Just to summarize the responses to my query for the benefit of others...

It seems that "swap generic" does, in fact, preclude using multiple
swap devices.  I now have a config line like:
	config          netbsd  root on wd1 swap on wd1 and wd0 and sd0
and I'm happily swapping to multiple drives.

** One important note that tripped me up:  the first item in the list
   of swap devices must be the one that corresponds to the root
   device.  I had reversed the order of wd0 and wd1, and nothing
   worked. 


However, nobody answered my real question, which was if this is a bug
or a feature.  Why should I have to configure my kernel to have root
on a specific volume just in order to be able to swap to multiple
devices?  Why can't swdevt[] contain all block devices (or even all
???b partitions of all block devices)?  Why does swdevt[] even have to
be initialized at all?  Why not dynamically allocate it when swapon()
is called?

Anyway, I guess I can keep a "generic" kernel around if I should ever
need to move things to another disk, so this problem isn't really
critical.  But it strikes me as annoying for seemingly arbitrary
reasons...


Tom Pavel

Stanford Linear Accelerator Center
pavel@slac.stanford.edu                 http://www.slac.stanford.edu/~pavel/