RE: 3w_9xxx drive/lenny kernel/AMD64 Problem/Solved
Thanks for everyone's help.
After capturing the bootup log for different kernels using a serial console,
it was clear that the AMD64 2.6.22 and 2.6.24 kernels were choking while
allocating resources for the pci busses. One of the symptoms (MAC address
of FF:FF ... for one of the internet controllers) match one fixed by a bios
update. My motherboard bios was 1.01 with the most recent one being 1.05.
Updating to the most recent motherboard bios, fixed the problem.
Some AMD86 related change between kernels 2.6.20 and 2.6.22 interacted badly
with the 1.01 bios on the Thunder K8WE (S2895) board. It seems to be ACPI
related. It many effected the PCI-100 and PCI-X 64bit connectors plus one
of the internet controllers.
Nick
-----Original Message-----
From: nicholas materer [mailto:materer@okstate.edu]
Sent: Thursday, July 03, 2008 6:15 PM
To: 'Peter Yorke'; debian-amd64@lists.debian.org
Subject: RE: 3w_9xxx drive/lenny kernel/AMD64 Problem
It looks like my problem is AMD64 specific. The
debian-LennyBeta2-i386-netinst.iso install image recognizes the card
correctly. However, the debian-LennyBeta2-amd64-netinst.iso install image
has the same problems that I have with booting my system after upgrading the
kernel to the 2.6.24 or 2.6.20 version.
In addition, I do not think that this problem is due to the 3w_9xxx driver.
I updated to the firmware on my card to the latest version with no success.
I complied latest the driver from the 3ware web site - some problem. I also
complied the driver from the 2.6.18 kernel source (had to make two trivial
patches due to a change in the scatterlist structure). They all failed with
the same errors - PCI Parity Error, PCI Abort and Controller Queue Error.
These error are printed from the routine twa_decode_bits in the driver -
"This function will print readable messages from status register errors".
I guess I have an AMD64 specific regression affecting the PCI-X buss on my
Thunder K8WE board. I guess that I need to boot off a non-raid boot disk
and collect the system log with both a i386 and an AMD64 system.
Nick
Reply to: