Re: [PATCH v2 3/5] mm: hugetlb: optimize replace_free_hugepage_folios()

[Andrew Morton <akpm@linux-foundation.org>]

On Wed, 14 Jan 2026 21:55:12 +0800 Kefeng Wang <wangkefeng.wang@huawei.com> wrote:

> If no free hugepage folios are available, there is no need to perform
> any replacement operations. Additionally, gigantic folios should not
> be replaced under any circumstances. Therefore, we only check for the
> presence of non-gigantic folios, also adding the gigantic folio check
> to avoid accidental replacement.
> 
> To optimize performance, we skip unnecessary iterations over pfn
> for compound pages and high-order buddy pages to save processing time.
> 
> A simple test on machine with 114G free memory, allocate 120 * 1G
> HugeTLB folios(104 successfully returned),
> 
>   time echo 120 > /sys/kernel/mm/hugepages/hugepages-1048576kB/nr_hugepages
> 
> Before: 0m0.602s
> After:  0m0.431s
> 

Thanks, I'll queue this as a -fix, below.

nit: the function has become rather a jumble of bailout styles:

:				if (order_is_gigantic(folio_order(folio)))
: 					return -ENOMEM;

This could have done ret=-ENOMEM;break;

: 
: 				ret = alloc_and_dissolve_hugetlb_folio(folio, &list);
: 				if (ret)
: 					break;

While this could have done `return ret;'!

Single-return-point is preferred style, I think.  Mainly to help avoid
leakage of resources and locks as the code evolves.

How does this look on top?

From: Andrew Morton <akpm@linux-foundation.org>
Subject: mm-hugetlb-optimize-replace_free_hugepage_folios-v2-fix
Date: Thu Jan 15 02:34:07 PM PST 2026

use single-return-point style, tweak comment

Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: Brendan Jackman <jackmanb@google.com>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Sidhartha Kumar <sidhartha.kumar@oracle.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <ziy@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---

 mm/hugetlb.c |   11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

--- a/mm/hugetlb.c~mm-hugetlb-optimize-replace_free_hugepage_folios-v2-fix
+++ a/mm/hugetlb.c
@@ -2834,10 +2834,15 @@ int replace_free_hugepage_folios(unsigne
 
 			nr = folio_nr_pages(folio) - folio_page_idx(folio, page);
 
-			/* Not to disrupt normal path by vainly holding hugetlb_lock */
+			/*
+			 * Don't disrupt normal path by vainly holding
+			 * hugetlb_lock
+			 */
 			if (folio_test_hugetlb(folio) && !folio_ref_count(folio)) {
-				if (order_is_gigantic(folio_order(folio)))
-					return -ENOMEM;
+				if (order_is_gigantic(folio_order(folio))) {
+					ret = -ENOMEM;
+					break;
+				}
 
 				ret = alloc_and_dissolve_hugetlb_folio(folio, &list);
 				if (ret)
_






From: Kefeng Wang <wangkefeng.wang@huawei.com>
Subject: mm: hugetlb: optimize replace_free_hugepage_folios()
Date: Wed, 14 Jan 2026 21:55:12 +0800

Turn back to use alloc_and_dissolve_hugetlb_folio() since Oscar pointed
out that the return value is different.  Then adding gigantic folio check
independently, also update changelog.

Link: https://lkml.kernel.org/r/20260114135512.2159799-1-wangkefeng.wang@huawei.com
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: Brendan Jackman <jackmanb@google.com>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Sidhartha Kumar <sidhartha.kumar@oracle.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <ziy@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---

 mm/hugetlb.c |   10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

--- a/mm/hugetlb.c~mm-hugetlb-optimize-replace_free_hugepage_folios-v2
+++ a/mm/hugetlb.c
@@ -2810,6 +2810,7 @@ int replace_free_hugepage_folios(unsigne
 	struct page *page;
 	struct hstate *h;
 	LIST_HEAD(list);
+	int ret = 0;
 
 	/* Avoid pfn iterations if no free non-gigantic huge pages */
 	for_each_hstate(h) {
@@ -2835,9 +2836,13 @@ int replace_free_hugepage_folios(unsigne
 
 			/* Not to disrupt normal path by vainly holding hugetlb_lock */
 			if (folio_test_hugetlb(folio) && !folio_ref_count(folio)) {
-				if (isolate_or_dissolve_huge_folio(folio, &list))
+				if (order_is_gigantic(folio_order(folio)))
 					return -ENOMEM;
 
+				ret = alloc_and_dissolve_hugetlb_folio(folio, &list);
+				if (ret)
+					break;
+
 				putback_movable_pages(&list);
 			}
 		} else if (PageBuddy(page)) {
@@ -2851,11 +2856,10 @@ int replace_free_hugepage_folios(unsigne
 			if (order <= MAX_PAGE_ORDER)
 				nr = 1UL << order;
 		}
-
 		start_pfn += nr;
 	}
 
-	return 0;
+	return ret;
 }
 
 void wait_for_freed_hugetlb_folios(void)
_