Description of problem: Version-Release number of selected component: gnome-shell-3.38.1-2.fc33 Additional info: reporter: libreport-2.14.0 backtrace_rating: 4 cgroup: 0::/user.slice/user-42.slice/session-c3.scope cmdline: /usr/bin/gnome-shell crash_function: vsubq_f32 executable: /usr/bin/gnome-shell journald_cursor: s=29f328ea0bce481a9492a0116e0809fe;i=1889e;b=0d1e78c7f04249b1b7c55d845adc965f;m=3d589251;t=5b1f5403467f0;x=18af5e9c83d176ed kernel: 5.9.0-36.fc34.armv7hl rootdir: / runlevel: N 5 type: CCpp uid: 42 Truncated backtrace: Thread no. 1 (10 frames) #0 vsubq_f32 at /usr/lib/gcc/armv7hl-redhat-linux-gnueabi/10/include/arm_neon.h:1960 #1 graphene_vec3_subtract at ../src/graphene-vectors.c:834 #2 _cogl_util_get_eye_planes_for_screen_poly.constprop.0 at ../clutter/clutter/clutter-stage.c:752 #3 setup_view_for_pick_or_paint at ../clutter/clutter/clutter-stage.c:828 #4 clutter_stage_do_paint_view at ../clutter/clutter/clutter-stage.c:850 #5 meta_stage_paint_view at ../src/backends/meta-stage.c:259 #6 clutter_stage_paint_view at ../clutter/clutter/clutter-stage.c:875 #7 paint_stage at ../clutter/clutter/cogl/clutter-stage-cogl.c:411 #8 clutter_stage_cogl_redraw_view_primary at ../clutter/clutter/cogl/clutter-stage-cogl.c:604 #9 clutter_stage_cogl_redraw_view at ../clutter/clutter/cogl/clutter-stage-cogl.c:699
Created attachment 1722575 [details] File: backtrace
Created attachment 1722576 [details] File: core_backtrace
Created attachment 1722577 [details] File: cpuinfo
Created attachment 1722578 [details] File: dso_list
Created attachment 1722579 [details] File: environ
Created attachment 1722580 [details] File: exploitable
Created attachment 1722581 [details] File: limits
Created attachment 1722582 [details] File: maps
Created attachment 1722583 [details] File: mountinfo
Created attachment 1722584 [details] File: open_fds
Created attachment 1722585 [details] File: proc_pid_status
Created attachment 1722586 [details] File: var_log_messages
Note that I'm using a non-default /proc/cpu/alignment option (echo 3 or 5) to generate the SIGBUS. I've noted that from kernel.log, from an earlier SIGBUS. -- oct. 18 18:02:04 jetson-tk1 kernel: Alignment trap: gnome-shell (2585) PC=0xb3441ed0 Instr=0xf4400aef Address=0x01a41458 FSR 0x811 oct. 18 18:02:04 jetson-tk1 kernel: Alignment trap: not handling instruction f4400aef at [<b3441ed0>] oct. 18 18:02:04 jetson-tk1 kernel: 8<--- cut here --- oct. 18 18:02:04 jetson-tk1 kernel: Unhandled fault: alignment exception (0x811) at 0x01a41458 oct. 18 18:02:04 jetson-tk1 kernel: pgd = f3dd6879 oct. 18 18:02:04 jetson-tk1 kernel: [01a41458] *pgd=ba323835 oct. 18 18:02:05 jetson-tk1 kernel: Alignment trap: gnome-shell (2585) PC=0xb3442008 Instr=0xf4612aef Address=0x01a41458 FSR 0x011 oct. 18 18:02:05 jetson-tk1 kernel: Alignment trap: not handling instruction f4612aef at [<b3442008>] oct. 18 18:02:05 jetson-tk1 kernel: 8<--- cut here --- oct. 18 18:02:05 jetson-tk1 kernel: Unhandled fault: alignment exception (0x011) at 0x01a41458 oct. 18 18:02:06 jetson-tk1 kernel: pgd = f3dd6879 oct. 18 18:02:06 jetson-tk1 kernel: [01a41458] *pgd=ba323835รจ --- Note sure how to read this, but the Instruction looks, but the Instruction address looks located in libEGL.so. I don't see libEGL been used directly in the bt... b332e000-b333a000 r-xp 00000000 00:1d 433142 /usr/lib/libEGL.so.1.1.0 b333a000-b334a000 ---p 0000c000 00:1d 433142 /usr/lib/libEGL.so.1.1.0 b334a000-b334b000 r--p 0000c000 00:1d 433142 /usr/lib/libEGL.so.1.1.0 b334b000-b334c000 rw-p 0000d000 00:1d 433142 /usr/lib/libEGL.so.1.1.0
re-assigned to gcc, as arm_neon.h is from there.
I've reported an issue in graphene upstream that is compiled with neon by default. This should have worked on the jetson-tk1 that has neon support, but for some reason, the neon code https://github.com/ebassi/graphene/issues/215 This was reported at rhbz#1937023 I also saw a previous rhbz related gcc for graphene neon been mis-compiled... *** This bug has been marked as a duplicate of bug 1937023 ***