In message <20011217132013.A21657@WintelKiller.T> you write:

|> > The disk is an IBM DNES-318350N which has worked flawlessly for about two
|> > years connected to an FSBE/S in a Sparc-2 running OpenBSD-2.3. It has a
|> Hmm. A SPARCStation 2 won't do synchronous/"Fast"-SCSI.

On an FSBE/S, it should. At least with SunOS 4.1.4, scsiinfo used to
tell me drives were 10 MB sync.

|> How are the drives/controllers detected by NetBSD (dmesg output)?

Don't remember the exact wording, but basically: vendor/type/rev IDs are
recognized at appropiate SCSI ID, and speed is negotiated to 10 MB sync.

|> Is jumper J0202 set on the FSBE/S?

Will check. BTW, the other two drives on the same bus work fine, even with
the "bad" one connected and in its locked-up state.

|> Does the drive work well with the onboard SCSI host adapter?

Haven't tried yet.

|> Does the drive work well when it's alone on the bus?

Will have to try. A bit difficult though, since I have no other working
Sparc-10 at hand, and all of this box's three buses are well populated.

|> Are the terminators of the "active" kind? (needed for Fast-SCSI)

Yes.

|> a) the end of the bus/last drive is terminated

Yes.

|> b) no other drive has termination on

Yes.

|> c) no drive delivers "termination power" (sounds evil, huh?) to the SCSI
|> bus (needed for the active termination)

Many recent drives don't have TERMPWR jumpers any more. Had they, I'd
have removed them first thing after unpacking the drive...

|> d) if you have some kind of multimeter/DMV you could check if the PTC "fuse"
|> of the FSBE/S is still working, if it's not enable termpower on one of the
|> drives (though very unlikely that they're dead)...

Well the TERMPWR indicator on the terminator is lit, so I assume TERMPWR
is supplied by _someone_.

|> If this all doesn't help you can disable synchronous negotiation for
|> some/all targets with flags in your kernel config file (see manpage for
|> esp(4)) - performance will degrade of course, but maybe you can get it
|> working.

Will take a look.

Thanks so far,

	Volker