search

StarLabsLtd / firmware

78 stars 5 forks source link

[*] Arch kernels enable simple drm without support for coreboots framebuffer #181

Open isomorphian opened 4 months ago

isomorphian commented 4 months ago

I have just received my StarLite V and so far have had nothing but issues. I have 5 other intel-based devices running Hyprland with no issues. On this device, running Hyprland, any graphical application will crash after about 30 seconds.

In addition, with either Gnome or KDE on wayland, the desktop environment will flicker uncontrollably before crashing after about 30 seconds.

Also, I am experiencing random kernel panics running the stock Arch kernel.

I have had somewhat better experience running X11. Apps will not crash, although every so often, upon starting the X server, the device will lock up completely, and not even allow me to switch virtual terminals. The only way out of this situation is to cut the power by holding the power button.

I am on the latest firmware that was just released to solve the intermittent touchscreen issue.

Sean-StarLabs commented 4 months ago

kernel 6.9 by any chance?

isomorphian commented 4 months ago

kernel 6.9 by any chance?

Yes, 6.9.7-arch1-1 installed with the "linux" package on arch

Sean-StarLabs commented 4 months ago

6.9 has bugs; should be fine if you use 6.8 instead

thmichel commented 4 months ago

Works well with Fedora Kernel 6.9.7 by the way.

isomorphian commented 4 months ago

Thank you for the information. I'm out of town for a few days but when I get back I will try downgrading my kernel.

I'm glad to hear things are working better on Fedora. I'd prefer to stay on Arch because that's what I'm most familiar with, but I'm glad to know that's an option.

stronnag commented 4 months ago

It continues with the latest Arch updates (6.9.8-zen1-1-zen), the journal reporting:

Jul 06 08:26:55 piglet kernel: i915 0000:00:02.0: [drm] *ERROR* Fault errors on pipe A: 0x00000080

Though not as frequently as previously.

PhilT commented 4 months ago

I'm using Wayland on NixOS with Kernel 6.9.7. No flickering problems that I've noticed so far. I do have animations off though but I'll check it out with them on.

isomorphian commented 4 months ago

Seems that the problem is in libwayland-client.so

I noticed that applications only crash after being interacted with via touchscreen. As long as I do not touch the screen at all, there are no problems. But if I interact with applications via touchscreen, they will continue to function for about 30 seconds before crashing.

I recorded a crash in Rnote, and a crash in Firefox.

Dmesg reports this message upon Rnote crashing: rnote[843]: segfault at 603aa950dc8d ip 0000790883a122af sp 00007ffe3dda1e40 error 4 in libwayland-client.so.0.23.0[790883a0d000+6000] likely on CPU 0 (core 0, socket 0)

Dmesg reports another, similar message upon Firefox crashing: traps: crashreporter[1349] general protection fault ip:764328dcfcc4 sp:7ffde78d7ff8 error:0 in libwayland-client.so.0.23.0[764328dce000+6000]

isomorphian commented 4 months ago

apologies for closing and reopening. I have never submitted an issue before, I did not realize how it worked.

isomorphian commented 4 months ago

Ack! Sorry !

isomorphian commented 4 months ago

I tried all three officially supported Arch Linux kernels:

linux 6.9.8 linux-zen 6.9.8 linux-lts 6.6.38

None of them make any difference.

Is it possible that there is some firmware that this device requires, which is not included in the linux-firmware package?

isomorphian commented 4 months ago

Interestingly, this does not affect all applications on my machine, but it does affect most of them.

Apps that crash upon touch input: Firefox Rnote LibreOffice Pavucontrol Thunar File Manager PCManFM (GTK3 version)

Apps that do not crash upon touch input: KeepassXC Blueman GNU Octave PCManFM (QT6 version) qBittorrent Discord

isomorphian commented 4 months ago

Just finished recovering my system from a kernel panic that interrupted a system upgrade. This machine really is not happy to be running Arch Linux.

isomorphian commented 4 months ago

I think the kernel panics during system upgrade could be an upstream issue with Arch, according to this forum post:

https://forum.endeavouros.com/t/kernel-panic-during-update-broke-my-system-thanks-god-and-linux-community-for-the-arch-chroot/51839/20?page=3

(they are talking about Endeavour, not Arch, but close enough to make me think it could be upstream issue)

Although I have never had this happen on any of my other machines that are also running Arch (and those installs are way more bloated than my StarLite) so I don't know.

isomorphian commented 4 months ago

I made the mistake of enabling gdm.service to see if starting Gnome that way would work. But I should have started it instead of enabling. Now I am trapped inside a completely unusable Gnome that freezes my system after a few seconds of runtime. Impossible to get to a terminal. Have to boot into recovery USB for the 50th time since I got this machine.

isomorphian commented 4 months ago

Both Gnome and KDE Plasma on wayland are completely unusable, only way to escape is to power off. This is not good. Wayland is the default option.

They start fine on X11 but have issues with the pointer not tracking the pen correctly.

isomorphian commented 4 months ago

i3 and awesomewm seem to work fine on X11 although the pen is lacking pressure-sensitivity.

Hyprland is broken, but I say that's probably their fault anyways.

Interestingly, Sway seems to work perfectly! :)

diggit commented 4 months ago

@isomorphian I am also on archlinux with same issues. In KDE/Gnome on wayland, workaround is to disable "ghost" output. Then the flickering stopped. SDDM probably also has some way of output configuration to apply the same same workaround.

Due to some reason, these DEs and SDDM (and probably more) see this Unknown-1 output and try to use it.

Contents of /sys/class/drm/, notice the coreboot8/simple-framebuffer.0

card0 -> ../../devices/platform/BOOT0000:00/coreboot8/simple-framebuffer.0/drm/card0/
card0-Unknown-1 -> ../../devices/platform/BOOT0000:00/coreboot8/simple-framebuffer.0/drm/card0/card0-Unknown-1/
card1 -> ../../devices/pci0000:00/0000:00:02.0/drm/card1/
card1-DP-1 -> ../../devices/pci0000:00/0000:00:02.0/drm/card1/card1-DP-1/
card1-DP-2 -> ../../devices/pci0000:00/0000:00:02.0/drm/card1/card1-DP-2/
card1-eDP-1 -> ../../devices/pci0000:00/0000:00:02.0/drm/card1/card1-eDP-1/
card1-HDMI-A-1 -> ../../devices/pci0000:00/0000:00:02.0/drm/card1/card1-HDMI-A-1/
card1-HDMI-A-2 -> ../../devices/pci0000:00/0000:00:02.0/drm/card1/card1-HDMI-A-2/
card1-HDMI-A-3 -> ../../devices/pci0000:00/0000:00:02.0/drm/card1/card1-HDMI-A-3/
renderD128 -> ../../devices/pci0000:00/0000:00:02.0/drm/renderD128/
isomorphian commented 4 months ago

@isomorphian

I am also on archlinux with same issues.

In KDE/Gnome on wayland, workaround is to disable "ghost" output. Then the flickering stopped. SDDM probably also has some way of output configuration to apply the same same workaround.

Due to some reason, these DEs and SDDM (and probably more) see this Unknown-1 output and try to use it.

Contents of /sys/class/drm/, notice the coreboot8/simple-framebuffer.0


card0 -> ../../devices/platform/BOOT0000:00/coreboot8/simple-framebuffer.0/drm/card0/

card0-Unknown-1 -> ../../devices/platform/BOOT0000:00/coreboot8/simple-framebuffer.0/drm/card0/card0-Unknown-1/

card1 -> ../../devices/pci0000:00/0000:00:02.0/drm/card1/

card1-DP-1 -> ../../devices/pci0000:00/0000:00:02.0/drm/card1/card1-DP-1/

card1-DP-2 -> ../../devices/pci0000:00/0000:00:02.0/drm/card1/card1-DP-2/

card1-eDP-1 -> ../../devices/pci0000:00/0000:00:02.0/drm/card1/card1-eDP-1/

card1-HDMI-A-1 -> ../../devices/pci0000:00/0000:00:02.0/drm/card1/card1-HDMI-A-1/

card1-HDMI-A-2 -> ../../devices/pci0000:00/0000:00:02.0/drm/card1/card1-HDMI-A-2/

card1-HDMI-A-3 -> ../../devices/pci0000:00/0000:00:02.0/drm/card1/card1-HDMI-A-3/

renderD128 -> ../../devices/pci0000:00/0000:00:02.0/drm/renderD128/

Interesting. I also noticed that Unknown-1 device, and I had assumed it had to do with the touchscreen. I'll experiment with this when I get home.

TheJackiMonster commented 4 months ago

I might have found a cause for this... potentially. I had some issue running into a framebuffer problem during boot (which appeared in log from dmesg). So I added the i915 module to /etc/mkinitcpio.conf for early KMS start. Now my issue during boot is gone but I encounter heavy flickering in my Wayland session. Didn't happen before that.

So my guess is that it's coming from this module...

_Edit: It seems like using the linux-lts kernel solves the flickering for me when using the early KMS. Otherwise I suspect it to be related to the PSR feature with selective regions. But disabling the option didn't do much for me._

stronnag commented 4 months ago

The Arch Testing / Zen kernels (6.9.x) do not work well on the Starlite V. The AUR LQX kernel (6.9.9) works very, very well (Wayland, Gnome). No flickering etc. Touch and GJS OSK are stable and predictable. Bit of pain to build the kernel from source, but it was worth it.

isomorphian commented 4 months ago

@stronnag Thank you! This solved all my problems.

I wonder what's different in the Liquorix kernel vs the official Arch kernels. Whatever it is, is working great!

TheJackiMonster commented 4 months ago

I wonder what's different in the Liquorix kernel vs the official Arch kernels. Whatever it is, is working great!

Tested the linux-rt kernel now and it seems to work fine. So that might be a potential option which is pre-built in the official repositories.

stronnag commented 4 months ago

I wonder what's different in the Liquorix kernel vs the official Arch kernels. Whatever it is, is working great!

Tested the linux-rt kernel now and it seems to work fine. So that might be a potential option which is pre-built in the official repositories.

Any Arch kernel prior to 6.9 will work. I'd prefer LTS or 6.8.9 over -rt any day,

isomorphian commented 4 months ago

I wonder what's different in the Liquorix kernel vs the official Arch kernels. Whatever it is, is working great!

Tested the linux-rt kernel now and it seems to work fine. So that might be a potential option which is pre-built in the official repositories.

Any Arch kernel prior to 6.9 will work. I'd prefer LTS or 6.8.9 over -rt any day,

Last time I tested LTS it had the same issues, but I'll give it another shot.

andrewwebber commented 4 months ago

Just adding my two pence:

Use Arch BTW with Sway/Warbar. Installed using archinstall and selecting minimal + sway. No complaints as Wayland works no less usable than on my desktop hardwares (e.g. full screen Slack crashes on both).

However I found it impossible to get VLC Player working on the Starlite. As a workaround mpv was recommended and worked perfectly.

DavidG-Ghb commented 2 months ago

Starlite V on Manjaro. Kernel 6.6 LTS works. 6.8 was working, but Manjaro have dropped support, so had to revert to 6.6. Current kernels 6.9, 6.10 and the experimental 6.11 all have the black flickering issues with Wayland. Slightly concerning that this issue remains in newer kernels, which suggests there could be issues with long term support for the Starlite V.

A workaround would be welcome.

Sean-StarLabs commented 2 months ago

On 24.08?

stronnag commented 2 months ago

This still occurs with Arch 6.9. 6.10 kernels on 24.08.

It does not occur with the Arch "lqx" 6.10 kernels (which, AFAIK, are build with Debian build flags). So it is probably an Arch (and derivatives) build issue rather than an upstream kernel issue.

DavidG-Ghb commented 2 months ago

Yes, tested on 24.08

isomorphian commented 2 months ago

I've been using lqx from the AUR with no problems, but the official arch kernels all have the same problems as when I originally posted this issue, with many graphical apps and environments causing screen flickering and kernel panics.

MRDGH2821 commented 1 month ago

In my distro hopping journey, I faced problems in Endeavour OS (which is arch based) So I fixed it in this way:

https://forum.endeavouros.com/t/unknown-1-display-causing-screen-flickering-on-main-display/60435

https://forum.endeavouros.com/t/after-updating-kernal-framebuffer-coreboot8-probe-with-driver-framebuffer-failed-with-error-17/60389

And later switched to Fedora 40 (which is working fine with kernel 6.10.10).

I feel like this is a common issue with coreboot related systems🤔

Waiting for a permanent fix because I wanna use Endeavour OS

diggit commented 1 month ago

@MRDGH2821 yeah that is what I wrote in my comment above https://github.com/StarLabsLtd/firmware/issues/181#issuecomment-2226431176

It's interesting, that sway (probably all wlroots based wayland compositors) don't even list nor enable this output, so no issues there.

MRDGH2821 commented 1 month ago

I freshly installed Endeavour OS in online mode & it installed kernel v6.11.2-arch1-1. So far no flickering & doesn't crash when external monitors are plugged in.

In other words, I didn't do any special steps. It just worked out of the box.

I feel that others can try kernel v6.11 from default source/upstream.

TheJackiMonster commented 1 month ago

I freshly installed Endeavour OS in online mode & it installed kernel v6.11.2-arch1-1. So far no flickering & doesn't crash when external monitors are plugged in.

In other words, I didn't do any special steps. It just worked out of the box.

I feel that others can try kernel v6.11 from default source/upstream.

Doesn't work for me on Archlinux. Using 6.11.2-arch1-1 still gives me flickering as soon as GDM (which is using Wayland) appears and logging into GNOME as Wayland session even results in it freezing so I have to press the power button for shutdown.

MRDGH2821 commented 1 month ago

Doesn't work for me on Archlinux. Using 6.11.2-arch1-1 still gives me flickering as soon as GDM (which is using Wayland) appears and logging into GNOME as Wayland session even results in it freezing so I have to press the power button for shutdown.

I see this inside /proc/cmdline

initrd=\0b6dd5cb07694ca283ac1472cda598c8\6.11.2-arch1-1\initrd nvme_load=YES nowatchdog rw root=UUID=a56e2122-bf54-4394-a71a-25f540c4cd52 rw root=UUID=a56e2122-bf54-4394-a71a-25f540c4cd52 systemd.machine_id=0b6dd5cb07694ca283ac1472cda598c8

Referred this https://wiki.archlinux.org/title/Kernel_parameters

I'm using EndeavourOS KDE Plasma 6.1.5 & Wayland if that matters

stronnag commented 1 month ago

I freshly installed Endeavour OS in online mode & it installed kernel v6.11.2-arch1-1. So far no flickering & doesn't crash when external monitors are plugged in. In other words, I didn't do any special steps. It just worked out of the box. I feel that others can try kernel v6.11 from default source/upstream.

Doesn't work for me on Archlinux. Using 6.11.2-arch1-1 still gives me flickering as soon as GDM (which is using Wayland) appears and logging into GNOME as Wayland session even results in it freezing so I have to press the power button for shutdown.

It does not freeze the computer, it freezes the screen. It is possible to log in via ssh and reboot into a known good (lqx?) kernel.

mixmastamyk commented 1 month ago

Been using Fedora 40 with Phosh (Wayland) for months without issue (since the Mint it came with was too old). First kernel was 6.8 I think, now on 6.10.x. Have updated the firmware twice, now on 24.08.

The root problem you're having does not appear to be Wayland. Might want to edit the title of this issue.

MRDGH2821 commented 1 month ago

I freshly installed Endeavour OS in online mode & it installed kernel v6.11.2-arch1-1. So far no flickering & doesn't crash when external monitors are plugged in.

In other words, I didn't do any special steps. It just worked out of the box.

I feel that others can try kernel v6.11 from default source/upstream.

My bad, next day I started my tablet & Framebuffer error came back.

It just works after fresh install. 2nd boot onwards it won't boot into login screen.

I switched to lqx kernel.

Sean-StarLabs commented 1 month ago

What version of wayland?

Sean-StarLabs commented 1 month ago

From my testing, d6de2690dd547a014309086a9cc435e2b58f091e does it, but let me know if not

stronnag commented 1 month ago

Lets see ...

# fwupdmgr --force install 24.09-coreboot-I5.cab
Decompressing…           [            /                          ]
No supported devices found

Anything else required?

stronnag commented 1 month ago

Lets see ...

# fwupdmgr --force install 24.09-coreboot-I5.cab
Decompressing…           [            /                          ]
No supported devices found

Anything else required?

Same result with Arch LTS (6.6.56) and Ach LQX (6.11.3) kernels.

Sean-StarLabs commented 1 month ago

Anything else required?

Don't know what you're using - see https://support.starlabs.systems/kb/firmware/getting-started

stronnag commented 1 month ago

As posted I'm using fwupdmgr and your latest publicly published firmware. So lets use flashrom instead ...

flashrom 1.4.0 (git:v1.4.0) on Linux 6.6.56-1-lts (x86_64)
flashrom is free software, get the source code at https://flashrom.org

Opened /dev/mtd0 successfully
Found Programmer flash chip "Opaque flash chip" (16384 kB, Programmer-specific) on internal.
Reading ich descriptor... done.
Using region: "bios".
linux_mtd_wp_read_cfg: ioctl: Operation not supported
Failed to unlock flash status reg with wp support.
Reading old flash chip contents... done.
Erase/write done from 600000 to ffffff

OK, messages not so encouraging, but let's see ...

It boots. 

$ cat  /sys/class/dmi/id/ec_firmware_release 
24.9

It appears to be the latest firmware.

It still flickers. Not quite as much as before, but still "not fit for purpose". Not fixed.

Sean-StarLabs commented 1 month ago

Okay, can you post:

stronnag commented 1 month ago

I will attempt to record a video, probably tomorrow.

MRDGH2821 commented 1 month ago

Just flashed 24.09 firmware. Still suffering from: framebuffer coreboot8: probe with driver framebuffer failed with error -17 I haven't done any of the steps in this, so can't confirm flickering issue.

Additionally I installed 2 kernels and here are my results: Kernel Version Boots? Additional notes
core/linux 6.11.3.arch-1 No Stuck on coreboot error
core/linux-lts 6.6.56-1 Yes While shutting down, it took long time (approx 3 mins)
liquorix/linux-lqx 6.11.3.lqx2-1 Yes
aur/linux-xanmod 6.11.3-2 Yes Takes 5+ hours to compile & install
aur/linux-cachyos 6.11.3-2 Yes Takes 6+ hours to compile & install

I'm using Endeavour OS, with no modifications done to kernel params. At max I installed binder kernel module to get Waydroid running (which unfortunately didn't run in any of the kernels)

stronnag commented 1 month ago

Interesting ...

Working Arch Linux-LQX kernel:

Oct 17 17:51:03 piglet kernel: efifb: probing for efifb
Oct 17 17:51:03 piglet kernel: efifb: framebuffer at 0x90000000, using 5400k, total 5400k
Oct 17 17:51:03 piglet kernel: efifb: mode is 1440x960x32, linelength=5760, pages=1
Oct 17 17:51:03 piglet kernel: efifb: scrolling: redraw
Oct 17 17:51:03 piglet kernel: efifb: Truecolor: size=0:8:8:8, shift=0:16:8:0

Flickering unusable Arch Linux / Arch Linux Zen kernels:

Oct 17 17:34:32 piglet kernel: [drm] Initialized simpledrm 1.0.0 for simple-framebuffer.0 on minor 0
Oct 17 17:34:32 piglet kernel: fbcon: Deferring console take-over
Oct 17 17:34:32 piglet kernel: simple-framebuffer simple-framebuffer.0: [drm] fb0: simpledrmdrmfb frame buffer device

[Additional note] None of the other three Intel GPU equipped computers here have any problem with current Arch (Zen or otherwise) kernels and simple-framebuffer. These working machines include a Starlite Mk4. It is also the case that none of the working machines use coreboot.

bkfirmen commented 1 month ago

I can also confirm that on arch there is flickering (Im using Gnome). Used EndeavourOS before and there was an issue after an update that I installed yesterday. If you logged out of Gnome you cant log in again into your user account. After a reboot also gdm throwed an error message. A fresh install of endeavour os also didnt fixed it. Logged in the first time worked but after that it was the same as before.