Re: preempt bug in set_pmd_pfn?

[Andi Kleen <ak@suse.de>]

On Wednesday 05 March 2008 01:19:36 Jeremy Fitzhardinge wrote:
> Andi Kleen wrote:
> > On Wednesday 05 March 2008 01:07:11 Jeremy Fitzhardinge wrote:
> >   
> >> Andi Kleen wrote:
> >>     
> >>>> Won't this leave a stale tlb on the old processor?
> >>>>     
> >>>>         
> >>> __set_fixmap should be only used in early boot, so it's always
> >>> on CPU 0
> >>>       
> >> vdso32-setup.c:map_compat_vdso() uses it to create the compat vdso 
> >> mapping, which typically happens on the first execve(),
> >>     
> >
> > First execve for 32bit binaries?
> >   
> 
> Yes, 32-bit kernel.

Ok I'm talking about 64bit. If you ask me something about 32bit please
always mention it especially.

> > Anyways __set_fixmap is __init and at the first  execve (unless you have initramfs) 
> > init.text should be already freed.
> 
> No, its not __init.  If it were I don't think there would have been much 
> booting going on for the last few months.

For 64bit my original statement was correct.

-Andi