Bouncing.  Interesting stuff.

I approach this from the naive clue-free side of things, as I'm not
knowledgeable on such matters, but if I understand things correctly,
all that needs to be done is for some regions of memory to be locked
exclusively for the purpose of bouncing data (I believe this is what
is referred to as a "bounce buffer" -- is this correct?).

Could it not be handled by using semaphores?  I.e.  the bounce buffer
has a couple of items at the start of it which denote pages, byte
remainder, destination, and the fact that the data has just been bounced
in.  The kernel checks this and sees the "BOUNCE" semaphore set,
bounces NPAGES+REMAINDER bytes from the buffer and unsets the BOUNCE sema-
phore.  There could be several of these.

The above model is an example constructed by someone who is, technically
speaking, on the outside looking in.  Bouncing seems to be a dark box
(not quite black) to me -- I can't quite see what needs to happen, but
I believe that my analysis isn't all that far off the mark.

...ODIMS? [TM]



				--*greywolf;
--
# On the 'Net:  Why are more and more fourth-level wizard(-wannabe)s trying
# to invoke ninth-level magic instead of taking the time to climb the other
# (quite essential) thirteen or fourteen levels so they can do this properly?
# Don't they know their console will crumble into dust if they do that?
# greywolf@captech.com