Subject: Re: R140 w/8 MB and multiuser
To: Ben Harris <email@example.com>
From: Ian Fry <Ian.Fry@sophos.com>
Date: 07/02/2001 12:56:10
On Mon, Jul 02, 2001 at 12:49:26PM +0100, Ben Harris wrote:
> On Mon, 2 Jul 2001, Ian Fry wrote:
> > On Mon, Jul 02, 2001 at 12:43:48PM +0100, Ben Harris wrote:
> > > On Mon, 2 Jul 2001, Ian Fry wrote:
> > > > > For the most part, buffering in the kernel should be in page-sized
> > > > > chunks these days, which is less than wonderful for us.
> > > > Indeed :-( Is there anyway we could get the kernel memory allocator
> > > > to sub-divide pages into 'virtual' ones - say 4 or 8k, rather than
> > > > 32?
> > > Not unless we can persuade the MEMC to do the same.
> > Drat! MEMC uses a fixed 32k page, doesn't it? So, that'd be a no go.
> No, but each MEMC can only support 128 pages, so if you reduce the page
> size, you reduce the amount of RAM it can address.
Ah, yes! That's it - I knew there was a limitation in MEMC, but couldn't
remember if it was the page size or number of pages.
> I'm also not sure if you can hook multiple MEMCs together with <32k pages.
I don't know either - presumably, they would all have to use the same page
size, even if you could.