[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openmosix-general
Subject:    [openMosix-general] Unable to handle kernel NULL pointer dereference at virtual address (was: stress
From:       CSights <csights () fastmail ! fm>
Date:       2006-05-20 13:39:52
Message-ID: 200605200939.52678.csights () fastmail ! fm
[Download RAW message or body]

Hi Roger,
	I also use Gentoo and have been having "Unable to handle kernel NULL pointer 
dereference at virtual address" problems with openMosix as well. I tracked 
down some earlier mentions of it on the mailing list.

        2004-12-16 someone using (non-official gentoo) kernel version 2.4.27 
was getting virtual pointer problems using perl[1]. Tab (of 2.6 fame) 
responded 'that was supposed to be fixed in 2.4.26'[2].

[1] http://sourceforge.net/mailarchive/message.php?msg_id=10329688
[2] http://sourceforge.net/mailarchive/message.php?msg_id=10329687

	Later (2006-03-15) someone using (non-official gentoo) openmosix 2.4.30 also 
reported the openmosix stress test freezing the machine [3].  There were a 
number of suggestions as to causes of the problem and a few suggested 
solutions (don't use tls, maybe reiserfs, use kernel 2.4.22) [4].

[3] http://sourceforge.net/mailarchive/message.php?msg_id=14983523
[4] http://sourceforge.net/mailarchive/message.php?msg_id=15125946

	My own NULL pointer problem was resolved by using official openmosix patch 
2.4.24-2 and gcc 3.3.6.
	Here is the test grid using my C++ program and two official openmosix patches 
(patched against the vanilla kernel). "success" means the program completed 
without NULL pointer for as many times as I wanted (around 10) and "failure" 
means a NULL pointer almost immediatedly.

om patch	gcc	result
2.4.24-2	3.3.6	success
2.4.24-2	3.4.5	failure
2.4.26		3.3.6	failure after extended testing
2.4.26		3.4.5	failure

	It appears that the NULL pointer problem depends both on the compiler version 
AND the kernel patch version.  om-patch 2.4.24-2 and gcc-3.3.6 "works for 
me", but if om-patch 2.4.24-2 is compiled with gcc-3.4.5 the NULL pointer 
problem appears after a few runs of the C++ program.
	Someone who knew enough about the difference in how the two versions of gcc 
produced code would probably be able to know where in the openmosix patch 
things are not quite right.
	If you find anything that works let us know!

Good luck,
	C.


-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd_______________________________________________
openMosix-general mailing list
openMosix-general@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/openmosix-general

[prev in list] [next in list] [prev in thread] [next in thread]