Arch Linux User Repository

Currently, kmscon-patched-git results in cpu usage of 100%.

PR 105 corrects this.

While matters are being straightened out upstream, you can apply the PR by editing the PKGBUILD to change the source to:

source=("git+https://github.com/Aetf/kmscon.git#commit=db524e9e0fc38d06faba1430dfeb9637649fec1b")

EDIT: SOLVED! nvidia_drm.fbdev=0 does the trick and allows efifb to be used, which is what was always used on my system and works fine. I was confused about the kernel option, just used fbdev=0 without the module prefix (and yes it DID feel wrong but I did it anyway lol). nvidia_drm.fbdev is simply broken, "you had one job" style.

Original post:

No matter what I try (fbdev=0, fbdev=1, --hwaccel, --no-hwaccel, not passing any hwaccel option, --drm, --no-drm, not passing any drm option, in all possible combinations), I can't get kmscon to work.

GeForce GTX 960 nvidia-dkms 550.54.14-1

lut 28 14:02:51 berbelek systemd[1]: Created slice Slice /system/kmsconvt.
lut 28 14:02:51 berbelek kmscon[2482]: [0000.000000] NOTICE: kmscon Revision v9.0.0-10-g20a2fc4 Dec 15 2023 16:20:47 (log_print_init() in ../kmscon/src/shl_log.c:509)
lut 28 14:02:51 berbelek kmscon[2482]: [0000.000085] INFO: conf: reading config file /etc/kmscon/kmscon.conf (conf_parse_file() in ../kmscon/src/conf.c:547)
lut 28 14:02:51 berbelek kmscon[2482]: [0000.004467] INFO: systemd: system booted with systemd, enabling multi-seat support (uterm_sd_new() in ../kmscon/src/uterm_systemd.c:67)
lut 28 14:02:51 berbelek kmscon[2482]: [0000.008422] NOTICE: vt: using tty /dev/tty4 (open_tty() in ../kmscon/src/uterm_vt.c:285)
lut 28 14:02:51 berbelek kmscon[2482]: [0000.101279] INFO: drm_shared: new drm device via /dev/dri/card0 (uterm_drm_video_init() in ../kmscon/src/uterm_drm_shared.c:501)
lut 28 14:02:51 berbelek kmscon[2482]: [0000.127123] INFO: video: new device 0x5715205a7820 (uterm_video_new() in ../kmscon/src/uterm_video.c:576)
lut 28 14:02:51 berbelek kmscon[2482]: [0000.127554] INFO: ignoring video device /dev/fb0 on seat seat0 as it is a DRM-fbdev device (app_seat_gpu_is_ignored() in ../kmscon/src/kmscon_main.c:287)
lut 28 14:02:52 berbelek kmscon[2482]: [0000.396007] INFO: video: new display 0x571520a2c650 (display_new() in ../kmscon/src/uterm_video.c:235)
lut 28 14:02:52 berbelek kmscon[2482]: [0000.396024] INFO: drm_shared: display 0x571520a2c650 DPMS is ON (bind_display() in ../kmscon/src/uterm_drm_shared.c:632)
lut 28 14:02:52 berbelek kmscon[2482]: [0000.397852] INFO: video: new display 0x571520a2b090 (display_new() in ../kmscon/src/uterm_video.c:235)
lut 28 14:02:52 berbelek kmscon[2482]: [0000.397867] INFO: drm_shared: display 0x571520a2b090 DPMS is ON (bind_display() in ../kmscon/src/uterm_drm_shared.c:632)
lut 28 14:02:52 berbelek kmscon[2482]: [0000.398156] INFO: uterm_drm3d_video: activating display 0x571520a2b090 to 1920x1200 (display_activate() in ../kmscon/src/uterm_drm3d_video.c:155)
lut 28 14:02:52 berbelek kmscon[2482]: [0000.399763] ERROR: uterm_drm3d_video: cannot create gbm surface (38): Function not implemented (display_activate() in ../kmscon/src/uterm_drm3d_video.c:171)
lut 28 14:02:52 berbelek kmscon[2482]: [0000.399773] INFO: uterm_drm3d_video: activating display 0x571520a2c650 to 1280x1024 (display_activate() in ../kmscon/src/uterm_drm3d_video.c:155)
lut 28 14:02:52 berbelek kmscon[2482]: [0000.569390] ERROR: uterm_drm3d_video: cannot create gbm surface (38): Function not implemented (display_activate() in ../kmscon/src/uterm_drm3d_video.c:171)

I enabled --debug in the service, and depending on the combination of options and kernel parameters there is or isn't this ERROR message pertaining to display_activate() not being implemented, but in any case the directly experienced symptom is consistent: switching to a ksmcon-managed virtual console with e.g. Ctrl+Alt+F4 always results in black screen NO SIGNAL GOODBYE on both monitors.

It may be worth noting that I am trying kmscon because the same symptom also manifests with the regular linux console - I simply don't HAVE it since driver 545 (last one that works is 535, and I don't want to be stuck with that). Could it be that the regular console and kmscon fail for the same deeper reason somewhere in the driver?

Don't understand why exactly but makepkg was failing for me (linking anything would fail because symbol "main" not found). Fixed by cd src/builddir; meson configure -Db_lto; cd ../.. ; makepkg -si.

It seems somehow disabling LTO fixed it for me (gcc not understanding with the linker?)

Some more news and update to my previous comment.

Just now, Antonio Rojas, an official Arch package maintainer has made a big "cleanup" in lot of official community packages. The were removed from the "community" repo, and transfered to the AUR, were they are now orphan. Maybe awaiting adoption.

This has happened to the, until today, official Archlinux packages for "kmscon" and "libtsm".

They were transfered to the AUR by Antonio Rojas, and are now orphaned as,

https://aur.archlinux.org/packages/kmscon
https://aur.archlinux.org/packages/libtsm

So maybe now the maintainer of this same PKGBUILD "kmscon-patched-git" would want to adopt them to keep everything under the same person. And clear to the end user.

Interesting news for us Archlinux console lovers (and for me rotated portrait monitor freak)

Found this repo, https://github.com/MacSlow/kmscon

This is my fork of kmscon, where I work on output-rotation (working already)
mouse copy&paste support (also working already) and gyro-sensor reading (wor
too, needs a bit more testing).

My branch is meant to be rebased against Atef's branch once I'm done with
everything.

It does seem to be quite active since January 2023. And he asks for testing. But it seems only Ubuntu .debs are available. Would be nice if we could test his work on Archlinux.

His name is Mirco Mueller and he talks about his work in his blog,

• https://blog.macslow.org/

And posts a lot of videos of kmscon on youtube

• https://www.youtube.com/watch?v=ybjmUFIXfUU&list=PLXvTBWcnTI1OHF7-Op-ASYsQKKg1TYb5Z

As a reminder to everyone interested in kmscon,

The original kmscon is by David Herrmann at https://www.freedesktop.org/wiki/Software/kmscon/

There has been no update or activity in the last 10 years, https://github.com/dvdhrm/kmscon, or https://cgit.freedesktop.org/~dvdhrm/kmscon/log/.

The official Archlinux package currently also uses this repo by Aetf, https://github.com/Aetf/kmscon, repo for the official package.

The other "kmscon-git" PKGBUILD, https://aur.archlinux.org/packages/kmscon-git, still uses the abandoned repo by David Herrman.

Just install libtsm-patched-git, it provides libtsm-patched

Thanks for providing this PKGBUILD. And for the work on the kmscon fork. Thought the project was dead. But then heard about somewhere that "kmscon" is "still maintained and updated" and wanted to test this version. Nice to see the arm arch. Want to test it on the raspberry pi archlinux arm.

But I cant install this on x86_64. Maybe the PKGBUILD needs un update.

depends=(systemd libdrm mesa libgl pango libxkbcommon xkeyboard-config libtsm-patched)

$ yay -S kmscon-patched-git

 -> Could not find all required packages:
    libtsm-patched (Wanted by: kmscon-patched-git)

A little bit confused. There is in fact a https://aur.archlinux.org/packages/libtsm-patched-git, but not "libtsm-patched".

@latipun

Added aarch64.

Also check out a95c9ae74b893330367991c31f6884a1c0f8f394, should be buildable now. Note that this still requires the latest libtsm built from the source.

Failed to update because of this error:

Dependency libtsm found: NO found 4.0.2 but need: '>=4.1.0'

Also, please add aarch64, I can build it in previous version with Raspberry Pi 4B.

leak makedepends meson