On Mon, Apr 24, 2000 at 08:24:19AM -0400, Chris Chiappa wrote:
> Without getting into the distro pissing contest, I really doubt there's any
> material difference in "BP6 stability" between distros. Certainly the
> majority of people on this list run Redhat, and I run Debian and we all
> get/have gotten lockups.
Right. It doesn't seem kernel dependant at all (except for IDE driver
capabilities/instabilities), the infamous hang is kernel agnostic it seems.
I'm typing this from my home machine BP6 based and reliable.
I have built last month 3 new BP6 boxen one is plain unreliable.
I built simultaneously an exactly similar machine, who locked up
earlier so I swapped the MB and this one is now reliable it seems.
I have tried half a dozen kernels on the unreliable box without
any success, including the Gentus distro SMP kernel. I have experienced
lockups with a load of 8 when testing the machine with make -j2
kernel compiles, large du, as well as with completely idle machines.
I have also switched rpmfind.net server to a BP6 it runs with an
uniproc kernel due to IDE/SMP problems, the load is around 3000 IDE
interrupts/sec and is up for 18 days now so it doesn't seems to exhibit
the lockup.
There is *something else* I have always switched APCI and other
power saving features off, switched back to SMP 1.1, etc ...
I doubt it's power supply related, rpmfind.net has just a 250W
power-supply, and 8 drives in it, and looks solid.
One interresting post I have seen was that swapping SCSI brand X
to another one Y fixed the reliablility problem for one person. Someone
also posted to the linux-SMP list the locations in the kernel code
near the place where the processors hung for him. Suspiciously one
of the CPUs was running some of the PCI low level code handling.
Another point worth looking at is that recent BP6 have the CPU thermal
sensor plugged as a blue sensor on top of 2 (long) pins. I had to
push them gently in order to be able to insert the CPUs correctly in the
sockets. Do they have to touch the CPU ? Anyway I have also forced
the CPU fans on in the BIOS.
Is there clear example of the lockup being fixed by upgrading the
BIOS ? I have seen a report taht the utility /sbin/hwclock could hang
the machine and that this could be cured by a BIOS update.
Also is there any way to find the BIOS version from a running system
I don't have physical access to the 3 boxen in the US.
Lots of questions, few answers. Maybe just asking for replacement
mobo until a solid one is found is the easiest method to fix this but
this ain't fun when the box is 3000 miles away <grin/>
Daniel
--
[EMAIL PROTECTED] | W3C, INRIA Rhone-Alpes | Today's Bookmarks :
Tel : +33 476 615 257 | 655, avenue de l'Europe | Linux XML libxml WWW
Fax : +33 476 615 207 | 38330 Montbonnot FRANCE | Gnome rpm2html rpmfind
http://www.w3.org/People/all#veillard%40w3.org | RPM badminton Kaffe
--
=- To unsubscribe, email [EMAIL PROTECTED] with the -=
=- body of "unsubscribe linux-abit". -=
