[PATCH 2/2] vmw_balloon: account 2 MiB pages in NR_BALLOON_PAGES
From: David Hildenbrand (Arm)
Date: Fri Nov 20 2026 - 05:35:01 EST
vmw_balloon keeps 2 MiB balloon-inflated pages on its own list, not
using balloon core that takes otherwise care of updating NR_BALLOON_PAGES
for base pages.
So far we didn't account 2 MiB pages under the "Balloon" counters, so
/proc/meminfo etc. would not reflect them.
Fix it by updating NR_BALLOON_PAGES when enqueuing and dequeuing for
these pages manually.
This has low urgency.
Fixes: 4d689474e1b2 ("balloon_compaction: update the NR_BALLOON_PAGES state")
Cc: stable@xxxxxxxxxxxxxxx
Signed-off-by: David Hildenbrand (Arm) <david@xxxxxxxxxx>
---
drivers/misc/vmw_balloon.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/drivers/misc/vmw_balloon.c b/drivers/misc/vmw_balloon.c
index 3edb934dd7fd5..47dbcb3be6ed6 100644
--- a/drivers/misc/vmw_balloon.c
+++ b/drivers/misc/vmw_balloon.c
@@ -1005,6 +1005,8 @@ static void vmballoon_enqueue_page_list(struct vmballoon *b,
list_for_each_entry(page, pages, lru) {
vmballoon_mark_page_offline(page, VMW_BALLOON_2M_PAGE);
+ mod_node_page_state(page_pgdat(page), NR_BALLOON_PAGES,
+ 1 << VMW_BALLOON_2M_ORDER);
}
list_splice_init(pages, &b->huge_pages);
@@ -1049,6 +1051,8 @@ static void vmballoon_dequeue_page_list(struct vmballoon *b,
spin_lock(&b->huge_pages_lock);
list_for_each_entry_safe(page, tmp, &b->huge_pages, lru) {
vmballoon_mark_page_online(page, VMW_BALLOON_2M_PAGE);
+ mod_node_page_state(page_pgdat(page), NR_BALLOON_PAGES,
+ -(1 << VMW_BALLOON_2M_ORDER));
list_move(&page->lru, pages);
if (++i == n_req_pages)
--
2.56.1