Re: [RFC v1 2/2] powerpc/64s: Add support for huge pfnmaps

Ritesh Harjani (IBM) [off-list ref] writes:

"Christophe Leroy (CS GROUP)" [off-list ref] writes:

quoted

Le 27/02/2026 à 07:16, Ritesh Harjani (IBM) a écrit :

quoted

This uses _RPAGE_SW2 bit for the PMD and PUDs similar to PTEs.
This also adds support for {pte,pmd,pud}_pgprot helpers needed for
follow_pfnmap APIs.

This allows us to extend the PFN mappings, e.g. PCI MMIO bars where
it can grow as large as 8GB or even bigger, to map at PMD / PUD level.
VFIO PCI core driver already supports fault handling at PMD / PUD level
for more efficient BAR mappings.

Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>

Reviewed-by: Christophe Leroy (CS GROUP) <chleroy@kernel.org>

Thanks for the review!

quoted

quoted

  #define __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
  extern int pmdp_set_access_flags(struct vm_area_struct *vma,

diff --git a/arch/powerpc/include/asm/pgtable.h b/arch/powerpc/include/asm/pgtable.h
index dcd3a88caaf6..2d27cb1c2334 100644
--- a/arch/powerpc/include/asm/pgtable.h
+++ b/arch/powerpc/include/asm/pgtable.h

@@ -63,6 +63,18 @@ static inline pgprot_t pte_pgprot(pte_t pte)
  	return __pgprot(pte_flags);
  }

+#define pmd_pgprot pmd_pgprot
+static inline pgprot_t pmd_pgprot(pmd_t pmd)
+{
+	return pte_pgprot(pmd_pte(pmd));
+}
+
+#define pud_pgprot pud_pgprot
+static inline pgprot_t pud_pgprot(pud_t pud)
+{
+	return pte_pgprot(pud_pte(pud));
+}
+

In v2 - I will add above under #ifdef CONFIG_PPC_BOOK3S_64 
to avoid build issues with 32-bit PPC.

On second thoughts, I am thinking maybe we should guard it with CONFIG_PPC64.  
Currently the build fails on 32-bit since no definitions of pmd_pte()
and pud_pte().  Though, we could open-code that, but I think as of
today, this only gets excercised from follow_pfnmap_start() which gates
it with VM_PFNMAP | VM_IO, which I think could only happen for THP which
is only true for book3s/64. 
But to keep the generic definitions of pXd_pgprot() and since pmd_pte()
and pud_pte() are anyways available on book3s/64 & nohash/64, so let's
just guard this with PPC64.

I will amend this change in RFC-v2 and will keep the RB from Christophe.


+#ifdef CONFIG_PPC64
+#define pmd_pgprot pmd_pgprot
+static inline pgprot_t pmd_pgprot(pmd_t pmd)
+{
+       return pte_pgprot(pmd_pte(pmd));
+}
+
+#define pud_pgprot pud_pgprot
+static inline pgprot_t pud_pgprot(pud_t pud)
+{
+       return pte_pgprot(pud_pte(pud));
+}
+#endif /* CONFIG_PPC64 */
+


-ritesh