Blank screen when starting X after upgrading from 4.10 to 4.13.0-26

[Tim Starling]

I'm following the HWE kernels in Ubuntu 16.04, so my kernel was upgraded today from 4.10.0-42 to 4.13.0-26. After rebooting, the splash screen and disk encryption passphrase prompt showed correctly, but when X started, the screen went blank. Logs showed everything running as normal despite the blank screen. Rebooting with the "nomodeset" kernel parameter fixed the issue but is not suitable as a permanent workaround since it disables my external monitor. Rebooting back into 4.10 also fixed the issue.

I am using a ThinkPad X240 laptop, current HWE X.org packages. I tried both intel and fbdev X drivers, both gave a blank screen.

lsb_release: Ubuntu 16.04.3 LTS

dpkg-query -W:

linux-signed-image-generic-hwe-16.04 4.13.0.26.46
xserver-xorg-hwe-16.04 1:7.7+16ubuntu3~16.04.1
xserver-xorg-video-intel-hwe-16.04 2:2.99.917+git20170309-0ubuntu1~16.04.1
xserver-xorg-video-fbdev-hwe-16.04 1:0.4.4-1build6~16.04.1

In Xorg.0.log with fbdev, I saw:

[ 23.646] (EE) FBDEV(0): FBIOPUTCMAP: Invalid argument

repeated 239 times.

Typical kernel messages attached (from kern.log, reformatted to look like dmesg).

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

This bug is missing log files that will aid in diagnosing the problem. While running an Ubuntu kernel (not a mainline or third-party kernel) please enter the following command in a terminal window:

apport-collect 1742602

and then change the status of the bug to 'Confirmed'.

If, due to the nature of the issue you have encountered, you are unable to run this command, please add a comment stating that fact and change the bug status to 'Confirmed'.

This change has been made by an automated script, maintained by the Ubuntu Kernel Team.

[Tim Starling (tstarling)]

I ran apport-collect, selected "keep", and edited the resulting file to remove private information. But I couldn't find any way to upload such a file to an existing bug report.

[Tim Starling (tstarling)]

Here's the packed apport file.

[Marco Esposito (antani125)]

I also have the same problem:

black screen after reboot into 4.13.26.46
video driver: nvidia 387.26 (from CUDA ppa)

Restarting with 4.10 solves the problem

[William Horner (willhorner)]

I have just had the same issue.

Blank and frozen screen after booting in to 4.13.26 for the first time.
The ubuntu colours are visible and the 16.04 logo also appears bottom left.
The desktop was previously configured to run in 1920x1080. However, the screen changes to 4:3 ratio. No mouse or keys so unable to confirm the actual resolution. I assume it is a default.
I followed another report before finding this one so deleted the .config directory. Ubuntu booted up to a desktop in default resolution. However, the screen then locked when changing the resolution.
Ubuntu 16.04 desktop running inside KVM. So no special graphics installed to speak of. The image is encrypted with LVM inside the qcow file. Requires a password to boot after the grub menu.
KVM hypervisor still running 4.10. Took a deep breath, rebooted and brought the hypervisor up on 4.13. Still the same issues.
ssh on to the desktop machine continues to work fine even when the graphics are blank and frozen.
Followed the workaround above, booted in to 4.10 via grub and everything springs back to life as expected. Able to put the desktop back to 1920x1080. Further reboots into 4.10 maintain that resolution.

[Tim Starling (tstarling)]

I booted various mainline kernels from http://kernel.ubuntu.com/~kernel-ppa/mainline/?C=N;O=D

* 4.11.12: works
* 4.12.14: fails
* 4.13.16: fails
* 4.14.2: works
* 4.14.13: works
* 4.15-rc7: works

So it would seem that the issue was introduced upstream in 4.12 and fixed in 4.14.

[Joseph Salisbury (jsalisbury)]

We can perform a "Reverse" kernel bisect to identify the exact commit that fixes this issue. We first would need to narrow down the versions a little further.

Can you see if v4.14-rc1 fixes this issue:
http://kernel.ubuntu.com/~kernel-ppa/mainline/v4.14-rc1/

To perform the bisect, we need to identify the last bad kernel and the first good one.

[Tim Starling (tstarling)]

4.14-rc1 does fix the issue. So that just leaves 12600 commits to choose from, according to git log v4.13-rc7..v4.14-rc1. I guess I would need to recompile the kernel for each bisection iteration? Are we looking for a drm/i915 commit? There's not quite so many of those (281).

[Joseph Salisbury (jsalisbury)]

I can perform the bisect and build the kernels for you. Can you first test 4.13 final to confirm it has the bug? If it does, I'll start a bisect between 4.13 final and 4.14-rc1.

4.13 final can be downloaded from:
http://kernel.ubuntu.com/~kernel-ppa/mainline/v4.13/

[Tim Starling (tstarling)]

4.13.0-041300 (4.13 final): broken

[Joseph Salisbury (jsalisbury)]

I started a "Reverse" kernel bisect between v4.13 final and v4.14-rc1. The kernel bisect will require testing of about 13 test kernels.

I built the first test kernel, up to the following commit:
aae3dbb4776e7916b6cd442d00159bea27a695c1

The test kernel can be downloaded from:
http://kernel.ubuntu.com/~jsalisbury/lp1742602

Can you test that kernel and report back if it has the bug or not? I will build the next test kernel based on your test results.

Thanks in advance

[Tim Starling (tstarling)]

aae3dbb4776e7916b6cd442d00159bea27a695c1: fixed

[Joseph Salisbury (jsalisbury)]

I built the next test kernel, up to the following commit:
bf1d6b2c76eda86159519bf5c427b1fa8f51f733

The test kernel can be downloaded from:
http://kernel.ubuntu.com/~jsalisbury/lp1742602

Can you test that kernel and report back if it has the bug or not? I will build the next test kernel based on your test results.

Thanks in advance

[Tim Starling (tstarling)]

bf1d6b2c76eda86159519bf5c427b1fa8f51f733: fixed

Curiously, there was a kernel error with this kernel which I don't think was there before:

[ 5.505217] [drm:intel_set_cpu_fifo_underrun_reporting [i915]] *ERROR* uncleared fifo underrun on pipe A
[ 5.505308] [drm:intel_cpu_fifo_underrun_irq_handler [i915]] *ERROR* CPU pipe A FIFO underrun

[Joseph Salisbury (jsalisbury)]

I built the next test kernel, up to the following commit:
edc2988c548db05e33b921fed15821010bc74895

The test kernel can be downloaded from:
http://kernel.ubuntu.com/~jsalisbury/lp1742602

Can you test that kernel and report back if it has the bug or not? I will build the next test kernel based on your test results.

Thanks in advance

[Tim Starling (tstarling)]

edc2988c548db05e33b921fed15821010bc74895: fixed

[Joseph Salisbury (jsalisbury)]

I built the next test kernel, up to the following commit:
e5fa05b96b02f45a91a364d2121451a4da05cd84

The test kernel can be downloaded from:
http://kernel.ubuntu.com/~jsalisbury/lp1742602

Can you test that kernel and report back if it has the bug or not? I will build the next test kernel based on your test results.

Thanks in advance

[Tim Starling (tstarling)]

e5fa05b96b02f45a91a364d2121451a4da05cd84: fixed

I discovered that the issue is not reproducible with the external monitor unplugged, i.e. the trigger seems to be lightdm starting in dual monitor mode. That invalidates the test I did at bf1d6b2c76eda86159519bf5c427b1fa8f51f733 , which I think was with the monitor unplugged. But since e5fa05b96b02f45a91a364d2121451a4da05cd84 is definitely fixed, i.e. lightdm starts correctly even with the monitor plugged in, immediately after confirming the bug in a known-bad kernel, I think we're still on the same range, so we don't need to retest any kernels.

[Joseph Salisbury (jsalisbury)]

I built the next test kernel, up to the following commit:
b9c55b6e2cc4369b0688961fa5de0e057f3ec0c4

The test kernel can be downloaded from:
http://kernel.ubuntu.com/~jsalisbury/lp1742602

Can you test that kernel and report back if it has the bug or not? I will build the next test kernel based on your test results.

Thanks in advance

[Tim Starling (tstarling)]

b9c55b6e2cc4369b0688961fa5de0e057f3ec0c4: fixed

I noticed that in the broken kernel, plymouth does not switch on the external monitor. Normally plymouth shows a cloned display.

Looking at the remaining commits, there was a merge of tag "drm-intel-next-2017-07-17" which had a very large number of i915 commits. You may want to manually choose commits either side of that merge, to hurry things along, since I don't think git-bisect will descend into the branch until the merge commit is fully isolated.

[Tim Starling (tstarling)]

In other words, I'd like packaged kernels for 2d62c799f8ffac4f7ffba6a4e7f148827dfc24c7 and 5771a8c08880cdca3bfb4a3fc6d309d6bba20877 . The expectation is that that will confirm that the fix came from drm-intel. Assuming that's confirmed then we would have to do a bisection of the drm-intel-next branch back to the merge base (whatever that is). Maybe the tag v4.12-rc3?

Next week I will be in San Francisco, which will be good for timezones, but bad for spare time and access to external monitors.

[William Horner (willhorner)]

Thanks for all this good testing Tim and Joseph.
If I can help to keep things moving while Tim is travelling, let me know, although might need a little guidance initially... if there are some good instructions you can point me to on how to get the test kernel in and to boot...
Tim: to your observation I see the issue (assuming it is the same one) with a VM with no dual monitor. But I do see the issue fix itself and reappear based on the screen resolution. Perhaps does that observation shed any further light on it or fit with your external monitor results?

[Joseph Salisbury (jsalisbury)]

I built a test kernel up to including:

2d62c799f8ffac4f7ffba6a4e7f148827dfc24c7

This is the merge tag. If this kernel is good, I'll build a test kernel up to the commit before this tag.

This kernel is available from:
http://kernel.ubuntu.com/~jsalisbury/lp1742602

[Furton Gyula (furtongy)]

The blank secreen problem still standing in 4.13.0-32 kernel too.
Linux Mint, 18.2, Cinnamon

[JuanJo Ciarlante (jjo)]

FYI this is also happening for me, LTS 16.04.3 + HWE (kernel and xorg pkgs),
Thinkpad x270 w/ Integrated Graphics Chipset: Intel(R) HD Graphics 620.

[Gamunu Balagalla (gamunu)]

Looks like this is occurring on hybrid graphics systems. Not only Ubuntu, Arch, Fedora all affected by this bug.
I have two Dell Latitude 5580 and Dell Latitude e5570 laptops. One has Intel+Nvidia hybrid graphic system and other one has Intel+ATi hybrid graphic system. Both having the same issue HDMi goes to sleep immediately after X server starts.
The only desktop environment I found fixed this is Linux Mint Cinnamon and Mate both initially had the issue but after updating issue is gone. Looks like the bug is in the kernel.

[Roland Lohner (loczaj)]

The blank screen problem is still present in my case on Ubuntu 18.04 using a Dell 5570 machine.
Upgrading to kernel 4.17.2-041702 has solved the problem.

[Joseph Salisbury (jsalisbury)]

Would it be possible for you to test the latest 4.15 upstream stable kernel?

[0] http://kernel.ubuntu.com/~kernel-ppa/mainline/v4.15.18/

[Tim Starling (tstarling)]

Sorry about not finishing the bisection. I was dubious about the prospect of finding a nice clean patch to cherry pick in that drm-intel branch. But this is superseded now by the upgrade of linux-generic-hwe-16.04 to 4.15, which does not have the same issue.

I recommend closing this bug as fixed, and discussing the unrelated blank screen issues elsewhere. A blank screen is a failure mode, not a specific bug. Anyone seeing a blank screen in 4.14 or later is not experiencing the same issue. This was an i915 bug in the 4.13 kernel.

[Andy Whitcroft (apw)]

This bug was nominated against a series that is no longer supported, ie artful. The bug task representing the artful nomination is being closed as Won't Fix.

This change has been made by an automated script, maintained by the Ubuntu Kernel Team.

[Daniel (enoch85)]

I am also affected. Running Ubuntu 18.04 MATE i686 with 1 VGA and 1 DP.

The DP screen goes blank randomly for 1 second and then comes back.

Running kernel 4.10.17 from mainline.

Any fix for this sometime soon?

Thanks!

[Matteus (galanonym)]

After reinstalling to 20.04 this bug no longer affects me.