Logging panic messages

[Tony Arcieri]

On Fri, Feb 04, 2005 at 04:28:54PM -0500, Chuck Swiger wrote:
> You might set hw.physmem to a smaller value in loader.conf to fit within 
> the amount of swapspace which is available:
> 
>         set hw.physmem=<value>                  MAXMEM (i386 only)
> 
>                 Limits the amount of physical memory space available to
>                 the system to <value> bytes.  <value> may have a k, M or G
>                 suffix to indicate kilobytes, megabytes and gigabytes
>                 respectively.  Note that the current i386 architecture
>                 limits this value to 4GB.

Thanks for the suggestion, I'll give that a try, and see if I can get
it to generate a corefile the next time it crashes.

> last and dmesg don't give you anything, hmm?

No, and neither do the logfiles, but I've read this is intentional to prevent
the kernel from executing complex kernel code like the VFS after a crash
to prevent deadlocking.

> That's unforunate, hmm, you might try leaving an ssh session logged in doing
> a tail -f on /var/log/messages and see whether you can get anything from
> that.

All networking hangs whenever it crashes.

I think my best bet is to try to generate a kernel core file now, but I was
wondering if there had ever been consideration of using the dump infrastructure
to record KTR dumps rather than an image of the kernel core.  It seems like
KTR would be a wonderful facility to debug the cause of an OS crash, but
the current infrastructure makes it relatively impossible to use in a situation
like mine.

Tony Arcieri