Re: [RFC] mm/pgtable/debug: Add test validating architecture page table helpers

From: Matthew Wilcox
Date: Thu Jul 25 2019 - 18:57:03 EST


On Thu, Jul 25, 2019 at 10:58:12PM +0100, Russell King - ARM Linux admin wrote:
> On Thu, Jul 25, 2019 at 02:42:22PM -0700, Matthew Wilcox wrote:
> > On Thu, Jul 25, 2019 at 10:38:58PM +0100, Russell King - ARM Linux admin wrote:
> > > On Thu, Jul 25, 2019 at 07:39:21AM -0700, Matthew Wilcox wrote:
> > > > But 'page' isn't necessarily PMD-aligned. I don't think we can rely on
> > > > architectures doing the right thing if asked to make a PMD for a randomly
> > > > aligned page.
> > > >
> > > > How about finding the physical address of something like kernel_init(),
> > > > and using the corresponding pte/pmd/pud/p4d/pgd that encompasses that
> > > > address? It's also better to pass in the pfn/page rather than using global
> > > > variables to communicate to the test functions.
> > >
> > > There are architectures (32-bit ARM) where the kernel is mapped using
> > > section mappings, and we don't expect the Linux page table walking to
> > > work for section mappings.
> >
> > This test doesn't go so far as to insert the PTE/PMD/PUD/... into the
> > page tables. It merely needs an appropriately aligned PFN to create a
> > PTE/PMD/PUD/... from.
>
> Well, in any case,
>
> c085ac68 t kernel_init
>
> so I'm not sure that would be an improvement.

I said "the corresponding pte/pmd/pud/p4d/pgd that encompasses that address"

So for a PTE, you'd use PFN 0xc085a000, for a PMD, you'd use PFN 0xc0000000
and for a PGD, you'd use PFN 0 (assuming 9 bits per level of table).