Charles,

I was certainly under the impression that when Mark implemented the
jumpslot format that we designed, it did *not* involve any I-cache
flushing apart from one initial flush the first time the fixup code was
entered (from memory, but I can't be sure on this), ie one cache flush per
shared library linked. I don't have a source tree to hand however to check
this. Could you be so kind as to take a moment to illustrate, via e-mail,
with example code, what it is that you think is happening in the code?

I take your point about reentrant jumpslots. Could it be possible at the
moment for an alarm signal to be delivered in the midst of a jumpslot
fillout which could cause a trash if reentrance is required? Given we
don't have proper user threads, presumably this is the oly occasion under
which this could occur?