kodi segmentation fault [solved]

seth · 2022-04-09 19:11:36

Yeah, BayTrail HD Graphics (and from the copredump on the crocus driver) => try mesa-amber.

davidx · 2022-04-09 21:26:36

for me it was python again. i downgraded and ignore listed python. I'll keep it like that until there is an new update, unfortunately.

alexela · 2022-04-10 10:38:34

So before having read your answer, i tried to both
$ export LD_PRELOAD=/usr/lib/dri/crocus_dri.so
and
# export LD_PRELOAD=/usr/lib/dri/crocus_dri.so
with wesa22 without any succes
(the /usr/local/lib… path doesnt exist on my machine)

Then i tried mesa-amber, but it still crashes with segfault. With mesa-amber, glxinfo complains :
«Error : couldn't find a RGB GLX visual or fbconfig»
And glxgears does not work anymore.

I tried to launch kodi from a wayland session (sway) instead of X11 (XFCE) and i just have a black screen, with no crash it seems (top shows a kodi.bin process is living and there is no crashlog reported…)

I can perform other tests or provide log file if needed, but i dont know what else could i do…

Thanks,
Alex

renegat · 2022-04-10 11:19:24

seth wrote:

1. it's not a "stub"

Well from what I've read about mesa, gallium, crocus, vulcan and all that very confusing stuff you must not expect very much development for this mesa-amber branch in future as it is only a collection of DRI drivers stripped of all vulcan and gallium drivers. If you consinder 'stub' is the wrong word - sorry.

seth wrote:

2. there's a guy w/ a gen7 chip right above you …

I'am totally confused about Intels 'gen'-numbering. Seems like CPUs, chipsets and graphic-cards have their own 'gens'.
According to mesa 22 release notes intel graphics since gen. 7 should be supported. Haswell graphics are Gen. 7 according to the wikipedia link you gave.

seth wrote:

3. "HD4000" is not an "Iris Graphics" chip, https://en.wikipedia.org/wiki/Intel_Gra … gy#Haswell

Totally right. It is an Intel HD-Card. Not UHD, not IRIS.
Nevertheless it should be supported by mesa 22 through crocus

seth wrote:

4. check your xorg log, there's likely an error about a missing /usr/lib/dri/i965_dri.so

Yes, the error is there. But only when using xf85-video-intel driver!

I see the following in xorg-log using intel driver on HD615 (Gen9):

(II) intel(0): [DRI2] Setup complete
(II) intel(0): [DRI2]   DRI driver: i965
(II) intel(0): [DRI2]   VDPAU driver: va_gl
...
(II) Initializing extension GLX
(EE) AIGLX error: dlopen of /usr/lib/dri/i965_dri.so failed (/usr/lib/dri/i965_dri.so: cannot open shared object file: No such file or directory)
(EE) AIGLX error: unable to load driver i965
(II) IGLX: Loaded and initialized swrast

modeset driver on HD615 (Gen9):

(II) modeset(0): [DRI2] Setup complete
[  4840.252] (II) modeset(0): [DRI2]   DRI driver: iris
[  4840.252] (II) modeset(0): [DRI2]   VDPAU driver: va_gl

with no AIGLX errors.

seth wrote:

=> you need mesa-amber and yeah, vainfo *likely* fails because of that as well. glxgears *probably* runs on the software rasterizer and glxinfo would reveal that.

No, you dont need mesa-amber. With your given

LD_PRELOAD=/usr/lib/dri/crocus_dri.so

everything works fine with Intel Gen.7 (Haswell) graphics cards : glxgears, glxinfo, vainfo and kodi!

There was a workaround for Intel graphics given in the wiki to set

MESA_LOADER_DRIVER_OVERRIDE=i965

environment variable in order to prevent weird display errors with Gen. 9 (KabyLake) graphics in qt-applications and esp. when using LXQt.
This is not needed anymore, the display-errors do not appear even without this setting.
You'll have to remove it to make Kodi work again.

Intel HD610/615:

  glxinfo -B
name of display: :0
display: :0  screen: 0
direct rendering: Yes
Extended renderer info (GLX_MESA_query_renderer):
    Vendor: Intel (0x8086)
    Device: Mesa Intel(R) HD Graphics 615 (KBL GT2) (0x591e)
    Version: 22.0.1
    Accelerated: yes
    Video memory: 3072MB
    Unified memory: yes
    Preferred profile: core (0x1)
    Max core profile version: 4.6
    Max compat profile version: 4.6
    Max GLES1 profile version: 1.1
    Max GLES[23] profile version: 3.2
OpenGL vendor string: Intel
OpenGL renderer string: Mesa Intel(R) HD Graphics 615 (KBL GT2)
OpenGL core profile version string: 4.6 (Core Profile) Mesa 22.0.1
OpenGL core profile shading language version string: 4.60
OpenGL core profile context flags: (none)
OpenGL core profile profile mask: core profile

OpenGL version string: 4.6 (Compatibility Profile) Mesa 22.0.1
OpenGL shading language version string: 4.60
OpenGL context flags: (none)
OpenGL profile mask: compatibility profile

OpenGL ES profile version string: OpenGL ES 3.2 Mesa 22.0.1
OpenGL ES profile shading language version string: OpenGL ES GLSL ES 3.20

For Haswell graphics Gen.7:

glxinfo -B
name of display: :0
display: :0  screen: 0
direct rendering: Yes
Extended renderer info (GLX_MESA_query_renderer):
    Vendor: Intel (0x8086)
    Device: Mesa Intel(R) HD Graphics 5000 (HSW GT3) (0xa26)
    Version: 22.0.1
    Accelerated: yes
    Video memory: 1536MB
    Unified memory: yes
    Preferred profile: core (0x1)
    Max core profile version: 4.6
    Max compat profile version: 3.1
    Max GLES1 profile version: 1.1
    Max GLES[23] profile version: 3.2
OpenGL vendor string: Intel
OpenGL renderer string: Mesa Intel(R) HD Graphics 5000 (HSW GT3)
OpenGL core profile version string: 4.6 (Core Profile) Mesa 22.0.1
OpenGL core profile shading language version string: 4.60
OpenGL core profile context flags: (none)
OpenGL core profile profile mask: core profile

OpenGL version string: 3.1 Mesa 22.0.1
OpenGL shading language version string: 1.40
OpenGL context flags: (none)

OpenGL ES profile version string: OpenGL ES 3.2 Mesa 22.0.1
OpenGL ES profile shading language version string: OpenGL ES GLSL ES 3.20

Thank you for your support!
Perhaps someone should update the wiki.
I cannot do this - everything I write there is revertet immediately on principle.

Last edited by renegat (2022-04-11 10:11:38)

alexela · 2022-04-10 11:53:44

So i rebooted with mesa-amber and
MESA_LOADER_DRIVER_OVERRIDE=i965
line in my /etc/environment.

Still those segfault errors (and no more glxgears…)

seth · 2022-04-10 11:56:10

alexela wrote:

Then i tried mesa-amber, but it still crashes with segfault. With mesa-amber, glxinfo complains :
«Error : couldn't find a RGB GLX visual or fbconfig»
And glxgears does not work anymore.

Please post the xorg log for that setup.

renegat wrote:

No, you dont need mesa-amber. With your given
LD_PRELOAD=/usr/lib/dri/crocus_dri.so
everything works fine with Intel Gen.7 (Haswell) graphics cards : glxgears, glxinfo, vainfo and kodi!

I found that workround later and previously just rolled w/ "crocus isn't reliable enough" - what still might be the case.

I cannot do this - everything I write there is revertet immediately on principle.

Not sure about the principle, esp. here, but have you tried opening a thread in the discussion of the page?

alexela · 2022-04-10 12:05:25

Here is the return of cat /var/log/Xorg.0.log : https://pastebin.com/6v66B20d

There is
[ 370.637] (II) modeset(0): [DRI2] DRI driver: i965
that would say it uses this old driver that used to work…

Last edited by alexela (2022-04-10 12:06:37)

seth · 2022-04-10 12:47:15

GLX and dri look fine, what's the actual, complete output of "glxinfo -B"?

alexela · 2022-04-10 13:47:15

Just these too lines :
name of display: :0.0
Error: couldn't find a RGB GLX visual or fbconfig

If i try to run glxgears, then i just have:
Error: couldn't get an RGB, Doubled-buffered visual

Thanks again for your help,
Alex

Last edited by alexela (2022-04-10 13:50:08)

seth · 2022-04-10 13:56:36

Let's see what libraries/objects are actually loaded.

strace -f -o /tmp/glxinfo.strace glxinfo

and post that file.

alexela · 2022-04-10 14:01:33

This command just outputs the same two lines as above

name of display: :0.0
Error: couldn't find a RGB GLX visual or fbconfig

seth · 2022-04-10 14:07:43

It'll also generate a file /tmp/glxinfo.strace - with the strace inside.

alexela · 2022-04-10 14:38:49

Sorry. Here it is : https://pastebin.com/mApR4cFw

seth · 2022-04-10 15:09:11

It want's to open libGLX_mesa.so.0 which isn't provided by mesa-amber

echo $__GLX_VENDOR_LIBRARY_NAME
__GLX_VENDOR_LIBRARY_NAME=amber glxinfo -B

alexela · 2022-04-10 15:33:02

And is it expected or mesa-amber should provide it ? if so, should i report this bug to the package maintainer of mesa-amber ? or to mesa-amber «main stream» (dont even know who is it) ?

If this is the default behavior of mesa-amber, cant i go back to some old mesa version maybe setting an environment variable to specify to use an old driver ?

Last edited by alexela (2022-04-10 15:34:59)

seth · 2022-04-10 15:50:44

libglvnd should™ not try to load libGLX_mesa or in very doubt resort to libGLX_amber - that's why I asked for the environment and whether you can control it to use the proper plugin.
But you're not the only one w/ this resolution issue, https://bugs.archlinux.org/task/74364

alexela · 2022-04-10 16:10:38

Ok, so it seems you understand the issue. If i do so, i guess i just have to wait for a new mesa-amber package with the good symlink names. Or it there an easy way for me to correct it or is it «hard coded» in the library and only a compilation can solve it ?

Thanks for the time you gave me,
Alex

seth · 2022-04-10 19:49:51

You can either symlink libGLX_mesa.so to resolve libGLX_amber.so but I'd suggest to just try and "export __GLX_VENDOR_LIBRARY_NAME=amber" instead (see #39)

alexela · 2022-04-10 20:04:16

So i created the following simlink in /usr/lib/ :
ln -s libGLX_amber.so.0 libGLX_mesa.so.0

and know the error has changed ! (but it still does not work, but at least, it does something)

Here is the new strace for glxinfo : https://pastebin.com/ggxna2jp where we can see the simlink seems to work : «openat(AT_FDCWD, "/usr/lib/libGLX_mesa.so.0", O_RDONLY|O_CLOEXEC) = 4» at line 200.

I now have just read your post about the environment variable __GLX_VENDOR_LIBRARY_NAME, but it does nothing different : glxinfo -B show the same error as with only the simlinks… («X Error or failed request : BadValue (integer parametre out of range…»)

Last edited by alexela (2022-04-10 20:24:07)

seth · 2022-04-10 20:39:28

You probably didn't restart X11 after the symlinking and you should (if we go w/ the symlink as workaround) symlink all libGLX_amber.so* libraries - not just the .0 version.

alexela · 2022-04-10 20:51:25

Ok, i created the 3 simlinks .so .so.0 and .so.0.0.0 and i restarted lightdm (i guess it restart X.org) and now glxgears works (one thing fixed, thanks) !

But kodi still has seg fault

here is the coredump info : https://pastebin.com/5QkUNu1c

Last edited by alexela (2022-04-10 20:53:49)

seth · 2022-04-10 21:26:10

Great. So we've completely swapped the GL implementation, but are left w/ the original error…
While this happens when initializing the KODI render system, the consntant trigger seems to be the logging… so https://bbs.archlinux.org/viewtopic.php … 3#p2030663 ?

alexela · 2022-04-10 21:52:37

I tried to downgrade both kodi and python-3 but it does not help. In the kodi_crashlog i can read this at the end :

2022-04-10 23:38:18.020 T:1421 INFO <general>: VDPAU::CreateContext - creating decoder context
2022-04-10 23:38:18.110 T:1421 INFO <general>: vdp_device = 0x000493e1 vdp_st = 0x0
2022-04-10 23:38:18.110 T:1421 INFO <general>: VDPAU Decoder capabilities:
2022-04-10 23:38:18.110 T:1421 INFO <general>: name level macbs width height
2022-04-10 23:38:18.110 T:1421 INFO <general>: ------------------------------------
2022-04-10 23:38:18.110 T:1421 INFO <general>: Skipped 1 duplicate messages..
2022-04-10 23:38:18.110 T:1421 INFO <general>: VDPAU::Close - closing decoder context
2022-04-10 23:38:18.125 T:1421 INFO <general>: CRenderSystemGL::InitRenderSystem - Version: (null), Major: 0, Minor: 0
2022-04-10 23:38:18.125 T:1421 INFO <general>: GL_VENDOR = NULL
2022-04-10 23:38:18.125 T:1421 INFO <general>: GL_RENDERER = NULL
2022-04-10 23:38:18.125 T:1421 INFO <general>: GL_VERSION = NULL
2022-04-10 23:38:18.125 T:1421 INFO <general>: GL_SHADING_LANGUAGE_VERSION = NULL

so it seem vdpau doesnot see all the effort we've made for having GL to work. I ask for vdpauinfo and here is the answer : https://pastebin.com/BmajBDGq
It looks like nothing is supported…

alexela · 2022-04-10 22:21:28

i found a workaround here : https://github.com/xbmc/xbmc/issues/18076

i just mv libvdpau_va_gl.so.1 to libvdpau_va_gl.so.1_backup and kodi is here again !!!

I'm gonna mark this thread as solved.

Thanks for your help,
Alex

alexela · 2022-04-10 22:29:22

… and i can update everything and it still works !

Arch Linux

#26 2022-04-09 19:11:36

Re: kodi segmentation fault [solved]

#27 2022-04-09 21:26:36

Re: kodi segmentation fault [solved]

#28 2022-04-10 10:38:34

Re: kodi segmentation fault [solved]

#29 2022-04-10 11:19:24

Re: kodi segmentation fault [solved]

#30 2022-04-10 11:53:44

Re: kodi segmentation fault [solved]

#31 2022-04-10 11:56:10

Re: kodi segmentation fault [solved]

#32 2022-04-10 12:05:25

Re: kodi segmentation fault [solved]

#33 2022-04-10 12:47:15

Re: kodi segmentation fault [solved]

#34 2022-04-10 13:47:15

Re: kodi segmentation fault [solved]

#35 2022-04-10 13:56:36

Re: kodi segmentation fault [solved]

#36 2022-04-10 14:01:33

Re: kodi segmentation fault [solved]

#37 2022-04-10 14:07:43

Re: kodi segmentation fault [solved]

#38 2022-04-10 14:38:49

Re: kodi segmentation fault [solved]

#39 2022-04-10 15:09:11

Re: kodi segmentation fault [solved]

#40 2022-04-10 15:33:02

Re: kodi segmentation fault [solved]

#41 2022-04-10 15:50:44

Re: kodi segmentation fault [solved]

#42 2022-04-10 16:10:38

Re: kodi segmentation fault [solved]

#43 2022-04-10 19:49:51

Re: kodi segmentation fault [solved]

#44 2022-04-10 20:04:16

Re: kodi segmentation fault [solved]

#45 2022-04-10 20:39:28

Re: kodi segmentation fault [solved]

#46 2022-04-10 20:51:25

Re: kodi segmentation fault [solved]

#47 2022-04-10 21:26:10

Re: kodi segmentation fault [solved]

#48 2022-04-10 21:52:37

Re: kodi segmentation fault [solved]

#49 2022-04-10 22:21:28

Re: kodi segmentation fault [solved]

#50 2022-04-10 22:29:22

Re: kodi segmentation fault [solved]

Board footer