Bug 24070 - After a 507-package update on 25 December, Plasma desktop display is unusable.
Summary: After a 507-package update on 25 December, Plasma desktop display is unusable.
Status: RESOLVED FIXED
Alias: None
Product: Mageia
Classification: Unclassified
Component: RPM Packages (show other bugs)
Version: Cauldron
Hardware: All Linux
Priority: release_blocker critical
Target Milestone: ---
Assignee: KDE maintainers
QA Contact:
URL:
Whiteboard:
Keywords: 7beta1, 7beta2
Depends on:
Blocks:
 
Reported: 2018-12-25 22:45 CET by Thomas Andrews
Modified: 2019-02-16 22:51 CET (History)
10 users (show)

See Also:
Source RPM:
CVE:
Status comment:


Attachments
Journal of bad desktop (165.40 KB, text/plain)
2018-12-26 20:56 CET, Thomas Andrews
Details

Description Thomas Andrews 2018-12-25 22:45:10 CET
Description of problem:
After a massive 507-package update on 25 December 2018, my Plasma display has become unusable. 

It starts with the greeter, which pops up an Android-like virtual keyboard that can be used with the mouse for entering the password. One can still enter the password using the keyboard, but there is no indication that characters are being entered.

That's the minor part of the problem. Much more severe is the display that follows the login. It looks normal, at first, but if you go to actually do something the windows that pop up show only pieces of the Mageia background screen (the one that shows the cauldron filling with bubbles before the Plasma desktop shows). If, for example, you go to call up MCC using the panel icon, the password request window will pop up, but nothing will show but background. Type in the password and another window will pop up (for MCC) but it will be the same.

Same thing if I go to use the menu icon, or run something from a desktop icon. The only way to shut back down again is to push the pwer button to get the log out screen up, and go from there.

Affected hardware: HP Probook 6550b, first generation i3, 8GB RAM, integrated Intel graphics. Before this update, Mageia 7 was working very,very well indeed.

Personally, I suspect a conflict between SDDM and the Intel video driver, but I don't know enough to do any more than speculate.

This install has just Plasma installed, so I can't log in to something else to get logs and stuff. Attempting to use run level 3 (by simply adding "3" to the kernel options) fails with a blank screen.

There is a 32-bit M6 install on the hard drive, so I might be able to use that to gather information - assuming I have detailed instructions of how to do so.
Manuel Hiebel 2018-12-25 22:56:10 CET

Assignee: bugsquad => kde

Comment 1 Béat E 2018-12-26 12:14:56 CET
I can confirm this bug. I notice the same problem on my Asus Zenbook (64-bit). I think it is related to the update to KDE 18.12. I think that the package qtvirtualkeyboard5 was installed with this update that was not installed beforehand. 

Only KDE (and its login screen) is affected. I can use an Xfce session without problems, even use KDE applications in that environment.

I found a discussion about a similar problem (the virtual keyboard on the login screen):
https://forum.kde.org/viewtopic.php?f=252&t=143311

CC: (none) => ed1

Comment 2 Marja Van Waes 2018-12-26 12:18:28 CET
(In reply to Thomas Andrews from comment #0)

> 
> There is a 32-bit M6 install on the hard drive, so I might be able to use
> that to gather information - assuming I have detailed instructions of how to
> do so.

You might first want to try, in the bootloader screen, to add " 3" to the kernel parameters of the plasma-update install, to login in text mode.

I'd first try to run "urpmi --auto-update", in case the KDE upgrade wasn't complete. If, after reboot, the issue persists, then you reboot again with (" 3" added) to fetch the logs of the previous boot 
 As root: journalctl -b-1 > journalctl.txt

If the above doesn't work, then I think you need a Mageia 7beta1 Live iso, because Mga6 used different compression for the logs and can't (IINM) handle the Mga7 journal log files.

Please mount your installed Mga7 root partition.
There's a directory with a very long hexadecimal name in
 /that/partition's/var/log/journal/

please run, as root,

  journalctl -D /path/into/that/hexadecimal/directory/ > journal.txt

and then attach journal.txt to this bug report.
(Compress the journal with "xz journal.txt" if it's too large to attach.)

CC: (none) => marja11
Keywords: (none) => NEEDINFO

Comment 3 Ulrich Beckmann 2018-12-26 13:04:18 CET
Same here on a Asus K50J
00:02.0 VGA compatible controller [0300]: Intel Corporation Mobile 4 Series Chipset Integrated Graphics Controller [8086:2a42] (rev 09)
        Subsystem: ASUSTeK Computer Inc. Device [1043:1862]
        Kernel driver in use: i915
        Kernel modules: i915

The upgrade was complete, and the KDE painel is unsuable. Another machine with AMD/ATI graphics looks ok.

The problem with the virtual keyboard only occurs with the Mageia theme. Breeze works ok.

Ulrich

CC: (none) => bequimao.de

Comment 4 Thomas Andrews 2018-12-26 15:08:00 CET
It's become clear that we are dealing with two issues here, the virtual keyboard, and the display. I've heard from others that have only seen the keyboard problem, so that must be less hardware-specific, and probably should be split off into another bug report. 

The display problem, on the other hand, so far has only been seen on Intel graphics. On my system, the "videocard" is identified only as "Core Processor Integrated Graphics Controller," and it's using the Intel 810 or later" module. More specs on the processor: Intel Core i3 350M, 2.26GHz.

Marja, I tried adding "3" to the kernel parameters, and the boot appeared to fail with a blank screen. Perhaps I didn't wait long enough. I shall try again, this time also removing "splash quiet" to see what I come up with.

I have also been able to boot into "recovery mode," so perhaps I can get the log that way. Unfortunately, I have almost zero experience in working in "recovery mode."
Comment 5 Thomas Andrews 2018-12-26 17:59:51 CET
Replacing "splash quiet" with "3" results in a hung boot. After ""Starting Hold until boot process finishes up..." and "Starting Terminate Plymouth Boot Screen..." I see several messages concerning wlo1 and iwlwifi, ending with "IFWLOG: register target." There it sits, cursor blinking.

Another issue, perhaps?

Will come back to this later. For now, some family activities call for my attention.
Comment 6 Jüri Ivask 2018-12-26 19:33:01 CET
Duplicate of 24060 ?

CC: (none) => jyri2000

Comment 7 Thomas Andrews 2018-12-26 20:56:05 CET
Created attachment 10605 [details]
Journal of bad desktop

This was created by logging in using "recovery mode."
Comment 8 Béat E 2018-12-26 20:59:53 CET
In my case the flickering desktop is the same as in bug 24060 because the workaround mentioned there worked for me.
Ulrich Beckmann 2018-12-26 21:03:52 CET

See Also: (none) => https://bugs.mageia.org/show_bug.cgi?id=24060

Comment 9 Thomas Andrews 2018-12-26 21:28:56 CET
Using MCC in "recovery mode" to change the video driver to "modesetting" eliminated the display problem.

Using "urpmi --auto-update" from within recovery mode did not find any updates waiting, but I don't believe I had an Internet connection in that mode. Once I booted to a "normal" desktop using the modesetting driver, MCC reported several Plasma-related updates waiting, one of them being sddm. Getting them now...
Comment 10 Thomas Andrews 2018-12-26 21:49:58 CET
The updates made no difference in this issue. The "Intel 810 or later" driver still messes up, while the modesetting driver is OK.

Funky virtual keyboard is still there, too.
William Kenney 2018-12-27 00:42:06 CET

CC: (none) => wilcal.int

Comment 11 William Kenney 2018-12-27 00:50:29 CET
On real hardware, M7, Plasma, 64-bit

Package(s) under test:
Plasma

My platform:

Intel Core i5-4460 Haswell Quad-Core 3.2GHz LGA 115
Gigabyte GA-B85M-D3H LGA 1150 Intel B85 chipset
Integrated Graphics Processor - Intel HD Graphics support
Audito chipset - Realtek ALC892, 7.1 channels
Corsair Vengeance 8GB ( 2 x 4GB ) 240-pin DDR3 SDRAM 1600

lspci -k
00:02.0 VGA compatible controller: Intel Corporation Xeon E3-1200
v3/4th Gen Core Processor Integrated Graphics Controller (rev 06)
Subsystem: Gigabyte Technology Co., Ltd Device d000
Kernel driver in use: i915
Kernel modules: i915

A really spectacular crash. Screwed the whole system up into a
completely unusable format. It's been awhile since we've seen
one this bad. Cool. For some reason the login screen turned
into an on screen keyboard. I've never seen that before.

I'll wait for this one to get resolved then do a netinstall
to a blank drive to get going again.
Comment 12 William Kenney 2018-12-27 00:53:37 CET
The M7 development was going so smoothly. We needed one or
two of these to make it exciting again.
Comment 13 William Kenney 2018-12-27 04:09:04 CET
In a Vbox client, M7, Mate, 64-bit

Package(s) under test:
Mate

All seems to be fine with the desktop. I can update then auto reboot
back to a usable Mate desktop. But, if I log out to the log in screen,
regardless of its theme, the login screen is locked up.

No virtual keyboard.

lspci -k
--------
00:02.0 VGA compatible controller: Intel Corporation Xeon E3-1200
v3/4th Gen Core Processor Integrated Graphics Controller (rev 06)
Subsystem: Gigabyte Technology Co., Ltd Device d000
Kernel driver in use: i915
Kernel modules: i915
Comment 14 William Kenney 2018-12-27 19:18:21 CET
On real hardware, M7, Plasma, 64-bit

Works very nicely here after a netinstall to a blank drive:

Intel Core i7-2600K Sandy Bridge 3.4GHz overclocked to 3.8GHz LGA 1155 95W Quad-Core
GIGABYTE GA-Z68X-UD3-B3 LGA 1155 Intel Z68 SATA 6Gb/s USB 3.0 ATX Intel Motherboard
GIGABYTE GV-N440D3-1GI GeForce GT 440 (Fermi) 1GB 128-bit DDR3 PCI Express 2.0 x16
RTL8111/8168B PCI Express 1Gbit Ethernet
CORSAIR Vengeance 16GB (4 x 4GB) 240-Pin DDR3 SDRAM DDR3 1600 (PC3 12800)
OCZ Vertex 4 VTX4-25SAT3-128G 2.5" 128GB SATA III

01:00.0 VGA compatible controller: NVIDIA Corporation GF108 [GeForce GT 440] (rev a1)
Subsystem: Gigabyte Technology Co., Ltd Device 3518
Kernel driver in use: nvidia
Kernel modules: nvidiafb, nouveau, nvidia_drm, nvidia390

Updates, goes in and out of the desktop to the login screen,
reboots cleanly back to a working disktop.
Comment 15 William Kenney 2018-12-27 20:21:43 CET
On real hardware, M7, Plasma, 64-bit

netinstall to my Intel:

Intel Core i5-4460 Haswell Quad-Core 3.2GHz LGA 115
Gigabyte GA-B85M-D3H LGA 1150 Intel B85 chipset
Integrated Graphics Processor - Intel HD Graphics support
Audito chipset - Realtek ALC892, 7.1 channels
Corsair Vengeance 8GB ( 2 x 4GB ) 240-pin DDR3 SDRAM 1600

lspci -k
00:02.0 VGA compatible controller: Intel Corporation Xeon E3-1200
v3/4th Gen Core Processor Integrated Graphics Controller (rev 06)
Subsystem: Gigabyte Technology Co., Ltd Device d000
Kernel driver in use: i915
Kernel modules: i915

platform. Installs cleanly.
Boots to a completely unusable desktop.
Comment 16 Maurice Batey 2018-12-27 20:42:30 CET
Well, here on my HP 450g2 Core i5 Probook with a UEFI-booted 64-bit Plasma Mageia-7 on real h/w, today's 590-package update had no detrimental effect, though neither of the Kmail & Kompozer problems has gone away.

$ lspci -k
00:00.0 Host bridge: Intel Corporation Broadwell-U Host Bridge -OPI (rev 09)
        Subsystem: Hewlett-Packard Company Device 2248
        Kernel driver in use: bdw_uncore
00:02.0 VGA compatible controller: Intel Corporation HD Graphics 5500 (rev 09)
        Subsystem: Hewlett-Packard Company Device 2248
        Kernel driver in use: i915
        Kernel modules: i915

/\/\aurice

CC: (none) => maurice

Comment 17 Antony Baker 2018-12-27 20:51:44 CET
I have the on-screen keyboard, but have not noticed anything else out of the ordinary.

Real hardware, Cauldron, Plasma, 64-bit

System: Host: localhost Kernel: 4.19.12-desktop-2.mga7 x86_64 bits: 64 compiler: gcc v: 8.2.1 Desktop: KDE Plasma 5.14.4 
tk: Qt 5.12.0 wm: kwin_x11 dm: SDDM Distro: Mageia 7 mga7


Graphics: Device-1: NVIDIA GP107 [GeForce GTX 1050 Ti] driver: nvidia v: 410.78 bus ID: 01:00.0 chip ID: 10de:1c82 
Display: x11 server: Mageia X.org 1.20.3 driver: nvidia,v4l compositor: kwin_x11 resolution: 1920x1080~60Hz 
OpenGL: renderer: GeForce GTX 1050 Ti/PCIe/SSE2 v: 4.6.0 NVIDIA 410.78 direct render: Yes

CC: (none) => jasperodus

Comment 18 Nicolas Lécureuil 2018-12-27 21:40:41 CET
can you kill plasmashell and start it under konsole to see debugs ?

CC: (none) => mageia

Comment 19 Béat E 2018-12-28 11:16:01 CET
I uninstalled lib64qt5virtualkeyboard5 (and the dependent  
packages lib64qt5hunspellinputmethod5-5.12.0-1.mga7.x86_64 
and qtvirtualkeyboard5-5.12.0-1.mga7.x86_64). Now the virtual keyboard on the login screen is gone.
Comment 20 Thomas Andrews 2018-12-28 17:24:21 CET
(In reply to Béat E from comment #19)
> I uninstalled lib64qt5virtualkeyboard5 (and the dependent  
> packages lib64qt5hunspellinputmethod5-5.12.0-1.mga7.x86_64 
> and qtvirtualkeyboard5-5.12.0-1.mga7.x86_64). Now the virtual keyboard on
> the login screen is gone.

The virtual keyboard problem was taken care of by a change in /etc/sddm.conf with the last update of sddm. Unfortunately, the user has to authorize the change before it could be applied. That was the "inspection" notice you should have seen after receiving that sddm update. If you did nothing, the new sddm.conf was saved as /etc/sddm.conf.rpmnew and your old sddm.conf is still being used. To use the new conf file, assuming you do have a /etc/sddm.conf.rpmnew, as root rename /etc/sddm.conf to something like sddm.conf.old, and then rename sddm.conf.rpmnew to sddm.conf. The virtual keyboard should not show after that.

Those who install from newer isos won't see the problem at all.
Comment 21 Thomas Andrews 2018-12-28 17:40:59 CET
Unfortunately, the latest round of updates doesn't do anything to fix the Intel video problem on the Probook 6550b.

Moreover, I decided to try updating a Core 2 Duo machine that is also using Intel graphics, but that hasn't been updated for a couple of weeks or so. Not only does that machine now exhibit the problem, it seems that while it looks like I can get into recovery mode as root, I'm not authorized to run MCC while there. 

So, I have no idea of how to change the video driver to modesetting so I can use the system again. Looks like a new install is in order, with a different DE, until we have new isos.
Comment 22 Martin Whitaker 2018-12-28 20:39:58 CET
(In reply to Thomas Andrews from comment #21)
> So, I have no idea of how to change the video driver to modesetting so I can
> use the system again. Looks like a new install is in order, with a different
> DE, until we have new isos.

If you can edit the file /etc/X11/xorg.conf, in the "Device" section change the line

  Driver "intel"

to

  Driver "modesetting"

Alternatively, you should have IceWM installed as well as Plasma - try using that to make the changes (it should appear in the drop-down list in the SDDM login screen).

You could also try the workaround described in bug 24060 comment 3.

CC: (none) => mageia

Comment 23 Thomas Andrews 2018-12-28 21:20:55 CET
(In reply to Martin Whitaker from comment #22)

> 
> If you can edit the file /etc/X11/xorg.conf, in the "Device" section change
> the line
> 
>   Driver "intel"
> 
> to
> 
>   Driver "modesetting"
> 
Hmmm. That reminds me of something. A while back - I think with Mageia 4 - there was an update (a kernel? video driver? I dunno.) that "broke" KDE on some Intel display hardware. The fix/workaround was adding a line to xorg.conf. Something about the acceleration method as I recall, but I don't remember details. I do remember that the next release didn't need it any more.

I wonder if something like that would help here?
Comment 24 Martin Whitaker 2018-12-28 21:37:54 CET
(In reply to Nicolas Lécureuil from comment #18)
> can you kill plasmashell and start it under konsole to see debugs ?

Messages are identical when running with/without the workaround. No obvious errors reported in the system and Xorg logs. Using kernel-linus makes no difference.

Keywords: NEEDINFO => (none)

Comment 25 Thomas Andrews 2018-12-28 22:24:11 CET
(In reply to Martin Whitaker from comment #22)

> 
> If you can edit the file /etc/X11/xorg.conf, in the "Device" section change
> the line
> 
>   Driver "intel"
> 
> to
> 
>   Driver "modesetting"
> 
> Alternatively, you should have IceWM installed as well as Plasma - try using
> that to make the changes (it should appear in the drop-down list in the SDDM
> login screen).
> 
Merely changing xorg.conf from within a Mageia 6 install didn't do the trick - no idea why.

IceWM seems to have problems too - several things not working - but I was able to click up a terminal and run MCC. Changed the video driver that way, and the problem went away. 

I will hold off on starting a bug report on the IceWM problems for now, probably waiting until this one is solved and new isos come out to see if they are still there.
JanKusanagi 2018-12-29 01:20:08 CET

CC: (none) => jan-bugs

Comment 26 Thomas Andrews 2018-12-29 21:52:48 CET
(In reply to Martin Whitaker from comment #22)
> 
> You could also try the workaround described in bug 24060 comment 3.

That also eliminated the problem on both of my affected systems.
Comment 27 William Kenney 2018-12-30 18:36:19 CET
A very big update to the M7 Repo today so I gave a Vbox client netinstall a go.
Installs cleanly and boots to a working desktop. Lots of things work correctly,
a big improvement from just a few days ago.

For some reason the keyboard, not the mouse, becomes inoperative. And that for
both a wireless ( Logitech ) keyboard and mouse, and a wired ( USB ) pair.
Note. The mouse continues to work, the keyboard not. A reboot of the Client
may, or may not, get the keyboard working again. Only to stop working.
Comment 28 Thomas Andrews 2018-12-30 20:32:06 CET
(In reply to Thomas Andrews from comment #26)
> (In reply to Martin Whitaker from comment #22)
> > 
> > You could also try the workaround described in bug 24060 comment 3.
> 
> That also eliminated the problem on both of my affected systems.

An interesting tidbit of information:

I was installing Plasma on one of my nvidia340 (Geforce 9800 GT) systems from the Live iso to check something else out, and I happened to miss booting with the non-free driver, so I did the install with nouveau. During the first boot, when the install finished up, everything was OK. But when I booted into the system after getting over 1000 updates the display was showing symptoms very like those seen on Intel systems in this bug.

On a hunch, I applied the workaround, and rebooted to a working desktop showing no symptoms whatsoever!

Now I'm wondering what other older systems could be "fixed" with this workaround...
Comment 29 Thomas Andrews 2018-12-31 16:37:56 CET
Updates are still coming along, as is the nature of Cauldron. Over 100 this morning.

If some of those updates might help with this bug, please let us know so that those of us who have applied the workaround can test by removing it again. 

Otherwise, I, for one, will leave the workaround in place, so that my systems remain usable.
Comment 30 Ulrich Beckmann 2019-01-01 13:50:11 CET
(In reply to Nicolas Lécureuil from comment #18)
> can you kill plasmashell and start it under konsole to see debugs ?

I tried this and got the following output

$ plasmashell

qt.qpa.xcb: could not connect to display
qt.qpa.plugin: Could not load the Qt platform plugin "xcb" in "" even though it was found.
This application failed to start because no Qt platform plugin could be initialized. Reinstalling the application may fix this problem.

Available platform plugins are: wayland-org.kde.kwin.qpa, eglfs, linuxfb, minimal, minimalegl, offscreen, vnc, xcb.
Comment 31 Thomas Andrews 2019-01-04 22:31:07 CET
As expected, still valid after an install from the Round 1 beta 2 isos. (We need a 7beta2 keyword)

Using a blank area of a hard drive on one of my affected machines, I did a "standard" Plasma install from the 32-bit Beta 2 CI. After the reboot and login, Plasma was unusable. 

This time, instead of using the above workaround, I logged into IceWM, ran systemsettings5 from a terminal, and told it to change the compositor to XRender. That did the trick. Plasma was fine after that.

The Compositor page of Mageia 6's System Settings GUI notes that OpenGL has crashed KWin in the past, and suggests the change to XRender as one alternative if it does. 

They say the cause is most likely a driver bug. But without knowing anything about such things, it seems to me strange that the same bug would suddenly appear in both the Intel and nouveau drivers, at the same time.
William Kenney 2019-01-06 15:24:02 CET

Priority: Normal => release_blocker

William Kenney 2019-01-06 15:25:00 CET

Keywords: (none) => 7beta1

Thomas Andrews 2019-01-06 17:09:53 CET

Keywords: (none) => 7beta2

Comment 32 William Kenney 2019-01-07 00:12:35 CET
On real hardware, M7, Plasma, 64-bit

Intel Core i5-4460 Haswell Quad-Core 3.2GHz LGA 115
Gigabyte GA-B85M-D3H LGA 1150 Intel B85 chipset
Integrated Graphics Processor - Intel HD Graphics support
Audito chipset - Realtek ALC892, 7.1 channels
Corsair Vengeance 8GB ( 2 x 4GB ) 240-pin DDR3 SDRAM 1600

[wilcal@localhost ~]$ lspci -k
00:02.0 VGA compatible controller: Intel Corporation Xeon E3-1200 v3/4th Gen Core Processor Integrated Graphics Controller (rev 06)
        Subsystem: Gigabyte Technology Co., Ltd Device d000
        Kernel driver in use: i915
        Kernel modules: i915

Install using:

Mageia-7-beta2-Live-Plasma-x86_64.iso  1/4/19
eccafd237f551eeaf4829763142f38ae

Boots too an unstable Plasma desktop
fall back to login page
login with Icewm which is much more stable
open ~/.config/kwinrc  with kwrite
Change
[Compositing]
OpenGLIsUnsafe=false
to
[Compositing]
OpenGLIsUnsafe=true

Log out of Icewm
Login to Plasma

Plasma desktop is now stable and usable.
Complete install and update and successfully boot back to a
stable and working desktop
Comment 33 Thomas Andrews 2019-01-24 17:14:56 CET
Good news! 

It appears that a Plasma update currently being tested fixes this bug.
Comment 34 Thomas Andrews 2019-02-16 22:51:18 CET
As of Round 3 of the 7beta2 test isos, this bug has been resolved.

Status: NEW => RESOLVED
Resolution: (none) => FIXED


Note You need to log in before you can comment on or make changes to this bug.