> On Mon, 3 Nov 1997, Neilski wrote:
> > If lots of pages are free, but none of them happen to be in the DMA-able
> > region, then __get_free_pages() will fail to return DMA-able pages - as far as
> > I can see... This happened to me today, I reckon.
> >
> > Is this a bug or a feature ?
>
> Definitely a bug, and a long standing one at that.
>
> I'm currently re-working mm/page_alloc.c to give page-colouring support,
> along with lazy-coalescing, a simple fragmentation control, and the
> ability to quickly re-alloc "cache-hot" pages. I'm also looking at the
> DMA problem. (The colouring, lazy-coalescing, and "hot" handling are
> working well in my limited test environment. The fragmentation control
> needs some more work, and I'm only maintaining DMA counts at the moment -
> but its getting there).
>
Hi Mark!
Very interesting, indeed.
Could you, please, send patches on the list (or mail them, if you
prefer).
Even very alpha one's would be great, just to experiment with them.
I'm very interested in all new features you mentioned, and am
investigating things on my own. My primary interest is in that
"fragmentation issues", but lazy coalescing in page allocator is also
on my list to play with, it's just I didn't have enough time.
Regards,
-- Posted by Zlatko Calusic E-mail: <Zlatko.Calusic@CARNet.hr> --------------------------------------------------------------------- Is this a machine? I don't talk to machines! [Click]