Description of problem: Version-Release number of selected component: xorg-x11-server-Xorg-1.20.14-23.fc39 Additional info: reporter: libreport-2.17.11 cgroup: 0::/system.slice/lightdm.service kernel: 6.5.0-0.rc0.20230628git6aeadf7896bf.2.fc39.x86_64 package: xorg-x11-server-Xorg-1.20.14-23.fc39 uid: 0 crash_function: __pthread_mutex_lock_full backtrace_rating: 4 runlevel: N 5 executable: /usr/libexec/Xorg journald_cursor: s=248aa2e59483464baf4cda2e539a2bbb;i=1206cb;b=f3de8567c1574752948d972b808025e3;m=2900fa27;t=5ff552fba3242;x=585f2d3e2a87700a reason: Xorg killed by SIGABRT cmdline: /usr/libexec/Xorg -core -noreset :0 -seat seat0 -auth /run/lightdm/root/:0 -nolisten tcp vt1 -novtswitch comment: type: CCpp rootdir: / Truncated backtrace: Thread no. 1 (18 frames) #0 __pthread_mutex_lock_full at pthread_mutex_lock.c:358 #1 sigset at ../sysdeps/posix/sigset.c:35 #2 get_rounding_mode at ../sysdeps/generic/get-rounding-mode.h:118 #3 round_and_return at ../stdlib/strtod_l.c:216 #9 __sigismember at ../sysdeps/unix/sysv/linux/sigsetops.h:96 #10 sigset at ../sysdeps/posix/sigset.c:67 #12 get_rounding_mode at ../sysdeps/generic/get-rounding-mode.h:118 #13 round_and_return at ../stdlib/strtod_l.c:216 #14 _IO_acquire_lock_fct at /usr/src/debug/glibc-2.37.9000-15.fc39.x86_64/libio/libioP.h:981 #15 _IO_peekc_locked at peekc.c:37 #16 _nl_find_msg at dcigettext.c:885 #17 dixGetPrivateAddr at dri2/../../../include/privates.h:121 #19 dixLookupPrivate at dri2/../../../include/privates.h:164 #20 DRI2Authenticate at dri2/dri2.c:1367 #21 ProcDRI2Authenticate at dri2/dri2ext.c:156 #22 ProcDRI2Dispatch at dri2/dri2ext.c:609 #23 Dispatch at ../../dix/dispatch.c:478 #24 dix_main at ../../dix/main.c:276
Created attachment 1973970 [details] File: limits
Created attachment 1973971 [details] File: dso_list
Created attachment 1973972 [details] File: core_backtrace
Created attachment 1973973 [details] File: open_fds
Created attachment 1973974 [details] File: os_info
Created attachment 1973975 [details] File: backtrace
Created attachment 1973976 [details] File: environ
Created attachment 1973977 [details] File: proc_pid_status
Created attachment 1973978 [details] File: cpuinfo
Created attachment 1973979 [details] File: mountinfo
Created attachment 1973980 [details] File: maps
Occurs on an up-to-date rawhide VM (F37 host) when login in Mate. Cannot run GUI since ~Q4 2022. Thanks for investigating.
Can you please attach the Xorg logs of the system with the crash?
Created attachment 1974028 [details] Xorg server log
Nothing stands out in those logs (I do not even see the crash there). FWIW, nothing has changed in the Xorg server for a very long time, so I suspect some other change might be causing this. * How reproducible is this? * Can you try with the "modesetting" driver instead of "qxl"?
> Nothing stands out in those logs (I do not even see the crash there). That's why my crash description is not more detailed :-( I cannot even be sure the initial cause is in xorg server, but it does abort! As it occurs in rawhide, I did not report it earlier (and also because reporting from text mode is not very ergonomic!). It started failing after an update around oct or nov 2022. This is a quite old rawide instance, but is up-to-date. I did not try creating a new instance. > How reproducible is this? Almost always. I succeeded (< 1%) starting it some months ago by switching the window focus to another one, but had'nt anymore successful try since then. I did not really bound the success to my mouse handling, and think it was just occasional "luck". > Can you try with the "modesetting" driver instead of "qxl"? Tried the modesetting driver using a config file without success. I also tried with other video "hardware" from virtmgr: no luck. Some more info: - I tried creating a brand new user (to avoid possible user-defined outdated files) and login into it: still crashes. - It sometimes has the time to show the panel or even the background before crashing. Seems to depend on concurrency and disk cache state. - If I hide the window during login (minimize or switch to a text screen) and wait a long time, it seems to crash only at the time I bring it back (this might be an impression). - The login screen (lighdm+gtkgreeter) is OK. - If I start it with "startx", it crashes immediately.
AFAICS, it seems to fail in DRI2Authenticate(). Does it make any difference if you disable DRI2 entirely? From "man xorg.conf", in the "ServerFlag" section of your xorg.conf: Option "DRI2" "boolean" enable or disable DRI2. DRI2 is disabled by default.
> Does it make any difference if you disable DRI2 entirely? With Option "DRI2" "off", it works ! And moreover, the crazy thing is it has almost completely unblocked the situation: after returning to the previous state (not disabling DRI2), the crash only occurs less than 10% of logins, even after cold rebooting the host. I don't say the problem has gone, I just say it's hard now to reproduce it and I still do not have any hint about a procedure to reproduce or avoid. Thanks for your work on it.
This bug appears to have been reported against 'rawhide' during the Fedora Linux 39 development cycle. Changing version to 39.