Hello,

I have searched these and many forums to help with some desktop freezing, which some folk seem to be having issues with; however, their solutions do not seem to work for me, and some never have been resolved. I am running Chakra GNU/Linux with a NVIDIA 560GTX and the latest NVIDIA proprietary drivers. I have tried older and open source drivers with less success. Seemingly random, the desktop freezes every so often where I can still move the mouse - there there is a significant delay. If I am listening to music, the music continues to play without any problems. I can "force" a desktop freeze if I begin to mesh a heavy computational model, which starts to use some RAM (about 5GB out of my available 32 GB). I have tried a few things, such as enabling the persistence mode via:

Code: Select all

/usr/bin/nvidia-smi -pm 1

which seemingly helped for some time. However, the freezes have come back with a vengeance. I did not upgrade kernels in that time frame and am otherwise unsure what might have happened. I freeze more often if I had suspended to RAM than from a fresh reboot. I have also tried turning of V-sync, which has not helped. Many other forums are also very much data from 2010 and 2011. I am hoping to get some help with this and hopefully resolve the issue.

Thank you.

This definitely seems to be a IO related issue, or a memory utilisation problem. Linux systems usually only behave in this manner when under heavy IO load (such as when swapping, or copying amount of data) or due to a lack of free disk space, particularly on /run (previously /var/run/), /var/tmp, or /tmp.

Does the system eventually recover from this state if given enough time?

Hello,

I have an available 700GB of hard disk memory and usually, running a reactor physics code Serpent (http://montecarlo.vtt.fi/), I have no issues with memory usage. I have reached the 32GB limit a few times, and the system goes become fairly unstable and Serpent will kill itself and the system will return to its normal state.

However, with the other code that I use (STAR-CCM+ by Cd-adapco (http://www.cd-adapco.com/products/star_ccm_plus/), I don't use more than 5GB and the system freezes completely. It does not recover even after 8 hours (the music still plays just fine, though).

Also, it still does freeze without using STAR-CCM+, particularly when I return from standby to RAM after sometime - I have not been able to determine what causes the freeze.

Thanks.

Hello,

I believe that I have a theory. When I use Chromium/Firefox while I am meshing, I freeze - when I do not, for example, right now I am using Rekonq, I do not freeze. I kept htop and top open to watch the processes and when the computer bogs down, like it is about to freeze, kwin takes 100% CPU (that's one of six cores). So, I'm thinking, since Rekonq plays nicer with KDE, perhaps there is an issue with Chromium/Firefox that causes kwin to crash and freeze the desktop? The refresh rate on htop/top is not high enough to capture resources during the actual freeze.

I do lose all keyboard input, so I am unable to save, close programs, change tty to reset the window manager, etc. Is there any way to confirm this, or is it maybe outlandish?

Thanks.

Hmm, that is quite unusual behaviour. One thing you could check is the number of processing threads / processes being used, because the CPU time scheduler could be overloaded if too many are being used.

This will usually be reflected by the system registering a high load, with full CPU utilisation.
If the application in question uses GPU based acceleration, then it is also possible that it is monopolising the GPU, preventing KDE / X from being able to display anything. If this is the case, then if you login remotely the system should still be quite responsive.

Hello,

Only about 35% of my total processing power is used (pretty much just 2 out of 6 CPUs) - that's one for the software and the second for kwin.

The GPU does not appear to be under much use, but

Code: Select all

nvidia-smi

is the best I can do:

Code: Select all

Sun Mar 31 17:31:04 2013       
+------------------------------------------------------+                       
| NVIDIA-SMI 4.313.26   Driver Version: 313.26         |                       
|-------------------------------+----------------------+----------------------+
| GPU  Name                     | Bus-Id        Disp.  | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap| Memory-Usage         | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 560          | 0000:01:00.0     N/A |                  N/A |
| 43%   43C  N/A     N/A /  N/A |  21%  214MB / 1023MB |     N/A      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Compute processes:                                               GPU Memory |
|  GPU       PID  Process name                                     Usage      |
|=============================================================================|
|    0            Not Supported                                               |
+-----------------------------------------------------------------------------+


Jesse

Hello,

And, the number of threads is 308, with all but two sleeping (those currently being htop and STAR-CCM+)... I'll closely monitor it when it freezes again and report back with the change. I actually have to get some results now, so I'm going to try to keep it stable.

Many thanks,

Jesse

Okay, that nvidia-smi output does not indicate any particular stress on the GPU unfortunately. It may be worth checking the CPU temperature as well however. The different workloads may be triggering different areas of stress in your setup.

Hi,

The temperatures remain pretty stable on my machine, but when it freezes, I can't see much. I did happen to be able to swap tty when it froze, and kwin was hung at 100% for 27 hours (while waiting for a simulation to complete) and swapping back - I could no longer do anything on the machine and had to hard reboot.

Have you tried disabling Desktop Effects during the meshing runs?

It may be that the simulation application uses the Nvidia GPU in a way which severely interferes with the operation of KWin (which would cause the apparent "freeze" of the desktop due to the way compositors work).

Hello,

Thank you for all the help. I give that a shot and report back by the end of the week if the computer hasn't froze - I experience it twice daily, depending on my activities. If it works out, then I'll mark it solved. As for what I did (for others that might run into this problem), simply:

Code: Select all

alt+shift+F12

Cheers.

Hello,

Over the course of the week, the desktop froze 37 times. I logged what my activity was and the correlation is not clear. Some times it was swapping workspaces and others creating a new tab in Chrome. I did very little computational work with the machine because of the instability. I am wondering if there is a generated log that I could generate or that is already generated that can possibly lead me to the cause.

The computer freezes about 30-45 minutes into being operated. I will reformat and reinstall everything next weekend, but in the meantime is there any other potential fixes or mean to log what might be going on?

Many thanks.

Given the increasing nature of these freezes, it may be wise to perform a hardware test and check that everything is operating properly. In particular i'd recommend checking temperatures (preferably in the BIOS) and running memory tests.

The behaviour you describe is quite unusual unfortunately, particularly considered your system is not swapping, or involved in heavy disk activity. Can you use another computer to keep an eye on your problematic computer (via SSH for instance) to see if the problem only affects the graphical desktop - or if it affects the entire system. If it is restricted purely to graphics (in which case the remote connection should be unaffected by the instability) then this could very well be a bug within X or the Nvidia drivers, or a hardware fault in the Nvidia card.

Hello,

I think that I have solved the issue. I reverted to an older "lts" kernel for Chakra (I don't think there was a new one between that previous post and last of mine). Things are running just fine with a few hangs (lasting 15-45 seconds) that would normally be a freeze, I'm assuming. Without effects, I don't experience any hangs at all and I'm quite okay with that. I will mark solved if this trends continues for the week.

In regard to your questions: ssh did work just fine and all I had to do was kill kwin. Unfortunately, that would make things mostly unusable - I couldn't figure out how to start kwin again to get things displaying properly. Temperatures are normal and I'm assuming the memory is okay, I have 32GB that I'd rather not wait on a memtest for... but for the cases that I run, if there are memory problems, it would crash with relevant errors.

Thanks for all the help.

Jesse

With regards to starting KWin up again - running "kwin --replace" in KRunner (accessible via Alt + F2) should start KWin.

That killing KWin caused the desktop to unfreeze confirmed my suspicion that it is graphics related - your setup may have exposed a bug in either a kernel subsystem or a previously unintended interaction between a subsystem and the Nvidia driver.