Red Hat Bugzilla – Bug 123994
Missing modules cause installation to fail
Last modified: 2007-11-30 17:10:43 EST
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.6) Gecko/20040126
Description of problem:
See bug #123982 for the premises and the preceding problem.
On a machine where /etc/modeprobe.conf contained the lines
alias ieee1394_controller ohci1394
alias scsi_hostadapter sbp2
the system became unbootable after an upgrade from FC2beta3 to FC2.
A series of problems combined to cause this:
- The modules in question are not part of the shipping kernel.
- Not finding the modules, the kernel %post script failed and
exited without finishing the kernel installation (no initrd
and no grub/menu.lst modification).
- Anaconda did not catch the error and removed (or had already
removed) the previous kernel.
Commenting out the offending lines in modeprobe.conf and trying to
reinstall the kernel from a rescue environment failed as well. Only
when the two lines were completely deleted from modeprobe.conf did the
kernel installation succeed.
Version-Release number of selected component (if applicable):
2.6.5-1.358 and several earlier
Steps to Reproduce:
1. Install drivers that are not part of the shipping kernel.
2. Try to upgrade the system.
Actual Results: Useless system.
Expected Results: A system that can boot:
The kernel %post script should ignore modules that it can't find and
it should create an initrd with the rest, rather than just fail.
The kernel %post script should ignore comments in modeprobe.conf.
Anaconda should be smarter than to just accept a kernel installation
failure and deliver an unbootable system.
The ohci1394 and sbp2 modules should be part of the shipping kernel
too, now that a kazillion other yet less common drivers are. These two
are part of the source anyway, so why not build them?
Note that this will happen with *any* non-default drivers, not only
with the ones that happened to be on this system.
If I'm not completely mistaken, the kernel %post problem affects RHEL too.
Of course, where I say "the kernel %post script should", I should
really say "mkinitrd should". The problem cascades all the way down...
Put something in CVS that should help here, although I'm not convinced
I fully like it.
Marking resolved as per comment #2. Assuming that Jeremy eventually came to
like, perhaps even love, his fix.