Why is it looking for /root/dev ?

Tom H tomh0665 at gmail.com
Sun May 2 19:37:07 UTC 2010


On Sun, May 2, 2010 at 5:45 AM, Mike Yates <ubuntu at fonehelp.co.uk> wrote:
> On Sun, 2 May 2010 00:29:41 -0400 Tom H wrote:
>> On Sat, May 1, 2010 at 5:18 PM, Mike Yates<ubuntu at fonehelp.co.uk>  wrote:
>>
>>> I've upgraded my 9.10 to 10.04 on-line but it will only work with the
>>> older kernel (2.6.31-21) not 2.6.32-21, which locks up completely while
>>> attempting to start a recovery shell (busybox started but no prompt).
>>> Among the messages above that are three curious ones, failing to find
>>> /root/dev, /root/proc and /root/sys, also "no /sbin/init.007".
>>> As I said, switching to the older kernel, with all the same arguments on
>>> the same filesystem, works perfectly.
>>> Is the initrd corrupt?
>>> I've tried "apt-get install --reinstall linux-image-2.6.32-21-generic"
>>> to no avail.
>>>
>> No idea what "no /sbin/init.007" means but the "/root/..." messages
>> mean to me that your initrd is not mounting your partition(s) and it
>> is therefore failing when it tries to move them at the end of its
>> scripts (init-bottom IIRC).
>>
>> Are you using lvm or mdadm?
>>
>> The "... reiinstall ..." should have had the same effect but you
>> should also try deleting and recreating your initrd with
>> "update-initramfs -d ..." and "update-initramfs -c ...".
>>
> Well, Tom, the recreate of initrd ran OK but the same error occurred.
> Although usually just after Busybox started, it sometimes froze at
> different times
> and in these, other messages were visible, such as "not yet mounted, use
> bootdelay".
> Before trying that grub parameter, I took a good look at /etc/fstab and
> found that the root partition was type ext4 (perhaps slower mounting in
> v2.6.32 ??) and that it was at line 34 of the file because I had sorted
> my many testing drives and partitions into physical order (/ =
> /dev/sdb5) and there were lots of installation comments.
>
> So, I just moved the root-mounting line of /etc/fstab to the top and
> it's fixed !!
>
> I'm still curious about those messages I received. Perhaps there's
> something historical about putting dev under root but I've been using
> Linux since 1993 and I've never seen it.
> The number appended to /sbin/init seems to be random, init.6700 last
> time, so perhaps an "artifact" of "rootless messaging", but a little
> worrying perhaps.

Line 34! I am not sure whether rootdelay is meant for such a case but maybe...

The root in /root/dev, etc is not the final /root.

When the initrd "boots" it mounts /dev, etc and before passing control
to the kernel, these mounts are moved to something like (I would have
to unpack my initrd to check the exact variable) ${root}/dev, etc or
${rootmount}/dev, etc so the /root/dev, etc errors seem somewhat
inaccurate but the reason is clear, ${root} or ${rootmount} (or
whatever the correct invocation is) is not available so /dev, etc
cannot be moved over.

The /sbin/init message is also puzzling because AFAIK init is at /init
not /sbin/init in the initrd unless it is referring to the actual
init...




More information about the ubuntu-users mailing list