Ubuntu intel-i915-kdms error on update

MAFoElffen · ‎09-16-2023

I am running Ubuntu 22.04.3 LTS. Linux Kernel 6.2.0-32.

Everything was running great until today.

During an update process today, I am now getting a dkms error during the update returned by intel-i915-dkms. Checking the module make log, it says: "intel-i915-dkms kernel version parse failed"

lshw says the i915 driver is still loading. lsmod has the following

Code:

~$ sudo lsmod | grep i915
[sudo] password for mafoelffen:
i915 3883008 3
drm_buddy 20480 1 i915
ttm 110592 1 i915
drm_display_helper 212992 1 i915
cec 94208 2 drm_display_helper,i915
drm_kms_helper 249856 3 drm_display_helper,nvidia_drm,i915
drm 696320 12 drm_kms_helper,drm_display_helper,nvidia,drm_buddy,nvidia_drm,i915,ttm
i2c_algo_bit 16384 2 igb,i915
video 69632 2 i915,nvidia_modeset

It's not really an installation issues, as it was installed and working previously.

Any ideas?

MAFoElffen · ‎09-17-2023

I noticed that the update involved libc6...

mafoelffen@msi-ubuntu:~$ apt list libc6* --installed
Listing... Done
libc6-dbg/jammy-updates,now 2.35-0ubuntu3.3 amd64 [installed,automatic]
libc6-dev/jammy-updates,now 2.35-0ubuntu3.3 amd64 [installed,automatic]
libc6/jammy-updates,now 2.35-0ubuntu3.3 amd64 [installed,automatic]
libc6/jammy-updates,now 2.35-0ubuntu3.3 i386 [installed,automatic]

MAFoElffen · ‎09-18-2023

Crash report (/var/crash/intel-i915-dkms.0.crash):

ProblemType: Package
DKMSBuildLog:
 DKMS make.log for intel-i915-dkms-1.23.6.24.230425.29 for kernel 6.2.0-32-generic (x86_64)
 Sat Sep 16 05:18:19 PM PDT 2023
 Generating local configuration database from kernel ...Kernel version parse failed!
 make: *** [Makefile:49: olddefconfig] Error 1
 Generating local configuration database from kernel ...Kernel version parse failed!
 make[1]: *** [Makefile:49: modules] Error 1
 make: *** [Makefile:36: default] Error 2
DKMSKernelVersion: 6.2.0-32-generic
Date: Sat Sep 16 17:18:20 2023
Package: intel-i915-dkms 1.23.6.24.230425.29+i44-1
PackageVersion: 1.23.6.24.230425.29+i44-1
SourcePackage: intel-i915-dkms
Title: intel-i915-dkms 1.23.6.24.230425.29+i44-1: intel-i915-dkms kernel module failed to build

Attached Intel Graphics Bug Report.

system-info report can be viewed at: https://paste.ubuntu.com/p/9FY5XgWX4d/

Launchpad bug at: https://bugs.launchpad.net/ubuntu/+source/glibc/+bug/2036452

Jose_Intel · ‎09-19-2023

Hello @MAFoElffen

Thank you for posting on the Intel️® communities.

We will be more than happy to assist you, please allow us to check the issue internally. As soon as we have any update we will let you know.

Best regards,

Jose B.

Intel Customer Support Technician

MAFoElffen · ‎09-19-2023

@Jose_Intel --

Thank you for your response. I am patient, and I understand support structures.

If you need me to submit anything, in the way of log/journal extracts, traces, other information, I am happy to submit them.

My level of experience is in supporting Linux, Unix and Windows for over 34 years, and DEV Testing for almost as long. I have been supporting the Linux Graphics Layer since 2005. So I come with a few skills to cooperate in trying to resolve this. (LOL)

MAFoElffen · ‎09-19-2023

Launchpad was able to reproduce the error and crash: https://bugs.launchpad.net/ubuntu/+source/glibc/+bug/2036452/comments/5

They are marking it as invalid on their end, as, quoted by them:

This should be investigated by the Intel team, as the failure really doesn't tell us much about what happened. 
Marking as Invalid as evidently the glibc SRU isn't at fault.

Jose_Intel · ‎09-21-2023

Hello MAFoElffen

Thank you for patiently waiting.

Here we have some useful information that might help: Unable to Install Intel® Arc™ Graphics Driver on Ubuntu*

Please let me know the results.

Best regards,

Jose B.

Intel Customer Support Technician

MAFoElffen · ‎09-21-2023

Did you forward me the correct link? That would be 9 steps backwards, do you not think? (take this in stride, I cannot stop laughing...)

Those instructions are if you were not able to install the drivers for Ubuntu 20.04 LTS... That is recommending "Summary: Install 5.17.0-1020-oem"

The updated and revised Intel Install Instructions for client, for Ubuntu 22.04 LTS-- RE: https://dgpu-docs.intel.com/driver/client/overview.html are for Ubuntu 22.04 and 23.04, kernels of at least 5.19 on 22.04 with Intel recommending by installing the HWE (Hardware Enablement Stack), and quoted from those instructions: "Use of the standard upstream Ubuntu 23.04 kernel is the easiest install path."

The current kernel in 22.04 with HWE is 6.2.0-33. Your package 'intel-i915-dkms' still builds for Linux kernels 6.2.0-31 and 6.2.0-32, and before libc6 updated, the module built for kernel 6.2.0-33, but now not.

The linked instructions say to downgrade the kernel all the way back to Ubuntu (20.04 HWE series) 5.17-1020-OEM. That does not make sense to you does it? It doe snot make sense to me. Especially since Intel recommends using Ubuntu 23.04, and "that is the kernel series" in standard 23.04, Linux kernel 6.2-33 is now the current kernel for 23.04, and for 22.04 w/ the HWE.

Those instructions, underlying in the wording, also refer to asking "Me" at the Ubuntu Community for help...

Those recommendations as an answer would mean that Intel's published instructions and recommendations are wrong. And that the working model is 5 years back, right? I have more faith in Intel to consider that as true.

Ask the upper tier to look at this again, and their own instructions. There is a problem in the code intel-i915-dkms, as it relates to Linux Kernel 6.2.0-33 and current GLIB, confirmed by Launchpad. It fails while parsing the kernel version during the make...

There is no emotion in these words. I am working with you. I will work with your upper tier(s), if need be. What I have said are just the facts. I usually work with New Users through System Architects.

Current kernel testing for release is now 6.5 which I am now testing for Ubuntu 23.10 (dev). If this having problems with 6.2, it will surely have troubles getting to there for Ubuntu 24.04 LTS'es release in 7 months. We start the dev testing for 24.04 in about a month, right after 23.10 releases. See the bigger picture now? (LOL)

EDIT:

New assigned Intel Case # is case number 05993346.

Jose_Intel · ‎09-22-2023

Hello MAFoElffen

Thank you for your reply.

I am sorry for the misunderstanding, the article refers to that kernel version because that was the latest version when it was created. Now the latest is 6.5.4, and that is the recommended.

We understand you have opened a chat with us, and we will continue to help you through that channel now. We will therefore close this community case. Thank you.

Best regards,

Jose B.

Intel Customer Support Technician

knayana-gse · ‎12-20-2023

Hello @MAFoElffen ,

intel-i915-dkms 1.23.6.24.230425.29+i44-1, this package doesn't have 6.2 support.

Use latest dkms package: intel-i915-dkms/1.23.6.37.230425.49.

tag: I915_23WW39.5_682.38_23.6.37_230425.49

MAFoElffen · ‎12-20-2023

@knayana-gse--

That package is not in the Intel Ubuntu jammy repository

>>>

mafoelffen@msi-ubuntu:~$ sudo apt list intel-i915-dkms -a

Listing... Done

intel- i915-dkms/unknown,unknown 1.23.8.20.230810.22+i33-1 all

intel-i915-dkms/unknown,unknown 1.23.7.17.230608.24+i37-1 all

intel-i915-dkms/unknown,unknown 1.23.6.24.230425.29+i44-1 all

intel-i915-dkms/unknown,unknown 1.23.5.19.230406.21.5.17.0.1034+i38-1 all

>>>

Is that package version in Intel's Lunar, Mantic or Noble repo?

So how do I either overload apt to get that from a different Intel repo, or do you have the URL so I can download the .deb file with wget?

After that, to correct that not being in the jammy repo, and since Jammy's kernel series went to 6.2.0-X series as it's LTS kernels... Don't you think that is a "Bug" that the packages in the Jammy repo no longer support the kernels it uses as LTS? Just a thought...

By the way, the open Intel Support Case for this is now: 05993346

I'll tell the Intel Support Engineers working on this what you found...

MAFoElffen · ‎12-20-2023

@knayana-gse--

The other Intel Support Team emailed me back and asked me to continue here for now with you, and to see what we can get working.

The communication was a bit burred, as I still think that the in-between person doesn't fully understand that I don't care about just the iGPU using the i915 in-kernel driver for the host.

As you understand, this was supposed to be a replacement for Intel GVT-g for shared virtual pass-through to VM's... That doesn't happen with just i915 in-kernel.

Luis did share this link with me: https://dgpu-docs.intel.com/driver/client/overview.html#client-install-options

Where do we go from here to make this happen? What can I do, and/or try on my end?

I figure if I can get that package version, even if I have to extract it to modify the checks and depends, then repackage it for jammy (locally)... I can try to make it work for these kernels in Ubuntu 22.04.3, where Ubuntu changed the LTS kernel series to 6.2.0-x... If we get something working, then either of us could submit the patches.

The other team says/implied that what we come up with here, they will update the Intel Doc's for the installation instructions. At least, that is my understanding of what they said on that.

If I am mistaken on any of that, please tell me so we are on the same understanding.

MAFoElffen · ‎12-20-2023

@knayana-gse--

I can't browse the Intel Repo's, so I'm going to probably need a URL to download it...

Are there any other depends on other package versions that will be needed with that? I'm thinking that's probably going to be on the version of intel-fw-gpu(?)

Either that or tell me which repo it is in, and I'll temporarily edit the sources to grab it and any depends that go along with that... To download those package(s).

EDIT. I don't see that anywhere. Now waiting on you.

knayana-gse · ‎12-20-2023

Hi,

Would you please try with latest release based on tag intel-i915-dkms/1.23.6.37.230425.49.

https://dgpu-docs.intel.com/releases/production_682.38_20230929.html

Please be aware that ubuntu22.04 hwe is moved to 6.5 kernel version, which is not supported yet to please make sure to check against 6.2 kernel.

MAFoElffen · ‎12-21-2023

@knayana-gse--

Trying that Link now...

Quote (knayana-gse):

>>> Please be aware that ubuntu22.04 hwe is moved to 6.5 kernel version, which is not supported yet to please make sure to check against 6.2 kernel.

Not yet. Not true currently. LTS Kernel Series for Ubuntu Jammy LTS point release 22.04.3 is 6.2.-X Kernel series. Who knows what is in the future point releases, but currently not now.

Code:

>>>

Listing... Done
mafoelffen@msi-ubuntu:~$ apt list *hwe*22.04* --installed
Listing... Done
linux-generic-hwe-22.04/jammy-updates,jammy-security,now 6.2.0.39.40~22.04.16 amd64 [installed]
linux-headers-generic-hwe-22.04/jammy-updates,jammy-security,now 6.2.0.39.40~22.04.16 amd64 [installed,automatic]
linux-image-generic-hwe-22.04/jammy-updates,jammy-security,now 6.2.0.39.40~22.04.16 amd64 [installed,automatic]
mafoelffen@msi-ubuntu:~$ uname -r
6.2.0-37-generic

>>>

I believe you might be confused. 22.04 HWE & Lunar 23.04 are 6.2. Mantic 23.10 & DEV-Noble 24.04, which I am in the process (separately) of testing now, is currently using and slated to be LTS Kernel series 6.5. I'm not worrying about that yet. We are still working things out for that in the next 4 months before release. It's harder to hit a moving target that changes whole lots right now, from day-to-day. I know personally on that.

DEV HWE stack for 24.04 is not out yet. We are currently testing kernel 6.7 RC and some other pieces of the stack... But who knows what will be in the HWE stack yet.

*** Being able to do Intel OOT i915 and share that between VM's brings a lot of added value to KVM guests. If we can get that working again, that's a big score for Intel!!!

MAFoElffen · ‎12-21-2023

Okay... I'm just going to deal with "Just the facts Ma'am..."

That page says it was updated on: 2023-09-29

For Ubuntu 22.04, it says to use: intel-i915-dkms=1.23.6.37.230425.49

You say to use: intel-i915-dkms=1.23.6.37.230425.49, probably based on that document, telling me that I should not be using version 1.23.6.24.230425.29+i44-1, which I had installed back in September, when the problem originally occurred...

Here is what is in the Intel Ubuntu jammy repo's today:

>>>

mafoelffen@msi-ubuntu:~$ apt list -a intel-i915-dkms
Listing... Done
intel-i915-dkms/unknown,unknown 1.23.9.11.231003.15+i19-1 all
intel-i915-dkms/unknown,unknown 1.23.8.20.230810.22+i33-1 all
intel-i915-dkms/unknown,unknown 1.23.7.17.230608.24+i37-1 all
intel-i915-dkms/unknown,unknown 1.23.6.24.230425.29+i44-1 all
intel-i915-dkms/unknown,unknown 1.23.5.19.230406.21.5.17.0.1034+i38-1 all

>>>

You can see that that package version is not only not in that repository, that there are 3 package version newer. That sort of indicates to me, that not only is that document currently out-of-date, that it was wrong with the package version back then. You are seeing that right?

That list is sorted as newest to oldest...

I can tell you that version 1.23.8.20.230810.22+i33-1 still failed...

NOTICE that version 1.23.9.11.231003.15+i19-1 was not there yesterday in my query posted above... That version is NEW today.

I can confirm that installing 'that new version from today':

>>>

apt install -y intel-i915-dkms=1.23.9.11.231003.15+i19-1

>>>

Succeeds and works... See attached results.

NOW -- I'll tell the other Intel team to update the doc's (They said they would). And I'll update the bug report at LaunchPad...

MAFoElffen · ‎12-31-2023

@knayana-gse---

I asked Luis if he wants to reopen the last support case or wants me to open a new one... Was premature.

>>> intel-i915-dkms/unknown,unknown 1.23.9.11.231003.15+i19-1 all

The last update to intel-i915-dkms, builds the modules for kernels 6.2.0-x without crashing. The VM's see the iGPU resources from the host. All was well until the reboot.

That machine is ZFS-On-Root, and this new version breaks zfs-linux: zed.

Error message on boot was "No device /dev/zd0: file or device not found" Only pool that was imported was rpool. 5 other pools not imported, but would import. zfs-zed.service would not start.

Would boot successfully on 5.15.0-x kernels.

Removed package intel-i915-dkms and would boot fine from the 6.2.0-x kernels.

I was so excited that it was working... Before finding that fatal error. Dang.

MAFoElffen · ‎01-03-2024

@knayana-gse---

Please contact Luis for me. He emailed me and said he is keeping Intel Customer Support Case # 05993346 open, because it is the same driver and a problem with the fix... He said he would give Jianshui and the team a heads up.

Problem with that, when I emailed him back via reply, with that Case # in the title, the intel Support email system kicks it out of the system, saying it's not tied to anything. So am currently caught in Intel Support limbo... Will post what I said here:

To reproduce, Install Ubuntu Desktop 22.04 LTS with ZFS as the install method. I have 5 pools (rpool, bpool, hpool, dpool, & kpool), but it will fail with just the defautl 2 pools (rpool & bpool). Reboot. Log in. Install the Intel OOT drivers. All seems to work... Albeit, a problem just with some of the 'libva' libraries just when you try to run vainfo. --> xpu-smi & clinfo seem to show everything is working okay. Notes, the default install of desktop is which the HWE stack installed. ensure you have both linux-image-generic & linux-image-gerneric-hwe-22.04 installed

>>> apt list linux-image-generic-* --installed

For test purposes, so that you can get to the Grub2 menu, edit the /etc/default/grub file with these edits

>>> GRUB_TIMEOUT_STYLE=menu
>>> GRUB_TIMEOUT=5
>>> GRUB_RECORDFAIL_TIMEOUT=5
>>> GRUB_CMDLINE_LINUX_DEFAULT="--- --verbose debug drm.debug=0xe intel_iommu=on"

Save. Pick up the changes:

>>> sudo update-grub

Reboot...

With any kernel in the 6.2.0-X series: Will give an error saying something along the lines of "device /dev/sd0: file or device not found"... Then continue booting, taking a long time on udisk, the ending up at a root maintenance prompt. (Instead of the desktop).

If you do

>>> zpool list

It will show that only rpool is imported. (bpool is not present) If you run

>>> zpool import bpool

>>> zpool list

All are imported fine. No matter what you do, if you boot from the ... Will result in the same.

Why? If you do

>>> systemctl status zfs-zed.service

'zed' is inactive. It never starts. It will not start manually, and will not restart. <-- On a normal boot sequence with ZFS, the zed service should start, and you should see it import of each pool in the status... That is not happening at all.

If you boot any of the 5.15.0 series kernels, it boots normally...

I installed the OOT drivers on the second Ubuntu 22.04 LTS installation that is installed on LVM2, with no ZFS... same machine, just different location on it. It boots fine on 6.2.0 kernels.

### End of that email that was kicked back. ###

I will extract the syslogs. Do you all want then emailed (if so to where)? Posted as an attachment? Or uploaded to a pastebin and post the URL?

Felipe_Intel · ‎01-08-2024

Hello @MAFoElffen I sent you a new message over email. Try to replay to it. If you still get the error, let me know we will create a new case for you. I already shared this info with the team.

Regards

Luis Felipe

PaulMD · ‎05-18-2024

@Felipe_IntelCan you please post the sequence on the forums instead of maintaining sidechannels please? I actually have been debugging some problems that in hindsight might have also been caused by ZFS-on-root and I'd like to know what the problems are too.

And I actually had the same problem where your support window appears to have drifted behind the actual OS's support state. I figured this was probably the cause and dug up this thread.