A day or two ago, I wrote to current-users about an argument mismatch
between physio() as defined in kern/kern_physio.c and physio() as
called by swread() (and swwrite(), incidentally).  mrg sent a patch to
vm_swap.c that amounts to this (context removed, line numbers not
changed to compensate):

>    *** 271,277 ****
>    ! 	return (physio(swstrategy, NULL, B_READ, minphys, uio));
>    --- 271,277 ----
>    ! 	return (physio(swstrategy, NULL, dev, B_READ, minphys, uio));
>    ***************
>    *** 280,286 ****
>    ! 	return (physio(swstrategy, NULL, B_WRITE, minphys, uio));
>    --- 280,286 ----
>    ! 	return (physio(swstrategy, NULL, dev, B_WRITE, minphys, uio));

I actually did dive in, "machete in hand", and made exactly that
change, after reading over what physio() did with the dev argument and
deciding it was sufficiently noncritical that I could risk trying it.

It works for me too.  I'll send-pr it, to make sure it doesn't slip
between the cracks.  (I did a fresh sup this morning and it hadn't been
fixed in that sup, so it's at risk for getting lost.)

					der Mouse

			    mouse@collatz.mcrcim.mcgill.edu