[PATCH 0/3] Some ptdesc cleanups

[PATCH 0/3] Some ptdesc cleanups

[Matthew Wilcox (Oracle)]

The first two patches here are preparation for splitting struct ptdesc
from struct page and struct folio.  I think their only dependency is on
the memdesc_flags_t patches from August which is in mm-new.  The third
patch is just something I noticed while working on the code.

Matthew Wilcox (Oracle) (3):
  ptdesc: Convert __page_flags to pt_flags
  ptdesc: Remove references to folios from __pagetable_ctor() and
    pagetable_dtor()
  ptdesc: Remove ptdesc_to_virt()

 arch/arm/mm/mmu.c        |  2 +-
 arch/s390/mm/pgalloc.c   |  6 +++---
 include/linux/mm.h       | 39 +++++++++++++++++++++++++--------------
 include/linux/mm_types.h |  6 +++---
 4 files changed, 32 insertions(+), 21 deletions(-)

-- 
2.47.2

[PATCH 1/3] ptdesc: Convert __page_flags to pt_flags

[Matthew Wilcox (Oracle)]

Use the new memdesc_flags_t type to show that these are the same
bits as page/folio/slab and thesefore have the zone/node/section
information in them.  Remove a use of ptdesc_folio() by converting
pagetable_is_reserved() to use test_bit() directly.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
---
 include/linux/mm.h       | 7 ++++++-
 include/linux/mm_types.h | 6 +++---
 2 files changed, 9 insertions(+), 4 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index 02fddbc38647..46eb2a7c4fb7 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -2902,6 +2902,11 @@ static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long a
 }
 #endif /* CONFIG_MMU */
 
+enum pt_flags {
+	PT_reserved = PG_reserved,
+	/* High bits are used for zone/node/section */
+};
+
 static inline struct ptdesc *virt_to_ptdesc(const void *x)
 {
 	return page_ptdesc(virt_to_page(x));
@@ -2919,7 +2924,7 @@ static inline void *ptdesc_address(const struct ptdesc *pt)
 
 static inline bool pagetable_is_reserved(struct ptdesc *pt)
 {
-	return folio_test_reserved(ptdesc_folio(pt));
+	return test_bit(PT_reserved, &pt->pt_flags.f);
 }
 
 /**
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index f7047483dd60..d3566bed5817 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -522,7 +522,7 @@ FOLIO_MATCH(compound_head, _head_3);
 
 /**
  * struct ptdesc -    Memory descriptor for page tables.
- * @__page_flags:     Same as page flags. Powerpc only.
+ * @pt_flags: enum pt_flags plus zone/node/section.
  * @pt_rcu_head:      For freeing page table pages.
  * @pt_list:          List of used page tables. Used for s390 gmap shadow pages
  *                    (which are not linked into the user page tables) and x86
@@ -544,7 +544,7 @@ FOLIO_MATCH(compound_head, _head_3);
  * understanding of the issues.
  */
 struct ptdesc {
-	unsigned long __page_flags;
+	memdesc_flags_t pt_flags;
 
 	union {
 		struct rcu_head pt_rcu_head;
@@ -582,7 +582,7 @@ struct ptdesc {
 
 #define TABLE_MATCH(pg, pt)						\
 	static_assert(offsetof(struct page, pg) == offsetof(struct ptdesc, pt))
-TABLE_MATCH(flags, __page_flags);
+TABLE_MATCH(flags, pt_flags);
 TABLE_MATCH(compound_head, pt_list);
 TABLE_MATCH(compound_head, _pt_pad_1);
 TABLE_MATCH(mapping, __page_mapping);
-- 
2.47.2

Re: [PATCH 1/3] ptdesc: Convert __page_flags to pt_flags

[David Hildenbrand]

On 08.09.25 19:11, Matthew Wilcox (Oracle) wrote:
> Use the new memdesc_flags_t type to show that these are the same
> bits as page/folio/slab and thesefore have the zone/node/section
> information in them.  Remove a use of ptdesc_folio() by converting
> pagetable_is_reserved() to use test_bit() directly.
> 
> Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
> ---

Acked-by: David Hildenbrand <david@redhat.com>

-- 
Cheers

David / dhildenb

[PATCH 2/3] ptdesc: Remove references to folios from __pagetable_ctor() and pagetable_dtor()

[Matthew Wilcox (Oracle)]

In preparation for splitting struct ptdesc from struct page and struct
folio, remove mentions of struct folio from these functions.  Introduce
ptdesc_nr_pages() to avoid using lruvec_stat_add/sub_folio()

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
---
 include/linux/mm.h | 21 +++++++++++++--------
 1 file changed, 13 insertions(+), 8 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index 46eb2a7c4fb7..3a339bd7a2f9 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -2065,9 +2065,9 @@ static inline long folio_nr_pages(const struct folio *folio)
  * page.  compound_nr() can be called on a tail page, and is defined to
  * return 1 in that case.
  */
-static inline long compound_nr(struct page *page)
+static inline long compound_nr(const struct page *page)
 {
-	struct folio *folio = (struct folio *)page;
+	const struct folio *folio = (struct folio *)page;
 
 	if (!test_bit(PG_head, &folio->flags.f))
 		return 1;
@@ -3034,21 +3034,26 @@ static inline bool ptlock_init(struct ptdesc *ptdesc) { return true; }
 static inline void ptlock_free(struct ptdesc *ptdesc) {}
 #endif /* defined(CONFIG_SPLIT_PTE_PTLOCKS) */
 
+static inline unsigned long ptdesc_nr_pages(const struct ptdesc *ptdesc)
+{
+	return compound_nr(ptdesc_page(ptdesc));
+}
+
 static inline void __pagetable_ctor(struct ptdesc *ptdesc)
 {
-	struct folio *folio = ptdesc_folio(ptdesc);
+	pg_data_t *pgdat = NODE_DATA(memdesc_nid(ptdesc->pt_flags));
 
-	__folio_set_pgtable(folio);
-	lruvec_stat_add_folio(folio, NR_PAGETABLE);
+	__SetPageTable(ptdesc_page(ptdesc));
+	mod_node_page_state(pgdat, NR_PAGETABLE, ptdesc_nr_pages(ptdesc));
 }
 
 static inline void pagetable_dtor(struct ptdesc *ptdesc)
 {
-	struct folio *folio = ptdesc_folio(ptdesc);
+	pg_data_t *pgdat = NODE_DATA(memdesc_nid(ptdesc->pt_flags));
 
 	ptlock_free(ptdesc);
-	__folio_clear_pgtable(folio);
-	lruvec_stat_sub_folio(folio, NR_PAGETABLE);
+	__ClearPageTable(ptdesc_page(ptdesc));
+	mod_node_page_state(pgdat, NR_PAGETABLE, -ptdesc_nr_pages(ptdesc));
 }
 
 static inline void pagetable_dtor_free(struct ptdesc *ptdesc)
-- 
2.47.2

Re: [PATCH 2/3] ptdesc: Remove references to folios from __pagetable_ctor() and pagetable_dtor()

[David Hildenbrand]

On 08.09.25 19:11, Matthew Wilcox (Oracle) wrote:
> In preparation for splitting struct ptdesc from struct page and struct
> folio, remove mentions of struct folio from these functions.  Introduce
> ptdesc_nr_pages() to avoid using lruvec_stat_add/sub_folio()
> 
> Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
> ---

Acked-by: David Hildenbrand <david@redhat.com>

-- 
Cheers

David / dhildenb

[PATCH 3/3] ptdesc: Remove ptdesc_to_virt()

[Matthew Wilcox (Oracle)]

This has the same effect as ptdesc_address() so convert the callers to
use that and delete the function.  Add kernel-doc for ptdesc_address().

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
---
 arch/arm/mm/mmu.c      |  2 +-
 arch/s390/mm/pgalloc.c |  6 +++---
 include/linux/mm.h     | 11 ++++++-----
 3 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/arch/arm/mm/mmu.c b/arch/arm/mm/mmu.c
index edb7f56b7c91..8bac96e205ac 100644
--- a/arch/arm/mm/mmu.c
+++ b/arch/arm/mm/mmu.c
@@ -737,7 +737,7 @@ static void *__init late_alloc(unsigned long sz)
 
 	if (!ptdesc || !pagetable_pte_ctor(NULL, ptdesc))
 		BUG();
-	return ptdesc_to_virt(ptdesc);
+	return ptdesc_address(ptdesc);
 }
 
 static pte_t * __init arm_pte_alloc(pmd_t *pmd, unsigned long addr,
diff --git a/arch/s390/mm/pgalloc.c b/arch/s390/mm/pgalloc.c
index d2f6f1f6d2fc..f56ee9aeac83 100644
--- a/arch/s390/mm/pgalloc.c
+++ b/arch/s390/mm/pgalloc.c
@@ -21,7 +21,7 @@ unsigned long *crst_table_alloc(struct mm_struct *mm)
 
 	if (!ptdesc)
 		return NULL;
-	table = ptdesc_to_virt(ptdesc);
+	table = ptdesc_address(ptdesc);
 	__arch_set_page_dat(table, 1UL << CRST_ALLOC_ORDER);
 	return table;
 }
@@ -119,7 +119,7 @@ struct ptdesc *page_table_alloc_pgste(struct mm_struct *mm)
 
 	ptdesc = pagetable_alloc(GFP_KERNEL, 0);
 	if (ptdesc) {
-		table = (u64 *)ptdesc_to_virt(ptdesc);
+		table = (u64 *)ptdesc_address(ptdesc);
 		__arch_set_page_dat(table, 1);
 		memset64(table, _PAGE_INVALID, PTRS_PER_PTE);
 		memset64(table + PTRS_PER_PTE, 0, PTRS_PER_PTE);
@@ -146,7 +146,7 @@ unsigned long *page_table_alloc(struct mm_struct *mm)
 		pagetable_free(ptdesc);
 		return NULL;
 	}
-	table = ptdesc_to_virt(ptdesc);
+	table = ptdesc_address(ptdesc);
 	__arch_set_page_dat(table, 1);
 	memset64((u64 *)table, _PAGE_INVALID, PTRS_PER_PTE);
 	memset64((u64 *)table + PTRS_PER_PTE, 0, PTRS_PER_PTE);
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 3a339bd7a2f9..2c3b99f60297 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -2912,11 +2912,12 @@ static inline struct ptdesc *virt_to_ptdesc(const void *x)
 	return page_ptdesc(virt_to_page(x));
 }
 
-static inline void *ptdesc_to_virt(const struct ptdesc *pt)
-{
-	return page_to_virt(ptdesc_page(pt));
-}
-
+/**
+ * ptdesc_address - Virtual address of page table.
+ * @pt: Page table descriptor.
+ *
+ * Return: The first byte of the page table described by @pt.
+ */
 static inline void *ptdesc_address(const struct ptdesc *pt)
 {
 	return folio_address(ptdesc_folio(pt));
-- 
2.47.2

Re: [PATCH 3/3] ptdesc: Remove ptdesc_to_virt()

[David Hildenbrand]

> -
> +/**
> + * ptdesc_address - Virtual address of page table.
> + * @pt: Page table descriptor.
> + *
> + * Return: The first byte of the page table described by @pt.
> + */
>   static inline void *ptdesc_address(const struct ptdesc *pt)
>   {
>   	return folio_address(ptdesc_folio(pt));

I assume you'll tackle the ptdesc_folio() stuff separately.

Acked-by: David Hildenbrand <david@redhat.com>

-- 
Cheers

David / dhildenb