Skip to contentSkip to conversation

mm/page_isolation: clarify FIXME around shrink_slab() in memory hotplug

PATCH In review Manish Kumar 3 participants 3 messages Oct 15, 2025 – Nov 11, 2025

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 mm/page_isolation: clarify FIXME around shrink_slab() in memory hotplug

The existing FIXME comment notes that memory hotplug doesn't invoke shrink_slab() directly. This patch adds context explaining that this is an intentional design choice to avoid recursion or deadlocks in the memory reclaim path, as slab shrinking is handled by vmscan.

---
 mm/page_isolation.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
mm/page_isolation.c +8 -2
index b2fc5266e3d2..2ca20c3f0a97 100644
@@ -176,10 +176,16 @@ static int set_migratetype_isolate(struct page *page, int migratetype, int isol_
 
 	/*
 	 * FIXME: Now, memory hotplug doesn't call shrink_slab() by itself.
-	 * We just check MOVABLE pages.
+	 *
+	 * This is an intentional limitation: invoking shrink_slab() from a
+	 * hotplug path can cause reclaim recursion or deadlock if the normal
+	 * memory reclaim (vmscan) path is already active. Slab shrinking is
+	 * handled by the vmscan reclaim code under normal operation, so hotplug
+	 * avoids direct calls into shrink_slab() to prevent reentrancy issues.
+	 *
+	 * We therefore only check MOVABLE pages here.
 	 *
 	 * Pass the intersection of [start_pfn, end_pfn) and the page's pageblock
-	 * to avoid redundant checks.
 	 */
 	check_unmovable_start = max(page_to_pfn(page), start_pfn);
 	check_unmovable_end = min(pageblock_end_pfn(page_to_pfn(page)),
Signature · 1 line
2.43.0
#1 lore raw
Quoted 15 lines · Manish Kumar

The existing FIXME comment notes that memory hotplug doesn't invoke shrink_slab() directly. This patch adds context explaining that this is an intentional design choice to avoid recursion or deadlocks in the memory reclaim path, as slab shrinking is handled by vmscan.

---
  mm/page_isolation.c | 10 ++++++++--
  1 file changed, 8 insertions(+), 2 deletions(-)
mm/page_isolation.c +0 -0
index b2fc5266e3d2..2ca20c3f0a97 100644

mm/page_isolation.c static int set_migratetype_isolate(struct page *page, int migratetype, int isol_ · line 184

Quoted 10 lines
mm/page_isolation.c +6 -1
@@ -176,10 +176,16 @@ static int set_migratetype_isolate(struct page *page, int migratetype, int isol_
 
  	/*
  	 * FIXME: Now, memory hotplug doesn't call shrink_slab() by itself.
-	 * We just check MOVABLE pages.
+	 *
+	 * This is an intentional limitation: invoking shrink_slab() from a
+	 * hotplug path can cause reclaim recursion or deadlock if the normal
+	 * memory reclaim (vmscan) path is already active. Slab shrinking is
+	 * handled by the vmscan reclaim code under normal operation, so hotplug
+	 * avoids direct calls into shrink_slab() to prevent reentrancy issues.

This is the first time I hear about this reentrance issue.

How did you come up with this explanation?

Signature · 3 lines
Cheers

David / dhildenb
maintainer
#2 lore raw
Quoted 29 lines · On Mon, 20 Oct 2025 21:26:42 +0200 David Hildenbrand <[email protected]> wrote:
Quoted 15 lines · Manish Kumar

The existing FIXME comment notes that memory hotplug doesn't invoke shrink_slab() directly. This patch adds context explaining that this is an intentional design choice to avoid recursion or deadlocks in the memory reclaim path, as slab shrinking is handled by vmscan.

---
  mm/page_isolation.c | 10 ++++++++--
  1 file changed, 8 insertions(+), 2 deletions(-)
mm/page_isolation.c +0 -0
index b2fc5266e3d2..2ca20c3f0a97 100644

mm/page_isolation.c static int set_migratetype_isolate(struct page *page, int migratetype, int isol_ · line 184

Quoted 10 lines
mm/page_isolation.c +6 -1
@@ -176,10 +176,16 @@ static int set_migratetype_isolate(struct page *page, int migratetype, int isol_
 
  	/*
  	 * FIXME: Now, memory hotplug doesn't call shrink_slab() by itself.
-	 * We just check MOVABLE pages.
+	 *
+	 * This is an intentional limitation: invoking shrink_slab() from a
+	 * hotplug path can cause reclaim recursion or deadlock if the normal
+	 * memory reclaim (vmscan) path is already active. Slab shrinking is
+	 * handled by the vmscan reclaim code under normal operation, so hotplug
+	 * avoids direct calls into shrink_slab() to prevent reentrancy issues.

This is the first time I hear about this reentrance issue.

Quoted 2 lines

How did you come up with this explanation?

-ETIMEDOUT, I'll drop this patch.