[PATCH 0/9] Convert x86 pgd_lists to ptdescs

[PATCH 0/9] Convert x86 pgd_lists to ptdescs

[Vishal Moola]

This continues the ptdesc conversions for x86[1]. This is a necessary step
in preparation for the separate allocation of ptdescs from struct page.
Once ptdescs are not wrappers, the pgd_list should be used to find our
ptdesc, not page.

In fact, after patch 4 we can successfully boot a 32-bit x86 kernel
that separately allocates ptdescs.

All of the patches are rather straightforward, I just kept them split up
for ease of review. There are no intended functional changes as ptdescs
are currently still wrappers around struct page.

[1] https://lore.kernel.org/linux-mm/20260211195233.368497-1-vishal.moola@gmail.com/

----------------

I've compiled and booted both 64 and 32 bit on a kernel with separately
allocated ptdescs. I ensured the codepaths were hit for all the changes
except Xen.

This applies cleanly to the tip bot. I'm thinking this can go through
Dave's mm.

Vishal Moola (9):
  x86/mm/pat: Use IS_ENABLED() instead of ifdef
  x86/mm/pat: Convert __set_pmd_pte() to ptdescs
  x86/mm/pat: Convert collapse_pmd_page() to ptdescs
  x86/mm: Convert arch_sync_kernel_mappings() to ptdescs
  x86/mm: Convert sync_global_pgds_l5() to ptdescs
  x86/mm: Convert sync_global_pgds_l4() to ptdescs
  x86/mm: Convert pgd_page_get_mm() to ptdescs
  x86/xen: Convert xen_mm_pin_all() to ptdescs
  x86/xen: Convert xen_mm_unpin_all() to ptdescs

 arch/x86/include/asm/pgtable.h       |  2 +-
 arch/x86/include/asm/pgtable_types.h |  2 +-
 arch/x86/mm/fault.c                  |  8 ++++----
 arch/x86/mm/init_64.c                | 16 ++++++++--------
 arch/x86/mm/pat/set_memory.c         | 17 ++++++++---------
 arch/x86/mm/pgtable.c                |  4 ++--
 arch/x86/xen/mmu_pv.c                | 22 +++++++++++-----------
 7 files changed, 35 insertions(+), 36 deletions(-)

-- 
2.54.0

[PATCH 1/9] x86/mm/pat: Use IS_ENABLED() instead of ifdef

[Vishal Moola]

Use IS_ENABLED() to check if we are on 32 bit. This standardizes this
check with the other 32 bit check in the file. No functional changes.

Signed-off-by: Vishal Moola <vishal.moola@gmail.com>
---
 arch/x86/mm/pat/set_memory.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/arch/x86/mm/pat/set_memory.c b/arch/x86/mm/pat/set_memory.c
index d023a40a1e03..3b9032a3f315 100644
--- a/arch/x86/mm/pat/set_memory.c
+++ b/arch/x86/mm/pat/set_memory.c
@@ -888,8 +888,8 @@ static void __set_pmd_pte(pte_t *kpte, unsigned long address, pte_t pte)
 {
 	/* change init_mm */
 	set_pte_atomic(kpte, pte);
-#ifdef CONFIG_X86_32
-	{
+
+	if (IS_ENABLED(CONFIG_X86_32)) {
 		struct page *page;
 
 		list_for_each_entry(page, &pgd_list, lru) {
@@ -905,7 +905,6 @@ static void __set_pmd_pte(pte_t *kpte, unsigned long address, pte_t pte)
 			set_pte_atomic((pte_t *)pmd, pte);
 		}
 	}
-#endif
 }
 
 static pgprot_t pgprot_clear_protnone_bits(pgprot_t prot)
-- 
2.54.0

[PATCH 2/9] x86/mm/pat: Convert __set_pmd_pte() to ptdescs

[Vishal Moola]

Convert __set_pmd_pte() to ptdescs in preparation for the eventual
splitting of ptdescs from struct page.

Signed-off-by: Vishal Moola <vishal.moola@gmail.com>
---
 arch/x86/mm/pat/set_memory.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/arch/x86/mm/pat/set_memory.c b/arch/x86/mm/pat/set_memory.c
index 3b9032a3f315..93cb71451ce2 100644
--- a/arch/x86/mm/pat/set_memory.c
+++ b/arch/x86/mm/pat/set_memory.c
@@ -890,15 +890,15 @@ static void __set_pmd_pte(pte_t *kpte, unsigned long address, pte_t pte)
 	set_pte_atomic(kpte, pte);
 
 	if (IS_ENABLED(CONFIG_X86_32)) {
-		struct page *page;
+		struct ptdesc *ptdesc;
 
-		list_for_each_entry(page, &pgd_list, lru) {
+		list_for_each_entry(ptdesc, &pgd_list, pt_list) {
 			pgd_t *pgd;
 			p4d_t *p4d;
 			pud_t *pud;
 			pmd_t *pmd;
 
-			pgd = (pgd_t *)page_address(page) + pgd_index(address);
+			pgd = (pgd_t *)ptdesc_address(ptdesc) + pgd_index(address);
 			p4d = p4d_offset(pgd, address);
 			pud = pud_offset(p4d, address);
 			pmd = pmd_offset(pud, address);
-- 
2.54.0

[PATCH 3/9] x86/mm/pat: Convert collapse_pmd_page() to ptdescs

[Vishal Moola]

Convert collapse_pmd_page() to ptdescs in preparation for the
eventual splitting of ptdescs from struct page.

Signed-off-by: Vishal Moola <vishal.moola@gmail.com>
---
 arch/x86/mm/pat/set_memory.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/arch/x86/mm/pat/set_memory.c b/arch/x86/mm/pat/set_memory.c
index 93cb71451ce2..45623d4c24c9 100644
--- a/arch/x86/mm/pat/set_memory.c
+++ b/arch/x86/mm/pat/set_memory.c
@@ -1297,11 +1297,11 @@ static int collapse_pmd_page(pmd_t *pmd, unsigned long addr,
 	list_add(&page_ptdesc(pmd_page(old_pmd))->pt_list, pgtables);
 
 	if (IS_ENABLED(CONFIG_X86_32)) {
-		struct page *page;
+		struct ptdesc *ptdesc;
 
 		/* Update all PGD tables to use the same large page */
-		list_for_each_entry(page, &pgd_list, lru) {
-			pgd_t *pgd = (pgd_t *)page_address(page) + pgd_index(addr);
+		list_for_each_entry(ptdesc, &pgd_list, pt_list) {
+			pgd_t *pgd = (pgd_t *)ptdesc_address(ptdesc) + pgd_index(addr);
 			p4d_t *p4d = p4d_offset(pgd, addr);
 			pud_t *pud = pud_offset(p4d, addr);
 			pmd_t *pmd = pmd_offset(pud, addr);
-- 
2.54.0

[PATCH 4/9] x86/mm: Convert arch_sync_kernel_mappings() to ptdescs

[Vishal Moola]

Convert arch_sync_kernel_mappings() to ptdescs in preparation for
the eventual splitting of ptdescs from struct page.

Following this patch, we can successfully boot a 32-bit x86 kernel with
separately allocated ptdescs.

Signed-off-by: Vishal Moola <vishal.moola@gmail.com>
---
 arch/x86/mm/fault.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
index 63de8e8684f2..6c72d6668416 100644
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -269,17 +269,17 @@ void arch_sync_kernel_mappings(unsigned long start, unsigned long end)
 	for (addr = start & PMD_MASK;
 	     addr >= TASK_SIZE_MAX && addr < VMALLOC_END;
 	     addr += PMD_SIZE) {
-		struct page *page;
+		struct ptdesc *ptdesc;
 
 		spin_lock(&pgd_lock);
-		list_for_each_entry(page, &pgd_list, lru) {
+		list_for_each_entry(ptdesc, &pgd_list, pt_list) {
 			spinlock_t *pgt_lock;
 
 			/* the pgt_lock only for Xen */
-			pgt_lock = &pgd_page_get_mm(page)->page_table_lock;
+			pgt_lock = &pgd_page_get_mm(ptdesc_page(ptdesc))->page_table_lock;
 
 			spin_lock(pgt_lock);
-			vmalloc_sync_one(page_address(page), addr);
+			vmalloc_sync_one(ptdesc_address(ptdesc), addr);
 			spin_unlock(pgt_lock);
 		}
 		spin_unlock(&pgd_lock);
-- 
2.54.0

[PATCH 5/9] x86/mm: Convert sync_global_pgds_l5() to ptdescs

[Vishal Moola]

Convert sync_global_pgds_l5() to ptdescs in preparation for the
eventual splitting of ptdescs from struct page.

Signed-off-by: Vishal Moola <vishal.moola@gmail.com>
---
 arch/x86/mm/init_64.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
index df27379d4bb4..e89b51aea2cf 100644
--- a/arch/x86/mm/init_64.c
+++ b/arch/x86/mm/init_64.c
@@ -136,7 +136,7 @@ static void sync_global_pgds_l5(unsigned long start, unsigned long end)
 
 	for (addr = start; addr <= end; addr = ALIGN(addr + 1, PGDIR_SIZE)) {
 		const pgd_t *pgd_ref = pgd_offset_k(addr);
-		struct page *page;
+		struct ptdesc *ptdesc;
 
 		/* Check for overflow */
 		if (addr < start)
@@ -146,13 +146,13 @@ static void sync_global_pgds_l5(unsigned long start, unsigned long end)
 			continue;
 
 		spin_lock(&pgd_lock);
-		list_for_each_entry(page, &pgd_list, lru) {
+		list_for_each_entry(ptdesc, &pgd_list, pt_list) {
 			pgd_t *pgd;
 			spinlock_t *pgt_lock;
 
-			pgd = (pgd_t *)page_address(page) + pgd_index(addr);
+			pgd = (pgd_t *)ptdesc_address(ptdesc) + pgd_index(addr);
 			/* the pgt_lock only for Xen */
-			pgt_lock = &pgd_page_get_mm(page)->page_table_lock;
+			pgt_lock = &pgd_page_get_mm(ptdesc_page(ptdesc))->page_table_lock;
 			spin_lock(pgt_lock);
 
 			if (!pgd_none(*pgd_ref) && !pgd_none(*pgd))
-- 
2.54.0

[PATCH 6/9] x86/mm: Convert sync_global_pgds_l4() to ptdescs

[Vishal Moola]

Convert sync_global_pgds_l4() to ptdescs in preparation for the
eventual splitting of ptdescs from struct page.

Signed-off-by: Vishal Moola <vishal.moola@gmail.com>
---
 arch/x86/mm/init_64.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
index e89b51aea2cf..a6ccb83f2d26 100644
--- a/arch/x86/mm/init_64.c
+++ b/arch/x86/mm/init_64.c
@@ -174,7 +174,7 @@ static void sync_global_pgds_l4(unsigned long start, unsigned long end)
 	for (addr = start; addr <= end; addr = ALIGN(addr + 1, PGDIR_SIZE)) {
 		pgd_t *pgd_ref = pgd_offset_k(addr);
 		const p4d_t *p4d_ref;
-		struct page *page;
+		struct ptdesc *ptdesc;
 
 		/*
 		 * With folded p4d, pgd_none() is always false, we need to
@@ -187,15 +187,15 @@ static void sync_global_pgds_l4(unsigned long start, unsigned long end)
 			continue;
 
 		spin_lock(&pgd_lock);
-		list_for_each_entry(page, &pgd_list, lru) {
+		list_for_each_entry(ptdesc, &pgd_list, pt_list) {
 			pgd_t *pgd;
 			p4d_t *p4d;
 			spinlock_t *pgt_lock;
 
-			pgd = (pgd_t *)page_address(page) + pgd_index(addr);
+			pgd = (pgd_t *)ptdesc_address(ptdesc) + pgd_index(addr);
 			p4d = p4d_offset(pgd, addr);
 			/* the pgt_lock only for Xen */
-			pgt_lock = &pgd_page_get_mm(page)->page_table_lock;
+			pgt_lock = &pgd_page_get_mm(ptdesc_page(ptdesc))->page_table_lock;
 			spin_lock(pgt_lock);
 
 			if (!p4d_none(*p4d_ref) && !p4d_none(*p4d))
-- 
2.54.0

[PATCH 7/9] x86/mm: Convert pgd_page_get_mm() to ptdescs

[Vishal Moola]

Convert pgd_page_get_mm() to ptdescs. Define struct ptdesc in our
pgtable_types so that our declarations recognize ptdesc as an
appropriate page table type.

Now that all callers are using ptdescs, we can pass in that
ptdesc to get the underlying mm_struct.

Signed-off-by: Vishal Moola <vishal.moola@gmail.com>
---
 arch/x86/include/asm/pgtable.h       | 2 +-
 arch/x86/include/asm/pgtable_types.h | 2 +-
 arch/x86/mm/fault.c                  | 2 +-
 arch/x86/mm/init_64.c                | 4 ++--
 arch/x86/mm/pgtable.c                | 4 ++--
 5 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/arch/x86/include/asm/pgtable.h b/arch/x86/include/asm/pgtable.h
index 2187e9cfcefa..20e8a6ddd2df 100644
--- a/arch/x86/include/asm/pgtable.h
+++ b/arch/x86/include/asm/pgtable.h
@@ -50,7 +50,7 @@ void ptdump_walk_user_pgd_level_checkwx(void);
 extern spinlock_t pgd_lock;
 extern struct list_head pgd_list;
 
-extern struct mm_struct *pgd_page_get_mm(struct page *page);
+struct mm_struct *pgd_page_get_mm(struct ptdesc *pt);
 
 extern pmdval_t early_pmd_flags;
 
diff --git a/arch/x86/include/asm/pgtable_types.h b/arch/x86/include/asm/pgtable_types.h
index 2ec250ba467e..95c7f453db17 100644
--- a/arch/x86/include/asm/pgtable_types.h
+++ b/arch/x86/include/asm/pgtable_types.h
@@ -512,7 +512,7 @@ static inline pgprot_t pgprot_large_2_4k(pgprot_t pgprot)
 	return __pgprot(protval_large_2_4k(pgprot_val(pgprot)));
 }
 
-
+struct ptdesc;
 typedef struct page *pgtable_t;
 
 extern pteval_t __supported_pte_mask;
diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
index 6c72d6668416..4d78c78a0bcc 100644
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -276,7 +276,7 @@ void arch_sync_kernel_mappings(unsigned long start, unsigned long end)
 			spinlock_t *pgt_lock;
 
 			/* the pgt_lock only for Xen */
-			pgt_lock = &pgd_page_get_mm(ptdesc_page(ptdesc))->page_table_lock;
+			pgt_lock = &pgd_page_get_mm(ptdesc)->page_table_lock;
 
 			spin_lock(pgt_lock);
 			vmalloc_sync_one(ptdesc_address(ptdesc), addr);
diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
index a6ccb83f2d26..a49be75dd713 100644
--- a/arch/x86/mm/init_64.c
+++ b/arch/x86/mm/init_64.c
@@ -152,7 +152,7 @@ static void sync_global_pgds_l5(unsigned long start, unsigned long end)
 
 			pgd = (pgd_t *)ptdesc_address(ptdesc) + pgd_index(addr);
 			/* the pgt_lock only for Xen */
-			pgt_lock = &pgd_page_get_mm(ptdesc_page(ptdesc))->page_table_lock;
+			pgt_lock = &pgd_page_get_mm(ptdesc)->page_table_lock;
 			spin_lock(pgt_lock);
 
 			if (!pgd_none(*pgd_ref) && !pgd_none(*pgd))
@@ -195,7 +195,7 @@ static void sync_global_pgds_l4(unsigned long start, unsigned long end)
 			pgd = (pgd_t *)ptdesc_address(ptdesc) + pgd_index(addr);
 			p4d = p4d_offset(pgd, addr);
 			/* the pgt_lock only for Xen */
-			pgt_lock = &pgd_page_get_mm(ptdesc_page(ptdesc))->page_table_lock;
+			pgt_lock = &pgd_page_get_mm(ptdesc)->page_table_lock;
 			spin_lock(pgt_lock);
 
 			if (!p4d_none(*p4d_ref) && !p4d_none(*p4d))
diff --git a/arch/x86/mm/pgtable.c b/arch/x86/mm/pgtable.c
index da7f0a03cf90..070e3d310eca 100644
--- a/arch/x86/mm/pgtable.c
+++ b/arch/x86/mm/pgtable.c
@@ -74,9 +74,9 @@ static void pgd_set_mm(pgd_t *pgd, struct mm_struct *mm)
 	virt_to_ptdesc(pgd)->pt_mm = mm;
 }
 
-struct mm_struct *pgd_page_get_mm(struct page *page)
+struct mm_struct *pgd_page_get_mm(struct ptdesc *pt)
 {
-	return page_ptdesc(page)->pt_mm;
+	return pt->pt_mm;
 }
 
 static void pgd_ctor(struct mm_struct *mm, pgd_t *pgd)
-- 
2.54.0

[PATCH 8/9] x86/xen: Convert xen_mm_pin_all() to ptdescs

[Vishal Moola]

Convert xen_mm_pin_all() to ptdescs in preparation for the eventual
splitting of ptdescs from struct page. Continue checking PagePinned
through the underlying page as we do not have a per-memdesc api for
page flags yet.

Signed-off-by: Vishal Moola <vishal.moola@gmail.com>
---
 arch/x86/xen/mmu_pv.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/arch/x86/xen/mmu_pv.c b/arch/x86/xen/mmu_pv.c
index 3eee5f84f8a7..b2f6e7b7c545 100644
--- a/arch/x86/xen/mmu_pv.c
+++ b/arch/x86/xen/mmu_pv.c
@@ -828,15 +828,15 @@ static void xen_pgd_pin(struct mm_struct *mm)
  */
 void xen_mm_pin_all(void)
 {
-	struct page *page;
+	struct ptdesc *ptdesc;
 
 	spin_lock(&init_mm.page_table_lock);
 	spin_lock(&pgd_lock);
 
-	list_for_each_entry(page, &pgd_list, lru) {
-		if (!PagePinned(page)) {
-			__xen_pgd_pin(&init_mm, (pgd_t *)page_address(page));
-			SetPageSavePinned(page);
+	list_for_each_entry(ptdesc, &pgd_list, pt_list) {
+		if (!PagePinned(ptdesc_page(ptdesc))) {
+			__xen_pgd_pin(&init_mm, (pgd_t *)ptdesc_address(ptdesc));
+			SetPageSavePinned(ptdesc_page(ptdesc));
 		}
 	}
 
-- 
2.54.0

[PATCH 9/9] x86/xen: Convert xen_mm_unpin_all() to ptdescs

[Vishal Moola]

Convert xen_mm_unpin_all() to ptdescs in preparation for the eventual
splitting of ptdescs from struct page. Continue checking PagePinned
through the underlying page as we do not have a per-memdesc api yet.

Signed-off-by: Vishal Moola <vishal.moola@gmail.com>
---
 arch/x86/xen/mmu_pv.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/arch/x86/xen/mmu_pv.c b/arch/x86/xen/mmu_pv.c
index b2f6e7b7c545..c4731c1e842d 100644
--- a/arch/x86/xen/mmu_pv.c
+++ b/arch/x86/xen/mmu_pv.c
@@ -939,16 +939,16 @@ static void xen_pgd_unpin(struct mm_struct *mm)
  */
 void xen_mm_unpin_all(void)
 {
-	struct page *page;
+	struct ptdesc *ptdesc;
 
 	spin_lock(&init_mm.page_table_lock);
 	spin_lock(&pgd_lock);
 
-	list_for_each_entry(page, &pgd_list, lru) {
-		if (PageSavePinned(page)) {
-			BUG_ON(!PagePinned(page));
-			__xen_pgd_unpin(&init_mm, (pgd_t *)page_address(page));
-			ClearPageSavePinned(page);
+	list_for_each_entry(ptdesc, &pgd_list, pt_list) {
+		if (PageSavePinned(ptdesc_page(ptdesc))) {
+			BUG_ON(!PagePinned(ptdesc_page(ptdesc)));
+			__xen_pgd_unpin(&init_mm, (pgd_t *)ptdesc_address(ptdesc));
+			ClearPageSavePinned(ptdesc_page(ptdesc));
 		}
 	}
 
-- 
2.54.0

Re: [PATCH 0/9] Convert x86 pgd_lists to ptdescs

[Mike Rapoport]

On Wed, May 27, 2026 at 03:03:07PM +0100, Vishal Moola wrote:
> This continues the ptdesc conversions for x86[1]. This is a necessary step
> in preparation for the separate allocation of ptdescs from struct page.
> Once ptdescs are not wrappers, the pgd_list should be used to find our
> ptdesc, not page.
> 
> In fact, after patch 4 we can successfully boot a 32-bit x86 kernel
> that separately allocates ptdescs.
> 
> All of the patches are rather straightforward, I just kept them split up
> for ease of review. There are no intended functional changes as ptdescs
> are currently still wrappers around struct page.

Looks straightforward to me indeed :)

Acked-by: Mike Rapoport (Microsoft) <rppt@kernel.org>

> [1] https://lore.kernel.org/linux-mm/20260211195233.368497-1-vishal.moola@gmail.com/
> 
> ----------------
> 
> I've compiled and booted both 64 and 32 bit on a kernel with separately
> allocated ptdescs. I ensured the codepaths were hit for all the changes
> except Xen.
> 
> This applies cleanly to the tip bot. I'm thinking this can go through
> Dave's mm.
> 
> Vishal Moola (9):
>   x86/mm/pat: Use IS_ENABLED() instead of ifdef
>   x86/mm/pat: Convert __set_pmd_pte() to ptdescs
>   x86/mm/pat: Convert collapse_pmd_page() to ptdescs
>   x86/mm: Convert arch_sync_kernel_mappings() to ptdescs
>   x86/mm: Convert sync_global_pgds_l5() to ptdescs
>   x86/mm: Convert sync_global_pgds_l4() to ptdescs
>   x86/mm: Convert pgd_page_get_mm() to ptdescs
>   x86/xen: Convert xen_mm_pin_all() to ptdescs
>   x86/xen: Convert xen_mm_unpin_all() to ptdescs
> 
>  arch/x86/include/asm/pgtable.h       |  2 +-
>  arch/x86/include/asm/pgtable_types.h |  2 +-
>  arch/x86/mm/fault.c                  |  8 ++++----
>  arch/x86/mm/init_64.c                | 16 ++++++++--------
>  arch/x86/mm/pat/set_memory.c         | 17 ++++++++---------
>  arch/x86/mm/pgtable.c                |  4 ++--
>  arch/x86/xen/mmu_pv.c                | 22 +++++++++++-----------
>  7 files changed, 35 insertions(+), 36 deletions(-)
> 
> -- 
> 2.54.0
> 

-- 
Sincerely yours,
Mike.

Re: [PATCH 0/9] Convert x86 pgd_lists to ptdescs

[William Kucharski]

Looks great!

Reviewed-by: William Kucharski <william.kucharski@linux.dev>

> On May 27, 2026, at 08:03, Vishal Moola <vishal.moola@gmail.com> wrote:
> 
> This continues the ptdesc conversions for x86[1]. This is a necessary step
> in preparation for the separate allocation of ptdescs from struct page.
> Once ptdescs are not wrappers, the pgd_list should be used to find our
> ptdesc, not page.
> 
> In fact, after patch 4 we can successfully boot a 32-bit x86 kernel
> that separately allocates ptdescs.
> 
> All of the patches are rather straightforward, I just kept them split up
> for ease of review. There are no intended functional changes as ptdescs
> are currently still wrappers around struct page.
> 
> [1] https://lore.kernel.org/linux-mm/20260211195233.368497-1-vishal.moola@gmail.com/
> 
> ----------------
> 
> I've compiled and booted both 64 and 32 bit on a kernel with separately
> allocated ptdescs. I ensured the codepaths were hit for all the changes
> except Xen.
> 
> This applies cleanly to the tip bot. I'm thinking this can go through
> Dave's mm.
> 
> Vishal Moola (9):
>  x86/mm/pat: Use IS_ENABLED() instead of ifdef
>  x86/mm/pat: Convert __set_pmd_pte() to ptdescs
>  x86/mm/pat: Convert collapse_pmd_page() to ptdescs
>  x86/mm: Convert arch_sync_kernel_mappings() to ptdescs
>  x86/mm: Convert sync_global_pgds_l5() to ptdescs
>  x86/mm: Convert sync_global_pgds_l4() to ptdescs
>  x86/mm: Convert pgd_page_get_mm() to ptdescs
>  x86/xen: Convert xen_mm_pin_all() to ptdescs
>  x86/xen: Convert xen_mm_unpin_all() to ptdescs
> 
> arch/x86/include/asm/pgtable.h       |  2 +-
> arch/x86/include/asm/pgtable_types.h |  2 +-
> arch/x86/mm/fault.c                  |  8 ++++----
> arch/x86/mm/init_64.c                | 16 ++++++++--------
> arch/x86/mm/pat/set_memory.c         | 17 ++++++++---------
> arch/x86/mm/pgtable.c                |  4 ++--
> arch/x86/xen/mmu_pv.c                | 22 +++++++++++-----------
> 7 files changed, 35 insertions(+), 36 deletions(-)
> 
> -- 
> 2.54.0
> 
>