Answering a question about NetBSD using low memory for the buffer cache
vs. a merged VM/Buffer cache scheme that is very dynamic:

It is possible to "prefer" lower memory for the VM/buffer cache, like NetBSD
hard-allocates for its buffers.  That would not be an issue of the bounce
code, but of the memory allocation code for the buffer cache, and the
appropriate hooks in the VM system, and the drivers (for bounce requirement
info.)

There are still the issues of buffers that could not be allocated from the
DMAable memory -- is it worth it to move pages around just to be able to
do I/O to/from them? (probably not, but it might be fun to write the page
migration code).  Or what about physical I/O that is done directly to/from the
user's address space?  Those pages would not be DMA preferred pages either.

It would be likely on a 20MB machine or even a 32MB machine that you would
not be able to measure a striking perf difference between a fully optimal
scheme and a simple, straightforward bouncing scheme.  I do suggest a
SCSI device independent scheme, even though I know that many are interested
primarily in the aha1542 type adapters, some have bt545s's also.

John Dyson
dyson@freebsd.org