[Unstable Update] 2021-01-17 - Kernels, Powerline-Go, Mutter, Python

Hi, sorry for the late reaction,
FYI,

  1. a few days ago, after doing a minor update (no kernel …), I rebooted to find: /bin /usr/bin empty …
    I had to reinstall the chrooted dead-man with pacman … seems all good .
  2. I have tried kernel 5.11, some games trigger a HardLockup with the Nvidia drivers during game launching.

FYI, this is an installation with the .21 rc iso, it worked perfectly for me.(and I noticed a big gain in speed installation)

I’ve got dosfstools: local (4.2-1) is newer than core (4.1-3.0) ?

1 Like

Mesa-git needs to be updated again, conflict with libelf package.

yes installing lib32-libelf (0.183-3) breaks dependency 'lib32-elfutils' required by lib32-mesa-git

the latest update with plasma 5.21 is a bit messed, but easy to solve. When I updated the package plasma-desktop-primex (5.20.5) prevented the installation of plasma-desktop 5.21 and there were various issues of course because of this.
So I just manually installed plasma-desktop 5.21 (it removes plasma-desktop-primex)
An alternative solution can be to just wait until the primex package reach the 5.21

normally that package should have not been installed the first place …

The manjaro installation is quite recent, it was either installed by default or it was suggested in the first update after installed the system because I didn’t install it manually.
Anyway I hope this info can be useful to someone

please rebuild mesa-git and lib32-mesa-git

I didn’t have this conflict. Since I don’t have fingerprint reader, I’m fine with plasma-desktop and this package was correctly updated.

Hi, about the lockup with kernel 511 and nvidia:

It is not a hard lockup, I made an ssh connection while playing,
As the screen hung, and the keyboard didnt answer anymore (even capslock dead)

I had a “dmesg -w” running: zero error messages,

tail -f of Xorg.log not abvious “crash error” but
the thign I see is: 20 thousands lines of:
[ 2048.203] (EE) modeset(0): present flip failed
[ 2048.204] (WW) modeset(0): flip queue failed: Invalid argument
[ 2048.204] (WW) modeset(0): Page flip failed: Invalid argument

The game’s log is fine too …

It is a laptop with prime-run … nvidia/intel,

rtx2060 and i7-9XXXXX with 16G of RAM , the rtx has 6G of VRAM.

This second test is with the latest 511 manjaro kernel : 5.11.0-1-MANJARO .

And with kernel 510 I have a very performant and stable system, even arma3 with proton gets more than a 100fps … that is twice as fast as under windobe.

I took a very “simple” openGL /SDL game for this test: xonotic :wink:

update:
I have deactivated openGL’s flipping and UBB : same problem and the same kind of logs.
The error message apears even before the game is launched.
I do have compositing disabled to make sure it is not an openGL provider confusing issue.

I have same error in my log with 5.11 but not with 5.10. I think is connected with assync page flyp enabled in 5.11

    [ 2048.203] (EE) modeset(0): present flip failed
    [ 2048.204] (WW) modeset(0): flip queue failed: Invalid argument
    [ 2048.204] (WW) modeset(0): Page flip failed: Invalid argument

Probably not, I see the news on phoronix, it is about intel GPU, but yet maybe the kernel has “to know about it” to trigger it in the video driver, and that extra code ; doesnt work well with nvidia … or hybrid laptop setups …

If you have any infos , please share: I am looking into the kernel docs to disable this by command line parameter and test, it would be quicked than to recompile the kernel with that mod. , but if I have to, I surely will …

I have second laptop with hybrid cards, but second card Is AMD, there problem is same and only on fullscreen, so I am pretty sure it is about iGPU

Edit: I looked trough patches and I can not see way to disable it with parameter which is plain stupid idea, anyway enablement is trough 3 rows in last patch

https://lists.freedesktop.org/archives/dri-devel/2020-July/273128.html

Nice try :stuck_out_tongue:

I have recompiled the 511 with no “flipping” , same error message && behaviour,

That is because: the message is not in dmesg/journal but in Xorg’s log !

Hence it is a Xorg driver problem, not a kernel module problem: nvidia nor intel.

I have jus done a grep on the kernel source and nvidia module, no trace what so ever about these text strings …

And the only place where this bug is spoken about, is in the “bugzilla” of freedektop …

I’ll keep you posted.

Thank you

Ok, the problem must be in xorg/intel:

By disabling glamor; by forcing accel method of modesetting to “SNA”

I have no lockup and no error message.

This could mean, there was a hidden bug in this, triggered by a new “thing” in 511.

:slight_smile:

Sorry:
While installing the moded i915, my source file got overwritten by the original one …

but that yet pushed me to dig enought to find what is it doing weird …

thus finding the glamor bug(?), which is the one trying to profit from this fonction,
because it is the one calling the function that throws the messages …
see files pageflip.c present.c in xorg-server]$ cd ./src/xserver/hw/xfree86/drivers/modesetting/

So in drivers/gpu/drm/i915/display/intel_display.c

mode_config->async_page_flip = false; replace this line with the one that says …true :slight_smile:

So the cause is glamor not working well with new async new code, glamor must be dependent on some sort of synchroneous traitement (like atomic fonctions …or “rt”)

basically: if you disable async it works with glamor .
if you disable glamor, it works without touching the kernel .

Thank you @phush0

1 Like

FYI
Same issue with 5.11.1-1 .

lol change log of 5.11.1 is 5 rows one bug with xen fixed :smiley:

The bug technically is in the Xorg module, not really the kernel.

I am filing right now a bug report at freedesktop.

We will see what they say .
:wink:

I have updated the ticket with a more sensible WorkAround:
setting PageFlip to 0 (zero) in the “Xorg.conf”

1 Like

FYI,

Yesterday I think, this package got installed automatically … :stuck_out_tongue:

I have no FP reader on this PC .