8:52 a.m.
Hi,
I made an update from FC2 to FC3, and now, my machine will no more boot:
The boot stops when executing the init scripts:
having problems with the /dev directory (cannot open /dev//null and/or
/dev//console). I'm not using the FC3 kernel, but the vanilla kernel
2.6.9. When using the FC3 kernel, all is OK. I think, it has
to do with udev (in my vanilla kernel, hotplug is configered
correctly!).
Regards
Joachim Backes
--
Joachim Backes <backes(a)rhrk.uni-kl.de>
University of Kaiserslautern,Computer Center,
High Performance Computing,
D-67653 Kaiserslautern, PO Box 3049, Germany
--------------------------------------------------
Phone: +49-631-205-2438, FAX: +49-631-205-3056
http://hlrwm.rhrk.uni-kl.de/home/staff/backes.html
8:55 a.m.
On Wed, 2004-11-10 at 15:52 +0100, Joachim Backes wrote:
Hi,
I made an update from FC2 to FC3, and now, my machine will no more boot:
The boot stops when executing the init scripts:
having problems with the /dev directory (cannot open /dev//null and/or
/dev//console). I'm not using the FC3 kernel, but the vanilla kernel
2.6.9. When using the FC3 kernel, all is OK. I think, it has
to do with udev (in my vanilla kernel, hotplug is configered
correctly!).
did you also make an initrd?
(if you used "make install" in the kernel tree one should have been
created for you automatically, as well as it being added to grub
automatically)
12:34 p.m.
On Wed, 2004-11-10 at 15:52 +0100, Joachim Backes wrote:
>Hi,
>
>I made an update from FC2 to FC3, and now, my machine will no more boot:
>The boot stops when executing the init scripts:
>having problems with the /dev directory (cannot open /dev//null and/or
>/dev//console). I'm not using the FC3 kernel, but the vanilla kernel
>2.6.9. When using the FC3 kernel, all is OK. I think, it has
>to do with udev (in my vanilla kernel, hotplug is configered
>correctly!).
I've seen this as well- have been unable to get FC3* to run
under a vanilla kernel.org kernel. When i disable the 'quiet' flag
in grub, the boot process aborts with something like 'unable to open
/dev/console'.
I never reported it because I thought i was missing something
obvious- i guess I'm not the only one.
Arjan van de Ven wrote:
did you also make an initrd?
(if you used "make install" in the kernel tree one should have been
created for you automatically, as well as it being added to grub
automatically)
I know I did when I tried it, except i run mkinitrd manually.
Could this have something to do with udev? I noticed that if
I enabled the deprecated /dev FS, it broke in a slightly different
manner. I suspect the problem is the FC3 environment expects a
magic combination of kernel options- I just have not been able to
stumble upon them yet...
ron
12:37 p.m.
On Wed, 2004-11-10 at 12:34 -0600, RON FLORY wrote:
I know I did when I tried it, except i run mkinitrd manually.
Could this have something to do with udev? I noticed that if
I enabled the deprecated /dev FS, it broke in a slightly different
manner. I suspect the problem is the FC3 environment expects a
magic combination of kernel options- I just have not been able to
stumble upon them yet...
hmm you need tmpfs enabled but beyond that... can't really think of
anything.
1:38 p.m.
On Wed, 2004-11-10 at 12:34 -0600, RON FLORY wrote:
> I know I did when I tried it, except i run mkinitrd manually.
>
> Could this have something to do with udev? I noticed that if
>I enabled the deprecated /dev FS, it broke in a slightly different
>manner. I suspect the problem is the FC3 environment expects a
>magic combination of kernel options- I just have not been able to
>stumble upon them yet...
Arjan van de Ven wrote:
hmm you need tmpfs enabled but beyond that... can't really think
of
anything.
yeah, I just tried it again. Looking at .config, tmpfs is
definitely enabled, but the boot process aborts with:
"Warning: unable to open an initial console."
and thats the last thing it does.
I've tried this on perhaps 3 different makes/models of
x86 systems, all with same results. In each case 'console'
is either local laptop or generic x86 HW. I know this will
be simple when it finally makes sense...
ron
P.S. Also, what has to be enabled in the kernel to support
the grub 'root=LABEL=/' directive? I always have to
supply literal 'root=/dev/hdax' with kernel.org kernels.
Thanks.
ron
2:34 p.m.
On Wed, 2004-11-10 at 13:38 -0600, RON FLORY wrote:
> P.S. Also, what has to be enabled in the kernel to support
> the grub 'root=LABEL=/' directive? I always have to
Arjan van de Ven wrote:
nothing. That is purely done and used by the initrd.
are you *sure* you are properly using an initrd ???
using 2.6.9 from kernel.org:
while in /usr/src/linux-2.6.9:
mkinitrd -f -v /boot/initrd-2.6.9.img 2.6.9 ??
is this bogus?
ron
P.S. as a side note, i remember the LABEL=/ directive not
working with kernel.org 2.4 kernels since RH8, and i
never really *had* to bother with building initrds then.
I just assumed some RH proprietary stuff was not present
and didn't worry too much about it.
ron
2:47 p.m.
using 2.6.9 from kernel.org:
while in /usr/src/linux-2.6.9:
mkinitrd -f -v /boot/initrd-2.6.9.img 2.6.9 ??
is this bogus?
no but you have to tell your bootloader about it as well otherwise it
just goes unused.
P.S. as a side note, i remember the LABEL=/ directive not
working with kernel.org 2.4 kernels since RH8, and i
never really *had* to bother with building initrds then.
I just assumed some RH proprietary stuff was not present
and didn't worry too much about it.
well that's the wrong assumption; LABEL= has never been a kernel side
thing...
3:29 p.m.
> using 2.6.9 from kernel.org:
>
> while in /usr/src/linux-2.6.9:
>
> mkinitrd -f -v /boot/initrd-2.6.9.img 2.6.9 ??
>
> is this bogus?
Arjan van de Ven wrote:
no but you have to tell your bootloader about it as well otherwise
it
just goes unused.
Like?
title Test (2.6.9)
root (hd0,1)
kernel /vmlinuz-2.6.9 ro root=/dev/hda4 rhgb
initrd /initrd-2.6.9.img
except, substituting 'root=LABEL=/' for 'root=LABEL=/' keeps
the bootloader from finding an initrd. As it is now, it just can't
find a console ;)
Yet, the below target works when the FC3 kernel source is built
and prepared with mkinitrd:
title Fedora Core (2.6.9-prep)
root (hd0,1)
kernel /vmlinuz-2.6.9-prep ro root=LABEL=/ rhgb
initrd /initrd-2.6.9-prep.img
> P.S. as a side note, i remember the LABEL=/ directive not
> working with kernel.org 2.4 kernels since RH8, and i
> never really *had* to bother with building initrds then.
> I just assumed some RH proprietary stuff was not present
> and didn't worry too much about it.
well that's the wrong assumption; LABEL= has never been a kernel side
thing...
OK, I believe you- it just an 'unknown' i can work around without
too much trouble. I am a bit curious why it behaves differently tho.
ron
2:07 a.m.
On Wed, 2004-11-10 at 15:29 -0600, RON FLORY wrote:
title Test (2.6.9)
root (hd0,1)
kernel /vmlinuz-2.6.9 ro root=/dev/hda4 rhgb
initrd /initrd-2.6.9.img
except, substituting 'root=LABEL=/' for 'root=LABEL=/' keeps
the bootloader from finding an initrd. As it is now, it just can't
find a console ;)
odd.. makes me wonder if you disabled initrd support in the kernel
config...
Do you see the nash banner printed at all on the screen ?
(which is about the first thing the initrd will do)
8:10 a.m.
New subject: FC3 does not boot with an own vanilla kernel ( *SOLVED* )
On Wed, 2004-11-10 at 15:29 -0600, RON FLORY wrote:
>title Test (2.6.9)
> root (hd0,1)
> kernel /vmlinuz-2.6.9 ro root=/dev/hda4 rhgb
> initrd /initrd-2.6.9.img
>
> except, substituting 'root=LABEL=/' for 'root=LABEL=/' keeps
>the bootloader from finding an initrd. As it is now, it just can't
>find a console ;)
Arjan van de Ven wrote:
odd.. makes me wonder if you disabled initrd support in the kernel
config...
BINGO- although its not quite that simple or obvious.
It seems the initrd support option depends upon
Device Drivers -> Block Devices -> RAM disk support
which defaults to N or M on a 2.6.9 kernel.org kernel after
'make mrproper' etc. Historically I tend to set this to M.
Also, the most important option:
"initial RAM disk (initrd) support"
is disabled and not even visible unless "RAM disk support"
is explicitly enabled (i.e., set to "Y"), which is not the case
on a stock kernel. It makes this option pretty hard to find
unless one knows exactly where it is and how to unhide it...
Anyway, enabling both these switches solved both the stock
kernel.org kernel and and LABEL=/ issues.
Thanks for the hint- I suspected it would be something painfully
simple, but not too obvious. I'm sure this issue will resurface
again and again.
ron
1:42 p.m.
I hit the same problem when updating from dev to udev the first time,
just run /sbin/udevstart to create the necessary /dev entries before you
reboot. You will likely need to umount /dev if you boot an FC kernel
though.
Tom
--
T h o m a s Z e h e t b a u e r ( TZ251 )
PGP encrypted mail preferred - KeyID 96FFCB89
finger thomasz(a)hostmaster.org for key
Die Post ist eine Institution zur verteuerten Verlangsamung der
Briefzustellung mit dem Ziel der Selbstabholung gegen zehnfache Gebühr.
C. Northcote Parkinson
6:14 p.m.
On Wednesday 10 November 2004 19:42, Thomas Zehetbauer wrote:
I hit the same problem when updating from dev to udev the first
time,
just run /sbin/udevstart to create the necessary /dev entries before you
reboot. You will likely need to umount /dev if you boot an FC kernel
though.
I didn't understand this.
What exactly do you mean by the last sentence?
What command do you give, and when?
--
Timothy Murphy
e-mail (<80k only): tim /at/ birdsnest.maths.tcd.ie
tel: +353-86-2336090, +353-1-2842366
s-mail: School of Mathematics, Trinity College, Dublin 2, Ireland
7104
days inactive
7106
days old
13 comments
5 participants
participants (5)
-
Arjan van de Ven -
Joachim Backes -
ron flory -
Thomas Zehetbauer -
Timothy Murphy