[PATCH 0/4] netfilter: nf_tables: dynamic set updates

[PATCH 0/4] netfilter: nf_tables: dynamic set updates

[Patrick McHardy]

The following patches contain a small bug fix for set selection with
timeouts, as well as support for dynamic set updates.

A new "dynset" expression type is added to dynamically add or update
set elements. The name is not my first choice, however due to clashes
with the regular set attributes simply "set" wasn't possible.
Userspace exposes this in a nicer fashion.

As preparation, set size accounting is changed to use an atomic_t
for the number of elements to avoid races, additionally set bindings
are changed to not assume every binding is related to a lookup or map
operation.

Please apply, thanks!


Patrick McHardy (4):
  netfilter: nf_tables: fix set selection when timeouts are requested
  netfilter: nf_tables: prepare set element accounting for async updates
  netfilter: nf_tables: support different set binding types
  netfilter: nf_tables: add support for dynamic set updates

 include/net/netfilter/nf_tables.h        |  23 +++-
 include/net/netfilter/nf_tables_core.h   |   3 +
 include/uapi/linux/netfilter/nf_tables.h |  27 ++++
 net/netfilter/Makefile                   |   2 +-
 net/netfilter/nf_tables_api.c            |  44 ++++---
 net/netfilter/nf_tables_core.c           |   7 +
 net/netfilter/nft_dynset.c               | 218 +++++++++++++++++++++++++++++++
 net/netfilter/nft_hash.c                 |  40 +++++-
 net/netfilter/nft_lookup.c               |   2 +
 9 files changed, 345 insertions(+), 21 deletions(-)
 create mode 100644 net/netfilter/nft_dynset.c

-- 
2.1.0

[PATCH 1/4] netfilter: nf_tables: fix set selection when timeouts are requested

[Patrick McHardy]

The NFT_SET_TIMEOUT flag is ignore in nft_select_set_ops, which may
lead to selection of a set implementation that doesn't actually
support timeouts.

Signed-off-by: Patrick McHardy <kaber@trash.net>
---
 net/netfilter/nf_tables_api.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c
index 3aa92b3..0dab872 100644
--- a/net/netfilter/nf_tables_api.c
+++ b/net/netfilter/nf_tables_api.c
@@ -2159,7 +2159,7 @@ nft_select_set_ops(const struct nlattr * const nla[],
 	features = 0;
 	if (nla[NFTA_SET_FLAGS] != NULL) {
 		features = ntohl(nla_get_be32(nla[NFTA_SET_FLAGS]));
-		features &= NFT_SET_INTERVAL | NFT_SET_MAP;
+		features &= NFT_SET_INTERVAL | NFT_SET_MAP | NFT_SET_TIMEOUT;
 	}
 
 	bops	   = NULL;
-- 
2.1.0

[PATCH 2/4] netfilter: nf_tables: prepare set element accounting for async updates

[Patrick McHardy]

Use atomic operations for the element count to avoid races with async
updates.

To properly handle the transactional semantics during netlink updates,
deleted but not yet committed elements are accounted for seperately and
are treated as being already removed. This means for the duration of
a netlink transaction, the limit might be exceeded by the amount of
elements deleted. Set implementations must be prepared to handle this.

Signed-off-by: Patrick McHardy <kaber@trash.net>
---
 include/net/netfilter/nf_tables.h |  4 +++-
 net/netfilter/nf_tables_api.c     | 21 ++++++++++++---------
 net/netfilter/nft_hash.c          |  3 ++-
 3 files changed, 17 insertions(+), 11 deletions(-)

diff --git a/include/net/netfilter/nf_tables.h b/include/net/netfilter/nf_tables.h
index a785699..7464233 100644
--- a/include/net/netfilter/nf_tables.h
+++ b/include/net/netfilter/nf_tables.h
@@ -258,6 +258,7 @@ void nft_unregister_set(struct nft_set_ops *ops);
  * 	@dtype: data type (verdict or numeric type defined by userspace)
  * 	@size: maximum set size
  * 	@nelems: number of elements
+ * 	@ndeact: number of deactivated elements queued for removal
  * 	@timeout: default timeout value in msecs
  * 	@gc_int: garbage collection interval in msecs
  *	@policy: set parameterization (see enum nft_set_policies)
@@ -275,7 +276,8 @@ struct nft_set {
 	u32				ktype;
 	u32				dtype;
 	u32				size;
-	u32				nelems;
+	atomic_t			nelems;
+	u32				ndeact;
 	u64				timeout;
 	u32				gc_int;
 	u16				policy;
diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c
index 0dab872..27d1bf5 100644
--- a/net/netfilter/nf_tables_api.c
+++ b/net/netfilter/nf_tables_api.c
@@ -3238,9 +3238,6 @@ static int nft_add_set_elem(struct nft_ctx *ctx, struct nft_set *set,
 	u32 flags;
 	int err;
 
-	if (set->size && set->nelems == set->size)
-		return -ENFILE;
-
 	err = nla_parse_nested(nla, NFTA_SET_ELEM_MAX, attr,
 			       nft_set_elem_policy);
 	if (err < 0)
@@ -3391,11 +3388,15 @@ static int nf_tables_newsetelem(struct sock *nlsk, struct sk_buff *skb,
 		return -EBUSY;
 
 	nla_for_each_nested(attr, nla[NFTA_SET_ELEM_LIST_ELEMENTS], rem) {
+		if (set->size &&
+		    !atomic_add_unless(&set->nelems, 1, set->size + set->ndeact))
+			return -ENFILE;
+
 		err = nft_add_set_elem(&ctx, set, attr);
-		if (err < 0)
+		if (err < 0) {
+			atomic_dec(&set->nelems);
 			break;
-
-		set->nelems++;
+		}
 	}
 	return err;
 }
@@ -3477,7 +3478,7 @@ static int nf_tables_delsetelem(struct sock *nlsk, struct sk_buff *skb,
 		if (err < 0)
 			break;
 
-		set->nelems--;
+		set->ndeact++;
 	}
 	return err;
 }
@@ -3810,6 +3811,8 @@ static int nf_tables_commit(struct sk_buff *skb)
 						 &te->elem,
 						 NFT_MSG_DELSETELEM, 0);
 			te->set->ops->remove(te->set, &te->elem);
+			atomic_dec(&te->set->nelems);
+			te->set->ndeact--;
 			break;
 		}
 	}
@@ -3913,16 +3916,16 @@ static int nf_tables_abort(struct sk_buff *skb)
 			nft_trans_destroy(trans);
 			break;
 		case NFT_MSG_NEWSETELEM:
-			nft_trans_elem_set(trans)->nelems--;
 			te = (struct nft_trans_elem *)trans->data;
 
 			te->set->ops->remove(te->set, &te->elem);
+			atomic_dec(&te->set->nelems);
 			break;
 		case NFT_MSG_DELSETELEM:
 			te = (struct nft_trans_elem *)trans->data;
 
-			nft_trans_elem_set(trans)->nelems++;
 			te->set->ops->activate(te->set, &te->elem);
+			te->set->ndeact--;
 
 			nft_trans_destroy(trans);
 			break;
diff --git a/net/netfilter/nft_hash.c b/net/netfilter/nft_hash.c
index 2a00da9..3379fb0 100644
--- a/net/netfilter/nft_hash.c
+++ b/net/netfilter/nft_hash.c
@@ -203,7 +203,7 @@ out:
 
 static void nft_hash_gc(struct work_struct *work)
 {
-	const struct nft_set *set;
+	struct nft_set *set;
 	struct nft_hash_elem *he;
 	struct nft_hash *priv;
 	struct nft_set_gc_batch *gcb = NULL;
@@ -237,6 +237,7 @@ static void nft_hash_gc(struct work_struct *work)
 		if (gcb == NULL)
 			goto out;
 		rhashtable_remove_fast(&priv->ht, &he->node, nft_hash_params);
+		atomic_dec(&set->nelems);
 		nft_set_gc_batch_add(gcb, he);
 	}
 out:
-- 
2.1.0

[PATCH 3/4] netfilter: nf_tables: support different set binding types

[Patrick McHardy]

Currently a set binding is assumed to be related to a lookup and, in
case of maps, a data load.

In order to use bindings for set updates, the loop detection checks
must be restricted to map operations only. Add a flags member to the
binding struct to hold the set "action" flags such as NFT_SET_MAP,
and perform loop detection based on these.

Signed-off-by: Patrick McHardy <kaber@trash.net>
---
 include/net/netfilter/nf_tables.h |  2 ++
 net/netfilter/nf_tables_api.c     | 11 ++++++++---
 net/netfilter/nft_lookup.c        |  2 ++
 3 files changed, 12 insertions(+), 3 deletions(-)

diff --git a/include/net/netfilter/nf_tables.h b/include/net/netfilter/nf_tables.h
index 7464233..e7e6365 100644
--- a/include/net/netfilter/nf_tables.h
+++ b/include/net/netfilter/nf_tables.h
@@ -316,6 +316,7 @@ static inline unsigned long nft_set_gc_interval(const struct nft_set *set)
  *
  *	@list: set bindings list node
  *	@chain: chain containing the rule bound to the set
+ *	@flags: set action flags
  *
  *	A set binding contains all information necessary for validation
  *	of new elements added to a bound set.
@@ -323,6 +324,7 @@ static inline unsigned long nft_set_gc_interval(const struct nft_set *set)
 struct nft_set_binding {
 	struct list_head		list;
 	const struct nft_chain		*chain;
+	u32				flags;
 };
 
 int nf_tables_bind_set(const struct nft_ctx *ctx, struct nft_set *set,
diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c
index 27d1bf5..90b8984 100644
--- a/net/netfilter/nf_tables_api.c
+++ b/net/netfilter/nf_tables_api.c
@@ -2811,12 +2811,13 @@ int nf_tables_bind_set(const struct nft_ctx *ctx, struct nft_set *set,
 	if (!list_empty(&set->bindings) && set->flags & NFT_SET_ANONYMOUS)
 		return -EBUSY;
 
-	if (set->flags & NFT_SET_MAP) {
+	if (binding->flags & NFT_SET_MAP) {
 		/* If the set is already bound to the same chain all
 		 * jumps are already validated for that chain.
 		 */
 		list_for_each_entry(i, &set->bindings, list) {
-			if (i->chain == binding->chain)
+			if (binding->flags & NFT_SET_MAP &&
+			    i->chain == binding->chain)
 				goto bind;
 		}
 
@@ -3312,6 +3313,9 @@ static int nft_add_set_elem(struct nft_ctx *ctx, struct nft_set *set,
 				.chain	= (struct nft_chain *)binding->chain,
 			};
 
+			if (!(binding->flags & NFT_SET_MAP))
+				continue;
+
 			err = nft_validate_data_load(&bind_ctx, dreg,
 						     &data, d2.type);
 			if (err < 0)
@@ -4063,7 +4067,8 @@ static int nf_tables_check_loops(const struct nft_ctx *ctx,
 			continue;
 
 		list_for_each_entry(binding, &set->bindings, list) {
-			if (binding->chain != chain)
+			if (!(binding->flags & NFT_SET_MAP) ||
+			    binding->chain != chain)
 				continue;
 
 			iter.skip 	= 0;
diff --git a/net/netfilter/nft_lookup.c b/net/netfilter/nft_lookup.c
index a5f30b8..d8cf86f 100644
--- a/net/netfilter/nft_lookup.c
+++ b/net/netfilter/nft_lookup.c
@@ -92,6 +92,8 @@ static int nft_lookup_init(const struct nft_ctx *ctx,
 	} else if (set->flags & NFT_SET_MAP)
 		return -EINVAL;
 
+	priv->binding.flags = set->flags & NFT_SET_MAP;
+
 	err = nf_tables_bind_set(ctx, set, &priv->binding);
 	if (err < 0)
 		return err;
-- 
2.1.0

[PATCH 4/4] netfilter: nf_tables: add support for dynamic set updates

[Patrick McHardy]

Add a new "dynset" expression for dynamic set updates.

A new set op ->update() is added which, for non existant elements,
invokes an initialization callback and inserts the new element.
For both new or existing elements the extenstion pointer is returned
to the caller to optionally perform timer updates or other actions.

Element removal is not supported so far, however that seems to be a
rather exotic need and can be added later on.

Signed-off-by: Patrick McHardy <kaber@trash.net>
---
 include/net/netfilter/nf_tables.h        |  17 +++
 include/net/netfilter/nf_tables_core.h   |   3 +
 include/uapi/linux/netfilter/nf_tables.h |  27 ++++
 net/netfilter/Makefile                   |   2 +-
 net/netfilter/nf_tables_api.c            |  10 +-
 net/netfilter/nf_tables_core.c           |   7 +
 net/netfilter/nft_dynset.c               | 218 +++++++++++++++++++++++++++++++
 net/netfilter/nft_hash.c                 |  37 ++++++
 8 files changed, 315 insertions(+), 6 deletions(-)

diff --git a/include/net/netfilter/nf_tables.h b/include/net/netfilter/nf_tables.h
index e7e6365..38c3496 100644
--- a/include/net/netfilter/nf_tables.h
+++ b/include/net/netfilter/nf_tables.h
@@ -196,6 +196,7 @@ struct nft_set_estimate {
 };
 
 struct nft_set_ext;
+struct nft_expr;
 
 /**
  *	struct nft_set_ops - nf_tables set operations
@@ -218,6 +219,15 @@ struct nft_set_ops {
 	bool				(*lookup)(const struct nft_set *set,
 						  const struct nft_data *key,
 						  const struct nft_set_ext **ext);
+	bool				(*update)(struct nft_set *set,
+						  const struct nft_data *key,
+						  void *(*new)(struct nft_set *,
+							       const struct nft_expr *,
+							       struct nft_data []),
+						  const struct nft_expr *expr,
+						  struct nft_data data[],
+						  const struct nft_set_ext **ext);
+
 	int				(*insert)(const struct nft_set *set,
 						  const struct nft_set_elem *elem);
 	void				(*activate)(const struct nft_set *set,
@@ -466,6 +476,11 @@ static inline struct nft_set_ext *nft_set_elem_ext(const struct nft_set *set,
 	return elem + set->ops->elemsize;
 }
 
+void *nft_set_elem_init(const struct nft_set *set,
+			const struct nft_set_ext_tmpl *tmpl,
+			const struct nft_data *key,
+			const struct nft_data *data,
+			u64 timeout, gfp_t gfp);
 void nft_set_elem_destroy(const struct nft_set *set, void *elem);
 
 /**
@@ -845,6 +860,8 @@ static inline u8 nft_genmask_cur(const struct net *net)
 	return 1 << ACCESS_ONCE(net->nft.gencursor);
 }
 
+#define NFT_GENMASK_ANY		((1 << 0) | (1 << 1))
+
 /*
  * Set element transaction helpers
  */
diff --git a/include/net/netfilter/nf_tables_core.h b/include/net/netfilter/nf_tables_core.h
index a75fc8e..c6f400c 100644
--- a/include/net/netfilter/nf_tables_core.h
+++ b/include/net/netfilter/nf_tables_core.h
@@ -31,6 +31,9 @@ void nft_cmp_module_exit(void);
 int nft_lookup_module_init(void);
 void nft_lookup_module_exit(void);
 
+int nft_dynset_module_init(void);
+void nft_dynset_module_exit(void);
+
 int nft_bitwise_module_init(void);
 void nft_bitwise_module_exit(void);
 
diff --git a/include/uapi/linux/netfilter/nf_tables.h b/include/uapi/linux/netfilter/nf_tables.h
index 83441cc..0b87b2f 100644
--- a/include/uapi/linux/netfilter/nf_tables.h
+++ b/include/uapi/linux/netfilter/nf_tables.h
@@ -515,6 +515,33 @@ enum nft_lookup_attributes {
 };
 #define NFTA_LOOKUP_MAX		(__NFTA_LOOKUP_MAX - 1)
 
+enum nft_dynset_ops {
+	NFT_DYNSET_OP_ADD,
+	NFT_DYNSET_OP_UPDATE,
+};
+
+/**
+ * enum nft_dynset_attributes - dynset expression attributes
+ *
+ * @NFTA_DYNSET_SET_NAME: name of set the to add data to (NLA_STRING)
+ * @NFTA_DYNSET_SET_ID: uniquely identifier of the set in the transaction (NLA_U32)
+ * @NFTA_DYNSET_OP: operation (NLA_U32)
+ * @NFTA_DYNSET_SREG_KEY: source register of the key (NLA_U32)
+ * @NFTA_DYNSET_SREG_DATA: source register of the data (NLA_U32)
+ * @NFTA_DYNSET_TIMEOUT: timeout value for the new element (NLA_U64)
+ */
+enum nft_dynset_attributes {
+	NFTA_DYNSET_UNSPEC,
+	NFTA_DYNSET_SET_NAME,
+	NFTA_DYNSET_SET_ID,
+	NFTA_DYNSET_OP,
+	NFTA_DYNSET_SREG_KEY,
+	NFTA_DYNSET_SREG_DATA,
+	NFTA_DYNSET_TIMEOUT,
+	__NFTA_DYNSET_MAX,
+};
+#define NFTA_DYNSET_MAX		(__NFTA_DYNSET_MAX - 1)
+
 /**
  * enum nft_payload_bases - nf_tables payload expression offset bases
  *
diff --git a/net/netfilter/Makefile b/net/netfilter/Makefile
index 89f73a9..a87d8b8 100644
--- a/net/netfilter/Makefile
+++ b/net/netfilter/Makefile
@@ -70,7 +70,7 @@ obj-$(CONFIG_NETFILTER_SYNPROXY) += nf_synproxy_core.o
 
 # nf_tables
 nf_tables-objs += nf_tables_core.o nf_tables_api.o
-nf_tables-objs += nft_immediate.o nft_cmp.o nft_lookup.o
+nf_tables-objs += nft_immediate.o nft_cmp.o nft_lookup.o nft_dynset.o
 nf_tables-objs += nft_bitwise.o nft_byteorder.o nft_payload.o
 
 obj-$(CONFIG_NF_TABLES)		+= nf_tables.o
diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c
index 90b8984..598e53e 100644
--- a/net/netfilter/nf_tables_api.c
+++ b/net/netfilter/nf_tables_api.c
@@ -3183,11 +3183,11 @@ static struct nft_trans *nft_trans_elem_alloc(struct nft_ctx *ctx,
 	return trans;
 }
 
-static void *nft_set_elem_init(const struct nft_set *set,
-			       const struct nft_set_ext_tmpl *tmpl,
-			       const struct nft_data *key,
-			       const struct nft_data *data,
-			       u64 timeout, gfp_t gfp)
+void *nft_set_elem_init(const struct nft_set *set,
+			const struct nft_set_ext_tmpl *tmpl,
+			const struct nft_data *key,
+			const struct nft_data *data,
+			u64 timeout, gfp_t gfp)
 {
 	struct nft_set_ext *ext;
 	void *elem;
diff --git a/net/netfilter/nf_tables_core.c b/net/netfilter/nf_tables_core.c
index ef4dfcb..7caf08a 100644
--- a/net/netfilter/nf_tables_core.c
+++ b/net/netfilter/nf_tables_core.c
@@ -239,8 +239,14 @@ int __init nf_tables_core_module_init(void)
 	if (err < 0)
 		goto err6;
 
+	err = nft_dynset_module_init();
+	if (err < 0)
+		goto err7;
+
 	return 0;
 
+err7:
+	nft_payload_module_exit();
 err6:
 	nft_byteorder_module_exit();
 err5:
@@ -257,6 +263,7 @@ err1:
 
 void nf_tables_core_module_exit(void)
 {
+	nft_dynset_module_exit();
 	nft_payload_module_exit();
 	nft_byteorder_module_exit();
 	nft_bitwise_module_exit();
diff --git a/net/netfilter/nft_dynset.c b/net/netfilter/nft_dynset.c
new file mode 100644
index 0000000..eeb72de
--- /dev/null
+++ b/net/netfilter/nft_dynset.c
@@ -0,0 +1,218 @@
+/*
+ * Copyright (c) 2015 Patrick McHardy <kaber@trash.net>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/netlink.h>
+#include <linux/netfilter.h>
+#include <linux/netfilter/nf_tables.h>
+#include <net/netfilter/nf_tables.h>
+#include <net/netfilter/nf_tables_core.h>
+
+struct nft_dynset {
+	struct nft_set			*set;
+	struct nft_set_ext_tmpl		tmpl;
+	enum nft_dynset_ops		op:8;
+	enum nft_registers		sreg_key:8;
+	enum nft_registers		sreg_data:8;
+	u64				timeout;
+	struct nft_set_binding		binding;
+};
+
+static void *nft_dynset_new(struct nft_set *set, const struct nft_expr *expr,
+			    struct nft_data data[NFT_REG_MAX + 1])
+{
+	const struct nft_dynset *priv = nft_expr_priv(expr);
+	u64 timeout;
+	void *elem;
+
+	if (set->size && !atomic_add_unless(&set->nelems, 1, set->size))
+		return NULL;
+
+	timeout = priv->timeout ? : set->timeout;
+	elem = nft_set_elem_init(set, &priv->tmpl,
+				 &data[priv->sreg_key], &data[priv->sreg_data],
+				 timeout, GFP_ATOMIC);
+	if (elem == NULL) {
+		if (set->size)
+			atomic_dec(&set->nelems);
+	}
+	return elem;
+}
+
+static void nft_dynset_eval(const struct nft_expr *expr,
+			    struct nft_data data[NFT_REG_MAX + 1],
+			    const struct nft_pktinfo *pkt)
+{
+	const struct nft_dynset *priv = nft_expr_priv(expr);
+	struct nft_set *set = priv->set;
+	const struct nft_set_ext *ext;
+	u64 timeout;
+
+	if (set->ops->update(set, &data[priv->sreg_key], nft_dynset_new,
+			     expr, data, &ext)) {
+		if (priv->op == NFT_DYNSET_OP_UPDATE &&
+		    nft_set_ext_exists(ext, NFT_SET_EXT_EXPIRATION)) {
+			timeout = priv->timeout ? : set->timeout;
+			*nft_set_ext_expiration(ext) = jiffies + timeout;
+			return;
+		}
+	}
+
+	data[NFT_REG_VERDICT].verdict = NFT_BREAK;
+}
+
+static const struct nla_policy nft_dynset_policy[NFTA_DYNSET_MAX + 1] = {
+	[NFTA_DYNSET_SET_NAME]	= { .type = NLA_STRING },
+	[NFTA_DYNSET_SET_ID]	= { .type = NLA_U32 },
+	[NFTA_DYNSET_OP]	= { .type = NLA_U32 },
+	[NFTA_DYNSET_SREG_KEY]	= { .type = NLA_U32 },
+	[NFTA_DYNSET_SREG_DATA]	= { .type = NLA_U32 },
+	[NFTA_DYNSET_TIMEOUT]	= { .type = NLA_U64 },
+};
+
+static int nft_dynset_init(const struct nft_ctx *ctx,
+			   const struct nft_expr *expr,
+			   const struct nlattr * const tb[])
+{
+	struct nft_dynset *priv = nft_expr_priv(expr);
+	struct nft_set *set;
+	u64 timeout;
+	int err;
+
+	if (tb[NFTA_DYNSET_SET_NAME] == NULL ||
+	    tb[NFTA_DYNSET_OP] == NULL ||
+	    tb[NFTA_DYNSET_SREG_KEY] == NULL)
+		return -EINVAL;
+
+	set = nf_tables_set_lookup(ctx->table, tb[NFTA_DYNSET_SET_NAME]);
+	if (IS_ERR(set)) {
+		if (tb[NFTA_DYNSET_SET_ID])
+			set = nf_tables_set_lookup_byid(ctx->net,
+							tb[NFTA_DYNSET_SET_ID]);
+		if (IS_ERR(set))
+			return PTR_ERR(set);
+	}
+
+	if (set->flags & NFT_SET_CONSTANT)
+		return -EBUSY;
+
+	priv->op = ntohl(nla_get_be32(tb[NFTA_DYNSET_OP]));
+	switch (priv->op) {
+	case NFT_DYNSET_OP_ADD:
+		break;
+	case NFT_DYNSET_OP_UPDATE:
+		if (!(set->flags & NFT_SET_TIMEOUT))
+			return -EOPNOTSUPP;
+		break;
+	default:
+		return -EOPNOTSUPP;
+	}
+
+	timeout = 0;
+	if (tb[NFTA_DYNSET_TIMEOUT] != NULL) {
+		if (!(set->flags & NFT_SET_TIMEOUT))
+			return -EINVAL;
+		timeout = be64_to_cpu(nla_get_be64(tb[NFTA_DYNSET_TIMEOUT]));
+	}
+
+	priv->sreg_key = ntohl(nla_get_be32(tb[NFTA_DYNSET_SREG_KEY]));
+	err = nft_validate_input_register(priv->sreg_key);
+	if (err < 0)
+		return err;
+
+	if (tb[NFTA_DYNSET_SREG_DATA] != NULL) {
+		if (!(set->flags & NFT_SET_MAP))
+			return -EINVAL;
+		if (set->dtype == NFT_DATA_VERDICT)
+			return -EOPNOTSUPP;
+
+		priv->sreg_data = ntohl(nla_get_be32(tb[NFTA_DYNSET_SREG_DATA]));
+		err = nft_validate_input_register(priv->sreg_data);
+		if (err < 0)
+			return err;
+	} else if (set->flags & NFT_SET_MAP)
+		return -EINVAL;
+
+	nft_set_ext_prepare(&priv->tmpl);
+	nft_set_ext_add_length(&priv->tmpl, NFT_SET_EXT_KEY, set->klen);
+	if (set->flags & NFT_SET_MAP)
+		nft_set_ext_add_length(&priv->tmpl, NFT_SET_EXT_DATA, set->dlen);
+	if (set->flags & NFT_SET_TIMEOUT) {
+		if (timeout || set->timeout)
+			nft_set_ext_add(&priv->tmpl, NFT_SET_EXT_EXPIRATION);
+	}
+
+	priv->timeout = timeout;
+
+	err = nf_tables_bind_set(ctx, set, &priv->binding);
+	if (err < 0)
+		return err;
+
+	priv->set = set;
+	return 0;
+}
+
+static void nft_dynset_destroy(const struct nft_ctx *ctx,
+			       const struct nft_expr *expr)
+{
+	struct nft_dynset *priv = nft_expr_priv(expr);
+
+	nf_tables_unbind_set(ctx, priv->set, &priv->binding);
+}
+
+static int nft_dynset_dump(struct sk_buff *skb, const struct nft_expr *expr)
+{
+	const struct nft_dynset *priv = nft_expr_priv(expr);
+
+	if (nla_put_be32(skb, NFTA_DYNSET_SREG_KEY, htonl(priv->sreg_key)))
+		goto nla_put_failure;
+	if (priv->set->flags & NFT_SET_MAP &&
+	    nla_put_be32(skb, NFTA_DYNSET_SREG_DATA, htonl(priv->sreg_data)))
+		goto nla_put_failure;
+	if (nla_put_be32(skb, NFTA_DYNSET_OP, htonl(priv->op)))
+		goto nla_put_failure;
+	if (nla_put_string(skb, NFTA_DYNSET_SET_NAME, priv->set->name))
+		goto nla_put_failure;
+	if (nla_put_be64(skb, NFTA_DYNSET_TIMEOUT, cpu_to_be64(priv->timeout)))
+		goto nla_put_failure;
+	return 0;
+
+nla_put_failure:
+	return -1;
+}
+
+static struct nft_expr_type nft_dynset_type;
+static const struct nft_expr_ops nft_dynset_ops = {
+	.type		= &nft_dynset_type,
+	.size		= NFT_EXPR_SIZE(sizeof(struct nft_dynset)),
+	.eval		= nft_dynset_eval,
+	.init		= nft_dynset_init,
+	.destroy	= nft_dynset_destroy,
+	.dump		= nft_dynset_dump,
+};
+
+static struct nft_expr_type nft_dynset_type __read_mostly = {
+	.name		= "dynset",
+	.ops		= &nft_dynset_ops,
+	.policy		= nft_dynset_policy,
+	.maxattr	= NFTA_DYNSET_MAX,
+	.owner		= THIS_MODULE,
+};
+
+int __init nft_dynset_module_init(void)
+{
+	return nft_register_expr(&nft_dynset_type);
+}
+
+void nft_dynset_module_exit(void)
+{
+	nft_unregister_expr(&nft_dynset_type);
+}
diff --git a/net/netfilter/nft_hash.c b/net/netfilter/nft_hash.c
index 3379fb0..a91194e 100644
--- a/net/netfilter/nft_hash.c
+++ b/net/netfilter/nft_hash.c
@@ -90,6 +90,42 @@ static bool nft_hash_lookup(const struct nft_set *set,
 	return !!he;
 }
 
+static bool nft_hash_update(struct nft_set *set, const struct nft_data *key,
+			    void *(*new)(struct nft_set *,
+					 const struct nft_expr *,
+					 struct nft_data []),
+			    const struct nft_expr *expr,
+			    struct nft_data data[],
+			    const struct nft_set_ext **ext)
+{
+	struct nft_hash *priv = nft_set_priv(set);
+	struct nft_hash_elem *he;
+	struct nft_hash_cmp_arg arg = {
+		.genmask = NFT_GENMASK_ANY,
+		.set	 = set,
+		.key	 = key,
+	};
+
+	he = rhashtable_lookup_fast(&priv->ht, &arg, nft_hash_params);
+	if (he != NULL)
+		goto out;
+
+	he = new(set, expr, data);
+	if (he == NULL)
+		goto err1;
+	if (rhashtable_lookup_insert_key(&priv->ht, &arg, &he->node,
+					 nft_hash_params))
+		goto err2;
+out:
+	*ext = &he->ext;
+	return true;
+
+err2:
+	nft_set_elem_destroy(set, he);
+err1:
+	return false;
+}
+
 static int nft_hash_insert(const struct nft_set *set,
 			   const struct nft_set_elem *elem)
 {
@@ -335,6 +371,7 @@ static struct nft_set_ops nft_hash_ops __read_mostly = {
 	.deactivate	= nft_hash_deactivate,
 	.remove		= nft_hash_remove,
 	.lookup		= nft_hash_lookup,
+	.update		= nft_hash_update,
 	.walk		= nft_hash_walk,
 	.features	= NFT_SET_MAP | NFT_SET_TIMEOUT,
 	.owner		= THIS_MODULE,
-- 
2.1.0

Re: [PATCH 0/4] netfilter: nf_tables: dynamic set updates

[Pablo Neira Ayuso]

On Sun, Apr 05, 2015 at 02:41:04PM +0200, Patrick McHardy wrote:
> The following patches contain a small bug fix for set selection with
> timeouts, as well as support for dynamic set updates.
> 
> A new "dynset" expression type is added to dynamically add or update
> set elements. The name is not my first choice, however due to clashes
> with the regular set attributes simply "set" wasn't possible.
> Userspace exposes this in a nicer fashion.
> 
> As preparation, set size accounting is changed to use an atomic_t
> for the number of elements to avoid races, additionally set bindings
> are changed to not assume every binding is related to a lookup or map
> operation.

Series applied, thanks Patrick.