Re: [PATCH] docs/mm: expand vma doc to highlight pte freeing, non-vma traversal

[Jann Horn <jannh@google.com>]

On Tue, Jun 3, 2025 at 12:45 PM Lorenzo Stoakes
<lorenzo.stoakes@oracle.com> wrote:
> On Tue, Jun 03, 2025 at 12:25:36AM +0200, Jann Horn wrote:
> > The one way in which I think this is currently kinda yolo/broken is
> > that vmap_try_huge_pud() can end up freeing page tables via
> > pud_free_pmd_page(), while holding no MM locks AFAICS, so that could
> > race with the ptdump debug logic such that ptdump walks into freed
> > page tables?
>
> But those mappings would be kernel mappings? How could ptdump walk into
> those?

/sys/kernel/debug/page_tables/kernel dumps kernel page tables. And I
think /sys/kernel/debug/page_tables/current_kernel dumps page tables
for the entire address space including both userspace and kernel.

> > (I think we should take all the vma locks in that ptdump code and get
> > rid of this weird exception instead of documenting it.)
>
> We really need to be sure that there aren't some weird architectures doing
> weird things or circumstances where this is meaningful.
>
> I mean people went to great lengths to make this possible, I find it hard
> to believe there aren't some _weird_ real world use cases.

FWIW, looking through the git logs for the x86 version of it, it seems
to mainly be used by developers of arch-specific code trying to
debug/validate kernel behavior.