Subject: Setting dumpdev as part of the boot process?
To: None <>
From: Jarle Greipsland <>
List: tech-userlevel
Date: 05/12/2002 16:01:13

I've been experimenting with setting up RAID partitions on a
couple of systems, and used the information found on as a very useful
resource.  A system configured as described in that document ends
up having the root file system on raid0a, while the swap
partition resides on raid1b (and is set up in /etc/fstab).  Since
raid0 != raid1, a kernel with the default {root,dump,swap}=?
configuration file entries will not by itself find the
dump device.  This presumably also holds for configurations like
root on sd0a, swap on sd1b.

I figure this can be solved in (at least) 3 ways:

o Lock down the dump device in the kernel configuration.  This
  will make the kernel less general.  Granted, building a new
  kernel is no big thing, but one cannot then directly copy the
  kernel to a slightly different system, or make significant
  changes to the partitioning and devices on the current system,
  and have it automatically work correctly.

o Make the dumpdev variable available through sysctl(2).  Then
  sysctl(8) and /etc/sysctl.conf can be used to set up the dump

o Create a separate /etc/rc.d/dumpdev script that will use
  'swapctl -D $dumpdevice' to set up the device.  Should
  include a 'BEFORE: savecore' statement. 

What's the correct approach?  Or maybe there already is a
different way to do this?