Yup, thanks. I found that, but in the event I fixed the problem by
rearranging things so that the aic7xxx now has its own IRQ rather than sharing
one, which seemed an all-round better bet.
BTW, I'm getting these when I read from the aic7xxx driver's /proc file:
aic7xxx: possible overflow at loop 0:8
aic7xxx: possible overflow at loop 0:8
aic7xxx: possible overflow at loop 3:8
p.