Bug 943297 – media-libs/mesa-24.2.x - Segmentation fault at loading drm at starting X on BigEndian machines

• Product: Gentoo Linux
• Status: UNCONFIRMED ---

[Description ernsteiswuerfel 2024-11-11 21:55:16 UTC]

Created attachment 908548 [details]
Xorg.0.log (ppc64, w. mesa 24.2.6-r2)

mesa-24.1.x is fine on my ppc/ppc64 machines but 24.2.x segfaults at starting X:

[...]
(II) RADEON(0): [DRI2] Setup complete
(II) RADEON(0): [DRI2]   DRI driver: r600
(II) RADEON(0): [DRI2]   VDPAU driver: r600
(II) RADEON(0): Front buffer size: 0K
(II) RADEON(0): VRAM usage limit set to 225568K
(EE) 
(EE) Backtrace:
(EE) 0: /usr/bin/X (xorg_backtrace+0xffe98c88) [0x10806c390]
(EE) 1: /usr/bin/X (0x107ed0000+0x1a1cac) [0x108071cac]
(EE) 2: linux-vdso64.so.1 (__kernel_sigtramp_rt64+0x0) [0x3fffbcc22424]
(EE) 3: /usr/lib64/xorg/modules/drivers/radeon_drv.so (0x3fffbae90000+0x5d83c) [0x3fffbaeed83c]
(EE) 4: /usr/bin/X (0x107ed0000+0x1c894c) [0x10809894c]
(EE) 5: /usr/bin/X (AddScreen+0xffd8df10) [0x107f518c8]
(EE) 6: /usr/bin/X (InitOutput+0xffec43ac) [0x10809ab2c]
(EE) 7: /usr/bin/X (0x107ed0000+0x86d3c) [0x107f56d3c]
(EE) 8: /usr/bin/X (0x107ed0000+0x39c20) [0x107f09c20]
(EE) 9: /usr/lib64/libc.so.6 (0x3fffbbeb0000+0x26b84) [0x3fffbbed6b84]
(EE) 10: /usr/lib64/libc.so.6 (__libc_start_main+0xffe0d4f8) [0x3fffbbed6de8]
(EE) 
(EE) Segmentation fault at address 0xc
(EE) 
error:
(EE) Caught signal 11 (Segmentation fault). Server aborting


The issue has already been reported upstream: https://gitlab.freedesktop.org/mesa/mesa/-/issues/12048

It has also been bisected to the following commit:
 # git bisect bad
20b34007014953f5bce7c0073879320c706273a7 is the first bad commit
commit 20b34007014953f5bce7c0073879320c706273a7 (HEAD)
Author: Mike Blumenkrantz <[e-mail]>
Date:   Mon Jul 22 19:18:00 2024 -0400

    dril: rework config creation
    
    the original implementation of config selection had a number of flaws:
    * using eglChooseConfigs with lots of loops, which was okay for filtering but
      also added considerable complexity and made it difficult to correctly
      get all the configs
    * not adding enough configs; there were a lot more color and zs formats
      which weren't in the base config list
    * double buffer configs were never created
    * srgb configs were also never created
    
    there will now be fewer configs than there were pre-DRIL, but this is only
    because accum buffers are now gone and not because anything of value is
    missing
    
    Fixes: 3de62b2f9a6 ("gallium/dril: Compatibility stub for the legacy DRI loader interface")
    
    Acked-by: Daniel Stone <[e-mail]>
    Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30311>
    (cherry picked from commit ec7afd2c24c6afa6a5c80da4f80e0b799580e12a)

 .pick_status.json                      |   2 +-
 src/gallium/targets/dril/dril_target.c | 199 +++++++++++++++++++++++++++++++++---------------
 2 files changed, 140 insertions(+), 61 deletions(-)


The 1st upstream report was on a ppc32 system with a Nvidia GeForce FX Go5200. I encountered the issue on a ppc64 system with a Radeon HD4550 where I did the bisect (https://gitlab.freedesktop.org/mesa/mesa/-/issues/11933#note_2626416).

Likely all BigEndian arches are affected.

[Comment 1 ernsteiswuerfel 2024-11-11 21:58:32 UTC]

Created attachment 908549 [details]
emerge --info

[Comment 2 yesi 2024-11-13 11:19:09 UTC]

I met the same problem: https://forums.gentoo.org/viewtopic-p-8846020.html#8846020
I solved masking this version =media-libs/mesa-24.2.6-r2.

[Comment 3 yesi 2024-11-13 13:30:21 UTC]

I have a amd64 arch.

[Comment 4 ernsteiswuerfel 2024-11-13 15:01:15 UTC]

(In reply to yesi from comment #3)
> I have a amd64 arch.
If it's the same issue remains to be seen, even if the segfault looks similar. ;)

The segfault states mesa-amber is used on your system? If so, remove mesa-amber and re-emerge mesa-24.1.7-r1. This should work (perhaps you already did that). Then try mesa-24.2.6-r2 again. If you run into the same issue doing a git-bisect would be really helpful!

If it's really the same issue, traceable back to the same commit (as I did the mesa upstream bug), it would be really interesting what's causing it on an amd64 system with Intel iGPUI!

[Comment 5 yesi 2024-11-14 15:50:07 UTC]

You are right.
mesa-amber is the culprit.
I remove it and un-mask the last version of mesa.
Thanks.

[Comment 6 Sam James 2024-11-30 05:15:34 UTC]

(In reply to yesi from comment #2)
> I met the same problem:
> https://forums.gentoo.org/viewtopic-p-8846020.html#8846020
> I solved masking this version =media-libs/mesa-24.2.6-r2.

This issue is bug 945373 now.