I have committed the dma changes to the netbsd-current cvs tree that
allowed the scsi bus to work.  I also upgraded the dma driver to
handle premature dma transfer shutdowns, and changed the receive
ethernet driver to correctly restart when they occur.  This should fix
pr 11175.

This means that the next68k netbsd-current sources run again.  It also
may mean that scsi will work in -current, but I didn't enable it by
default since I haven't done any testing on it since last December.
The esp (scsi) driver will probably need work to handle dma aborts.

I have pretty much used up my time to work on the port for the next
month, but hopefully it is running again so that others can hack on
it.  The following tasks are on my todo list, but probably won't
happen right away without help: 

  . fix the netbsd-1-5 branch to avoid oversize packet errors.  This
    either means to pull up the changes just made to -current, or to
    hack the netbsd-1-5 sources to truncate the packets before handing
    them up.  If the latter works, it's probably the right change for
    the release branch.

  . move -current to elf, following the lead set by the rest of the
    m68k ports.

  . test the esp scsi driver, make it stable.

  . work on disklabel changes to interoperate with nextstep.
    see pr 11938

  . look into status of non-512 byte sector support.

  . poke volunteers about X.  There is a dma channel reserved for
    video dma transfers, that might make performance not suck.

If anyone wants to be useful, the upgrade to elf shouldn't be too
hard.  There should be people who can help on the port-m68k mailing
list.

Otherwise, I could use reports from anyone who tries the scsi driver
with -current.

Thanks,
Darrin