If you've been experiencing what could best be termed as `mysterious
lossage' with recent kernels, you may have been bitten by a bug that I
tracked down earlier today to the epprobe() routine.

Basically, it was trying to write to a port in each EISA slot before
identifying the card, which is wrong, and which, on an ISA-only
machine, ended up wrapping around to one of the registers on the first
DMA controller.  This could likely cause a variety of problems; the
one I saw was a vm_fault() panic in tsleep() right after configuring
root.

If you think you might have run into this, please try again, and sorry
for the inconvenience (though I'm not responsible for that code).


------------------------------------------------------------------------------