Re: [PATCH v2 2/3] x86/mm/pat: fix VM_PAT handling in COW mappings

[David Hildenbrand]

On 04.04.24 00:12, Andrew Morton wrote:
> On Wed,  3 Apr 2024 23:21:30 +0200 David Hildenbrand <david@xxxxxxxxxx> wrote:
>
> > PAT handling won't do the right thing in COW mappings: the first PTE
> > (or, in fact, all PTEs) can be replaced during write faults to point at
> > anon folios. Reliably recovering the correct PFN and cachemode using
> > follow_phys() from PTEs will not work in COW mappings.
> >
> > ...
> >
> > Reported-by: Wupeng Ma <mawupeng1@xxxxxxxxxx>
> > Closes: https://lkml.kernel.org/r/20240227122814.3781907-1-mawupeng1@xxxxxxxxxx
> > Fixes: b1a86e15dc03 ("x86, pat: remove the dependency on 'vm_pgoff' in track/untrack pfn vma routines")
> > Fixes: 5899329b1910 ("x86: PAT: implement track/untrack of pfnmap regions for x86 - v3")
>
> These are really old.  Should we backport this?

I was asking that question myself.

With the reproducer, the worst thing that happens on most systems is the 
warning. On !RAM and with PAT, there could be memory leaks and other 
surprises.

Likely, we should just backport it to stable. Should not be too hard to 
backport to stable kernels I guess/hope.

--
Cheers,

David / dhildenb