Hi, short update: I uploaded lbacache-15apr2004.zip to
http://www.coli.uni-sb.de/~eric/stuff/soft/
... instead of doing a panic shutdown it now checks if an
access wraps around the end of the floppy track. If this
is the case, the access bypasses the cache and the cache
contents for that drive are flushed to ensure consistency.
This version also displays a warning message whenever such
problems are detected.

First tests by Tyler give the idea that the cache now
handles 1680k floppy disks properly (Question: Even if
you load it while a 1680k disk is in use? Int 13.8
geometry at cache init time must not be > 1.44 MB or
the drive will not be cached...). However, it is
possible that you find a way to confuse the cache enough to cause
data errors when copying data to/from floppy when
you use mixed sizes of disks or even format disks
to other sizes without changing the disk (changing
as in "the drive door sensor notices something").

Happy testing...

At least the move from 4k to 8k element size did not
seem to cause problems (even better, it means less RAM
used and faster processing of cache miss events), but
I did not get reports about whether and how much the
cache hit percentage changed between 22oct2003 and 15apr2004
versions of LBAcache for a real life test of your
choice (e.g. compiling the FreeDOS kernel).

If the bigger element size did not hurt cache hit
percentage even for relatively small caches, fine.
Otherwise, I can add a command line option to select
the element size when loading LBAcache.

Eric