From 68983a354a655c35d3fb204489d383a2a051fda7 Mon Sep 17 00:00:00 2001
From: Manoj Basapathi <manojbm@codeaurora.org>
Date: Thu, 6 Feb 2020 16:37:29 +0530
Subject: netfilter: xtables: Add snapshot of hardidletimer target

This is a snapshot of hardidletimer netfilter target.

This patch implements a hardidletimer Xtables target that can be
used to identify when interfaces have been idle for a certain period
of time.

Timers are identified by labels and are created when a rule is set
with a new label. The rules also take a timeout value (in seconds) as
an option. If more than one rule uses the same timer label, the timer
will be restarted whenever any of the rules get a hit.

One entry for each timer is created in sysfs. This attribute contains
the timer remaining for the timer to expire. The attributes are
located under the xt_idletimer class:

/sys/class/xt_idletimer/timers/<label>

When the timer expires, the target module sends a sysfs notification
to the userspace, which can then decide what to do (eg. disconnect to
save power)

Compared to IDLETIMER, HARDIDLETIMER can send notifications when
CPU is in suspend too, to notify the timer expiry.

v1->v2: Moved all functionality into IDLETIMER module to avoid
code duplication per comment from Florian.

Signed-off-by: Manoj Basapathi <manojbm@codeaurora.org>
Signed-off-by: Subash Abhinov Kasiviswanathan <subashab@codeaurora.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
 include/uapi/linux/netfilter/xt_IDLETIMER.h | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

(limited to 'include')

diff --git a/include/uapi/linux/netfilter/xt_IDLETIMER.h b/include/uapi/linux/netfilter/xt_IDLETIMER.h
index 3c586a19baea..434e6506abaa 100644
--- a/include/uapi/linux/netfilter/xt_IDLETIMER.h
+++ b/include/uapi/linux/netfilter/xt_IDLETIMER.h
@@ -1,4 +1,3 @@
-/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
 /*
  * linux/include/linux/netfilter/xt_IDLETIMER.h
  *
@@ -33,6 +32,7 @@
 #include <linux/types.h>
 
 #define MAX_IDLETIMER_LABEL_SIZE 28
+#define XT_IDLETIMER_ALARM 0x01
 
 struct idletimer_tg_info {
 	__u32 timeout;
@@ -43,4 +43,14 @@ struct idletimer_tg_info {
 	struct idletimer_tg *timer __attribute__((aligned(8)));
 };
 
+struct idletimer_tg_info_v1 {
+	__u32 timeout;
+
+	char label[MAX_IDLETIMER_LABEL_SIZE];
+
+	__u8 timer_type;
+
+	/* for kernel module internal use only */
+	struct idletimer_tg *timer __attribute__((aligned(8)));
+};
 #endif
-- 
cgit v1.2.3-59-g8ed1b


From 925d844696d9287f841d6b3e0ed62a35fb175970 Mon Sep 17 00:00:00 2001
From: Xin Long <lucien.xin@gmail.com>
Date: Mon, 10 Feb 2020 13:41:22 +0800
Subject: netfilter: nft_tunnel: add support for geneve opts

Like vxlan and erspan opts, geneve opts should also be supported in
nft_tunnel. The difference is geneve RFC (draft-ietf-nvo3-geneve-14)
allows a geneve packet to carry multiple geneve opts. So with this
patch, nftables/libnftnl would do:

  # nft add table ip filter
  # nft add chain ip filter input { type filter hook input priority 0 \; }
  # nft add tunnel filter geneve_02 { type geneve\; id 2\; \
    ip saddr 192.168.1.1\; ip daddr 192.168.1.2\; \
    sport 9000\; dport 9001\; dscp 1234\; ttl 64\; flags 1\; \
    opts \"1:1:34567890,2:2:12121212,3:3:1212121234567890\"\; }
  # nft list tunnels table filter
    table ip filter {
    	tunnel geneve_02 {
    		id 2
    		ip saddr 192.168.1.1
    		ip daddr 192.168.1.2
    		sport 9000
    		dport 9001
    		tos 18
    		ttl 64
    		flags 1
    		geneve opts 1:1:34567890,2:2:12121212,3:3:1212121234567890
    	}
    }

v1->v2:
  - no changes, just post it separately.

Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
 include/uapi/linux/netfilter/nf_tables.h |  10 +++
 net/netfilter/nft_tunnel.c               | 110 +++++++++++++++++++++++++++----
 2 files changed, 108 insertions(+), 12 deletions(-)

(limited to 'include')

diff --git a/include/uapi/linux/netfilter/nf_tables.h b/include/uapi/linux/netfilter/nf_tables.h
index 065218a20bb7..9c3d2d04d6a1 100644
--- a/include/uapi/linux/netfilter/nf_tables.h
+++ b/include/uapi/linux/netfilter/nf_tables.h
@@ -1770,6 +1770,7 @@ enum nft_tunnel_opts_attributes {
 	NFTA_TUNNEL_KEY_OPTS_UNSPEC,
 	NFTA_TUNNEL_KEY_OPTS_VXLAN,
 	NFTA_TUNNEL_KEY_OPTS_ERSPAN,
+	NFTA_TUNNEL_KEY_OPTS_GENEVE,
 	__NFTA_TUNNEL_KEY_OPTS_MAX
 };
 #define NFTA_TUNNEL_KEY_OPTS_MAX	(__NFTA_TUNNEL_KEY_OPTS_MAX - 1)
@@ -1791,6 +1792,15 @@ enum nft_tunnel_opts_erspan_attributes {
 };
 #define NFTA_TUNNEL_KEY_ERSPAN_MAX	(__NFTA_TUNNEL_KEY_ERSPAN_MAX - 1)
 
+enum nft_tunnel_opts_geneve_attributes {
+	NFTA_TUNNEL_KEY_GENEVE_UNSPEC,
+	NFTA_TUNNEL_KEY_GENEVE_CLASS,
+	NFTA_TUNNEL_KEY_GENEVE_TYPE,
+	NFTA_TUNNEL_KEY_GENEVE_DATA,
+	__NFTA_TUNNEL_KEY_GENEVE_MAX
+};
+#define NFTA_TUNNEL_KEY_GENEVE_MAX	(__NFTA_TUNNEL_KEY_GENEVE_MAX - 1)
+
 enum nft_tunnel_flags {
 	NFT_TUNNEL_F_ZERO_CSUM_TX	= (1 << 0),
 	NFT_TUNNEL_F_DONT_FRAGMENT	= (1 << 1),
diff --git a/net/netfilter/nft_tunnel.c b/net/netfilter/nft_tunnel.c
index 764e88682a81..30be5787fbde 100644
--- a/net/netfilter/nft_tunnel.c
+++ b/net/netfilter/nft_tunnel.c
@@ -11,6 +11,7 @@
 #include <net/ip_tunnels.h>
 #include <net/vxlan.h>
 #include <net/erspan.h>
+#include <net/geneve.h>
 
 struct nft_tunnel {
 	enum nft_tunnel_keys	key:8;
@@ -144,6 +145,7 @@ struct nft_tunnel_opts {
 	union {
 		struct vxlan_metadata	vxlan;
 		struct erspan_metadata	erspan;
+		u8	data[IP_TUNNEL_OPTS_MAX];
 	} u;
 	u32	len;
 	__be16	flags;
@@ -301,9 +303,53 @@ static int nft_tunnel_obj_erspan_init(const struct nlattr *attr,
 	return 0;
 }
 
+static const struct nla_policy nft_tunnel_opts_geneve_policy[NFTA_TUNNEL_KEY_GENEVE_MAX + 1] = {
+	[NFTA_TUNNEL_KEY_GENEVE_CLASS]	= { .type = NLA_U16 },
+	[NFTA_TUNNEL_KEY_GENEVE_TYPE]	= { .type = NLA_U8 },
+	[NFTA_TUNNEL_KEY_GENEVE_DATA]	= { .type = NLA_BINARY, .len = 128 },
+};
+
+static int nft_tunnel_obj_geneve_init(const struct nlattr *attr,
+				      struct nft_tunnel_opts *opts)
+{
+	struct geneve_opt *opt = (struct geneve_opt *)opts->u.data + opts->len;
+	struct nlattr *tb[NFTA_TUNNEL_KEY_GENEVE_MAX + 1];
+	int err, data_len;
+
+	err = nla_parse_nested(tb, NFTA_TUNNEL_KEY_GENEVE_MAX, attr,
+			       nft_tunnel_opts_geneve_policy, NULL);
+	if (err < 0)
+		return err;
+
+	if (!tb[NFTA_TUNNEL_KEY_GENEVE_CLASS] ||
+	    !tb[NFTA_TUNNEL_KEY_GENEVE_TYPE] ||
+	    !tb[NFTA_TUNNEL_KEY_GENEVE_DATA])
+		return -EINVAL;
+
+	attr = tb[NFTA_TUNNEL_KEY_GENEVE_DATA];
+	data_len = nla_len(attr);
+	if (data_len % 4)
+		return -EINVAL;
+
+	opts->len += sizeof(*opt) + data_len;
+	if (opts->len > IP_TUNNEL_OPTS_MAX)
+		return -EINVAL;
+
+	memcpy(opt->opt_data, nla_data(attr), data_len);
+	opt->length = data_len / 4;
+	opt->opt_class = nla_get_be16(tb[NFTA_TUNNEL_KEY_GENEVE_CLASS]);
+	opt->type = nla_get_u8(tb[NFTA_TUNNEL_KEY_GENEVE_TYPE]);
+	opts->flags = TUNNEL_GENEVE_OPT;
+
+	return 0;
+}
+
 static const struct nla_policy nft_tunnel_opts_policy[NFTA_TUNNEL_KEY_OPTS_MAX + 1] = {
+	[NFTA_TUNNEL_KEY_OPTS_UNSPEC]	= {
+		.strict_start_type = NFTA_TUNNEL_KEY_OPTS_GENEVE },
 	[NFTA_TUNNEL_KEY_OPTS_VXLAN]	= { .type = NLA_NESTED, },
 	[NFTA_TUNNEL_KEY_OPTS_ERSPAN]	= { .type = NLA_NESTED, },
+	[NFTA_TUNNEL_KEY_OPTS_GENEVE]	= { .type = NLA_NESTED, },
 };
 
 static int nft_tunnel_obj_opts_init(const struct nft_ctx *ctx,
@@ -311,22 +357,43 @@ static int nft_tunnel_obj_opts_init(const struct nft_ctx *ctx,
 				    struct ip_tunnel_info *info,
 				    struct nft_tunnel_opts *opts)
 {
-	struct nlattr *tb[NFTA_TUNNEL_KEY_OPTS_MAX + 1];
-	int err;
+	int err, rem, type = 0;
+	struct nlattr *nla;
 
-	err = nla_parse_nested_deprecated(tb, NFTA_TUNNEL_KEY_OPTS_MAX, attr,
-					  nft_tunnel_opts_policy, NULL);
+	err = nla_validate_nested_deprecated(attr, NFTA_TUNNEL_KEY_OPTS_MAX,
+					     nft_tunnel_opts_policy, NULL);
 	if (err < 0)
 		return err;
 
-	if (tb[NFTA_TUNNEL_KEY_OPTS_VXLAN]) {
-		err = nft_tunnel_obj_vxlan_init(tb[NFTA_TUNNEL_KEY_OPTS_VXLAN],
-						opts);
-	} else if (tb[NFTA_TUNNEL_KEY_OPTS_ERSPAN]) {
-		err = nft_tunnel_obj_erspan_init(tb[NFTA_TUNNEL_KEY_OPTS_ERSPAN],
-						 opts);
-	} else {
-		return -EOPNOTSUPP;
+	nla_for_each_attr(nla, nla_data(attr), nla_len(attr), rem) {
+		switch (nla_type(nla)) {
+		case NFTA_TUNNEL_KEY_OPTS_VXLAN:
+			if (type)
+				return -EINVAL;
+			err = nft_tunnel_obj_vxlan_init(nla, opts);
+			if (err)
+				return err;
+			type = TUNNEL_VXLAN_OPT;
+			break;
+		case NFTA_TUNNEL_KEY_OPTS_ERSPAN:
+			if (type)
+				return -EINVAL;
+			err = nft_tunnel_obj_erspan_init(nla, opts);
+			if (err)
+				return err;
+			type = TUNNEL_ERSPAN_OPT;
+			break;
+		case NFTA_TUNNEL_KEY_OPTS_GENEVE:
+			if (type && type != TUNNEL_GENEVE_OPT)
+				return -EINVAL;
+			err = nft_tunnel_obj_geneve_init(nla, opts);
+			if (err)
+				return err;
+			type = TUNNEL_GENEVE_OPT;
+			break;
+		default:
+			return -EOPNOTSUPP;
+		}
 	}
 
 	return err;
@@ -518,6 +585,25 @@ static int nft_tunnel_opts_dump(struct sk_buff *skb,
 			break;
 		}
 		nla_nest_end(skb, inner);
+	} else if (opts->flags & TUNNEL_GENEVE_OPT) {
+		struct geneve_opt *opt;
+		int offset = 0;
+
+		inner = nla_nest_start_noflag(skb, NFTA_TUNNEL_KEY_OPTS_GENEVE);
+		if (!inner)
+			goto failure;
+		while (opts->len > offset) {
+			opt = (struct geneve_opt *)opts->u.data + offset;
+			if (nla_put_be16(skb, NFTA_TUNNEL_KEY_GENEVE_CLASS,
+					 opt->opt_class) ||
+			    nla_put_u8(skb, NFTA_TUNNEL_KEY_GENEVE_TYPE,
+				       opt->type) ||
+			    nla_put(skb, NFTA_TUNNEL_KEY_GENEVE_DATA,
+				    opt->length * 4, opt->opt_data))
+				goto inner_failure;
+			offset += sizeof(*opt) + opt->length * 4;
+		}
+		nla_nest_end(skb, inner);
 	}
 	nla_nest_end(skb, nest);
 	return 0;
-- 
cgit v1.2.3-59-g8ed1b


From e32a4dc6512ce3c1a1920531246e7037896e510a Mon Sep 17 00:00:00 2001
From: Florian Westphal <fw@strlen.de>
Date: Tue, 18 Feb 2020 11:59:26 +0100
Subject: netfilter: nf_tables: make sets built-in

Placing nftables set support in an extra module is pointless:

1. nf_tables needs dynamic registeration interface for sake of one module
2. nft heavily relies on sets, e.g. even simple rule like
   "nft ... tcp dport { 80, 443 }" will not work with _SETS=n.

IOW, either nftables isn't used or both nf_tables and nf_tables_set
modules are needed anyway.

With extra module:
 307K net/netfilter/nf_tables.ko
  79K net/netfilter/nf_tables_set.ko

   text  data  bss     dec filename
 146416  3072  545  150033 nf_tables.ko
  35496  1817    0   37313 nf_tables_set.ko

This patch:
 373K net/netfilter/nf_tables.ko

 178563  4049  545  183157 nf_tables.ko

Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
 include/net/netfilter/nf_tables.h  |  6 ------
 net/netfilter/Kconfig              |  8 --------
 net/netfilter/Makefile             |  9 +++------
 net/netfilter/nf_tables_api.c      | 41 +++++++++++---------------------------
 net/netfilter/nf_tables_set_core.c | 31 ----------------------------
 5 files changed, 15 insertions(+), 80 deletions(-)
 delete mode 100644 net/netfilter/nf_tables_set_core.c

(limited to 'include')

diff --git a/include/net/netfilter/nf_tables.h b/include/net/netfilter/nf_tables.h
index 4170c033d461..9a5f41028736 100644
--- a/include/net/netfilter/nf_tables.h
+++ b/include/net/netfilter/nf_tables.h
@@ -397,9 +397,6 @@ struct nft_set_type {
 };
 #define to_set_type(o) container_of(o, struct nft_set_type, ops)
 
-int nft_register_set(struct nft_set_type *type);
-void nft_unregister_set(struct nft_set_type *type);
-
 /**
  * 	struct nft_set - nf_tables set instance
  *
@@ -1253,9 +1250,6 @@ void nft_trace_notify(struct nft_traceinfo *info);
 #define MODULE_ALIAS_NFT_EXPR(name) \
 	MODULE_ALIAS("nft-expr-" name)
 
-#define MODULE_ALIAS_NFT_SET() \
-	MODULE_ALIAS("nft-set")
-
 #define MODULE_ALIAS_NFT_OBJ(type) \
 	MODULE_ALIAS("nft-obj-" __stringify(type))
 
diff --git a/net/netfilter/Kconfig b/net/netfilter/Kconfig
index 91efae88e8c2..468fea1aebba 100644
--- a/net/netfilter/Kconfig
+++ b/net/netfilter/Kconfig
@@ -455,14 +455,6 @@ config NF_TABLES
 	  To compile it as a module, choose M here.
 
 if NF_TABLES
-
-config NF_TABLES_SET
-	tristate "Netfilter nf_tables set infrastructure"
-	help
-	  This option enables the nf_tables set infrastructure that allows to
-	  look up for elements in a set and to build one-way mappings between
-	  matchings and actions.
-
 config NF_TABLES_INET
 	depends on IPV6
 	select NF_TABLES_IPV4
diff --git a/net/netfilter/Makefile b/net/netfilter/Makefile
index 3f572e5a975e..4fff7d0e2d27 100644
--- a/net/netfilter/Makefile
+++ b/net/netfilter/Makefile
@@ -78,14 +78,11 @@ nf_tables-objs := nf_tables_core.o nf_tables_api.o nft_chain_filter.o \
 		  nf_tables_trace.o nft_immediate.o nft_cmp.o nft_range.o \
 		  nft_bitwise.o nft_byteorder.o nft_payload.o nft_lookup.o \
 		  nft_dynset.o nft_meta.o nft_rt.o nft_exthdr.o \
-		  nft_chain_route.o nf_tables_offload.o
-
-nf_tables_set-objs := nf_tables_set_core.o \
-		      nft_set_hash.o nft_set_bitmap.o nft_set_rbtree.o \
-		      nft_set_pipapo.o
+		  nft_chain_route.o nf_tables_offload.o \
+		  nft_set_hash.o nft_set_bitmap.o nft_set_rbtree.o \
+		  nft_set_pipapo.o
 
 obj-$(CONFIG_NF_TABLES)		+= nf_tables.o
-obj-$(CONFIG_NF_TABLES_SET)	+= nf_tables_set.o
 obj-$(CONFIG_NFT_COMPAT)	+= nft_compat.o
 obj-$(CONFIG_NFT_CONNLIMIT)	+= nft_connlimit.o
 obj-$(CONFIG_NFT_NUMGEN)	+= nft_numgen.o
diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c
index 38c680f28f15..f26a9b638d6c 100644
--- a/net/netfilter/nf_tables_api.c
+++ b/net/netfilter/nf_tables_api.c
@@ -3266,25 +3266,14 @@ static int nf_tables_delrule(struct net *net, struct sock *nlsk,
 /*
  * Sets
  */
-
-static LIST_HEAD(nf_tables_set_types);
-
-int nft_register_set(struct nft_set_type *type)
-{
-	nfnl_lock(NFNL_SUBSYS_NFTABLES);
-	list_add_tail_rcu(&type->list, &nf_tables_set_types);
-	nfnl_unlock(NFNL_SUBSYS_NFTABLES);
-	return 0;
-}
-EXPORT_SYMBOL_GPL(nft_register_set);
-
-void nft_unregister_set(struct nft_set_type *type)
-{
-	nfnl_lock(NFNL_SUBSYS_NFTABLES);
-	list_del_rcu(&type->list);
-	nfnl_unlock(NFNL_SUBSYS_NFTABLES);
-}
-EXPORT_SYMBOL_GPL(nft_unregister_set);
+static const struct nft_set_type *nft_set_types[] = {
+	&nft_set_hash_fast_type,
+	&nft_set_hash_type,
+	&nft_set_rhash_type,
+	&nft_set_bitmap_type,
+	&nft_set_rbtree_type,
+	&nft_set_pipapo_type,
+};
 
 #define NFT_SET_FEATURES	(NFT_SET_INTERVAL | NFT_SET_MAP | \
 				 NFT_SET_TIMEOUT | NFT_SET_OBJECT | \
@@ -3310,15 +3299,11 @@ nft_select_set_ops(const struct nft_ctx *ctx,
 	struct nft_set_estimate est, best;
 	const struct nft_set_type *type;
 	u32 flags = 0;
+	int i;
 
 	lockdep_assert_held(&ctx->net->nft.commit_mutex);
 	lockdep_nfnl_nft_mutex_not_held();
-#ifdef CONFIG_MODULES
-	if (list_empty(&nf_tables_set_types)) {
-		if (nft_request_module(ctx->net, "nft-set") == -EAGAIN)
-			return ERR_PTR(-EAGAIN);
-	}
-#endif
+
 	if (nla[NFTA_SET_FLAGS] != NULL)
 		flags = ntohl(nla_get_be32(nla[NFTA_SET_FLAGS]));
 
@@ -3327,7 +3312,8 @@ nft_select_set_ops(const struct nft_ctx *ctx,
 	best.lookup = ~0;
 	best.space  = ~0;
 
-	list_for_each_entry(type, &nf_tables_set_types, list) {
+	for (i = 0; i < ARRAY_SIZE(nft_set_types); i++) {
+		type = nft_set_types[i];
 		ops = &type->ops;
 
 		if (!nft_set_ops_candidate(type, flags))
@@ -4312,7 +4298,6 @@ const struct nft_set_ext_type nft_set_ext_types[] = {
 		.align	= __alignof__(u32),
 	},
 };
-EXPORT_SYMBOL_GPL(nft_set_ext_types);
 
 /*
  * Set elements
@@ -5365,7 +5350,6 @@ void nft_set_gc_batch_release(struct rcu_head *rcu)
 		nft_set_elem_destroy(gcb->head.set, gcb->elems[i], true);
 	kfree(gcb);
 }
-EXPORT_SYMBOL_GPL(nft_set_gc_batch_release);
 
 struct nft_set_gc_batch *nft_set_gc_batch_alloc(const struct nft_set *set,
 						gfp_t gfp)
@@ -5378,7 +5362,6 @@ struct nft_set_gc_batch *nft_set_gc_batch_alloc(const struct nft_set *set,
 	gcb->head.set = set;
 	return gcb;
 }
-EXPORT_SYMBOL_GPL(nft_set_gc_batch_alloc);
 
 /*
  * Stateful objects
diff --git a/net/netfilter/nf_tables_set_core.c b/net/netfilter/nf_tables_set_core.c
deleted file mode 100644
index 586b621007eb..000000000000
--- a/net/netfilter/nf_tables_set_core.c
+++ /dev/null
@@ -1,31 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0 */
-#include <linux/module.h>
-#include <net/netfilter/nf_tables_core.h>
-
-static int __init nf_tables_set_module_init(void)
-{
-	nft_register_set(&nft_set_hash_fast_type);
-	nft_register_set(&nft_set_hash_type);
-	nft_register_set(&nft_set_rhash_type);
-	nft_register_set(&nft_set_bitmap_type);
-	nft_register_set(&nft_set_rbtree_type);
-	nft_register_set(&nft_set_pipapo_type);
-
-	return 0;
-}
-
-static void __exit nf_tables_set_module_exit(void)
-{
-	nft_unregister_set(&nft_set_pipapo_type);
-	nft_unregister_set(&nft_set_rbtree_type);
-	nft_unregister_set(&nft_set_bitmap_type);
-	nft_unregister_set(&nft_set_rhash_type);
-	nft_unregister_set(&nft_set_hash_type);
-	nft_unregister_set(&nft_set_hash_fast_type);
-}
-
-module_init(nf_tables_set_module_init);
-module_exit(nf_tables_set_module_exit);
-
-MODULE_LICENSE("GPL");
-MODULE_ALIAS_NFT_SET();
-- 
cgit v1.2.3-59-g8ed1b


From 24d19826fcbd97144908ae32019ee67d358c5879 Mon Sep 17 00:00:00 2001
From: Florian Westphal <fw@strlen.de>
Date: Tue, 18 Feb 2020 11:59:27 +0100
Subject: netfilter: nf_tables: make all set structs const

They do not need to be writeable anymore.

v2: remove left-over __read_mostly annotation in set_pipapo.c (Stefano)

Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
 include/net/netfilter/nf_tables.h      |  4 ----
 include/net/netfilter/nf_tables_core.h | 12 ++++++------
 net/netfilter/nf_tables_api.c          | 14 ++------------
 net/netfilter/nft_set_bitmap.c         |  3 +--
 net/netfilter/nft_set_hash.c           |  9 +++------
 net/netfilter/nft_set_pipapo.c         |  3 +--
 net/netfilter/nft_set_rbtree.c         |  3 +--
 7 files changed, 14 insertions(+), 34 deletions(-)

(limited to 'include')

diff --git a/include/net/netfilter/nf_tables.h b/include/net/netfilter/nf_tables.h
index 9a5f41028736..d913cdb6a27b 100644
--- a/include/net/netfilter/nf_tables.h
+++ b/include/net/netfilter/nf_tables.h
@@ -385,14 +385,10 @@ struct nft_set_ops {
  *      struct nft_set_type - nf_tables set type
  *
  *      @ops: set ops for this type
- *      @list: used internally
- *      @owner: module reference
  *      @features: features supported by the implementation
  */
 struct nft_set_type {
 	const struct nft_set_ops	ops;
-	struct list_head		list;
-	struct module			*owner;
 	u32				features;
 };
 #define to_set_type(o) container_of(o, struct nft_set_type, ops)
diff --git a/include/net/netfilter/nf_tables_core.h b/include/net/netfilter/nf_tables_core.h
index 29e7e1021267..3e30cc5d195b 100644
--- a/include/net/netfilter/nf_tables_core.h
+++ b/include/net/netfilter/nf_tables_core.h
@@ -69,12 +69,12 @@ extern const struct nft_expr_ops nft_payload_fast_ops;
 extern struct static_key_false nft_counters_enabled;
 extern struct static_key_false nft_trace_enabled;
 
-extern struct nft_set_type nft_set_rhash_type;
-extern struct nft_set_type nft_set_hash_type;
-extern struct nft_set_type nft_set_hash_fast_type;
-extern struct nft_set_type nft_set_rbtree_type;
-extern struct nft_set_type nft_set_bitmap_type;
-extern struct nft_set_type nft_set_pipapo_type;
+extern const struct nft_set_type nft_set_rhash_type;
+extern const struct nft_set_type nft_set_hash_type;
+extern const struct nft_set_type nft_set_hash_fast_type;
+extern const struct nft_set_type nft_set_rbtree_type;
+extern const struct nft_set_type nft_set_bitmap_type;
+extern const struct nft_set_type nft_set_pipapo_type;
 
 struct nft_expr;
 struct nft_regs;
diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c
index f26a9b638d6c..3bdf2d0259f6 100644
--- a/net/netfilter/nf_tables_api.c
+++ b/net/netfilter/nf_tables_api.c
@@ -3344,11 +3344,6 @@ nft_select_set_ops(const struct nft_ctx *ctx,
 			break;
 		}
 
-		if (!try_module_get(type->owner))
-			continue;
-		if (bops != NULL)
-			module_put(to_set_type(bops)->owner);
-
 		bops = ops;
 		best = est;
 	}
@@ -4047,10 +4042,8 @@ static int nf_tables_newset(struct net *net, struct sock *nlsk,
 		size = ops->privsize(nla, &desc);
 
 	set = kvzalloc(sizeof(*set) + size + udlen, GFP_KERNEL);
-	if (!set) {
-		err = -ENOMEM;
-		goto err1;
-	}
+	if (!set)
+		return -ENOMEM;
 
 	name = nla_strdup(nla[NFTA_SET_NAME], GFP_KERNEL);
 	if (!name) {
@@ -4109,8 +4102,6 @@ err3:
 	kfree(set->name);
 err2:
 	kvfree(set);
-err1:
-	module_put(to_set_type(ops)->owner);
 	return err;
 }
 
@@ -4120,7 +4111,6 @@ static void nft_set_destroy(struct nft_set *set)
 		return;
 
 	set->ops->destroy(set);
-	module_put(to_set_type(set->ops)->owner);
 	kfree(set->name);
 	kvfree(set);
 }
diff --git a/net/netfilter/nft_set_bitmap.c b/net/netfilter/nft_set_bitmap.c
index 87e8d9ba0c9b..1cb2e67e6e03 100644
--- a/net/netfilter/nft_set_bitmap.c
+++ b/net/netfilter/nft_set_bitmap.c
@@ -293,8 +293,7 @@ static bool nft_bitmap_estimate(const struct nft_set_desc *desc, u32 features,
 	return true;
 }
 
-struct nft_set_type nft_set_bitmap_type __read_mostly = {
-	.owner		= THIS_MODULE,
+const struct nft_set_type nft_set_bitmap_type = {
 	.ops		= {
 		.privsize	= nft_bitmap_privsize,
 		.elemsize	= offsetof(struct nft_bitmap_elem, ext),
diff --git a/net/netfilter/nft_set_hash.c b/net/netfilter/nft_set_hash.c
index d350a7cd3af0..4d3f147e8d8d 100644
--- a/net/netfilter/nft_set_hash.c
+++ b/net/netfilter/nft_set_hash.c
@@ -662,8 +662,7 @@ static bool nft_hash_fast_estimate(const struct nft_set_desc *desc, u32 features
 	return true;
 }
 
-struct nft_set_type nft_set_rhash_type __read_mostly = {
-	.owner		= THIS_MODULE,
+const struct nft_set_type nft_set_rhash_type = {
 	.features	= NFT_SET_MAP | NFT_SET_OBJECT |
 			  NFT_SET_TIMEOUT | NFT_SET_EVAL,
 	.ops		= {
@@ -686,8 +685,7 @@ struct nft_set_type nft_set_rhash_type __read_mostly = {
 	},
 };
 
-struct nft_set_type nft_set_hash_type __read_mostly = {
-	.owner		= THIS_MODULE,
+const struct nft_set_type nft_set_hash_type = {
 	.features	= NFT_SET_MAP | NFT_SET_OBJECT,
 	.ops		= {
 		.privsize       = nft_hash_privsize,
@@ -706,8 +704,7 @@ struct nft_set_type nft_set_hash_type __read_mostly = {
 	},
 };
 
-struct nft_set_type nft_set_hash_fast_type __read_mostly = {
-	.owner		= THIS_MODULE,
+const struct nft_set_type nft_set_hash_fast_type = {
 	.features	= NFT_SET_MAP | NFT_SET_OBJECT,
 	.ops		= {
 		.privsize       = nft_hash_privsize,
diff --git a/net/netfilter/nft_set_pipapo.c b/net/netfilter/nft_set_pipapo.c
index 4fc0c924ed5d..34a1678cf290 100644
--- a/net/netfilter/nft_set_pipapo.c
+++ b/net/netfilter/nft_set_pipapo.c
@@ -2081,8 +2081,7 @@ static void nft_pipapo_gc_init(const struct nft_set *set)
 	priv->last_gc = jiffies;
 }
 
-struct nft_set_type nft_set_pipapo_type __read_mostly = {
-	.owner		= THIS_MODULE,
+const struct nft_set_type nft_set_pipapo_type = {
 	.features	= NFT_SET_INTERVAL | NFT_SET_MAP | NFT_SET_OBJECT |
 			  NFT_SET_TIMEOUT,
 	.ops		= {
diff --git a/net/netfilter/nft_set_rbtree.c b/net/netfilter/nft_set_rbtree.c
index 5000b938ab1e..172ef8189f99 100644
--- a/net/netfilter/nft_set_rbtree.c
+++ b/net/netfilter/nft_set_rbtree.c
@@ -481,8 +481,7 @@ static bool nft_rbtree_estimate(const struct nft_set_desc *desc, u32 features,
 	return true;
 }
 
-struct nft_set_type nft_set_rbtree_type __read_mostly = {
-	.owner		= THIS_MODULE,
+const struct nft_set_type nft_set_rbtree_type = {
 	.features	= NFT_SET_INTERVAL | NFT_SET_MAP | NFT_SET_OBJECT | NFT_SET_TIMEOUT,
 	.ops		= {
 		.privsize	= nft_rbtree_privsize,
-- 
cgit v1.2.3-59-g8ed1b


From 6daf14140129d30207ed6a0a69851fa6a3636bda Mon Sep 17 00:00:00 2001
From: "Gustavo A. R. Silva" <gustavo@embeddedor.com>
Date: Thu, 20 Feb 2020 07:59:14 -0600
Subject: netfilter: Replace zero-length array with flexible-array member

The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:

struct foo {
        int stuff;
        struct boo array[];
};

By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.

Also, notice that, dynamic memory allocations won't be affected by
this change:

"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]

Lastly, fix checkpatch.pl warning
WARNING: __aligned(size) is preferred over __attribute__((aligned(size)))
in net/bridge/netfilter/ebtables.c

This issue was found with the help of Coccinelle.

[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour")

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
 include/linux/netfilter/ipset/ip_set.h          | 2 +-
 include/linux/netfilter/x_tables.h              | 8 ++++----
 include/linux/netfilter_arp/arp_tables.h        | 2 +-
 include/linux/netfilter_bridge/ebtables.h       | 2 +-
 include/linux/netfilter_ipv4/ip_tables.h        | 2 +-
 include/linux/netfilter_ipv6/ip6_tables.h       | 2 +-
 include/net/netfilter/nf_conntrack_extend.h     | 2 +-
 include/net/netfilter/nf_conntrack_timeout.h    | 2 +-
 include/net/netfilter/nf_tables.h               | 6 +++---
 include/uapi/linux/netfilter_bridge/ebt_among.h | 2 +-
 net/bridge/netfilter/ebtables.c                 | 2 +-
 net/ipv4/netfilter/arp_tables.c                 | 4 ++--
 net/ipv4/netfilter/ip_tables.c                  | 4 ++--
 net/ipv6/netfilter/ip6_tables.c                 | 4 ++--
 net/netfilter/ipset/ip_set_bitmap_ip.c          | 2 +-
 net/netfilter/ipset/ip_set_bitmap_ipmac.c       | 2 +-
 net/netfilter/ipset/ip_set_bitmap_port.c        | 2 +-
 net/netfilter/ipset/ip_set_hash_gen.h           | 4 ++--
 net/netfilter/nfnetlink_acct.c                  | 2 +-
 net/netfilter/nft_set_pipapo.c                  | 2 +-
 net/netfilter/xt_hashlimit.c                    | 2 +-
 net/netfilter/xt_recent.c                       | 4 ++--
 22 files changed, 32 insertions(+), 32 deletions(-)

(limited to 'include')

diff --git a/include/linux/netfilter/ipset/ip_set.h b/include/linux/netfilter/ipset/ip_set.h
index 5448c8b443db..ab192720e2d6 100644
--- a/include/linux/netfilter/ipset/ip_set.h
+++ b/include/linux/netfilter/ipset/ip_set.h
@@ -98,7 +98,7 @@ struct ip_set_counter {
 
 struct ip_set_comment_rcu {
 	struct rcu_head rcu;
-	char str[0];
+	char str[];
 };
 
 struct ip_set_comment {
diff --git a/include/linux/netfilter/x_tables.h b/include/linux/netfilter/x_tables.h
index 1b261c51b3a3..5da88451853b 100644
--- a/include/linux/netfilter/x_tables.h
+++ b/include/linux/netfilter/x_tables.h
@@ -264,7 +264,7 @@ struct xt_table_info {
 	unsigned int stacksize;
 	void ***jumpstack;
 
-	unsigned char entries[0] __aligned(8);
+	unsigned char entries[] __aligned(8);
 };
 
 int xt_register_target(struct xt_target *target);
@@ -464,7 +464,7 @@ struct compat_xt_entry_match {
 		} kernel;
 		u_int16_t match_size;
 	} u;
-	unsigned char data[0];
+	unsigned char data[];
 };
 
 struct compat_xt_entry_target {
@@ -480,7 +480,7 @@ struct compat_xt_entry_target {
 		} kernel;
 		u_int16_t target_size;
 	} u;
-	unsigned char data[0];
+	unsigned char data[];
 };
 
 /* FIXME: this works only on 32 bit tasks
@@ -494,7 +494,7 @@ struct compat_xt_counters {
 struct compat_xt_counters_info {
 	char name[XT_TABLE_MAXNAMELEN];
 	compat_uint_t num_counters;
-	struct compat_xt_counters counters[0];
+	struct compat_xt_counters counters[];
 };
 
 struct _compat_xt_align {
diff --git a/include/linux/netfilter_arp/arp_tables.h b/include/linux/netfilter_arp/arp_tables.h
index e98028f00e47..7d3537c40ec9 100644
--- a/include/linux/netfilter_arp/arp_tables.h
+++ b/include/linux/netfilter_arp/arp_tables.h
@@ -67,7 +67,7 @@ struct compat_arpt_entry {
 	__u16 next_offset;
 	compat_uint_t comefrom;
 	struct compat_xt_counters counters;
-	unsigned char elems[0];
+	unsigned char elems[];
 };
 
 static inline struct xt_entry_target *
diff --git a/include/linux/netfilter_bridge/ebtables.h b/include/linux/netfilter_bridge/ebtables.h
index 162f59d0d17a..2f5c4e6ecd8a 100644
--- a/include/linux/netfilter_bridge/ebtables.h
+++ b/include/linux/netfilter_bridge/ebtables.h
@@ -85,7 +85,7 @@ struct ebt_table_info {
 	/* room to maintain the stack used for jumping from and into udc */
 	struct ebt_chainstack **chainstack;
 	char *entries;
-	struct ebt_counter counters[0] ____cacheline_aligned;
+	struct ebt_counter counters[] ____cacheline_aligned;
 };
 
 struct ebt_table {
diff --git a/include/linux/netfilter_ipv4/ip_tables.h b/include/linux/netfilter_ipv4/ip_tables.h
index e9e1ed74cdf1..b394bd4f68a3 100644
--- a/include/linux/netfilter_ipv4/ip_tables.h
+++ b/include/linux/netfilter_ipv4/ip_tables.h
@@ -76,7 +76,7 @@ struct compat_ipt_entry {
 	__u16 next_offset;
 	compat_uint_t comefrom;
 	struct compat_xt_counters counters;
-	unsigned char elems[0];
+	unsigned char elems[];
 };
 
 /* Helper functions */
diff --git a/include/linux/netfilter_ipv6/ip6_tables.h b/include/linux/netfilter_ipv6/ip6_tables.h
index 78ab959c4575..8225f7821a29 100644
--- a/include/linux/netfilter_ipv6/ip6_tables.h
+++ b/include/linux/netfilter_ipv6/ip6_tables.h
@@ -43,7 +43,7 @@ struct compat_ip6t_entry {
 	__u16 next_offset;
 	compat_uint_t comefrom;
 	struct compat_xt_counters counters;
-	unsigned char elems[0];
+	unsigned char elems[];
 };
 
 static inline struct xt_entry_target *
diff --git a/include/net/netfilter/nf_conntrack_extend.h b/include/net/netfilter/nf_conntrack_extend.h
index 5ae5295aa46d..e1e588387103 100644
--- a/include/net/netfilter/nf_conntrack_extend.h
+++ b/include/net/netfilter/nf_conntrack_extend.h
@@ -45,7 +45,7 @@ enum nf_ct_ext_id {
 struct nf_ct_ext {
 	u8 offset[NF_CT_EXT_NUM];
 	u8 len;
-	char data[0];
+	char data[];
 };
 
 static inline bool __nf_ct_ext_exist(const struct nf_ct_ext *ext, u8 id)
diff --git a/include/net/netfilter/nf_conntrack_timeout.h b/include/net/netfilter/nf_conntrack_timeout.h
index 6dd72396f534..659b0ea25b4d 100644
--- a/include/net/netfilter/nf_conntrack_timeout.h
+++ b/include/net/netfilter/nf_conntrack_timeout.h
@@ -14,7 +14,7 @@
 struct nf_ct_timeout {
 	__u16			l3num;
 	const struct nf_conntrack_l4proto *l4proto;
-	char			data[0];
+	char			data[];
 };
 
 struct ctnl_timeout {
diff --git a/include/net/netfilter/nf_tables.h b/include/net/netfilter/nf_tables.h
index d913cdb6a27b..01f48b5e6bf2 100644
--- a/include/net/netfilter/nf_tables.h
+++ b/include/net/netfilter/nf_tables.h
@@ -224,7 +224,7 @@ int nft_validate_register_store(const struct nft_ctx *ctx,
  */
 struct nft_userdata {
 	u8			len;
-	unsigned char		data[0];
+	unsigned char		data[];
 };
 
 /**
@@ -565,7 +565,7 @@ struct nft_set_ext_tmpl {
 struct nft_set_ext {
 	u8	genmask;
 	u8	offset[NFT_SET_EXT_NUM];
-	char	data[0];
+	char	data[];
 };
 
 static inline void nft_set_ext_prepare(struct nft_set_ext_tmpl *tmpl)
@@ -1375,7 +1375,7 @@ struct nft_trans {
 	int				msg_type;
 	bool				put_net;
 	struct nft_ctx			ctx;
-	char				data[0];
+	char				data[];
 };
 
 struct nft_trans_rule {
diff --git a/include/uapi/linux/netfilter_bridge/ebt_among.h b/include/uapi/linux/netfilter_bridge/ebt_among.h
index 9acf757bc1f7..73b26a280c4f 100644
--- a/include/uapi/linux/netfilter_bridge/ebt_among.h
+++ b/include/uapi/linux/netfilter_bridge/ebt_among.h
@@ -40,7 +40,7 @@ struct ebt_mac_wormhash_tuple {
 struct ebt_mac_wormhash {
 	int table[257];
 	int poolsize;
-	struct ebt_mac_wormhash_tuple pool[0];
+	struct ebt_mac_wormhash_tuple pool[];
 };
 
 #define ebt_mac_wormhash_size(x) ((x) ? sizeof(struct ebt_mac_wormhash) \
diff --git a/net/bridge/netfilter/ebtables.c b/net/bridge/netfilter/ebtables.c
index e1256e03a9a8..78db58c7aec2 100644
--- a/net/bridge/netfilter/ebtables.c
+++ b/net/bridge/netfilter/ebtables.c
@@ -1561,7 +1561,7 @@ struct compat_ebt_entry_mwt {
 		compat_uptr_t ptr;
 	} u;
 	compat_uint_t match_size;
-	compat_uint_t data[0] __attribute__ ((aligned (__alignof__(struct compat_ebt_replace))));
+	compat_uint_t data[] __aligned(__alignof__(struct compat_ebt_replace));
 };
 
 /* account for possible padding between match_size and ->data */
diff --git a/net/ipv4/netfilter/arp_tables.c b/net/ipv4/netfilter/arp_tables.c
index f1f78a742b36..b167f4a5b684 100644
--- a/net/ipv4/netfilter/arp_tables.c
+++ b/net/ipv4/netfilter/arp_tables.c
@@ -1057,7 +1057,7 @@ struct compat_arpt_replace {
 	u32				underflow[NF_ARP_NUMHOOKS];
 	u32				num_counters;
 	compat_uptr_t			counters;
-	struct compat_arpt_entry	entries[0];
+	struct compat_arpt_entry	entries[];
 };
 
 static inline void compat_release_entry(struct compat_arpt_entry *e)
@@ -1383,7 +1383,7 @@ static int compat_copy_entries_to_user(unsigned int total_size,
 struct compat_arpt_get_entries {
 	char name[XT_TABLE_MAXNAMELEN];
 	compat_uint_t size;
-	struct compat_arpt_entry entrytable[0];
+	struct compat_arpt_entry entrytable[];
 };
 
 static int compat_get_entries(struct net *net,
diff --git a/net/ipv4/netfilter/ip_tables.c b/net/ipv4/netfilter/ip_tables.c
index 10b91ebdf213..c2670eaa74e6 100644
--- a/net/ipv4/netfilter/ip_tables.c
+++ b/net/ipv4/netfilter/ip_tables.c
@@ -1211,7 +1211,7 @@ struct compat_ipt_replace {
 	u32			underflow[NF_INET_NUMHOOKS];
 	u32			num_counters;
 	compat_uptr_t		counters;	/* struct xt_counters * */
-	struct compat_ipt_entry	entries[0];
+	struct compat_ipt_entry	entries[];
 };
 
 static int
@@ -1562,7 +1562,7 @@ compat_do_ipt_set_ctl(struct sock *sk,	int cmd, void __user *user,
 struct compat_ipt_get_entries {
 	char name[XT_TABLE_MAXNAMELEN];
 	compat_uint_t size;
-	struct compat_ipt_entry entrytable[0];
+	struct compat_ipt_entry entrytable[];
 };
 
 static int
diff --git a/net/ipv6/netfilter/ip6_tables.c b/net/ipv6/netfilter/ip6_tables.c
index c973ace208c5..e27393498ecb 100644
--- a/net/ipv6/netfilter/ip6_tables.c
+++ b/net/ipv6/netfilter/ip6_tables.c
@@ -1227,7 +1227,7 @@ struct compat_ip6t_replace {
 	u32			underflow[NF_INET_NUMHOOKS];
 	u32			num_counters;
 	compat_uptr_t		counters;	/* struct xt_counters * */
-	struct compat_ip6t_entry entries[0];
+	struct compat_ip6t_entry entries[];
 };
 
 static int
@@ -1571,7 +1571,7 @@ compat_do_ip6t_set_ctl(struct sock *sk, int cmd, void __user *user,
 struct compat_ip6t_get_entries {
 	char name[XT_TABLE_MAXNAMELEN];
 	compat_uint_t size;
-	struct compat_ip6t_entry entrytable[0];
+	struct compat_ip6t_entry entrytable[];
 };
 
 static int
diff --git a/net/netfilter/ipset/ip_set_bitmap_ip.c b/net/netfilter/ipset/ip_set_bitmap_ip.c
index 0a2196f59106..486959f70cf3 100644
--- a/net/netfilter/ipset/ip_set_bitmap_ip.c
+++ b/net/netfilter/ipset/ip_set_bitmap_ip.c
@@ -46,7 +46,7 @@ struct bitmap_ip {
 	u8 netmask;		/* subnet netmask */
 	struct timer_list gc;	/* garbage collection */
 	struct ip_set *set;	/* attached to this ip_set */
-	unsigned char extensions[0]	/* data extensions */
+	unsigned char extensions[]	/* data extensions */
 		__aligned(__alignof__(u64));
 };
 
diff --git a/net/netfilter/ipset/ip_set_bitmap_ipmac.c b/net/netfilter/ipset/ip_set_bitmap_ipmac.c
index 739e343efaf6..2310a316e0af 100644
--- a/net/netfilter/ipset/ip_set_bitmap_ipmac.c
+++ b/net/netfilter/ipset/ip_set_bitmap_ipmac.c
@@ -49,7 +49,7 @@ struct bitmap_ipmac {
 	size_t memsize;		/* members size */
 	struct timer_list gc;	/* garbage collector */
 	struct ip_set *set;	/* attached to this ip_set */
-	unsigned char extensions[0]	/* MAC + data extensions */
+	unsigned char extensions[]	/* MAC + data extensions */
 		__aligned(__alignof__(u64));
 };
 
diff --git a/net/netfilter/ipset/ip_set_bitmap_port.c b/net/netfilter/ipset/ip_set_bitmap_port.c
index b49978dd810d..e56ced66f202 100644
--- a/net/netfilter/ipset/ip_set_bitmap_port.c
+++ b/net/netfilter/ipset/ip_set_bitmap_port.c
@@ -37,7 +37,7 @@ struct bitmap_port {
 	size_t memsize;		/* members size */
 	struct timer_list gc;	/* garbage collection */
 	struct ip_set *set;	/* attached to this ip_set */
-	unsigned char extensions[0]	/* data extensions */
+	unsigned char extensions[]	/* data extensions */
 		__aligned(__alignof__(u64));
 };
 
diff --git a/net/netfilter/ipset/ip_set_hash_gen.h b/net/netfilter/ipset/ip_set_hash_gen.h
index e52d7b7597a0..1ee43752d6d3 100644
--- a/net/netfilter/ipset/ip_set_hash_gen.h
+++ b/net/netfilter/ipset/ip_set_hash_gen.h
@@ -76,7 +76,7 @@ struct hbucket {
 	DECLARE_BITMAP(used, AHASH_MAX_TUNED);
 	u8 size;		/* size of the array */
 	u8 pos;			/* position of the first free entry */
-	unsigned char value[0]	/* the array of the values */
+	unsigned char value[]	/* the array of the values */
 		__aligned(__alignof__(u64));
 };
 
@@ -109,7 +109,7 @@ struct htable {
 	u8 htable_bits;		/* size of hash table == 2^htable_bits */
 	u32 maxelem;		/* Maxelem per region */
 	struct ip_set_region *hregion;	/* Region locks and ext sizes */
-	struct hbucket __rcu *bucket[0]; /* hashtable buckets */
+	struct hbucket __rcu *bucket[]; /* hashtable buckets */
 };
 
 #define hbucket(h, i)		((h)->bucket[i])
diff --git a/net/netfilter/nfnetlink_acct.c b/net/netfilter/nfnetlink_acct.c
index 2481470dec36..5827117f2635 100644
--- a/net/netfilter/nfnetlink_acct.c
+++ b/net/netfilter/nfnetlink_acct.c
@@ -33,7 +33,7 @@ struct nf_acct {
 	refcount_t		refcnt;
 	char			name[NFACCT_NAME_MAX];
 	struct rcu_head		rcu_head;
-	char			data[0];
+	char			data[];
 };
 
 struct nfacct_filter {
diff --git a/net/netfilter/nft_set_pipapo.c b/net/netfilter/nft_set_pipapo.c
index d752a39b97d6..26395c8188b1 100644
--- a/net/netfilter/nft_set_pipapo.c
+++ b/net/netfilter/nft_set_pipapo.c
@@ -433,7 +433,7 @@ struct nft_pipapo_match {
 	unsigned long * __percpu *scratch;
 	size_t bsize_max;
 	struct rcu_head rcu;
-	struct nft_pipapo_field f[0];
+	struct nft_pipapo_field f[];
 };
 
 /* Current working bitmap index, toggled between field matches */
diff --git a/net/netfilter/xt_hashlimit.c b/net/netfilter/xt_hashlimit.c
index 8c835ad63729..9c5cfd74a0ee 100644
--- a/net/netfilter/xt_hashlimit.c
+++ b/net/netfilter/xt_hashlimit.c
@@ -132,7 +132,7 @@ struct xt_hashlimit_htable {
 	const char *name;
 	struct net *net;
 
-	struct hlist_head hash[0];	/* hashtable itself */
+	struct hlist_head hash[];	/* hashtable itself */
 };
 
 static int
diff --git a/net/netfilter/xt_recent.c b/net/netfilter/xt_recent.c
index 225a7ab6d79a..19bef176145e 100644
--- a/net/netfilter/xt_recent.c
+++ b/net/netfilter/xt_recent.c
@@ -71,7 +71,7 @@ struct recent_entry {
 	u_int8_t		ttl;
 	u_int8_t		index;
 	u_int16_t		nstamps;
-	unsigned long		stamps[0];
+	unsigned long		stamps[];
 };
 
 struct recent_table {
@@ -82,7 +82,7 @@ struct recent_table {
 	unsigned int		entries;
 	u8			nstamps_max_mask;
 	struct list_head	lru_list;
-	struct list_head	iphash[0];
+	struct list_head	iphash[];
 };
 
 struct recent_net {
-- 
cgit v1.2.3-59-g8ed1b


From cfab6dbd0ecf342fc904952d8565c8d80c741a63 Mon Sep 17 00:00:00 2001
From: wenxu <wenxu@ucloud.cn>
Date: Mon, 24 Feb 2020 05:22:54 +0100
Subject: netfilter: flowtable: add tunnel match offload support

This patch support both ipv4 and ipv6 tunnel_id, tunnel_src and
tunnel_dst match for flowtable offload

Signed-off-by: wenxu <wenxu@ucloud.cn>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
 include/net/netfilter/nf_flow_table.h |  6 ++++
 net/netfilter/nf_flow_table_offload.c | 61 +++++++++++++++++++++++++++++++++--
 2 files changed, 65 insertions(+), 2 deletions(-)

(limited to 'include')

diff --git a/include/net/netfilter/nf_flow_table.h b/include/net/netfilter/nf_flow_table.h
index 6890f1ca3e31..f523ea87b6ae 100644
--- a/include/net/netfilter/nf_flow_table.h
+++ b/include/net/netfilter/nf_flow_table.h
@@ -19,11 +19,17 @@ enum flow_offload_tuple_dir;
 struct nf_flow_key {
 	struct flow_dissector_key_meta			meta;
 	struct flow_dissector_key_control		control;
+	struct flow_dissector_key_control		enc_control;
 	struct flow_dissector_key_basic			basic;
 	union {
 		struct flow_dissector_key_ipv4_addrs	ipv4;
 		struct flow_dissector_key_ipv6_addrs	ipv6;
 	};
+	struct flow_dissector_key_keyid			enc_key_id;
+	union {
+		struct flow_dissector_key_ipv4_addrs	enc_ipv4;
+		struct flow_dissector_key_ipv6_addrs	enc_ipv6;
+	};
 	struct flow_dissector_key_tcp			tcp;
 	struct flow_dissector_key_ports			tp;
 } __aligned(BITS_PER_LONG / 8); /* Ensure that we can do comparisons as longs. */
diff --git a/net/netfilter/nf_flow_table_offload.c b/net/netfilter/nf_flow_table_offload.c
index f60f01e929b8..3101b35eac80 100644
--- a/net/netfilter/nf_flow_table_offload.c
+++ b/net/netfilter/nf_flow_table_offload.c
@@ -28,11 +28,61 @@ struct flow_offload_work {
 	(__match)->dissector.offset[__type] =		\
 		offsetof(struct nf_flow_key, __field)
 
+static void nf_flow_rule_lwt_match(struct nf_flow_match *match,
+				   struct ip_tunnel_info *tun_info)
+{
+	struct nf_flow_key *mask = &match->mask;
+	struct nf_flow_key *key = &match->key;
+	unsigned int enc_keys;
+
+	if (!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX))
+		return;
+
+	NF_FLOW_DISSECTOR(match, FLOW_DISSECTOR_KEY_ENC_CONTROL, enc_control);
+	NF_FLOW_DISSECTOR(match, FLOW_DISSECTOR_KEY_ENC_KEYID, enc_key_id);
+	key->enc_key_id.keyid = tunnel_id_to_key32(tun_info->key.tun_id);
+	mask->enc_key_id.keyid = 0xffffffff;
+	enc_keys = BIT(FLOW_DISSECTOR_KEY_ENC_KEYID) |
+		   BIT(FLOW_DISSECTOR_KEY_ENC_CONTROL);
+
+	if (ip_tunnel_info_af(tun_info) == AF_INET) {
+		NF_FLOW_DISSECTOR(match, FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
+				  enc_ipv4);
+		key->enc_ipv4.src = tun_info->key.u.ipv4.dst;
+		key->enc_ipv4.dst = tun_info->key.u.ipv4.src;
+		if (key->enc_ipv4.src)
+			mask->enc_ipv4.src = 0xffffffff;
+		if (key->enc_ipv4.dst)
+			mask->enc_ipv4.dst = 0xffffffff;
+		enc_keys |= BIT(FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS);
+		key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
+	} else {
+		memcpy(&key->enc_ipv6.src, &tun_info->key.u.ipv6.dst,
+		       sizeof(struct in6_addr));
+		memcpy(&key->enc_ipv6.dst, &tun_info->key.u.ipv6.src,
+		       sizeof(struct in6_addr));
+		if (memcmp(&key->enc_ipv6.src, &in6addr_any,
+			   sizeof(struct in6_addr)))
+			memset(&key->enc_ipv6.src, 0xff,
+			       sizeof(struct in6_addr));
+		if (memcmp(&key->enc_ipv6.dst, &in6addr_any,
+			   sizeof(struct in6_addr)))
+			memset(&key->enc_ipv6.dst, 0xff,
+			       sizeof(struct in6_addr));
+		enc_keys |= BIT(FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS);
+		key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
+	}
+
+	match->dissector.used_keys |= enc_keys;
+}
+
 static int nf_flow_rule_match(struct nf_flow_match *match,
-			      const struct flow_offload_tuple *tuple)
+			      const struct flow_offload_tuple *tuple,
+			      struct dst_entry *other_dst)
 {
 	struct nf_flow_key *mask = &match->mask;
 	struct nf_flow_key *key = &match->key;
+	struct ip_tunnel_info *tun_info;
 
 	NF_FLOW_DISSECTOR(match, FLOW_DISSECTOR_KEY_META, meta);
 	NF_FLOW_DISSECTOR(match, FLOW_DISSECTOR_KEY_CONTROL, control);
@@ -42,6 +92,11 @@ static int nf_flow_rule_match(struct nf_flow_match *match,
 	NF_FLOW_DISSECTOR(match, FLOW_DISSECTOR_KEY_TCP, tcp);
 	NF_FLOW_DISSECTOR(match, FLOW_DISSECTOR_KEY_PORTS, tp);
 
+	if (other_dst->lwtstate) {
+		tun_info = lwt_tun_info(other_dst->lwtstate);
+		nf_flow_rule_lwt_match(match, tun_info);
+	}
+
 	key->meta.ingress_ifindex = tuple->iifidx;
 	mask->meta.ingress_ifindex = 0xffffffff;
 
@@ -480,6 +535,7 @@ nf_flow_offload_rule_alloc(struct net *net,
 	const struct flow_offload *flow = offload->flow;
 	const struct flow_offload_tuple *tuple;
 	struct nf_flow_rule *flow_rule;
+	struct dst_entry *other_dst;
 	int err = -ENOMEM;
 
 	flow_rule = kzalloc(sizeof(*flow_rule), GFP_KERNEL);
@@ -495,7 +551,8 @@ nf_flow_offload_rule_alloc(struct net *net,
 	flow_rule->rule->match.key = &flow_rule->match.key;
 
 	tuple = &flow->tuplehash[dir].tuple;
-	err = nf_flow_rule_match(&flow_rule->match, tuple);
+	other_dst = flow->tuplehash[!dir].tuple.dst_cache;
+	err = nf_flow_rule_match(&flow_rule->match, tuple, other_dst);
 	if (err < 0)
 		goto err_flow_match;
 
-- 
cgit v1.2.3-59-g8ed1b


From 7400b063969bdca4a06cd97f1294d765c8eecbe1 Mon Sep 17 00:00:00 2001
From: Stefano Brivio <sbrivio@redhat.com>
Date: Sat, 7 Mar 2020 17:52:36 +0100
Subject: nft_set_pipapo: Introduce AVX2-based lookup implementation

If the AVX2 set is available, we can exploit the repetitive
characteristic of this algorithm to provide a fast, vectorised
version by using 256-bit wide AVX2 operations for bucket loads and
bitwise intersections.

In most cases, this implementation consistently outperforms rbtree
set instances despite the fact they are configured to use a given,
single, ranged data type out of the ones used for performance
measurements by the nft_concat_range.sh kselftest.

That script, injecting packets directly on the ingoing device path
with pktgen, reports, averaged over five runs on a single AMD Epyc
7402 thread (3.35GHz, 768 KiB L1D$, 12 MiB L2$), the figures below.
CONFIG_RETPOLINE was not set here.

Note that this is not a fair comparison over hash and rbtree set
types: non-ranged entries (used to have a reference for hash types)
would be matched faster than this, and matching on a single field
only (which is the case for rbtree) is also significantly faster.

However, it's not possible at the moment to choose this set type
for non-ranged entries, and the current implementation also needs
a few minor adjustments in order to match on less than two fields.

 ---------------.-----------------------------------.------------.
 AMD Epyc 7402  |          baselines, Mpps          | this patch |
  1 thread      |___________________________________|____________|
  3.35GHz       |        |        |        |        |            |
  768KiB L1D$   | netdev |  hash  | rbtree |        |            |
 ---------------|  hook  |   no   | single |        |   pipapo   |
 type   entries |  drop  | ranges | field  | pipapo |    AVX2    |
 ---------------|--------|--------|--------|--------|------------|
 net,port       |        |        |        |        |            |
          1000  |   19.0 |   10.4 |    3.8 |    4.0 | 7.5   +87% |
 ---------------|--------|--------|--------|--------|------------|
 port,net       |        |        |        |        |            |
           100  |   18.8 |   10.3 |    5.8 |    6.3 | 8.1   +29% |
 ---------------|--------|--------|--------|--------|------------|
 net6,port      |        |        |        |        |            |
          1000  |   16.4 |    7.6 |    1.8 |    2.1 | 4.8  +128% |
 ---------------|--------|--------|--------|--------|------------|
 port,proto     |        |        |        |        |            |
         30000  |   19.6 |   11.6 |    3.9 |    0.5 | 2.6  +420% |
 ---------------|--------|--------|--------|--------|------------|
 net6,port,mac  |        |        |        |        |            |
            10  |   16.5 |    5.4 |    4.3 |    3.4 | 4.7   +38% |
 ---------------|--------|--------|--------|--------|------------|
 net6,port,mac, |        |        |        |        |            |
 proto    1000  |   16.5 |    5.7 |    1.9 |    1.4 | 3.6   +26% |
 ---------------|--------|--------|--------|--------|------------|
 net,mac        |        |        |        |        |            |
          1000  |   19.0 |    8.4 |    3.9 |    2.5 | 6.4  +156% |
 ---------------'--------'--------'--------'--------'------------'

A similar strategy could be easily reused to implement specialised
versions for other SIMD sets, and I plan to post at least a NEON
version at a later time.

Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
 include/net/netfilter/nf_tables_core.h |    1 +
 net/netfilter/Makefile                 |    6 +
 net/netfilter/nf_tables_api.c          |    3 +
 net/netfilter/nft_set_pipapo.c         |   24 +
 net/netfilter/nft_set_pipapo_avx2.c    | 1222 ++++++++++++++++++++++++++++++++
 net/netfilter/nft_set_pipapo_avx2.h    |   14 +
 6 files changed, 1270 insertions(+)
 create mode 100644 net/netfilter/nft_set_pipapo_avx2.c
 create mode 100644 net/netfilter/nft_set_pipapo_avx2.h

(limited to 'include')

diff --git a/include/net/netfilter/nf_tables_core.h b/include/net/netfilter/nf_tables_core.h
index 3e30cc5d195b..78516de14d31 100644
--- a/include/net/netfilter/nf_tables_core.h
+++ b/include/net/netfilter/nf_tables_core.h
@@ -75,6 +75,7 @@ extern const struct nft_set_type nft_set_hash_fast_type;
 extern const struct nft_set_type nft_set_rbtree_type;
 extern const struct nft_set_type nft_set_bitmap_type;
 extern const struct nft_set_type nft_set_pipapo_type;
+extern const struct nft_set_type nft_set_pipapo_avx2_type;
 
 struct nft_expr;
 struct nft_regs;
diff --git a/net/netfilter/Makefile b/net/netfilter/Makefile
index 4fff7d0e2d27..292e71dc7ba4 100644
--- a/net/netfilter/Makefile
+++ b/net/netfilter/Makefile
@@ -82,6 +82,12 @@ nf_tables-objs := nf_tables_core.o nf_tables_api.o nft_chain_filter.o \
 		  nft_set_hash.o nft_set_bitmap.o nft_set_rbtree.o \
 		  nft_set_pipapo.o
 
+ifdef CONFIG_X86_64
+ifneq (,$(findstring -DCONFIG_AS_AVX2=1,$(KBUILD_CFLAGS)))
+nf_tables-objs += nft_set_pipapo_avx2.o
+endif
+endif
+
 obj-$(CONFIG_NF_TABLES)		+= nf_tables.o
 obj-$(CONFIG_NFT_COMPAT)	+= nft_compat.o
 obj-$(CONFIG_NFT_CONNLIMIT)	+= nft_connlimit.o
diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c
index 3bdf2d0259f6..7187b0d76d9a 100644
--- a/net/netfilter/nf_tables_api.c
+++ b/net/netfilter/nf_tables_api.c
@@ -3272,6 +3272,9 @@ static const struct nft_set_type *nft_set_types[] = {
 	&nft_set_rhash_type,
 	&nft_set_bitmap_type,
 	&nft_set_rbtree_type,
+#if defined(CONFIG_X86_64) && defined(CONFIG_AS_AVX2)
+	&nft_set_pipapo_avx2_type,
+#endif
 	&nft_set_pipapo_type,
 };
 
diff --git a/net/netfilter/nft_set_pipapo.c b/net/netfilter/nft_set_pipapo.c
index 141e0ab26d3c..1e8dd5dccdf7 100644
--- a/net/netfilter/nft_set_pipapo.c
+++ b/net/netfilter/nft_set_pipapo.c
@@ -339,6 +339,7 @@
 #include <linux/bitmap.h>
 #include <linux/bitops.h>
 
+#include "nft_set_pipapo_avx2.h"
 #include "nft_set_pipapo.h"
 
 /* Current working bitmap index, toggled between field matches */
@@ -2174,3 +2175,26 @@ const struct nft_set_type nft_set_pipapo_type = {
 		.elemsize	= offsetof(struct nft_pipapo_elem, ext),
 	},
 };
+
+#if defined(CONFIG_X86_64) && defined(CONFIG_AS_AVX2)
+const struct nft_set_type nft_set_pipapo_avx2_type = {
+	.features	= NFT_SET_INTERVAL | NFT_SET_MAP | NFT_SET_OBJECT |
+			  NFT_SET_TIMEOUT,
+	.ops		= {
+		.lookup		= nft_pipapo_avx2_lookup,
+		.insert		= nft_pipapo_insert,
+		.activate	= nft_pipapo_activate,
+		.deactivate	= nft_pipapo_deactivate,
+		.flush		= nft_pipapo_flush,
+		.remove		= nft_pipapo_remove,
+		.walk		= nft_pipapo_walk,
+		.get		= nft_pipapo_get,
+		.privsize	= nft_pipapo_privsize,
+		.estimate	= nft_pipapo_avx2_estimate,
+		.init		= nft_pipapo_init,
+		.destroy	= nft_pipapo_destroy,
+		.gc_init	= nft_pipapo_gc_init,
+		.elemsize	= offsetof(struct nft_pipapo_elem, ext),
+	},
+};
+#endif
diff --git a/net/netfilter/nft_set_pipapo_avx2.c b/net/netfilter/nft_set_pipapo_avx2.c
new file mode 100644
index 000000000000..f6e20154d2b7
--- /dev/null
+++ b/net/netfilter/nft_set_pipapo_avx2.c
@@ -0,0 +1,1222 @@
+// SPDX-License-Identifier: GPL-2.0-only
+
+/* PIPAPO: PIle PAcket POlicies: AVX2 packet lookup routines
+ *
+ * Copyright (c) 2019-2020 Red Hat GmbH
+ *
+ * Author: Stefano Brivio <sbrivio@redhat.com>
+ */
+
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/netlink.h>
+#include <linux/netfilter.h>
+#include <linux/netfilter/nf_tables.h>
+#include <net/netfilter/nf_tables_core.h>
+#include <uapi/linux/netfilter/nf_tables.h>
+#include <linux/bitmap.h>
+#include <linux/bitops.h>
+
+#include <linux/compiler.h>
+#include <asm/fpu/api.h>
+
+#include "nft_set_pipapo_avx2.h"
+#include "nft_set_pipapo.h"
+
+#define NFT_PIPAPO_LONGS_PER_M256	(XSAVE_YMM_SIZE / BITS_PER_LONG)
+
+/* Load from memory into YMM register with non-temporal hint ("stream load"),
+ * that is, don't fetch lines from memory into the cache. This avoids pushing
+ * precious packet data out of the cache hierarchy, and is appropriate when:
+ *
+ * - loading buckets from lookup tables, as they are not going to be used
+ *   again before packets are entirely classified
+ *
+ * - loading the result bitmap from the previous field, as it's never used
+ *   again
+ */
+#define NFT_PIPAPO_AVX2_LOAD(reg, loc)					\
+	asm volatile("vmovntdqa %0, %%ymm" #reg : : "m" (loc))
+
+/* Stream a single lookup table bucket into YMM register given lookup table,
+ * group index, value of packet bits, bucket size.
+ */
+#define NFT_PIPAPO_AVX2_BUCKET_LOAD4(reg, lt, group, v, bsize)		\
+	NFT_PIPAPO_AVX2_LOAD(reg,					\
+			     lt[((group) * NFT_PIPAPO_BUCKETS(4) +	\
+				 (v)) * (bsize)])
+#define NFT_PIPAPO_AVX2_BUCKET_LOAD8(reg, lt, group, v, bsize)		\
+	NFT_PIPAPO_AVX2_LOAD(reg,					\
+			     lt[((group) * NFT_PIPAPO_BUCKETS(8) +	\
+				 (v)) * (bsize)])
+
+/* Bitwise AND: the staple operation of this algorithm */
+#define NFT_PIPAPO_AVX2_AND(dst, a, b)					\
+	asm volatile("vpand %ymm" #a ", %ymm" #b ", %ymm" #dst)
+
+/* Jump to label if @reg is zero */
+#define NFT_PIPAPO_AVX2_NOMATCH_GOTO(reg, label)			\
+	asm_volatile_goto("vptest %%ymm" #reg ", %%ymm" #reg ";"	\
+			  "je %l[" #label "]" : : : : label)
+
+/* Store 256 bits from YMM register into memory. Contrary to bucket load
+ * operation, we don't bypass the cache here, as stored matching results
+ * are always used shortly after.
+ */
+#define NFT_PIPAPO_AVX2_STORE(loc, reg)					\
+	asm volatile("vmovdqa %%ymm" #reg ", %0" : "=m" (loc))
+
+/* Zero out a complete YMM register, @reg */
+#define NFT_PIPAPO_AVX2_ZERO(reg)					\
+	asm volatile("vpxor %ymm" #reg ", %ymm" #reg ", %ymm" #reg)
+
+/* Current working bitmap index, toggled between field matches */
+static DEFINE_PER_CPU(bool, nft_pipapo_avx2_scratch_index);
+
+/**
+ * nft_pipapo_avx2_prepare() - Prepare before main algorithm body
+ *
+ * This zeroes out ymm15, which is later used whenever we need to clear a
+ * memory location, by storing its content into memory.
+ */
+static void nft_pipapo_avx2_prepare(void)
+{
+	NFT_PIPAPO_AVX2_ZERO(15);
+}
+
+/**
+ * nft_pipapo_avx2_fill() - Fill a bitmap region with ones
+ * @data:	Base memory area
+ * @start:	First bit to set
+ * @len:	Count of bits to fill
+ *
+ * This is nothing else than a version of bitmap_set(), as used e.g. by
+ * pipapo_refill(), tailored for the microarchitectures using it and better
+ * suited for the specific usage: it's very likely that we'll set a small number
+ * of bits, not crossing a word boundary, and correct branch prediction is
+ * critical here.
+ *
+ * This function doesn't actually use any AVX2 instruction.
+ */
+static void nft_pipapo_avx2_fill(unsigned long *data, int start, int len)
+{
+	int offset = start % BITS_PER_LONG;
+	unsigned long mask;
+
+	data += start / BITS_PER_LONG;
+
+	if (likely(len == 1)) {
+		*data |= BIT(offset);
+		return;
+	}
+
+	if (likely(len < BITS_PER_LONG || offset)) {
+		if (likely(len + offset <= BITS_PER_LONG)) {
+			*data |= GENMASK(len - 1 + offset, offset);
+			return;
+		}
+
+		*data |= ~0UL << offset;
+		len -= BITS_PER_LONG - offset;
+		data++;
+
+		if (len <= BITS_PER_LONG) {
+			mask = ~0UL >> (BITS_PER_LONG - len);
+			*data |= mask;
+			return;
+		}
+	}
+
+	memset(data, 0xff, len / BITS_PER_BYTE);
+	data += len / BITS_PER_LONG;
+
+	len %= BITS_PER_LONG;
+	if (len)
+		*data |= ~0UL >> (BITS_PER_LONG - len);
+}
+
+/**
+ * nft_pipapo_avx2_refill() - Scan bitmap, select mapping table item, set bits
+ * @offset:	Start from given bitmap (equivalent to bucket) offset, in longs
+ * @map:	Bitmap to be scanned for set bits
+ * @dst:	Destination bitmap
+ * @mt:		Mapping table containing bit set specifiers
+ * @len:	Length of bitmap in longs
+ * @last:	Return index of first set bit, if this is the last field
+ *
+ * This is an alternative implementation of pipapo_refill() suitable for usage
+ * with AVX2 lookup routines: we know there are four words to be scanned, at
+ * a given offset inside the map, for each matching iteration.
+ *
+ * This function doesn't actually use any AVX2 instruction.
+ *
+ * Return: first set bit index if @last, index of first filled word otherwise.
+ */
+static int nft_pipapo_avx2_refill(int offset, unsigned long *map,
+				  unsigned long *dst,
+				  union nft_pipapo_map_bucket *mt, bool last)
+{
+	int ret = -1;
+
+#define NFT_PIPAPO_AVX2_REFILL_ONE_WORD(x)				\
+	do {								\
+		while (map[(x)]) {					\
+			int r = __builtin_ctzl(map[(x)]);		\
+			int i = (offset + (x)) * BITS_PER_LONG + r;	\
+									\
+			if (last)					\
+				return i;				\
+									\
+			nft_pipapo_avx2_fill(dst, mt[i].to, mt[i].n);	\
+									\
+			if (ret == -1)					\
+				ret = mt[i].to;				\
+									\
+			map[(x)] &= ~(1UL << r);			\
+		}							\
+	} while (0)
+
+	NFT_PIPAPO_AVX2_REFILL_ONE_WORD(0);
+	NFT_PIPAPO_AVX2_REFILL_ONE_WORD(1);
+	NFT_PIPAPO_AVX2_REFILL_ONE_WORD(2);
+	NFT_PIPAPO_AVX2_REFILL_ONE_WORD(3);
+#undef NFT_PIPAPO_AVX2_REFILL_ONE_WORD
+
+	return ret;
+}
+
+/**
+ * nft_pipapo_avx2_lookup_4b_2() - AVX2-based lookup for 2 four-bit groups
+ * @map:	Previous match result, used as initial bitmap
+ * @fill:	Destination bitmap to be filled with current match result
+ * @f:		Field, containing lookup and mapping tables
+ * @offset:	Ignore buckets before the given index, no bits are filled there
+ * @pkt:	Packet data, pointer to input nftables register
+ * @first:	If this is the first field, don't source previous result
+ * @last:	Last field: stop at the first match and return bit index
+ *
+ * Load buckets from lookup table corresponding to the values of each 4-bit
+ * group of packet bytes, and perform a bitwise intersection between them. If
+ * this is the first field in the set, simply AND the buckets together
+ * (equivalent to using an all-ones starting bitmap), use the provided starting
+ * bitmap otherwise. Then call nft_pipapo_avx2_refill() to generate the next
+ * working bitmap, @fill.
+ *
+ * This is used for 8-bit fields (i.e. protocol numbers).
+ *
+ * Out-of-order (and superscalar) execution is vital here, so it's critical to
+ * avoid false data dependencies. CPU and compiler could (mostly) take care of
+ * this on their own, but the operation ordering is explicitly given here with
+ * a likely execution order in mind, to highlight possible stalls. That's why
+ * a number of logically distinct operations (i.e. loading buckets, intersecting
+ * buckets) are interleaved.
+ *
+ * Return: -1 on no match, rule index of match if @last, otherwise first long
+ * word index to be checked next (i.e. first filled word).
+ */
+static int nft_pipapo_avx2_lookup_4b_2(unsigned long *map, unsigned long *fill,
+				       struct nft_pipapo_field *f, int offset,
+				       const u8 *pkt, bool first, bool last)
+{
+	int i, ret = -1, m256_size = f->bsize / NFT_PIPAPO_LONGS_PER_M256, b;
+	u8 pg[2] = { pkt[0] >> 4, pkt[0] & 0xf };
+	unsigned long *lt = f->lt, bsize = f->bsize;
+
+	lt += offset * NFT_PIPAPO_LONGS_PER_M256;
+	for (i = offset; i < m256_size; i++, lt += NFT_PIPAPO_LONGS_PER_M256) {
+		int i_ul = i * NFT_PIPAPO_LONGS_PER_M256;
+
+		if (first) {
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(0, lt, 0, pg[0], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(1, lt, 1, pg[1], bsize);
+			NFT_PIPAPO_AVX2_AND(4, 0, 1);
+		} else {
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(0, lt, 0, pg[0], bsize);
+			NFT_PIPAPO_AVX2_LOAD(2, map[i_ul]);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(1, lt, 1, pg[1], bsize);
+			NFT_PIPAPO_AVX2_NOMATCH_GOTO(2, nothing);
+			NFT_PIPAPO_AVX2_AND(3, 0, 1);
+			NFT_PIPAPO_AVX2_AND(4, 2, 3);
+		}
+
+		NFT_PIPAPO_AVX2_NOMATCH_GOTO(4, nomatch);
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 4);
+
+		b = nft_pipapo_avx2_refill(i_ul, &map[i_ul], fill, f->mt, last);
+		if (last)
+			return b;
+
+		if (unlikely(ret == -1))
+			ret = b / XSAVE_YMM_SIZE;
+
+		continue;
+nomatch:
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 15);
+nothing:
+		;
+	}
+
+	return ret;
+}
+
+/**
+ * nft_pipapo_avx2_lookup_4b_4() - AVX2-based lookup for 4 four-bit groups
+ * @map:	Previous match result, used as initial bitmap
+ * @fill:	Destination bitmap to be filled with current match result
+ * @f:		Field, containing lookup and mapping tables
+ * @offset:	Ignore buckets before the given index, no bits are filled there
+ * @pkt:	Packet data, pointer to input nftables register
+ * @first:	If this is the first field, don't source previous result
+ * @last:	Last field: stop at the first match and return bit index
+ *
+ * See nft_pipapo_avx2_lookup_4b_2().
+ *
+ * This is used for 16-bit fields (i.e. ports).
+ *
+ * Return: -1 on no match, rule index of match if @last, otherwise first long
+ * word index to be checked next (i.e. first filled word).
+ */
+static int nft_pipapo_avx2_lookup_4b_4(unsigned long *map, unsigned long *fill,
+				       struct nft_pipapo_field *f, int offset,
+				       const u8 *pkt, bool first, bool last)
+{
+	int i, ret = -1, m256_size = f->bsize / NFT_PIPAPO_LONGS_PER_M256, b;
+	u8 pg[4] = { pkt[0] >> 4, pkt[0] & 0xf, pkt[1] >> 4, pkt[1] & 0xf };
+	unsigned long *lt = f->lt, bsize = f->bsize;
+
+	lt += offset * NFT_PIPAPO_LONGS_PER_M256;
+	for (i = offset; i < m256_size; i++, lt += NFT_PIPAPO_LONGS_PER_M256) {
+		int i_ul = i * NFT_PIPAPO_LONGS_PER_M256;
+
+		if (first) {
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(0, lt, 0, pg[0], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(1, lt, 1, pg[1], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(2, lt, 2, pg[2], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(3, lt, 3, pg[3], bsize);
+			NFT_PIPAPO_AVX2_AND(4, 0, 1);
+			NFT_PIPAPO_AVX2_AND(5, 2, 3);
+			NFT_PIPAPO_AVX2_AND(7, 4, 5);
+		} else {
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(0, lt, 0, pg[0], bsize);
+
+			NFT_PIPAPO_AVX2_LOAD(1, map[i_ul]);
+
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(2, lt, 1, pg[1], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(3, lt, 2, pg[2], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(4, lt, 3, pg[3], bsize);
+			NFT_PIPAPO_AVX2_AND(5, 0, 1);
+
+			NFT_PIPAPO_AVX2_NOMATCH_GOTO(1, nothing);
+
+			NFT_PIPAPO_AVX2_AND(6, 2, 3);
+			NFT_PIPAPO_AVX2_AND(7, 4, 5);
+			/* Stall */
+			NFT_PIPAPO_AVX2_AND(7, 6, 7);
+		}
+
+		/* Stall */
+		NFT_PIPAPO_AVX2_NOMATCH_GOTO(7, nomatch);
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 7);
+
+		b = nft_pipapo_avx2_refill(i_ul, &map[i_ul], fill, f->mt, last);
+		if (last)
+			return b;
+
+		if (unlikely(ret == -1))
+			ret = b / XSAVE_YMM_SIZE;
+
+		continue;
+nomatch:
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 15);
+nothing:
+		;
+	}
+
+	return ret;
+}
+
+/**
+ * nft_pipapo_avx2_lookup_4b_8() - AVX2-based lookup for 8 four-bit groups
+ * @map:	Previous match result, used as initial bitmap
+ * @fill:	Destination bitmap to be filled with current match result
+ * @f:		Field, containing lookup and mapping tables
+ * @offset:	Ignore buckets before the given index, no bits are filled there
+ * @pkt:	Packet data, pointer to input nftables register
+ * @first:	If this is the first field, don't source previous result
+ * @last:	Last field: stop at the first match and return bit index
+ *
+ * See nft_pipapo_avx2_lookup_4b_2().
+ *
+ * This is used for 32-bit fields (i.e. IPv4 addresses).
+ *
+ * Return: -1 on no match, rule index of match if @last, otherwise first long
+ * word index to be checked next (i.e. first filled word).
+ */
+static int nft_pipapo_avx2_lookup_4b_8(unsigned long *map, unsigned long *fill,
+				       struct nft_pipapo_field *f, int offset,
+				       const u8 *pkt, bool first, bool last)
+{
+	u8 pg[8] = {  pkt[0] >> 4,  pkt[0] & 0xf,  pkt[1] >> 4,  pkt[1] & 0xf,
+		      pkt[2] >> 4,  pkt[2] & 0xf,  pkt[3] >> 4,  pkt[3] & 0xf,
+		   };
+	int i, ret = -1, m256_size = f->bsize / NFT_PIPAPO_LONGS_PER_M256, b;
+	unsigned long *lt = f->lt, bsize = f->bsize;
+
+	lt += offset * NFT_PIPAPO_LONGS_PER_M256;
+	for (i = offset; i < m256_size; i++, lt += NFT_PIPAPO_LONGS_PER_M256) {
+		int i_ul = i * NFT_PIPAPO_LONGS_PER_M256;
+
+		if (first) {
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(0,  lt, 0, pg[0], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(1,  lt, 1, pg[1], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(2,  lt, 2, pg[2], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(3,  lt, 3, pg[3], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(4,  lt, 4, pg[4], bsize);
+			NFT_PIPAPO_AVX2_AND(5,   0,  1);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(6,  lt, 5, pg[5], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(7,  lt, 6, pg[6], bsize);
+			NFT_PIPAPO_AVX2_AND(8,   2,  3);
+			NFT_PIPAPO_AVX2_AND(9,   4,  5);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(10, lt, 7, pg[7], bsize);
+			NFT_PIPAPO_AVX2_AND(11,  6,  7);
+			NFT_PIPAPO_AVX2_AND(12,  8,  9);
+			NFT_PIPAPO_AVX2_AND(13, 10, 11);
+
+			/* Stall */
+			NFT_PIPAPO_AVX2_AND(1,  12, 13);
+		} else {
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(0,  lt, 0, pg[0], bsize);
+			NFT_PIPAPO_AVX2_LOAD(1, map[i_ul]);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(2,  lt, 1, pg[1], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(3,  lt, 2, pg[2], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(4,  lt, 3, pg[3], bsize);
+
+			NFT_PIPAPO_AVX2_NOMATCH_GOTO(1, nothing);
+
+			NFT_PIPAPO_AVX2_AND(5,   0,  1);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(6,  lt, 4, pg[4], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(7,  lt, 5, pg[5], bsize);
+			NFT_PIPAPO_AVX2_AND(8,   2,  3);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(9,  lt, 6, pg[6], bsize);
+			NFT_PIPAPO_AVX2_AND(10,  4,  5);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD4(11, lt, 7, pg[7], bsize);
+			NFT_PIPAPO_AVX2_AND(12,  6,  7);
+			NFT_PIPAPO_AVX2_AND(13,  8,  9);
+			NFT_PIPAPO_AVX2_AND(14, 10, 11);
+
+			/* Stall */
+			NFT_PIPAPO_AVX2_AND(1,  12, 13);
+			NFT_PIPAPO_AVX2_AND(1,   1, 14);
+		}
+
+		NFT_PIPAPO_AVX2_NOMATCH_GOTO(1, nomatch);
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 1);
+
+		b = nft_pipapo_avx2_refill(i_ul, &map[i_ul], fill, f->mt, last);
+		if (last)
+			return b;
+
+		if (unlikely(ret == -1))
+			ret = b / XSAVE_YMM_SIZE;
+
+		continue;
+
+nomatch:
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 15);
+nothing:
+		;
+	}
+
+	return ret;
+}
+
+/**
+ * nft_pipapo_avx2_lookup_4b_12() - AVX2-based lookup for 12 four-bit groups
+ * @map:	Previous match result, used as initial bitmap
+ * @fill:	Destination bitmap to be filled with current match result
+ * @f:		Field, containing lookup and mapping tables
+ * @offset:	Ignore buckets before the given index, no bits are filled there
+ * @pkt:	Packet data, pointer to input nftables register
+ * @first:	If this is the first field, don't source previous result
+ * @last:	Last field: stop at the first match and return bit index
+ *
+ * See nft_pipapo_avx2_lookup_4b_2().
+ *
+ * This is used for 48-bit fields (i.e. MAC addresses/EUI-48).
+ *
+ * Return: -1 on no match, rule index of match if @last, otherwise first long
+ * word index to be checked next (i.e. first filled word).
+ */
+static int nft_pipapo_avx2_lookup_4b_12(unsigned long *map, unsigned long *fill,
+				        struct nft_pipapo_field *f, int offset,
+				        const u8 *pkt, bool first, bool last)
+{
+	u8 pg[12] = {  pkt[0] >> 4,  pkt[0] & 0xf,  pkt[1] >> 4,  pkt[1] & 0xf,
+		       pkt[2] >> 4,  pkt[2] & 0xf,  pkt[3] >> 4,  pkt[3] & 0xf,
+		       pkt[4] >> 4,  pkt[4] & 0xf,  pkt[5] >> 4,  pkt[5] & 0xf,
+		    };
+	int i, ret = -1, m256_size = f->bsize / NFT_PIPAPO_LONGS_PER_M256, b;
+	unsigned long *lt = f->lt, bsize = f->bsize;
+
+	lt += offset * NFT_PIPAPO_LONGS_PER_M256;
+	for (i = offset; i < m256_size; i++, lt += NFT_PIPAPO_LONGS_PER_M256) {
+		int i_ul = i * NFT_PIPAPO_LONGS_PER_M256;
+
+		if (!first)
+			NFT_PIPAPO_AVX2_LOAD(0, map[i_ul]);
+
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(1,  lt,  0,  pg[0], bsize);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(2,  lt,  1,  pg[1], bsize);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(3,  lt,  2,  pg[2], bsize);
+
+		if (!first) {
+			NFT_PIPAPO_AVX2_NOMATCH_GOTO(0, nothing);
+			NFT_PIPAPO_AVX2_AND(1, 1, 0);
+		}
+
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(4,  lt,  3,  pg[3], bsize);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(5,  lt,  4,  pg[4], bsize);
+		NFT_PIPAPO_AVX2_AND(6,   2,  3);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(7,  lt,  5,  pg[5], bsize);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(8,  lt,  6,  pg[6], bsize);
+		NFT_PIPAPO_AVX2_AND(9,   1,  4);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(10, lt,  7,  pg[7], bsize);
+		NFT_PIPAPO_AVX2_AND(11,  5,  6);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(12, lt,  8,  pg[8], bsize);
+		NFT_PIPAPO_AVX2_AND(13,  7,  8);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(14, lt,  9,  pg[9], bsize);
+
+		NFT_PIPAPO_AVX2_AND(0,   9, 10);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(1,  lt, 10,  pg[10], bsize);
+		NFT_PIPAPO_AVX2_AND(2,  11, 12);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(3,  lt, 11,  pg[11], bsize);
+		NFT_PIPAPO_AVX2_AND(4,  13, 14);
+		NFT_PIPAPO_AVX2_AND(5,   0,  1);
+
+		NFT_PIPAPO_AVX2_AND(6,   2,  3);
+
+		/* Stalls */
+		NFT_PIPAPO_AVX2_AND(7,   4,  5);
+		NFT_PIPAPO_AVX2_AND(8,   6,  7);
+
+		NFT_PIPAPO_AVX2_NOMATCH_GOTO(8, nomatch);
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 8);
+
+		b = nft_pipapo_avx2_refill(i_ul, &map[i_ul], fill, f->mt, last);
+		if (last)
+			return b;
+
+		if (unlikely(ret == -1))
+			ret = b / XSAVE_YMM_SIZE;
+
+		continue;
+nomatch:
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 15);
+nothing:
+		;
+	}
+
+	return ret;
+}
+
+/**
+ * nft_pipapo_avx2_lookup_4b_32() - AVX2-based lookup for 32 four-bit groups
+ * @map:	Previous match result, used as initial bitmap
+ * @fill:	Destination bitmap to be filled with current match result
+ * @f:		Field, containing lookup and mapping tables
+ * @offset:	Ignore buckets before the given index, no bits are filled there
+ * @pkt:	Packet data, pointer to input nftables register
+ * @first:	If this is the first field, don't source previous result
+ * @last:	Last field: stop at the first match and return bit index
+ *
+ * See nft_pipapo_avx2_lookup_4b_2().
+ *
+ * This is used for 128-bit fields (i.e. IPv6 addresses).
+ *
+ * Return: -1 on no match, rule index of match if @last, otherwise first long
+ * word index to be checked next (i.e. first filled word).
+ */
+static int nft_pipapo_avx2_lookup_4b_32(unsigned long *map, unsigned long *fill,
+					struct nft_pipapo_field *f, int offset,
+					const u8 *pkt, bool first, bool last)
+{
+	u8 pg[32] = {  pkt[0] >> 4,  pkt[0] & 0xf,  pkt[1] >> 4,  pkt[1] & 0xf,
+		       pkt[2] >> 4,  pkt[2] & 0xf,  pkt[3] >> 4,  pkt[3] & 0xf,
+		       pkt[4] >> 4,  pkt[4] & 0xf,  pkt[5] >> 4,  pkt[5] & 0xf,
+		       pkt[6] >> 4,  pkt[6] & 0xf,  pkt[7] >> 4,  pkt[7] & 0xf,
+		       pkt[8] >> 4,  pkt[8] & 0xf,  pkt[9] >> 4,  pkt[9] & 0xf,
+		      pkt[10] >> 4, pkt[10] & 0xf, pkt[11] >> 4, pkt[11] & 0xf,
+		      pkt[12] >> 4, pkt[12] & 0xf, pkt[13] >> 4, pkt[13] & 0xf,
+		      pkt[14] >> 4, pkt[14] & 0xf, pkt[15] >> 4, pkt[15] & 0xf,
+		    };
+	int i, ret = -1, m256_size = f->bsize / NFT_PIPAPO_LONGS_PER_M256, b;
+	unsigned long *lt = f->lt, bsize = f->bsize;
+
+	lt += offset * NFT_PIPAPO_LONGS_PER_M256;
+	for (i = offset; i < m256_size; i++, lt += NFT_PIPAPO_LONGS_PER_M256) {
+		int i_ul = i * NFT_PIPAPO_LONGS_PER_M256;
+
+		if (!first)
+			NFT_PIPAPO_AVX2_LOAD(0, map[i_ul]);
+
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(1,  lt,  0,  pg[0], bsize);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(2,  lt,  1,  pg[1], bsize);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(3,  lt,  2,  pg[2], bsize);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(4,  lt,  3,  pg[3], bsize);
+		if (!first) {
+			NFT_PIPAPO_AVX2_NOMATCH_GOTO(0, nothing);
+			NFT_PIPAPO_AVX2_AND(1, 1, 0);
+		}
+
+		NFT_PIPAPO_AVX2_AND(5,   2,  3);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(6,  lt,  4,  pg[4], bsize);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(7,  lt,  5,  pg[5], bsize);
+		NFT_PIPAPO_AVX2_AND(8,   1,  4);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(9,  lt,  6,  pg[6], bsize);
+		NFT_PIPAPO_AVX2_AND(10,  5,  6);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(11, lt,  7,  pg[7], bsize);
+		NFT_PIPAPO_AVX2_AND(12,  7,  8);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(13, lt,  8,  pg[8], bsize);
+		NFT_PIPAPO_AVX2_AND(14,  9, 10);
+
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(0,  lt,  9,  pg[9], bsize);
+		NFT_PIPAPO_AVX2_AND(1,  11, 12);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(2,  lt, 10, pg[10], bsize);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(3,  lt, 11, pg[11], bsize);
+		NFT_PIPAPO_AVX2_AND(4,  13, 14);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(5,  lt, 12, pg[12], bsize);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(6,  lt, 13, pg[13], bsize);
+		NFT_PIPAPO_AVX2_AND(7,   0,  1);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(8,  lt, 14, pg[14], bsize);
+		NFT_PIPAPO_AVX2_AND(9,   2,  3);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(10, lt, 15, pg[15], bsize);
+		NFT_PIPAPO_AVX2_AND(11,  4,  5);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(12, lt, 16, pg[16], bsize);
+		NFT_PIPAPO_AVX2_AND(13,  6,  7);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(14, lt, 17, pg[17], bsize);
+
+		NFT_PIPAPO_AVX2_AND(0,   8,  9);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(1,  lt, 18, pg[18], bsize);
+		NFT_PIPAPO_AVX2_AND(2,  10, 11);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(3,  lt, 19, pg[19], bsize);
+		NFT_PIPAPO_AVX2_AND(4,  12, 13);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(5,  lt, 20, pg[20], bsize);
+		NFT_PIPAPO_AVX2_AND(6,  14,  0);
+		NFT_PIPAPO_AVX2_AND(7,   1,  2);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(8,  lt, 21, pg[21], bsize);
+		NFT_PIPAPO_AVX2_AND(9,   3,  4);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(10, lt, 22, pg[22], bsize);
+		NFT_PIPAPO_AVX2_AND(11,  5,  6);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(12, lt, 23, pg[23], bsize);
+		NFT_PIPAPO_AVX2_AND(13,  7,  8);
+
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(14, lt, 24, pg[24], bsize);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(0,  lt, 25, pg[25], bsize);
+		NFT_PIPAPO_AVX2_AND(1,   9, 10);
+		NFT_PIPAPO_AVX2_AND(2,  11, 12);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(3,  lt, 26, pg[26], bsize);
+		NFT_PIPAPO_AVX2_AND(4,  13, 14);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(5,  lt, 27, pg[27], bsize);
+		NFT_PIPAPO_AVX2_AND(6,   0,  1);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(7,  lt, 28, pg[28], bsize);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(8,  lt, 29, pg[29], bsize);
+		NFT_PIPAPO_AVX2_AND(9,   2,  3);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(10, lt, 30, pg[30], bsize);
+		NFT_PIPAPO_AVX2_AND(11,  4,  5);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD4(12, lt, 31, pg[31], bsize);
+
+		NFT_PIPAPO_AVX2_AND(0,   6,  7);
+		NFT_PIPAPO_AVX2_AND(1,   8,  9);
+		NFT_PIPAPO_AVX2_AND(2,  10, 11);
+		NFT_PIPAPO_AVX2_AND(3,  12,  0);
+
+		/* Stalls */
+		NFT_PIPAPO_AVX2_AND(4,   1,  2);
+		NFT_PIPAPO_AVX2_AND(5,   3,  4);
+
+		NFT_PIPAPO_AVX2_NOMATCH_GOTO(5, nomatch);
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 5);
+
+		b = nft_pipapo_avx2_refill(i_ul, &map[i_ul], fill, f->mt, last);
+		if (last)
+			return b;
+
+		if (unlikely(ret == -1))
+			ret = b / XSAVE_YMM_SIZE;
+
+		continue;
+nomatch:
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 15);
+nothing:
+		;
+	}
+
+	return ret;
+}
+
+/**
+ * nft_pipapo_avx2_lookup_8b_1() - AVX2-based lookup for one eight-bit group
+ * @map:	Previous match result, used as initial bitmap
+ * @fill:	Destination bitmap to be filled with current match result
+ * @f:		Field, containing lookup and mapping tables
+ * @offset:	Ignore buckets before the given index, no bits are filled there
+ * @pkt:	Packet data, pointer to input nftables register
+ * @first:	If this is the first field, don't source previous result
+ * @last:	Last field: stop at the first match and return bit index
+ *
+ * See nft_pipapo_avx2_lookup_4b_2().
+ *
+ * This is used for 8-bit fields (i.e. protocol numbers).
+ *
+ * Return: -1 on no match, rule index of match if @last, otherwise first long
+ * word index to be checked next (i.e. first filled word).
+ */
+static int nft_pipapo_avx2_lookup_8b_1(unsigned long *map, unsigned long *fill,
+				       struct nft_pipapo_field *f, int offset,
+				       const u8 *pkt, bool first, bool last)
+{
+	int i, ret = -1, m256_size = f->bsize / NFT_PIPAPO_LONGS_PER_M256, b;
+	unsigned long *lt = f->lt, bsize = f->bsize;
+
+	lt += offset * NFT_PIPAPO_LONGS_PER_M256;
+	for (i = offset; i < m256_size; i++, lt += NFT_PIPAPO_LONGS_PER_M256) {
+		int i_ul = i * NFT_PIPAPO_LONGS_PER_M256;
+
+		if (first) {
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(2, lt, 0, pkt[0], bsize);
+		} else {
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(0, lt, 0, pkt[0], bsize);
+			NFT_PIPAPO_AVX2_LOAD(1, map[i_ul]);
+			NFT_PIPAPO_AVX2_AND(2, 0, 1);
+			NFT_PIPAPO_AVX2_NOMATCH_GOTO(1, nothing);
+		}
+
+		NFT_PIPAPO_AVX2_NOMATCH_GOTO(2, nomatch);
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 2);
+
+		b = nft_pipapo_avx2_refill(i_ul, &map[i_ul], fill, f->mt, last);
+		if (last)
+			return b;
+
+		if (unlikely(ret == -1))
+			ret = b / XSAVE_YMM_SIZE;
+
+		continue;
+nomatch:
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 15);
+nothing:
+		;
+	}
+
+	return ret;
+}
+
+/**
+ * nft_pipapo_avx2_lookup_8b_2() - AVX2-based lookup for 2 eight-bit groups
+ * @map:	Previous match result, used as initial bitmap
+ * @fill:	Destination bitmap to be filled with current match result
+ * @f:		Field, containing lookup and mapping tables
+ * @offset:	Ignore buckets before the given index, no bits are filled there
+ * @pkt:	Packet data, pointer to input nftables register
+ * @first:	If this is the first field, don't source previous result
+ * @last:	Last field: stop at the first match and return bit index
+ *
+ * See nft_pipapo_avx2_lookup_4b_2().
+ *
+ * This is used for 16-bit fields (i.e. ports).
+ *
+ * Return: -1 on no match, rule index of match if @last, otherwise first long
+ * word index to be checked next (i.e. first filled word).
+ */
+static int nft_pipapo_avx2_lookup_8b_2(unsigned long *map, unsigned long *fill,
+				       struct nft_pipapo_field *f, int offset,
+				       const u8 *pkt, bool first, bool last)
+{
+	int i, ret = -1, m256_size = f->bsize / NFT_PIPAPO_LONGS_PER_M256, b;
+	unsigned long *lt = f->lt, bsize = f->bsize;
+
+	lt += offset * NFT_PIPAPO_LONGS_PER_M256;
+	for (i = offset; i < m256_size; i++, lt += NFT_PIPAPO_LONGS_PER_M256) {
+		int i_ul = i * NFT_PIPAPO_LONGS_PER_M256;
+
+		if (first) {
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(0, lt, 0, pkt[0], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(1, lt, 1, pkt[1], bsize);
+			NFT_PIPAPO_AVX2_AND(4, 0, 1);
+		} else {
+			NFT_PIPAPO_AVX2_LOAD(0, map[i_ul]);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(1, lt, 0, pkt[0], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(2, lt, 1, pkt[1], bsize);
+
+			/* Stall */
+			NFT_PIPAPO_AVX2_AND(3, 0, 1);
+			NFT_PIPAPO_AVX2_NOMATCH_GOTO(0, nothing);
+			NFT_PIPAPO_AVX2_AND(4, 3, 2);
+		}
+
+		/* Stall */
+		NFT_PIPAPO_AVX2_NOMATCH_GOTO(4, nomatch);
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 4);
+
+		b = nft_pipapo_avx2_refill(i_ul, &map[i_ul], fill, f->mt, last);
+		if (last)
+			return b;
+
+		if (unlikely(ret == -1))
+			ret = b / XSAVE_YMM_SIZE;
+
+		continue;
+nomatch:
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 15);
+nothing:
+		;
+	}
+
+	return ret;
+}
+
+/**
+ * nft_pipapo_avx2_lookup_8b_4() - AVX2-based lookup for 4 eight-bit groups
+ * @map:	Previous match result, used as initial bitmap
+ * @fill:	Destination bitmap to be filled with current match result
+ * @f:		Field, containing lookup and mapping tables
+ * @offset:	Ignore buckets before the given index, no bits are filled there
+ * @pkt:	Packet data, pointer to input nftables register
+ * @first:	If this is the first field, don't source previous result
+ * @last:	Last field: stop at the first match and return bit index
+ *
+ * See nft_pipapo_avx2_lookup_4b_2().
+ *
+ * This is used for 32-bit fields (i.e. IPv4 addresses).
+ *
+ * Return: -1 on no match, rule index of match if @last, otherwise first long
+ * word index to be checked next (i.e. first filled word).
+ */
+static int nft_pipapo_avx2_lookup_8b_4(unsigned long *map, unsigned long *fill,
+				       struct nft_pipapo_field *f, int offset,
+				       const u8 *pkt, bool first, bool last)
+{
+	int i, ret = -1, m256_size = f->bsize / NFT_PIPAPO_LONGS_PER_M256, b;
+	unsigned long *lt = f->lt, bsize = f->bsize;
+
+	lt += offset * NFT_PIPAPO_LONGS_PER_M256;
+	for (i = offset; i < m256_size; i++, lt += NFT_PIPAPO_LONGS_PER_M256) {
+		int i_ul = i * NFT_PIPAPO_LONGS_PER_M256;
+
+		if (first) {
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(0,  lt, 0, pkt[0], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(1,  lt, 1, pkt[1], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(2,  lt, 2, pkt[2], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(3,  lt, 3, pkt[3], bsize);
+
+			/* Stall */
+			NFT_PIPAPO_AVX2_AND(4, 0, 1);
+			NFT_PIPAPO_AVX2_AND(5, 2, 3);
+			NFT_PIPAPO_AVX2_AND(0, 4, 5);
+		} else {
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(0,  lt, 0, pkt[0], bsize);
+			NFT_PIPAPO_AVX2_LOAD(1, map[i_ul]);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(2,  lt, 1, pkt[1], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(3,  lt, 2, pkt[2], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(4,  lt, 3, pkt[3], bsize);
+
+			NFT_PIPAPO_AVX2_AND(5, 0, 1);
+			NFT_PIPAPO_AVX2_NOMATCH_GOTO(1, nothing);
+			NFT_PIPAPO_AVX2_AND(6, 2, 3);
+
+			/* Stall */
+			NFT_PIPAPO_AVX2_AND(7, 4, 5);
+			NFT_PIPAPO_AVX2_AND(0, 6, 7);
+		}
+
+		NFT_PIPAPO_AVX2_NOMATCH_GOTO(0, nomatch);
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 0);
+
+		b = nft_pipapo_avx2_refill(i_ul, &map[i_ul], fill, f->mt, last);
+		if (last)
+			return b;
+
+		if (unlikely(ret == -1))
+			ret = b / XSAVE_YMM_SIZE;
+
+		continue;
+
+nomatch:
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 15);
+nothing:
+		;
+	}
+
+	return ret;
+}
+
+/**
+ * nft_pipapo_avx2_lookup_8b_6() - AVX2-based lookup for 6 eight-bit groups
+ * @map:	Previous match result, used as initial bitmap
+ * @fill:	Destination bitmap to be filled with current match result
+ * @f:		Field, containing lookup and mapping tables
+ * @offset:	Ignore buckets before the given index, no bits are filled there
+ * @pkt:	Packet data, pointer to input nftables register
+ * @first:	If this is the first field, don't source previous result
+ * @last:	Last field: stop at the first match and return bit index
+ *
+ * See nft_pipapo_avx2_lookup_4b_2().
+ *
+ * This is used for 48-bit fields (i.e. MAC addresses/EUI-48).
+ *
+ * Return: -1 on no match, rule index of match if @last, otherwise first long
+ * word index to be checked next (i.e. first filled word).
+ */
+static int nft_pipapo_avx2_lookup_8b_6(unsigned long *map, unsigned long *fill,
+				       struct nft_pipapo_field *f, int offset,
+				       const u8 *pkt, bool first, bool last)
+{
+	int i, ret = -1, m256_size = f->bsize / NFT_PIPAPO_LONGS_PER_M256, b;
+	unsigned long *lt = f->lt, bsize = f->bsize;
+
+	lt += offset * NFT_PIPAPO_LONGS_PER_M256;
+	for (i = offset; i < m256_size; i++, lt += NFT_PIPAPO_LONGS_PER_M256) {
+		int i_ul = i * NFT_PIPAPO_LONGS_PER_M256;
+
+		if (first) {
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(0,  lt, 0, pkt[0], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(1,  lt, 1, pkt[1], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(2,  lt, 2, pkt[2], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(3,  lt, 3, pkt[3], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(4,  lt, 4, pkt[4], bsize);
+
+			NFT_PIPAPO_AVX2_AND(5, 0, 1);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(6,  lt, 6, pkt[5], bsize);
+			NFT_PIPAPO_AVX2_AND(7, 2, 3);
+
+			/* Stall */
+			NFT_PIPAPO_AVX2_AND(0, 4, 5);
+			NFT_PIPAPO_AVX2_AND(1, 6, 7);
+			NFT_PIPAPO_AVX2_AND(4, 0, 1);
+		} else {
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(0,  lt, 0, pkt[0], bsize);
+			NFT_PIPAPO_AVX2_LOAD(1, map[i_ul]);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(2,  lt, 1, pkt[1], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(3,  lt, 2, pkt[2], bsize);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(4,  lt, 3, pkt[3], bsize);
+
+			NFT_PIPAPO_AVX2_AND(5, 0, 1);
+			NFT_PIPAPO_AVX2_NOMATCH_GOTO(1, nothing);
+
+			NFT_PIPAPO_AVX2_AND(6, 2, 3);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(7,  lt, 4, pkt[4], bsize);
+			NFT_PIPAPO_AVX2_AND(0, 4, 5);
+			NFT_PIPAPO_AVX2_BUCKET_LOAD8(1,  lt, 5, pkt[5], bsize);
+			NFT_PIPAPO_AVX2_AND(2, 6, 7);
+
+			/* Stall */
+			NFT_PIPAPO_AVX2_AND(3, 0, 1);
+			NFT_PIPAPO_AVX2_AND(4, 2, 3);
+		}
+
+		NFT_PIPAPO_AVX2_NOMATCH_GOTO(4, nomatch);
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 4);
+
+		b = nft_pipapo_avx2_refill(i_ul, &map[i_ul], fill, f->mt, last);
+		if (last)
+			return b;
+
+		if (unlikely(ret == -1))
+			ret = b / XSAVE_YMM_SIZE;
+
+		continue;
+
+nomatch:
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 15);
+nothing:
+		;
+	}
+
+	return ret;
+}
+
+/**
+ * nft_pipapo_avx2_lookup_8b_16() - AVX2-based lookup for 16 eight-bit groups
+ * @map:	Previous match result, used as initial bitmap
+ * @fill:	Destination bitmap to be filled with current match result
+ * @f:		Field, containing lookup and mapping tables
+ * @offset:	Ignore buckets before the given index, no bits are filled there
+ * @pkt:	Packet data, pointer to input nftables register
+ * @first:	If this is the first field, don't source previous result
+ * @last:	Last field: stop at the first match and return bit index
+ *
+ * See nft_pipapo_avx2_lookup_4b_2().
+ *
+ * This is used for 128-bit fields (i.e. IPv6 addresses).
+ *
+ * Return: -1 on no match, rule index of match if @last, otherwise first long
+ * word index to be checked next (i.e. first filled word).
+ */
+static int nft_pipapo_avx2_lookup_8b_16(unsigned long *map, unsigned long *fill,
+					struct nft_pipapo_field *f, int offset,
+					const u8 *pkt, bool first, bool last)
+{
+	int i, ret = -1, m256_size = f->bsize / NFT_PIPAPO_LONGS_PER_M256, b;
+	unsigned long *lt = f->lt, bsize = f->bsize;
+
+	lt += offset * NFT_PIPAPO_LONGS_PER_M256;
+	for (i = offset; i < m256_size; i++, lt += NFT_PIPAPO_LONGS_PER_M256) {
+		int i_ul = i * NFT_PIPAPO_LONGS_PER_M256;
+
+		if (!first)
+			NFT_PIPAPO_AVX2_LOAD(0, map[i_ul]);
+
+		NFT_PIPAPO_AVX2_BUCKET_LOAD8(1, lt,  0,  pkt[0], bsize);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD8(2, lt,  1,  pkt[1], bsize);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD8(3, lt,  2,  pkt[2], bsize);
+		if (!first) {
+			NFT_PIPAPO_AVX2_NOMATCH_GOTO(0, nothing);
+			NFT_PIPAPO_AVX2_AND(1, 1, 0);
+		}
+		NFT_PIPAPO_AVX2_BUCKET_LOAD8(4, lt,  3,  pkt[3], bsize);
+
+		NFT_PIPAPO_AVX2_BUCKET_LOAD8(5, lt,  4,  pkt[4], bsize);
+		NFT_PIPAPO_AVX2_AND(6, 1, 2);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD8(7, lt,  5,  pkt[5], bsize);
+		NFT_PIPAPO_AVX2_AND(0, 3, 4);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD8(1, lt,  6,  pkt[6], bsize);
+
+		NFT_PIPAPO_AVX2_BUCKET_LOAD8(2, lt,  7,  pkt[7], bsize);
+		NFT_PIPAPO_AVX2_AND(3, 5, 6);
+		NFT_PIPAPO_AVX2_AND(4, 0, 1);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD8(5, lt,  8,  pkt[8], bsize);
+
+		NFT_PIPAPO_AVX2_AND(6, 2, 3);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD8(7, lt,  9,  pkt[9], bsize);
+		NFT_PIPAPO_AVX2_AND(0, 4, 5);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD8(1, lt, 10, pkt[10], bsize);
+		NFT_PIPAPO_AVX2_AND(2, 6, 7);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD8(3, lt, 11, pkt[11], bsize);
+		NFT_PIPAPO_AVX2_AND(4, 0, 1);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD8(5, lt, 12, pkt[12], bsize);
+		NFT_PIPAPO_AVX2_AND(6, 2, 3);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD8(7, lt, 13, pkt[13], bsize);
+		NFT_PIPAPO_AVX2_AND(0, 4, 5);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD8(1, lt, 14, pkt[14], bsize);
+		NFT_PIPAPO_AVX2_AND(2, 6, 7);
+		NFT_PIPAPO_AVX2_BUCKET_LOAD8(3, lt, 15, pkt[15], bsize);
+		NFT_PIPAPO_AVX2_AND(4, 0, 1);
+
+		/* Stall */
+		NFT_PIPAPO_AVX2_AND(5, 2, 3);
+		NFT_PIPAPO_AVX2_AND(6, 4, 5);
+
+		NFT_PIPAPO_AVX2_NOMATCH_GOTO(6, nomatch);
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 6);
+
+		b = nft_pipapo_avx2_refill(i_ul, &map[i_ul], fill, f->mt, last);
+		if (last)
+			return b;
+
+		if (unlikely(ret == -1))
+			ret = b / XSAVE_YMM_SIZE;
+
+		continue;
+
+nomatch:
+		NFT_PIPAPO_AVX2_STORE(map[i_ul], 15);
+nothing:
+		;
+	}
+
+	return ret;
+}
+
+/**
+ * nft_pipapo_avx2_lookup_slow() - Fallback function for uncommon field sizes
+ * @map:	Previous match result, used as initial bitmap
+ * @fill:	Destination bitmap to be filled with current match result
+ * @f:		Field, containing lookup and mapping tables
+ * @offset:	Ignore buckets before the given index, no bits are filled there
+ * @pkt:	Packet data, pointer to input nftables register
+ * @first:	If this is the first field, don't source previous result
+ * @last:	Last field: stop at the first match and return bit index
+ *
+ * This function should never be called, but is provided for the case the field
+ * size doesn't match any of the known data types. Matching rate is
+ * substantially lower than AVX2 routines.
+ *
+ * Return: -1 on no match, rule index of match if @last, otherwise first long
+ * word index to be checked next (i.e. first filled word).
+ */
+static int nft_pipapo_avx2_lookup_slow(unsigned long *map, unsigned long *fill,
+					struct nft_pipapo_field *f, int offset,
+					const u8 *pkt, bool first, bool last)
+{
+	unsigned long *lt = f->lt, bsize = f->bsize;
+	int i, ret = -1, b;
+
+	lt += offset * NFT_PIPAPO_LONGS_PER_M256;
+
+	if (first)
+		memset(map, 0xff, bsize * sizeof(*map));
+
+	for (i = offset; i < bsize; i++) {
+		if (f->bb == 8)
+			pipapo_and_field_buckets_8bit(f, map, pkt);
+		else
+			pipapo_and_field_buckets_4bit(f, map, pkt);
+		NFT_PIPAPO_GROUP_BITS_ARE_8_OR_4;
+
+		b = pipapo_refill(map, bsize, f->rules, fill, f->mt, last);
+
+		if (last)
+			return b;
+
+		if (ret == -1)
+			ret = b / XSAVE_YMM_SIZE;
+	}
+
+	return ret;
+}
+
+/**
+ * nft_pipapo_avx2_estimate() - Set size, space and lookup complexity
+ * @desc:	Set description, element count and field description used
+ * @features:	Flags: NFT_SET_INTERVAL needs to be there
+ * @est:	Storage for estimation data
+ *
+ * Return: true if set is compatible and AVX2 available, false otherwise.
+ */
+bool nft_pipapo_avx2_estimate(const struct nft_set_desc *desc, u32 features,
+			      struct nft_set_estimate *est)
+{
+	if (!(features & NFT_SET_INTERVAL) || desc->field_count <= 1)
+		return false;
+
+	if (!boot_cpu_has(X86_FEATURE_AVX2) || !boot_cpu_has(X86_FEATURE_AVX))
+		return false;
+
+	est->size = pipapo_estimate_size(desc);
+	if (!est->size)
+		return false;
+
+	est->lookup = NFT_SET_CLASS_O_LOG_N;
+
+	est->space = NFT_SET_CLASS_O_N;
+
+	return true;
+}
+
+/**
+ * nft_pipapo_avx2_lookup() - Lookup function for AVX2 implementation
+ * @net:	Network namespace
+ * @set:	nftables API set representation
+ * @elem:	nftables API element representation containing key data
+ * @ext:	nftables API extension pointer, filled with matching reference
+ *
+ * For more details, see DOC: Theory of Operation in nft_set_pipapo.c.
+ *
+ * This implementation exploits the repetitive characteristic of the algorithm
+ * to provide a fast, vectorised version using the AVX2 SIMD instruction set.
+ *
+ * Return: true on match, false otherwise.
+ */
+bool nft_pipapo_avx2_lookup(const struct net *net, const struct nft_set *set,
+			    const u32 *key, const struct nft_set_ext **ext)
+{
+	struct nft_pipapo *priv = nft_set_priv(set);
+	unsigned long *res, *fill, *scratch;
+	u8 genmask = nft_genmask_cur(net);
+	const u8 *rp = (const u8 *)key;
+	struct nft_pipapo_match *m;
+	struct nft_pipapo_field *f;
+	bool map_index;
+	int i, ret = 0;
+
+	m = rcu_dereference(priv->match);
+
+	/* This also protects access to all data related to scratch maps */
+	kernel_fpu_begin();
+
+	scratch = *raw_cpu_ptr(m->scratch_aligned);
+	if (unlikely(!scratch)) {
+		kernel_fpu_end();
+		return false;
+	}
+	map_index = raw_cpu_read(nft_pipapo_avx2_scratch_index);
+
+	res  = scratch + (map_index ? m->bsize_max : 0);
+	fill = scratch + (map_index ? 0 : m->bsize_max);
+
+	/* Starting map doesn't need to be set for this implementation */
+
+	nft_pipapo_avx2_prepare();
+
+next_match:
+	nft_pipapo_for_each_field(f, i, m) {
+		bool last = i == m->field_count - 1, first = !i;
+
+#define NFT_SET_PIPAPO_AVX2_LOOKUP(b, n)				\
+		(ret = nft_pipapo_avx2_lookup_##b##b_##n(res, fill, f,	\
+							 ret, rp,	\
+							 first, last))
+
+		if (likely(f->bb == 8)) {
+			if (f->groups == 1) {
+				NFT_SET_PIPAPO_AVX2_LOOKUP(8, 1);
+			} else if (f->groups == 2) {
+				NFT_SET_PIPAPO_AVX2_LOOKUP(8, 2);
+			} else if (f->groups == 4) {
+				NFT_SET_PIPAPO_AVX2_LOOKUP(8, 4);
+			} else if (f->groups == 6) {
+				NFT_SET_PIPAPO_AVX2_LOOKUP(8, 6);
+			} else if (f->groups == 16) {
+				NFT_SET_PIPAPO_AVX2_LOOKUP(8, 16);
+			} else {
+				ret = nft_pipapo_avx2_lookup_slow(res, fill, f,
+								  ret, rp,
+								  first, last);
+			}
+		} else {
+			if (f->groups == 2) {
+				NFT_SET_PIPAPO_AVX2_LOOKUP(4, 2);
+			} else if (f->groups == 4) {
+				NFT_SET_PIPAPO_AVX2_LOOKUP(4, 4);
+			} else if (f->groups == 8) {
+				NFT_SET_PIPAPO_AVX2_LOOKUP(4, 8);
+			} else if (f->groups == 12) {
+				NFT_SET_PIPAPO_AVX2_LOOKUP(4, 12);
+			} else if (f->groups == 32) {
+				NFT_SET_PIPAPO_AVX2_LOOKUP(4, 32);
+			} else {
+				ret = nft_pipapo_avx2_lookup_slow(res, fill, f,
+								  ret, rp,
+								  first, last);
+			}
+		}
+		NFT_PIPAPO_GROUP_BITS_ARE_8_OR_4;
+
+#undef NFT_SET_PIPAPO_AVX2_LOOKUP
+
+		if (ret < 0)
+			goto out;
+
+		if (last) {
+			*ext = &f->mt[ret].e->ext;
+			if (unlikely(nft_set_elem_expired(*ext) ||
+				     !nft_set_elem_active(*ext, genmask))) {
+				ret = 0;
+				goto next_match;
+			}
+
+			goto out;
+		}
+
+		swap(res, fill);
+		rp += NFT_PIPAPO_GROUPS_PADDED_SIZE(f);
+	}
+
+out:
+	if (i % 2)
+		raw_cpu_write(nft_pipapo_avx2_scratch_index, !map_index);
+	kernel_fpu_end();
+
+	return ret >= 0;
+}
diff --git a/net/netfilter/nft_set_pipapo_avx2.h b/net/netfilter/nft_set_pipapo_avx2.h
new file mode 100644
index 000000000000..396caf7bfca8
--- /dev/null
+++ b/net/netfilter/nft_set_pipapo_avx2.h
@@ -0,0 +1,14 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+#ifndef _NFT_SET_PIPAPO_AVX2_H
+
+#ifdef CONFIG_AS_AVX2
+#include <asm/fpu/xstate.h>
+#define NFT_PIPAPO_ALIGN	(XSAVE_YMM_SIZE / BITS_PER_BYTE)
+
+bool nft_pipapo_avx2_lookup(const struct net *net, const struct nft_set *set,
+			    const u32 *key, const struct nft_set_ext **ext);
+bool nft_pipapo_avx2_estimate(const struct nft_set_desc *desc, u32 features,
+			      struct nft_set_estimate *est);
+#endif /* CONFIG_AS_AVX2 */
+
+#endif /* _NFT_SET_PIPAPO_AVX2_H */
-- 
cgit v1.2.3-59-g8ed1b


From a7fc936804084145e2a6374e23744defdc948e09 Mon Sep 17 00:00:00 2001
From: Pablo Neira Ayuso <pablo@netfilter.org>
Date: Wed, 11 Mar 2020 15:30:12 +0100
Subject: netfilter: nf_tables: add nft_set_elem_expr_alloc()

Add helper function to create stateful expression.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
 include/net/netfilter/nf_tables.h |  4 ++++
 net/netfilter/nf_tables_api.c     | 30 ++++++++++++++++++++++++++++++
 net/netfilter/nft_dynset.c        | 15 ++-------------
 3 files changed, 36 insertions(+), 13 deletions(-)

(limited to 'include')

diff --git a/include/net/netfilter/nf_tables.h b/include/net/netfilter/nf_tables.h
index 01f48b5e6bf2..52b872a9a2cd 100644
--- a/include/net/netfilter/nf_tables.h
+++ b/include/net/netfilter/nf_tables.h
@@ -666,6 +666,10 @@ static inline struct nft_object **nft_set_ext_obj(const struct nft_set_ext *ext)
 	return nft_set_ext(ext, NFT_SET_EXT_OBJREF);
 }
 
+struct nft_expr *nft_set_elem_expr_alloc(const struct nft_ctx *ctx,
+					 const struct nft_set *set,
+					 const struct nlattr *attr);
+
 void *nft_set_elem_init(const struct nft_set *set,
 			const struct nft_set_ext_tmpl *tmpl,
 			const u32 *key, const u32 *key_end, const u32 *data,
diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c
index 7187b0d76d9a..1605a728eb2c 100644
--- a/net/netfilter/nf_tables_api.c
+++ b/net/netfilter/nf_tables_api.c
@@ -4779,6 +4779,36 @@ static struct nft_trans *nft_trans_elem_alloc(struct nft_ctx *ctx,
 	return trans;
 }
 
+struct nft_expr *nft_set_elem_expr_alloc(const struct nft_ctx *ctx,
+					 const struct nft_set *set,
+					 const struct nlattr *attr)
+{
+	struct nft_expr *expr;
+	int err;
+
+	expr = nft_expr_init(ctx, attr);
+	if (IS_ERR(expr))
+		return expr;
+
+	err = -EOPNOTSUPP;
+	if (!(expr->ops->type->flags & NFT_EXPR_STATEFUL))
+		goto err_set_elem_expr;
+
+	if (expr->ops->type->flags & NFT_EXPR_GC) {
+		if (set->flags & NFT_SET_TIMEOUT)
+			goto err_set_elem_expr;
+		if (!set->ops->gc_init)
+			goto err_set_elem_expr;
+		set->ops->gc_init(set);
+	}
+
+	return expr;
+
+err_set_elem_expr:
+	nft_expr_destroy(ctx, expr);
+	return ERR_PTR(err);
+}
+
 void *nft_set_elem_init(const struct nft_set *set,
 			const struct nft_set_ext_tmpl *tmpl,
 			const u32 *key, const u32 *key_end,
diff --git a/net/netfilter/nft_dynset.c b/net/netfilter/nft_dynset.c
index 683785225a3e..e106cf1c5b8b 100644
--- a/net/netfilter/nft_dynset.c
+++ b/net/netfilter/nft_dynset.c
@@ -206,21 +206,10 @@ static int nft_dynset_init(const struct nft_ctx *ctx,
 		if (!(set->flags & NFT_SET_EVAL))
 			return -EINVAL;
 
-		priv->expr = nft_expr_init(ctx, tb[NFTA_DYNSET_EXPR]);
+		priv->expr = nft_set_elem_expr_alloc(ctx, set,
+						     tb[NFTA_DYNSET_EXPR]);
 		if (IS_ERR(priv->expr))
 			return PTR_ERR(priv->expr);
-
-		err = -EOPNOTSUPP;
-		if (!(priv->expr->ops->type->flags & NFT_EXPR_STATEFUL))
-			goto err1;
-
-		if (priv->expr->ops->type->flags & NFT_EXPR_GC) {
-			if (set->flags & NFT_SET_TIMEOUT)
-				goto err1;
-			if (!set->ops->gc_init)
-				goto err1;
-			set->ops->gc_init(set);
-		}
 	}
 
 	nft_set_ext_prepare(&priv->tmpl);
-- 
cgit v1.2.3-59-g8ed1b


From 795a6d6b42244d0aabec645c6ddd05c29cfd39d6 Mon Sep 17 00:00:00 2001
From: Pablo Neira Ayuso <pablo@netfilter.org>
Date: Wed, 11 Mar 2020 15:30:13 +0100
Subject: netfilter: nf_tables: statify nft_expr_init()

Not exposed anymore to modules, statify this function.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
 include/net/netfilter/nf_tables.h | 2 --
 net/netfilter/nf_tables_api.c     | 4 ++--
 2 files changed, 2 insertions(+), 4 deletions(-)

(limited to 'include')

diff --git a/include/net/netfilter/nf_tables.h b/include/net/netfilter/nf_tables.h
index 52b872a9a2cd..025ec148ea4f 100644
--- a/include/net/netfilter/nf_tables.h
+++ b/include/net/netfilter/nf_tables.h
@@ -846,8 +846,6 @@ static inline void *nft_expr_priv(const struct nft_expr *expr)
 	return (void *)expr->data;
 }
 
-struct nft_expr *nft_expr_init(const struct nft_ctx *ctx,
-			       const struct nlattr *nla);
 void nft_expr_destroy(const struct nft_ctx *ctx, struct nft_expr *expr);
 int nft_expr_dump(struct sk_buff *skb, unsigned int attr,
 		  const struct nft_expr *expr);
diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c
index 1605a728eb2c..bc7a33f0cd06 100644
--- a/net/netfilter/nf_tables_api.c
+++ b/net/netfilter/nf_tables_api.c
@@ -2523,8 +2523,8 @@ static void nf_tables_expr_destroy(const struct nft_ctx *ctx,
 	module_put(type->owner);
 }
 
-struct nft_expr *nft_expr_init(const struct nft_ctx *ctx,
-			       const struct nlattr *nla)
+static struct nft_expr *nft_expr_init(const struct nft_ctx *ctx,
+				      const struct nlattr *nla)
 {
 	struct nft_expr_info info;
 	struct nft_expr *expr;
-- 
cgit v1.2.3-59-g8ed1b


From 76adfafecab5e4aae460e9781720b20d115bb096 Mon Sep 17 00:00:00 2001
From: Pablo Neira Ayuso <pablo@netfilter.org>
Date: Wed, 11 Mar 2020 15:30:15 +0100
Subject: netfilter: nf_tables: add nft_set_elem_update_expr() helper function

This helper function runs the eval path of the stateful expression
of an existing set element.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
 include/net/netfilter/nf_tables.h | 12 ++++++++++++
 net/netfilter/nft_dynset.c        |  8 +-------
 2 files changed, 13 insertions(+), 7 deletions(-)

(limited to 'include')

diff --git a/include/net/netfilter/nf_tables.h b/include/net/netfilter/nf_tables.h
index 025ec148ea4f..5d80e09f8148 100644
--- a/include/net/netfilter/nf_tables.h
+++ b/include/net/netfilter/nf_tables.h
@@ -890,6 +890,18 @@ static inline struct nft_userdata *nft_userdata(const struct nft_rule *rule)
 	return (void *)&rule->data[rule->dlen];
 }
 
+static inline void nft_set_elem_update_expr(const struct nft_set_ext *ext,
+					    struct nft_regs *regs,
+					    const struct nft_pktinfo *pkt)
+{
+	struct nft_expr *expr;
+
+	if (nft_set_ext_exists(ext, NFT_SET_EXT_EXPR)) {
+		expr = nft_set_ext_expr(ext);
+		expr->ops->eval(expr, regs, pkt);
+	}
+}
+
 /*
  * The last pointer isn't really necessary, but the compiler isn't able to
  * determine that the result of nft_expr_last() is always the same since it
diff --git a/net/netfilter/nft_dynset.c b/net/netfilter/nft_dynset.c
index e106cf1c5b8b..46ab28ec4b53 100644
--- a/net/netfilter/nft_dynset.c
+++ b/net/netfilter/nft_dynset.c
@@ -81,7 +81,6 @@ void nft_dynset_eval(const struct nft_expr *expr,
 	const struct nft_dynset *priv = nft_expr_priv(expr);
 	struct nft_set *set = priv->set;
 	const struct nft_set_ext *ext;
-	const struct nft_expr *sexpr;
 	u64 timeout;
 
 	if (priv->op == NFT_DYNSET_OP_DELETE) {
@@ -91,18 +90,13 @@ void nft_dynset_eval(const struct nft_expr *expr,
 
 	if (set->ops->update(set, &regs->data[priv->sreg_key], nft_dynset_new,
 			     expr, regs, &ext)) {
-		sexpr = NULL;
-		if (nft_set_ext_exists(ext, NFT_SET_EXT_EXPR))
-			sexpr = nft_set_ext_expr(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) = get_jiffies_64() + timeout;
 		}
 
-		if (sexpr != NULL)
-			sexpr->ops->eval(sexpr, regs, pkt);
+		nft_set_elem_update_expr(ext, regs, pkt);
 
 		if (priv->invert)
 			regs->verdict.code = NFT_BREAK;
-- 
cgit v1.2.3-59-g8ed1b


From b030f194aed290705426c62e501201c0739405c5 Mon Sep 17 00:00:00 2001
From: Lukas Wunner <lukas@wunner.de>
Date: Wed, 11 Mar 2020 12:59:01 +0100
Subject: netfilter: Rename ingress hook include file

Prepare for addition of a netfilter egress hook by renaming
<linux/netfilter_ingress.h> to <linux/netfilter_netdev.h>.

The egress hook also necessitates a refactoring of the include file,
but that is done in a separate commit to ease reviewing.

No functional change intended.

Signed-off-by: Lukas Wunner <lukas@wunner.de>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
 include/linux/netfilter_ingress.h | 58 ---------------------------------------
 include/linux/netfilter_netdev.h  | 58 +++++++++++++++++++++++++++++++++++++++
 net/core/dev.c                    |  2 +-
 3 files changed, 59 insertions(+), 59 deletions(-)
 delete mode 100644 include/linux/netfilter_ingress.h
 create mode 100644 include/linux/netfilter_netdev.h

(limited to 'include')

diff --git a/include/linux/netfilter_ingress.h b/include/linux/netfilter_ingress.h
deleted file mode 100644
index a13774be2eb5..000000000000
--- a/include/linux/netfilter_ingress.h
+++ /dev/null
@@ -1,58 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0 */
-#ifndef _NETFILTER_INGRESS_H_
-#define _NETFILTER_INGRESS_H_
-
-#include <linux/netfilter.h>
-#include <linux/netdevice.h>
-
-#ifdef CONFIG_NETFILTER_INGRESS
-static inline bool nf_hook_ingress_active(const struct sk_buff *skb)
-{
-#ifdef CONFIG_JUMP_LABEL
-	if (!static_key_false(&nf_hooks_needed[NFPROTO_NETDEV][NF_NETDEV_INGRESS]))
-		return false;
-#endif
-	return rcu_access_pointer(skb->dev->nf_hooks_ingress);
-}
-
-/* caller must hold rcu_read_lock */
-static inline int nf_hook_ingress(struct sk_buff *skb)
-{
-	struct nf_hook_entries *e = rcu_dereference(skb->dev->nf_hooks_ingress);
-	struct nf_hook_state state;
-	int ret;
-
-	/* Must recheck the ingress hook head, in the event it became NULL
-	 * after the check in nf_hook_ingress_active evaluated to true.
-	 */
-	if (unlikely(!e))
-		return 0;
-
-	nf_hook_state_init(&state, NF_NETDEV_INGRESS,
-			   NFPROTO_NETDEV, skb->dev, NULL, NULL,
-			   dev_net(skb->dev), NULL);
-	ret = nf_hook_slow(skb, &state, e, 0);
-	if (ret == 0)
-		return -1;
-
-	return ret;
-}
-
-static inline void nf_hook_ingress_init(struct net_device *dev)
-{
-	RCU_INIT_POINTER(dev->nf_hooks_ingress, NULL);
-}
-#else /* CONFIG_NETFILTER_INGRESS */
-static inline int nf_hook_ingress_active(struct sk_buff *skb)
-{
-	return 0;
-}
-
-static inline int nf_hook_ingress(struct sk_buff *skb)
-{
-	return 0;
-}
-
-static inline void nf_hook_ingress_init(struct net_device *dev) {}
-#endif /* CONFIG_NETFILTER_INGRESS */
-#endif /* _NETFILTER_INGRESS_H_ */
diff --git a/include/linux/netfilter_netdev.h b/include/linux/netfilter_netdev.h
new file mode 100644
index 000000000000..a13774be2eb5
--- /dev/null
+++ b/include/linux/netfilter_netdev.h
@@ -0,0 +1,58 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _NETFILTER_INGRESS_H_
+#define _NETFILTER_INGRESS_H_
+
+#include <linux/netfilter.h>
+#include <linux/netdevice.h>
+
+#ifdef CONFIG_NETFILTER_INGRESS
+static inline bool nf_hook_ingress_active(const struct sk_buff *skb)
+{
+#ifdef CONFIG_JUMP_LABEL
+	if (!static_key_false(&nf_hooks_needed[NFPROTO_NETDEV][NF_NETDEV_INGRESS]))
+		return false;
+#endif
+	return rcu_access_pointer(skb->dev->nf_hooks_ingress);
+}
+
+/* caller must hold rcu_read_lock */
+static inline int nf_hook_ingress(struct sk_buff *skb)
+{
+	struct nf_hook_entries *e = rcu_dereference(skb->dev->nf_hooks_ingress);
+	struct nf_hook_state state;
+	int ret;
+
+	/* Must recheck the ingress hook head, in the event it became NULL
+	 * after the check in nf_hook_ingress_active evaluated to true.
+	 */
+	if (unlikely(!e))
+		return 0;
+
+	nf_hook_state_init(&state, NF_NETDEV_INGRESS,
+			   NFPROTO_NETDEV, skb->dev, NULL, NULL,
+			   dev_net(skb->dev), NULL);
+	ret = nf_hook_slow(skb, &state, e, 0);
+	if (ret == 0)
+		return -1;
+
+	return ret;
+}
+
+static inline void nf_hook_ingress_init(struct net_device *dev)
+{
+	RCU_INIT_POINTER(dev->nf_hooks_ingress, NULL);
+}
+#else /* CONFIG_NETFILTER_INGRESS */
+static inline int nf_hook_ingress_active(struct sk_buff *skb)
+{
+	return 0;
+}
+
+static inline int nf_hook_ingress(struct sk_buff *skb)
+{
+	return 0;
+}
+
+static inline void nf_hook_ingress_init(struct net_device *dev) {}
+#endif /* CONFIG_NETFILTER_INGRESS */
+#endif /* _NETFILTER_INGRESS_H_ */
diff --git a/net/core/dev.c b/net/core/dev.c
index d84541c24446..b1ce1c942b54 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -135,7 +135,7 @@
 #include <linux/if_macvlan.h>
 #include <linux/errqueue.h>
 #include <linux/hrtimer.h>
-#include <linux/netfilter_ingress.h>
+#include <linux/netfilter_netdev.h>
 #include <linux/crash_dump.h>
 #include <linux/sctp.h>
 #include <net/udp_tunnel.h>
-- 
cgit v1.2.3-59-g8ed1b


From 5418d3881e1f5d2cf9c1076eb8bd85770393a0e8 Mon Sep 17 00:00:00 2001
From: Lukas Wunner <lukas@wunner.de>
Date: Wed, 11 Mar 2020 12:59:02 +0100
Subject: netfilter: Generalize ingress hook

Prepare for addition of a netfilter egress hook by generalizing the
ingress hook introduced by commit e687ad60af09 ("netfilter: add
netfilter ingress hook after handle_ing() under unique static key").

In particular, rename and refactor the ingress hook's static inlines
such that they can be reused for an egress hook.

No functional change intended.

Signed-off-by: Lukas Wunner <lukas@wunner.de>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
 include/linux/netfilter_netdev.h | 45 +++++++++++++++++++++++++++-------------
 net/core/dev.c                   |  2 +-
 2 files changed, 32 insertions(+), 15 deletions(-)

(limited to 'include')

diff --git a/include/linux/netfilter_netdev.h b/include/linux/netfilter_netdev.h
index a13774be2eb5..49e26479642e 100644
--- a/include/linux/netfilter_netdev.h
+++ b/include/linux/netfilter_netdev.h
@@ -1,34 +1,37 @@
 /* SPDX-License-Identifier: GPL-2.0 */
-#ifndef _NETFILTER_INGRESS_H_
-#define _NETFILTER_INGRESS_H_
+#ifndef _NETFILTER_NETDEV_H_
+#define _NETFILTER_NETDEV_H_
 
 #include <linux/netfilter.h>
 #include <linux/netdevice.h>
 
-#ifdef CONFIG_NETFILTER_INGRESS
-static inline bool nf_hook_ingress_active(const struct sk_buff *skb)
+#ifdef CONFIG_NETFILTER
+static __always_inline bool nf_hook_netdev_active(enum nf_dev_hooks hooknum,
+					  struct nf_hook_entries __rcu *hooks)
 {
 #ifdef CONFIG_JUMP_LABEL
-	if (!static_key_false(&nf_hooks_needed[NFPROTO_NETDEV][NF_NETDEV_INGRESS]))
+	if (!static_key_false(&nf_hooks_needed[NFPROTO_NETDEV][hooknum]))
 		return false;
 #endif
-	return rcu_access_pointer(skb->dev->nf_hooks_ingress);
+	return rcu_access_pointer(hooks);
 }
 
 /* caller must hold rcu_read_lock */
-static inline int nf_hook_ingress(struct sk_buff *skb)
+static __always_inline int nf_hook_netdev(struct sk_buff *skb,
+					  enum nf_dev_hooks hooknum,
+					  struct nf_hook_entries __rcu *hooks)
 {
-	struct nf_hook_entries *e = rcu_dereference(skb->dev->nf_hooks_ingress);
+	struct nf_hook_entries *e = rcu_dereference(hooks);
 	struct nf_hook_state state;
 	int ret;
 
-	/* Must recheck the ingress hook head, in the event it became NULL
-	 * after the check in nf_hook_ingress_active evaluated to true.
+	/* Must recheck the hook head, in the event it became NULL
+	 * after the check in nf_hook_netdev_active evaluated to true.
 	 */
 	if (unlikely(!e))
 		return 0;
 
-	nf_hook_state_init(&state, NF_NETDEV_INGRESS,
+	nf_hook_state_init(&state, hooknum,
 			   NFPROTO_NETDEV, skb->dev, NULL, NULL,
 			   dev_net(skb->dev), NULL);
 	ret = nf_hook_slow(skb, &state, e, 0);
@@ -37,10 +40,26 @@ static inline int nf_hook_ingress(struct sk_buff *skb)
 
 	return ret;
 }
+#endif /* CONFIG_NETFILTER */
 
-static inline void nf_hook_ingress_init(struct net_device *dev)
+static inline void nf_hook_netdev_init(struct net_device *dev)
 {
+#ifdef CONFIG_NETFILTER_INGRESS
 	RCU_INIT_POINTER(dev->nf_hooks_ingress, NULL);
+#endif
+}
+
+#ifdef CONFIG_NETFILTER_INGRESS
+static inline bool nf_hook_ingress_active(const struct sk_buff *skb)
+{
+	return nf_hook_netdev_active(NF_NETDEV_INGRESS,
+				     skb->dev->nf_hooks_ingress);
+}
+
+static inline int nf_hook_ingress(struct sk_buff *skb)
+{
+	return nf_hook_netdev(skb, NF_NETDEV_INGRESS,
+			      skb->dev->nf_hooks_ingress);
 }
 #else /* CONFIG_NETFILTER_INGRESS */
 static inline int nf_hook_ingress_active(struct sk_buff *skb)
@@ -52,7 +71,5 @@ static inline int nf_hook_ingress(struct sk_buff *skb)
 {
 	return 0;
 }
-
-static inline void nf_hook_ingress_init(struct net_device *dev) {}
 #endif /* CONFIG_NETFILTER_INGRESS */
 #endif /* _NETFILTER_INGRESS_H_ */
diff --git a/net/core/dev.c b/net/core/dev.c
index b1ce1c942b54..13d562f67e9c 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -9846,7 +9846,7 @@ struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
 	if (!dev->ethtool_ops)
 		dev->ethtool_ops = &default_ethtool_ops;
 
-	nf_hook_ingress_init(dev);
+	nf_hook_netdev_init(dev);
 
 	return dev;
 
-- 
cgit v1.2.3-59-g8ed1b


From 8537f78647c072bdb1a5dbe32e1c7e5b13ff1258 Mon Sep 17 00:00:00 2001
From: Lukas Wunner <lukas@wunner.de>
Date: Wed, 11 Mar 2020 12:59:03 +0100
Subject: netfilter: Introduce egress hook

Commit e687ad60af09 ("netfilter: add netfilter ingress hook after
handle_ing() under unique static key") introduced the ability to
classify packets on ingress.

Allow the same on egress.  Position the hook immediately before a packet
is handed to tc and then sent out on an interface, thereby mirroring the
ingress order.  This order allows marking packets in the netfilter
egress hook and subsequently using the mark in tc.  Another benefit of
this order is consistency with a lot of existing documentation which
says that egress tc is performed after netfilter hooks.

Egress hooks already exist for the most common protocols, such as
NF_INET_LOCAL_OUT or NF_ARP_OUT, and those are to be preferred because
they are executed earlier during packet processing.  However for more
exotic protocols, there is currently no provision to apply netfilter on
egress.  A common workaround is to enslave the interface to a bridge and
use ebtables, or to resort to tc.  But when the ingress hook was
introduced, consensus was that users should be given the choice to use
netfilter or tc, whichever tool suits their needs best:
https://lore.kernel.org/netdev/20150430153317.GA3230@salvia/
This hook is also useful for NAT46/NAT64, tunneling and filtering of
locally generated af_packet traffic such as dhclient.

There have also been occasional user requests for a netfilter egress
hook in the past, e.g.:
https://www.spinics.net/lists/netfilter/msg50038.html

Performance measurements with pktgen surprisingly show a speedup rather
than a slowdown with this commit:

* Without this commit:
  Result: OK: 34240933(c34238375+d2558) usec, 100000000 (60byte,0frags)
  2920481pps 1401Mb/sec (1401830880bps) errors: 0

* With this commit:
  Result: OK: 33997299(c33994193+d3106) usec, 100000000 (60byte,0frags)
  2941410pps 1411Mb/sec (1411876800bps) errors: 0

* Without this commit + tc egress:
  Result: OK: 39022386(c39019547+d2839) usec, 100000000 (60byte,0frags)
  2562631pps 1230Mb/sec (1230062880bps) errors: 0

* With this commit + tc egress:
  Result: OK: 37604447(c37601877+d2570) usec, 100000000 (60byte,0frags)
  2659259pps 1276Mb/sec (1276444320bps) errors: 0

* With this commit + nft egress:
  Result: OK: 41436689(c41434088+d2600) usec, 100000000 (60byte,0frags)
  2413320pps 1158Mb/sec (1158393600bps) errors: 0

Tested on a bare-metal Core i7-3615QM, each measurement was performed
three times to verify that the numbers are stable.

Commands to perform a measurement:
modprobe pktgen
echo "add_device lo@3" > /proc/net/pktgen/kpktgend_3
samples/pktgen/pktgen_bench_xmit_mode_queue_xmit.sh -i 'lo@3' -n 100000000

Commands for testing tc egress:
tc qdisc add dev lo clsact
tc filter add dev lo egress protocol ip prio 1 u32 match ip dst 4.3.2.1/32

Commands for testing nft egress:
nft add table netdev t
nft add chain netdev t co \{ type filter hook egress device lo priority 0 \; \}
nft add rule netdev t co ip daddr 4.3.2.1/32 drop

All testing was performed on the loopback interface to avoid distorting
measurements by the packet handling in the low-level Ethernet driver.

Signed-off-by: Lukas Wunner <lukas@wunner.de>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
 include/linux/netdevice.h        |  4 ++++
 include/linux/netfilter_netdev.h | 27 +++++++++++++++++++++++++++
 include/uapi/linux/netfilter.h   |  1 +
 net/core/dev.c                   | 23 ++++++++++++++++++++---
 net/netfilter/Kconfig            |  8 ++++++++
 net/netfilter/core.c             | 24 ++++++++++++++++++++----
 net/netfilter/nft_chain_filter.c |  4 +++-
 7 files changed, 83 insertions(+), 8 deletions(-)

(limited to 'include')

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 654808bfad83..15f1e32b430c 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -1751,6 +1751,7 @@ enum netdev_priv_flags {
  *	@xps_maps:	XXX: need comments on this one
  *	@miniq_egress:		clsact qdisc specific data for
  *				egress processing
+ *	@nf_hooks_egress:	netfilter hooks executed for egress packets
  *	@qdisc_hash:		qdisc hash table
  *	@watchdog_timeo:	Represents the timeout that is used by
  *				the watchdog (see dev_watchdog())
@@ -2026,6 +2027,9 @@ struct net_device {
 #ifdef CONFIG_NET_CLS_ACT
 	struct mini_Qdisc __rcu	*miniq_egress;
 #endif
+#ifdef CONFIG_NETFILTER_EGRESS
+	struct nf_hook_entries __rcu *nf_hooks_egress;
+#endif
 
 #ifdef CONFIG_NET_SCHED
 	DECLARE_HASHTABLE	(qdisc_hash, 4);
diff --git a/include/linux/netfilter_netdev.h b/include/linux/netfilter_netdev.h
index 49e26479642e..92d3611a782e 100644
--- a/include/linux/netfilter_netdev.h
+++ b/include/linux/netfilter_netdev.h
@@ -47,6 +47,9 @@ static inline void nf_hook_netdev_init(struct net_device *dev)
 #ifdef CONFIG_NETFILTER_INGRESS
 	RCU_INIT_POINTER(dev->nf_hooks_ingress, NULL);
 #endif
+#ifdef CONFIG_NETFILTER_EGRESS
+	RCU_INIT_POINTER(dev->nf_hooks_egress, NULL);
+#endif
 }
 
 #ifdef CONFIG_NETFILTER_INGRESS
@@ -72,4 +75,28 @@ static inline int nf_hook_ingress(struct sk_buff *skb)
 	return 0;
 }
 #endif /* CONFIG_NETFILTER_INGRESS */
+
+#ifdef CONFIG_NETFILTER_EGRESS
+static inline bool nf_hook_egress_active(const struct sk_buff *skb)
+{
+	return nf_hook_netdev_active(NF_NETDEV_EGRESS,
+				     skb->dev->nf_hooks_egress);
+}
+
+static inline int nf_hook_egress(struct sk_buff *skb)
+{
+	return nf_hook_netdev(skb, NF_NETDEV_EGRESS,
+			      skb->dev->nf_hooks_egress);
+}
+#else /* CONFIG_NETFILTER_EGRESS */
+static inline int nf_hook_egress_active(struct sk_buff *skb)
+{
+	return 0;
+}
+
+static inline int nf_hook_egress(struct sk_buff *skb)
+{
+	return 0;
+}
+#endif /* CONFIG_NETFILTER_EGRESS */
 #endif /* _NETFILTER_INGRESS_H_ */
diff --git a/include/uapi/linux/netfilter.h b/include/uapi/linux/netfilter.h
index ca9e63d6e0e4..d1616574c54f 100644
--- a/include/uapi/linux/netfilter.h
+++ b/include/uapi/linux/netfilter.h
@@ -50,6 +50,7 @@ enum nf_inet_hooks {
 
 enum nf_dev_hooks {
 	NF_NETDEV_INGRESS,
+	NF_NETDEV_EGRESS,
 	NF_NETDEV_NUMHOOKS
 };
 
diff --git a/net/core/dev.c b/net/core/dev.c
index 13d562f67e9c..a2da72a77c20 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -3773,6 +3773,7 @@ EXPORT_SYMBOL(dev_loopback_xmit);
 static struct sk_buff *
 sch_handle_egress(struct sk_buff *skb, int *ret, struct net_device *dev)
 {
+#ifdef CONFIG_NET_CLS_ACT
 	struct mini_Qdisc *miniq = rcu_dereference_bh(dev->miniq_egress);
 	struct tcf_result cl_res;
 
@@ -3806,11 +3807,24 @@ sch_handle_egress(struct sk_buff *skb, int *ret, struct net_device *dev)
 	default:
 		break;
 	}
-
+#endif /* CONFIG_NET_CLS_ACT */
 	return skb;
 }
 #endif /* CONFIG_NET_EGRESS */
 
+static inline int nf_egress(struct sk_buff *skb)
+{
+	if (nf_hook_egress_active(skb)) {
+		int ret;
+
+		rcu_read_lock();
+		ret = nf_hook_egress(skb);
+		rcu_read_unlock();
+		return ret;
+	}
+	return 0;
+}
+
 #ifdef CONFIG_XPS
 static int __get_xps_queue_idx(struct net_device *dev, struct sk_buff *skb,
 			       struct xps_dev_maps *dev_maps, unsigned int tci)
@@ -3997,13 +4011,16 @@ static int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev)
 	qdisc_pkt_len_init(skb);
 #ifdef CONFIG_NET_CLS_ACT
 	skb->tc_at_ingress = 0;
-# ifdef CONFIG_NET_EGRESS
+#endif
+#ifdef CONFIG_NET_EGRESS
 	if (static_branch_unlikely(&egress_needed_key)) {
+		if (nf_egress(skb) < 0)
+			goto out;
+
 		skb = sch_handle_egress(skb, &rc, dev);
 		if (!skb)
 			goto out;
 	}
-# endif
 #endif
 	/* If device/qdisc don't need skb->dst, release it right now while
 	 * its hot in this cpu cache.
diff --git a/net/netfilter/Kconfig b/net/netfilter/Kconfig
index 468fea1aebba..f4c68f60f241 100644
--- a/net/netfilter/Kconfig
+++ b/net/netfilter/Kconfig
@@ -10,6 +10,14 @@ config NETFILTER_INGRESS
 	  This allows you to classify packets from ingress using the Netfilter
 	  infrastructure.
 
+config NETFILTER_EGRESS
+	bool "Netfilter egress support"
+	default y
+	select NET_EGRESS
+	help
+	  This allows you to classify packets before transmission using the
+	  Netfilter infrastructure.
+
 config NETFILTER_NETLINK
 	tristate
 
diff --git a/net/netfilter/core.c b/net/netfilter/core.c
index 78f046ec506f..85e9c959aba7 100644
--- a/net/netfilter/core.c
+++ b/net/netfilter/core.c
@@ -306,6 +306,12 @@ nf_hook_entry_head(struct net *net, int pf, unsigned int hooknum,
 		if (dev && dev_net(dev) == net)
 			return &dev->nf_hooks_ingress;
 	}
+#endif
+#ifdef CONFIG_NETFILTER_EGRESS
+	if (hooknum == NF_NETDEV_EGRESS) {
+		if (dev && dev_net(dev) == net)
+			return &dev->nf_hooks_egress;
+	}
 #endif
 	WARN_ON_ONCE(1);
 	return NULL;
@@ -318,11 +324,13 @@ static int __nf_register_net_hook(struct net *net, int pf,
 	struct nf_hook_entries __rcu **pp;
 
 	if (pf == NFPROTO_NETDEV) {
-#ifndef CONFIG_NETFILTER_INGRESS
-		if (reg->hooknum == NF_NETDEV_INGRESS)
+		if ((!IS_ENABLED(CONFIG_NETFILTER_INGRESS) &&
+		     reg->hooknum == NF_NETDEV_INGRESS) ||
+		    (!IS_ENABLED(CONFIG_NETFILTER_EGRESS) &&
+		     reg->hooknum == NF_NETDEV_EGRESS))
 			return -EOPNOTSUPP;
-#endif
-		if (reg->hooknum != NF_NETDEV_INGRESS ||
+		if ((reg->hooknum != NF_NETDEV_INGRESS &&
+		     reg->hooknum != NF_NETDEV_EGRESS) ||
 		    !reg->dev || dev_net(reg->dev) != net)
 			return -EINVAL;
 	}
@@ -348,6 +356,10 @@ static int __nf_register_net_hook(struct net *net, int pf,
 	if (pf == NFPROTO_NETDEV && reg->hooknum == NF_NETDEV_INGRESS)
 		net_inc_ingress_queue();
 #endif
+#ifdef CONFIG_NETFILTER_EGRESS
+	if (pf == NFPROTO_NETDEV && reg->hooknum == NF_NETDEV_EGRESS)
+		net_inc_egress_queue();
+#endif
 #ifdef CONFIG_JUMP_LABEL
 	static_key_slow_inc(&nf_hooks_needed[pf][reg->hooknum]);
 #endif
@@ -406,6 +418,10 @@ static void __nf_unregister_net_hook(struct net *net, int pf,
 		if (pf == NFPROTO_NETDEV && reg->hooknum == NF_NETDEV_INGRESS)
 			net_dec_ingress_queue();
 #endif
+#ifdef CONFIG_NETFILTER_EGRESS
+		if (pf == NFPROTO_NETDEV && reg->hooknum == NF_NETDEV_EGRESS)
+			net_dec_egress_queue();
+#endif
 #ifdef CONFIG_JUMP_LABEL
 		static_key_slow_dec(&nf_hooks_needed[pf][reg->hooknum]);
 #endif
diff --git a/net/netfilter/nft_chain_filter.c b/net/netfilter/nft_chain_filter.c
index c78d01bc02e9..67ce6dbb5496 100644
--- a/net/netfilter/nft_chain_filter.c
+++ b/net/netfilter/nft_chain_filter.c
@@ -277,9 +277,11 @@ static const struct nft_chain_type nft_chain_filter_netdev = {
 	.name		= "filter",
 	.type		= NFT_CHAIN_T_DEFAULT,
 	.family		= NFPROTO_NETDEV,
-	.hook_mask	= (1 << NF_NETDEV_INGRESS),
+	.hook_mask	= (1 << NF_NETDEV_INGRESS) |
+			  (1 << NF_NETDEV_EGRESS),
 	.hooks		= {
 		[NF_NETDEV_INGRESS]	= nft_do_chain_netdev,
+		[NF_NETDEV_EGRESS]	= nft_do_chain_netdev,
 	},
 };
 
-- 
cgit v1.2.3-59-g8ed1b