fc5 kernels after 1788 have excessive, kernel overhead

[Dave Jones]

On Mon, Jan 02, 2006 at 11:29:26PM +0000, Andy Burns wrote:
 > Dave Jones wrote:
 > 
 > >Does acpi=off on the smp kernel make any difference?
 > 
 > acpi=off with the SMP kernel doesn't break it
 > acpi=off with the UP kernel doesn't mend it
 > 
 > >Any chance you can hook up another pc, and get a serial capture
 > >of the failing boot ?
 > 
 > Below is the UP log, followed by SMP log for good measure
 > 
 > Incidentally, is it correct that the UP kernel only approximates NX, yet 
 > the SMP kernel makes proper use of NX?

If the CPU has hardware NX support , then the SMP kernel does take
advantage of it, whilst the UP one has to fall back on segment-limit
approach to 'emulate' it.  The reason for this is that the SMP
kernel has support for physical address extensions (PAE) (Ie, >4GB RAM)
and NX is dependant on that.  The UP kernel doesn't have PAE
because a kernel compiled with that supported doesn't boot on
systems that lack it, (Celerons, early VIA, some mobile Pentium)

 > Would disabling HT in BIOS affect it?

You tell me :)  I can't think of any reason why this has suddenly
broken that would be related to that though.

		Dave