> Are the problems with IRQs on single-processor Intel machines that we were
> having around 2.1.30 thought to be fixed in 2.1.33? I'm seeing some SCSI
> problems that look remarkably similar to the ones I was getting then, and
> there does seem to be an interrupt component to it (things lock up until I
> type at the keyboard, for example).
Ah, no, it looks like this was a false alarm. I changed to a different
driver, and my SCSI problems are gone. It looks like they were due to
bugs in the AM53C974 driver. The wait-for-keypress thing was because when
it gets confused, it actually deliberately waits for your to press a key,
inside its interrupt routine (!), presumably as a debugging aid.
Whoops.
phil