[PATCH] mm/huge_memory: only get folio_order() once during __folio_split()

[PATCH] mm/huge_memory: only get folio_order() once during __folio_split()

[Wei Yang]

Before splitting folio, its order keeps the same.

It is only necessary to get folio_order() once.

Also rename order to old_order to represent the original folio order.

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dev Jain <dev.jain@arm.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Lance Yang <lance.yang@linux.dev>
---
 mm/huge_memory.c | 17 ++++++++---------
 1 file changed, 8 insertions(+), 9 deletions(-)

diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 960fe50f2d89..9454fb7d635e 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -3786,7 +3786,7 @@ static int __folio_split(struct folio *folio, unsigned int new_order,
 	bool is_anon = folio_test_anon(folio);
 	struct address_space *mapping = NULL;
 	struct anon_vma *anon_vma = NULL;
-	int order = folio_order(folio);
+	int old_order = folio_order(folio);
 	struct folio *new_folio, *next;
 	int nr_shmem_dropped = 0;
 	int remap_flags = 0;
@@ -3800,7 +3800,7 @@ static int __folio_split(struct folio *folio, unsigned int new_order,
 	if (folio != page_folio(split_at) || folio != page_folio(lock_at))
 		return -EINVAL;
 
-	if (new_order >= folio_order(folio))
+	if (new_order >= old_order)
 		return -EINVAL;
 
 	if (uniform_split && !uniform_split_supported(folio, new_order, true))
@@ -3871,7 +3871,7 @@ static int __folio_split(struct folio *folio, unsigned int new_order,
 
 		if (uniform_split) {
 			xas_set_order(&xas, folio->index, new_order);
-			xas_split_alloc(&xas, folio, folio_order(folio), gfp);
+			xas_split_alloc(&xas, folio, old_order, gfp);
 			if (xas_error(&xas)) {
 				ret = xas_error(&xas);
 				goto out;
@@ -3926,7 +3926,7 @@ static int __folio_split(struct folio *folio, unsigned int new_order,
 		struct lruvec *lruvec;
 		int expected_refs;
 
-		if (folio_order(folio) > 1) {
+		if (old_order > 1) {
 			if (!list_empty(&folio->_deferred_list)) {
 				ds_queue->split_queue_len--;
 				/*
@@ -3939,8 +3939,7 @@ static int __folio_split(struct folio *folio, unsigned int new_order,
 			}
 			if (folio_test_partially_mapped(folio)) {
 				folio_clear_partially_mapped(folio);
-				mod_mthp_stat(folio_order(folio),
-					      MTHP_STAT_NR_ANON_PARTIALLY_MAPPED, -1);
+				mod_mthp_stat(old_order, MTHP_STAT_NR_ANON_PARTIALLY_MAPPED, -1);
 			}
 		}
 		split_queue_unlock(ds_queue);
@@ -4061,7 +4060,7 @@ static int __folio_split(struct folio *folio, unsigned int new_order,
 	if (!ret && is_anon && !folio_is_device_private(folio))
 		remap_flags = RMP_USE_SHARED_ZEROPAGE;
 
-	remap_page(folio, 1 << order, remap_flags);
+	remap_page(folio, 1 << old_order, remap_flags);
 
 	/*
 	 * Unlock all after-split folios except the one containing
@@ -4092,9 +4091,9 @@ static int __folio_split(struct folio *folio, unsigned int new_order,
 		i_mmap_unlock_read(mapping);
 out:
 	xas_destroy(&xas);
-	if (order == HPAGE_PMD_ORDER)
+	if (old_order == HPAGE_PMD_ORDER)
 		count_vm_event(!ret ? THP_SPLIT_PAGE : THP_SPLIT_PAGE_FAILED);
-	count_mthp_stat(order, !ret ? MTHP_STAT_SPLIT : MTHP_STAT_SPLIT_FAILED);
+	count_mthp_stat(old_order, !ret ? MTHP_STAT_SPLIT : MTHP_STAT_SPLIT_FAILED);
 	return ret;
 }
 
-- 
2.34.1

Re: [PATCH] mm/huge_memory: only get folio_order() once during __folio_split()

[Lance Yang]

On 2025/10/10 22:11, Wei Yang wrote:
> Before splitting folio, its order keeps the same.
> 
> It is only necessary to get folio_order() once.
> 
> Also rename order to old_order to represent the original folio order.

Nice cleanup! I was thinking of doing the same thing :)

Getting the order once at the beginning makes sense to me.

Acked-by: Lance Yang <lance.yang@linux.dev>

Re: [PATCH] mm/huge_memory: only get folio_order() once during __folio_split()

[David Hildenbrand]

On 10.10.25 16:11, Wei Yang wrote:
> Before splitting folio, its order keeps the same.
> 
> It is only necessary to get folio_order() once.
> 
> Also rename order to old_order to represent the original folio order.
> 
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
> Cc: Zi Yan <ziy@nvidia.com>
> Cc: Dev Jain <dev.jain@arm.com>
> Cc: David Hildenbrand <david@redhat.com>
> Cc: Lance Yang <lance.yang@linux.dev>
> ---

LGTM

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

-- 
Cheers

David / dhildenb