Red Hat Bugzilla – Bug 485771
6.0.76 breaks root=/dev/VG/Vol
Last modified: 2009-02-16 15:54:04 EST
Description of problem:
After an update, newly installed kernels stopped booting due to broken
initrd; root volume cannot be found. This regressed from 6.0.75.
Version-Release number of selected component (if applicable):
100%, given the configuration
Steps to Reproduce:
1. Install Fedora with root on a logical volume.
2. Install mkinitrd 6.0.76 or later.
3. Run mkinitrd
initrd-xxxxx.img fails to find root volume
When running mkinitrd, the following messages are seen:
[root@niphredil linux-2.6.29-rc5-tip]# mkinitrd /boot/initrd-2.6.29-rc5-tip.img 2.6.29-rc5-tip
WARNING: /sys/block/dm-0 is a not a block sysfs path, skipping
WARNING: /sys/block/dm-1 is a not a block sysfs path, skipping
Pay no mind to the self-built kernel, it's exactly the same with any
kernel (first noticed with kernel-2.6.29-0.112.rc4.git3.fc11.x86_64).
The root argument:
[zaitcev@niphredil tmp]$ cat /proc/cmdline
ro root=/dev/N1/Fedora selinux=0 libusual.bias="ub"
Forcing 6.0.75 with rpm -U --oldpackage fixes the problem.
* Wed Feb 04 2009 Hans de Goede <firstname.lastname@example.org> - 6.0.76-1
- Fix configuration of network interfaces for network boot (#481078)
- Fix iscsi chap password being seen as ********
- Handle lv root specified as /dev/dm-X properly (#471729)
- Initial commit of EXTLINUX (no menu) support
- Do not call dm_resolve_name on dmraidsets names, sometimes it fails, and
it is not necessary, this fixes the dmraid boot failure seen on
some systems (#476818)
- Make nash mount support relatime (#296361)
- FIX: nash unable to find dm devs by uuid or label (#480667)
At the face of it, it's either bug 471729 or bug 480667.
Also! The bug 483423 has similar symptoms, but it's for a different
release and seems to occur in F10 GA. So, probably not the same.
Please don't dup until Hans reproduces this bug and finds the issue.
Are you on irc? Anyways can you do an ls -l /sys/block ?
What you see there should be symlinks, unless you've got "deprecated sysfs" (or something similar) enabled in your kernel config, in which case yes the new mkinitrd will break.
I think this is your problem, because mkinitrd does a readlink, so normally the message should never show the /sys/block/dm-0, unless that is not a symlink, which means you're using the deprecated sysfs layout option.
p.s. I'm hansg on both freenode and RHAT irc
Thanks a lot, I'll verify and close as NOTABUG. Maybe I tried to run
yum update while under a kernel with obsolete layout.