Perhaps we need a defrag memory function for use with GPF_DMA, that
pages out adjacent pages, and then allocates them, glueing them together
to produce a bigger area. Perhaps this should be done over time (at a
low priority) so that bigger pages are continually produced. I suspect
that getting a large (96k) DMAable piece of memory on a moderately loaded
box after a few weeks would be next to nil, that's a bad thing IMHO.
> Marek
>
Bryn
-- PGP key pass phrase forgotten, \ Overload -- core meltdown sequence again :( | initiated. / This space is intentionally left | blank, apart from this text ;-) \____________________________________