Merge tag 'nf-26-03-05' of https://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf

Florian Westphal says:

====================
netfilter: updates for net

1) Inseo An reported a bug with the set element handling in nf_tables:
   When set cannot accept more elements, we unlink and immediately free
   an element that was inserted into a public data structure, freeing it
   without waiting for RCU grace period.  Fix this by doing the
   increment earlier and by deferring possible unlink-and-free to the
   existing abort path, which performs the needed synchronize_rcu before
   free.  From Pablo Neira Ayuso. This is an ancient bug, dating back to
   kernel 4.10.

2) syzbot reported WARN_ON() splat in nf_tables that occurs on memory
   allocation failure.  Fix this by a new iterator annotation:
   The affected walker does not need to clone the data structure and
   can just use the live version if no clone exists yet.
   Also from Pablo.  This bug existed since 6.10 days.

3) Ancient forever bug in nft_pipapo data structure:
   The garbage collection logic to remove expired elements is broken.
   We must unlink from data structure and can only hand the freeing
   to call_rcu after the clone/live pointers of the data structures
   have been swapped.  Else, readers can observe the free'd element.
   Reported by Yiming Qian.

* tag 'nf-26-03-05' of https://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
  netfilter: nft_set_pipapo: split gc into unlink and reclaim phase
  netfilter: nf_tables: clone set on flush only
  netfilter: nf_tables: unconditionally bump set->nelems before insertion
====================

Link: https://patch.msgid.link/20260305122635.23525-1-fw@strlen.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

6 files changed, 92 insertions, 33 deletions

diff --git a/include/net/netfilter/nf_tables.h b/include/net/netfilter/nf_tables.h
index 426534a711b0..e2d2bfc1f989 100644
--- a/include/net/netfilter/nf_tables.h
+++ b/include/net/netfilter/nf_tables.h
@@ -320,11 +320,13 @@ static inline void *nft_elem_priv_cast(const struct nft_elem_priv *priv)
  * @NFT_ITER_UNSPEC: unspecified, to catch errors
  * @NFT_ITER_READ: read-only iteration over set elements
  * @NFT_ITER_UPDATE: iteration under mutex to update set element state
+ * @NFT_ITER_UPDATE_CLONE: clone set before iteration under mutex to update element
  */
 enum nft_iter_type {
 	NFT_ITER_UNSPEC,
 	NFT_ITER_READ,
 	NFT_ITER_UPDATE,
+	NFT_ITER_UPDATE_CLONE,
 };
 
 struct nft_set;
@@ -1861,6 +1863,11 @@ struct nft_trans_gc {
 	struct rcu_head		rcu;
 };
 
+static inline int nft_trans_gc_space(const struct nft_trans_gc *trans)
+{
+	return NFT_TRANS_GC_BATCHCOUNT - trans->count;
+}
+
 static inline void nft_ctx_update(struct nft_ctx *ctx,
 				  const struct nft_trans *trans)
 {
diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c
index fd7f7e4e2a43..1862bd7fe804 100644
--- a/net/netfilter/nf_tables_api.c
+++ b/net/netfilter/nf_tables_api.c
@@ -833,6 +833,11 @@ static void nft_map_catchall_deactivate(const struct nft_ctx *ctx,
 	}
 }
 
+/* Use NFT_ITER_UPDATE iterator even if this may be called from the preparation
+ * phase, the set clone might already exist from a previous command, or it might
+ * be a set that is going away and does not require a clone. The netns and
+ * netlink release paths also need to work on the live set.
+ */
 static void nft_map_deactivate(const struct nft_ctx *ctx, struct nft_set *set)
 {
 	struct nft_set_iter iter = {
@@ -7170,6 +7175,7 @@ static int nft_add_set_elem(struct nft_ctx *ctx, struct nft_set *set,
 	struct nft_data_desc desc;
 	enum nft_registers dreg;
 	struct nft_trans *trans;
+	bool set_full = false;
 	u64 expiration;
 	u64 timeout;
 	int err, i;
@@ -7461,10 +7467,18 @@ static int nft_add_set_elem(struct nft_ctx *ctx, struct nft_set *set,
 	if (err < 0)
 		goto err_elem_free;
 
+	if (!(flags & NFT_SET_ELEM_CATCHALL)) {
+		unsigned int max = nft_set_maxsize(set), nelems;
+
+		nelems = atomic_inc_return(&set->nelems);
+		if (nelems > max)
+			set_full = true;
+	}
+
 	trans = nft_trans_elem_alloc(ctx, NFT_MSG_NEWSETELEM, set);
 	if (trans == NULL) {
 		err = -ENOMEM;
-		goto err_elem_free;
+		goto err_set_size;
 	}
 
 	ext->genmask = nft_genmask_cur(ctx->net);
@@ -7516,7 +7530,7 @@ static int nft_add_set_elem(struct nft_ctx *ctx, struct nft_set *set,
 
 						ue->priv = elem_priv;
 						nft_trans_commit_list_add_elem(ctx->net, trans);
-						goto err_elem_free;
+						goto err_set_size;
 					}
 				}
 			}
@@ -7534,23 +7548,16 @@ static int nft_add_set_elem(struct nft_ctx *ctx, struct nft_set *set,
 		goto err_element_clash;
 	}
 
-	if (!(flags & NFT_SET_ELEM_CATCHALL)) {
-		unsigned int max = nft_set_maxsize(set);
-
-		if (!atomic_add_unless(&set->nelems, 1, max)) {
-			err = -ENFILE;
-			goto err_set_full;
-		}
-	}
-
 	nft_trans_container_elem(trans)->elems[0].priv = elem.priv;
 	nft_trans_commit_list_add_elem(ctx->net, trans);
-	return 0;
 
-err_set_full:
-	nft_setelem_remove(ctx->net, set, elem.priv);
+	return set_full ? -ENFILE : 0;
+
 err_element_clash:
 	kfree(trans);
+err_set_size:
+	if (!(flags & NFT_SET_ELEM_CATCHALL))
+		atomic_dec(&set->nelems);
 err_elem_free:
 	nf_tables_set_elem_destroy(ctx, set, elem.priv);
 err_parse_data:
@@ -7901,9 +7908,12 @@ static int nft_set_catchall_flush(const struct nft_ctx *ctx,
 
 static int nft_set_flush(struct nft_ctx *ctx, struct nft_set *set, u8 genmask)
 {
+	/* The set backend might need to clone the set, do it now from the
+	 * preparation phase, use NFT_ITER_UPDATE_CLONE iterator type.
+	 */
 	struct nft_set_iter iter = {
 		.genmask	= genmask,
-		.type		= NFT_ITER_UPDATE,
+		.type		= NFT_ITER_UPDATE_CLONE,
 		.fn		= nft_setelem_flush,
 	};
 
@@ -10483,11 +10493,6 @@ static void nft_trans_gc_queue_work(struct nft_trans_gc *trans)
 	schedule_work(&trans_gc_work);
 }
 
-static int nft_trans_gc_space(struct nft_trans_gc *trans)
-{
-	return NFT_TRANS_GC_BATCHCOUNT - trans->count;
-}
-
 struct nft_trans_gc *nft_trans_gc_queue_async(struct nft_trans_gc *gc,
 					      unsigned int gc_seq, gfp_t gfp)
 {
diff --git a/net/netfilter/nft_set_hash.c b/net/netfilter/nft_set_hash.c
index 739b992bde59..b0e571c8e3f3 100644
--- a/net/netfilter/nft_set_hash.c
+++ b/net/netfilter/nft_set_hash.c
@@ -374,6 +374,7 @@ static void nft_rhash_walk(const struct nft_ctx *ctx, struct nft_set *set,
 {
 	switch (iter->type) {
 	case NFT_ITER_UPDATE:
+	case NFT_ITER_UPDATE_CLONE:
 		/* only relevant for netlink dumps which use READ type */
 		WARN_ON_ONCE(iter->skip != 0);
 
diff --git a/net/netfilter/nft_set_pipapo.c b/net/netfilter/nft_set_pipapo.c
index 7ef4b44471d3..a34632ae6048 100644
--- a/net/netfilter/nft_set_pipapo.c
+++ b/net/netfilter/nft_set_pipapo.c
@@ -1680,11 +1680,11 @@ static void nft_pipapo_gc_deactivate(struct net *net, struct nft_set *set,
 }
 
 /**
- * pipapo_gc() - Drop expired entries from set, destroy start and end elements
+ * pipapo_gc_scan() - Drop expired entries from set and link them to gc list
  * @set:	nftables API set representation
  * @m:		Matching data
  */
-static void pipapo_gc(struct nft_set *set, struct nft_pipapo_match *m)
+static void pipapo_gc_scan(struct nft_set *set, struct nft_pipapo_match *m)
 {
 	struct nft_pipapo *priv = nft_set_priv(set);
 	struct net *net = read_pnet(&set->net);
@@ -1697,6 +1697,8 @@ static void pipapo_gc(struct nft_set *set, struct nft_pipapo_match *m)
 	if (!gc)
 		return;
 
+	list_add(&gc->list, &priv->gc_head);
+
 	while ((rules_f0 = pipapo_rules_same_key(m->f, first_rule))) {
 		union nft_pipapo_map_bucket rulemap[NFT_PIPAPO_MAX_FIELDS];
 		const struct nft_pipapo_field *f;
@@ -1724,9 +1726,13 @@ static void pipapo_gc(struct nft_set *set, struct nft_pipapo_match *m)
 		 * NFT_SET_ELEM_DEAD_BIT.
 		 */
 		if (__nft_set_elem_expired(&e->ext, tstamp)) {
-			gc = nft_trans_gc_queue_sync(gc, GFP_KERNEL);
-			if (!gc)
-				return;
+			if (!nft_trans_gc_space(gc)) {
+				gc = nft_trans_gc_alloc(set, 0, GFP_KERNEL);
+				if (!gc)
+					return;
+
+				list_add(&gc->list, &priv->gc_head);
+			}
 
 			nft_pipapo_gc_deactivate(net, set, e);
 			pipapo_drop(m, rulemap);
@@ -1740,10 +1746,30 @@ static void pipapo_gc(struct nft_set *set, struct nft_pipapo_match *m)
 		}
 	}
 
-	gc = nft_trans_gc_catchall_sync(gc);
+	priv->last_gc = jiffies;
+}
+
+/**
+ * pipapo_gc_queue() - Free expired elements
+ * @set:	nftables API set representation
+ */
+static void pipapo_gc_queue(struct nft_set *set)
+{
+	struct nft_pipapo *priv = nft_set_priv(set);
+	struct nft_trans_gc *gc, *next;
+
+	/* always do a catchall cycle: */
+	gc = nft_trans_gc_alloc(set, 0, GFP_KERNEL);
 	if (gc) {
+		gc = nft_trans_gc_catchall_sync(gc);
+		if (gc)
+			nft_trans_gc_queue_sync_done(gc);
+	}
+
+	/* always purge queued gc elements. */
+	list_for_each_entry_safe(gc, next, &priv->gc_head, list) {
+		list_del(&gc->list);
 		nft_trans_gc_queue_sync_done(gc);
-		priv->last_gc = jiffies;
 	}
 }
 
@@ -1797,6 +1823,10 @@ static void pipapo_reclaim_match(struct rcu_head *rcu)
  *
  * We also need to create a new working copy for subsequent insertions and
  * deletions.
+ *
+ * After the live copy has been replaced by the clone, we can safely queue
+ * expired elements that have been collected by pipapo_gc_scan() for
+ * memory reclaim.
  */
 static void nft_pipapo_commit(struct nft_set *set)
 {
@@ -1807,7 +1837,7 @@ static void nft_pipapo_commit(struct nft_set *set)
 		return;
 
 	if (time_after_eq(jiffies, priv->last_gc + nft_set_gc_interval(set)))
-		pipapo_gc(set, priv->clone);
+		pipapo_gc_scan(set, priv->clone);
 
 	old = rcu_replace_pointer(priv->match, priv->clone,
 				  nft_pipapo_transaction_mutex_held(set));
@@ -1815,6 +1845,8 @@ static void nft_pipapo_commit(struct nft_set *set)
 
 	if (old)
 		call_rcu(&old->rcu, pipapo_reclaim_match);
+
+	pipapo_gc_queue(set);
 }
 
 static void nft_pipapo_abort(const struct nft_set *set)
@@ -2144,13 +2176,20 @@ static void nft_pipapo_walk(const struct nft_ctx *ctx, struct nft_set *set,
 	const struct nft_pipapo_match *m;
 
 	switch (iter->type) {
-	case NFT_ITER_UPDATE:
+	case NFT_ITER_UPDATE_CLONE:
 		m = pipapo_maybe_clone(set);
 		if (!m) {
 			iter->err = -ENOMEM;
 			return;
 		}
-
+		nft_pipapo_do_walk(ctx, set, m, iter);
+		break;
+	case NFT_ITER_UPDATE:
+		if (priv->clone)
+			m = priv->clone;
+		else
+			m = rcu_dereference_protected(priv->match,
+						      nft_pipapo_transaction_mutex_held(set));
 		nft_pipapo_do_walk(ctx, set, m, iter);
 		break;
 	case NFT_ITER_READ:
@@ -2272,6 +2311,7 @@ static int nft_pipapo_init(const struct nft_set *set,
 		f->mt = NULL;
 	}
 
+	INIT_LIST_HEAD(&priv->gc_head);
 	rcu_assign_pointer(priv->match, m);
 
 	return 0;
@@ -2321,6 +2361,8 @@ static void nft_pipapo_destroy(const struct nft_ctx *ctx,
 	struct nft_pipapo *priv = nft_set_priv(set);
 	struct nft_pipapo_match *m;
 
+	WARN_ON_ONCE(!list_empty(&priv->gc_head));
+
 	m = rcu_dereference_protected(priv->match, true);
 
 	if (priv->clone) {
diff --git a/net/netfilter/nft_set_pipapo.h b/net/netfilter/nft_set_pipapo.h
index eaab422aa56a..9aee9a9eaeb7 100644
--- a/net/netfilter/nft_set_pipapo.h
+++ b/net/netfilter/nft_set_pipapo.h
@@ -156,12 +156,14 @@ struct nft_pipapo_match {
  * @clone:	Copy where pending insertions and deletions are kept
  * @width:	Total bytes to be matched for one packet, including padding
  * @last_gc:	Timestamp of last garbage collection run, jiffies
+ * @gc_head:	list of nft_trans_gc to queue up for mem reclaim
  */
 struct nft_pipapo {
 	struct nft_pipapo_match __rcu *match;
 	struct nft_pipapo_match *clone;
 	int width;
 	unsigned long last_gc;
+	struct list_head gc_head;
 };
 
 struct nft_pipapo_elem;
diff --git a/net/netfilter/nft_set_rbtree.c b/net/netfilter/nft_set_rbtree.c
index 3f02e4478216..ee3d4f5b9ff7 100644
--- a/net/netfilter/nft_set_rbtree.c
+++ b/net/netfilter/nft_set_rbtree.c
@@ -861,13 +861,15 @@ static void nft_rbtree_walk(const struct nft_ctx *ctx,
 	struct nft_rbtree *priv = nft_set_priv(set);
 
 	switch (iter->type) {
-	case NFT_ITER_UPDATE:
-		lockdep_assert_held(&nft_pernet(ctx->net)->commit_mutex);
-
+	case NFT_ITER_UPDATE_CLONE:
 		if (nft_array_may_resize(set) < 0) {
 			iter->err = -ENOMEM;
 			break;
 		}
+		fallthrough;
+	case NFT_ITER_UPDATE:
+		lockdep_assert_held(&nft_pernet(ctx->net)->commit_mutex);
+
 		nft_rbtree_do_walk(ctx, set, iter);
 		break;
 	case NFT_ITER_READ: