symlinks managed by kernel postinst are different from zipl-installer and livefs-rootfs

[Dimitri John Ledkov]

On Ubuntu/Debian like systems /etc/kernel-img.conf should be created by the "installer". It is currently done by base-installer/live-installer/ubiquity but not curtin, but it should be. At the moment we require kernel-img.conf and we do not have the correct per-arch built-in defaults for its settings in all of our kernels. It is not a config file, nor is it created by livecd-rootfs (after all our squashfs might be containers, and never live to install kernels and bootloaders).

One day we might fix our kernels to not require kernel-img.conf, but until then curtin should be generating the right one. Making a merge proposal to fix this in curtin by mimicking what base-installer did.

==

Symlinks are not managed correctly.

Last installed and configured kernel, prior to purging -5- was -6-, yet symlinks were not updated to -6- when that happened.

root@devac03:~# apt-get remove --purge linux-headers-4.3.0-5 linux-headers-4.3.0-5-generic linux-image-4.3.0-5-generic linux-image-extra-4.3.0-5-generic
Reading package lists... Done
Building dependency tree
Reading state information... Done
The following packages will be REMOVED:
  linux-headers-4.3.0-5* linux-headers-4.3.0-5-generic* linux-image-4.3.0-5-generic*
  linux-image-extra-4.3.0-5-generic*
0 upgraded, 0 newly installed, 4 to remove and 0 not upgraded.
After this operation, 131 MB disk space will be freed.
Do you want to continue? [Y/n] Y
(Reading database ... 92073 files and directories currently installed.)
Removing linux-headers-4.3.0-5-generic (4.3.0-5.16) ...
Removing linux-headers-4.3.0-5 (4.3.0-5.16) ...
Removing linux-image-extra-4.3.0-5-generic (4.3.0-5.16) ...
run-parts: executing /etc/kernel/postinst.d/apt-auto-removal 4.3.0-5-generic /boot/vmlinuz-4.3.0-5-generic
run-parts: executing /etc/kernel/postinst.d/initramfs-tools 4.3.0-5-generic /boot/vmlinuz-4.3.0-5-generic
update-initramfs: Generating /boot/initrd.img-4.3.0-5-generic
Using config file '/etc/zipl.conf'
Building bootmap in '/boot/'
Building menu 'zipl-automatic-menu'
Adding #1: IPL section 'ubuntu' (default)
Preparing boot device: dasda (0200).
Done.
run-parts: executing /etc/kernel/postinst.d/pm-utils 4.3.0-5-generic /boot/vmlinuz-4.3.0-5-generic
run-parts: executing /etc/kernel/postinst.d/zz-zipl 4.3.0-5-generic /boot/vmlinuz-4.3.0-5-generic
Using config file '/etc/zipl.conf'
Building bootmap in '/boot/'
Building menu 'zipl-automatic-menu'
Adding #1: IPL section 'ubuntu' (default)
Preparing boot device: dasda (0200).
Done.
Purging configuration files for linux-image-extra-4.3.0-5-generic (4.3.0-5.16) ...
Removing linux-image-4.3.0-5-generic (4.3.0-5.16) ...
WARN: Proceeding with removing running kernel image.
Examining /etc/kernel/postrm.d .
run-parts: executing /etc/kernel/postrm.d/initramfs-tools 4.3.0-5-generic /boot/vmlinuz-4.3.0-5-generic
update-initramfs: Deleting /boot/initrd.img-4.3.0-5-generic
run-parts: executing /etc/kernel/postrm.d/zz-zipl 4.3.0-5-generic /boot/vmlinuz-4.3.0-5-generic
Using config file '/etc/zipl.conf'
Error: Image file '/boot/vmlinuz' in section 'ubuntu': No such file or directory
run-parts: /etc/kernel/postrm.d/zz-zipl exited with return code 1
Failed to process /etc/kernel/postrm.d at /var/lib/dpkg/info/linux-image-4.3.0-5-generic.postrm line 328.
dpkg: error processing package linux-image-4.3.0-5-generic (--purge):
 subprocess installed post-removal script returned error exit status 1
Errors were encountered while processing:
 linux-image-4.3.0-5-generic
E: Sub-process /usr/bin/dpkg returned an error code (1)
root@devac03:~# ls -latr /boot
total 24208
drwx------ 2 root root 16384 Dec 9 16:38 lost+found
lrwxrwxrwx 1 root root 26 Jan 6 16:23 initrd.img -> initrd.img-4.3.0-5-generic
lrwxrwxrwx 1 root root 23 Jan 6 16:24 vmlinuz -> vmlinuz-4.3.0-5-generic
-rw------- 1 root root 13026048 Jan 11 21:36 vmlinuz-4.3.0-6-generic
-rw------- 1 root root 2446124 Jan 11 21:36 System.map-4.3.0-6-generic
-rw-r--r-- 1 root root 63422 Jan 11 21:36 config-4.3.0-6-generic
-rw-r--r-- 1 root root 517933 Jan 11 21:36 abi-4.3.0-6-generic
drwxr-xr-x 22 root root 4096 Jan 14 13:03 ..
-rw-r--r-- 1 root root 8574889 Jan 14 13:03 initrd.img-4.3.0-6-generic
-rw------- 1 root root 69632 Jan 14 13:41 bootmap
drwxr-xr-x 3 root root 4096 Jan 14 13:41 .
root@devac03:~# dpkg -l | grep 4.3.0
ii iproute 1:4.3.0-1ubuntu1 all transitional dummy package for iproute2
ii iproute2 4.3.0-1ubuntu1 s390x networking and traffic control tools
ii linux-generic 4.3.0.6.7 s390x Complete Generic Linux kernel and headers
ii linux-headers-4.3.0-6 4.3.0-6.17 all Header files related to Linux kernel version 4.3.0
ii linux-headers-4.3.0-6-generic 4.3.0-6.17 s390x Linux kernel headers for version 4.3.0 on System 390x SMP
ii linux-headers-generic 4.3.0.6.7 s390x Generic Linux kernel headers
rc linux-image-4.3.0-2-generic 4.3.0-2.11 s390x Linux kernel image for version 4.3.0 on System 390x SMP
pH linux-image-4.3.0-5-generic 4.3.0-5.16 s390x Linux kernel image for version 4.3.0 on System 390x SMP
ii linux-image-4.3.0-6-generic 4.3.0-6.17 s390x Linux kernel image for version 4.3.0 on System 390x SMP
rc linux-image-extra-4.3.0-2-generic 4.3.0-2.11 s390x Linux kernel extra modules for version 4.3.0 on System 390x SMP
ii linux-image-extra-4.3.0-6-generic 4.3.0-6.17 s390x Linux kernel extra modules for version 4.3.0 on System 390x SMP
ii linux-image-generic 4.3.0.6.7 s390x Generic Linux kernel image
ii linux-libc-dev:s390x 4.3.0-6.17 s390x Linux Kernel Headers for development
ii xfslibs-dev 4.3.0 s390x XFS filesystem-specific static libraries and headers
ii xfsprogs 4.3.0 s390x Utilities for managing the XFS filesystem

[Dimitri John Ledkov (xnox)]

http://paste.ubuntu.com/14495976/

[Dimitri John Ledkov (xnox)]

Ok, here is a better scenario

I have purged -6, and installed -5.

Then i have installed -6 and expected for symlinks to change from -5 -> -6, however they did not.

http://paste.ubuntu.com/14496011/

[Dimitri John Ledkov (xnox)]

i am an idiot.

symlinks are updated in /
yet my zipl is using weird symlinks in /boot/

[Dimitri John Ledkov (xnox)]

Aha

# cat /etc/kernel-img.conf
# Kernel image management overrides
# See kernel-img.conf(5) for details
do_symlinks = yes
link_in_boot = yes

[Paride Legovini (paride)]

It seems that we're hitting this issue again with the subiquity installer. The installer fails at the "installing kernel" stage, when zipl is called. These are the last lines of the installer log (retrieving the full log is not trivial because of LP: #1823797):

Setting up linux-image-5.0.0-8-generic (5.0.0-8.9) ...
I: /vmlinuz.old is now a symlink to boot/vmlinuz-5.0.0-8-generic
I: /initrd.img.old is now a symlink to boot/initrd.img-5.0.0-8-generic
I: /vmlinuz is now a symlink to boot/vmlinuz-5.0.0-8-generic
I: /initrd.img is now a symlink to boot/initrd.img-5.0.0-8-generic
Not invoking zipl: initrd doesn't exist yet
/etc/kernel/postinst.d/zz-zipl:
Not invoking zipl: initrd doesn't exist yet
Setting up linux-generic (5.0.0.8.9) ...
/etc/kernel/postinst.d/initramfs-tools:
update-initramfs: Generating /boot/initrd.img-5.0.0-8-generic
Not invoking zipl: initrd doesn't exist yet
/etc/kernel/postinst.d/zz-zipl:
Not invoking zipl: initrd doesn't exist yet
Running command ['unshare', '--fork', '--pid', '--', 'chroot', '/target', 'zipl']
Error: Image file '/boot/vmlinuz' in section 'ubuntu': No such file or directory

The qcow2 disk image can be mounted like this:

$ sudo modprobe nbd max_part=8
$ sudo qemu-nbd --connect=/dev/nbd0 /path/to/CE-install-test.qcow2
$ sudo mount /dev/nbd0p1 /mnt/guest

and it contains the expected target filesystem, but the vmlinuz and initrd.img symlink are in / instead of being in /boot, where zipl expects them:

lrwxrwxrwx 1 root root 31 Apr 9 08:11 initrd.img ->
boot/initrd.img-5.0.0-8-generic
lrwxrwxrwx 1 root root 28 Apr 9 08:11 vmlinuz ->
boot/vmlinuz-5.0.0-8-generic

Moving the symlinks in /boot and chrooting in the target partition makes zipl work:

$ for i in sys dev proc; do sudo mount -o bind /$i /mnt/guest/$i; done
$ sudo chroot /mnt/guest/
groups: cannot find name for group ID 129
root@s1lp04:/# rm initrd.img* vmlinuz*
root@s1lp04:/# cd /boot/
root@s1lp04:/boot# ln -s vmlinuz-5.0.0-8-generic vmlinuz
root@s1lp04:/boot# ln -s initrd.img-5.0.0-8-generic initrd.img
root@s1lp04:/boot# cd ..
root@s1lp04:/# zipl -V
Using config file '/etc/zipl.conf'
Warning: Could not remove temporary file : No such file or
directoryfree(): double free detected in tcache 2
Aborted

(It still errors, but we go past the "No such file or directory" error, and this error is probably related to the chroot.)

It seems that the location of the vmlinuz and initrd symlinks can be configured with the /etc/kernel-img.conf file. On the host system we have:

$ cat /etc/kernel-img.conf
# Kernel image management overrides
# See kernel-img.conf(5) for details
do_symlinks = yes
do_bootloader = yes
do_initrd = yes
link_in_boot = yes

and the symlinks are indeed creat...

[Dimitri John Ledkov (xnox)]

I think last time it was fixed by fixing base-installer.
http://launchpadlibrarian.net/235185956/base-installer_1.158ubuntu1_1.158ubuntu2.diff.gz

but this time around, we need to decide who/what/where should write out kernel-img.conf.

because we do not have kernel-common by default.

[Server Team CI bot (server-team-bot)]

This bug is fixed with commit 08bf6ff7 to curtin on branch master.
To view that commit see the following URL:
https://git.launchpad.net/curtin/commit/?id=08bf6ff7

[Dan Watkins (oddbloke)]

This bug is believed to be fixed in curtin in version 19.1. If this is still a problem for you, please make a comment and set the state back to New

Thank you.