> Hans Lermen wrote:
>
> > - one writeprotected for LIDT
> > - one writeable for the kernel.
>
> But AFAIK, only arch/i386/kernel/irq.c writes to the idt entries
> (in 2.0.x that is), and it only updates entries >= 0x20.
right.
> So instead of write-protecting the entire ldt page and setting
> up an alias page, we should be able to keep the 2.0.32 scheme,
> but *write protect* instead of *unmap* the lower page.
> I think this is a simpler solution than the alias-page trick.
In principal you are right, but the 'alias' version would require
_one_ page of mem only while your solution needs _two_ pages.
( admitted, I forgot to do __freepage() for the lower page in my patch )
Hans
<lermen@fgan.de>