Kernel 4.13.0-25 broke nvidia driver

[Lastique]

After updating to linux-image-4.13.0-25-generic the binary nvidia driver 387.22 cannot be built. As a result the system does not boot into the desktop environment (in my case, KDE).

I've attached the driver build log.

ProblemType: Bug
DistroRelease: Ubuntu 17.10
Package: linux-image-4.13.0-25-generic 4.13.0-25.29
ProcVersionSignature: Ubuntu 4.13.0-21.24-generic 4.13.13
Uname: Linux 4.13.0-21-generic x86_64
NonfreeKernelModules: nvidia_uvm nvidia_drm nvidia_modeset nvidia
ApportVersion: 2.20.7-0ubuntu3.7
Architecture: amd64
AudioDevicesInUse:
 USER PID ACCESS COMMAND
 /dev/snd/controlC0: lastique 3049 F.... pulseaudio
 /dev/snd/controlC2: lastique 3049 F.... pulseaudio
 /dev/snd/controlC3: lastique 3049 F.... pulseaudio
 /dev/snd/controlC1: lastique 3049 F.... pulseaudio
CurrentDesktop: KDE
Date: Wed Jan 10 02:00:46 2018
InstallationDate: Installed on 2015-05-01 (984 days ago)
InstallationMedia: Kubuntu 15.04 "Vivid Vervet" - Release amd64 (20150422)
MachineType: System manufacturer System Product Name
ProcFB:

ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-4.13.0-21-generic root=UUID=a03f1835-52f9-4409-9da7-fe45770637ae ro quiet splash nomdmonddf nomdmonisw
RelatedPackageVersions:
 linux-restricted-modules-4.13.0-21-generic N/A
 linux-backports-modules-4.13.0-21-generic N/A
 linux-firmware 1.169.1
RfKill:
 0: hci0: Bluetooth
  Soft blocked: no
  Hard blocked: no
SourcePackage: linux
UpgradeStatus: Upgraded to artful on 2017-11-03 (67 days ago)
dmi.bios.date: 11/09/2012
dmi.bios.vendor: American Megatrends Inc.
dmi.bios.version: 3603
dmi.board.asset.tag: To be filled by O.E.M.
dmi.board.name: P8Z68-V PRO
dmi.board.vendor: ASUSTeK Computer INC.
dmi.board.version: Rev 1.xx
dmi.chassis.asset.tag: Asset-1234567890
dmi.chassis.type: 3
dmi.chassis.vendor: Chassis Manufacture
dmi.chassis.version: Chassis Version
dmi.modalias: dmi:bvnAmericanMegatrendsInc.:bvr3603:bd11/09/2012:svnSystemmanufacturer:pnSystemProductName:pvrSystemVersion:rvnASUSTeKComputerINC.:rnP8Z68-VPRO:rvrRev1.xx:cvnChassisManufacture:ct3:cvrChassisVersion:
dmi.product.family: To be filled by O.E.M.
dmi.product.name: System Product Name
dmi.product.version: System Version
dmi.sys.vendor: System manufacturer

[Ubuntu Kernel Bot (ubuntu-kernel-bot)]

This change was made by a bot.

[NJ (njspam+ubuntulaunchpad)]

Not sure if this is related. Caught this before I upgraded. System is now freezing and I can't use an old kernel as I get the following:

NVRM: API mismatch: the client has the version 384.111, but
NVRM: this kernel module has the version 384.90. Please
NVRM: make sure that this kernel module and all NVIDIA driver

From the upgrade (term.log)

Setting up nvidia-384 (384.111-0ubuntu0.17.10.1) ...
update-alternatives: renaming x86_64-linux-gnu_nvidia_app_profile slave link from /usr/share/nvidia/nvidia-application-profiles-384.90-rc to /usr/share/nvidia/nvidia-application-profiles-384.111-rc
update-alternatives: renaming x86_64-linux-gnu_nvidia_app_profile_keys slave link from /usr/share/nvidia/nvidia-application-profiles-384.90-key-documentation to /usr/share/nvidia/nvidia-application-profiles-384.111-key-documentation
update-alternatives: updating alternative /usr/lib/nvidia-384/ld.so.conf because link group x86_64-linux-gnu_gl_conf has changed slave links
dpkg: error: version '-' has bad syntax: revision number is empty
dpkg: error: version '-' has bad syntax: revision number is empty
dpkg: error: version '-' has bad syntax: revision number is empty
update-initramfs: deferring update (trigger activated)
INFO:Enable nvidia-384
DEBUG:Parsing /usr/share/ubuntu-drivers-common/quirks/dell_latitude
DEBUG:Parsing /usr/share/ubuntu-drivers-common/quirks/put_your_quirks_here
DEBUG:Parsing /usr/share/ubuntu-drivers-common/quirks/lenovo_thinkpad
Loading new nvidia-384-384.111 DKMS files...
Building for 4.13.0-21-generic 4.13.0-25-generic
Building for architecture x86_64
Building initial module for 4.13.0-21-generic
Done.

[monkeybrain2012 (kammon101)]

Still the same in 4.13.0-26. Ubuntu 16.04.3, cannot boot into desktop after kernel upgraded, boot into the previous kernel it all works again.

[hero (theherodied)]

I had the same issue at first. What worked for me was to make sure the following was installed:

linux-headers-generic
linux-image-4.13.0-26-generic
linux-image-extra-4.13.0-26-generic

Then run driver-manager and switch to a different Nvidia driver and wait for it to finish. Then switch it back to the one you wanted and reboot. For mine I switched to nvidia-381 and then back to nvidia-387.

[hero (theherodied)]

I run Linux Mint so it might be a little different. Not sure if Ubuntu has the driver-manager package.

[hero (theherodied)]

Above should have said linux-headers-4.13.0-26-generic like the rest of the packages.

[Joseph Salisbury (jsalisbury)]

Would it be possible for you to test the latest upstream kernel? Refer to https://wiki.ubuntu.com/KernelMainlineBuilds . Please test the latest v4.15 kernel[0].

If this bug is fixed in the mainline kernel, please add the following tag 'kernel-fixed-upstream'.

If the mainline kernel does not fix this bug, please add the tag: 'kernel-bug-exists-upstream'.

Once testing of the upstream kernel is complete, please mark this bug as "Confirmed".

Thanks in advance.

[0] http://kernel.ubuntu.com/~kernel-ppa/mainline/v4.15-rc7

[Matt W. (mw1234)]

This bug is also occurring for me when installing nvidia-387 off Nvidia's 16.04 repository when using 4.13.0-26-generic. Attached is /var/lib/dkms/nvidia-387/387.26/build/make.log.

[monkeybrain2012 (kammon101)]

I tested with the mainline kernel4.15rc7, problem still not fixed, still can't boot into desktop. I tried reinstalling the Nvidia driver in kernel 4.10 but build for 4.15 has error, I did it with synaptic and the window closed too soon for me to catch the error message.

[Matt W. (mw1234)]

I also installed and booted 4.15-rc7 and attempted to install nvidia-387 and got another crash in the DKMS build process.

However I installed nvidia-384 (this would be the just-released 384.111) on 4.13.0-26-generic and it works fine.

[Felipe Veas (fjveas)]

This seems to be related to https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1742095 on which I found a workaround for the DKMS build process.

[Po-Hsu Lin (cypressyew)]

Possible duplicate bug 1742160

[fgr1986 (fernando-garca)]

The same happens with driver 340 and 384. Older kernels work fine.
For now I have disabled nvidia's driver, and the open-sourced driver works.

[Lastique (andysem)]

I have fixed the problem by installing nvidia driver 390.12.

[db (dbec)]

I confirm this with 4.13.0-26. Does not occur after reversion to 4.10.0-42 or earlier kernels.

I have not tested the new 390 series driver. 387.26 + 4.10.0-42 combination works.

[db (dbec)]

Some additional information: when installing to a bare new installation of 16.04.3 LTS upgraded to the new kernel (4.13.xxx), installation of NVIDIA's CUDA 9 packages fails due to a module build failure, similar to the description of bug 1742160 https://bugs.launchpad.net/ubuntu/+source/nvidia-graphics-drivers-384/+bug/1742160

[Hrafn Malmquist (jabbi)]

I have an Nvidia GTX 1060 3gb and after upgrading to kernel 4.13.0-26 I could not log into desktop.

Resolution at the login was very low. Running 4.13.0-1x worked.

What fixed the problem for me was upgrading the Nvidia drivers from 381 => 384.

[jessepage1989 (jessepage1989)]

Same thing happened to me. When I tried to reboot with previous kernel the mouse and keyboard would disappear. Removing Nvidia driver got me back to the desktop.
In my case original nvidia drivers was the cause and switching back to open source nouveau helped. First, I removed all stuff from nvidia:

1. sudo apt-get remove --purge 'nvidia-*'
2. sudo apt-get install ubuntu-desktop
3. sudo rm /etc/X11/xorg.conf
4. echo 'nouveau' | sudo tee -a /etc/modules
(source: How can I uninstall a nvidia driver completely ?)

After that I could login, but unity didn't load, there was no launcher etc. It turn out I had to re-enable Unity plugin:

5. sudo apt-get install compizconfig-settings-manager
6. DISPLAY=:0 ccsm
7. Find the Unity plugin and enable it.
(source: Unity doesn't load, no Launcher, no Dash appears)

Found this solution in one of the Ubuntu forums. Steps 6 and 7 didn't apply to me so I just rebooted after and I could log in.

[Sudipan (sudipan)]

Any solution to this problem yet? Does installing any other version of Ubuntu helps? I was in process of installing Tensorflow with GPU support when I faced this situation. If anyone has found a solution, kindly let us know.

[Raymond Wan (rwan)]

For what it's worth, I'm using a GeForce GT 620 and Ubuntu 17.10 and I can't see the KDE login screen. I get a black screen but when I forcefully shut the computer down, I get the KDE logo as if I'm being logged out.

I upgraded to the latest NVidia driver from the NVidia web site (384.111 was the most recent one offered to me) and then booting back to 4.13.0-21 works, as others have observed. That is, I'm not using the standard NVidia driver packages.

I picked this option since it's easier to boot to a slightly older kernel; going back to an older NVidia driver is a bit harder...

I guess there is something between 4.13.0-21 and 4.13.0-25? I'm not doing any GPU work, so I don't know if that is fine.

[Raymond Wan (rwan)]

I spoke too soon! The recent 4.13.0-31 kernel appears to have fixed the problem for me, at least.

[J L (baduser)]

I know it might sound stupid, but how do I get a terminal session as I am not able to log in under X11 at all when this issue occurs. It boots straight away into bios.

[Bharata Kalbuaji (barbarian1803)]

I just upgraded the kernel to 4.13.0-31 but the problem still persists. I am using Kubuntu 17.10 and after I reboot, I can not start the X server. If I try to start it, it shows message that it can not find opengl 2. Any idea how to fix this? I install the nvidia driver via repository.

[Kai-Heng Feng (kaihengfeng)]

Please attach the output of `dkms status`.

[Frederik Fromm (mrpious)]

I have the same problem. Since 4.13.0-25 I can't login into desktop. Here is my dkms status output:
root@sam:~# dkms status
bbswitch, 0.8, 4.11.8-041108-generic, x86_64: installed
bbswitch, 0.8, 4.13.0-17-generic, x86_64: installed
bbswitch, 0.8, 4.13.0-21-generic, x86_64: installed
bbswitch, 0.8, 4.13.0-25-generic, x86_64: installed
bbswitch, 0.8, 4.13.0-31-generic, x86_64: installed
bbswitch, 0.8, 4.13.0-32-generic, x86_64: installed
nvidia-381, 381.22, 4.13.0-17-generic, x86_64: installed
nvidia-381, 381.22, 4.13.0-21-generic, x86_64: installed

nvidia stopped at 21... hmmm...

[Frederik Fromm (mrpious)]

OK, just installed the nvidia driver from http://de.download.nvidia.com/XFree86/Linux-x86_64/384.111/NVIDIA-Linux-x86_64-384.111.run by hitting ctrl+alt+F2 after booting into login screen to switch to a console, login to root, stop lightdm, execute the nvidia update. There were some errors during install but now I have

bbswitch, 0.8, 4.11.8-041108-generic, x86_64: installed
bbswitch, 0.8, 4.13.0-17-generic, x86_64: installed
bbswitch, 0.8, 4.13.0-21-generic, x86_64: installed
bbswitch, 0.8, 4.13.0-25-generic, x86_64: installed
bbswitch, 0.8, 4.13.0-31-generic, x86_64: installed
bbswitch, 0.8, 4.13.0-32-generic, x86_64: installed
nvidia-381, 381.22, 4.13.0-17-generic, x86_64: installed
nvidia-381, 381.22, 4.13.0-21-generic, x86_64: installed
nvidia, 384.111, 4.13.0-32-generic, x86_64: installed

and it's booting into desktop. Works for me...

[Kai-Heng Feng (kaihengfeng)]

Can `sudo dkms install nvidia-381/381.22 -k 4.13.0-25-generic` build the driver correctly?
If not, please attach the make.log in its error message.

[Frederik Fromm (mrpious)]

No didn't work:
dkms install nvidia-381/381.22 -k 4.13.0-25-generic

Kernel preparation unnecessary for this kernel. Skipping...

Building module:
cleaning build area...
'make' -j8 NV_EXCLUDE_BUILD_MODULES='' KERNEL_UNAME=4.13.0-25-generic IGNORE_XEN_PRESENCE=1 IGNORE_CC_MISMATCH=1 SYSSRC=/lib/modules/4.13.0-25-generic/build LD=/usr/bin/ld.bfd modules.....(bad exit status: 2)
Error! Bad return status for module build on kernel: 4.13.0-25-generic (x86_64)
Consult /var/lib/dkms/nvidia-381/381.22/build/make.log for more information.

[Frederik Fromm (mrpious)]

See make.log

[Kai-Heng Feng (kaihengfeng)]

Ok, just bootstrapped a Artful system, looks like you should remove nvidia-381 and use nvidia-384.

nvidia-384 is built correctly on 4.13.0-32-generic.

[Bharata Kalbuaji (barbarian1803)]

Is nvidia-390 working? I am using the latest kernel but cannot install nvidia-390

[Kai-Heng Feng (kaihengfeng)]

> On 2 Feb 2018, at 6:37 PM, Bharata Kalbuaji <email address hidden> wrote:
>
> Is nvidia-390 working? I am using the latest kernel but cannot install
> nvidia-390

Is the kernel from archive.ubuntu.com <http://archive.ubuntu.com/>?

>
> --
> You received this bug notification because you are subscribed to linux
> in Ubuntu.
> https://bugs.launchpad.net/bugs/1742302
>
> Title:
> Kernel 4.13.0-25 broke nvidia driver
>
> Status in linux package in Ubuntu:
> Confirmed
>
> Bug description:
> After updating to linux-image-4.13.0-25-generic the binary nvidia
> driver 387.22 cannot be built. As a result the system does not boot
> into the desktop environment (in my case, KDE).
>
> I've attached the driver build log.
>
> ProblemType: Bug
> DistroRelease: Ubuntu 17.10
> Package: linux-image-4.13.0-25-generic 4.13.0-25.29
> ProcVersionSignature: Ubuntu 4.13.0-21.24-generic 4.13.13
> Uname: Linux 4.13.0-21-generic x86_64
> NonfreeKernelModules: nvidia_uvm nvidia_drm nvidia_modeset nvidia
> ApportVersion: 2.20.7-0ubuntu3.7
> Architecture: amd64
> AudioDevicesInUse:
> USER PID ACCESS COMMAND
> /dev/snd/controlC0: lastique 3049 F.... pulseaudio
> /dev/snd/controlC2: lastique 3049 F.... pulseaudio
> /dev/snd/controlC3: lastique 3049 F.... pulseaudio
> /dev/snd/controlC1: lastique 3049 F.... pulseaudio
> CurrentDesktop: KDE
> Date: Wed Jan 10 02:00:46 2018
> InstallationDate: Installed on 2015-05-01 (984 days ago)
> InstallationMedia: Kubuntu 15.04 "Vivid Vervet" - Release amd64 (20150422)
> MachineType: System manufacturer System Product Name
> ProcFB:
>
> ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-4.13.0-21-generic root=UUID=a03f1835-52f9-4409-9da7-fe45770637ae ro quiet splash nomdmonddf nomdmonisw
> RelatedPackageVersions:
> linux-restricted-modules-4.13.0-21-generic N/A
> linux-backports-modules-4.13.0-21-generic N/A
> linux-firmware 1.169.1
> RfKill:
> 0: hci0: Bluetooth
> Soft blocked: no
> Hard blocked: no
> SourcePackage: linux
> UpgradeStatus: Upgraded to artful on 2017-11-03 (67 days ago)
> dmi.bios.date: 11/09/2012
> dmi.bios.vendor: American Megatrends Inc.
> dmi.bios.version: 3603
> dmi.board.asset.tag: To be filled by O.E.M.
> dmi.board.name: P8Z68-V PRO
> dmi.board.vendor: ASUSTeK Computer INC.
> dmi.board.version: Rev 1.xx
> dmi.chassis.asset.tag: Asset-1234567890
> dmi.chassis.type: 3
> dmi.chassis.vendor: Chassis Manufacture
> dmi.chassis.version: Chassis Version
> dmi.modalias: dmi:bvnAmericanMegatrendsInc.:bvr3603:bd11/09/2012:svnSystemmanufacturer:pnSystemProductName:pvrSystemVersion:rvnASUSTeKComputerINC.:rnP8Z68-VPRO:rvrRev1.xx:cvnChassisManufacture:ct3:cvrChassisVersion:
> dmi.product.family: To be filled by O.E.M.
> dmi.product.name: System Product Name
> dmi.product.version: System Version
> dmi.sys.vendor: System manufacturer
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1742302/+subscriptions

[Mark Valor (mkvalor)]

I can confirm Kai-Heng Feng's comment, installing apt package nvidia-384 works fine.

My system has kernel versions 4.13.0-25-generic and 4.13.0-32-generic, with the additional headers and extras packages for both. When I install packages nvidia-387 or nvidia-390, the next boot results in a black screen and the need to reboot to root console to fix this situation by uninstalling the package. However, installing nvidia-384 results in the ability to boot to gnome desktop normally. After doing so with either kernel, I can confirm the nvidia driver is loaded from the output of:

$ lspci -k | grep -A 2 -E "(VGA|3D)"

Then, if I run:

$ __GL_SYNC_TO_VBLANK=0 glxgears

I see that the framerate for my 1080 ti card is over 28K fps (vs a lower rate when the nouveau driver is loaded).

Finally -- I had no trouble booting to desktop after removing the apt package and installing the proprietary driver package version 390.25 (same minor version as the apt package) provided directly from nvidia's website for Linux 64-bit. I'm not recommending this, but pointing out that the driver scripts and software from the vendor do not have the problem exhibited by the apt package for versions 387 and 390, so there is most likely some problem with the post-install commands customized by the apt packages. I've reverted to the nvidia-384 apt package for the time being.

If you do try the installer from the nvidia website, however, be sure you have 'cc' configured to run gcc and not clang, or else the compilation of the kernel module will fail. (I couldn't get around this by setting env vars)

[Po-Hsu Lin (cypressyew)]

4.13 EOL.
Closing this bug.