I've got two virtually identical PCs. Both are PentiumPro Digital Celebris GL 6180s. Both run SuSE 7.1; one as a server and one as a workstation I'm putting together for my daughter. The workstation install is brand new as of Saturday but the server's been in duty for a several months. I've updated the packages on both and that's about all. Both run reiser file systems and both have 100MB reiser /boot partitions. Neither have any exotic hardware. The only differences worth noting are:

* The servers got 64MB RAM and the workstation's got 104MB
* The workstation is running more packages than the server
* I recently started using pd-ksh on the server but still have bash on the workstation

Saturday evening, following SuSE's instructions, I successfully upgraded the workstation's kernel from 2.4. to 2.7 via RPM (rpm -Uvh --nodeps --force kernel.rpm). Afterward, I added resierfs module support by editing /etc/rc.config and running mk_initrd. After modifying lilo, the new kernel booted perfectly.

I wasn't so lucky with the server. Following the same steps, I got the following error after running mk_initrd:

creating initrd "//boot/initrd" for kernel "//boot/vmlinuz" (2.4.7-4GB)
no static shell

When I try to boot with the kernel, I get an error indicating the reiser file systems cannot be mounted:

RAMDISK: Compressed image found at block 0
Uncompressing ......... done.
Freeing initrd memory: 438k freed
VFS: Mounted root (ext2 filesystem)
Loading module usbcore ...
insmod: usbcore: no module by that name found
fatfs: bogus logical sector size 0
fatfs: bogus logical sector size 0
kernel panic: VFS: Unable to mount root fs on 03:08

I ran this from scratch several times (after blowing away the new kernel) and got the same error each time. I strongly suspect my korn shell failed to run mk_initrd properly but am not sure why. I tinkered with running mk_initrd with bash to no avail.

I guess the question is whether anyone is familiar enough with mk_initrd, bash and pd-ksh to know whether or not I can expect this to work under pd-ksh or knows what else might be wrong. I use the korn shell at work and would really love to continue using it here at home but am now unsure whether or not I'll run into other problems down the line.

Comments? Thanks.

Quote:

I guess the question is whether anyone is familiar enough with mk_initrd, bash and pd-ksh to know whether or not I can expect this to work under pd-ksh or knows what else might be wrong. I use the korn shell at work and would really love to continue using it here at home but am now unsure whether or not I'll run into other problems down the line. Comments? Thanks.

None from any actual experience, but it's pretty much a bash world when you talk about linux.

As for your problem, it sounds like copying the kernel from the working PC to the other should fix things...


John

Good idea, John. I'll copy the kernel - should work fine. Meanwhile, I'd really like to understand what's behind this problem. There's squat about it on the web (that I've found, anyway). Would still like to hear from anyone that's got any ideas.

Thanks

Did not work. Got precisely as far as before which confirms that the resierfs module is not being passed to the kernel.

Will try loading via lilo...

FWIW, I did a string search in initrd for "reiser" and it popped up in there. You could try copying that over too, but eventually you might as well copy over all of /boot...


John

I'll cp the workstations initrd over under a different name and diff the two files to see what's up. If it looks promising, I've give that a shot. Got nothing to lose since the box is still running on the 2.4.0 kernel.

Worse come to worse, I'll compile the server's kernel from scratch (2.4.14!!!) and be done with it.

The binary rpm upgrade appeals to me because it's something I can easily teach my daughter to do. I'd never gone that route before and was curious about how rpm would handle it.

Thanks for hanging in there, John.