Red Hat Bugzilla – Bug 133684
unable to open an initial console - ppc only
Last modified: 2007-11-30 17:10:50 EST
Description of problem: "unable to open an initial console" is
displayed, then a few SELinux context messages, then it just hangs
there, and booting does not go further. This is similar to Bug #132862
however it _only_ affects PowerPC with the 2509 tree - I did an x86
install, and that went perfectly
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Perform fresh install on iMac DV SE (powerpc)
See error message as in Bug #132862, and the booting hangs there
/sbin/start_udev runs, device nodes get created, and I can boot again
When I booted into rescue mode, chrooted to /mnt/sysimage, executing
/sbin/start_udev meant things worked well for me (I got all the
devices in /dev that I needed). This is how I ran yabootconfig, and so on.
But for some reason, during a normal bootup, its not being picked up
in a timely fashion. Maybe there's some rc scripts that need changing,
I haven't poked at this issue yet.
Also, comment #5 in bug #132862 was tried, but that wasn't the cause
of the problem. This wasn't an upgrade to begin with, but a fresh
which kernel version?
Latest as of Rawhide then
Reproduced with today's rawhide on peach.cambridge.redhat.com
Can reproduce with the .603 kernel, on a fresh rawhide latest install,
from the 0910 tree.
Trying dwmw2's suggestion on irc to mknod /mnt/sysimage/dev/console c
5 1 makes it boot further (yay!), but more errors are thrown, as per
the graphic attachment
Created attachment 104977 [details]
shot of the breakage
after adding /dev/console manually. Otherwise, there is just the error that
/dev/console isn't found as per the bug report, and it doesn't boot further
Disable the bloody stupid mount-by-label. Use device names in your
fstab and on your kernel command line.
Mount by label is never going to help you much on a desktop-class
system. It's only really appropriate on big servers where you're
likely to be moving discs around. On a desktop-class system it'll only
offer one more likely point of failure, like this.
this all looks like /dev is not a mounted tmpfs, which should happen
in initramfs, produced by mkinitrd.
I've gotten this, too. The reason (and thus the fix) was that
yaboot.conf did not contain an "initrd=" line. Adding that makes it
work (rawhide as of 24 Oct. 2004)
Ralf, in comment #10, fresh install on the 28th Oct on iMac DV SE,
yaboot.conf had an initrd line (i.e. the default one now has it too -
I remember adding it before), and it still broke !
Though I should add, /dev/console is being created. When logged into
"linux rescue" mode, I can see quite the few entries in
/mnt/sysimage/dev (~195 or so). So why it still complains, I'm wondering
SELinux is enabled, just in case that makes a difference (selinux=0
when booting didn't seem to)
Rawhide latest (okay, its just FC3), and the install went fine, got
the usualy sillyness for unable to find inital console, even with the
initrd in /etc/yaboot.conf. Tried with the sysmap= option too
Following Harald's udev page to actually start without an initrd, I
did that, rebooted with selinux=0, and the mac's happy
So for some reason, yaboot isn't recognising the initrd properly ? I
guess its becoming less of a udev "bug", and might only seem prevalent
in this iMac DV SE (i'll fiddle w/it on an iBook later)
Okay, I opened this bug, and its fixed in the testing tree that we
released as FC-3 that sits on fedoraproject.org so I'm going to close
this as currentrelease as I don't think it bothers anyone, any longer.
yaboot boots fine when an intird= is specified.