Linux 2.6 Kernel Panics


log in

Advanced search

Questions and Answers : Unix/Linux : Linux 2.6 Kernel Panics

Author Message
tRaV
Send message
Joined: 19 Dec 99
Posts: 2
Credit: 5,642
RAC: 0
United States
Message 6758 - Posted: 12 Jul 2004, 2:55:11 UTC

Has anyone been successful in getting the setiathome client to run on the linux 2.6.x kernel? Mine seems to panic within several hours of my firing up boinc (well, any version of setiathome, actually) - possibly when the first work unit gets finished. I tried to resolve the issue by compiling boinc on my 2.6 box, but I realized that the binary that actually does the work (and is causing kernel panics) is downloaded from the project after I attach to the project.

Granted, this shouldn't be possible (especially as a non-root user) and represents an issue with the 2.6 kernel, but I'm working on debugging that separately (trying to setup a serial console to capture the panic output). I just wondered if it was specific to my system (which was upgraded to FC2 from RH9), or if anyone else had tried it successfully on the 2.6 kernel?

Unfortunately, I probably won't be able to contribute any CPU cycles from this system until I find a way to capture the kernel error and get it fixed.

thanx,
tRaV

Darren
Volunteer tester
Avatar
Send message
Joined: 2 Jul 99
Posts: 259
Credit: 276,590
RAC: 61
United States
Message 6782 - Posted: 12 Jul 2004, 3:56:22 UTC

"I just wondered if it was specific to my system (which was upgraded to FC2 from RH9), or if anyone else had tried it successfully on the 2.6 kernel?"

Can't help you at all with the kernel panic (when mine panics, I panic even more), but I do run the software without problems on 3 of my hosts using Mandrake 10 with kernel 2.6.3, so I can at least let you know it's not the kernel in general.


Bill
Send message
Joined: 13 Jun 99
Posts: 1
Credit: 664,181
RAC: 0
United States
Message 8472 - Posted: 15 Jul 2004, 23:38:05 UTC

I'm running BOINC on FC2 (2.6.6) with no observed problems.
Could be a harware issue, bad memory, running hot, etc or it could be a problem if you upgraded from RH9 rather than a clean install.

Bill

Brad & Pam
Send message
Joined: 29 Jul 02
Posts: 1
Credit: 1,758
RAC: 0
United States
Message 8570 - Posted: 16 Jul 2004, 4:54:39 UTC

I also run boinc on 2.6.7 kernel here. Only been running it for about 24 hours now but it runs fine.

Also you said that there might be an issue with the 2.6 kernel because the boinc client downloaded the seti@home client and was using it. Not really, it installs it as the user you are running boinc as and in that user's home directory. Look for it here:

/path/to/boinc/projects/setiathome.berkeley.edu/setiathome_3.08_i686-pc-linux-gnu

its perfectly legal for a program to change files associated with its user. If it were to put it in /usr/bin or something similar, then you would have a problem.

The kernel panic youre getting is pretty odd though. The only time I can remember seeing a kernel panic is screwed up drivers or screwed up hardware. Im sure there can be other causes though. Can you give any more details about the error?

tRaV
Send message
Joined: 19 Dec 99
Posts: 2
Credit: 5,642
RAC: 0
United States
Message 8692 - Posted: 16 Jul 2004, 13:07:35 UTC - in response to Message 8570.

> Also you said that there might be an issue with the 2.6 kernel because the
> boinc client downloaded the seti@home client and was using it. Not really, it
> installs it as the user you are running boinc as and in that user's home
> directory. Look for it here:

My concern had been that, while I could compile the boinc binary myself, the seti@home client is a pre-compiled binary that gets downloaded from the project based upon boinc's assessment of my architecture. But, that's probably not the issue, since other folks have been successful in running the same binary with the 2.6.x kernel.

> Can you give any more details about the error?

It appeared to be an I/O related error. I filed it on bugzilla.redhat.com as Bug ID 127756. The entire console output is there as an attachment, and the most updated info on the issue should be there.

My two most hopeful theories are currently either that my choice of filesystem and software raid1 usage might be digging up a kernel bug, or that my upgrade path from RH9 to FC2 might have convoluted my system enough to jack it up.

I'm planning on taking Warren's advice and trying the latest rawhide kernel, since I'm still working off of the latest kernel from the released updates, as opposed to the development repository.

Darren
Volunteer tester
Avatar
Send message
Joined: 2 Jul 99
Posts: 259
Credit: 276,590
RAC: 61
United States
Message 8754 - Posted: 16 Jul 2004, 16:21:41 UTC - in response to Message 8692.


> My concern had been that, while I could compile the boinc binary myself, the
> seti@home client is a pre-compiled binary that gets downloaded from the
> project based upon boinc's assessment of my architecture. But, that's probably
> not the issue, since other folks have been successful in running the same
> binary with the 2.6.x kernel.

If you're compiling your own boinc, you also have to compile your own seti. It's not possible to use the pre-compiled seti with a self compiled boinc.

You have to extract both files into a common directory, which will create a boinc and a seti_boinc subdirectory. Then you must compile the boinc client first, followed by the seti client (as seti will need files from the boinc directory). After this, you will have to create your own app_info.xml file to allow boinc to ask for work correct for the seti version you compiled.

Put the boinc_client executable in your main boinc directory, then put both the seti executable you compiled and the app_info.xml you created into your ~/projects/setiathome.berkeley.edu directory under your main boinc directory.

The seti source comes from here - http://boinc.berkeley.edu/seti_source/ - and an example of the app_info.xml file can be found here - http://boinc.berkeley.edu/anonymous_platform.php


Questions and Answers : Unix/Linux : Linux 2.6 Kernel Panics

Copyright © 2014 University of California