> Found the problem. There's a typo in mesh.c, it would return status i=
nfo
> like SCSI_CHECK in xs->xs_status, but scsipi looks for it in xs->stat=
us
> - with the result that no one ever requested sense data.
> With this fixed both drives spin up automatically and I guess a bunch=
 of
> other weirdnesses ( like occasional bogus 'drive offline' messages wh=
en
> probing ) will just disappear.

I suspect there's a similar bug lurking in the sii driver used on
DECstation machines.  At least I'm experiencing a similar symptom: on
our local 3100, the secondary drive (non-system disk) isn't spun up
automatically by NetBSD either, requiring a manual spinning via
scsictl and a second reboot after a power cycle...

Regards,

- H=E5vard