A Bug In The Intel Linux Graphics Driver Is Causing KDE's Plasma 5 To Crash

[byECHO]

On Fedora 22 KDE spin I had repeated crashes with Nouveau and nVidia with KDE 5. I've switched to GNOME, and so far the stability is flawless.

So how much of these bugs are "KDE exposes graphics driver flaws" and how much are "KDE makes incorrect graphics calls"? I have no idea.

It is actually caused by QML. KDE 5 mostly uses QML for rendering so It is not KDE's fault.

[Michael_S]

On Fedora 22 KDE spin I had repeated crashes with Nouveau and nVidia with KDE 5. I've switched to GNOME, and so far the stability is flawless.

So how much of these bugs are "KDE exposes graphics driver flaws" and how much are "KDE makes incorrect graphics calls"? I have no idea.

[byECHO]

I actually found a better workaround than using UXA. I set xorg to use DRI version 1 instead of 2 which it used by default and reverted to the not git releases of xf86-video-intel and mesa. So far I have had 0 crashes with DRI1 and SNA, and much better performance than using UXA (I can't notice any difference in performance compared with the default DRI2).
I put this in /etc/X11/xorg.conf.d/20-intel.conf

Section "Device"
Identifier "Intel Graphics"
Driver "intel"
Option "AccelMethod" "sna"
Option "DRI" "1"
EndSection

I have a Haswell gpu. My system became unusable after I set dri to 1. Also, in my case, I can't notice any performance difference between UXA and SNA. I'm using Arch Linux with linux-ck kernel. But I can confirm that changing Accel Method to UXA fixes the issue.

[sarfarazahmad]

I actually found a better workaround than using UXA. I set xorg to use DRI version 1 instead of 2 which it used by default and reverted to the not git releases of xf86-video-intel and mesa. So far I have had 0 crashes with DRI1 and SNA, and much better performance than using UXA (I can't notice any difference in performance compared with the default DRI2).
I put this in /etc/X11/xorg.conf.d/20-intel.conf

Section "Device"
Identifier "Intel Graphics"
Driver "intel"
Option "AccelMethod" "sna"
Option "DRI" "1"
EndSection

Thanks been seeing plasma crash quite often on manjaro 0.8.13 myself. guess its just not that often as reported by one arch user. Will head home and try this out and report back !

[persept]

I actually found a better workaround than using UXA. I set xorg to use DRI version 1 instead of 2 which it used by default and reverted to the not git releases of xf86-video-intel and mesa. So far I have had 0 crashes with DRI1 and SNA, and much better performance than using UXA (I can't notice any difference in performance compared with the default DRI2).
I put this in /etc/X11/xorg.conf.d/20-intel.conf

Section "Device"
   Identifier  "Intel Graphics"
   Driver      "intel"
   Option      "AccelMethod"  "sna"
   Option "DRI" "1"
EndSection

[e8hffff]

26
Is that why KDE5 was so crash and burn on my Nuc5i5

[Kano]

It seems that ickle does not use KDE that often. In mesa 9 there was a change that made in patched kwin from KDE 4.4 instable as well. At least compile problems with Intel ddx get fixed very fast in case it happens.

[BlueJayofEvil]

There has been a pervasive issue for months with plasma 5 locking up and requiring "killall plasmashell ; plasmashell" and it's certainly not limited to Intel cards... I have an ATI Radeon - same issue. This has been reported and documented ad nauseam and the answer has been "oh you have an intel card"... so yeah, there might indeed be an issue with an Intel driver, but plasma 5 has bigger issues than this... I'm getting the feeling the plasma folks are thinking "oh, it's just intel" - hope that isn't the case, cause it certainly isn't.

I'm having the same issue on my laptop using radeonsi (AMD APU system running openSUSE Tumbleweed x86-64 with Pontostroy's repos and latest kernel snapshot). The issue manifests most often when moving/manipulating widgets on the desktop, some seem to be more "volatile" than others.

[bug77]

There has been a pervasive issue for months with plasma 5 locking up and requiring "killall plasmashell ; plasmashell" and it's certainly not limited to Intel cards... I have an ATI Radeon - same issue. This has been reported and documented ad nauseam and the answer has been "oh you have an intel card"... so yeah, there might indeed be an issue with an Intel driver, but plasma 5 has bigger issues than this... I'm getting the feeling the plasma folks are thinking "oh, it's just intel" - hope that isn't the case, cause it certainly isn't.

If intel driver causes a large chunk of the crashes, I can see how developers might be tempted to lump other unrelated crashed into the same category.

[yoshi314]

Intel has awful track record wrt stability. I remember reporting a BSOD on kms init bug that dragged since 3.0 kernel will it was fixed in 3.5. Then it came back and went. Disabling kms forced me to use vesa driver, as intel was already 'kms or nothing' at that point.

One update hard locked the pc during xorg startup, until it was hotfixed.

there are numerous bugs where switching between sna or other acceleration mitigates the issue. I mean, what the hell? Is the hardware that whimsical?