optimized-for-athlon memcopy instructions
are asking too much from your system. Ppro optimizations are probably
less aggressive, but I would not trust critical data to it either, if
the athlon codes result in clear and apparent integrity problems.
> raid(!) arrays. since the arrays were stopped clean
> i didn't understand why it was doing this. i have
> /usr and /var on raid1 and i have /space on raid0.
Uh-oh. Raid code in 2.4 makes extensive use of cpu-specific optimization
tricks, iirc.
> this of course leaves a horrible taste in my
> mouth. im just glad 2.4 didn't trash my drives.
> even if it did i'd only lost a week of data
> luckily i decided to stop trusting the local
> machine due to an asus motherboard frying my
> filesystems a couple weeks ago(have since replaced
> it) and now save most of my data directly to
> NFS shares on my main server.
You already fried a motherboard in this machine? I'm not really
surprised even, noticing the list below. Consider replacing your power
supply too. A faulty one can wreak havoc, while you are suspecting all
the devices that break down due to the out of spec supply. Even the
best manufacturers have monday mornings, too.
> ohwell....i guess this experience just makes me
> more glad that linux 2.2 is solid.
It certainly had a lot more field testing than 2.4
> my system:
> woody(1 week old)
> athlon 1300mhz
> 768mb sdram
> tyan KT133A-based motherboard(forgot the part#)
> dual IBM 20GB ide drives
> promise ata100 controller(non raid)
> adaptec aha2940UW scsi
> 4x8x CD-R (SCSI)
> Nvidia Geforce2 MX 200 (64MB) AGP
> Intel eepro100
> Soundblaster PCI 128
> Hauppauge WinTV Go!(bt848 i believe)
> 450watt PC Power & Cooling power supply
> ~650watt (1100VA) UPS
> Linux 2.2.19(ide patch, nvidia kernel drivers,
> bttv drivers)
Are you in California?
Cheers,
Joost
Reply to: