Removing all dkms, or installing and compiling all dkms are the only two
viable options.
Incomplete installs or mix&match of both can lead to issues.
More recently zfs upstream switched to not shipping many small modules
precisely because of issues like these.
** Changed in: zfs-linux (Ubuntu)
Assignee: Dimitri John Ledkov (xnox) => (unassigned)
** Changed in: zfs-linux (Ubuntu)
Status: Triaged => Invalid
--
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to zfs-linux in Ubuntu.
https://bugs.launchpad.net/bugs/1920956
Title:
ZFS module cannot be loaded after Kernel 5.4.0-67-generic release AND
if the system has zfs-dkms and spl-dkms packages installed
Status in zfs-linux package in Ubuntu:
Invalid
Bug description:
WARNING: The current bug will only occur under the following
conditions:
1- You have a system running on Bionic Release (18.04). Other systems seems
to be unaffected for now;
2- This system have 'spl-dkms' and 'zfs-dkms' packages installed. This
triggers both SPL and ZFS to be recompiled from source if we have a new Kernel
Image being installed;
3- You didn't patched your system kernel to either
"linux-modules-5.4.0-67-generic" or "linux-modules-5.4.0-1039-aws". You are
running an earlier kernel version, which can be from 5.3.0-XXX or 5.4.0-60 for
example and you need to upgrade your system to the latest 'kernel-image'
release.
ALTERNATIVELY:
1- When creating a new OS image based on the Bionic and it's already running
kernel 5.4.0-67 or 5.4.0-1030-aws, you decide to install 'spl-dkms' and
'zfs-dkms'.
SCENARIO:
We have systems running on Xenial, Bionic, and Focal Fossa releases. We also
have the 'Unattended Upgrades' on and we use ZFS.
Every time that Ubuntu releases kernel updates, they are updated on
our systems and both the new kernel and its modules will be active on
the next reboot. Or if you run your workload on AWS and uses Canonical
images, this is happening also on the image tagged as
'ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20210224' on
AWS Ireland / eu-west-1 Region (AMI ID = ami-0d75330b9efa7072d) for
fresh installs.
Some days ago, Ubuntu releases a new Linux Kernel update for the 5.4.0
release, and this is happening on both "linux-
modules-5.4.0-67-generic" and "linux-modules-5.4.0-1039-aws" (this
might be true on GKE, Azure and other public cloud providers kernel
releases, but I didn't test them).
I have noticed that both EC2 instances that were rebooted after the
kernel upgrade or if we are baking a new OS image based on the latest
Canonical images I was unable to get the ZFS module loaded. The error
message is showed below:
$ sudo modprobe zfs
modprobe: ERROR: could not insert 'zfs': Invalid argument
When checking the 'dmesg', we can see the errors below (or even when
booting up if you have the 'zfs-load-module.service' active on your
system).
[ 51.358828] icp: disagrees about version of symbol __kstat_create
[ 51.358830] icp: Unknown symbol __kstat_create (err -22)
[ 51.358925] icp: disagrees about version of symbol __kstat_install
[ 51.358927] icp: Unknown symbol __kstat_install (err -22)
[ 79.196638] icp: disagrees about version of symbol __kstat_delete
[ 79.196643] icp: Unknown symbol __kstat_delete (err -22)
[ 79.196687] icp: disagrees about version of symbol __kstat_create
[ 79.196689] icp: Unknown symbol __kstat_create (err -22)
[ 79.196765] icp: disagrees about version of symbol __kstat_install
[ 79.196766] icp: Unknown symbol __kstat_install (err -22)
On Bionic release, both th ZFS and the SPL modules comes with version
'0.7.5'.
$ dpkg -l |grep 0.7.5-1ubuntu16.9
ii libnvpair1linux 0.7.5-1ubuntu16.9
amd64 Solaris name-value library for Linux
ii libuutil1linux 0.7.5-1ubuntu16.9
amd64 Solaris userland utility library for Linux
ii libzfs2linux 0.7.5-1ubuntu16.9
amd64 OpenZFS filesystem library for Linux
ii libzpool2linux 0.7.5-1ubuntu16.9
amd64 OpenZFS pool library for Linux
ii zfs-dkms 0.7.5-1ubuntu16.9 all
OpenZFS filesystem kernel modules for Linux
ii zfs-initramfs 0.7.5-1ubuntu16.9 all
OpenZFS root filesystem capabilities for Linux - initramfs
ii zfs-zed 0.7.5-1ubuntu16.9
amd64 OpenZFS Event Daemon
ii zfsutils-linux 0.7.5-1ubuntu16.9
amd64 command-line tools to manage OpenZFS filesystems
On systems running kernel 5.4.0 already have the modules compiled on
version 0.8.1, which can vary depending on the active kernel.
(test) wbraga@bastion-host-i-07c42966be34bd44e:~$ modinfo zfs |head -2
filename: /lib/modules/5.3.0-1030-aws/kernel/zfs/zfs.ko
version: 0.8.1-1ubuntu14.4
(test) wbraga@bastion-host-i-07c42966be34bd44e:~$ modinfo spl |head -2
filename: /lib/modules/5.3.0-1030-aws/kernel/zfs/spl.ko
version: 0.8.1-1ubuntu14.4
(test) wbraga@bastion-host-i-07c42966be34bd44e:~$ uname -sr
Linux 5.3.0-1030-aws
(test) wbraga@bastion-host-i-07c42966be34bd44e:~$
From what I understood, both the SPL and the ZFS DKMS scripts would
not replace the modules already installed on the systems that comes
with 'linux-modules-5.3.0-XXXX' if they are equal or newer than the
compiled one, as shown below:
(...)
zavl.ko:
Running module version sanity check.
Error! Module version 0.7.5-1ubuntu16.11 for zavl.ko
is not newer than what is already found in kernel 5.4.0-64-generic
(0.8.3-1ubuntu12.5).
You may override by specifying --force.
(...)
However, some of the modules will be installed in another path (in
this case, I am showing the modules that were installed):
(...)
splat.ko:
Running module version sanity check.
- Original module
- No original module exists within this kernel
- Installation
- Installing to /lib/modules/5.4.0-67-generic/updates/dkms/
(...)
zpios.ko:
Running module version sanity check.
- Original module
- No original module exists within this kernel
- Installation
- Installing to /lib/modules/5.4.0-67-generic/updates/dkms/
(...)
icp.ko:
Running module version sanity check.
- Original module
- Installation
- Installing to /lib/modules/5.4.0-67-generic/updates/dkms/
After all these three modules are installed, I cannot load ZFS. I would
assume that the issue lies on the duplicated 'icp.ko' module under
'/lib/modules/5.4.0-67-generic/updates/dkms/' directory.
$ ls -l /lib/modules/5.4.0-67-generic/updates/dkms/
total 632
-rw-r--r-- 1 root root 317744 Mar 23 14:55 icp.ko
-rw-r--r-- 1 root root 289680 Mar 23 14:52 splat.ko
-rw-r--r-- 1 root root 36848 Mar 23 14:55 zpios.ko
$
It appears that newer kernels (such as '5.4.0-67-generic') have ZFS
version '0.8.3-1ubuntu12.6', while on previous kernel release (such as
5.4.0-64-generic) the ZFS version is '0.8.3-1ubuntu12.5'. Something
may have changed that now confuses modprobe to properly load the
kernel.
A workaround solution for this issue is to remove both 'spl-dkms' and
'zfs-dkms' packages. apt/dkg will remove the compiled modules and we
can load the module back.
STEPS TO REPRODUCE THE ISSUE:
1- Fire it up a Bionic machine. You can test it out on Vagrant for example.
Ensure the kernel is something one or two versions below.
2- Bump your system to kernel 5.4.0-60 (for example). You can achieve this by
running:
sudo apt install linux-image-5.4.0-60-generic linux-
headers-5.4.0-60-generic linux-modules-5.4.0-60-generic
3- Reboot your system
4- Install ZFS packages, including both DKMS.
sudo apt install libnvpair1linux libuutil1linux libzfs2linux
libzpool2linux spl spl-dkms zfs-dkms zfs-initramfs zfs-zed zfsutils-
linux.
5- Install a new kernel release, such as 5.4.0-64. Check whether spl-
dkms and zfs-dkms are triggered.
sudo apt install linux-image-5.4.0-64-generic linux-
headers-5.4.0-64-generic linux-modules-5.4.0-64-generic
6- reboot (you will be in in kernel 5.4.0-64)
7- Check whether you can load zfs with 'sudo modprobe zfs'
8- Now install kernel 5.4.0-67.
sudo apt install linux-image-5.4.0-67-generic linux-
headers-5.4.0-67-generic linux-modules-5.4.0-67-generic.
9- Reboot your system. You will be in Kernel '5.4.0-67'. The ZFS
module won't come up (check 'dmesg').
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/zfs-linux/+bug/1920956/+subscriptions
--
Mailing list: https://launchpad.net/~kernel-packages
Post to : kernel-packages@lists.launchpad.net
Unsubscribe : https://launchpad.net/~kernel-packages
More help : https://help.launchpad.net/ListHelp
