reopen 469312 !
found 469312 0.92a
found 469312 0.85h
notfixed 469312 0.91d
severity 469312 grave
tag 469312 +etch
thanks

Sorry, but this bug is neither fixed in Sid nor in Etch (where it is
also present). In Etch the bug breaks dist-upgrades from Sarge to Etch
at least if previously a devfs kernel, lilo and md was used. Just had
such a case yesterday on one of our DHCP servers.

And since such systems usually continue to use the /dev/md/* naming
scheme because udev seems to recognise a former devfs system, this
still breaks systems and can render them unbootable (therefore
severity grave) if the admin doesn't notice the mentioned syntax
errors.

The line which obviously is causing the syntax error is the following
which is used when update-initramfs parses lilo.conf:

boot="/dev/$(awk "/^${MD}/{print substr($5, 1, 3)}"

It's line 207 in Sid and line 166 in Etch.

Slashes contained in $MD need be escaped properly to fix this problem.

There's one more line which potentially has a similar bug:

&& dev=$(awk "/${groot}/{ print $NF}" /boot/grub/device.map)

It's line 195 in Sid and line 154 in Etch.

Since is primarily an escaping bug which should be easy and not
critical to fix, I would be happy, if this could be fixed in Etch with
perhaps 4.0r4, too.

Kind regards, Axel Beckert
--
Axel Beckert <beckert@phys.ethz.ch> support: +41 44 633 2668
IT Support Group, HPR E 86.1 voice: +41 44 633 4189
Departement Physik, ETH Zurich fax: +41 44 633 1239
CH-8093 Zurich, Switzerland http://nic.phys.ethz.ch/



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

On Fri, May 30, 2008 at 01:22:12PM +0200, Axel Beckert wrote:
>
> Sorry, but this bug is neither fixed in Sid nor in Etch (where it is
> also present). In Etch the bug breaks dist-upgrades from Sarge to Etch
> at least if previously a devfs kernel, lilo and md was used. Just had
> such a case yesterday on one of our DHCP servers.
>
> And since such systems usually continue to use the /dev/md/* naming
> scheme because udev seems to recognise a former devfs system, this
> still breaks systems and can render them unbootable (therefore
> severity grave) if the admin doesn't notice the mentioned syntax
> errors.
>
> The line which obviously is causing the syntax error is the following
> which is used when update-initramfs parses lilo.conf:
>
> boot="/dev/$(awk "/^${MD}/{print substr($5, 1, 3)}"
>
> It's line 207 in Sid and line 166 in Etch.

thanks for the useful report, fixed in latest trunk.
i'll upload during the day and will propose this fix for an etch update
too as another fix has been proposed for that too.

> Slashes contained in $MD need be escaped properly to fix this problem.
>
> There's one more line which potentially has a similar bug:
>
> && dev=$(awk "/${groot}/{ print $NF}" /boot/grub/device.map)
>
> It's line 195 in Sid and line 154 in Etch.

no there it can't happen due to groot beeing a grub naming style
variable aka fdX or hdX

no slash.

> Since is primarily an escaping bug which should be easy and not
> critical to fix, I would be happy, if this could be fixed in Etch with
> perhaps 4.0r4, too.
>
> Kind regards, Axel Beckert

see
http://git.debian.org/?p=kernel/initramfs-tools.git;a=summary

best regards

--
maks




--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

Hi,

On Fri, May 30, 2008 at 02:11:23PM +0200, maximilian attems wrote:
> thanks for the useful report, fixed in latest trunk.

thanks for the very fast response and fix. :-)

> > There's one more line which potentially has a similar bug:
> >
> > && dev=$(awk "/${groot}/{ print $NF}" /boot/grub/device.map)
> >
> > It's line 195 in Sid and line 154 in Etch.
>
> no there it can't happen due to groot beeing a grub naming style
> variable aka fdX or hdX

Ok, I haven't looked at that very close. Just did a grep awk over the
script and looked for variables in regexps. :-)

> http://git.debian.org/?p=kernel/initramfs-tools.git;a=summary

Ok, that's also a way to fix it. But I wonder if that could make
problems with other device names with slashes like LVM logical volumes
or other (e.g. crypted) device manager devices and such...

Kind regards, Axel Beckert
--
Axel Beckert <beckert@phys.ethz.ch> support: +41 44 633 2668
IT Support Group, HPR E 86.1 voice: +41 44 633 4189
Departement Physik, ETH Zurich fax: +41 44 633 1239
CH-8093 Zurich, Switzerland http://nic.phys.ethz.ch/



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

On Fri, May 30, 2008 at 03:03:51PM +0200, Axel Beckert wrote:
> > http://git.debian.org/?p=kernel/initramfs-tools.git;a=summary
>
> Ok, that's also a way to fix it. But I wonder if that could make
> problems with other device names with slashes like LVM logical volumes
> or other (e.g. crypted) device manager devices and such...

never seen a lvm2 named "md", you are right that it is possible,
but hell that is catched by no existing /proc/mdstat, so..

sunny greetings

--
maks



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org