On Mon, 2008-09-22 at 07:03 -0300, Edesio Costa e Silva wrote:
> On Mon, Sep 22, 2008 at 10:53:38AM +0200, Bernd Petrovitsch wrote:
> > On Mon, 2008-09-22 at 07:54 +0900, MinChan Kim wrote:
> > > * Rik van Riel (riel@xxxxxxxxxxx) wrote:
> > [....]
> > > > The second option is to rewrite part of the VM so less
> > > > metadata is kept for page cache. Specifically, you will
> > > > have to write code to reclaim buffer heads from page
> > > > cache pages.
> > >
> > > Do You mean current page reclaim mechanism have a problem ?
> >
> > Only on insane installations.
> >
> > > As you know, kernel already have reclaimed fs-specific metadata.
> > > What is a problem?
> >
> > That people run a 32bit x86 kernel with 64GB RAM on x86_64 hardware
> > (obviously - 32bit cannot address 64GB RAM anyways) with no good reason.
>
> On Intel Pentium III and later it can, using PAE and/or PSE 36 bit
> physical addresses.
Hmm, (even) if 64GB RAM is physically possible[0], the question is: What
must the hardware/MMU/... and their drivers do to address the range
beyond 4GB. And how much performance do you loose through that.
Especially as the simple solution is: run a 64bit kernel (with a pure
32bit userspace if you wish).
Bernd
[0]: I know how - ahemm - "elegant" the access of > 1MB in DOS-6.22
times were - A20 line (IIRC), EMS, XMS.
--
Firmix Software GmbH http://www.firmix.at/
mobil: +43 664 4416156 fax: +43 1 7890849-55
Embedded Linux Development and Services
--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx
Please read the FAQ at http://kernelnewbies.org/FAQ
