slab: Avoid race on slab->obj_exts in alloc_slab_obj_exts

PATCH In review Hao Ge 4 participants 4 messages Oct 21, 2025

Subsystems (3)

Stable branch Memory management Slab allocator

lore

Share & tools

lore

Hao Ge

author

2025-10-21 01:04 UTC #0 lore raw

PATCH v3 1/1 slab: Avoid race on slab->obj_exts in alloc_slab_obj_exts

From: Hao Ge <[email protected]>

If two competing threads enter alloc_slab_obj_exts() and one of them fails to allocate the object extension vector, it might override the valid slab->obj_exts allocated by the other thread with OBJEXTS_ALLOC_FAIL. This will cause the thread that lost this race and expects a valid pointer to dereference a NULL pointer later on.

Update slab->obj_exts atomically using cmpxchg() to avoid slab->obj_exts overrides by racing threads.

Thanks for Vlastimil and Suren's help with debugging.

Fixes f7381b911640 ("slab: mark slab->obj_exts allocation failures unconditionally")

Cc <[email protected]>

Signed-off-by Hao Ge <[email protected]>

---
v3: According to Suren's suggestion, simplify the commit message and the code comments.
Thanks for Suren.

v2: Incorporate handling for the scenario where, if mark_failed_objexts_alloc wins the race,
the other process (that previously succeeded in allocation) will lose the race, based on Suren's suggestion. Add Suggested-by: Suren Baghdasaryan <[email protected]>

---
 mm/slub.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

mm/slub.c +6 -3

index 2e4340c75be2..d4403341c9df 100644
@@ -2054,7 +2054,7 @@ static inline void mark_objexts_empty(struct slabobj_ext *obj_exts)
 
 static inline void mark_failed_objexts_alloc(struct slab *slab)
 {
-	slab->obj_exts = OBJEXTS_ALLOC_FAIL;
+	cmpxchg(&slab->obj_exts, 0, OBJEXTS_ALLOC_FAIL);
 }
 
 static inline void handle_failed_objexts_alloc(unsigned long obj_exts,

@@ -2136,6 +2136,7 @@ int alloc_slab_obj_exts(struct slab *slab, struct kmem_cache *s,
 #ifdef CONFIG_MEMCG
 	new_exts |= MEMCG_DATA_OBJEXTS;
 #endif
+retry:
 	old_exts = READ_ONCE(slab->obj_exts);
 	handle_failed_objexts_alloc(old_exts, vec, objects);
 	if (new_slab) {

@@ -2145,8 +2146,7 @@ int alloc_slab_obj_exts(struct slab *slab, struct kmem_cache *s,
 		 * be simply assigned.
 		 */
 		slab->obj_exts = new_exts;
-	} else if ((old_exts & ~OBJEXTS_FLAGS_MASK) ||
-		   cmpxchg(&slab->obj_exts, old_exts, new_exts) != old_exts) {
+	} else if (old_exts & ~OBJEXTS_FLAGS_MASK) {
 		/*
 		 * If the slab is already in use, somebody can allocate and
 		 * assign slabobj_exts in parallel. In this case the existing

@@ -2158,6 +2158,9 @@ int alloc_slab_obj_exts(struct slab *slab, struct kmem_cache *s,
 		else
 			kfree(vec);
 		return 0;
+	} else if (cmpxchg(&slab->obj_exts, old_exts, new_exts) != old_exts) {
+		/* Retry if a racing thread changed slab->obj_exts from under us. */
+		goto retry;
 	}
 
 	if (allow_spin)

Signature · 1 line

2.25.1

Harry Yoo

Reviewed-by on v3 1/1

2025-10-21 02:16 UTC #1 lore raw

Quoted 17 lines · Hao Ge

From: Hao Ge <[email protected]>
If two competing threads enter alloc_slab_obj_exts() and one of them fails to allocate the object extension vector, it might override the valid slab->obj_exts allocated by the other thread with OBJEXTS_ALLOC_FAIL. This will cause the thread that lost this race and expects a valid pointer to dereference a NULL pointer later on.
Update slab->obj_exts atomically using cmpxchg() to avoid slab->obj_exts overrides by racing threads.
Thanks for Vlastimil and Suren's help with debugging.
Fixes f7381b911640 ("slab: mark slab->obj_exts allocation failures unconditionally")
Cc <[email protected]>
Suggested-by Suren Baghdasaryan <[email protected]>
Signed-off-by Hao Ge <[email protected]>

Quoted 8 lines

---
v3: According to Suren's suggestion, simplify the commit message and the code comments.
Thanks for Suren.
v2: Incorporate handling for the scenario where, if mark_failed_objexts_alloc wins the race,
the other process (that previously succeeded in allocation) will lose the race, based on Suren's suggestion. Add Suggested-by: Suren Baghdasaryan <[email protected]>
---

Looks good to me, thanks for fixing this!

Reviewed-by Harry Yoo <[email protected]>

Signature · 2 lines

Cheers,
Harry / Hyeonggon

Vlastimil Babka

2025-10-21 07:04 UTC #2 lore raw

In reply to Hao Ge

Quoted 11 lines · Hao Ge

From: Hao Ge <[email protected]>
If two competing threads enter alloc_slab_obj_exts() and one of them fails to allocate the object extension vector, it might override the valid slab->obj_exts allocated by the other thread with OBJEXTS_ALLOC_FAIL. This will cause the thread that lost this race and expects a valid pointer to dereference a NULL pointer later on.
Update slab->obj_exts atomically using cmpxchg() to avoid slab->obj_exts overrides by racing threads.

Quoted 6 lines

Thanks for Vlastimil and Suren's help with debugging.
Fixes f7381b911640 ("slab: mark slab->obj_exts allocation failures unconditionally")
Cc <[email protected]>
Suggested-by Suren Baghdasaryan <[email protected]>
Signed-off-by Hao Ge <[email protected]>

Added to slab/for-next-fixes, thanks!

Quoted 52 lines

---
v3: According to Suren's suggestion, simplify the commit message and the code comments.
Thanks for Suren.

---
 mm/slub.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

mm/slub.c +6 -3

index 2e4340c75be2..d4403341c9df 100644
@@ -2054,7 +2054,7 @@ static inline void mark_objexts_empty(struct slabobj_ext *obj_exts)
 
 static inline void mark_failed_objexts_alloc(struct slab *slab)
 {
-	slab->obj_exts = OBJEXTS_ALLOC_FAIL;
+	cmpxchg(&slab->obj_exts, 0, OBJEXTS_ALLOC_FAIL);
 }
 
 static inline void handle_failed_objexts_alloc(unsigned long obj_exts,

@@ -2136,6 +2136,7 @@ int alloc_slab_obj_exts(struct slab *slab, struct kmem_cache *s,
 #ifdef CONFIG_MEMCG
 	new_exts |= MEMCG_DATA_OBJEXTS;
 #endif
+retry:
 	old_exts = READ_ONCE(slab->obj_exts);
 	handle_failed_objexts_alloc(old_exts, vec, objects);
 	if (new_slab) {

@@ -2145,8 +2146,7 @@ int alloc_slab_obj_exts(struct slab *slab, struct kmem_cache *s,
 		 * be simply assigned.
 		 */
 		slab->obj_exts = new_exts;
-	} else if ((old_exts & ~OBJEXTS_FLAGS_MASK) ||
-		   cmpxchg(&slab->obj_exts, old_exts, new_exts) != old_exts) {
+	} else if (old_exts & ~OBJEXTS_FLAGS_MASK) {
 		/*
 		 * If the slab is already in use, somebody can allocate and
 		 * assign slabobj_exts in parallel. In this case the existing

@@ -2158,6 +2158,9 @@ int alloc_slab_obj_exts(struct slab *slab, struct kmem_cache *s,
 		else
 			kfree(vec);
 		return 0;
+	} else if (cmpxchg(&slab->obj_exts, old_exts, new_exts) != old_exts) {
+		/* Retry if a racing thread changed slab->obj_exts from under us. */
+		goto retry;
 	}
 
 	if (allow_spin)

Suren Baghdasaryan

Reviewed-by

2025-10-21 13:25 UTC #3 lore raw

Quoted 13 lines · Vlastimil Babka

Quoted 11 lines · Hao Ge
From: Hao Ge <[email protected]>
If two competing threads enter alloc_slab_obj_exts() and one of them fails to allocate the object extension vector, it might override the valid slab->obj_exts allocated by the other thread with OBJEXTS_ALLOC_FAIL. This will cause the thread that lost this race and expects a valid pointer to dereference a NULL pointer later on.
Update slab->obj_exts atomically using cmpxchg() to avoid slab->obj_exts overrides by racing threads.

Quoted 6 lines

Quoted 6 lines
Thanks for Vlastimil and Suren's help with debugging.
Fixes f7381b911640 ("slab: mark slab->obj_exts allocation failures unconditionally")
Cc <[email protected]>
Suggested-by Suren Baghdasaryan <[email protected]>
Signed-off-by Hao Ge <[email protected]>

Reviewed-by Suren Baghdasaryan <[email protected]>

Thanks for the fix, Hao!

Quoted 56 lines

Added to slab/for-next-fixes, thanks!

Quoted 52 lines

---
v3: According to Suren's suggestion, simplify the commit message and the code comments.
Thanks for Suren.

---
 mm/slub.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

mm/slub.c +6 -3

index 2e4340c75be2..d4403341c9df 100644
@@ -2054,7 +2054,7 @@ static inline void mark_objexts_empty(struct slabobj_ext *obj_exts)
 
 static inline void mark_failed_objexts_alloc(struct slab *slab)
 {
-     slab->obj_exts = OBJEXTS_ALLOC_FAIL;
+     cmpxchg(&slab->obj_exts, 0, OBJEXTS_ALLOC_FAIL);
 }
 
 static inline void handle_failed_objexts_alloc(unsigned long obj_exts,

@@ -2136,6 +2136,7 @@ int alloc_slab_obj_exts(struct slab *slab, struct kmem_cache *s,
 #ifdef CONFIG_MEMCG
      new_exts |= MEMCG_DATA_OBJEXTS;
 #endif
+retry:
      old_exts = READ_ONCE(slab->obj_exts);
      handle_failed_objexts_alloc(old_exts, vec, objects);
      if (new_slab) {

@@ -2145,8 +2146,7 @@ int alloc_slab_obj_exts(struct slab *slab, struct kmem_cache *s,
               * be simply assigned.
               */
              slab->obj_exts = new_exts;
-     } else if ((old_exts & ~OBJEXTS_FLAGS_MASK) ||
-                cmpxchg(&slab->obj_exts, old_exts, new_exts) != old_exts) {
+     } else if (old_exts & ~OBJEXTS_FLAGS_MASK) {
              /*
               * If the slab is already in use, somebody can allocate and
               * assign slabobj_exts in parallel. In this case the existing

@@ -2158,6 +2158,9 @@ int alloc_slab_obj_exts(struct slab *slab, struct kmem_cache *s,
              else
                      kfree(vec);
              return 0;
+     } else if (cmpxchg(&slab->obj_exts, old_exts, new_exts) != old_exts) {
+             /* Retry if a racing thread changed slab->obj_exts from under us. */
+             goto retry;
      }
 
      if (allow_spin)