2 * Copyright (c) 2007-2009 Patrick McHardy <kaber@trash.net>
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License version 2 as
6 * published by the Free Software Foundation.
8 * Development of this code funded by Astaro AG (http://www.astaro.com/)
11 #include <linux/module.h>
12 #include <linux/init.h>
13 #include <linux/list.h>
14 #include <linux/skbuff.h>
15 #include <linux/netlink.h>
16 #include <linux/netfilter.h>
17 #include <linux/netfilter/nfnetlink.h>
18 #include <linux/netfilter/nf_tables.h>
19 #include <net/netfilter/nf_tables_core.h>
20 #include <net/netfilter/nf_tables.h>
21 #include <net/net_namespace.h>
24 static LIST_HEAD(nf_tables_expressions);
27 * nft_register_afinfo - register nf_tables address family info
29 * @afi: address family info to register
31 * Register the address family for use with nf_tables. Returns zero on
32 * success or a negative errno code otherwise.
34 int nft_register_afinfo(struct net *net, struct nft_af_info *afi)
36 INIT_LIST_HEAD(&afi->tables);
37 nfnl_lock(NFNL_SUBSYS_NFTABLES);
38 list_add_tail_rcu(&afi->list, &net->nft.af_info);
39 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
42 EXPORT_SYMBOL_GPL(nft_register_afinfo);
44 static void __nft_release_afinfo(struct net *net, struct nft_af_info *afi);
47 * nft_unregister_afinfo - unregister nf_tables address family info
49 * @afi: address family info to unregister
51 * Unregister the address family for use with nf_tables.
53 void nft_unregister_afinfo(struct net *net, struct nft_af_info *afi)
55 nfnl_lock(NFNL_SUBSYS_NFTABLES);
56 __nft_release_afinfo(net, afi);
57 list_del_rcu(&afi->list);
58 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
60 EXPORT_SYMBOL_GPL(nft_unregister_afinfo);
62 static struct nft_af_info *nft_afinfo_lookup(struct net *net, int family)
64 struct nft_af_info *afi;
66 list_for_each_entry(afi, &net->nft.af_info, list) {
67 if (afi->family == family)
73 static struct nft_af_info *
74 nf_tables_afinfo_lookup(struct net *net, int family, bool autoload)
76 struct nft_af_info *afi;
78 afi = nft_afinfo_lookup(net, family);
83 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
84 request_module("nft-afinfo-%u", family);
85 nfnl_lock(NFNL_SUBSYS_NFTABLES);
86 afi = nft_afinfo_lookup(net, family);
88 return ERR_PTR(-EAGAIN);
91 return ERR_PTR(-EAFNOSUPPORT);
94 static void nft_ctx_init(struct nft_ctx *ctx,
96 const struct sk_buff *skb,
97 const struct nlmsghdr *nlh,
98 struct nft_af_info *afi,
99 struct nft_table *table,
100 struct nft_chain *chain,
101 const struct nlattr * const *nla)
108 ctx->portid = NETLINK_CB(skb).portid;
109 ctx->report = nlmsg_report(nlh);
110 ctx->seq = nlh->nlmsg_seq;
113 static struct nft_trans *nft_trans_alloc(struct nft_ctx *ctx, int msg_type,
116 struct nft_trans *trans;
118 trans = kzalloc(sizeof(struct nft_trans) + size, GFP_KERNEL);
122 trans->msg_type = msg_type;
128 static void nft_trans_destroy(struct nft_trans *trans)
130 list_del(&trans->list);
134 static int nft_register_basechain(struct nft_base_chain *basechain,
135 unsigned int hook_nops)
137 struct net *net = read_pnet(&basechain->pnet);
139 if (basechain->flags & NFT_BASECHAIN_DISABLED)
142 return nf_register_net_hooks(net, basechain->ops, hook_nops);
145 static void nft_unregister_basechain(struct nft_base_chain *basechain,
146 unsigned int hook_nops)
148 struct net *net = read_pnet(&basechain->pnet);
150 if (basechain->flags & NFT_BASECHAIN_DISABLED)
153 nf_unregister_net_hooks(net, basechain->ops, hook_nops);
156 static int nf_tables_register_hooks(const struct nft_table *table,
157 struct nft_chain *chain,
158 unsigned int hook_nops)
160 if (table->flags & NFT_TABLE_F_DORMANT ||
161 !(chain->flags & NFT_BASE_CHAIN))
164 return nft_register_basechain(nft_base_chain(chain), hook_nops);
167 static void nf_tables_unregister_hooks(const struct nft_table *table,
168 struct nft_chain *chain,
169 unsigned int hook_nops)
171 if (table->flags & NFT_TABLE_F_DORMANT ||
172 !(chain->flags & NFT_BASE_CHAIN))
175 nft_unregister_basechain(nft_base_chain(chain), hook_nops);
178 /* Internal table flags */
179 #define NFT_TABLE_INACTIVE (1 << 15)
181 static int nft_trans_table_add(struct nft_ctx *ctx, int msg_type)
183 struct nft_trans *trans;
185 trans = nft_trans_alloc(ctx, msg_type, sizeof(struct nft_trans_table));
189 if (msg_type == NFT_MSG_NEWTABLE)
190 ctx->table->flags |= NFT_TABLE_INACTIVE;
192 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
196 static int nft_deltable(struct nft_ctx *ctx)
200 err = nft_trans_table_add(ctx, NFT_MSG_DELTABLE);
204 list_del_rcu(&ctx->table->list);
208 static int nft_trans_chain_add(struct nft_ctx *ctx, int msg_type)
210 struct nft_trans *trans;
212 trans = nft_trans_alloc(ctx, msg_type, sizeof(struct nft_trans_chain));
216 if (msg_type == NFT_MSG_NEWCHAIN)
217 ctx->chain->flags |= NFT_CHAIN_INACTIVE;
219 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
223 static int nft_delchain(struct nft_ctx *ctx)
227 err = nft_trans_chain_add(ctx, NFT_MSG_DELCHAIN);
232 list_del_rcu(&ctx->chain->list);
238 nft_rule_is_active(struct net *net, const struct nft_rule *rule)
240 return (rule->genmask & nft_genmask_cur(net)) == 0;
244 nft_rule_is_active_next(struct net *net, const struct nft_rule *rule)
246 return (rule->genmask & nft_genmask_next(net)) == 0;
250 nft_rule_activate_next(struct net *net, struct nft_rule *rule)
252 /* Now inactive, will be active in the future */
253 rule->genmask = nft_genmask_cur(net);
257 nft_rule_deactivate_next(struct net *net, struct nft_rule *rule)
259 rule->genmask = nft_genmask_next(net);
262 static inline void nft_rule_clear(struct net *net, struct nft_rule *rule)
264 rule->genmask &= ~nft_genmask_next(net);
268 nf_tables_delrule_deactivate(struct nft_ctx *ctx, struct nft_rule *rule)
270 /* You cannot delete the same rule twice */
271 if (nft_rule_is_active_next(ctx->net, rule)) {
272 nft_rule_deactivate_next(ctx->net, rule);
279 static struct nft_trans *nft_trans_rule_add(struct nft_ctx *ctx, int msg_type,
280 struct nft_rule *rule)
282 struct nft_trans *trans;
284 trans = nft_trans_alloc(ctx, msg_type, sizeof(struct nft_trans_rule));
288 nft_trans_rule(trans) = rule;
289 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
294 static int nft_delrule(struct nft_ctx *ctx, struct nft_rule *rule)
296 struct nft_trans *trans;
299 trans = nft_trans_rule_add(ctx, NFT_MSG_DELRULE, rule);
303 err = nf_tables_delrule_deactivate(ctx, rule);
305 nft_trans_destroy(trans);
312 static int nft_delrule_by_chain(struct nft_ctx *ctx)
314 struct nft_rule *rule;
317 list_for_each_entry(rule, &ctx->chain->rules, list) {
318 err = nft_delrule(ctx, rule);
325 /* Internal set flag */
326 #define NFT_SET_INACTIVE (1 << 15)
328 static int nft_trans_set_add(struct nft_ctx *ctx, int msg_type,
331 struct nft_trans *trans;
333 trans = nft_trans_alloc(ctx, msg_type, sizeof(struct nft_trans_set));
337 if (msg_type == NFT_MSG_NEWSET && ctx->nla[NFTA_SET_ID] != NULL) {
338 nft_trans_set_id(trans) =
339 ntohl(nla_get_be32(ctx->nla[NFTA_SET_ID]));
340 set->flags |= NFT_SET_INACTIVE;
342 nft_trans_set(trans) = set;
343 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
348 static int nft_delset(struct nft_ctx *ctx, struct nft_set *set)
352 err = nft_trans_set_add(ctx, NFT_MSG_DELSET, set);
356 list_del_rcu(&set->list);
366 static struct nft_table *nft_table_lookup(const struct nft_af_info *afi,
367 const struct nlattr *nla)
369 struct nft_table *table;
371 list_for_each_entry(table, &afi->tables, list) {
372 if (!nla_strcmp(nla, table->name))
378 static struct nft_table *nf_tables_table_lookup(const struct nft_af_info *afi,
379 const struct nlattr *nla)
381 struct nft_table *table;
384 return ERR_PTR(-EINVAL);
386 table = nft_table_lookup(afi, nla);
390 return ERR_PTR(-ENOENT);
393 static inline u64 nf_tables_alloc_handle(struct nft_table *table)
395 return ++table->hgenerator;
398 static const struct nf_chain_type *chain_type[AF_MAX][NFT_CHAIN_T_MAX];
400 static const struct nf_chain_type *
401 __nf_tables_chain_type_lookup(int family, const struct nlattr *nla)
405 for (i = 0; i < NFT_CHAIN_T_MAX; i++) {
406 if (chain_type[family][i] != NULL &&
407 !nla_strcmp(nla, chain_type[family][i]->name))
408 return chain_type[family][i];
413 static const struct nf_chain_type *
414 nf_tables_chain_type_lookup(const struct nft_af_info *afi,
415 const struct nlattr *nla,
418 const struct nf_chain_type *type;
420 type = __nf_tables_chain_type_lookup(afi->family, nla);
423 #ifdef CONFIG_MODULES
425 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
426 request_module("nft-chain-%u-%.*s", afi->family,
427 nla_len(nla), (const char *)nla_data(nla));
428 nfnl_lock(NFNL_SUBSYS_NFTABLES);
429 type = __nf_tables_chain_type_lookup(afi->family, nla);
431 return ERR_PTR(-EAGAIN);
434 return ERR_PTR(-ENOENT);
437 static const struct nla_policy nft_table_policy[NFTA_TABLE_MAX + 1] = {
438 [NFTA_TABLE_NAME] = { .type = NLA_STRING,
439 .len = NFT_TABLE_MAXNAMELEN - 1 },
440 [NFTA_TABLE_FLAGS] = { .type = NLA_U32 },
443 static int nf_tables_fill_table_info(struct sk_buff *skb, struct net *net,
444 u32 portid, u32 seq, int event, u32 flags,
445 int family, const struct nft_table *table)
447 struct nlmsghdr *nlh;
448 struct nfgenmsg *nfmsg;
450 event |= NFNL_SUBSYS_NFTABLES << 8;
451 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct nfgenmsg), flags);
453 goto nla_put_failure;
455 nfmsg = nlmsg_data(nlh);
456 nfmsg->nfgen_family = family;
457 nfmsg->version = NFNETLINK_V0;
458 nfmsg->res_id = htons(net->nft.base_seq & 0xffff);
460 if (nla_put_string(skb, NFTA_TABLE_NAME, table->name) ||
461 nla_put_be32(skb, NFTA_TABLE_FLAGS, htonl(table->flags)) ||
462 nla_put_be32(skb, NFTA_TABLE_USE, htonl(table->use)))
463 goto nla_put_failure;
469 nlmsg_trim(skb, nlh);
473 static int nf_tables_table_notify(const struct nft_ctx *ctx, int event)
479 !nfnetlink_has_listeners(ctx->net, NFNLGRP_NFTABLES))
483 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
487 err = nf_tables_fill_table_info(skb, ctx->net, ctx->portid, ctx->seq,
488 event, 0, ctx->afi->family, ctx->table);
494 err = nfnetlink_send(skb, ctx->net, ctx->portid, NFNLGRP_NFTABLES,
495 ctx->report, GFP_KERNEL);
498 nfnetlink_set_err(ctx->net, ctx->portid, NFNLGRP_NFTABLES,
504 static int nf_tables_dump_tables(struct sk_buff *skb,
505 struct netlink_callback *cb)
507 const struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
508 const struct nft_af_info *afi;
509 const struct nft_table *table;
510 unsigned int idx = 0, s_idx = cb->args[0];
511 struct net *net = sock_net(skb->sk);
512 int family = nfmsg->nfgen_family;
515 cb->seq = net->nft.base_seq;
517 list_for_each_entry_rcu(afi, &net->nft.af_info, list) {
518 if (family != NFPROTO_UNSPEC && family != afi->family)
521 list_for_each_entry_rcu(table, &afi->tables, list) {
525 memset(&cb->args[1], 0,
526 sizeof(cb->args) - sizeof(cb->args[0]));
527 if (nf_tables_fill_table_info(skb, net,
528 NETLINK_CB(cb->skb).portid,
532 afi->family, table) < 0)
535 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
546 static int nf_tables_gettable(struct net *net, struct sock *nlsk,
547 struct sk_buff *skb, const struct nlmsghdr *nlh,
548 const struct nlattr * const nla[])
550 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
551 const struct nft_af_info *afi;
552 const struct nft_table *table;
553 struct sk_buff *skb2;
554 int family = nfmsg->nfgen_family;
557 if (nlh->nlmsg_flags & NLM_F_DUMP) {
558 struct netlink_dump_control c = {
559 .dump = nf_tables_dump_tables,
561 return netlink_dump_start(nlsk, skb, nlh, &c);
564 afi = nf_tables_afinfo_lookup(net, family, false);
568 table = nf_tables_table_lookup(afi, nla[NFTA_TABLE_NAME]);
570 return PTR_ERR(table);
571 if (table->flags & NFT_TABLE_INACTIVE)
574 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
578 err = nf_tables_fill_table_info(skb2, net, NETLINK_CB(skb).portid,
579 nlh->nlmsg_seq, NFT_MSG_NEWTABLE, 0,
584 return nlmsg_unicast(nlsk, skb2, NETLINK_CB(skb).portid);
591 static int nf_tables_table_enable(const struct nft_af_info *afi,
592 struct nft_table *table)
594 struct nft_chain *chain;
597 list_for_each_entry(chain, &table->chains, list) {
598 if (!(chain->flags & NFT_BASE_CHAIN))
601 err = nft_register_basechain(nft_base_chain(chain), afi->nops);
609 list_for_each_entry(chain, &table->chains, list) {
610 if (!(chain->flags & NFT_BASE_CHAIN))
616 nft_unregister_basechain(nft_base_chain(chain), afi->nops);
621 static void nf_tables_table_disable(const struct nft_af_info *afi,
622 struct nft_table *table)
624 struct nft_chain *chain;
626 list_for_each_entry(chain, &table->chains, list) {
627 if (chain->flags & NFT_BASE_CHAIN)
628 nft_unregister_basechain(nft_base_chain(chain),
633 static int nf_tables_updtable(struct nft_ctx *ctx)
635 struct nft_trans *trans;
639 if (!ctx->nla[NFTA_TABLE_FLAGS])
642 flags = ntohl(nla_get_be32(ctx->nla[NFTA_TABLE_FLAGS]));
643 if (flags & ~NFT_TABLE_F_DORMANT)
646 if (flags == ctx->table->flags)
649 trans = nft_trans_alloc(ctx, NFT_MSG_NEWTABLE,
650 sizeof(struct nft_trans_table));
654 if ((flags & NFT_TABLE_F_DORMANT) &&
655 !(ctx->table->flags & NFT_TABLE_F_DORMANT)) {
656 nft_trans_table_enable(trans) = false;
657 } else if (!(flags & NFT_TABLE_F_DORMANT) &&
658 ctx->table->flags & NFT_TABLE_F_DORMANT) {
659 ret = nf_tables_table_enable(ctx->afi, ctx->table);
661 ctx->table->flags &= ~NFT_TABLE_F_DORMANT;
662 nft_trans_table_enable(trans) = true;
668 nft_trans_table_update(trans) = true;
669 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
672 nft_trans_destroy(trans);
676 static int nf_tables_newtable(struct net *net, struct sock *nlsk,
677 struct sk_buff *skb, const struct nlmsghdr *nlh,
678 const struct nlattr * const nla[])
680 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
681 const struct nlattr *name;
682 struct nft_af_info *afi;
683 struct nft_table *table;
684 int family = nfmsg->nfgen_family;
689 afi = nf_tables_afinfo_lookup(net, family, true);
693 name = nla[NFTA_TABLE_NAME];
694 table = nf_tables_table_lookup(afi, name);
696 if (PTR_ERR(table) != -ENOENT)
697 return PTR_ERR(table);
702 if (table->flags & NFT_TABLE_INACTIVE)
704 if (nlh->nlmsg_flags & NLM_F_EXCL)
706 if (nlh->nlmsg_flags & NLM_F_REPLACE)
709 nft_ctx_init(&ctx, net, skb, nlh, afi, table, NULL, nla);
710 return nf_tables_updtable(&ctx);
713 if (nla[NFTA_TABLE_FLAGS]) {
714 flags = ntohl(nla_get_be32(nla[NFTA_TABLE_FLAGS]));
715 if (flags & ~NFT_TABLE_F_DORMANT)
720 if (!try_module_get(afi->owner))
724 table = kzalloc(sizeof(*table), GFP_KERNEL);
728 nla_strlcpy(table->name, name, NFT_TABLE_MAXNAMELEN);
729 INIT_LIST_HEAD(&table->chains);
730 INIT_LIST_HEAD(&table->sets);
731 table->flags = flags;
733 nft_ctx_init(&ctx, net, skb, nlh, afi, table, NULL, nla);
734 err = nft_trans_table_add(&ctx, NFT_MSG_NEWTABLE);
738 list_add_tail_rcu(&table->list, &afi->tables);
743 module_put(afi->owner);
748 static int nft_flush_table(struct nft_ctx *ctx)
751 struct nft_chain *chain, *nc;
752 struct nft_set *set, *ns;
754 list_for_each_entry(chain, &ctx->table->chains, list) {
757 err = nft_delrule_by_chain(ctx);
762 list_for_each_entry_safe(set, ns, &ctx->table->sets, list) {
763 if (set->flags & NFT_SET_ANONYMOUS &&
764 !list_empty(&set->bindings))
767 err = nft_delset(ctx, set);
772 list_for_each_entry_safe(chain, nc, &ctx->table->chains, list) {
775 err = nft_delchain(ctx);
780 err = nft_deltable(ctx);
785 static int nft_flush(struct nft_ctx *ctx, int family)
787 struct nft_af_info *afi;
788 struct nft_table *table, *nt;
789 const struct nlattr * const *nla = ctx->nla;
792 list_for_each_entry(afi, &ctx->net->nft.af_info, list) {
793 if (family != AF_UNSPEC && afi->family != family)
797 list_for_each_entry_safe(table, nt, &afi->tables, list) {
798 if (nla[NFTA_TABLE_NAME] &&
799 nla_strcmp(nla[NFTA_TABLE_NAME], table->name) != 0)
804 err = nft_flush_table(ctx);
813 static int nf_tables_deltable(struct net *net, struct sock *nlsk,
814 struct sk_buff *skb, const struct nlmsghdr *nlh,
815 const struct nlattr * const nla[])
817 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
818 struct nft_af_info *afi;
819 struct nft_table *table;
820 int family = nfmsg->nfgen_family;
823 nft_ctx_init(&ctx, net, skb, nlh, NULL, NULL, NULL, nla);
824 if (family == AF_UNSPEC || nla[NFTA_TABLE_NAME] == NULL)
825 return nft_flush(&ctx, family);
827 afi = nf_tables_afinfo_lookup(net, family, false);
831 table = nf_tables_table_lookup(afi, nla[NFTA_TABLE_NAME]);
833 return PTR_ERR(table);
838 return nft_flush_table(&ctx);
841 static void nf_tables_table_destroy(struct nft_ctx *ctx)
843 BUG_ON(ctx->table->use > 0);
846 module_put(ctx->afi->owner);
849 int nft_register_chain_type(const struct nf_chain_type *ctype)
853 nfnl_lock(NFNL_SUBSYS_NFTABLES);
854 if (chain_type[ctype->family][ctype->type] != NULL) {
858 chain_type[ctype->family][ctype->type] = ctype;
860 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
863 EXPORT_SYMBOL_GPL(nft_register_chain_type);
865 void nft_unregister_chain_type(const struct nf_chain_type *ctype)
867 nfnl_lock(NFNL_SUBSYS_NFTABLES);
868 chain_type[ctype->family][ctype->type] = NULL;
869 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
871 EXPORT_SYMBOL_GPL(nft_unregister_chain_type);
877 static struct nft_chain *
878 nf_tables_chain_lookup_byhandle(const struct nft_table *table, u64 handle)
880 struct nft_chain *chain;
882 list_for_each_entry(chain, &table->chains, list) {
883 if (chain->handle == handle)
887 return ERR_PTR(-ENOENT);
890 static struct nft_chain *nf_tables_chain_lookup(const struct nft_table *table,
891 const struct nlattr *nla)
893 struct nft_chain *chain;
896 return ERR_PTR(-EINVAL);
898 list_for_each_entry(chain, &table->chains, list) {
899 if (!nla_strcmp(nla, chain->name))
903 return ERR_PTR(-ENOENT);
906 static const struct nla_policy nft_chain_policy[NFTA_CHAIN_MAX + 1] = {
907 [NFTA_CHAIN_TABLE] = { .type = NLA_STRING },
908 [NFTA_CHAIN_HANDLE] = { .type = NLA_U64 },
909 [NFTA_CHAIN_NAME] = { .type = NLA_STRING,
910 .len = NFT_CHAIN_MAXNAMELEN - 1 },
911 [NFTA_CHAIN_HOOK] = { .type = NLA_NESTED },
912 [NFTA_CHAIN_POLICY] = { .type = NLA_U32 },
913 [NFTA_CHAIN_TYPE] = { .type = NLA_STRING },
914 [NFTA_CHAIN_COUNTERS] = { .type = NLA_NESTED },
917 static const struct nla_policy nft_hook_policy[NFTA_HOOK_MAX + 1] = {
918 [NFTA_HOOK_HOOKNUM] = { .type = NLA_U32 },
919 [NFTA_HOOK_PRIORITY] = { .type = NLA_U32 },
920 [NFTA_HOOK_DEV] = { .type = NLA_STRING,
921 .len = IFNAMSIZ - 1 },
924 static int nft_dump_stats(struct sk_buff *skb, struct nft_stats __percpu *stats)
926 struct nft_stats *cpu_stats, total;
932 memset(&total, 0, sizeof(total));
933 for_each_possible_cpu(cpu) {
934 cpu_stats = per_cpu_ptr(stats, cpu);
936 seq = u64_stats_fetch_begin_irq(&cpu_stats->syncp);
937 pkts = cpu_stats->pkts;
938 bytes = cpu_stats->bytes;
939 } while (u64_stats_fetch_retry_irq(&cpu_stats->syncp, seq));
941 total.bytes += bytes;
943 nest = nla_nest_start(skb, NFTA_CHAIN_COUNTERS);
945 goto nla_put_failure;
947 if (nla_put_be64(skb, NFTA_COUNTER_PACKETS, cpu_to_be64(total.pkts),
949 nla_put_be64(skb, NFTA_COUNTER_BYTES, cpu_to_be64(total.bytes),
951 goto nla_put_failure;
953 nla_nest_end(skb, nest);
960 static int nf_tables_fill_chain_info(struct sk_buff *skb, struct net *net,
961 u32 portid, u32 seq, int event, u32 flags,
962 int family, const struct nft_table *table,
963 const struct nft_chain *chain)
965 struct nlmsghdr *nlh;
966 struct nfgenmsg *nfmsg;
968 event |= NFNL_SUBSYS_NFTABLES << 8;
969 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct nfgenmsg), flags);
971 goto nla_put_failure;
973 nfmsg = nlmsg_data(nlh);
974 nfmsg->nfgen_family = family;
975 nfmsg->version = NFNETLINK_V0;
976 nfmsg->res_id = htons(net->nft.base_seq & 0xffff);
978 if (nla_put_string(skb, NFTA_CHAIN_TABLE, table->name))
979 goto nla_put_failure;
980 if (nla_put_be64(skb, NFTA_CHAIN_HANDLE, cpu_to_be64(chain->handle),
982 goto nla_put_failure;
983 if (nla_put_string(skb, NFTA_CHAIN_NAME, chain->name))
984 goto nla_put_failure;
986 if (chain->flags & NFT_BASE_CHAIN) {
987 const struct nft_base_chain *basechain = nft_base_chain(chain);
988 const struct nf_hook_ops *ops = &basechain->ops[0];
991 nest = nla_nest_start(skb, NFTA_CHAIN_HOOK);
993 goto nla_put_failure;
994 if (nla_put_be32(skb, NFTA_HOOK_HOOKNUM, htonl(ops->hooknum)))
995 goto nla_put_failure;
996 if (nla_put_be32(skb, NFTA_HOOK_PRIORITY, htonl(ops->priority)))
997 goto nla_put_failure;
998 if (basechain->dev_name[0] &&
999 nla_put_string(skb, NFTA_HOOK_DEV, basechain->dev_name))
1000 goto nla_put_failure;
1001 nla_nest_end(skb, nest);
1003 if (nla_put_be32(skb, NFTA_CHAIN_POLICY,
1004 htonl(basechain->policy)))
1005 goto nla_put_failure;
1007 if (nla_put_string(skb, NFTA_CHAIN_TYPE, basechain->type->name))
1008 goto nla_put_failure;
1010 if (nft_dump_stats(skb, nft_base_chain(chain)->stats))
1011 goto nla_put_failure;
1014 if (nla_put_be32(skb, NFTA_CHAIN_USE, htonl(chain->use)))
1015 goto nla_put_failure;
1017 nlmsg_end(skb, nlh);
1021 nlmsg_trim(skb, nlh);
1025 static int nf_tables_chain_notify(const struct nft_ctx *ctx, int event)
1027 struct sk_buff *skb;
1031 !nfnetlink_has_listeners(ctx->net, NFNLGRP_NFTABLES))
1035 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
1039 err = nf_tables_fill_chain_info(skb, ctx->net, ctx->portid, ctx->seq,
1040 event, 0, ctx->afi->family, ctx->table,
1047 err = nfnetlink_send(skb, ctx->net, ctx->portid, NFNLGRP_NFTABLES,
1048 ctx->report, GFP_KERNEL);
1051 nfnetlink_set_err(ctx->net, ctx->portid, NFNLGRP_NFTABLES,
1057 static int nf_tables_dump_chains(struct sk_buff *skb,
1058 struct netlink_callback *cb)
1060 const struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
1061 const struct nft_af_info *afi;
1062 const struct nft_table *table;
1063 const struct nft_chain *chain;
1064 unsigned int idx = 0, s_idx = cb->args[0];
1065 struct net *net = sock_net(skb->sk);
1066 int family = nfmsg->nfgen_family;
1069 cb->seq = net->nft.base_seq;
1071 list_for_each_entry_rcu(afi, &net->nft.af_info, list) {
1072 if (family != NFPROTO_UNSPEC && family != afi->family)
1075 list_for_each_entry_rcu(table, &afi->tables, list) {
1076 list_for_each_entry_rcu(chain, &table->chains, list) {
1080 memset(&cb->args[1], 0,
1081 sizeof(cb->args) - sizeof(cb->args[0]));
1082 if (nf_tables_fill_chain_info(skb, net,
1083 NETLINK_CB(cb->skb).portid,
1087 afi->family, table, chain) < 0)
1090 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1102 static int nf_tables_getchain(struct net *net, struct sock *nlsk,
1103 struct sk_buff *skb, const struct nlmsghdr *nlh,
1104 const struct nlattr * const nla[])
1106 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1107 const struct nft_af_info *afi;
1108 const struct nft_table *table;
1109 const struct nft_chain *chain;
1110 struct sk_buff *skb2;
1111 int family = nfmsg->nfgen_family;
1114 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1115 struct netlink_dump_control c = {
1116 .dump = nf_tables_dump_chains,
1118 return netlink_dump_start(nlsk, skb, nlh, &c);
1121 afi = nf_tables_afinfo_lookup(net, family, false);
1123 return PTR_ERR(afi);
1125 table = nf_tables_table_lookup(afi, nla[NFTA_CHAIN_TABLE]);
1127 return PTR_ERR(table);
1128 if (table->flags & NFT_TABLE_INACTIVE)
1131 chain = nf_tables_chain_lookup(table, nla[NFTA_CHAIN_NAME]);
1133 return PTR_ERR(chain);
1134 if (chain->flags & NFT_CHAIN_INACTIVE)
1137 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1141 err = nf_tables_fill_chain_info(skb2, net, NETLINK_CB(skb).portid,
1142 nlh->nlmsg_seq, NFT_MSG_NEWCHAIN, 0,
1143 family, table, chain);
1147 return nlmsg_unicast(nlsk, skb2, NETLINK_CB(skb).portid);
1154 static const struct nla_policy nft_counter_policy[NFTA_COUNTER_MAX + 1] = {
1155 [NFTA_COUNTER_PACKETS] = { .type = NLA_U64 },
1156 [NFTA_COUNTER_BYTES] = { .type = NLA_U64 },
1159 static struct nft_stats __percpu *nft_stats_alloc(const struct nlattr *attr)
1161 struct nlattr *tb[NFTA_COUNTER_MAX+1];
1162 struct nft_stats __percpu *newstats;
1163 struct nft_stats *stats;
1166 err = nla_parse_nested(tb, NFTA_COUNTER_MAX, attr, nft_counter_policy);
1168 return ERR_PTR(err);
1170 if (!tb[NFTA_COUNTER_BYTES] || !tb[NFTA_COUNTER_PACKETS])
1171 return ERR_PTR(-EINVAL);
1173 newstats = netdev_alloc_pcpu_stats(struct nft_stats);
1174 if (newstats == NULL)
1175 return ERR_PTR(-ENOMEM);
1177 /* Restore old counters on this cpu, no problem. Per-cpu statistics
1178 * are not exposed to userspace.
1181 stats = this_cpu_ptr(newstats);
1182 stats->bytes = be64_to_cpu(nla_get_be64(tb[NFTA_COUNTER_BYTES]));
1183 stats->pkts = be64_to_cpu(nla_get_be64(tb[NFTA_COUNTER_PACKETS]));
1189 static void nft_chain_stats_replace(struct nft_base_chain *chain,
1190 struct nft_stats __percpu *newstats)
1192 if (newstats == NULL)
1196 struct nft_stats __percpu *oldstats =
1197 nft_dereference(chain->stats);
1199 rcu_assign_pointer(chain->stats, newstats);
1201 free_percpu(oldstats);
1203 rcu_assign_pointer(chain->stats, newstats);
1206 static void nf_tables_chain_destroy(struct nft_chain *chain)
1208 BUG_ON(chain->use > 0);
1210 if (chain->flags & NFT_BASE_CHAIN) {
1211 struct nft_base_chain *basechain = nft_base_chain(chain);
1213 module_put(basechain->type->owner);
1214 free_percpu(basechain->stats);
1215 if (basechain->ops[0].dev != NULL)
1216 dev_put(basechain->ops[0].dev);
1223 static int nf_tables_newchain(struct net *net, struct sock *nlsk,
1224 struct sk_buff *skb, const struct nlmsghdr *nlh,
1225 const struct nlattr * const nla[])
1227 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1228 const struct nlattr * uninitialized_var(name);
1229 struct nft_af_info *afi;
1230 struct nft_table *table;
1231 struct nft_chain *chain;
1232 struct nft_base_chain *basechain = NULL;
1233 struct nlattr *ha[NFTA_HOOK_MAX + 1];
1234 int family = nfmsg->nfgen_family;
1235 struct net_device *dev = NULL;
1236 u8 policy = NF_ACCEPT;
1239 struct nft_stats __percpu *stats;
1244 create = nlh->nlmsg_flags & NLM_F_CREATE ? true : false;
1246 afi = nf_tables_afinfo_lookup(net, family, true);
1248 return PTR_ERR(afi);
1250 table = nf_tables_table_lookup(afi, nla[NFTA_CHAIN_TABLE]);
1252 return PTR_ERR(table);
1255 name = nla[NFTA_CHAIN_NAME];
1257 if (nla[NFTA_CHAIN_HANDLE]) {
1258 handle = be64_to_cpu(nla_get_be64(nla[NFTA_CHAIN_HANDLE]));
1259 chain = nf_tables_chain_lookup_byhandle(table, handle);
1261 return PTR_ERR(chain);
1263 chain = nf_tables_chain_lookup(table, name);
1264 if (IS_ERR(chain)) {
1265 if (PTR_ERR(chain) != -ENOENT)
1266 return PTR_ERR(chain);
1271 if (nla[NFTA_CHAIN_POLICY]) {
1272 if ((chain != NULL &&
1273 !(chain->flags & NFT_BASE_CHAIN)))
1276 if (chain == NULL &&
1277 nla[NFTA_CHAIN_HOOK] == NULL)
1280 policy = ntohl(nla_get_be32(nla[NFTA_CHAIN_POLICY]));
1290 if (chain != NULL) {
1291 struct nft_stats *stats = NULL;
1292 struct nft_trans *trans;
1294 if (chain->flags & NFT_CHAIN_INACTIVE)
1296 if (nlh->nlmsg_flags & NLM_F_EXCL)
1298 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1301 if (nla[NFTA_CHAIN_HANDLE] && name &&
1302 !IS_ERR(nf_tables_chain_lookup(table, nla[NFTA_CHAIN_NAME])))
1305 if (nla[NFTA_CHAIN_COUNTERS]) {
1306 if (!(chain->flags & NFT_BASE_CHAIN))
1309 stats = nft_stats_alloc(nla[NFTA_CHAIN_COUNTERS]);
1311 return PTR_ERR(stats);
1314 nft_ctx_init(&ctx, net, skb, nlh, afi, table, chain, nla);
1315 trans = nft_trans_alloc(&ctx, NFT_MSG_NEWCHAIN,
1316 sizeof(struct nft_trans_chain));
1317 if (trans == NULL) {
1322 nft_trans_chain_stats(trans) = stats;
1323 nft_trans_chain_update(trans) = true;
1325 if (nla[NFTA_CHAIN_POLICY])
1326 nft_trans_chain_policy(trans) = policy;
1328 nft_trans_chain_policy(trans) = -1;
1330 if (nla[NFTA_CHAIN_HANDLE] && name) {
1331 nla_strlcpy(nft_trans_chain_name(trans), name,
1332 NFT_CHAIN_MAXNAMELEN);
1334 list_add_tail(&trans->list, &net->nft.commit_list);
1338 if (table->use == UINT_MAX)
1341 if (nla[NFTA_CHAIN_HOOK]) {
1342 const struct nf_chain_type *type;
1343 struct nf_hook_ops *ops;
1345 u32 hooknum, priority;
1347 type = chain_type[family][NFT_CHAIN_T_DEFAULT];
1348 if (nla[NFTA_CHAIN_TYPE]) {
1349 type = nf_tables_chain_type_lookup(afi,
1350 nla[NFTA_CHAIN_TYPE],
1353 return PTR_ERR(type);
1356 err = nla_parse_nested(ha, NFTA_HOOK_MAX, nla[NFTA_CHAIN_HOOK],
1360 if (ha[NFTA_HOOK_HOOKNUM] == NULL ||
1361 ha[NFTA_HOOK_PRIORITY] == NULL)
1364 hooknum = ntohl(nla_get_be32(ha[NFTA_HOOK_HOOKNUM]));
1365 if (hooknum >= afi->nhooks)
1367 priority = ntohl(nla_get_be32(ha[NFTA_HOOK_PRIORITY]));
1369 if (!(type->hook_mask & (1 << hooknum)))
1371 if (!try_module_get(type->owner))
1373 hookfn = type->hooks[hooknum];
1375 if (afi->flags & NFT_AF_NEEDS_DEV) {
1376 char ifname[IFNAMSIZ];
1378 if (!ha[NFTA_HOOK_DEV]) {
1379 module_put(type->owner);
1383 nla_strlcpy(ifname, ha[NFTA_HOOK_DEV], IFNAMSIZ);
1384 dev = dev_get_by_name(net, ifname);
1386 module_put(type->owner);
1389 } else if (ha[NFTA_HOOK_DEV]) {
1390 module_put(type->owner);
1394 basechain = kzalloc(sizeof(*basechain), GFP_KERNEL);
1395 if (basechain == NULL) {
1396 module_put(type->owner);
1403 strncpy(basechain->dev_name, dev->name, IFNAMSIZ);
1405 if (nla[NFTA_CHAIN_COUNTERS]) {
1406 stats = nft_stats_alloc(nla[NFTA_CHAIN_COUNTERS]);
1407 if (IS_ERR(stats)) {
1408 module_put(type->owner);
1412 return PTR_ERR(stats);
1414 basechain->stats = stats;
1416 stats = netdev_alloc_pcpu_stats(struct nft_stats);
1417 if (stats == NULL) {
1418 module_put(type->owner);
1424 rcu_assign_pointer(basechain->stats, stats);
1427 write_pnet(&basechain->pnet, net);
1428 basechain->type = type;
1429 chain = &basechain->chain;
1431 for (i = 0; i < afi->nops; i++) {
1432 ops = &basechain->ops[i];
1434 ops->hooknum = hooknum;
1435 ops->priority = priority;
1437 ops->hook = afi->hooks[ops->hooknum];
1441 if (afi->hook_ops_init)
1442 afi->hook_ops_init(ops, i);
1445 chain->flags |= NFT_BASE_CHAIN;
1446 basechain->policy = policy;
1448 chain = kzalloc(sizeof(*chain), GFP_KERNEL);
1453 INIT_LIST_HEAD(&chain->rules);
1454 chain->handle = nf_tables_alloc_handle(table);
1455 chain->table = table;
1456 nla_strlcpy(chain->name, name, NFT_CHAIN_MAXNAMELEN);
1458 err = nf_tables_register_hooks(table, chain, afi->nops);
1462 nft_ctx_init(&ctx, net, skb, nlh, afi, table, chain, nla);
1463 err = nft_trans_chain_add(&ctx, NFT_MSG_NEWCHAIN);
1468 list_add_tail_rcu(&chain->list, &table->chains);
1471 nf_tables_unregister_hooks(table, chain, afi->nops);
1473 nf_tables_chain_destroy(chain);
1477 static int nf_tables_delchain(struct net *net, struct sock *nlsk,
1478 struct sk_buff *skb, const struct nlmsghdr *nlh,
1479 const struct nlattr * const nla[])
1481 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1482 struct nft_af_info *afi;
1483 struct nft_table *table;
1484 struct nft_chain *chain;
1485 int family = nfmsg->nfgen_family;
1488 afi = nf_tables_afinfo_lookup(net, family, false);
1490 return PTR_ERR(afi);
1492 table = nf_tables_table_lookup(afi, nla[NFTA_CHAIN_TABLE]);
1494 return PTR_ERR(table);
1496 chain = nf_tables_chain_lookup(table, nla[NFTA_CHAIN_NAME]);
1498 return PTR_ERR(chain);
1502 nft_ctx_init(&ctx, net, skb, nlh, afi, table, chain, nla);
1504 return nft_delchain(&ctx);
1512 * nft_register_expr - register nf_tables expr type
1515 * Registers the expr type for use with nf_tables. Returns zero on
1516 * success or a negative errno code otherwise.
1518 int nft_register_expr(struct nft_expr_type *type)
1520 nfnl_lock(NFNL_SUBSYS_NFTABLES);
1521 if (type->family == NFPROTO_UNSPEC)
1522 list_add_tail_rcu(&type->list, &nf_tables_expressions);
1524 list_add_rcu(&type->list, &nf_tables_expressions);
1525 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
1528 EXPORT_SYMBOL_GPL(nft_register_expr);
1531 * nft_unregister_expr - unregister nf_tables expr type
1534 * Unregisters the expr typefor use with nf_tables.
1536 void nft_unregister_expr(struct nft_expr_type *type)
1538 nfnl_lock(NFNL_SUBSYS_NFTABLES);
1539 list_del_rcu(&type->list);
1540 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
1542 EXPORT_SYMBOL_GPL(nft_unregister_expr);
1544 static const struct nft_expr_type *__nft_expr_type_get(u8 family,
1547 const struct nft_expr_type *type;
1549 list_for_each_entry(type, &nf_tables_expressions, list) {
1550 if (!nla_strcmp(nla, type->name) &&
1551 (!type->family || type->family == family))
1557 static const struct nft_expr_type *nft_expr_type_get(u8 family,
1560 const struct nft_expr_type *type;
1563 return ERR_PTR(-EINVAL);
1565 type = __nft_expr_type_get(family, nla);
1566 if (type != NULL && try_module_get(type->owner))
1569 #ifdef CONFIG_MODULES
1571 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
1572 request_module("nft-expr-%u-%.*s", family,
1573 nla_len(nla), (char *)nla_data(nla));
1574 nfnl_lock(NFNL_SUBSYS_NFTABLES);
1575 if (__nft_expr_type_get(family, nla))
1576 return ERR_PTR(-EAGAIN);
1578 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
1579 request_module("nft-expr-%.*s",
1580 nla_len(nla), (char *)nla_data(nla));
1581 nfnl_lock(NFNL_SUBSYS_NFTABLES);
1582 if (__nft_expr_type_get(family, nla))
1583 return ERR_PTR(-EAGAIN);
1586 return ERR_PTR(-ENOENT);
1589 static const struct nla_policy nft_expr_policy[NFTA_EXPR_MAX + 1] = {
1590 [NFTA_EXPR_NAME] = { .type = NLA_STRING },
1591 [NFTA_EXPR_DATA] = { .type = NLA_NESTED },
1594 static int nf_tables_fill_expr_info(struct sk_buff *skb,
1595 const struct nft_expr *expr)
1597 if (nla_put_string(skb, NFTA_EXPR_NAME, expr->ops->type->name))
1598 goto nla_put_failure;
1600 if (expr->ops->dump) {
1601 struct nlattr *data = nla_nest_start(skb, NFTA_EXPR_DATA);
1603 goto nla_put_failure;
1604 if (expr->ops->dump(skb, expr) < 0)
1605 goto nla_put_failure;
1606 nla_nest_end(skb, data);
1615 int nft_expr_dump(struct sk_buff *skb, unsigned int attr,
1616 const struct nft_expr *expr)
1618 struct nlattr *nest;
1620 nest = nla_nest_start(skb, attr);
1622 goto nla_put_failure;
1623 if (nf_tables_fill_expr_info(skb, expr) < 0)
1624 goto nla_put_failure;
1625 nla_nest_end(skb, nest);
1632 struct nft_expr_info {
1633 const struct nft_expr_ops *ops;
1634 struct nlattr *tb[NFT_EXPR_MAXATTR + 1];
1637 static int nf_tables_expr_parse(const struct nft_ctx *ctx,
1638 const struct nlattr *nla,
1639 struct nft_expr_info *info)
1641 const struct nft_expr_type *type;
1642 const struct nft_expr_ops *ops;
1643 struct nlattr *tb[NFTA_EXPR_MAX + 1];
1646 err = nla_parse_nested(tb, NFTA_EXPR_MAX, nla, nft_expr_policy);
1650 type = nft_expr_type_get(ctx->afi->family, tb[NFTA_EXPR_NAME]);
1652 return PTR_ERR(type);
1654 if (tb[NFTA_EXPR_DATA]) {
1655 err = nla_parse_nested(info->tb, type->maxattr,
1656 tb[NFTA_EXPR_DATA], type->policy);
1660 memset(info->tb, 0, sizeof(info->tb[0]) * (type->maxattr + 1));
1662 if (type->select_ops != NULL) {
1663 ops = type->select_ops(ctx,
1664 (const struct nlattr * const *)info->tb);
1676 module_put(type->owner);
1680 static int nf_tables_newexpr(const struct nft_ctx *ctx,
1681 const struct nft_expr_info *info,
1682 struct nft_expr *expr)
1684 const struct nft_expr_ops *ops = info->ops;
1689 err = ops->init(ctx, expr, (const struct nlattr **)info->tb);
1701 static void nf_tables_expr_destroy(const struct nft_ctx *ctx,
1702 struct nft_expr *expr)
1704 if (expr->ops->destroy)
1705 expr->ops->destroy(ctx, expr);
1706 module_put(expr->ops->type->owner);
1709 struct nft_expr *nft_expr_init(const struct nft_ctx *ctx,
1710 const struct nlattr *nla)
1712 struct nft_expr_info info;
1713 struct nft_expr *expr;
1716 err = nf_tables_expr_parse(ctx, nla, &info);
1721 expr = kzalloc(info.ops->size, GFP_KERNEL);
1725 err = nf_tables_newexpr(ctx, &info, expr);
1731 module_put(info.ops->type->owner);
1733 return ERR_PTR(err);
1736 void nft_expr_destroy(const struct nft_ctx *ctx, struct nft_expr *expr)
1738 nf_tables_expr_destroy(ctx, expr);
1746 static struct nft_rule *__nf_tables_rule_lookup(const struct nft_chain *chain,
1749 struct nft_rule *rule;
1751 // FIXME: this sucks
1752 list_for_each_entry(rule, &chain->rules, list) {
1753 if (handle == rule->handle)
1757 return ERR_PTR(-ENOENT);
1760 static struct nft_rule *nf_tables_rule_lookup(const struct nft_chain *chain,
1761 const struct nlattr *nla)
1764 return ERR_PTR(-EINVAL);
1766 return __nf_tables_rule_lookup(chain, be64_to_cpu(nla_get_be64(nla)));
1769 static const struct nla_policy nft_rule_policy[NFTA_RULE_MAX + 1] = {
1770 [NFTA_RULE_TABLE] = { .type = NLA_STRING },
1771 [NFTA_RULE_CHAIN] = { .type = NLA_STRING,
1772 .len = NFT_CHAIN_MAXNAMELEN - 1 },
1773 [NFTA_RULE_HANDLE] = { .type = NLA_U64 },
1774 [NFTA_RULE_EXPRESSIONS] = { .type = NLA_NESTED },
1775 [NFTA_RULE_COMPAT] = { .type = NLA_NESTED },
1776 [NFTA_RULE_POSITION] = { .type = NLA_U64 },
1777 [NFTA_RULE_USERDATA] = { .type = NLA_BINARY,
1778 .len = NFT_USERDATA_MAXLEN },
1781 static int nf_tables_fill_rule_info(struct sk_buff *skb, struct net *net,
1782 u32 portid, u32 seq, int event,
1783 u32 flags, int family,
1784 const struct nft_table *table,
1785 const struct nft_chain *chain,
1786 const struct nft_rule *rule)
1788 struct nlmsghdr *nlh;
1789 struct nfgenmsg *nfmsg;
1790 const struct nft_expr *expr, *next;
1791 struct nlattr *list;
1792 const struct nft_rule *prule;
1793 int type = event | NFNL_SUBSYS_NFTABLES << 8;
1795 nlh = nlmsg_put(skb, portid, seq, type, sizeof(struct nfgenmsg),
1798 goto nla_put_failure;
1800 nfmsg = nlmsg_data(nlh);
1801 nfmsg->nfgen_family = family;
1802 nfmsg->version = NFNETLINK_V0;
1803 nfmsg->res_id = htons(net->nft.base_seq & 0xffff);
1805 if (nla_put_string(skb, NFTA_RULE_TABLE, table->name))
1806 goto nla_put_failure;
1807 if (nla_put_string(skb, NFTA_RULE_CHAIN, chain->name))
1808 goto nla_put_failure;
1809 if (nla_put_be64(skb, NFTA_RULE_HANDLE, cpu_to_be64(rule->handle),
1811 goto nla_put_failure;
1813 if ((event != NFT_MSG_DELRULE) && (rule->list.prev != &chain->rules)) {
1814 prule = list_entry(rule->list.prev, struct nft_rule, list);
1815 if (nla_put_be64(skb, NFTA_RULE_POSITION,
1816 cpu_to_be64(prule->handle),
1818 goto nla_put_failure;
1821 list = nla_nest_start(skb, NFTA_RULE_EXPRESSIONS);
1823 goto nla_put_failure;
1824 nft_rule_for_each_expr(expr, next, rule) {
1825 if (nft_expr_dump(skb, NFTA_LIST_ELEM, expr) < 0)
1826 goto nla_put_failure;
1828 nla_nest_end(skb, list);
1831 struct nft_userdata *udata = nft_userdata(rule);
1832 if (nla_put(skb, NFTA_RULE_USERDATA, udata->len + 1,
1834 goto nla_put_failure;
1837 nlmsg_end(skb, nlh);
1841 nlmsg_trim(skb, nlh);
1845 static int nf_tables_rule_notify(const struct nft_ctx *ctx,
1846 const struct nft_rule *rule,
1849 struct sk_buff *skb;
1853 !nfnetlink_has_listeners(ctx->net, NFNLGRP_NFTABLES))
1857 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
1861 err = nf_tables_fill_rule_info(skb, ctx->net, ctx->portid, ctx->seq,
1862 event, 0, ctx->afi->family, ctx->table,
1869 err = nfnetlink_send(skb, ctx->net, ctx->portid, NFNLGRP_NFTABLES,
1870 ctx->report, GFP_KERNEL);
1873 nfnetlink_set_err(ctx->net, ctx->portid, NFNLGRP_NFTABLES,
1879 static int nf_tables_dump_rules(struct sk_buff *skb,
1880 struct netlink_callback *cb)
1882 const struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
1883 const struct nft_af_info *afi;
1884 const struct nft_table *table;
1885 const struct nft_chain *chain;
1886 const struct nft_rule *rule;
1887 unsigned int idx = 0, s_idx = cb->args[0];
1888 struct net *net = sock_net(skb->sk);
1889 int family = nfmsg->nfgen_family;
1892 cb->seq = net->nft.base_seq;
1894 list_for_each_entry_rcu(afi, &net->nft.af_info, list) {
1895 if (family != NFPROTO_UNSPEC && family != afi->family)
1898 list_for_each_entry_rcu(table, &afi->tables, list) {
1899 list_for_each_entry_rcu(chain, &table->chains, list) {
1900 list_for_each_entry_rcu(rule, &chain->rules, list) {
1901 if (!nft_rule_is_active(net, rule))
1906 memset(&cb->args[1], 0,
1907 sizeof(cb->args) - sizeof(cb->args[0]));
1908 if (nf_tables_fill_rule_info(skb, net, NETLINK_CB(cb->skb).portid,
1911 NLM_F_MULTI | NLM_F_APPEND,
1912 afi->family, table, chain, rule) < 0)
1915 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1929 static int nf_tables_getrule(struct net *net, struct sock *nlsk,
1930 struct sk_buff *skb, const struct nlmsghdr *nlh,
1931 const struct nlattr * const nla[])
1933 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1934 const struct nft_af_info *afi;
1935 const struct nft_table *table;
1936 const struct nft_chain *chain;
1937 const struct nft_rule *rule;
1938 struct sk_buff *skb2;
1939 int family = nfmsg->nfgen_family;
1942 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1943 struct netlink_dump_control c = {
1944 .dump = nf_tables_dump_rules,
1946 return netlink_dump_start(nlsk, skb, nlh, &c);
1949 afi = nf_tables_afinfo_lookup(net, family, false);
1951 return PTR_ERR(afi);
1953 table = nf_tables_table_lookup(afi, nla[NFTA_RULE_TABLE]);
1955 return PTR_ERR(table);
1956 if (table->flags & NFT_TABLE_INACTIVE)
1959 chain = nf_tables_chain_lookup(table, nla[NFTA_RULE_CHAIN]);
1961 return PTR_ERR(chain);
1962 if (chain->flags & NFT_CHAIN_INACTIVE)
1965 rule = nf_tables_rule_lookup(chain, nla[NFTA_RULE_HANDLE]);
1967 return PTR_ERR(rule);
1969 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1973 err = nf_tables_fill_rule_info(skb2, net, NETLINK_CB(skb).portid,
1974 nlh->nlmsg_seq, NFT_MSG_NEWRULE, 0,
1975 family, table, chain, rule);
1979 return nlmsg_unicast(nlsk, skb2, NETLINK_CB(skb).portid);
1986 static void nf_tables_rule_destroy(const struct nft_ctx *ctx,
1987 struct nft_rule *rule)
1989 struct nft_expr *expr;
1992 * Careful: some expressions might not be initialized in case this
1993 * is called on error from nf_tables_newrule().
1995 expr = nft_expr_first(rule);
1996 while (expr->ops && expr != nft_expr_last(rule)) {
1997 nf_tables_expr_destroy(ctx, expr);
1998 expr = nft_expr_next(expr);
2003 #define NFT_RULE_MAXEXPRS 128
2005 static struct nft_expr_info *info;
2007 static int nf_tables_newrule(struct net *net, struct sock *nlsk,
2008 struct sk_buff *skb, const struct nlmsghdr *nlh,
2009 const struct nlattr * const nla[])
2011 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
2012 struct nft_af_info *afi;
2013 struct nft_table *table;
2014 struct nft_chain *chain;
2015 struct nft_rule *rule, *old_rule = NULL;
2016 struct nft_userdata *udata;
2017 struct nft_trans *trans = NULL;
2018 struct nft_expr *expr;
2021 unsigned int size, i, n, ulen = 0, usize = 0;
2024 u64 handle, pos_handle;
2026 create = nlh->nlmsg_flags & NLM_F_CREATE ? true : false;
2028 afi = nf_tables_afinfo_lookup(net, nfmsg->nfgen_family, create);
2030 return PTR_ERR(afi);
2032 table = nf_tables_table_lookup(afi, nla[NFTA_RULE_TABLE]);
2034 return PTR_ERR(table);
2036 chain = nf_tables_chain_lookup(table, nla[NFTA_RULE_CHAIN]);
2038 return PTR_ERR(chain);
2040 if (nla[NFTA_RULE_HANDLE]) {
2041 handle = be64_to_cpu(nla_get_be64(nla[NFTA_RULE_HANDLE]));
2042 rule = __nf_tables_rule_lookup(chain, handle);
2044 return PTR_ERR(rule);
2046 if (nlh->nlmsg_flags & NLM_F_EXCL)
2048 if (nlh->nlmsg_flags & NLM_F_REPLACE)
2053 if (!create || nlh->nlmsg_flags & NLM_F_REPLACE)
2055 handle = nf_tables_alloc_handle(table);
2057 if (chain->use == UINT_MAX)
2061 if (nla[NFTA_RULE_POSITION]) {
2062 if (!(nlh->nlmsg_flags & NLM_F_CREATE))
2065 pos_handle = be64_to_cpu(nla_get_be64(nla[NFTA_RULE_POSITION]));
2066 old_rule = __nf_tables_rule_lookup(chain, pos_handle);
2067 if (IS_ERR(old_rule))
2068 return PTR_ERR(old_rule);
2071 nft_ctx_init(&ctx, net, skb, nlh, afi, table, chain, nla);
2075 if (nla[NFTA_RULE_EXPRESSIONS]) {
2076 nla_for_each_nested(tmp, nla[NFTA_RULE_EXPRESSIONS], rem) {
2078 if (nla_type(tmp) != NFTA_LIST_ELEM)
2080 if (n == NFT_RULE_MAXEXPRS)
2082 err = nf_tables_expr_parse(&ctx, tmp, &info[n]);
2085 size += info[n].ops->size;
2089 /* Check for overflow of dlen field */
2091 if (size >= 1 << 12)
2094 if (nla[NFTA_RULE_USERDATA]) {
2095 ulen = nla_len(nla[NFTA_RULE_USERDATA]);
2097 usize = sizeof(struct nft_userdata) + ulen;
2101 rule = kzalloc(sizeof(*rule) + size + usize, GFP_KERNEL);
2105 nft_rule_activate_next(net, rule);
2107 rule->handle = handle;
2109 rule->udata = ulen ? 1 : 0;
2112 udata = nft_userdata(rule);
2113 udata->len = ulen - 1;
2114 nla_memcpy(udata->data, nla[NFTA_RULE_USERDATA], ulen);
2117 expr = nft_expr_first(rule);
2118 for (i = 0; i < n; i++) {
2119 err = nf_tables_newexpr(&ctx, &info[i], expr);
2123 expr = nft_expr_next(expr);
2126 if (nlh->nlmsg_flags & NLM_F_REPLACE) {
2127 if (nft_rule_is_active_next(net, old_rule)) {
2128 trans = nft_trans_rule_add(&ctx, NFT_MSG_DELRULE,
2130 if (trans == NULL) {
2134 nft_rule_deactivate_next(net, old_rule);
2136 list_add_tail_rcu(&rule->list, &old_rule->list);
2141 } else if (nlh->nlmsg_flags & NLM_F_APPEND)
2143 list_add_rcu(&rule->list, &old_rule->list);
2145 list_add_tail_rcu(&rule->list, &chain->rules);
2148 list_add_tail_rcu(&rule->list, &old_rule->list);
2150 list_add_rcu(&rule->list, &chain->rules);
2153 if (nft_trans_rule_add(&ctx, NFT_MSG_NEWRULE, rule) == NULL) {
2161 list_del_rcu(&rule->list);
2163 nf_tables_rule_destroy(&ctx, rule);
2165 for (i = 0; i < n; i++) {
2166 if (info[i].ops != NULL)
2167 module_put(info[i].ops->type->owner);
2172 static int nf_tables_delrule(struct net *net, struct sock *nlsk,
2173 struct sk_buff *skb, const struct nlmsghdr *nlh,
2174 const struct nlattr * const nla[])
2176 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
2177 struct nft_af_info *afi;
2178 struct nft_table *table;
2179 struct nft_chain *chain = NULL;
2180 struct nft_rule *rule;
2181 int family = nfmsg->nfgen_family, err = 0;
2184 afi = nf_tables_afinfo_lookup(net, family, false);
2186 return PTR_ERR(afi);
2188 table = nf_tables_table_lookup(afi, nla[NFTA_RULE_TABLE]);
2190 return PTR_ERR(table);
2192 if (nla[NFTA_RULE_CHAIN]) {
2193 chain = nf_tables_chain_lookup(table, nla[NFTA_RULE_CHAIN]);
2195 return PTR_ERR(chain);
2198 nft_ctx_init(&ctx, net, skb, nlh, afi, table, chain, nla);
2201 if (nla[NFTA_RULE_HANDLE]) {
2202 rule = nf_tables_rule_lookup(chain,
2203 nla[NFTA_RULE_HANDLE]);
2205 return PTR_ERR(rule);
2207 err = nft_delrule(&ctx, rule);
2209 err = nft_delrule_by_chain(&ctx);
2212 list_for_each_entry(chain, &table->chains, list) {
2214 err = nft_delrule_by_chain(&ctx);
2227 static LIST_HEAD(nf_tables_set_ops);
2229 int nft_register_set(struct nft_set_ops *ops)
2231 nfnl_lock(NFNL_SUBSYS_NFTABLES);
2232 list_add_tail_rcu(&ops->list, &nf_tables_set_ops);
2233 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
2236 EXPORT_SYMBOL_GPL(nft_register_set);
2238 void nft_unregister_set(struct nft_set_ops *ops)
2240 nfnl_lock(NFNL_SUBSYS_NFTABLES);
2241 list_del_rcu(&ops->list);
2242 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
2244 EXPORT_SYMBOL_GPL(nft_unregister_set);
2247 * Select a set implementation based on the data characteristics and the
2248 * given policy. The total memory use might not be known if no size is
2249 * given, in that case the amount of memory per element is used.
2251 static const struct nft_set_ops *
2252 nft_select_set_ops(const struct nlattr * const nla[],
2253 const struct nft_set_desc *desc,
2254 enum nft_set_policies policy)
2256 const struct nft_set_ops *ops, *bops;
2257 struct nft_set_estimate est, best;
2260 #ifdef CONFIG_MODULES
2261 if (list_empty(&nf_tables_set_ops)) {
2262 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
2263 request_module("nft-set");
2264 nfnl_lock(NFNL_SUBSYS_NFTABLES);
2265 if (!list_empty(&nf_tables_set_ops))
2266 return ERR_PTR(-EAGAIN);
2270 if (nla[NFTA_SET_FLAGS] != NULL) {
2271 features = ntohl(nla_get_be32(nla[NFTA_SET_FLAGS]));
2272 features &= NFT_SET_INTERVAL | NFT_SET_MAP | NFT_SET_TIMEOUT;
2279 list_for_each_entry(ops, &nf_tables_set_ops, list) {
2280 if ((ops->features & features) != features)
2282 if (!ops->estimate(desc, features, &est))
2286 case NFT_SET_POL_PERFORMANCE:
2287 if (est.class < best.class)
2289 if (est.class == best.class && est.size < best.size)
2292 case NFT_SET_POL_MEMORY:
2293 if (est.size < best.size)
2295 if (est.size == best.size && est.class < best.class)
2302 if (!try_module_get(ops->owner))
2305 module_put(bops->owner);
2314 return ERR_PTR(-EOPNOTSUPP);
2317 static const struct nla_policy nft_set_policy[NFTA_SET_MAX + 1] = {
2318 [NFTA_SET_TABLE] = { .type = NLA_STRING },
2319 [NFTA_SET_NAME] = { .type = NLA_STRING,
2320 .len = NFT_SET_MAXNAMELEN - 1 },
2321 [NFTA_SET_FLAGS] = { .type = NLA_U32 },
2322 [NFTA_SET_KEY_TYPE] = { .type = NLA_U32 },
2323 [NFTA_SET_KEY_LEN] = { .type = NLA_U32 },
2324 [NFTA_SET_DATA_TYPE] = { .type = NLA_U32 },
2325 [NFTA_SET_DATA_LEN] = { .type = NLA_U32 },
2326 [NFTA_SET_POLICY] = { .type = NLA_U32 },
2327 [NFTA_SET_DESC] = { .type = NLA_NESTED },
2328 [NFTA_SET_ID] = { .type = NLA_U32 },
2329 [NFTA_SET_TIMEOUT] = { .type = NLA_U64 },
2330 [NFTA_SET_GC_INTERVAL] = { .type = NLA_U32 },
2331 [NFTA_SET_USERDATA] = { .type = NLA_BINARY,
2332 .len = NFT_USERDATA_MAXLEN },
2335 static const struct nla_policy nft_set_desc_policy[NFTA_SET_DESC_MAX + 1] = {
2336 [NFTA_SET_DESC_SIZE] = { .type = NLA_U32 },
2339 static int nft_ctx_init_from_setattr(struct nft_ctx *ctx, struct net *net,
2340 const struct sk_buff *skb,
2341 const struct nlmsghdr *nlh,
2342 const struct nlattr * const nla[])
2344 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
2345 struct nft_af_info *afi = NULL;
2346 struct nft_table *table = NULL;
2348 if (nfmsg->nfgen_family != NFPROTO_UNSPEC) {
2349 afi = nf_tables_afinfo_lookup(net, nfmsg->nfgen_family, false);
2351 return PTR_ERR(afi);
2354 if (nla[NFTA_SET_TABLE] != NULL) {
2356 return -EAFNOSUPPORT;
2358 table = nf_tables_table_lookup(afi, nla[NFTA_SET_TABLE]);
2360 return PTR_ERR(table);
2363 nft_ctx_init(ctx, net, skb, nlh, afi, table, NULL, nla);
2367 struct nft_set *nf_tables_set_lookup(const struct nft_table *table,
2368 const struct nlattr *nla)
2370 struct nft_set *set;
2373 return ERR_PTR(-EINVAL);
2375 list_for_each_entry(set, &table->sets, list) {
2376 if (!nla_strcmp(nla, set->name))
2379 return ERR_PTR(-ENOENT);
2382 struct nft_set *nf_tables_set_lookup_byid(const struct net *net,
2383 const struct nlattr *nla)
2385 struct nft_trans *trans;
2386 u32 id = ntohl(nla_get_be32(nla));
2388 list_for_each_entry(trans, &net->nft.commit_list, list) {
2389 if (trans->msg_type == NFT_MSG_NEWSET &&
2390 id == nft_trans_set_id(trans))
2391 return nft_trans_set(trans);
2393 return ERR_PTR(-ENOENT);
2396 static int nf_tables_set_alloc_name(struct nft_ctx *ctx, struct nft_set *set,
2399 const struct nft_set *i;
2401 unsigned long *inuse;
2402 unsigned int n = 0, min = 0;
2404 p = strnchr(name, NFT_SET_MAXNAMELEN, '%');
2406 if (p[1] != 'd' || strchr(p + 2, '%'))
2409 inuse = (unsigned long *)get_zeroed_page(GFP_KERNEL);
2413 list_for_each_entry(i, &ctx->table->sets, list) {
2416 if (!sscanf(i->name, name, &tmp))
2418 if (tmp < min || tmp >= min + BITS_PER_BYTE * PAGE_SIZE)
2421 set_bit(tmp - min, inuse);
2424 n = find_first_zero_bit(inuse, BITS_PER_BYTE * PAGE_SIZE);
2425 if (n >= BITS_PER_BYTE * PAGE_SIZE) {
2426 min += BITS_PER_BYTE * PAGE_SIZE;
2427 memset(inuse, 0, PAGE_SIZE);
2430 free_page((unsigned long)inuse);
2433 snprintf(set->name, sizeof(set->name), name, min + n);
2434 list_for_each_entry(i, &ctx->table->sets, list) {
2435 if (!strcmp(set->name, i->name))
2441 static int nf_tables_fill_set(struct sk_buff *skb, const struct nft_ctx *ctx,
2442 const struct nft_set *set, u16 event, u16 flags)
2444 struct nfgenmsg *nfmsg;
2445 struct nlmsghdr *nlh;
2446 struct nlattr *desc;
2447 u32 portid = ctx->portid;
2450 event |= NFNL_SUBSYS_NFTABLES << 8;
2451 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct nfgenmsg),
2454 goto nla_put_failure;
2456 nfmsg = nlmsg_data(nlh);
2457 nfmsg->nfgen_family = ctx->afi->family;
2458 nfmsg->version = NFNETLINK_V0;
2459 nfmsg->res_id = htons(ctx->net->nft.base_seq & 0xffff);
2461 if (nla_put_string(skb, NFTA_SET_TABLE, ctx->table->name))
2462 goto nla_put_failure;
2463 if (nla_put_string(skb, NFTA_SET_NAME, set->name))
2464 goto nla_put_failure;
2465 if (set->flags != 0)
2466 if (nla_put_be32(skb, NFTA_SET_FLAGS, htonl(set->flags)))
2467 goto nla_put_failure;
2469 if (nla_put_be32(skb, NFTA_SET_KEY_TYPE, htonl(set->ktype)))
2470 goto nla_put_failure;
2471 if (nla_put_be32(skb, NFTA_SET_KEY_LEN, htonl(set->klen)))
2472 goto nla_put_failure;
2473 if (set->flags & NFT_SET_MAP) {
2474 if (nla_put_be32(skb, NFTA_SET_DATA_TYPE, htonl(set->dtype)))
2475 goto nla_put_failure;
2476 if (nla_put_be32(skb, NFTA_SET_DATA_LEN, htonl(set->dlen)))
2477 goto nla_put_failure;
2481 nla_put_be64(skb, NFTA_SET_TIMEOUT, cpu_to_be64(set->timeout),
2483 goto nla_put_failure;
2485 nla_put_be32(skb, NFTA_SET_GC_INTERVAL, htonl(set->gc_int)))
2486 goto nla_put_failure;
2488 if (set->policy != NFT_SET_POL_PERFORMANCE) {
2489 if (nla_put_be32(skb, NFTA_SET_POLICY, htonl(set->policy)))
2490 goto nla_put_failure;
2493 if (nla_put(skb, NFTA_SET_USERDATA, set->udlen, set->udata))
2494 goto nla_put_failure;
2496 desc = nla_nest_start(skb, NFTA_SET_DESC);
2498 goto nla_put_failure;
2500 nla_put_be32(skb, NFTA_SET_DESC_SIZE, htonl(set->size)))
2501 goto nla_put_failure;
2502 nla_nest_end(skb, desc);
2504 nlmsg_end(skb, nlh);
2508 nlmsg_trim(skb, nlh);
2512 static int nf_tables_set_notify(const struct nft_ctx *ctx,
2513 const struct nft_set *set,
2514 int event, gfp_t gfp_flags)
2516 struct sk_buff *skb;
2517 u32 portid = ctx->portid;
2521 !nfnetlink_has_listeners(ctx->net, NFNLGRP_NFTABLES))
2525 skb = nlmsg_new(NLMSG_GOODSIZE, gfp_flags);
2529 err = nf_tables_fill_set(skb, ctx, set, event, 0);
2535 err = nfnetlink_send(skb, ctx->net, portid, NFNLGRP_NFTABLES,
2536 ctx->report, gfp_flags);
2539 nfnetlink_set_err(ctx->net, portid, NFNLGRP_NFTABLES, err);
2543 static int nf_tables_dump_sets(struct sk_buff *skb, struct netlink_callback *cb)
2545 const struct nft_set *set;
2546 unsigned int idx, s_idx = cb->args[0];
2547 struct nft_af_info *afi;
2548 struct nft_table *table, *cur_table = (struct nft_table *)cb->args[2];
2549 struct net *net = sock_net(skb->sk);
2550 int cur_family = cb->args[3];
2551 struct nft_ctx *ctx = cb->data, ctx_set;
2557 cb->seq = net->nft.base_seq;
2559 list_for_each_entry_rcu(afi, &net->nft.af_info, list) {
2560 if (ctx->afi && ctx->afi != afi)
2564 if (afi->family != cur_family)
2569 list_for_each_entry_rcu(table, &afi->tables, list) {
2570 if (ctx->table && ctx->table != table)
2574 if (cur_table != table)
2580 list_for_each_entry_rcu(set, &table->sets, list) {
2585 ctx_set.table = table;
2587 if (nf_tables_fill_set(skb, &ctx_set, set,
2591 cb->args[2] = (unsigned long) table;
2592 cb->args[3] = afi->family;
2595 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2609 static int nf_tables_dump_sets_done(struct netlink_callback *cb)
2615 static int nf_tables_getset(struct net *net, struct sock *nlsk,
2616 struct sk_buff *skb, const struct nlmsghdr *nlh,
2617 const struct nlattr * const nla[])
2619 const struct nft_set *set;
2621 struct sk_buff *skb2;
2622 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
2625 /* Verify existence before starting dump */
2626 err = nft_ctx_init_from_setattr(&ctx, net, skb, nlh, nla);
2630 if (nlh->nlmsg_flags & NLM_F_DUMP) {
2631 struct netlink_dump_control c = {
2632 .dump = nf_tables_dump_sets,
2633 .done = nf_tables_dump_sets_done,
2635 struct nft_ctx *ctx_dump;
2637 ctx_dump = kmalloc(sizeof(*ctx_dump), GFP_KERNEL);
2638 if (ctx_dump == NULL)
2644 return netlink_dump_start(nlsk, skb, nlh, &c);
2647 /* Only accept unspec with dump */
2648 if (nfmsg->nfgen_family == NFPROTO_UNSPEC)
2649 return -EAFNOSUPPORT;
2650 if (!nla[NFTA_SET_TABLE])
2653 set = nf_tables_set_lookup(ctx.table, nla[NFTA_SET_NAME]);
2655 return PTR_ERR(set);
2656 if (set->flags & NFT_SET_INACTIVE)
2659 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2663 err = nf_tables_fill_set(skb2, &ctx, set, NFT_MSG_NEWSET, 0);
2667 return nlmsg_unicast(nlsk, skb2, NETLINK_CB(skb).portid);
2674 static int nf_tables_set_desc_parse(const struct nft_ctx *ctx,
2675 struct nft_set_desc *desc,
2676 const struct nlattr *nla)
2678 struct nlattr *da[NFTA_SET_DESC_MAX + 1];
2681 err = nla_parse_nested(da, NFTA_SET_DESC_MAX, nla, nft_set_desc_policy);
2685 if (da[NFTA_SET_DESC_SIZE] != NULL)
2686 desc->size = ntohl(nla_get_be32(da[NFTA_SET_DESC_SIZE]));
2691 static int nf_tables_newset(struct net *net, struct sock *nlsk,
2692 struct sk_buff *skb, const struct nlmsghdr *nlh,
2693 const struct nlattr * const nla[])
2695 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
2696 const struct nft_set_ops *ops;
2697 struct nft_af_info *afi;
2698 struct nft_table *table;
2699 struct nft_set *set;
2701 char name[NFT_SET_MAXNAMELEN];
2705 u32 ktype, dtype, flags, policy, gc_int;
2706 struct nft_set_desc desc;
2707 unsigned char *udata;
2711 if (nla[NFTA_SET_TABLE] == NULL ||
2712 nla[NFTA_SET_NAME] == NULL ||
2713 nla[NFTA_SET_KEY_LEN] == NULL ||
2714 nla[NFTA_SET_ID] == NULL)
2717 memset(&desc, 0, sizeof(desc));
2719 ktype = NFT_DATA_VALUE;
2720 if (nla[NFTA_SET_KEY_TYPE] != NULL) {
2721 ktype = ntohl(nla_get_be32(nla[NFTA_SET_KEY_TYPE]));
2722 if ((ktype & NFT_DATA_RESERVED_MASK) == NFT_DATA_RESERVED_MASK)
2726 desc.klen = ntohl(nla_get_be32(nla[NFTA_SET_KEY_LEN]));
2727 if (desc.klen == 0 || desc.klen > NFT_DATA_VALUE_MAXLEN)
2731 if (nla[NFTA_SET_FLAGS] != NULL) {
2732 flags = ntohl(nla_get_be32(nla[NFTA_SET_FLAGS]));
2733 if (flags & ~(NFT_SET_ANONYMOUS | NFT_SET_CONSTANT |
2734 NFT_SET_INTERVAL | NFT_SET_TIMEOUT |
2735 NFT_SET_MAP | NFT_SET_EVAL))
2737 /* Only one of both operations is supported */
2738 if ((flags & (NFT_SET_MAP | NFT_SET_EVAL)) ==
2739 (NFT_SET_MAP | NFT_SET_EVAL))
2744 if (nla[NFTA_SET_DATA_TYPE] != NULL) {
2745 if (!(flags & NFT_SET_MAP))
2748 dtype = ntohl(nla_get_be32(nla[NFTA_SET_DATA_TYPE]));
2749 if ((dtype & NFT_DATA_RESERVED_MASK) == NFT_DATA_RESERVED_MASK &&
2750 dtype != NFT_DATA_VERDICT)
2753 if (dtype != NFT_DATA_VERDICT) {
2754 if (nla[NFTA_SET_DATA_LEN] == NULL)
2756 desc.dlen = ntohl(nla_get_be32(nla[NFTA_SET_DATA_LEN]));
2757 if (desc.dlen == 0 || desc.dlen > NFT_DATA_VALUE_MAXLEN)
2760 desc.dlen = sizeof(struct nft_verdict);
2761 } else if (flags & NFT_SET_MAP)
2765 if (nla[NFTA_SET_TIMEOUT] != NULL) {
2766 if (!(flags & NFT_SET_TIMEOUT))
2768 timeout = be64_to_cpu(nla_get_be64(nla[NFTA_SET_TIMEOUT]));
2771 if (nla[NFTA_SET_GC_INTERVAL] != NULL) {
2772 if (!(flags & NFT_SET_TIMEOUT))
2774 gc_int = ntohl(nla_get_be32(nla[NFTA_SET_GC_INTERVAL]));
2777 policy = NFT_SET_POL_PERFORMANCE;
2778 if (nla[NFTA_SET_POLICY] != NULL)
2779 policy = ntohl(nla_get_be32(nla[NFTA_SET_POLICY]));
2781 if (nla[NFTA_SET_DESC] != NULL) {
2782 err = nf_tables_set_desc_parse(&ctx, &desc, nla[NFTA_SET_DESC]);
2787 create = nlh->nlmsg_flags & NLM_F_CREATE ? true : false;
2789 afi = nf_tables_afinfo_lookup(net, nfmsg->nfgen_family, create);
2791 return PTR_ERR(afi);
2793 table = nf_tables_table_lookup(afi, nla[NFTA_SET_TABLE]);
2795 return PTR_ERR(table);
2797 nft_ctx_init(&ctx, net, skb, nlh, afi, table, NULL, nla);
2799 set = nf_tables_set_lookup(table, nla[NFTA_SET_NAME]);
2801 if (PTR_ERR(set) != -ENOENT)
2802 return PTR_ERR(set);
2807 if (nlh->nlmsg_flags & NLM_F_EXCL)
2809 if (nlh->nlmsg_flags & NLM_F_REPLACE)
2814 if (!(nlh->nlmsg_flags & NLM_F_CREATE))
2817 ops = nft_select_set_ops(nla, &desc, policy);
2819 return PTR_ERR(ops);
2822 if (nla[NFTA_SET_USERDATA])
2823 udlen = nla_len(nla[NFTA_SET_USERDATA]);
2826 if (ops->privsize != NULL)
2827 size = ops->privsize(nla);
2830 set = kzalloc(sizeof(*set) + size + udlen, GFP_KERNEL);
2834 nla_strlcpy(name, nla[NFTA_SET_NAME], sizeof(set->name));
2835 err = nf_tables_set_alloc_name(&ctx, set, name);
2841 udata = set->data + size;
2842 nla_memcpy(udata, nla[NFTA_SET_USERDATA], udlen);
2845 INIT_LIST_HEAD(&set->bindings);
2846 write_pnet(&set->pnet, net);
2849 set->klen = desc.klen;
2851 set->dlen = desc.dlen;
2853 set->size = desc.size;
2854 set->policy = policy;
2857 set->timeout = timeout;
2858 set->gc_int = gc_int;
2860 err = ops->init(set, &desc, nla);
2864 err = nft_trans_set_add(&ctx, NFT_MSG_NEWSET, set);
2868 list_add_tail_rcu(&set->list, &table->sets);
2875 module_put(ops->owner);
2879 static void nft_set_destroy(struct nft_set *set)
2881 set->ops->destroy(set);
2882 module_put(set->ops->owner);
2886 static void nf_tables_set_destroy(const struct nft_ctx *ctx, struct nft_set *set)
2888 list_del_rcu(&set->list);
2889 nf_tables_set_notify(ctx, set, NFT_MSG_DELSET, GFP_ATOMIC);
2890 nft_set_destroy(set);
2893 static int nf_tables_delset(struct net *net, struct sock *nlsk,
2894 struct sk_buff *skb, const struct nlmsghdr *nlh,
2895 const struct nlattr * const nla[])
2897 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
2898 struct nft_set *set;
2902 if (nfmsg->nfgen_family == NFPROTO_UNSPEC)
2903 return -EAFNOSUPPORT;
2904 if (nla[NFTA_SET_TABLE] == NULL)
2907 err = nft_ctx_init_from_setattr(&ctx, net, skb, nlh, nla);
2911 set = nf_tables_set_lookup(ctx.table, nla[NFTA_SET_NAME]);
2913 return PTR_ERR(set);
2914 if (!list_empty(&set->bindings))
2917 return nft_delset(&ctx, set);
2920 static int nf_tables_bind_check_setelem(const struct nft_ctx *ctx,
2921 const struct nft_set *set,
2922 const struct nft_set_iter *iter,
2923 const struct nft_set_elem *elem)
2925 const struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv);
2926 enum nft_registers dreg;
2928 dreg = nft_type_to_reg(set->dtype);
2929 return nft_validate_register_store(ctx, dreg, nft_set_ext_data(ext),
2930 set->dtype == NFT_DATA_VERDICT ?
2931 NFT_DATA_VERDICT : NFT_DATA_VALUE,
2935 int nf_tables_bind_set(const struct nft_ctx *ctx, struct nft_set *set,
2936 struct nft_set_binding *binding)
2938 struct nft_set_binding *i;
2939 struct nft_set_iter iter;
2941 if (!list_empty(&set->bindings) && set->flags & NFT_SET_ANONYMOUS)
2944 if (binding->flags & NFT_SET_MAP) {
2945 /* If the set is already bound to the same chain all
2946 * jumps are already validated for that chain.
2948 list_for_each_entry(i, &set->bindings, list) {
2949 if (i->flags & NFT_SET_MAP &&
2950 i->chain == binding->chain)
2954 iter.genmask = nft_genmask_next(ctx->net);
2958 iter.fn = nf_tables_bind_check_setelem;
2960 set->ops->walk(ctx, set, &iter);
2962 /* Destroy anonymous sets if binding fails */
2963 if (set->flags & NFT_SET_ANONYMOUS)
2964 nf_tables_set_destroy(ctx, set);
2970 binding->chain = ctx->chain;
2971 list_add_tail_rcu(&binding->list, &set->bindings);
2975 void nf_tables_unbind_set(const struct nft_ctx *ctx, struct nft_set *set,
2976 struct nft_set_binding *binding)
2978 list_del_rcu(&binding->list);
2980 if (list_empty(&set->bindings) && set->flags & NFT_SET_ANONYMOUS &&
2981 !(set->flags & NFT_SET_INACTIVE))
2982 nf_tables_set_destroy(ctx, set);
2985 const struct nft_set_ext_type nft_set_ext_types[] = {
2986 [NFT_SET_EXT_KEY] = {
2987 .align = __alignof__(u32),
2989 [NFT_SET_EXT_DATA] = {
2990 .align = __alignof__(u32),
2992 [NFT_SET_EXT_EXPR] = {
2993 .align = __alignof__(struct nft_expr),
2995 [NFT_SET_EXT_FLAGS] = {
2997 .align = __alignof__(u8),
2999 [NFT_SET_EXT_TIMEOUT] = {
3001 .align = __alignof__(u64),
3003 [NFT_SET_EXT_EXPIRATION] = {
3004 .len = sizeof(unsigned long),
3005 .align = __alignof__(unsigned long),
3007 [NFT_SET_EXT_USERDATA] = {
3008 .len = sizeof(struct nft_userdata),
3009 .align = __alignof__(struct nft_userdata),
3012 EXPORT_SYMBOL_GPL(nft_set_ext_types);
3018 static const struct nla_policy nft_set_elem_policy[NFTA_SET_ELEM_MAX + 1] = {
3019 [NFTA_SET_ELEM_KEY] = { .type = NLA_NESTED },
3020 [NFTA_SET_ELEM_DATA] = { .type = NLA_NESTED },
3021 [NFTA_SET_ELEM_FLAGS] = { .type = NLA_U32 },
3022 [NFTA_SET_ELEM_TIMEOUT] = { .type = NLA_U64 },
3023 [NFTA_SET_ELEM_USERDATA] = { .type = NLA_BINARY,
3024 .len = NFT_USERDATA_MAXLEN },
3027 static const struct nla_policy nft_set_elem_list_policy[NFTA_SET_ELEM_LIST_MAX + 1] = {
3028 [NFTA_SET_ELEM_LIST_TABLE] = { .type = NLA_STRING },
3029 [NFTA_SET_ELEM_LIST_SET] = { .type = NLA_STRING },
3030 [NFTA_SET_ELEM_LIST_ELEMENTS] = { .type = NLA_NESTED },
3031 [NFTA_SET_ELEM_LIST_SET_ID] = { .type = NLA_U32 },
3034 static int nft_ctx_init_from_elemattr(struct nft_ctx *ctx, struct net *net,
3035 const struct sk_buff *skb,
3036 const struct nlmsghdr *nlh,
3037 const struct nlattr * const nla[])
3039 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
3040 struct nft_af_info *afi;
3041 struct nft_table *table;
3043 afi = nf_tables_afinfo_lookup(net, nfmsg->nfgen_family, false);
3045 return PTR_ERR(afi);
3047 table = nf_tables_table_lookup(afi, nla[NFTA_SET_ELEM_LIST_TABLE]);
3049 return PTR_ERR(table);
3051 nft_ctx_init(ctx, net, skb, nlh, afi, table, NULL, nla);
3055 static int nf_tables_fill_setelem(struct sk_buff *skb,
3056 const struct nft_set *set,
3057 const struct nft_set_elem *elem)
3059 const struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv);
3060 unsigned char *b = skb_tail_pointer(skb);
3061 struct nlattr *nest;
3063 nest = nla_nest_start(skb, NFTA_LIST_ELEM);
3065 goto nla_put_failure;
3067 if (nft_data_dump(skb, NFTA_SET_ELEM_KEY, nft_set_ext_key(ext),
3068 NFT_DATA_VALUE, set->klen) < 0)
3069 goto nla_put_failure;
3071 if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA) &&
3072 nft_data_dump(skb, NFTA_SET_ELEM_DATA, nft_set_ext_data(ext),
3073 set->dtype == NFT_DATA_VERDICT ? NFT_DATA_VERDICT : NFT_DATA_VALUE,
3075 goto nla_put_failure;
3077 if (nft_set_ext_exists(ext, NFT_SET_EXT_EXPR) &&
3078 nft_expr_dump(skb, NFTA_SET_ELEM_EXPR, nft_set_ext_expr(ext)) < 0)
3079 goto nla_put_failure;
3081 if (nft_set_ext_exists(ext, NFT_SET_EXT_FLAGS) &&
3082 nla_put_be32(skb, NFTA_SET_ELEM_FLAGS,
3083 htonl(*nft_set_ext_flags(ext))))
3084 goto nla_put_failure;
3086 if (nft_set_ext_exists(ext, NFT_SET_EXT_TIMEOUT) &&
3087 nla_put_be64(skb, NFTA_SET_ELEM_TIMEOUT,
3088 cpu_to_be64(*nft_set_ext_timeout(ext)),
3090 goto nla_put_failure;
3092 if (nft_set_ext_exists(ext, NFT_SET_EXT_EXPIRATION)) {
3093 unsigned long expires, now = jiffies;
3095 expires = *nft_set_ext_expiration(ext);
3096 if (time_before(now, expires))
3101 if (nla_put_be64(skb, NFTA_SET_ELEM_EXPIRATION,
3102 cpu_to_be64(jiffies_to_msecs(expires)),
3104 goto nla_put_failure;
3107 if (nft_set_ext_exists(ext, NFT_SET_EXT_USERDATA)) {
3108 struct nft_userdata *udata;
3110 udata = nft_set_ext_userdata(ext);
3111 if (nla_put(skb, NFTA_SET_ELEM_USERDATA,
3112 udata->len + 1, udata->data))
3113 goto nla_put_failure;
3116 nla_nest_end(skb, nest);
3124 struct nft_set_dump_args {
3125 const struct netlink_callback *cb;
3126 struct nft_set_iter iter;
3127 struct sk_buff *skb;
3130 static int nf_tables_dump_setelem(const struct nft_ctx *ctx,
3131 const struct nft_set *set,
3132 const struct nft_set_iter *iter,
3133 const struct nft_set_elem *elem)
3135 struct nft_set_dump_args *args;
3137 args = container_of(iter, struct nft_set_dump_args, iter);
3138 return nf_tables_fill_setelem(args->skb, set, elem);
3141 static int nf_tables_dump_set(struct sk_buff *skb, struct netlink_callback *cb)
3143 struct net *net = sock_net(skb->sk);
3144 const struct nft_set *set;
3145 struct nft_set_dump_args args;
3147 struct nlattr *nla[NFTA_SET_ELEM_LIST_MAX + 1];
3148 struct nfgenmsg *nfmsg;
3149 struct nlmsghdr *nlh;
3150 struct nlattr *nest;
3154 err = nlmsg_parse(cb->nlh, sizeof(struct nfgenmsg), nla,
3155 NFTA_SET_ELEM_LIST_MAX, nft_set_elem_list_policy);
3159 err = nft_ctx_init_from_elemattr(&ctx, net, cb->skb, cb->nlh,
3163 if (ctx.table->flags & NFT_TABLE_INACTIVE)
3166 set = nf_tables_set_lookup(ctx.table, nla[NFTA_SET_ELEM_LIST_SET]);
3168 return PTR_ERR(set);
3169 if (set->flags & NFT_SET_INACTIVE)
3172 event = NFT_MSG_NEWSETELEM;
3173 event |= NFNL_SUBSYS_NFTABLES << 8;
3174 portid = NETLINK_CB(cb->skb).portid;
3175 seq = cb->nlh->nlmsg_seq;
3177 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct nfgenmsg),
3180 goto nla_put_failure;
3182 nfmsg = nlmsg_data(nlh);
3183 nfmsg->nfgen_family = ctx.afi->family;
3184 nfmsg->version = NFNETLINK_V0;
3185 nfmsg->res_id = htons(ctx.net->nft.base_seq & 0xffff);
3187 if (nla_put_string(skb, NFTA_SET_ELEM_LIST_TABLE, ctx.table->name))
3188 goto nla_put_failure;
3189 if (nla_put_string(skb, NFTA_SET_ELEM_LIST_SET, set->name))
3190 goto nla_put_failure;
3192 nest = nla_nest_start(skb, NFTA_SET_ELEM_LIST_ELEMENTS);
3194 goto nla_put_failure;
3198 args.iter.genmask = nft_genmask_cur(ctx.net);
3199 args.iter.skip = cb->args[0];
3200 args.iter.count = 0;
3202 args.iter.fn = nf_tables_dump_setelem;
3203 set->ops->walk(&ctx, set, &args.iter);
3205 nla_nest_end(skb, nest);
3206 nlmsg_end(skb, nlh);
3208 if (args.iter.err && args.iter.err != -EMSGSIZE)
3209 return args.iter.err;
3210 if (args.iter.count == cb->args[0])
3213 cb->args[0] = args.iter.count;
3220 static int nf_tables_getsetelem(struct net *net, struct sock *nlsk,
3221 struct sk_buff *skb, const struct nlmsghdr *nlh,
3222 const struct nlattr * const nla[])
3224 const struct nft_set *set;
3228 err = nft_ctx_init_from_elemattr(&ctx, net, skb, nlh, nla);
3231 if (ctx.table->flags & NFT_TABLE_INACTIVE)
3234 set = nf_tables_set_lookup(ctx.table, nla[NFTA_SET_ELEM_LIST_SET]);
3236 return PTR_ERR(set);
3237 if (set->flags & NFT_SET_INACTIVE)
3240 if (nlh->nlmsg_flags & NLM_F_DUMP) {
3241 struct netlink_dump_control c = {
3242 .dump = nf_tables_dump_set,
3244 return netlink_dump_start(nlsk, skb, nlh, &c);
3249 static int nf_tables_fill_setelem_info(struct sk_buff *skb,
3250 const struct nft_ctx *ctx, u32 seq,
3251 u32 portid, int event, u16 flags,
3252 const struct nft_set *set,
3253 const struct nft_set_elem *elem)
3255 struct nfgenmsg *nfmsg;
3256 struct nlmsghdr *nlh;
3257 struct nlattr *nest;
3260 event |= NFNL_SUBSYS_NFTABLES << 8;
3261 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct nfgenmsg),
3264 goto nla_put_failure;
3266 nfmsg = nlmsg_data(nlh);
3267 nfmsg->nfgen_family = ctx->afi->family;
3268 nfmsg->version = NFNETLINK_V0;
3269 nfmsg->res_id = htons(ctx->net->nft.base_seq & 0xffff);
3271 if (nla_put_string(skb, NFTA_SET_TABLE, ctx->table->name))
3272 goto nla_put_failure;
3273 if (nla_put_string(skb, NFTA_SET_NAME, set->name))
3274 goto nla_put_failure;
3276 nest = nla_nest_start(skb, NFTA_SET_ELEM_LIST_ELEMENTS);
3278 goto nla_put_failure;
3280 err = nf_tables_fill_setelem(skb, set, elem);
3282 goto nla_put_failure;
3284 nla_nest_end(skb, nest);
3286 nlmsg_end(skb, nlh);
3290 nlmsg_trim(skb, nlh);
3294 static int nf_tables_setelem_notify(const struct nft_ctx *ctx,
3295 const struct nft_set *set,
3296 const struct nft_set_elem *elem,
3297 int event, u16 flags)
3299 struct net *net = ctx->net;
3300 u32 portid = ctx->portid;
3301 struct sk_buff *skb;
3304 if (!ctx->report && !nfnetlink_has_listeners(net, NFNLGRP_NFTABLES))
3308 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
3312 err = nf_tables_fill_setelem_info(skb, ctx, 0, portid, event, flags,
3319 err = nfnetlink_send(skb, net, portid, NFNLGRP_NFTABLES, ctx->report,
3323 nfnetlink_set_err(net, portid, NFNLGRP_NFTABLES, err);
3327 static struct nft_trans *nft_trans_elem_alloc(struct nft_ctx *ctx,
3329 struct nft_set *set)
3331 struct nft_trans *trans;
3333 trans = nft_trans_alloc(ctx, msg_type, sizeof(struct nft_trans_elem));
3337 nft_trans_elem_set(trans) = set;
3341 void *nft_set_elem_init(const struct nft_set *set,
3342 const struct nft_set_ext_tmpl *tmpl,
3343 const u32 *key, const u32 *data,
3344 u64 timeout, gfp_t gfp)
3346 struct nft_set_ext *ext;
3349 elem = kzalloc(set->ops->elemsize + tmpl->len, gfp);
3353 ext = nft_set_elem_ext(set, elem);
3354 nft_set_ext_init(ext, tmpl);
3356 memcpy(nft_set_ext_key(ext), key, set->klen);
3357 if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA))
3358 memcpy(nft_set_ext_data(ext), data, set->dlen);
3359 if (nft_set_ext_exists(ext, NFT_SET_EXT_EXPIRATION))
3360 *nft_set_ext_expiration(ext) =
3361 jiffies + msecs_to_jiffies(timeout);
3362 if (nft_set_ext_exists(ext, NFT_SET_EXT_TIMEOUT))
3363 *nft_set_ext_timeout(ext) = timeout;
3368 void nft_set_elem_destroy(const struct nft_set *set, void *elem)
3370 struct nft_set_ext *ext = nft_set_elem_ext(set, elem);
3372 nft_data_uninit(nft_set_ext_key(ext), NFT_DATA_VALUE);
3373 if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA))
3374 nft_data_uninit(nft_set_ext_data(ext), set->dtype);
3375 if (nft_set_ext_exists(ext, NFT_SET_EXT_EXPR))
3376 nf_tables_expr_destroy(NULL, nft_set_ext_expr(ext));
3380 EXPORT_SYMBOL_GPL(nft_set_elem_destroy);
3382 static int nft_setelem_parse_flags(const struct nft_set *set,
3383 const struct nlattr *attr, u32 *flags)
3388 *flags = ntohl(nla_get_be32(attr));
3389 if (*flags & ~NFT_SET_ELEM_INTERVAL_END)
3391 if (!(set->flags & NFT_SET_INTERVAL) &&
3392 *flags & NFT_SET_ELEM_INTERVAL_END)
3398 static int nft_add_set_elem(struct nft_ctx *ctx, struct nft_set *set,
3399 const struct nlattr *attr)
3401 struct nlattr *nla[NFTA_SET_ELEM_MAX + 1];
3402 struct nft_data_desc d1, d2;
3403 struct nft_set_ext_tmpl tmpl;
3404 struct nft_set_ext *ext;
3405 struct nft_set_elem elem;
3406 struct nft_set_binding *binding;
3407 struct nft_userdata *udata;
3408 struct nft_data data;
3409 enum nft_registers dreg;
3410 struct nft_trans *trans;
3416 err = nla_parse_nested(nla, NFTA_SET_ELEM_MAX, attr,
3417 nft_set_elem_policy);
3421 if (nla[NFTA_SET_ELEM_KEY] == NULL)
3424 nft_set_ext_prepare(&tmpl);
3426 err = nft_setelem_parse_flags(set, nla[NFTA_SET_ELEM_FLAGS], &flags);
3430 nft_set_ext_add(&tmpl, NFT_SET_EXT_FLAGS);
3432 if (set->flags & NFT_SET_MAP) {
3433 if (nla[NFTA_SET_ELEM_DATA] == NULL &&
3434 !(flags & NFT_SET_ELEM_INTERVAL_END))
3436 if (nla[NFTA_SET_ELEM_DATA] != NULL &&
3437 flags & NFT_SET_ELEM_INTERVAL_END)
3440 if (nla[NFTA_SET_ELEM_DATA] != NULL)
3445 if (nla[NFTA_SET_ELEM_TIMEOUT] != NULL) {
3446 if (!(set->flags & NFT_SET_TIMEOUT))
3448 timeout = be64_to_cpu(nla_get_be64(nla[NFTA_SET_ELEM_TIMEOUT]));
3449 } else if (set->flags & NFT_SET_TIMEOUT) {
3450 timeout = set->timeout;
3453 err = nft_data_init(ctx, &elem.key.val, sizeof(elem.key), &d1,
3454 nla[NFTA_SET_ELEM_KEY]);
3458 if (d1.type != NFT_DATA_VALUE || d1.len != set->klen)
3461 nft_set_ext_add_length(&tmpl, NFT_SET_EXT_KEY, d1.len);
3463 nft_set_ext_add(&tmpl, NFT_SET_EXT_EXPIRATION);
3464 if (timeout != set->timeout)
3465 nft_set_ext_add(&tmpl, NFT_SET_EXT_TIMEOUT);
3468 if (nla[NFTA_SET_ELEM_DATA] != NULL) {
3469 err = nft_data_init(ctx, &data, sizeof(data), &d2,
3470 nla[NFTA_SET_ELEM_DATA]);
3475 if (set->dtype != NFT_DATA_VERDICT && d2.len != set->dlen)
3478 dreg = nft_type_to_reg(set->dtype);
3479 list_for_each_entry(binding, &set->bindings, list) {
3480 struct nft_ctx bind_ctx = {
3482 .table = ctx->table,
3483 .chain = (struct nft_chain *)binding->chain,
3486 if (!(binding->flags & NFT_SET_MAP))
3489 err = nft_validate_register_store(&bind_ctx, dreg,
3496 nft_set_ext_add_length(&tmpl, NFT_SET_EXT_DATA, d2.len);
3499 /* The full maximum length of userdata can exceed the maximum
3500 * offset value (U8_MAX) for following extensions, therefor it
3501 * must be the last extension added.
3504 if (nla[NFTA_SET_ELEM_USERDATA] != NULL) {
3505 ulen = nla_len(nla[NFTA_SET_ELEM_USERDATA]);
3507 nft_set_ext_add_length(&tmpl, NFT_SET_EXT_USERDATA,
3512 elem.priv = nft_set_elem_init(set, &tmpl, elem.key.val.data, data.data,
3513 timeout, GFP_KERNEL);
3514 if (elem.priv == NULL)
3517 ext = nft_set_elem_ext(set, elem.priv);
3519 *nft_set_ext_flags(ext) = flags;
3521 udata = nft_set_ext_userdata(ext);
3522 udata->len = ulen - 1;
3523 nla_memcpy(&udata->data, nla[NFTA_SET_ELEM_USERDATA], ulen);
3526 trans = nft_trans_elem_alloc(ctx, NFT_MSG_NEWSETELEM, set);
3530 ext->genmask = nft_genmask_cur(ctx->net) | NFT_SET_ELEM_BUSY_MASK;
3531 err = set->ops->insert(set, &elem);
3535 nft_trans_elem(trans) = elem;
3536 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
3544 if (nla[NFTA_SET_ELEM_DATA] != NULL)
3545 nft_data_uninit(&data, d2.type);
3547 nft_data_uninit(&elem.key.val, d1.type);
3552 static int nf_tables_newsetelem(struct net *net, struct sock *nlsk,
3553 struct sk_buff *skb, const struct nlmsghdr *nlh,
3554 const struct nlattr * const nla[])
3556 const struct nlattr *attr;
3557 struct nft_set *set;
3561 if (nla[NFTA_SET_ELEM_LIST_ELEMENTS] == NULL)
3564 err = nft_ctx_init_from_elemattr(&ctx, net, skb, nlh, nla);
3568 set = nf_tables_set_lookup(ctx.table, nla[NFTA_SET_ELEM_LIST_SET]);
3570 if (nla[NFTA_SET_ELEM_LIST_SET_ID]) {
3571 set = nf_tables_set_lookup_byid(net,
3572 nla[NFTA_SET_ELEM_LIST_SET_ID]);
3575 return PTR_ERR(set);
3578 if (!list_empty(&set->bindings) && set->flags & NFT_SET_CONSTANT)
3581 nla_for_each_nested(attr, nla[NFTA_SET_ELEM_LIST_ELEMENTS], rem) {
3583 !atomic_add_unless(&set->nelems, 1, set->size + set->ndeact))
3586 err = nft_add_set_elem(&ctx, set, attr);
3588 atomic_dec(&set->nelems);
3595 static int nft_del_setelem(struct nft_ctx *ctx, struct nft_set *set,
3596 const struct nlattr *attr)
3598 struct nlattr *nla[NFTA_SET_ELEM_MAX + 1];
3599 struct nft_set_ext_tmpl tmpl;
3600 struct nft_data_desc desc;
3601 struct nft_set_elem elem;
3602 struct nft_set_ext *ext;
3603 struct nft_trans *trans;
3608 err = nla_parse_nested(nla, NFTA_SET_ELEM_MAX, attr,
3609 nft_set_elem_policy);
3614 if (nla[NFTA_SET_ELEM_KEY] == NULL)
3617 nft_set_ext_prepare(&tmpl);
3619 err = nft_setelem_parse_flags(set, nla[NFTA_SET_ELEM_FLAGS], &flags);
3623 nft_set_ext_add(&tmpl, NFT_SET_EXT_FLAGS);
3625 err = nft_data_init(ctx, &elem.key.val, sizeof(elem.key), &desc,
3626 nla[NFTA_SET_ELEM_KEY]);
3631 if (desc.type != NFT_DATA_VALUE || desc.len != set->klen)
3634 nft_set_ext_add_length(&tmpl, NFT_SET_EXT_KEY, desc.len);
3637 elem.priv = nft_set_elem_init(set, &tmpl, elem.key.val.data, NULL, 0,
3639 if (elem.priv == NULL)
3642 ext = nft_set_elem_ext(set, elem.priv);
3644 *nft_set_ext_flags(ext) = flags;
3646 trans = nft_trans_elem_alloc(ctx, NFT_MSG_DELSETELEM, set);
3647 if (trans == NULL) {
3652 priv = set->ops->deactivate(set, &elem);
3660 nft_trans_elem(trans) = elem;
3661 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
3669 nft_data_uninit(&elem.key.val, desc.type);
3674 static int nf_tables_delsetelem(struct net *net, struct sock *nlsk,
3675 struct sk_buff *skb, const struct nlmsghdr *nlh,
3676 const struct nlattr * const nla[])
3678 const struct nlattr *attr;
3679 struct nft_set *set;
3683 if (nla[NFTA_SET_ELEM_LIST_ELEMENTS] == NULL)
3686 err = nft_ctx_init_from_elemattr(&ctx, net, skb, nlh, nla);
3690 set = nf_tables_set_lookup(ctx.table, nla[NFTA_SET_ELEM_LIST_SET]);
3692 return PTR_ERR(set);
3693 if (!list_empty(&set->bindings) && set->flags & NFT_SET_CONSTANT)
3696 nla_for_each_nested(attr, nla[NFTA_SET_ELEM_LIST_ELEMENTS], rem) {
3697 err = nft_del_setelem(&ctx, set, attr);
3706 void nft_set_gc_batch_release(struct rcu_head *rcu)
3708 struct nft_set_gc_batch *gcb;
3711 gcb = container_of(rcu, struct nft_set_gc_batch, head.rcu);
3712 for (i = 0; i < gcb->head.cnt; i++)
3713 nft_set_elem_destroy(gcb->head.set, gcb->elems[i]);
3716 EXPORT_SYMBOL_GPL(nft_set_gc_batch_release);
3718 struct nft_set_gc_batch *nft_set_gc_batch_alloc(const struct nft_set *set,
3721 struct nft_set_gc_batch *gcb;
3723 gcb = kzalloc(sizeof(*gcb), gfp);
3726 gcb->head.set = set;
3729 EXPORT_SYMBOL_GPL(nft_set_gc_batch_alloc);
3731 static int nf_tables_fill_gen_info(struct sk_buff *skb, struct net *net,
3732 u32 portid, u32 seq)
3734 struct nlmsghdr *nlh;
3735 struct nfgenmsg *nfmsg;
3736 int event = (NFNL_SUBSYS_NFTABLES << 8) | NFT_MSG_NEWGEN;
3738 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct nfgenmsg), 0);
3740 goto nla_put_failure;
3742 nfmsg = nlmsg_data(nlh);
3743 nfmsg->nfgen_family = AF_UNSPEC;
3744 nfmsg->version = NFNETLINK_V0;
3745 nfmsg->res_id = htons(net->nft.base_seq & 0xffff);
3747 if (nla_put_be32(skb, NFTA_GEN_ID, htonl(net->nft.base_seq)))
3748 goto nla_put_failure;
3750 nlmsg_end(skb, nlh);
3754 nlmsg_trim(skb, nlh);
3758 static int nf_tables_gen_notify(struct net *net, struct sk_buff *skb, int event)
3760 struct nlmsghdr *nlh = nlmsg_hdr(skb);
3761 struct sk_buff *skb2;
3764 if (nlmsg_report(nlh) &&
3765 !nfnetlink_has_listeners(net, NFNLGRP_NFTABLES))
3769 skb2 = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
3773 err = nf_tables_fill_gen_info(skb2, net, NETLINK_CB(skb).portid,
3780 err = nfnetlink_send(skb2, net, NETLINK_CB(skb).portid,
3781 NFNLGRP_NFTABLES, nlmsg_report(nlh), GFP_KERNEL);
3784 nfnetlink_set_err(net, NETLINK_CB(skb).portid, NFNLGRP_NFTABLES,
3790 static int nf_tables_getgen(struct net *net, struct sock *nlsk,
3791 struct sk_buff *skb, const struct nlmsghdr *nlh,
3792 const struct nlattr * const nla[])
3794 struct sk_buff *skb2;
3797 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3801 err = nf_tables_fill_gen_info(skb2, net, NETLINK_CB(skb).portid,
3806 return nlmsg_unicast(nlsk, skb2, NETLINK_CB(skb).portid);
3812 static const struct nfnl_callback nf_tables_cb[NFT_MSG_MAX] = {
3813 [NFT_MSG_NEWTABLE] = {
3814 .call_batch = nf_tables_newtable,
3815 .attr_count = NFTA_TABLE_MAX,
3816 .policy = nft_table_policy,
3818 [NFT_MSG_GETTABLE] = {
3819 .call = nf_tables_gettable,
3820 .attr_count = NFTA_TABLE_MAX,
3821 .policy = nft_table_policy,
3823 [NFT_MSG_DELTABLE] = {
3824 .call_batch = nf_tables_deltable,
3825 .attr_count = NFTA_TABLE_MAX,
3826 .policy = nft_table_policy,
3828 [NFT_MSG_NEWCHAIN] = {
3829 .call_batch = nf_tables_newchain,
3830 .attr_count = NFTA_CHAIN_MAX,
3831 .policy = nft_chain_policy,
3833 [NFT_MSG_GETCHAIN] = {
3834 .call = nf_tables_getchain,
3835 .attr_count = NFTA_CHAIN_MAX,
3836 .policy = nft_chain_policy,
3838 [NFT_MSG_DELCHAIN] = {
3839 .call_batch = nf_tables_delchain,
3840 .attr_count = NFTA_CHAIN_MAX,
3841 .policy = nft_chain_policy,
3843 [NFT_MSG_NEWRULE] = {
3844 .call_batch = nf_tables_newrule,
3845 .attr_count = NFTA_RULE_MAX,
3846 .policy = nft_rule_policy,
3848 [NFT_MSG_GETRULE] = {
3849 .call = nf_tables_getrule,
3850 .attr_count = NFTA_RULE_MAX,
3851 .policy = nft_rule_policy,
3853 [NFT_MSG_DELRULE] = {
3854 .call_batch = nf_tables_delrule,
3855 .attr_count = NFTA_RULE_MAX,
3856 .policy = nft_rule_policy,
3858 [NFT_MSG_NEWSET] = {
3859 .call_batch = nf_tables_newset,
3860 .attr_count = NFTA_SET_MAX,
3861 .policy = nft_set_policy,
3863 [NFT_MSG_GETSET] = {
3864 .call = nf_tables_getset,
3865 .attr_count = NFTA_SET_MAX,
3866 .policy = nft_set_policy,
3868 [NFT_MSG_DELSET] = {
3869 .call_batch = nf_tables_delset,
3870 .attr_count = NFTA_SET_MAX,
3871 .policy = nft_set_policy,
3873 [NFT_MSG_NEWSETELEM] = {
3874 .call_batch = nf_tables_newsetelem,
3875 .attr_count = NFTA_SET_ELEM_LIST_MAX,
3876 .policy = nft_set_elem_list_policy,
3878 [NFT_MSG_GETSETELEM] = {
3879 .call = nf_tables_getsetelem,
3880 .attr_count = NFTA_SET_ELEM_LIST_MAX,
3881 .policy = nft_set_elem_list_policy,
3883 [NFT_MSG_DELSETELEM] = {
3884 .call_batch = nf_tables_delsetelem,
3885 .attr_count = NFTA_SET_ELEM_LIST_MAX,
3886 .policy = nft_set_elem_list_policy,
3888 [NFT_MSG_GETGEN] = {
3889 .call = nf_tables_getgen,
3893 static void nft_chain_commit_update(struct nft_trans *trans)
3895 struct nft_base_chain *basechain;
3897 if (nft_trans_chain_name(trans)[0])
3898 strcpy(trans->ctx.chain->name, nft_trans_chain_name(trans));
3900 if (!(trans->ctx.chain->flags & NFT_BASE_CHAIN))
3903 basechain = nft_base_chain(trans->ctx.chain);
3904 nft_chain_stats_replace(basechain, nft_trans_chain_stats(trans));
3906 switch (nft_trans_chain_policy(trans)) {
3909 basechain->policy = nft_trans_chain_policy(trans);
3914 static void nf_tables_commit_release(struct nft_trans *trans)
3916 switch (trans->msg_type) {
3917 case NFT_MSG_DELTABLE:
3918 nf_tables_table_destroy(&trans->ctx);
3920 case NFT_MSG_DELCHAIN:
3921 nf_tables_chain_destroy(trans->ctx.chain);
3923 case NFT_MSG_DELRULE:
3924 nf_tables_rule_destroy(&trans->ctx, nft_trans_rule(trans));
3926 case NFT_MSG_DELSET:
3927 nft_set_destroy(nft_trans_set(trans));
3929 case NFT_MSG_DELSETELEM:
3930 nft_set_elem_destroy(nft_trans_elem_set(trans),
3931 nft_trans_elem(trans).priv);
3937 static int nf_tables_commit(struct net *net, struct sk_buff *skb)
3939 struct nft_trans *trans, *next;
3940 struct nft_trans_elem *te;
3942 /* Bump generation counter, invalidate any dump in progress */
3943 while (++net->nft.base_seq == 0);
3945 /* A new generation has just started */
3946 net->nft.gencursor = nft_gencursor_next(net);
3948 /* Make sure all packets have left the previous generation before
3949 * purging old rules.
3953 list_for_each_entry_safe(trans, next, &net->nft.commit_list, list) {
3954 switch (trans->msg_type) {
3955 case NFT_MSG_NEWTABLE:
3956 if (nft_trans_table_update(trans)) {
3957 if (!nft_trans_table_enable(trans)) {
3958 nf_tables_table_disable(trans->ctx.afi,
3960 trans->ctx.table->flags |= NFT_TABLE_F_DORMANT;
3963 trans->ctx.table->flags &= ~NFT_TABLE_INACTIVE;
3965 nf_tables_table_notify(&trans->ctx, NFT_MSG_NEWTABLE);
3966 nft_trans_destroy(trans);
3968 case NFT_MSG_DELTABLE:
3969 nf_tables_table_notify(&trans->ctx, NFT_MSG_DELTABLE);
3971 case NFT_MSG_NEWCHAIN:
3972 if (nft_trans_chain_update(trans))
3973 nft_chain_commit_update(trans);
3975 trans->ctx.chain->flags &= ~NFT_CHAIN_INACTIVE;
3977 nf_tables_chain_notify(&trans->ctx, NFT_MSG_NEWCHAIN);
3978 nft_trans_destroy(trans);
3980 case NFT_MSG_DELCHAIN:
3981 nf_tables_chain_notify(&trans->ctx, NFT_MSG_DELCHAIN);
3982 nf_tables_unregister_hooks(trans->ctx.table,
3984 trans->ctx.afi->nops);
3986 case NFT_MSG_NEWRULE:
3987 nft_rule_clear(trans->ctx.net, nft_trans_rule(trans));
3988 nf_tables_rule_notify(&trans->ctx,
3989 nft_trans_rule(trans),
3991 nft_trans_destroy(trans);
3993 case NFT_MSG_DELRULE:
3994 list_del_rcu(&nft_trans_rule(trans)->list);
3995 nf_tables_rule_notify(&trans->ctx,
3996 nft_trans_rule(trans),
3999 case NFT_MSG_NEWSET:
4000 nft_trans_set(trans)->flags &= ~NFT_SET_INACTIVE;
4001 /* This avoids hitting -EBUSY when deleting the table
4002 * from the transaction.
4004 if (nft_trans_set(trans)->flags & NFT_SET_ANONYMOUS &&
4005 !list_empty(&nft_trans_set(trans)->bindings))
4006 trans->ctx.table->use--;
4008 nf_tables_set_notify(&trans->ctx, nft_trans_set(trans),
4009 NFT_MSG_NEWSET, GFP_KERNEL);
4010 nft_trans_destroy(trans);
4012 case NFT_MSG_DELSET:
4013 nf_tables_set_notify(&trans->ctx, nft_trans_set(trans),
4014 NFT_MSG_DELSET, GFP_KERNEL);
4016 case NFT_MSG_NEWSETELEM:
4017 te = (struct nft_trans_elem *)trans->data;
4019 te->set->ops->activate(te->set, &te->elem);
4020 nf_tables_setelem_notify(&trans->ctx, te->set,
4022 NFT_MSG_NEWSETELEM, 0);
4023 nft_trans_destroy(trans);
4025 case NFT_MSG_DELSETELEM:
4026 te = (struct nft_trans_elem *)trans->data;
4028 nf_tables_setelem_notify(&trans->ctx, te->set,
4030 NFT_MSG_DELSETELEM, 0);
4031 te->set->ops->remove(te->set, &te->elem);
4032 atomic_dec(&te->set->nelems);
4040 list_for_each_entry_safe(trans, next, &net->nft.commit_list, list) {
4041 list_del(&trans->list);
4042 nf_tables_commit_release(trans);
4045 nf_tables_gen_notify(net, skb, NFT_MSG_NEWGEN);
4050 static void nf_tables_abort_release(struct nft_trans *trans)
4052 switch (trans->msg_type) {
4053 case NFT_MSG_NEWTABLE:
4054 nf_tables_table_destroy(&trans->ctx);
4056 case NFT_MSG_NEWCHAIN:
4057 nf_tables_chain_destroy(trans->ctx.chain);
4059 case NFT_MSG_NEWRULE:
4060 nf_tables_rule_destroy(&trans->ctx, nft_trans_rule(trans));
4062 case NFT_MSG_NEWSET:
4063 nft_set_destroy(nft_trans_set(trans));
4065 case NFT_MSG_NEWSETELEM:
4066 nft_set_elem_destroy(nft_trans_elem_set(trans),
4067 nft_trans_elem(trans).priv);
4073 static int nf_tables_abort(struct net *net, struct sk_buff *skb)
4075 struct nft_trans *trans, *next;
4076 struct nft_trans_elem *te;
4078 list_for_each_entry_safe_reverse(trans, next, &net->nft.commit_list,
4080 switch (trans->msg_type) {
4081 case NFT_MSG_NEWTABLE:
4082 if (nft_trans_table_update(trans)) {
4083 if (nft_trans_table_enable(trans)) {
4084 nf_tables_table_disable(trans->ctx.afi,
4086 trans->ctx.table->flags |= NFT_TABLE_F_DORMANT;
4088 nft_trans_destroy(trans);
4090 list_del_rcu(&trans->ctx.table->list);
4093 case NFT_MSG_DELTABLE:
4094 list_add_tail_rcu(&trans->ctx.table->list,
4095 &trans->ctx.afi->tables);
4096 nft_trans_destroy(trans);
4098 case NFT_MSG_NEWCHAIN:
4099 if (nft_trans_chain_update(trans)) {
4100 free_percpu(nft_trans_chain_stats(trans));
4102 nft_trans_destroy(trans);
4104 trans->ctx.table->use--;
4105 list_del_rcu(&trans->ctx.chain->list);
4106 nf_tables_unregister_hooks(trans->ctx.table,
4108 trans->ctx.afi->nops);
4111 case NFT_MSG_DELCHAIN:
4112 trans->ctx.table->use++;
4113 list_add_tail_rcu(&trans->ctx.chain->list,
4114 &trans->ctx.table->chains);
4115 nft_trans_destroy(trans);
4117 case NFT_MSG_NEWRULE:
4118 trans->ctx.chain->use--;
4119 list_del_rcu(&nft_trans_rule(trans)->list);
4121 case NFT_MSG_DELRULE:
4122 trans->ctx.chain->use++;
4123 nft_rule_clear(trans->ctx.net, nft_trans_rule(trans));
4124 nft_trans_destroy(trans);
4126 case NFT_MSG_NEWSET:
4127 trans->ctx.table->use--;
4128 list_del_rcu(&nft_trans_set(trans)->list);
4130 case NFT_MSG_DELSET:
4131 trans->ctx.table->use++;
4132 list_add_tail_rcu(&nft_trans_set(trans)->list,
4133 &trans->ctx.table->sets);
4134 nft_trans_destroy(trans);
4136 case NFT_MSG_NEWSETELEM:
4137 te = (struct nft_trans_elem *)trans->data;
4139 te->set->ops->remove(te->set, &te->elem);
4140 atomic_dec(&te->set->nelems);
4142 case NFT_MSG_DELSETELEM:
4143 te = (struct nft_trans_elem *)trans->data;
4145 te->set->ops->activate(te->set, &te->elem);
4148 nft_trans_destroy(trans);
4155 list_for_each_entry_safe_reverse(trans, next,
4156 &net->nft.commit_list, list) {
4157 list_del(&trans->list);
4158 nf_tables_abort_release(trans);
4164 static const struct nfnetlink_subsystem nf_tables_subsys = {
4165 .name = "nf_tables",
4166 .subsys_id = NFNL_SUBSYS_NFTABLES,
4167 .cb_count = NFT_MSG_MAX,
4169 .commit = nf_tables_commit,
4170 .abort = nf_tables_abort,
4173 int nft_chain_validate_dependency(const struct nft_chain *chain,
4174 enum nft_chain_type type)
4176 const struct nft_base_chain *basechain;
4178 if (chain->flags & NFT_BASE_CHAIN) {
4179 basechain = nft_base_chain(chain);
4180 if (basechain->type->type != type)
4185 EXPORT_SYMBOL_GPL(nft_chain_validate_dependency);
4187 int nft_chain_validate_hooks(const struct nft_chain *chain,
4188 unsigned int hook_flags)
4190 struct nft_base_chain *basechain;
4192 if (chain->flags & NFT_BASE_CHAIN) {
4193 basechain = nft_base_chain(chain);
4195 if ((1 << basechain->ops[0].hooknum) & hook_flags)
4203 EXPORT_SYMBOL_GPL(nft_chain_validate_hooks);
4206 * Loop detection - walk through the ruleset beginning at the destination chain
4207 * of a new jump until either the source chain is reached (loop) or all
4208 * reachable chains have been traversed.
4210 * The loop check is performed whenever a new jump verdict is added to an
4211 * expression or verdict map or a verdict map is bound to a new chain.
4214 static int nf_tables_check_loops(const struct nft_ctx *ctx,
4215 const struct nft_chain *chain);
4217 static int nf_tables_loop_check_setelem(const struct nft_ctx *ctx,
4218 const struct nft_set *set,
4219 const struct nft_set_iter *iter,
4220 const struct nft_set_elem *elem)
4222 const struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv);
4223 const struct nft_data *data;
4225 if (nft_set_ext_exists(ext, NFT_SET_EXT_FLAGS) &&
4226 *nft_set_ext_flags(ext) & NFT_SET_ELEM_INTERVAL_END)
4229 data = nft_set_ext_data(ext);
4230 switch (data->verdict.code) {
4233 return nf_tables_check_loops(ctx, data->verdict.chain);
4239 static int nf_tables_check_loops(const struct nft_ctx *ctx,
4240 const struct nft_chain *chain)
4242 const struct nft_rule *rule;
4243 const struct nft_expr *expr, *last;
4244 const struct nft_set *set;
4245 struct nft_set_binding *binding;
4246 struct nft_set_iter iter;
4248 if (ctx->chain == chain)
4251 list_for_each_entry(rule, &chain->rules, list) {
4252 nft_rule_for_each_expr(expr, last, rule) {
4253 const struct nft_data *data = NULL;
4256 if (!expr->ops->validate)
4259 err = expr->ops->validate(ctx, expr, &data);
4266 switch (data->verdict.code) {
4269 err = nf_tables_check_loops(ctx,
4270 data->verdict.chain);
4279 list_for_each_entry(set, &ctx->table->sets, list) {
4280 if (!(set->flags & NFT_SET_MAP) ||
4281 set->dtype != NFT_DATA_VERDICT)
4284 list_for_each_entry(binding, &set->bindings, list) {
4285 if (!(binding->flags & NFT_SET_MAP) ||
4286 binding->chain != chain)
4289 iter.genmask = nft_genmask_next(ctx->net);
4293 iter.fn = nf_tables_loop_check_setelem;
4295 set->ops->walk(ctx, set, &iter);
4305 * nft_parse_register - parse a register value from a netlink attribute
4307 * @attr: netlink attribute
4309 * Parse and translate a register value from a netlink attribute.
4310 * Registers used to be 128 bit wide, these register numbers will be
4311 * mapped to the corresponding 32 bit register numbers.
4313 unsigned int nft_parse_register(const struct nlattr *attr)
4317 reg = ntohl(nla_get_be32(attr));
4319 case NFT_REG_VERDICT...NFT_REG_4:
4320 return reg * NFT_REG_SIZE / NFT_REG32_SIZE;
4322 return reg + NFT_REG_SIZE / NFT_REG32_SIZE - NFT_REG32_00;
4325 EXPORT_SYMBOL_GPL(nft_parse_register);
4328 * nft_dump_register - dump a register value to a netlink attribute
4330 * @skb: socket buffer
4331 * @attr: attribute number
4332 * @reg: register number
4334 * Construct a netlink attribute containing the register number. For
4335 * compatibility reasons, register numbers being a multiple of 4 are
4336 * translated to the corresponding 128 bit register numbers.
4338 int nft_dump_register(struct sk_buff *skb, unsigned int attr, unsigned int reg)
4340 if (reg % (NFT_REG_SIZE / NFT_REG32_SIZE) == 0)
4341 reg = reg / (NFT_REG_SIZE / NFT_REG32_SIZE);
4343 reg = reg - NFT_REG_SIZE / NFT_REG32_SIZE + NFT_REG32_00;
4345 return nla_put_be32(skb, attr, htonl(reg));
4347 EXPORT_SYMBOL_GPL(nft_dump_register);
4350 * nft_validate_register_load - validate a load from a register
4352 * @reg: the register number
4353 * @len: the length of the data
4355 * Validate that the input register is one of the general purpose
4356 * registers and that the length of the load is within the bounds.
4358 int nft_validate_register_load(enum nft_registers reg, unsigned int len)
4360 if (reg < NFT_REG_1 * NFT_REG_SIZE / NFT_REG32_SIZE)
4364 if (reg * NFT_REG32_SIZE + len > FIELD_SIZEOF(struct nft_regs, data))
4369 EXPORT_SYMBOL_GPL(nft_validate_register_load);
4372 * nft_validate_register_store - validate an expressions' register store
4374 * @ctx: context of the expression performing the load
4375 * @reg: the destination register number
4376 * @data: the data to load
4377 * @type: the data type
4378 * @len: the length of the data
4380 * Validate that a data load uses the appropriate data type for
4381 * the destination register and the length is within the bounds.
4382 * A value of NULL for the data means that its runtime gathered
4385 int nft_validate_register_store(const struct nft_ctx *ctx,
4386 enum nft_registers reg,
4387 const struct nft_data *data,
4388 enum nft_data_types type, unsigned int len)
4393 case NFT_REG_VERDICT:
4394 if (type != NFT_DATA_VERDICT)
4398 (data->verdict.code == NFT_GOTO ||
4399 data->verdict.code == NFT_JUMP)) {
4400 err = nf_tables_check_loops(ctx, data->verdict.chain);
4404 if (ctx->chain->level + 1 >
4405 data->verdict.chain->level) {
4406 if (ctx->chain->level + 1 == NFT_JUMP_STACK_SIZE)
4408 data->verdict.chain->level = ctx->chain->level + 1;
4414 if (reg < NFT_REG_1 * NFT_REG_SIZE / NFT_REG32_SIZE)
4418 if (reg * NFT_REG32_SIZE + len >
4419 FIELD_SIZEOF(struct nft_regs, data))
4422 if (data != NULL && type != NFT_DATA_VALUE)
4427 EXPORT_SYMBOL_GPL(nft_validate_register_store);
4429 static const struct nla_policy nft_verdict_policy[NFTA_VERDICT_MAX + 1] = {
4430 [NFTA_VERDICT_CODE] = { .type = NLA_U32 },
4431 [NFTA_VERDICT_CHAIN] = { .type = NLA_STRING,
4432 .len = NFT_CHAIN_MAXNAMELEN - 1 },
4435 static int nft_verdict_init(const struct nft_ctx *ctx, struct nft_data *data,
4436 struct nft_data_desc *desc, const struct nlattr *nla)
4438 struct nlattr *tb[NFTA_VERDICT_MAX + 1];
4439 struct nft_chain *chain;
4442 err = nla_parse_nested(tb, NFTA_VERDICT_MAX, nla, nft_verdict_policy);
4446 if (!tb[NFTA_VERDICT_CODE])
4448 data->verdict.code = ntohl(nla_get_be32(tb[NFTA_VERDICT_CODE]));
4450 switch (data->verdict.code) {
4452 switch (data->verdict.code & NF_VERDICT_MASK) {
4467 if (!tb[NFTA_VERDICT_CHAIN])
4469 chain = nf_tables_chain_lookup(ctx->table,
4470 tb[NFTA_VERDICT_CHAIN]);
4472 return PTR_ERR(chain);
4473 if (chain->flags & NFT_BASE_CHAIN)
4477 data->verdict.chain = chain;
4481 desc->len = sizeof(data->verdict);
4482 desc->type = NFT_DATA_VERDICT;
4486 static void nft_verdict_uninit(const struct nft_data *data)
4488 switch (data->verdict.code) {
4491 data->verdict.chain->use--;
4496 int nft_verdict_dump(struct sk_buff *skb, int type, const struct nft_verdict *v)
4498 struct nlattr *nest;
4500 nest = nla_nest_start(skb, type);
4502 goto nla_put_failure;
4504 if (nla_put_be32(skb, NFTA_VERDICT_CODE, htonl(v->code)))
4505 goto nla_put_failure;
4510 if (nla_put_string(skb, NFTA_VERDICT_CHAIN,
4512 goto nla_put_failure;
4514 nla_nest_end(skb, nest);
4521 static int nft_value_init(const struct nft_ctx *ctx,
4522 struct nft_data *data, unsigned int size,
4523 struct nft_data_desc *desc, const struct nlattr *nla)
4533 nla_memcpy(data->data, nla, len);
4534 desc->type = NFT_DATA_VALUE;
4539 static int nft_value_dump(struct sk_buff *skb, const struct nft_data *data,
4542 return nla_put(skb, NFTA_DATA_VALUE, len, data->data);
4545 static const struct nla_policy nft_data_policy[NFTA_DATA_MAX + 1] = {
4546 [NFTA_DATA_VALUE] = { .type = NLA_BINARY },
4547 [NFTA_DATA_VERDICT] = { .type = NLA_NESTED },
4551 * nft_data_init - parse nf_tables data netlink attributes
4553 * @ctx: context of the expression using the data
4554 * @data: destination struct nft_data
4555 * @size: maximum data length
4556 * @desc: data description
4557 * @nla: netlink attribute containing data
4559 * Parse the netlink data attributes and initialize a struct nft_data.
4560 * The type and length of data are returned in the data description.
4562 * The caller can indicate that it only wants to accept data of type
4563 * NFT_DATA_VALUE by passing NULL for the ctx argument.
4565 int nft_data_init(const struct nft_ctx *ctx,
4566 struct nft_data *data, unsigned int size,
4567 struct nft_data_desc *desc, const struct nlattr *nla)
4569 struct nlattr *tb[NFTA_DATA_MAX + 1];
4572 err = nla_parse_nested(tb, NFTA_DATA_MAX, nla, nft_data_policy);
4576 if (tb[NFTA_DATA_VALUE])
4577 return nft_value_init(ctx, data, size, desc,
4578 tb[NFTA_DATA_VALUE]);
4579 if (tb[NFTA_DATA_VERDICT] && ctx != NULL)
4580 return nft_verdict_init(ctx, data, desc, tb[NFTA_DATA_VERDICT]);
4583 EXPORT_SYMBOL_GPL(nft_data_init);
4586 * nft_data_uninit - release a nft_data item
4588 * @data: struct nft_data to release
4589 * @type: type of data
4591 * Release a nft_data item. NFT_DATA_VALUE types can be silently discarded,
4592 * all others need to be released by calling this function.
4594 void nft_data_uninit(const struct nft_data *data, enum nft_data_types type)
4596 if (type < NFT_DATA_VERDICT)
4599 case NFT_DATA_VERDICT:
4600 return nft_verdict_uninit(data);
4605 EXPORT_SYMBOL_GPL(nft_data_uninit);
4607 int nft_data_dump(struct sk_buff *skb, int attr, const struct nft_data *data,
4608 enum nft_data_types type, unsigned int len)
4610 struct nlattr *nest;
4613 nest = nla_nest_start(skb, attr);
4618 case NFT_DATA_VALUE:
4619 err = nft_value_dump(skb, data, len);
4621 case NFT_DATA_VERDICT:
4622 err = nft_verdict_dump(skb, NFTA_DATA_VERDICT, &data->verdict);
4629 nla_nest_end(skb, nest);
4632 EXPORT_SYMBOL_GPL(nft_data_dump);
4634 static int __net_init nf_tables_init_net(struct net *net)
4636 INIT_LIST_HEAD(&net->nft.af_info);
4637 INIT_LIST_HEAD(&net->nft.commit_list);
4638 net->nft.base_seq = 1;
4642 int __nft_release_basechain(struct nft_ctx *ctx)
4644 struct nft_rule *rule, *nr;
4646 BUG_ON(!(ctx->chain->flags & NFT_BASE_CHAIN));
4648 nf_tables_unregister_hooks(ctx->chain->table, ctx->chain,
4650 list_for_each_entry_safe(rule, nr, &ctx->chain->rules, list) {
4651 list_del(&rule->list);
4653 nf_tables_rule_destroy(ctx, rule);
4655 list_del(&ctx->chain->list);
4657 nf_tables_chain_destroy(ctx->chain);
4661 EXPORT_SYMBOL_GPL(__nft_release_basechain);
4663 /* Called by nft_unregister_afinfo() from __net_exit path, nfnl_lock is held. */
4664 static void __nft_release_afinfo(struct net *net, struct nft_af_info *afi)
4666 struct nft_table *table, *nt;
4667 struct nft_chain *chain, *nc;
4668 struct nft_rule *rule, *nr;
4669 struct nft_set *set, *ns;
4670 struct nft_ctx ctx = {
4675 list_for_each_entry_safe(table, nt, &afi->tables, list) {
4676 list_for_each_entry(chain, &table->chains, list)
4677 nf_tables_unregister_hooks(table, chain, afi->nops);
4678 /* No packets are walking on these chains anymore. */
4680 list_for_each_entry(chain, &table->chains, list) {
4682 list_for_each_entry_safe(rule, nr, &chain->rules, list) {
4683 list_del(&rule->list);
4685 nf_tables_rule_destroy(&ctx, rule);
4688 list_for_each_entry_safe(set, ns, &table->sets, list) {
4689 list_del(&set->list);
4691 nft_set_destroy(set);
4693 list_for_each_entry_safe(chain, nc, &table->chains, list) {
4694 list_del(&chain->list);
4696 nf_tables_chain_destroy(chain);
4698 list_del(&table->list);
4699 nf_tables_table_destroy(&ctx);
4703 static struct pernet_operations nf_tables_net_ops = {
4704 .init = nf_tables_init_net,
4707 static int __init nf_tables_module_init(void)
4711 info = kmalloc(sizeof(struct nft_expr_info) * NFT_RULE_MAXEXPRS,
4718 err = nf_tables_core_module_init();
4722 err = nfnetlink_subsys_register(&nf_tables_subsys);
4726 pr_info("nf_tables: (c) 2007-2009 Patrick McHardy <kaber@trash.net>\n");
4727 return register_pernet_subsys(&nf_tables_net_ops);
4729 nf_tables_core_module_exit();
4736 static void __exit nf_tables_module_exit(void)
4738 unregister_pernet_subsys(&nf_tables_net_ops);
4739 nfnetlink_subsys_unregister(&nf_tables_subsys);
4741 nf_tables_core_module_exit();
4745 module_init(nf_tables_module_init);
4746 module_exit(nf_tables_module_exit);
4748 MODULE_LICENSE("GPL");
4749 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
4750 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_NFTABLES);