On Tue, 15 Sep 1998, Helge Hafting wrote:
> In <Pine.GSO.3.95q.980914165148.14025D-100000@jupiter>, on 09/14/98
> at 04:56 PM, " Raymond A. Ingles" <firstname.lastname@example.org> said:
> > Unfortunately, the PC ISA DMA controller can't address memory beyond the
> >first 16MB. So, when a DMA buffer is requested, the kernel has to find a
> >continuous chunk of memory that is physically below 16MB. If it can't
> >find it, too bad.
> The kernel should then swap something below 16M out, or preferably push it
> above 16M. Anybody know why it doesn't do that? I find it unlikely that
> *all* memory below 16M should be locked for i/o or something.
I don't know all the details about why the kernel can't do that, but I'm
given to understand that performance would suffer, and some programs would
break. There was a discussion on linux-kernel a while ago, maybe it's
Ray Ingles (248)377-7735 email@example.com
Modern inductive method: 1) Devise hypothesis. 2) Apply for grant.
3) Perform experiments. 4) Revise data to fit hypothesis. 5) Publish.