Re: swsusp in 2.6.16: works fine w/o PSE...

["Rafael J. Wysocki" <rjw@sisk.pl>]

Hi,

On Thursday 25 May 2006 23:43, Pavel Machek wrote:
> > I'm CC-ing the two swsusp gurus ;)
> > 
> > > I was just feeling lucky and tried suspend-to-disk cycle
> > > on my VIA C3 machine, which lacks PSE which is marked as
> > > being required for swsusp to work.  After commenting out
> > > the PSE check in include/asm-i386/suspend.h and rebooting,
> > > I tried the whole cycle, several times, with real load
> > > (while running 3 kernel compile in parallel) and while
> > > IDLE... And surprizingly, it all worked flawlessly for
> > > me, without a single glitch...
> > > 
> > > So the question is: is PSE really needed nowadays?
> 
> I think so. Or can you prove that pagetables are not going to be
> overwritten in wrong order in !PSE case?
> 
> Look at x86-64 how !PSE case can be solved, but it is a bit of code.

Well, on i386 it'll have to be more complicated, because on x86_64 we use
2 MB pages for the temporary 1-1 mapping.

Greetings,
Rafael