Created attachment 178211 [details] Crashlog created by Google earth On FreeBSD 11.0-p5 Google Earth crashes directly after the start (After the splash screen I see the main window pop up for just a fraction of a second). This is the console output: Fontconfig error: "/etc/fonts/conf.d/10-scale-bitmap-fonts.conf", line 72: non-double matrix element Fontconfig error: "/etc/fonts/conf.d/10-scale-bitmap-fonts.conf", line 72: non-double matrix element Fontconfig warning: "/etc/fonts/conf.d/10-scale-bitmap-fonts.conf", line 80: saw unknown, expected number [1222/230006:ERROR:net_util.cc(2195)] Not implemented reached in bool net::HaveOnlyLoopbackAddresses() Failed to load "/compat/linux/opt/google/earth/free/libinput_plugin.so" because "/usr/lib/libstdc++.so.6: version `GLIBCXX_3.4.17' not found (required by ./libLeap.so)" [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230007:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230008:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230008:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230008:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. [1222/230008:ERROR:nss_ocsp.cc(581)] No URLRequestContext for OCSP handler. pthread create thread: Resource temporarily unavailable Google Earth has caught signal 6. We apologize for the inconvenience, but Google Earth has crashed. This is a bug in the program, and should never happen under normal circumstances. A bug report and debugging data have been written to this text file: /usr/home/yggdrasil/.googleearth/crashlogs/crashlog-585c4ce8.txt Please include this file if you submit a bug report to Google. Sometimes it is signal 11 instead of 6 The mentioned crashlog file is attached When running 10.3 google earth worked fine.
Do you install binary packages or build your own from ports? Which branch: quarterly or latest head? Do you use nvidia-driver? What is the output of "pkg info -x linux nvidia"?
Do you install binary packages or build your own from ports? Which branch: quarterly or latest head? Do you use nvidia-driver? What is the output of "pkg info -x linux nvidia"? There is no package, i build from latest repo. I use the nvidia driver, yes. pkg info -x linux nvidia linux-c6-dri-11.0.7_3 linux-c6-expat-2.0.1_3 linux-c6-fontconfig-2.8.0_1 linux-c6-libelf-0.164 linux-c6-libpciaccess-0.13.4_1 linux-c6-xorg-libs-7.4_6 linux_base-c6-6.8_6 linuxlibertine-g-20120116_1 nvidia-driver-367.44_2 nvidia-settings-355.11_3 nvidia-xconfig-367.35
(In reply to freebsd from comment #2) What is the output of 'md5 /compat/linux/usr/lib/libnvidia-tls.so.367.44'? There are two versions of this file and the package sometimes installs the wrong one.
e0cf316ee43c6ff25b1e5f128c73cb00
That's the right one. In bug 215345 somebody else also has a problem with the most recent Nvidia driver on FreeBSD 11 when trying to use linux-flashplayer. It only seems to happen with multithreaded programs. Could you install graphics/linux-c6-glx-utils and then run /compat/linux/usr/bin/glxgears to see if that works? What is the output of /compat/linux/usr/bin/glxinfo?
Created attachment 178323 [details] glxinfo
The gears run with 60 fps
Dmitry, do you know how to debug this further? There appears to be something wrong with multithreaded Linux programs and x11/nvidia-driver on FreeBSD 11.
Do you have linprocfs and tmpfs mounted as explained in the linux_base-c6 pkg-message (show using "pkg info -D linux_base-c6")? Can you also try the following commands: % /bin/sh $ ulimit -s 32768 $ googleearth
mount | grep linux linprocfs on /compat/linux/proc (linprocfs, local) tmpfs on /compat/linux/dev/shm (tmpfs, local) running with ulimit still crashes
Run 'ktrace -i googleearth' and then 'kdump -H > kdump.txt'. Then attach kdump.txt. (If it's too big you can compress it with xz.)
kdump.txt.xz is still to large, download it from here: http://schukraft.org/kdump.txt.xz Had to run it via sudo. Googleearth seemed to run way longer than before, getting to the point where it presents the Tips window and a popup complaining that it can't detect my graphics card. Thereafter it crashes as usual.
Thanks, can you upload the coredump as well?
It doesn't create a core file. kern.coredump is already 1, so how do I force this?
(In reply to freebsd from comment #14) Right, it can only be done with a debugger because the signal is captured by googleearth which then generates its own crash dump. From the kdump it looks like the problem is caused by virtual memory allocations that fail so I think you should try the ulimit thing again but with lower values: 8192, 4096, 2048. Make sure you do this from within a /bin/sh shell, because it has no effect in csh.
Lower values (in /bin/sh) doesn't do any good either :( If you tell me how to create a proper crashdump I'll do that.
(In reply to freebsd from comment #16) Try to limit the data segment as well: /bin/sh ulimit -s 8192 ulimit -d 262144 googleearth You can probably go all the way down to 8192 or so.
ulimit -s 2048 ulimit -d 8192 googleearth crashes as before
(In reply to freebsd from comment #18) Sigh, can you generate and upload another kdump.txt with these limits? /bin/sh ulimit -s 8192 ulimit -d 65536 ktrace -i googleearth kdump -H > kdump.txt
Sorry this took so long. Now running ktrace -i googleearth as root results in a strange error now: Warning: Unable to create symlink for lock '/root/.googleearth/instance-running-lock'. File exists. Deleting that file or the entire /root/.googleearth dir doesn't change the error. Running ktrace as a regular user seems not to work as it always fails with ktrace: ktrace.out: Operation not permitted
(In reply to freebsd from comment #20) You need to run the commands from a directory where your user has write access. ktrace writes to a file named ktrace.out in the current directory. kdump then processes that file into human readable text.
I'm running it as root, not sure how much more write access I can get ;) And it also worked earlier, not sure whats suddenly different.
(In reply to freebsd from comment #22) Does /compat/linux/root exist maybe?
Yes, it does. I didn't realise there's a different root for linux binaries. http://schukraft.org/kdump2.txt.xz
(In reply to freebsd from comment #24) For a Linux program the kernel first checks /compat/linux/path/to/file before checking /path/to/file. When the program tries to create a file or directory this can be a problem when you run as root because then the file may be created under /compat/linux. This never happens when you run as a regular user because then you don't have write access there. So remove /compat/linux/root and try to run googleearth as a regular user. Maybe you should also remove ~/.googleearth from your user home directory, just in case this is a problem with a corrupted cache or something. cd ~ /bin/sh ulimit -s 8192 ktrace -i googleearth kdump -H > kdump.txt
I rm'ed /compat/linux/root, /root/.googleearth, ~/.googleearth, then sudo /bin/sh cd ulimit -s 8192 ktrace -i googleearth kdump -H > kdump.txt http://schukraft.org/kdump3.txt.xz gives a running googleearth, but without the actual earth part rendering. It spouts a message about an "Unknown Graphics Card", then the main window pops up with everything working except the big map part. Running googleearth as a regular user still fails.
(In reply to freebsd from comment #26) Can you do exactly the same but with "/bin/sh" instead of "sudo /bin/sh"? Also, what is the output of "ls /compat/linux/dev"?
When you run googleearth as root it tries to remove /dev/nvidiactl and then recreate it, but this gets recreated as /compat/linux/dev/nvidiactl. Then it tries to use the latter which fails probably because it isn't a proper device. This means it will never work as root. You should remove /compat/linux/dev/nvidiactl (and anything else in that directory except "shm") and then we'll have to get it working as a regular user.
You were right about a nvidiactl file in /compat/linux/dev, but it still crashed when running as a regular user. But I can't run ktrace as a regular user: $ /bin./sh $ ulimit -s 8192 $ rm ktrace.out $ ls ktrace.out ls: ktrace.out: No such file or directory $ ktrace -i ls ktrace: ktrace.out: Operation not permitted
(In reply to freebsd from comment #29) Are you running from a directory where the user has write access? You can also specify a path with the -f flag like this: /bin/sh ulimit -s 8192 ktrace -i -f /tmp/ktrace.out googleearth kdump -H -f /tmp/ktrace.out > kdump.txt Another possibility is that you have set certain security related sysctl in /boot/loader.conf or /etc/sysctl.conf that prevent debugging by regular users.
I've updated x11/nvidia-driver. See if that changes anything.
Any news here?
Please reopen if this is still an issue.