Arch Linux

Cbhihe

Member

Registered: 2017-04-09

Posts: 219

[SOLVED] Issue w/ laptop built-in display on integrated Intel Graphics

[ Setup:  Arch on Dell xps15-9560 with 'gdm' on 'Xorg']

Hi: Two days ago, after a 2 month lapse in its updates, I updated a laptop's Archlinux system (with usual pacman -Syu)
Since then I get a vertically striated laptop display like this .
Note that everything else on the laptop seems to work fine.  Also the discrete GeForce GTX 1050 Nvidia GPU is not in use. The Nouveau driver is blacklisted and no Nvidia driver is installed.

$ more  /etc/modprobe.d/blacklist-nouveau.conf
blacklist nouveau

The screwed up display output appears after the Dell splash screen and a couple of slow black screen flickers. When I get to the hardly readable session login screen, I can log in fine. The temporarily fix to get rid of the striated output and recover a usable laptop screen consist in opening gdm's Display settings' window and either:

- select a different resolution, click apply then either confirm or cancel (it's indifferent) or
- go from single display to extended display or vice versa depending on whether an external display is connected to the laptop at boot or not, apply and, as before, either confirm or cancel the change (it's indifferent). 

From there everything apparently goes back to normal until next boot.  Whether I have an HDMI or a VGA external display attached or none makes no difference.

$ lspci | grep -i vga
00:02.0 VGA compatible controller: Intel Corporation HD Graphics 630 (rev 04)
$ grep -i upgraded /var/log/pacman.log | grep -e "^\[2020-11-02T20" | wc -l
522
$ grep -i upgraded /var/log/pacman.log | grep -i linux-headers | tail -2
[2020-08-29T09:59:04+0200] [ALPM] upgraded linux-headers (5.8.3.arch1-1 -> 5.8.5.arch1-1)
[2020-11-02T20:57:17+0100] [ALPM] upgraded linux-headers (5.8.5.arch1-1 -> 5.9.3.arch1-1)

I believe that the Intel driver in itself is not to blame as it (in particular) was not updated during that most recent update and there was no online chatter about this in recent weeks..

$ grep -i upgraded /var/log/pacman.log | grep -i video-intel | tail -2
[2020-04-25T23:12:18+0200] [ALPM] upgraded xf86-video-intel (1:2.99.917+904+gf2853658-1 -> 1:2.99.917+906+g846b53da-1)
[2020-05-12T08:44:58+0200] [ALPM] upgraded xf86-video-intel (1:2.99.917+906+g846b53da-1 -> 1:2.99.917+908+g7181c5a4-1)

This is the list of upgraded packages and the output of dmesg -xe. 
I could not detect any obvious issue except perhaps at delta [+0.001270] when /dev/nvme0n1p8 (root's and the kernel file system partition) is re-mounted. 
I see no obvious relationship though. Before the large upgrade the laptop never needed to re-mount root and also did not show a messed up stdout.   
Any help in the form of pointers or experience to share is most welcome...

Last edited by Cbhihe (2020-11-05 05:47:33)

---

I like strawberries, therefore I'm not a bot.

V1del

Forum Moderator

Registered: 2012-10-16

Posts: 21,659

Re: [SOLVED] Issue w/ laptop built-in display on integrated Intel Graphics

https://wiki.archlinux.org/index.php/Ke … _KMS_start to ensure the graphics driver is ready before attempts at using it are made, I've seen that reported a few times now with intel and the 5.9 kernel.

Also FWIW the xf86-video-intel driver is often problematic precisely because it's not really seeing any updates, so you might be generally better off trying without.

Last edited by V1del (2020-11-04 17:00:17)

Cbhihe

Member

Registered: 2017-04-09

Posts: 219

Re: [SOLVED] Issue w/ laptop built-in display on integrated Intel Graphics

Thank you. I was not aware of the possibility of NOT using the xf86-video-intel driver. I can try with early KMS start, but what should I do about the Intel driver ? A straightforward

$ sudo pacman -R xf86-video-intel

along with changing /etc/mkinitcpio.conf from 

MODULES="nvme crc32 libcrc32c crc32c_generic crc32c-intel crc32-pclmul crypto-crc32"

to

$grep "MODULES=" /etc/mkinitcpio.conf
MODULES="nvme intel_agp i915 crc32 libcrc32c crc32c_generic crc32c-intel crc32-pclmul crypto-crc32"
$ sudo mkinitcpio -p linux

... is the correct way of doing this ?   In particular I 'm not clear on whether placing nvme before or after intel_agp (to suppress the ACPI errors before module i915 is loaded) is important.
In fact, precisely to avoid ACPI errors, my boot loader's cmd-line is :

$ sudo grep -ie "linux_default" /etc/default/grub
GRUB_CMDLINE_LINUX_DEFAULT="quiet acpi_rev_override=1 acpi_osi=Linux"

... so just remove xf86-video-intel and insert new components into the array MODULES ?

(Sorry if the question is lame, but early KMS start is new to me. In fact up to now I had remained stuck at the X server way of managing display control .)

Last edited by Cbhihe (2020-11-05 05:45:18)

---

I like strawberries, therefore I'm not a bot.

V1del

Forum Moderator

Registered: 2012-10-16

Posts: 21,659

Re: [SOLVED] Issue w/ laptop built-in display on integrated Intel Graphics

Granted you haven't created any xorg config referencing it, yes that should suffice. Regarding the modules I doubt the order matters (... and you don't need intel_agp) here, this is simply about ensuring everything is up and available. But anyway how old is that mkinitpcio? Your "supposed" to use arrays e.g. ( i915 ) since quite a while (though afaik there should be a fallback still) and you can drop nvme since it should be a kernel built-in anyway so it doesn't really matter. But that's just a siderant and shouldn't inherently prevent things.

Cbhihe

Member

Registered: 2017-04-09

Posts: 219

Re: [SOLVED] Issue w/ laptop built-in display on integrated Intel Graphics

Hmmmh.... I think that the first Arch kernel was build in Oct. 2017 on that machine.

To tell you the truth, I just inherited 6 such boxes (only as far as their maintenance is concerned) so I am doing archeology on what was done and not done on each and as much updating as I can engineer without breaking things or reinstalling entire systems.

But anyway my guess is that mkinitpcio is pretty old, maybe even dating back to October 2017. In IT the worst is always possible ...    But even so there would be a silverlining to that: namely how durable and reliable a roll-out distro such as Arch is in terms of back-compatibility ... Not bad !
BTW, if rebuilding the initramfs does no modify its location, I imagine that there is also no need to regenerate the working grub menu ?

Checking the xorg conf for reference to Intel driver: ... all clear. Thanks again !

---

I like strawberries, therefore I'm not a bot.

V1del

Forum Moderator

Registered: 2012-10-16

Posts: 21,659

Re: [SOLVED] Issue w/ laptop built-in display on integrated Intel Graphics

yes the regenerated image should be picked up automatically.

phoric

Member

From: Seattle, WA

Registered: 2011-10-17

Posts: 22

Re: [SOLVED] Issue w/ laptop built-in display on integrated Intel Graphics

Just encountered the same thing on a Lenovo X1 Carbon 3rd Gen. After a kernel update, my display was glitched or tearing in Xorg.

While troubleshooting I noticed that quitting Xorg/LightDM and starting it again fixed the issue until the next reboot. And that when it failed the Xorg.0.log was showing the following error, and indicated that it was falling back to the vesa driver:

modeset(0): drmSetMaster failed: Device or resource busy

So I thought maybe it had something to do with Early KMS.

Adding "intel_agp i915" to MODULES in /etc/mkinitcpio.conf and regenerating fixed it for me.

Last edited by phoric (2020-11-11 22:50:57)