Re: [PATCH RFC 05/11] mm: convert PTE table entries in ptep_get()

On Thu, Jul 30, 2026 at 11:11:35AM +0100, Muhammad Usama Anjum wrote:
> On 30/07/2026 11:08 am, Alexander Gordeev wrote:
> > On Mon, Jul 27, 2026 at 05:46:56PM +0100, Muhammad Usama Anjum wrote:
> >> ptep_get() now accepts a pointer to hw_pte_t storage but must continue to
> >> return a logical pte_t value. Add __pte_from_hw() as the conversion hook
> >> and apply it in the generic ptep_get() implementation.
> >>
> >> Signed-off-by: Muhammad Usama Anjum <[email protected]>
> >> ---
> >>  include/linux/pgtable.h       | 2 +-
> >>  include/linux/pgtable_types.h | 1 +
> >>  2 files changed, 2 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/include/linux/pgtable.h b/include/linux/pgtable.h
> >> index dad80d264aac2..3003ac9cd81fe 100644
> >> --- a/include/linux/pgtable.h
> >> +++ b/include/linux/pgtable.h
> >> @@ -493,7 +493,7 @@ static inline int pudp_set_access_flags(struct vm_area_struct *vma,
> >>  #ifndef ptep_get
> >>  static inline pte_t ptep_get(hw_pte_t *ptep)
> >>  {
> >> -	return READ_ONCE(*ptep);
> >> +	return READ_ONCE(__pte_from_hw(*ptep));
> > 
> > Should it be something like?
> > 
> > 	return __pte_from_hw(READ_ONCE(*ptep));
> I'd thought about it. My thinking is that we should get the pte_t first before
> before calling READ_ONCE(). Please correct me what makes more sense?

I have a problem with "get the pte_t first" part ;) I think what we want
is atomically read the hw_pte_t from the HW table and then convert the
result to pte_t.

> > 
> >>  }
> >>  #endif
> >>  
> >> diff --git a/include/linux/pgtable_types.h b/include/linux/pgtable_types.h
> >> index ed5dba42a4f8c..61afe27307ec2 100644
> >> --- a/include/linux/pgtable_types.h
> >> +++ b/include/linux/pgtable_types.h
> >> @@ -6,6 +6,7 @@
> >>  
> >>  #ifndef __HAVE_ARCH_HW_PTE_T
> >>  #define hw_pte_t pte_t
> >> +#define __pte_from_hw(pte)	(pte)
> >>  #endif
> >>  
> >>  #endif /* _LINUX_PGTABLE_TYPES_H */
> >> -- 
> >> 2.47.3
> >>
> 
> -- 
> Thanks,
> Usama
>