Bug 2261845

Summary: mesa-24.0.0~rc2-5.fc40 breaks many things
Product: [Fedora] Fedora Reporter: Kevin Fenzi <kevin>
Component: mesaAssignee: José Expósito <jexposit>
Status: RELEASE_PENDING --- QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: medium Docs Contact:
Priority: unspecified    
Version: rawhideCC: ajax, bruno, bskeggs, igor.raits, jexposit, j, lyude, mdaenzer, megger, rhughes, rstrode, sevmek, tstellar, walter.pete
Target Milestone: ---   
Target Release: ---   
Hardware: Unspecified   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
journalctl -b none

Description Kevin Fenzi 2024-01-30 01:46:36 UTC
Upgrading to mesa-24.0.0~rc2-5.fc40 seems to break: gnome-shell, sddm_greeter, nheko and even Xorg using Xfce. ;(

Downgrading to -4 and things start working again fine. 

I can get backtraces if you want... 

A partial gnome-shell one: 

                Stack trace of thread 3835:
                #0  0x00007f34c147f64c _ZNK4llvm14ConstantFolder15FoldNoWrapBinOpENS_11Instru>
                #1  0x00007f34c13f3d8a LLVMBuildAdd (libLLVM-17.so + 0x9f3d8a)
                #2  0x00007f34de241386 si_llvm_build_vs_prolog (radeonsi_dri.so + 0x841386)
                #3  0x00007f34de246640 si_llvm_build_shader_part (radeonsi_dri.so + 0x846640)
                #4  0x00007f34de1ac4eb si_get_shader_part (radeonsi_dri.so + 0x7ac4eb)
                #5  0x00007f34de1ac832 si_get_vs_prolog (radeonsi_dri.so + 0x7ac832)
                #6  0x00007f34de1ce54c si_create_shader_variant (radeonsi_dri.so + 0x7ce54c)
                #7  0x00007f34de1eef2b _ZL23si_build_shader_variantP9si_shaderib (radeonsi_dr>
                #8  0x00007f34de1fd427 si_shader_select (radeonsi_dri.so + 0x7fd427)
                #9  0x00007f34de7755db _Z17si_update_shadersIL13amd_gfx_level14EL11si_has_tes>
                #10 0x00007f34de776873 _Z11si_draw_vboIL13amd_gfx_level14EL11si_has_tess0EL9s>
                #11 0x00007f34de60f6c5 _ZL17si_draw_rectangleP15blitter_contextPvPFS1_S0_Eiii>
                #12 0x00007f34deff3ed0 util_blitter_clear_custom.constprop.0 (radeonsi_dri.so>
                #13 0x00007f34de26992e si_clear.lto_priv.0 (radeonsi_dri.so + 0x86992e)
                #14 0x00007f34ddf9be53 tc_call_clear (radeonsi_dri.so + 0x59be53)
                #15 0x00007f34ddf9feb6 tc_batch_execute.lto_priv.0 (radeonsi_dri.so + 0x59feb>
                #16 0x00007f34df19b548 _tc_sync.isra.0 (radeonsi_dri.so + 0x179b548)
                #17 0x00007f34ddfb70cb tc_flush (radeonsi_dri.so + 0x5b70cb)
                #18 0x00007f34ddb20147 st_context_flush (radeonsi_dri.so + 0x120147)
                #19 0x00007f34dda46e69 dri_flush (radeonsi_dri.so + 0x46e69)
                #20 0x00007f34e420fd91 loader_dri3_swap_buffers_msc.constprop.0 (libGLX_mesa.>
                #21 0x00007f34e41f3a6b glXSwapBuffers (libGLX_mesa.so.0 + 0x38a6b)
                #22 0x00007f34feaa0c41 cogl_onscreen_glx_swap_buffers_with_damage.lto_priv.0 >
                #23 0x00007f34fea85b0e cogl_onscreen_swap_buffers_with_damage (libmutter-cogl>
                #24 0x00007f34feea008b swap_framebuffer (libmutter-14.so.0 + 0xa008b)
                #25 0x00007f34feea42c1 meta_stage_impl_redraw_view (libmutter-14.so.0 + 0xa42>
                #26 0x00007f34feefdd1a meta_stage_x11_redraw_view (libmutter-14.so.0 + 0xfdd1>
                #27 0x00007f34ff27f18c handle_frame_clock_frame (libmutter-clutter-14.so.0 + >
                #28 0x00007f34ff251edb frame_clock_source_dispatch (libmutter-clutter-14.so.0>
                #29 0x00007f34ff54426c g_main_context_dispatch_unlocked.lto_priv.0 (libglib-2>
                #30 0x00007f34ff5a52a8 g_main_context_iterate_unlocked.isra.0 (libglib-2.0.so>
                #31 0x00007f34ff54a227 g_main_loop_run (libglib-2.0.so.0 + 0x62227)
                #32 0x00007f34feed2f9a meta_context_run_main_loop (libmutter-14.so.0 + 0xd2f9>
                #33 0x00007f34fe3c0056 ffi_call_unix64 (libffi.so.8 + 0x9056)
                #34 0x00007f34fe3bc6a0 ffi_call_int.lto_priv.0 (libffi.so.8 + 0x56a0)
                #35 0x00007f34fe3bf4ee ffi_call (libffi.so.8 + 0x84ee)
                #36 0x00007f34ff33ed9e _ZN3Gjs8Function6invokeEP9JSContextRKN2JS8CallArgsENS3>
                #37 0x00007f34ff33f673 _ZN3Gjs8Function4callEP9JSContextjPN2JS5ValueE (libgjs>
                #38 0x00007f34fd52c6b8 _ZN2js23InternalCallOrConstructEP9JSContextRKN2JS8Call>
                #39 0x00007f34fd520593 _ZN2js9InterpretEP9JSContextRNS_8RunStateE (libmozjs-1>
                #40 0x00007f34fd52c1fb _ZN2js9RunScriptEP9JSContextRNS_8RunStateE (libmozjs-1>
                #41 0x00007f34fd52c58b _ZN2js23InternalCallOrConstructEP9JSContextRKN2JS8Call>
                #42 0x00007f34fd52cafd _ZN2js4CallEP9JSContextN2JS6HandleINS2_5ValueEEES5_RKN>
                #43 0x00007f34fd5a0a18 _ZN2JS4CallEP9JSContextNS_6HandleINS_5ValueEEES4_RKNS_>
                #44 0x00007f34ff372aa2 _ZN17GjsContextPrivate18run_main_loop_hookEv (libgjs.s>
                #45 0x00007f34ff37b08f gjs_context_eval_module (libgjs.so.0 + 0x9208f)
                #46 0x00007f34ff37b2b9 gjs_context_eval_module_file (libgjs.so.0 + 0x922b9)
                #47 0x00005581e3db8036 main (gnome-shell + 0x3036)
                #48 0x00007f34fec3d088 __libc_start_call_main (libc.so.6 + 0x2a088)
                #49 0x00007f34fec3d14b __libc_start_main@@GLIBC_2.34 (libc.so.6 + 0x2a14b)
                #50 0x00005581e3db8485 _start (gnome-shell + 0x3485)

Reproducible: Always

Comment 1 José Expósito 2024-01-30 09:15:50 UTC
It looks like it is related to LTO:
https://src.fedoraproject.org/rpms/mesa/pull-request/38

@mdaenzer I'm afraid we might have to disable it again :(

Comment 2 Michel Dänzer 2024-01-30 09:27:08 UTC
(In reply to José Expósito from comment #1)
> It looks like it is related to LTO:

So enabling LTO was the only change in -5? E.g. -4 was already built against the same LLVM version?

Comment 3 Michel Dänzer 2024-01-30 09:39:35 UTC
(In reply to Michel Dänzer from comment #2)
> So enabling LTO was the only change in -5? E.g. -4 was already built against
> the same LLVM version?

Looks like that's the case, so let's revert indeed.

Quite surprising, haven't hit this in my testing or seen anything like it reported upstream.

Comment 4 Kevin Fenzi 2024-01-30 17:20:23 UTC
My hardware is a frame.work amd laptop/radeon gui... happy to provide mode details.

I can't think of what might be specific to my setup here, I would think this would be happening to everyone...

Comment 5 Pete Walter 2024-01-30 21:46:00 UTC
Michel, will you handle the revert?

Comment 6 José Expósito 2024-01-31 09:24:45 UTC
I reverted the change:
https://src.fedoraproject.org/rpms/mesa/c/49e079331021182fda74270284b28b333ab04c5b?branch=rawhide

This issue should be fixed in mesa-24.0.0~rc2-6.

Comment 7 Yannick Defais 2024-02-01 11:11:19 UTC
Created attachment 2014360 [details]
journalctl -b

Process 1428 (sddm-greeter) of user 988 dumped core.

Comment 8 Yannick Defais 2024-02-01 11:12:15 UTC
Ir broke the graphical session for KDE in rawhide here. See attachment above.

Comment 9 Bruno Wolff III 2024-02-14 16:27:15 UTC
I'm not sure if it is the same thing, but I am having problems with mesa-24.0.0-2.fc40 on a machine with a radeon gpu, but not on another one with an intel gpu. Typically lightdm locks up before I can login to an xfce desktop. Because the affected machine is at work and I've been working from home a lot, I'm not sure when the problem started. ssh access worked fine. I'm currently using 23.3.5-1.fc39 as a work around. lshw says the gpu is a FirePro W2100.

Comment 10 Manfred Egger 2024-03-02 18:54:51 UTC
It seems the issue is still present as I experienced similar sddm-greater problems (segfaults) with KDE and radeon graphics.
I reported the issue against sddm as bug 2267124 -- which might be a possible duplicate.

Comment 11 Michel Dänzer 2024-03-04 10:02:42 UTC
The change which caused the issue reported here was reverted again in the very next build. Any issues you are seeing with newer Mesa packages must be separate.