Skip to contentSkip to conversation

mm/huge_memory: check new folio order when split a folio

PATCH Applied Zi Yan 3 participants 5 messages Mar 7, 2024

Subsystems (2)
lore
Share & tools
lore

Zi Yan

author
#0 lore raw

PATCH 1/2 mm/huge_memory: check new folio order when split a folio

From: Zi Yan <[email protected]>

A folio can only be split into lower orders. Check new_order to make sure it is smaller than input folio order.

Fixes c010d47f107f ("mm: thp: split huge page to any lower order pages")
---
 mm/huge_memory.c | 3 +++
 1 file changed, 3 insertions(+)
mm/huge_memory.c +3 -0
index a81a09236c16..57fca7bffd20 100644
@@ -3052,6 +3052,9 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
 	VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
 	VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
 
+	if (new_order >= folio_order(folio))
+		return -EINVAL;
+
 	/* Cannot split anonymous THP to order-1 */
 	if (new_order == 1 && folio_test_anon(folio)) {
 		VM_WARN_ONCE(1, "Cannot split to order-1 folio");
Signature · 1 line
2.43.0
author
#1 lore raw

PATCH 2/2 mm/huge_memory: skip invalid debugfs new_order input for folio split

From: Zi Yan <[email protected]>

User can put arbitrary new_order via debugfs for folio split test. Although new_order check is added to split_huge_page_to_list_order() in the prior commit, these two additional checks can avoid unnecessary folio locking and split_folio_to_order() calls.

---
 mm/huge_memory.c | 6 ++++++
 1 file changed, 6 insertions(+)
mm/huge_memory.c +6 -0
index 57fca7bffd20..9859aa4f7553 100644
@@ -3486,6 +3486,9 @@ static int split_huge_pages_pid(int pid, unsigned long vaddr_start,
 		if (!is_transparent_hugepage(folio))
 			goto next;
 
+		if (new_order >= folio_order(folio))
+			goto next;
+
 		total++;
 		/*
 		 * For folios with private, split_huge_page_to_list_to_order()
@@ -3553,6 +3556,9 @@ static int split_huge_pages_in_file(const char *file_path, pgoff_t off_start,
 		total++;
 		nr_pages = folio_nr_pages(folio);
 
+		if (new_order >= folio_order(folio))
+			goto next;
+
 		if (!folio_trylock(folio))
 			goto next;
 
Signature · 1 line
2.43.0
maintainer
#2 lore raw
In reply to Zi Yan
Quoted 4 lines · On Thu, 7 Mar 2024 13:18:53 -0500 Zi Yan <[email protected]> wrote:

From: Zi Yan <[email protected]>

A folio can only be split into lower orders. Check new_order to make sure it is smaller than input folio order.

It isn't clear what's being fixed here. Presumably something is passing in such folios, but what and where and why and what are the effects?

Might it be that these folios are being caused by the debugfs interface? Or something else?

So I'll add it, but I do think more information and context would improve the patch, please. Suitable Reported-by:, Closes: and Link:
tags, perhaps.

reviewer
#3 lore raw
Quoted 7 lines · Andrew Morton
Quoted 4 lines · On Thu, 7 Mar 2024 13:18:53 -0500 Zi Yan <[email protected]> wrote:

From: Zi Yan <[email protected]>

A folio can only be split into lower orders. Check new_order to make sure it is smaller than input folio order.

Quoted 6 lines

It isn't clear what's being fixed here. Presumably something is passing in such folios, but what and where and why and what are the effects?

Might it be that these folios are being caused by the debugfs interface? Or something else?

Since there is no new_order checks in debugfs before, any new_order can be passed via debugfs into split_huge_page_to_list_to_order().

I did not explicitly mention it here as the debugfs is added in the commit after the Fixes one.

Quoted 4 lines

So I'll add it, but I do think more information and context would improve the patch, please. Suitable Reported-by:, Closes: and Link:
tags, perhaps.

--
Best Regards,
Yan, Zi