> Does the tape drive work on that controller?  I could almost guess that

Yes and no.  See (kern/5647: ARCHIVE, Python: read errors "Incorrect
Length Indicator Set")  tape writes ok and does appear to read at
least some data (I've never let it finish because of the volume of ILI
errors poluting /var/log/messages).

> you have an IRQ misconfiguration, or something.  During device probes,
> polling is used, but it's all interrupt-driven after autoconfiguration
> is complete...

The fact that tape I/O does not time out, I assume rules out an IRQ
problem, but does not rule out some sort of problem with the
controller I guess.  The full dmesg output is included in both PR's
though the sd3 probe info is missing as I can't write it all down
before it scrolls off and the boot never gets far enought to be able
to have another look at it.

--sjg