Skip to contentSkip to conversation

mm/mremap: Use can_pte_batch_count() instead of folio_pte_batch() for pte batch

RFC PATCH zhangqilong 1 participant 1 message Oct 28, 2025 Memory management

lore
Share & tools
lore
#0 lore raw
Quoted 61 lines · Zhang Qilong

mm/mremap.c static int move_ptes(struct

Quoted 50 lines

In current mremap_folio_pte_batch(), 1) pte_batch_hint() always return one pte in non-ARM64 machine, it is not efficient. 2) Next, it need to acquire a folio to call the folio_pte_batch().

Due to new added can_pte_batch_count(), we just call it instead of folio_pte_batch(). And then rename mremap_folio_pte_batch() to mremap_pte_batch().

---
  mm/mremap.c | 16 +++-------------
  1 file changed, 3 insertions(+), 13 deletions(-)
mm/mremap.c index renamed from mm/mremap.c +0 -0

bd7314898ec5..d11f93f1622f 100644

mm/mremap.c +3 -12
@@ -169,27 +169,17 @@ static pte_t move_soft_dirty_pte(pte_t pte)
  		pte = pte_swp_mksoft_dirty(pte);
  #endif
  	return pte;
  }
 
-static int mremap_folio_pte_batch(struct vm_area_struct *vma,
 unsigned long addr,
+static int mremap_pte_batch(struct vm_area_struct *vma, unsigned long
+addr,
  		pte_t *ptep, pte_t pte, int max_nr)
  {
-	struct folio *folio;
-
  	if (max_nr == 1)
  		return 1;
 
-	/* Avoid expensive folio lookup if we stand no chance of benefit. */
-	if (pte_batch_hint(ptep, pte) == 1)
-		return 1;
-
-	folio = vm_normal_folio(vma, addr, pte);
-	if (!folio || !folio_test_large(folio))
-		return 1;
-
-	return folio_pte_batch(folio, ptep, pte, max_nr);
+	return can_pte_batch_count(vma, ptep, &pte, max_nr, 0);
  }
 
  static int move_ptes(struct pagetable_move_control *pmc,
  		unsigned long extent, pmd_t *old_pmd, pmd_t *new_pmd)
  {
@@ -278,11 +268,11 @@ static int move_ptes(struct

pagetable_move_control *pmc,

mm/mremap.c static int move_ptes(struct · line 283

Quoted 6 lines
mm/mremap.c +0 -1
  		 * make sure the physical page stays valid until
  		 * the TLB entry for the old mapping has been
  		 * flushed.
  		 */
  		if (pte_present(old_pte)) {
-			nr_ptes = mremap_folio_pte_batch(vma, old_addr,

old_ptep,

mm/mremap.c static int move_ptes(struct · line 274

Quoted 2 lines
mm/mremap.c +1 -0
+			nr_ptes = mremap_pte_batch(vma, old_addr, old_ptep,
  							 old_pte,

max_nr_ptes);

Quoted 6 lines

mm/mremap.c static int move_ptes(struct · line 277

Quoted 3 lines
mm/mremap.c +0 -0
  			force_flush = true;
  		}
  		pte = get_and_clear_ptes(mm, old_addr, old_ptep, nr_ptes);

get_and_clear_ptes() documents: "Clear present PTEs that map consecutive pages of the same folio, collecting dirty/accessed bits."

Oh, good catch. My focus was solely on the implementations of get_and_clear_ptes() and set_ptes() and regarding their multi-folio PTEs handling, and I missed this comment. get_and_clear_ptes() will collect dirty/accessed bits in batch ranges, and set in later.

Quoted 3 lines

And as can_pte_batch_count() will merge access/dirty bits, you would silently set ptes dirty/accessed that belong to other folios, which sounds very wrong.

Year, your analysis is very thorough. The access/dirty bit will be merged between neighboring batched folios due to get_and_clear_ptes().

If caller don't want to merge access/dirty bits between folios(means not ignore any bits), they should call folio_pte_batch() or the new interface with 'flag | FPB_RESPECT_DIRTY ' (access bit is default be respected).

Quoted 3 lines

Staring at the code, I wonder if there is also a problem with the write bit, have to dig into that.

The write bit is handled similarly to dirty bit. If called don't want merge write between folios, we could call the new interface with 'flag | FPB_RESPECT_WRITE', and pte_same() will compare the write bit with the next neighboring folio pte. If it's different, it will break.

Thanks for the in-depth review.

Quoted 5 lines

--
Cheers

David / dhildenb