I recently committed some changes to interrupt-initialization for
Turbochannel machines that've been pending since about January.  I've
tested these _thoroughly on a 5k/200 and 5k/240 (i.e., for all of the
last 3 months), less thoroughly on a 5k/25, and not at all on a
5k/1xx.  I'm pretty sure they're right, though.

I'm not entirely sure, but it looks like the console-code change Ted
retracted via e-mail still isn't in the CVS source.  This may
cause kernels built with my change to fail, with the same
symptom Todd Masco <cactus@hks.net> and I reported last week.


That aside, if someone with a 3min or xine could build a new kernel
and let me know if it boots, I'd appreciate it.  (The symptoms of
failure are that the kernel hangs very, very early on in the boot
process; it's hard to miss.)

This fix also fixes a bug similar to the interrupt-before-config
problem with 3100s; except the window is smaller, and it only really
affects turbochannel option cards that generate interrupts (i.e.,
mostly just me).