Solved Gnome3 crash (freeze) frequency problem

epopen

epopen

Hi.

I had upgraded to x11/gnome3 (before, x11/gnome2, never crashed), but encounter crash (freeze) frequently.

Crash(freeze) situation:
1.Gnome3 UI all of object freeze suddenly, exclude mouse cursor, still move(but click won't work).
2.Move mouse cursor continue, after several seconds it freeze same.
3.Ctrl-Alt-Del won't work, only one solution: Cut-off power by power bottom...

Under is some situation.
1.Wipe all of ports and re-install ports,
In the time, crash(freeze) period: In system on time 48hrs.
2. System used after 2 months, Update all of ports frequency.
Crash(freeze) period became short.
3.Finally, 3 times crash in 30minutes.
4.Back to situation 1

Look like software issue, but haven't useful (e.g. Error) message.
Before % startx, I tried execute script crash.log and check crash.log after restart system same haven't useful message.

Current/usr/local/etc/X11/xorg.conf at http://pastebin.com/cvHzYZC7
Note: It still use VESA video driver, because https://forums.freebsd.org/threads/keyboard-wont-work-in-xorg-after-vt-4-enabled.49096 still can't solve.

Ports list:
Code: 
xorg-7.7_2
xorg-apps-7.7_2
xorg-drivers-7.7_3
xorg-fonts-7.7_1
xorg-fonts-100dpi-7.7
xorg-fonts-75dpi-7.7
xorg-fonts-cyrillic-7.7
xorg-fonts-miscbitmaps-7.7
xorg-fonts-truetype-7.7_1
xorg-fonts-type1-7.7
xorg-libraries-7.7_2
xorg-macros-1.19.0
xorg-server-1.14.7_4,1
xf86-input-keyboard-1.8.0_7
xf86-input-mouse-1.9.0_6
xf86-video-intel-2.21.15_7
xf86-video-vesa-2.3.3_5
xf86bigfontproto-1.2.0
xf86dga-1.0.3_1
xf86dgaproto-2.1
xf86driproto-2.1.1
xf86miscproto-0.9.3
xf86vidmodeproto-2.3.1
I had been crazy about the problem.
Have any good suggest for debug the problem?

Thanks a lot.
 
getopt said:
Have you found any *core files?
Freezing might point to exhausted memory problems. Watch swapinfo.
A wrongly configured/selected driver might also be a problem.
Click to expand...
Hi, Thanks your reply

1.About core file, Only under 2 files.
Code: 
1289506816  4  1 21:46 firefox.core
664993792  4  4 22:21 npviewer.bin.core

2.About exhausted memory(swap?)
Under is swapinfo -h result
Code: 
Device  1K-blocks  Used  Avail Capacity
/dev/zvol/zpool/swap  8388608  137M  7.9G  2%
/dev/ada0p3  2097152  134M  1.9G  7%
Total  10485760  271M  9.7G  3%
Here, running maximum memory usage application is www/firefox like top result,
Code: 
  PID USERNAME  THR PRI NICE  SIZE  RES STATE  C  TIME  WCPU COMMAND
83282 XXX  61  36  0  1857M  1297M uwait  1  57:50  6.40% firefox
Even www/firefox running, swap usage 3% only. (Physical RAM=4GB @ amd64)

3.About wrongly configured/selected driver..
I agree, but can't narrow down which root cause from driver/gnome3/other ports/FreeBSD base now.
Because freeze suddenly and haven't any log about problem.

Crash kernel dump is useful method I think, but I haven't swap partition bigger physical RAM.

Thanks you a lot.
 
epopen said:
Hi.

I had upgraded to x11/gnome3 (before, x11/gnome2, never crashed), but encounter crash (freeze) frequently.

Crash(freeze) situation:
1.Gnome3 UI all of object freeze suddenly, exclude mouse cursor, still move(but click won't work).
2.Move mouse cursor continue, after several seconds it freeze same.
3.Ctrl-Alt-Del won't work, only one solution: Cut-off power by power bottom...
Click to expand...

Can you switch to another virtual terminal via Ctrl+Alt+F1~F8 key or ssh to your machine when the crash happens? If yes then I think this situation just means the window manager (gnome-shell) has crashed. All UI cannot have any response after gnome-shell crashes, including the Ctrl+Alt+Del popup dialog.

gnome-session should automatically restart gnome-shell after the crash. However, you have to wait for gnome-shell to completely dump its core, which can take several minutes.

This is a known problem and it usually happens when you are updating ports or packages on your system because of this bug https://bugzilla.gnome.org/show_bug.cgi?id=739424. GNOME 2 doesn't have this problem because gnome-shell only starts using the file monitor API differently in recent version (maybe 3.12 or 3.14).

A simple workaround is to disable the core dump feature by running sysctl kern.coredump=0. This makes the gnome-shell process terminate immediately when the problem occurs, so gnome-session can restart gnome-shell within seconds. There is a GNOME developer working on fixing this problem, but we don't know when it can be fixed.
 
lantw44 said:
Can you switch to another virtual terminal via Ctrl+Alt+F1~F8 key or ssh to your machine when the crash happens? If yes then I think this situation just means the window manager (gnome-shell) has crashed. All UI cannot have any response after gnome-shell crashes, including the Ctrl+Alt+Del popup dialog.

gnome-session should automatically restart gnome-shell after the crash. However, you have to wait for gnome-shell to completely dump its core, which can take several minutes.

This is a known problem and it usually happens when you are updating ports or packages on your system because of this bug https://bugzilla.gnome.org/show_bug.cgi?id=739424. GNOME 2 doesn't have this problem because gnome-shell only starts using the file monitor API differently in recent version (maybe 3.12 or 3.14).

A simple workaround is to disable the core dump feature by running sysctl kern.coredump=0. This makes the gnome-shell process terminate immediately when the problem occurs, so gnome-session can restart gnome-shell within seconds. There is a GNOME developer working on fixing this problem, but we don't know when it can be fixed.
Click to expand...

Thanks lantw44.
Yesterday system crash (screen=still picture, mouse frozen, all silent), I had been test your method.
1.Ctrl+Alt+F1~F8 key : No any effect.
2.Ctrl+Alt+Del : No any effect.
3.SSH
Before crash, connected to system by SSH(via iPhone SSH client) and run top for alive monitor.
When mouse cursor frozen, top update several seconds.
Finally, top frozen same, New connect denied.

I couldn't test about https://bugzilla.gnome.org/show_bug.cgi?id=739424
Because example program build fail about glib not found issue.

I think H/W like RAM defect?
Have recommended about deep/complete RAM test method?

Thanks all a lot!
 
Last edited by a moderator:
epopen said:
I think H/W like RAM defect?
Have recommended about deep/complete RAM test method?

Thanks all a lot!
Click to expand...

There are some RAM test programs available in ports. You can find them with echo /usr/ports/*/*memtest*. Is the system crash problem reproducible using other desktop environments?

I think there should be some messages printed on ttyv0 (Ctrl+Alt+F1), such as kernel panic and backtraces, but I don't know whether it is possible to see them using watch v0 in a ssh session.
 
lantw44 said:
There are some RAM test programs available in ports. You can find them with echo /usr/ports/*/*memtest*. Is the system crash problem reproducible using other desktop environments?

I think there should be some messages printed on ttyv0 (Ctrl+Alt+F1), such as kernel panic and backtraces, but I don't know whether it is possible to see them using watch v0 in a ssh session.
Click to expand...
Thanks lantw44.

I got twice crash between your and this message...
I had been try two memory test ports.
1.sysutils/memtest86+ Reboot when boot it's modules in boot prompt.
2.sysutils/memtest It executes OK, but it tests low address (0~200MB) memory only and test pass (mlock fail in message).

About ttyv0, I can't switch it and other all when screen freeze started (mouse still work several seconds).
About watch v0 in a ssh session, it works.

At freeze start, only browser's ordinary message in ttyv0 and stall, No any one about freeze.
New connect NG.

Thank you a lot.
 
You must log in or register to reply here.
Back
Top