It turns out I misunderstood the semantics for the no-rewind device.
in using /dev/nrst0, the tape remained "mounted" and hence the driver
didn't figure it out when I changed the tape. Using /dev/rst0 or 
using /dev/nrst0 + "mt offline" at the end of the first tape results
in the correct behavior.  man st(4) also helped clear it up.

BTW, "mt eew" does nasty things to my tape drive: bad, short dump
sessions, EOM reported after 10s of feet, etc. But I didn't try
to follow up on that. After reading st(4) and the driver comments,
perhaps "mt eew" should be disallowed for fixed blocksize drives?

	Sorry for any confusion.

	- Scott

Hauke Fath wrote:
> 
> At 22:49 Uhr +0200 3.10.2002, Manuel Bouyer wrote:
> >> Everything is fine until I insert the second tape: dump thinks it's already
> >> at the EOT and aborts (but it's certainly not, since I just rewound it).
> >
> >Did you try 'mt eew' before dump ? I had to use this for multidumps on
> >hexabyte ...
> 
> I don't think it is 'early warning for EOM' he gets. His is a QIC drive...
> QIC drives transparently write two filemarks at EOD, and st(4) doesn't play
> nice with this habit.
> 
> I had to hack st.c from NetBSD 1.3 to 1.5.x to get it to work with Tandberg
> QIC drives - the issue bites Amanda, too. 1.6 appears to fix the issue, so
> an upgrade would help here. Scott, if that is not an option, ask me for the
> 1.5 patches.
>