On Wed, 14 Jan 2009 18:34:43 +0100 Xav' wrote: > On Wed, 14 Jan 2009 08:14:01 -0500, David Relson > <rel...@osagesoftware.com> wrote: > > On Wed, 14 Jan 2009 14:01:28 +0100 > > Daniel Pielmeier wrote: > > > >> 2009/1/14 David Relson <rel...@osagesoftware.com>: > >> > > >> > I noticed the "taint" message and don't think that's right. How > >> > do I identify the tainted module? > >> > > >> > >> Jan 13 21:35:04 osage Pid: 13119, comm: rrdtool Tainted: G D W > >> 2.6.28-gentoo #11 > >> > >> rrdtool? > >> > >> > > It seems according to your trace that the Python script > system_health.py does not stop to do segfaults while it is executed > by cron, so maybe you can start checking this script to see if it > does not write anywhere in the memory and then cause the kernel to > "bug".
The LWN article is about RTSJ -- real-time services for java. Having done some work with real-time and embedded systems I can understand that access to physical memory is useful. However, I'm not running java and _not_ using RTSJ. As regards tainting, after a BUG the kernel sets its status to "tainted". After all, having encountered a BUG problem, there's no telling what else may be fubar'd. That explains one problem. system_health.py is python code that reads system info and generates "health" graphs for the computer. It's presently running under Python 2.6. Does python even have RTSJ-like capabilities??? Looking further at the log messages (see attached file): system_health.py - segfaults at 20:50, then runs fine at 20:55, 21:00, 21:05, 21:10, and 21 At 20:59:06 is a stack backtrace. This is 9 minutes after the system_health.py segfault and a successful run of system_health. The problem seems to be mmap related. Unfortunately it seems that part of the oops(?) is missing since the first line has a right paren without a preceding left paren. 21:20:01 -- another stack backtrace -- different from previous 21:27:58 -- another stack backtrace -- different from all previous ones 21:35:04 -- first mention of BUG ... and tainting Evidently bad things were happening for a while before the first BUG. So, back to an earlier question: To whom should these backtraces be sent -- in hope that they'll shed light on a previously encountered problem and aid in a fix being generated ?? Regards, David
