[PATCH] mm/mempolicy: convert mempolicy refcnt to refcount_t

[PATCH] mm/mempolicy: convert mempolicy refcnt to refcount_t

[Gregory Price (Meta)]

struct mempolicy uses an atomic_t as a pure reference counter.

Convert it to refcount_t so that overflow and underflow are caught
by the refcount saturation checks.

This silences the atomic_as_refcounter.cocci warning, which started
matching __mpol_put() when it switched kmem_cache_free() to kfree_rcu().

No functional change for correct callers.

Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202610061946.Bx22ZFrw-lkp@intel.com/
Signed-off-by: Gregory Price (Meta) <gourry@gourry.net>
---
 include/linux/mempolicy.h |  5 +++--
 mm/mempolicy.c            | 14 +++++++-------
 2 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/include/linux/mempolicy.h b/include/linux/mempolicy.h
index 65c732d440d2..8b7570a90e54 100644
--- a/include/linux/mempolicy.h
+++ b/include/linux/mempolicy.h
@@ -10,6 +10,7 @@
 #include <linux/mmzone.h>
 #include <linux/slab.h>
 #include <linux/rbtree.h>
+#include <linux/refcount.h>
 #include <linux/spinlock.h>
 #include <linux/node.h>
 #include <linux/nodemask.h>
@@ -45,7 +46,7 @@ struct mm_struct;
  * to 1, representing the caller of mpol_dup().
  */
 struct mempolicy {
-	atomic_t refcnt;
+	refcount_t refcnt;
 	unsigned short mode; 	/* See MPOL_* above */
 	unsigned short flags;	/* See set_mempolicy() MPOL_F_* above */
 	nodemask_t nodes;	/* interleave/bind/preferred/etc */
@@ -96,7 +97,7 @@ static inline struct mempolicy *mpol_dup(struct mempolicy *pol)
 static inline void mpol_get(struct mempolicy *pol)
 {
 	if (pol)
-		atomic_inc(&pol->refcnt);
+		refcount_inc(&pol->refcnt);
 }
 
 extern bool __mpol_equal(struct mempolicy *a, struct mempolicy *b);
diff --git a/mm/mempolicy.c b/mm/mempolicy.c
index 70298fded1b4..7c78c66f7f21 100644
--- a/mm/mempolicy.c
+++ b/mm/mempolicy.c
@@ -138,7 +138,7 @@ enum zone_type policy_zone = 0;
  * run-time system-wide default policy => local allocation
  */
 static struct mempolicy default_policy = {
-	.refcnt = ATOMIC_INIT(1), /* never free it */
+	.refcnt = REFCOUNT_INIT(1), /* never free it */
 	.mode = MPOL_LOCAL,
 };
 
@@ -492,7 +492,7 @@ static struct mempolicy *mpol_new(unsigned short mode, unsigned short flags,
 	policy = kmem_cache_alloc(policy_cache, GFP_KERNEL);
 	if (!policy)
 		return ERR_PTR(-ENOMEM);
-	atomic_set(&policy->refcnt, 1);
+	refcount_set(&policy->refcnt, 1);
 	policy->mode = mode;
 	policy->flags = flags;
 	policy->home_node = NUMA_NO_NODE;
@@ -503,7 +503,7 @@ static struct mempolicy *mpol_new(unsigned short mode, unsigned short flags,
 /* Slow path of a mpol destructor. */
 void __mpol_put(struct mempolicy *pol)
 {
-	if (!atomic_dec_and_test(&pol->refcnt))
+	if (!refcount_dec_and_test(&pol->refcnt))
 		return;
 	/*
 	 * Required to allow mmap_lock_speculative*() access, see for example
@@ -2871,7 +2871,7 @@ struct mempolicy *__mpol_dup(struct mempolicy *old)
 		nodemask_t mems = cpuset_mems_allowed(current);
 		mpol_rebind_policy(new, &mems);
 	}
-	atomic_set(&new->refcnt, 1);
+	refcount_set(&new->refcnt, 1);
 	return new;
 }
 
@@ -3188,7 +3188,7 @@ static int shared_policy_replace(struct shared_policy *sp, pgoff_t start,
 					goto alloc_new;
 
 				*mpol_new = *n->policy;
-				atomic_set(&mpol_new->refcnt, 1);
+				refcount_set(&mpol_new->refcnt, 1);
 				sp_node_init(n_new, end, n->end, mpol_new);
 				n->end = start;
 				sp_insert(sp, n_new);
@@ -3224,7 +3224,7 @@ static int shared_policy_replace(struct shared_policy *sp, pgoff_t start,
 	mpol_new = kmem_cache_alloc(policy_cache, GFP_KERNEL);
 	if (!mpol_new)
 		goto err_out;
-	atomic_set(&mpol_new->refcnt, 1);
+	refcount_set(&mpol_new->refcnt, 1);
 	goto restart;
 }
 
@@ -3380,7 +3380,7 @@ void __init numa_policy_init(void)
 
 	for_each_node(nid) {
 		preferred_node_policy[nid] = (struct mempolicy) {
-			.refcnt = ATOMIC_INIT(1),
+			.refcnt = REFCOUNT_INIT(1),
 			.mode = MPOL_PREFERRED,
 			.flags = MPOL_F_MOF | MPOL_F_MORON,
 			.nodes = nodemask_of_node(nid),
-- 
2.56.0

Re: [PATCH] mm/mempolicy: convert mempolicy refcnt to refcount_t

[SJ Park]

On Thu, 15 Oct 2026 19:13:22 -0400 "Gregory Price (Meta)" <gourry@gourry.net> wrote:

> struct mempolicy uses an atomic_t as a pure reference counter.
> 
> Convert it to refcount_t so that overflow and underflow are caught
> by the refcount saturation checks.
> 
> This silences the atomic_as_refcounter.cocci warning, which started
> matching __mpol_put() when it switched kmem_cache_free() to kfree_rcu().
> 
> No functional change for correct callers.

Looks correct to me.

> 
> Reported-by: kernel test robot <lkp@intel.com>
> Closes: https://lore.kernel.org/oe-kbuild-all/202610061946.Bx22ZFrw-lkp@intel.com/
> Signed-off-by: Gregory Price (Meta) <gourry@gourry.net>

Reviewed-by: SJ Park <sj@kernel.org>


Thanks,
SJ

[...]

Re: [PATCH] mm/mempolicy: convert mempolicy refcnt to refcount_t

[David Hildenbrand (Arm)]

On 10/16/26 01:13, Gregory Price (Meta) wrote:
> struct mempolicy uses an atomic_t as a pure reference counter.
> 
> Convert it to refcount_t so that overflow and underflow are caught
> by the refcount saturation checks.
> 
> This silences the atomic_as_refcounter.cocci warning, which started
> matching __mpol_put() when it switched kmem_cache_free() to kfree_rcu().
> 
> No functional change for correct callers.
> 
> Reported-by: kernel test robot <lkp@intel.com>
> Closes: https://lore.kernel.org/oe-kbuild-all/202610061946.Bx22ZFrw-lkp@intel.com/
> Signed-off-by: Gregory Price (Meta) <gourry@gourry.net>
> ---

Wondering if we want a Fixes? If this only fixes cocci warnings I guess not.

Acked-by: David Hildenbrand (Arm) <david@kernel.org>

-- 
Cheers,

David

Re: [PATCH] mm/mempolicy: convert mempolicy refcnt to refcount_t

[Gregory Price]

On Fri, Oct 16, 2026 at 09:54:33AM +0200, David Hildenbrand (Arm) wrote:
> On 10/16/26 01:13, Gregory Price (Meta) wrote:
> > struct mempolicy uses an atomic_t as a pure reference counter.
> > 
> > Convert it to refcount_t so that overflow and underflow are caught
> > by the refcount saturation checks.
> > 
> > This silences the atomic_as_refcounter.cocci warning, which started
> > matching __mpol_put() when it switched kmem_cache_free() to kfree_rcu().
> > 
> > No functional change for correct callers.
> > 
> > Reported-by: kernel test robot <lkp@intel.com>
> > Closes: https://lore.kernel.org/oe-kbuild-all/202610061946.Bx22ZFrw-lkp@intel.com/
> > Signed-off-by: Gregory Price (Meta) <gourry@gourry.net>
> > ---
> 
> Wondering if we want a Fixes? If this only fixes cocci warnings I guess not.
>

That was my thought.

> Acked-by: David Hildenbrand (Arm) <david@kernel.org>