Problems with kernel built from linux-source-2.6.16_2.6.16-3_all.deb
I've been following the debian-kernel email since the release of the
2.6.16 kernel, as well as searched the unofficial archives, for the
cause of the basic problem, and have found no solutions.
My platform is a Dell PowerEdge 2500, dual CPU, 0.5 GB RAM, 5 SCSI hard
disks on the integrated controller channel A, none on channel B; two
AHA-2944 controllers; one AHA-2940 controller with one SCSI disk. Boot
partition is /dev/sda1 and there are no IDE hard disks in the system.
The CD-ROM is IDE. I've been running Debian 3.1r0a, with kernel 2.6.8
without problems. The reasons for trying a newer kernel are legion but
not relevant to this (already too long) discussion.
Initially, I attempted to work with the 2.6.16 source from kernel.org,
and, when I had problems with it, followed with the download of the deb
(today), to get any possible Debian specific changes that might
(hopefully) resolve my issue.
The kernels I've built all panic when run. The first builds, using
kernel.org source, began with a VFS error: "can't find ext3 filesystem
on dev sda1"; from cramfs "wrong magic"; umount "devfs not mounted" (not
an issue, I think); pivot_root "No such file or directory"; /sbin/init
"cannot open dev/console: no such file" and the kernel panic.
I believed the VFS error regarding the ext3 filesystem had something to
do with the HBA scan order. The 2.6.8 Debian kernel sees the
controllers in the same order as the BIOS does (integrated A, B, 2944,
2940, 2944), the kernel.org derived kernel sees them in reverse. So I
munged around with the grub boot loader files, thinking this might be
part of the problem. And it did get rid of the VFS error. But, the
failure of init to find /dev/console continued. References found by
googling indicated that this could be caused by missing dev entries
needed in early stages of the boot process, so I booted on a rescue CD,
mounted sda1 and checked it. Too many files to count in /dev, and
/dev/console was there. I've tried changing parameters in the kernel
configuration, leaving things in I'd originally taken out, and so on.
I've lost count of the number of builds/boot-failed/reboot-stock I've
gone through. No luck in any of them.
That's when I decided to get the Debianized source and see if that would
work. No joy, same problem. Only this time, unlike all the others, the
stock 2.6.8 kernel is now also panicing, with the same problem of not
Thanks in advance for your patience in reading this and, hopefully, help
getting things back together.