Getting crashdumps to work (trying to submit crash report)

Hello,

Short version: Is there a list of reasons why crashdumps might not
work? I have dumpdev on a plain (non-gmirror/non-gstripe/etc) device
that is equal to RAM size, but I am not getting a dump for unknown
reasons.

Long version:

so a few months ago I started seeing page faults and/or VM page
related panics while bulk-building ports. I was updating CURRENT every
month or so a the time and using the bulk building as a test case for
ZFS stability issues.

While I have stopped seeing deadlocks and such, which is presumably
due to the fixes taht have gone in, I instead started seeing these
crashes.

I never got crash dumps and didn't know why, and was hoping to wait it
out a bit.

With the advent of 8.0 BETA1 I wanted to give it another shot. I
finally realized that having the dumpdev on gstripes/gmirrors was not
supported, so I reconfigured my system with a dumpdev on a plain
device. I also manually ran the dumpdev command.

After I finally made the box crash again (of course the *one* time I
had hoped to have proper dumpdev configuration it took several days),
I still don't have a dump, and I really don't know why.

Is there a checklist of things that need to be correct in order for
the kernel to do a dump?

Are there certain forms of panic where a dump won't be attempted (such
as in this case where it was some a VM page insert or similar)?

The problem is that I cannot get a stacktrace interactively since the
USB keyboard won't work after a panic, so I really want a
crashdump. But since it takes at least a day often, possibly several
days, to trigger the crash I'm running out of time if I am to produce
some kind of report before 8.0...

-- 
/ Peter Schuller

PGP userID: 0xE9758B7D or 'Peter Schuller <peter.schuller_at_infidyne.com>'
Key retrieval: Send an E-Mail to getpgpkey_at_scode.org
E-Mail: peter.schuller_at_infidyne.com Web: http://www.scode.org