Re: [Xen-changelog] This adds a patch for the vanilla kernel, to be pushed upstream some

Re: [Xen-changelog] This adds a patch for the vanilla kernel, to be pushed upstream some

[Rik van Riel]

On Mon, 11 Jul 2005, Xen patchbot -unstable wrote:

> --- a/linux-2.6-xen-sparse/arch/xen/i386/mm/pgtable.c	Mon Jul 11 15:43:12 2005
> +++ b/linux-2.6-xen-sparse/arch/xen/i386/mm/pgtable.c	Mon Jul 11 15:46:46 2005

>  	if (PTRS_PER_PMD == 1 || !pgd)
>  		return pgd;
>  
> +	if (!HAVE_SHARED_KERNEL_PMD) {
> +		/* alloc and copy kernel pmd */
> +		unsigned long flags;
> +		pgd_t *copy_pgd = pgd_offset_k(PAGE_OFFSET);
> +		pud_t *copy_pud = pud_offset(copy_pgd, PAGE_OFFSET);

I don't understand how the code inside this if condition
ever gets to run.  Surely the previous test should catch
the non-PAE case already ?

Or is this code a preparation for non-3:1 splits ?

-- 
The Theory of Escalating Commitment: "The cost of continuing mistakes is
borne by others, while the cost of admitting mistakes is borne by yourself."
  -- Joseph Stiglitz, Nobel Laureate in Economics

Re: Re: [Xen-changelog] This adds a patch for the vanilla kernel, to be pushed upstream some

[Rik van Riel]

On Tue, 12 Jul 2005, Keir Fraser wrote:
> On 12 Jul 2005, at 16:47, Rik van Riel wrote:
> 
> > I don't understand how the code inside this if condition
> > ever gets to run.  Surely the previous test should catch
> > the non-PAE case already ?
> > 
> > Or is this code a preparation for non-3:1 splits ?
> 
> That code path would get executed by native pae builds (ie. non-xen pae 
> builds).

Makes sense.  There are non-3:1 splits in native PAE builds.

-- 
The Theory of Escalating Commitment: "The cost of continuing mistakes is
borne by others, while the cost of admitting mistakes is borne by yourself."
  -- Joseph Stiglitz, Nobel Laureate in Economics

Re: Re: [Xen-changelog] This adds a patch for the vanilla kernel, to be pushed upstream some

[Keir Fraser]

On 12 Jul 2005, at 16:47, Rik van Riel wrote:

> I don't understand how the code inside this if condition
> ever gets to run.  Surely the previous test should catch
> the non-PAE case already ?
>
> Or is this code a preparation for non-3:1 splits ?

That code path would get executed by native pae builds (ie. non-xen pae 
builds).

It's a dead path on xen/i386 but we keep it around to keep the diff vs 
i386 small.

  -- Keir

Re: Re: [Xen-changelog] This adds a patch for the vanilla kernel, to be pushed upstream some

[Gerd Knorr]

Rik van Riel <riel@redhat.com> writes:

> On Mon, 11 Jul 2005, Xen patchbot -unstable wrote:
> 
> > --- a/linux-2.6-xen-sparse/arch/xen/i386/mm/pgtable.c	Mon Jul 11 15:43:12 2005
> > +++ b/linux-2.6-xen-sparse/arch/xen/i386/mm/pgtable.c	Mon Jul 11 15:46:46 2005
> 
> >  	if (PTRS_PER_PMD == 1 || !pgd)
> >  		return pgd;
> >  
> > +	if (!HAVE_SHARED_KERNEL_PMD) {
> > +		/* alloc and copy kernel pmd */
> > +		unsigned long flags;
> > +		pgd_t *copy_pgd = pgd_offset_k(PAGE_OFFSET);
> > +		pud_t *copy_pud = pud_offset(copy_pgd, PAGE_OFFSET);
> 
> I don't understand how the code inside this if condition
> ever gets to run.  Surely the previous test should catch
> the non-PAE case already ?

It's about PAE mode.  xenlinux in PAE mode does _not_ have a shared
PMD for the kernel address space.

  Gerd

-- 
panic("it works"); /* avoid being flooded with debug messages */