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/vmalloc.h>
17 #include <linux/netfilter.h>
18 #include <linux/netfilter/nfnetlink.h>
19 #include <linux/netfilter/nf_tables.h>
20 #include <net/netfilter/nf_flow_table.h>
21 #include <net/netfilter/nf_tables_core.h>
22 #include <net/netfilter/nf_tables.h>
23 #include <net/net_namespace.h>
26 static LIST_HEAD(nf_tables_expressions);
27 static LIST_HEAD(nf_tables_objects);
28 static LIST_HEAD(nf_tables_flowtables);
29 static u64 table_handle;
31 static void nft_ctx_init(struct nft_ctx *ctx,
33 const struct sk_buff *skb,
34 const struct nlmsghdr *nlh,
36 struct nft_table *table,
37 struct nft_chain *chain,
38 const struct nlattr * const *nla)
45 ctx->portid = NETLINK_CB(skb).portid;
46 ctx->report = nlmsg_report(nlh);
47 ctx->seq = nlh->nlmsg_seq;
50 static struct nft_trans *nft_trans_alloc_gfp(const struct nft_ctx *ctx,
51 int msg_type, u32 size, gfp_t gfp)
53 struct nft_trans *trans;
55 trans = kzalloc(sizeof(struct nft_trans) + size, gfp);
59 trans->msg_type = msg_type;
65 static struct nft_trans *nft_trans_alloc(const struct nft_ctx *ctx,
66 int msg_type, u32 size)
68 return nft_trans_alloc_gfp(ctx, msg_type, size, GFP_KERNEL);
71 static void nft_trans_destroy(struct nft_trans *trans)
73 list_del(&trans->list);
77 /* removal requests are queued in the commit_list, but not acted upon
78 * until after all new rules are in place.
80 * Therefore, nf_register_net_hook(net, &nat_hook) runs before pending
81 * nf_unregister_net_hook().
83 * nf_register_net_hook thus fails if a nat hook is already in place
84 * even if the conflicting hook is about to be removed.
86 * If collision is detected, search commit_log for DELCHAIN matching
87 * the new nat hooknum; if we find one collision is temporary:
89 * Either transaction is aborted (new/colliding hook is removed), or
90 * transaction is committed (old hook is removed).
92 static bool nf_tables_allow_nat_conflict(const struct net *net,
93 const struct nf_hook_ops *ops)
95 const struct nft_trans *trans;
101 list_for_each_entry(trans, &net->nft.commit_list, list) {
102 const struct nf_hook_ops *pending_ops;
103 const struct nft_chain *pending;
105 if (trans->msg_type != NFT_MSG_NEWCHAIN &&
106 trans->msg_type != NFT_MSG_DELCHAIN)
109 pending = trans->ctx.chain;
110 if (!nft_is_base_chain(pending))
113 pending_ops = &nft_base_chain(pending)->ops;
114 if (pending_ops->nat_hook &&
115 pending_ops->pf == ops->pf &&
116 pending_ops->hooknum == ops->hooknum) {
117 /* other hook registration already pending? */
118 if (trans->msg_type == NFT_MSG_NEWCHAIN)
128 static int nf_tables_register_hook(struct net *net,
129 const struct nft_table *table,
130 struct nft_chain *chain)
132 struct nf_hook_ops *ops;
135 if (table->flags & NFT_TABLE_F_DORMANT ||
136 !nft_is_base_chain(chain))
139 ops = &nft_base_chain(chain)->ops;
140 ret = nf_register_net_hook(net, ops);
141 if (ret == -EBUSY && nf_tables_allow_nat_conflict(net, ops)) {
142 ops->nat_hook = false;
143 ret = nf_register_net_hook(net, ops);
144 ops->nat_hook = true;
150 static void nf_tables_unregister_hook(struct net *net,
151 const struct nft_table *table,
152 struct nft_chain *chain)
154 if (table->flags & NFT_TABLE_F_DORMANT ||
155 !nft_is_base_chain(chain))
158 nf_unregister_net_hook(net, &nft_base_chain(chain)->ops);
161 static int nft_trans_table_add(struct nft_ctx *ctx, int msg_type)
163 struct nft_trans *trans;
165 trans = nft_trans_alloc(ctx, msg_type, sizeof(struct nft_trans_table));
169 if (msg_type == NFT_MSG_NEWTABLE)
170 nft_activate_next(ctx->net, ctx->table);
172 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
176 static int nft_deltable(struct nft_ctx *ctx)
180 err = nft_trans_table_add(ctx, NFT_MSG_DELTABLE);
184 nft_deactivate_next(ctx->net, ctx->table);
188 static int nft_trans_chain_add(struct nft_ctx *ctx, int msg_type)
190 struct nft_trans *trans;
192 trans = nft_trans_alloc(ctx, msg_type, sizeof(struct nft_trans_chain));
196 if (msg_type == NFT_MSG_NEWCHAIN)
197 nft_activate_next(ctx->net, ctx->chain);
199 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
203 static int nft_delchain(struct nft_ctx *ctx)
207 err = nft_trans_chain_add(ctx, NFT_MSG_DELCHAIN);
212 nft_deactivate_next(ctx->net, ctx->chain);
218 nf_tables_delrule_deactivate(struct nft_ctx *ctx, struct nft_rule *rule)
220 /* You cannot delete the same rule twice */
221 if (nft_is_active_next(ctx->net, rule)) {
222 nft_deactivate_next(ctx->net, rule);
229 static struct nft_trans *nft_trans_rule_add(struct nft_ctx *ctx, int msg_type,
230 struct nft_rule *rule)
232 struct nft_trans *trans;
234 trans = nft_trans_alloc(ctx, msg_type, sizeof(struct nft_trans_rule));
238 if (msg_type == NFT_MSG_NEWRULE && ctx->nla[NFTA_RULE_ID] != NULL) {
239 nft_trans_rule_id(trans) =
240 ntohl(nla_get_be32(ctx->nla[NFTA_RULE_ID]));
242 nft_trans_rule(trans) = rule;
243 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
248 static int nft_delrule(struct nft_ctx *ctx, struct nft_rule *rule)
250 struct nft_trans *trans;
253 trans = nft_trans_rule_add(ctx, NFT_MSG_DELRULE, rule);
257 err = nf_tables_delrule_deactivate(ctx, rule);
259 nft_trans_destroy(trans);
266 static int nft_delrule_by_chain(struct nft_ctx *ctx)
268 struct nft_rule *rule;
271 list_for_each_entry(rule, &ctx->chain->rules, list) {
272 err = nft_delrule(ctx, rule);
279 static int nft_trans_set_add(struct nft_ctx *ctx, int msg_type,
282 struct nft_trans *trans;
284 trans = nft_trans_alloc(ctx, msg_type, sizeof(struct nft_trans_set));
288 if (msg_type == NFT_MSG_NEWSET && ctx->nla[NFTA_SET_ID] != NULL) {
289 nft_trans_set_id(trans) =
290 ntohl(nla_get_be32(ctx->nla[NFTA_SET_ID]));
291 nft_activate_next(ctx->net, set);
293 nft_trans_set(trans) = set;
294 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
299 static int nft_delset(struct nft_ctx *ctx, struct nft_set *set)
303 err = nft_trans_set_add(ctx, NFT_MSG_DELSET, set);
307 nft_deactivate_next(ctx->net, set);
313 static int nft_trans_obj_add(struct nft_ctx *ctx, int msg_type,
314 struct nft_object *obj)
316 struct nft_trans *trans;
318 trans = nft_trans_alloc(ctx, msg_type, sizeof(struct nft_trans_obj));
322 if (msg_type == NFT_MSG_NEWOBJ)
323 nft_activate_next(ctx->net, obj);
325 nft_trans_obj(trans) = obj;
326 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
331 static int nft_delobj(struct nft_ctx *ctx, struct nft_object *obj)
335 err = nft_trans_obj_add(ctx, NFT_MSG_DELOBJ, obj);
339 nft_deactivate_next(ctx->net, obj);
345 static int nft_trans_flowtable_add(struct nft_ctx *ctx, int msg_type,
346 struct nft_flowtable *flowtable)
348 struct nft_trans *trans;
350 trans = nft_trans_alloc(ctx, msg_type,
351 sizeof(struct nft_trans_flowtable));
355 if (msg_type == NFT_MSG_NEWFLOWTABLE)
356 nft_activate_next(ctx->net, flowtable);
358 nft_trans_flowtable(trans) = flowtable;
359 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
364 static int nft_delflowtable(struct nft_ctx *ctx,
365 struct nft_flowtable *flowtable)
369 err = nft_trans_flowtable_add(ctx, NFT_MSG_DELFLOWTABLE, flowtable);
373 nft_deactivate_next(ctx->net, flowtable);
383 static struct nft_table *nft_table_lookup(const struct net *net,
384 const struct nlattr *nla,
385 u8 family, u8 genmask)
387 struct nft_table *table;
389 list_for_each_entry(table, &net->nft.tables, list) {
390 if (!nla_strcmp(nla, table->name) &&
391 table->family == family &&
392 nft_active_genmask(table, genmask))
398 static struct nft_table *nft_table_lookup_byhandle(const struct net *net,
399 const struct nlattr *nla,
402 struct nft_table *table;
404 list_for_each_entry(table, &net->nft.tables, list) {
405 if (be64_to_cpu(nla_get_be64(nla)) == table->handle &&
406 nft_active_genmask(table, genmask))
412 static struct nft_table *nf_tables_table_lookup(const struct net *net,
413 const struct nlattr *nla,
414 u8 family, u8 genmask)
416 struct nft_table *table;
419 return ERR_PTR(-EINVAL);
421 table = nft_table_lookup(net, nla, family, genmask);
425 return ERR_PTR(-ENOENT);
428 static struct nft_table *nf_tables_table_lookup_byhandle(const struct net *net,
429 const struct nlattr *nla,
432 struct nft_table *table;
435 return ERR_PTR(-EINVAL);
437 table = nft_table_lookup_byhandle(net, nla, genmask);
441 return ERR_PTR(-ENOENT);
444 static inline u64 nf_tables_alloc_handle(struct nft_table *table)
446 return ++table->hgenerator;
449 static const struct nft_chain_type *chain_type[NFPROTO_NUMPROTO][NFT_CHAIN_T_MAX];
451 static const struct nft_chain_type *
452 __nf_tables_chain_type_lookup(const struct nlattr *nla, u8 family)
456 for (i = 0; i < NFT_CHAIN_T_MAX; i++) {
457 if (chain_type[family][i] != NULL &&
458 !nla_strcmp(nla, chain_type[family][i]->name))
459 return chain_type[family][i];
464 static const struct nft_chain_type *
465 nf_tables_chain_type_lookup(const struct nlattr *nla, u8 family, bool autoload)
467 const struct nft_chain_type *type;
469 type = __nf_tables_chain_type_lookup(nla, family);
472 #ifdef CONFIG_MODULES
474 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
475 request_module("nft-chain-%u-%.*s", family,
476 nla_len(nla), (const char *)nla_data(nla));
477 nfnl_lock(NFNL_SUBSYS_NFTABLES);
478 type = __nf_tables_chain_type_lookup(nla, family);
480 return ERR_PTR(-EAGAIN);
483 return ERR_PTR(-ENOENT);
486 static const struct nla_policy nft_table_policy[NFTA_TABLE_MAX + 1] = {
487 [NFTA_TABLE_NAME] = { .type = NLA_STRING,
488 .len = NFT_TABLE_MAXNAMELEN - 1 },
489 [NFTA_TABLE_FLAGS] = { .type = NLA_U32 },
490 [NFTA_TABLE_HANDLE] = { .type = NLA_U64 },
493 static int nf_tables_fill_table_info(struct sk_buff *skb, struct net *net,
494 u32 portid, u32 seq, int event, u32 flags,
495 int family, const struct nft_table *table)
497 struct nlmsghdr *nlh;
498 struct nfgenmsg *nfmsg;
500 event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, event);
501 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct nfgenmsg), flags);
503 goto nla_put_failure;
505 nfmsg = nlmsg_data(nlh);
506 nfmsg->nfgen_family = family;
507 nfmsg->version = NFNETLINK_V0;
508 nfmsg->res_id = htons(net->nft.base_seq & 0xffff);
510 if (nla_put_string(skb, NFTA_TABLE_NAME, table->name) ||
511 nla_put_be32(skb, NFTA_TABLE_FLAGS, htonl(table->flags)) ||
512 nla_put_be32(skb, NFTA_TABLE_USE, htonl(table->use)) ||
513 nla_put_be64(skb, NFTA_TABLE_HANDLE, cpu_to_be64(table->handle),
515 goto nla_put_failure;
521 nlmsg_trim(skb, nlh);
525 static void nf_tables_table_notify(const struct nft_ctx *ctx, int event)
531 !nfnetlink_has_listeners(ctx->net, NFNLGRP_NFTABLES))
534 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
538 err = nf_tables_fill_table_info(skb, ctx->net, ctx->portid, ctx->seq,
539 event, 0, ctx->family, ctx->table);
545 nfnetlink_send(skb, ctx->net, ctx->portid, NFNLGRP_NFTABLES,
546 ctx->report, GFP_KERNEL);
549 nfnetlink_set_err(ctx->net, ctx->portid, NFNLGRP_NFTABLES, -ENOBUFS);
552 static int nf_tables_dump_tables(struct sk_buff *skb,
553 struct netlink_callback *cb)
555 const struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
556 const struct nft_table *table;
557 unsigned int idx = 0, s_idx = cb->args[0];
558 struct net *net = sock_net(skb->sk);
559 int family = nfmsg->nfgen_family;
562 cb->seq = net->nft.base_seq;
564 list_for_each_entry_rcu(table, &net->nft.tables, list) {
565 if (family != NFPROTO_UNSPEC && family != table->family)
571 memset(&cb->args[1], 0,
572 sizeof(cb->args) - sizeof(cb->args[0]));
573 if (!nft_is_active(net, table))
575 if (nf_tables_fill_table_info(skb, net,
576 NETLINK_CB(cb->skb).portid,
578 NFT_MSG_NEWTABLE, NLM_F_MULTI,
579 table->family, table) < 0)
582 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
592 static int nf_tables_gettable(struct net *net, struct sock *nlsk,
593 struct sk_buff *skb, const struct nlmsghdr *nlh,
594 const struct nlattr * const nla[],
595 struct netlink_ext_ack *extack)
597 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
598 u8 genmask = nft_genmask_cur(net);
599 const struct nft_table *table;
600 struct sk_buff *skb2;
601 int family = nfmsg->nfgen_family;
604 if (nlh->nlmsg_flags & NLM_F_DUMP) {
605 struct netlink_dump_control c = {
606 .dump = nf_tables_dump_tables,
608 return netlink_dump_start(nlsk, skb, nlh, &c);
611 table = nf_tables_table_lookup(net, nla[NFTA_TABLE_NAME], family,
614 return PTR_ERR(table);
616 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
620 err = nf_tables_fill_table_info(skb2, net, NETLINK_CB(skb).portid,
621 nlh->nlmsg_seq, NFT_MSG_NEWTABLE, 0,
626 return nlmsg_unicast(nlsk, skb2, NETLINK_CB(skb).portid);
633 static void nft_table_disable(struct net *net, struct nft_table *table, u32 cnt)
635 struct nft_chain *chain;
638 list_for_each_entry(chain, &table->chains, list) {
639 if (!nft_is_active_next(net, chain))
641 if (!nft_is_base_chain(chain))
644 if (cnt && i++ == cnt)
647 nf_unregister_net_hook(net, &nft_base_chain(chain)->ops);
651 static int nf_tables_table_enable(struct net *net, struct nft_table *table)
653 struct nft_chain *chain;
656 list_for_each_entry(chain, &table->chains, list) {
657 if (!nft_is_active_next(net, chain))
659 if (!nft_is_base_chain(chain))
662 err = nf_register_net_hook(net, &nft_base_chain(chain)->ops);
671 nft_table_disable(net, table, i);
675 static void nf_tables_table_disable(struct net *net, struct nft_table *table)
677 nft_table_disable(net, table, 0);
680 static int nf_tables_updtable(struct nft_ctx *ctx)
682 struct nft_trans *trans;
686 if (!ctx->nla[NFTA_TABLE_FLAGS])
689 flags = ntohl(nla_get_be32(ctx->nla[NFTA_TABLE_FLAGS]));
690 if (flags & ~NFT_TABLE_F_DORMANT)
693 if (flags == ctx->table->flags)
696 trans = nft_trans_alloc(ctx, NFT_MSG_NEWTABLE,
697 sizeof(struct nft_trans_table));
701 if ((flags & NFT_TABLE_F_DORMANT) &&
702 !(ctx->table->flags & NFT_TABLE_F_DORMANT)) {
703 nft_trans_table_enable(trans) = false;
704 } else if (!(flags & NFT_TABLE_F_DORMANT) &&
705 ctx->table->flags & NFT_TABLE_F_DORMANT) {
706 ret = nf_tables_table_enable(ctx->net, ctx->table);
708 ctx->table->flags &= ~NFT_TABLE_F_DORMANT;
709 nft_trans_table_enable(trans) = true;
715 nft_trans_table_update(trans) = true;
716 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
719 nft_trans_destroy(trans);
723 static int nf_tables_newtable(struct net *net, struct sock *nlsk,
724 struct sk_buff *skb, const struct nlmsghdr *nlh,
725 const struct nlattr * const nla[],
726 struct netlink_ext_ack *extack)
728 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
729 u8 genmask = nft_genmask_next(net);
730 const struct nlattr *name;
731 struct nft_table *table;
732 int family = nfmsg->nfgen_family;
737 name = nla[NFTA_TABLE_NAME];
738 table = nf_tables_table_lookup(net, name, family, genmask);
740 if (PTR_ERR(table) != -ENOENT)
741 return PTR_ERR(table);
743 if (nlh->nlmsg_flags & NLM_F_EXCL)
745 if (nlh->nlmsg_flags & NLM_F_REPLACE)
748 nft_ctx_init(&ctx, net, skb, nlh, family, table, NULL, nla);
749 return nf_tables_updtable(&ctx);
752 if (nla[NFTA_TABLE_FLAGS]) {
753 flags = ntohl(nla_get_be32(nla[NFTA_TABLE_FLAGS]));
754 if (flags & ~NFT_TABLE_F_DORMANT)
759 table = kzalloc(sizeof(*table), GFP_KERNEL);
763 table->name = nla_strdup(name, GFP_KERNEL);
764 if (table->name == NULL)
767 INIT_LIST_HEAD(&table->chains);
768 INIT_LIST_HEAD(&table->sets);
769 INIT_LIST_HEAD(&table->objects);
770 INIT_LIST_HEAD(&table->flowtables);
771 table->family = family;
772 table->flags = flags;
773 table->handle = ++table_handle;
775 nft_ctx_init(&ctx, net, skb, nlh, family, table, NULL, nla);
776 err = nft_trans_table_add(&ctx, NFT_MSG_NEWTABLE);
780 list_add_tail_rcu(&table->list, &net->nft.tables);
790 static int nft_flush_table(struct nft_ctx *ctx)
792 struct nft_flowtable *flowtable, *nft;
793 struct nft_chain *chain, *nc;
794 struct nft_object *obj, *ne;
795 struct nft_set *set, *ns;
798 list_for_each_entry(chain, &ctx->table->chains, list) {
799 if (!nft_is_active_next(ctx->net, chain))
804 err = nft_delrule_by_chain(ctx);
809 list_for_each_entry_safe(set, ns, &ctx->table->sets, list) {
810 if (!nft_is_active_next(ctx->net, set))
813 if (nft_set_is_anonymous(set) &&
814 !list_empty(&set->bindings))
817 err = nft_delset(ctx, set);
822 list_for_each_entry_safe(flowtable, nft, &ctx->table->flowtables, list) {
823 err = nft_delflowtable(ctx, flowtable);
828 list_for_each_entry_safe(obj, ne, &ctx->table->objects, list) {
829 err = nft_delobj(ctx, obj);
834 list_for_each_entry_safe(chain, nc, &ctx->table->chains, list) {
835 if (!nft_is_active_next(ctx->net, chain))
840 err = nft_delchain(ctx);
845 err = nft_deltable(ctx);
850 static int nft_flush(struct nft_ctx *ctx, int family)
852 struct nft_table *table, *nt;
853 const struct nlattr * const *nla = ctx->nla;
856 list_for_each_entry_safe(table, nt, &ctx->net->nft.tables, list) {
857 if (family != AF_UNSPEC && table->family != family)
860 ctx->family = table->family;
862 if (!nft_is_active_next(ctx->net, table))
865 if (nla[NFTA_TABLE_NAME] &&
866 nla_strcmp(nla[NFTA_TABLE_NAME], table->name) != 0)
871 err = nft_flush_table(ctx);
879 static int nf_tables_deltable(struct net *net, struct sock *nlsk,
880 struct sk_buff *skb, const struct nlmsghdr *nlh,
881 const struct nlattr * const nla[],
882 struct netlink_ext_ack *extack)
884 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
885 u8 genmask = nft_genmask_next(net);
886 struct nft_table *table;
887 int family = nfmsg->nfgen_family;
890 nft_ctx_init(&ctx, net, skb, nlh, 0, NULL, NULL, nla);
891 if (family == AF_UNSPEC ||
892 (!nla[NFTA_TABLE_NAME] && !nla[NFTA_TABLE_HANDLE]))
893 return nft_flush(&ctx, family);
895 if (nla[NFTA_TABLE_HANDLE])
896 table = nf_tables_table_lookup_byhandle(net,
897 nla[NFTA_TABLE_HANDLE],
900 table = nf_tables_table_lookup(net, nla[NFTA_TABLE_NAME],
904 return PTR_ERR(table);
906 if (nlh->nlmsg_flags & NLM_F_NONREC &&
913 return nft_flush_table(&ctx);
916 static void nf_tables_table_destroy(struct nft_ctx *ctx)
918 BUG_ON(ctx->table->use > 0);
920 kfree(ctx->table->name);
924 void nft_register_chain_type(const struct nft_chain_type *ctype)
926 if (WARN_ON(ctype->family >= NFPROTO_NUMPROTO))
929 nfnl_lock(NFNL_SUBSYS_NFTABLES);
930 if (WARN_ON(chain_type[ctype->family][ctype->type] != NULL)) {
931 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
934 chain_type[ctype->family][ctype->type] = ctype;
935 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
937 EXPORT_SYMBOL_GPL(nft_register_chain_type);
939 void nft_unregister_chain_type(const struct nft_chain_type *ctype)
941 nfnl_lock(NFNL_SUBSYS_NFTABLES);
942 chain_type[ctype->family][ctype->type] = NULL;
943 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
945 EXPORT_SYMBOL_GPL(nft_unregister_chain_type);
951 static struct nft_chain *
952 nf_tables_chain_lookup_byhandle(const struct nft_table *table, u64 handle,
955 struct nft_chain *chain;
957 list_for_each_entry(chain, &table->chains, list) {
958 if (chain->handle == handle &&
959 nft_active_genmask(chain, genmask))
963 return ERR_PTR(-ENOENT);
966 static struct nft_chain *nf_tables_chain_lookup(const struct nft_table *table,
967 const struct nlattr *nla,
970 struct nft_chain *chain;
973 return ERR_PTR(-EINVAL);
975 list_for_each_entry(chain, &table->chains, list) {
976 if (!nla_strcmp(nla, chain->name) &&
977 nft_active_genmask(chain, genmask))
981 return ERR_PTR(-ENOENT);
984 static const struct nla_policy nft_chain_policy[NFTA_CHAIN_MAX + 1] = {
985 [NFTA_CHAIN_TABLE] = { .type = NLA_STRING,
986 .len = NFT_TABLE_MAXNAMELEN - 1 },
987 [NFTA_CHAIN_HANDLE] = { .type = NLA_U64 },
988 [NFTA_CHAIN_NAME] = { .type = NLA_STRING,
989 .len = NFT_CHAIN_MAXNAMELEN - 1 },
990 [NFTA_CHAIN_HOOK] = { .type = NLA_NESTED },
991 [NFTA_CHAIN_POLICY] = { .type = NLA_U32 },
992 [NFTA_CHAIN_TYPE] = { .type = NLA_STRING },
993 [NFTA_CHAIN_COUNTERS] = { .type = NLA_NESTED },
996 static const struct nla_policy nft_hook_policy[NFTA_HOOK_MAX + 1] = {
997 [NFTA_HOOK_HOOKNUM] = { .type = NLA_U32 },
998 [NFTA_HOOK_PRIORITY] = { .type = NLA_U32 },
999 [NFTA_HOOK_DEV] = { .type = NLA_STRING,
1000 .len = IFNAMSIZ - 1 },
1003 static int nft_dump_stats(struct sk_buff *skb, struct nft_stats __percpu *stats)
1005 struct nft_stats *cpu_stats, total;
1006 struct nlattr *nest;
1011 memset(&total, 0, sizeof(total));
1012 for_each_possible_cpu(cpu) {
1013 cpu_stats = per_cpu_ptr(stats, cpu);
1015 seq = u64_stats_fetch_begin_irq(&cpu_stats->syncp);
1016 pkts = cpu_stats->pkts;
1017 bytes = cpu_stats->bytes;
1018 } while (u64_stats_fetch_retry_irq(&cpu_stats->syncp, seq));
1020 total.bytes += bytes;
1022 nest = nla_nest_start(skb, NFTA_CHAIN_COUNTERS);
1024 goto nla_put_failure;
1026 if (nla_put_be64(skb, NFTA_COUNTER_PACKETS, cpu_to_be64(total.pkts),
1027 NFTA_COUNTER_PAD) ||
1028 nla_put_be64(skb, NFTA_COUNTER_BYTES, cpu_to_be64(total.bytes),
1030 goto nla_put_failure;
1032 nla_nest_end(skb, nest);
1039 static int nf_tables_fill_chain_info(struct sk_buff *skb, struct net *net,
1040 u32 portid, u32 seq, int event, u32 flags,
1041 int family, const struct nft_table *table,
1042 const struct nft_chain *chain)
1044 struct nlmsghdr *nlh;
1045 struct nfgenmsg *nfmsg;
1047 event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, event);
1048 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct nfgenmsg), flags);
1050 goto nla_put_failure;
1052 nfmsg = nlmsg_data(nlh);
1053 nfmsg->nfgen_family = family;
1054 nfmsg->version = NFNETLINK_V0;
1055 nfmsg->res_id = htons(net->nft.base_seq & 0xffff);
1057 if (nla_put_string(skb, NFTA_CHAIN_TABLE, table->name))
1058 goto nla_put_failure;
1059 if (nla_put_be64(skb, NFTA_CHAIN_HANDLE, cpu_to_be64(chain->handle),
1061 goto nla_put_failure;
1062 if (nla_put_string(skb, NFTA_CHAIN_NAME, chain->name))
1063 goto nla_put_failure;
1065 if (nft_is_base_chain(chain)) {
1066 const struct nft_base_chain *basechain = nft_base_chain(chain);
1067 const struct nf_hook_ops *ops = &basechain->ops;
1068 struct nlattr *nest;
1070 nest = nla_nest_start(skb, NFTA_CHAIN_HOOK);
1072 goto nla_put_failure;
1073 if (nla_put_be32(skb, NFTA_HOOK_HOOKNUM, htonl(ops->hooknum)))
1074 goto nla_put_failure;
1075 if (nla_put_be32(skb, NFTA_HOOK_PRIORITY, htonl(ops->priority)))
1076 goto nla_put_failure;
1077 if (basechain->dev_name[0] &&
1078 nla_put_string(skb, NFTA_HOOK_DEV, basechain->dev_name))
1079 goto nla_put_failure;
1080 nla_nest_end(skb, nest);
1082 if (nla_put_be32(skb, NFTA_CHAIN_POLICY,
1083 htonl(basechain->policy)))
1084 goto nla_put_failure;
1086 if (nla_put_string(skb, NFTA_CHAIN_TYPE, basechain->type->name))
1087 goto nla_put_failure;
1089 if (basechain->stats && nft_dump_stats(skb, basechain->stats))
1090 goto nla_put_failure;
1093 if (nla_put_be32(skb, NFTA_CHAIN_USE, htonl(chain->use)))
1094 goto nla_put_failure;
1096 nlmsg_end(skb, nlh);
1100 nlmsg_trim(skb, nlh);
1104 static void nf_tables_chain_notify(const struct nft_ctx *ctx, int event)
1106 struct sk_buff *skb;
1110 !nfnetlink_has_listeners(ctx->net, NFNLGRP_NFTABLES))
1113 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
1117 err = nf_tables_fill_chain_info(skb, ctx->net, ctx->portid, ctx->seq,
1118 event, 0, ctx->family, ctx->table,
1125 nfnetlink_send(skb, ctx->net, ctx->portid, NFNLGRP_NFTABLES,
1126 ctx->report, GFP_KERNEL);
1129 nfnetlink_set_err(ctx->net, ctx->portid, NFNLGRP_NFTABLES, -ENOBUFS);
1132 static int nf_tables_dump_chains(struct sk_buff *skb,
1133 struct netlink_callback *cb)
1135 const struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
1136 const struct nft_table *table;
1137 const struct nft_chain *chain;
1138 unsigned int idx = 0, s_idx = cb->args[0];
1139 struct net *net = sock_net(skb->sk);
1140 int family = nfmsg->nfgen_family;
1143 cb->seq = net->nft.base_seq;
1145 list_for_each_entry_rcu(table, &net->nft.tables, list) {
1146 if (family != NFPROTO_UNSPEC && family != table->family)
1149 list_for_each_entry_rcu(chain, &table->chains, list) {
1153 memset(&cb->args[1], 0,
1154 sizeof(cb->args) - sizeof(cb->args[0]));
1155 if (!nft_is_active(net, chain))
1157 if (nf_tables_fill_chain_info(skb, net,
1158 NETLINK_CB(cb->skb).portid,
1162 table->family, table,
1166 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1177 static int nf_tables_getchain(struct net *net, struct sock *nlsk,
1178 struct sk_buff *skb, const struct nlmsghdr *nlh,
1179 const struct nlattr * const nla[],
1180 struct netlink_ext_ack *extack)
1182 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1183 u8 genmask = nft_genmask_cur(net);
1184 const struct nft_table *table;
1185 const struct nft_chain *chain;
1186 struct sk_buff *skb2;
1187 int family = nfmsg->nfgen_family;
1190 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1191 struct netlink_dump_control c = {
1192 .dump = nf_tables_dump_chains,
1194 return netlink_dump_start(nlsk, skb, nlh, &c);
1197 table = nf_tables_table_lookup(net, nla[NFTA_CHAIN_TABLE], family,
1200 return PTR_ERR(table);
1202 chain = nf_tables_chain_lookup(table, nla[NFTA_CHAIN_NAME], genmask);
1204 return PTR_ERR(chain);
1206 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1210 err = nf_tables_fill_chain_info(skb2, net, NETLINK_CB(skb).portid,
1211 nlh->nlmsg_seq, NFT_MSG_NEWCHAIN, 0,
1212 family, table, chain);
1216 return nlmsg_unicast(nlsk, skb2, NETLINK_CB(skb).portid);
1223 static const struct nla_policy nft_counter_policy[NFTA_COUNTER_MAX + 1] = {
1224 [NFTA_COUNTER_PACKETS] = { .type = NLA_U64 },
1225 [NFTA_COUNTER_BYTES] = { .type = NLA_U64 },
1228 static struct nft_stats __percpu *nft_stats_alloc(const struct nlattr *attr)
1230 struct nlattr *tb[NFTA_COUNTER_MAX+1];
1231 struct nft_stats __percpu *newstats;
1232 struct nft_stats *stats;
1235 err = nla_parse_nested(tb, NFTA_COUNTER_MAX, attr, nft_counter_policy,
1238 return ERR_PTR(err);
1240 if (!tb[NFTA_COUNTER_BYTES] || !tb[NFTA_COUNTER_PACKETS])
1241 return ERR_PTR(-EINVAL);
1243 newstats = netdev_alloc_pcpu_stats(struct nft_stats);
1244 if (newstats == NULL)
1245 return ERR_PTR(-ENOMEM);
1247 /* Restore old counters on this cpu, no problem. Per-cpu statistics
1248 * are not exposed to userspace.
1251 stats = this_cpu_ptr(newstats);
1252 stats->bytes = be64_to_cpu(nla_get_be64(tb[NFTA_COUNTER_BYTES]));
1253 stats->pkts = be64_to_cpu(nla_get_be64(tb[NFTA_COUNTER_PACKETS]));
1259 static void nft_chain_stats_replace(struct nft_base_chain *chain,
1260 struct nft_stats __percpu *newstats)
1262 struct nft_stats __percpu *oldstats;
1264 if (newstats == NULL)
1268 oldstats = nfnl_dereference(chain->stats, NFNL_SUBSYS_NFTABLES);
1269 rcu_assign_pointer(chain->stats, newstats);
1271 free_percpu(oldstats);
1273 rcu_assign_pointer(chain->stats, newstats);
1276 static void nf_tables_chain_destroy(struct nft_ctx *ctx)
1278 struct nft_chain *chain = ctx->chain;
1280 BUG_ON(chain->use > 0);
1282 if (nft_is_base_chain(chain)) {
1283 struct nft_base_chain *basechain = nft_base_chain(chain);
1285 if (basechain->type->free)
1286 basechain->type->free(ctx);
1287 module_put(basechain->type->owner);
1288 free_percpu(basechain->stats);
1289 if (basechain->stats)
1290 static_branch_dec(&nft_counters_enabled);
1299 struct nft_chain_hook {
1302 const struct nft_chain_type *type;
1303 struct net_device *dev;
1306 static int nft_chain_parse_hook(struct net *net,
1307 const struct nlattr * const nla[],
1308 struct nft_chain_hook *hook, u8 family,
1311 struct nlattr *ha[NFTA_HOOK_MAX + 1];
1312 const struct nft_chain_type *type;
1313 struct net_device *dev;
1316 err = nla_parse_nested(ha, NFTA_HOOK_MAX, nla[NFTA_CHAIN_HOOK],
1317 nft_hook_policy, NULL);
1321 if (ha[NFTA_HOOK_HOOKNUM] == NULL ||
1322 ha[NFTA_HOOK_PRIORITY] == NULL)
1325 hook->num = ntohl(nla_get_be32(ha[NFTA_HOOK_HOOKNUM]));
1326 hook->priority = ntohl(nla_get_be32(ha[NFTA_HOOK_PRIORITY]));
1328 type = chain_type[family][NFT_CHAIN_T_DEFAULT];
1329 if (nla[NFTA_CHAIN_TYPE]) {
1330 type = nf_tables_chain_type_lookup(nla[NFTA_CHAIN_TYPE],
1333 return PTR_ERR(type);
1335 if (!(type->hook_mask & (1 << hook->num)))
1338 if (type->type == NFT_CHAIN_T_NAT &&
1339 hook->priority <= NF_IP_PRI_CONNTRACK)
1342 if (!try_module_get(type->owner))
1348 if (family == NFPROTO_NETDEV) {
1349 char ifname[IFNAMSIZ];
1351 if (!ha[NFTA_HOOK_DEV]) {
1352 module_put(type->owner);
1356 nla_strlcpy(ifname, ha[NFTA_HOOK_DEV], IFNAMSIZ);
1357 dev = __dev_get_by_name(net, ifname);
1359 module_put(type->owner);
1363 } else if (ha[NFTA_HOOK_DEV]) {
1364 module_put(type->owner);
1371 static void nft_chain_release_hook(struct nft_chain_hook *hook)
1373 module_put(hook->type->owner);
1376 static int nf_tables_addchain(struct nft_ctx *ctx, u8 family, u8 genmask,
1377 u8 policy, bool create)
1379 const struct nlattr * const *nla = ctx->nla;
1380 struct nft_table *table = ctx->table;
1381 struct nft_base_chain *basechain;
1382 struct nft_stats __percpu *stats;
1383 struct net *net = ctx->net;
1384 struct nft_chain *chain;
1387 if (table->use == UINT_MAX)
1390 if (nla[NFTA_CHAIN_HOOK]) {
1391 struct nft_chain_hook hook;
1392 struct nf_hook_ops *ops;
1394 err = nft_chain_parse_hook(net, nla, &hook, family, create);
1398 basechain = kzalloc(sizeof(*basechain), GFP_KERNEL);
1399 if (basechain == NULL) {
1400 nft_chain_release_hook(&hook);
1404 if (hook.dev != NULL)
1405 strncpy(basechain->dev_name, hook.dev->name, IFNAMSIZ);
1407 if (nla[NFTA_CHAIN_COUNTERS]) {
1408 stats = nft_stats_alloc(nla[NFTA_CHAIN_COUNTERS]);
1409 if (IS_ERR(stats)) {
1410 nft_chain_release_hook(&hook);
1412 return PTR_ERR(stats);
1414 basechain->stats = stats;
1415 static_branch_inc(&nft_counters_enabled);
1418 basechain->type = hook.type;
1419 if (basechain->type->init)
1420 basechain->type->init(ctx);
1422 chain = &basechain->chain;
1424 ops = &basechain->ops;
1426 ops->hooknum = hook.num;
1427 ops->priority = hook.priority;
1429 ops->hook = hook.type->hooks[ops->hooknum];
1430 ops->dev = hook.dev;
1432 if (basechain->type->type == NFT_CHAIN_T_NAT)
1433 ops->nat_hook = true;
1435 chain->flags |= NFT_BASE_CHAIN;
1436 basechain->policy = policy;
1438 chain = kzalloc(sizeof(*chain), GFP_KERNEL);
1444 INIT_LIST_HEAD(&chain->rules);
1445 chain->handle = nf_tables_alloc_handle(table);
1446 chain->table = table;
1447 chain->name = nla_strdup(nla[NFTA_CHAIN_NAME], GFP_KERNEL);
1453 err = nf_tables_register_hook(net, table, chain);
1457 err = nft_trans_chain_add(ctx, NFT_MSG_NEWCHAIN);
1462 list_add_tail_rcu(&chain->list, &table->chains);
1466 nf_tables_unregister_hook(net, table, chain);
1468 nf_tables_chain_destroy(ctx);
1473 static int nf_tables_updchain(struct nft_ctx *ctx, u8 genmask, u8 policy,
1476 const struct nlattr * const *nla = ctx->nla;
1477 struct nft_table *table = ctx->table;
1478 struct nft_chain *chain = ctx->chain;
1479 struct nft_base_chain *basechain;
1480 struct nft_stats *stats = NULL;
1481 struct nft_chain_hook hook;
1482 const struct nlattr *name;
1483 struct nf_hook_ops *ops;
1484 struct nft_trans *trans;
1487 if (nla[NFTA_CHAIN_HOOK]) {
1488 if (!nft_is_base_chain(chain))
1491 err = nft_chain_parse_hook(ctx->net, nla, &hook, ctx->family,
1496 basechain = nft_base_chain(chain);
1497 if (basechain->type != hook.type) {
1498 nft_chain_release_hook(&hook);
1502 ops = &basechain->ops;
1503 if (ops->hooknum != hook.num ||
1504 ops->priority != hook.priority ||
1505 ops->dev != hook.dev) {
1506 nft_chain_release_hook(&hook);
1509 nft_chain_release_hook(&hook);
1512 if (nla[NFTA_CHAIN_HANDLE] &&
1513 nla[NFTA_CHAIN_NAME]) {
1514 struct nft_chain *chain2;
1516 chain2 = nf_tables_chain_lookup(table, nla[NFTA_CHAIN_NAME],
1518 if (!IS_ERR(chain2))
1522 if (nla[NFTA_CHAIN_COUNTERS]) {
1523 if (!nft_is_base_chain(chain))
1526 stats = nft_stats_alloc(nla[NFTA_CHAIN_COUNTERS]);
1528 return PTR_ERR(stats);
1531 trans = nft_trans_alloc(ctx, NFT_MSG_NEWCHAIN,
1532 sizeof(struct nft_trans_chain));
1533 if (trans == NULL) {
1538 nft_trans_chain_stats(trans) = stats;
1539 nft_trans_chain_update(trans) = true;
1541 if (nla[NFTA_CHAIN_POLICY])
1542 nft_trans_chain_policy(trans) = policy;
1544 nft_trans_chain_policy(trans) = -1;
1546 name = nla[NFTA_CHAIN_NAME];
1547 if (nla[NFTA_CHAIN_HANDLE] && name) {
1548 nft_trans_chain_name(trans) =
1549 nla_strdup(name, GFP_KERNEL);
1550 if (!nft_trans_chain_name(trans)) {
1556 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
1561 static int nf_tables_newchain(struct net *net, struct sock *nlsk,
1562 struct sk_buff *skb, const struct nlmsghdr *nlh,
1563 const struct nlattr * const nla[],
1564 struct netlink_ext_ack *extack)
1566 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1567 const struct nlattr * uninitialized_var(name);
1568 u8 genmask = nft_genmask_next(net);
1569 int family = nfmsg->nfgen_family;
1570 struct nft_table *table;
1571 struct nft_chain *chain;
1572 u8 policy = NF_ACCEPT;
1577 create = nlh->nlmsg_flags & NLM_F_CREATE ? true : false;
1579 table = nf_tables_table_lookup(net, nla[NFTA_CHAIN_TABLE], family,
1582 return PTR_ERR(table);
1585 name = nla[NFTA_CHAIN_NAME];
1587 if (nla[NFTA_CHAIN_HANDLE]) {
1588 handle = be64_to_cpu(nla_get_be64(nla[NFTA_CHAIN_HANDLE]));
1589 chain = nf_tables_chain_lookup_byhandle(table, handle, genmask);
1591 return PTR_ERR(chain);
1593 chain = nf_tables_chain_lookup(table, name, genmask);
1594 if (IS_ERR(chain)) {
1595 if (PTR_ERR(chain) != -ENOENT)
1596 return PTR_ERR(chain);
1601 if (nla[NFTA_CHAIN_POLICY]) {
1602 if (chain != NULL &&
1603 !nft_is_base_chain(chain))
1606 if (chain == NULL &&
1607 nla[NFTA_CHAIN_HOOK] == NULL)
1610 policy = ntohl(nla_get_be32(nla[NFTA_CHAIN_POLICY]));
1620 nft_ctx_init(&ctx, net, skb, nlh, family, table, chain, nla);
1622 if (chain != NULL) {
1623 if (nlh->nlmsg_flags & NLM_F_EXCL)
1625 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1628 return nf_tables_updchain(&ctx, genmask, policy, create);
1631 return nf_tables_addchain(&ctx, family, genmask, policy, create);
1634 static int nf_tables_delchain(struct net *net, struct sock *nlsk,
1635 struct sk_buff *skb, const struct nlmsghdr *nlh,
1636 const struct nlattr * const nla[],
1637 struct netlink_ext_ack *extack)
1639 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1640 u8 genmask = nft_genmask_next(net);
1641 struct nft_table *table;
1642 struct nft_chain *chain;
1643 struct nft_rule *rule;
1644 int family = nfmsg->nfgen_family;
1650 table = nf_tables_table_lookup(net, nla[NFTA_CHAIN_TABLE], family,
1653 return PTR_ERR(table);
1655 if (nla[NFTA_CHAIN_HANDLE]) {
1656 handle = be64_to_cpu(nla_get_be64(nla[NFTA_CHAIN_HANDLE]));
1657 chain = nf_tables_chain_lookup_byhandle(table, handle, genmask);
1659 chain = nf_tables_chain_lookup(table, nla[NFTA_CHAIN_NAME], genmask);
1662 return PTR_ERR(chain);
1664 if (nlh->nlmsg_flags & NLM_F_NONREC &&
1668 nft_ctx_init(&ctx, net, skb, nlh, family, table, chain, nla);
1671 list_for_each_entry(rule, &chain->rules, list) {
1672 if (!nft_is_active_next(net, rule))
1676 err = nft_delrule(&ctx, rule);
1681 /* There are rules and elements that are still holding references to us,
1682 * we cannot do a recursive removal in this case.
1687 return nft_delchain(&ctx);
1695 * nft_register_expr - register nf_tables expr type
1698 * Registers the expr type for use with nf_tables. Returns zero on
1699 * success or a negative errno code otherwise.
1701 int nft_register_expr(struct nft_expr_type *type)
1703 nfnl_lock(NFNL_SUBSYS_NFTABLES);
1704 if (type->family == NFPROTO_UNSPEC)
1705 list_add_tail_rcu(&type->list, &nf_tables_expressions);
1707 list_add_rcu(&type->list, &nf_tables_expressions);
1708 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
1711 EXPORT_SYMBOL_GPL(nft_register_expr);
1714 * nft_unregister_expr - unregister nf_tables expr type
1717 * Unregisters the expr typefor use with nf_tables.
1719 void nft_unregister_expr(struct nft_expr_type *type)
1721 nfnl_lock(NFNL_SUBSYS_NFTABLES);
1722 list_del_rcu(&type->list);
1723 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
1725 EXPORT_SYMBOL_GPL(nft_unregister_expr);
1727 static const struct nft_expr_type *__nft_expr_type_get(u8 family,
1730 const struct nft_expr_type *type;
1732 list_for_each_entry(type, &nf_tables_expressions, list) {
1733 if (!nla_strcmp(nla, type->name) &&
1734 (!type->family || type->family == family))
1740 static const struct nft_expr_type *nft_expr_type_get(u8 family,
1743 const struct nft_expr_type *type;
1746 return ERR_PTR(-EINVAL);
1748 type = __nft_expr_type_get(family, nla);
1749 if (type != NULL && try_module_get(type->owner))
1752 #ifdef CONFIG_MODULES
1754 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
1755 request_module("nft-expr-%u-%.*s", family,
1756 nla_len(nla), (char *)nla_data(nla));
1757 nfnl_lock(NFNL_SUBSYS_NFTABLES);
1758 if (__nft_expr_type_get(family, nla))
1759 return ERR_PTR(-EAGAIN);
1761 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
1762 request_module("nft-expr-%.*s",
1763 nla_len(nla), (char *)nla_data(nla));
1764 nfnl_lock(NFNL_SUBSYS_NFTABLES);
1765 if (__nft_expr_type_get(family, nla))
1766 return ERR_PTR(-EAGAIN);
1769 return ERR_PTR(-ENOENT);
1772 static const struct nla_policy nft_expr_policy[NFTA_EXPR_MAX + 1] = {
1773 [NFTA_EXPR_NAME] = { .type = NLA_STRING },
1774 [NFTA_EXPR_DATA] = { .type = NLA_NESTED },
1777 static int nf_tables_fill_expr_info(struct sk_buff *skb,
1778 const struct nft_expr *expr)
1780 if (nla_put_string(skb, NFTA_EXPR_NAME, expr->ops->type->name))
1781 goto nla_put_failure;
1783 if (expr->ops->dump) {
1784 struct nlattr *data = nla_nest_start(skb, NFTA_EXPR_DATA);
1786 goto nla_put_failure;
1787 if (expr->ops->dump(skb, expr) < 0)
1788 goto nla_put_failure;
1789 nla_nest_end(skb, data);
1798 int nft_expr_dump(struct sk_buff *skb, unsigned int attr,
1799 const struct nft_expr *expr)
1801 struct nlattr *nest;
1803 nest = nla_nest_start(skb, attr);
1805 goto nla_put_failure;
1806 if (nf_tables_fill_expr_info(skb, expr) < 0)
1807 goto nla_put_failure;
1808 nla_nest_end(skb, nest);
1815 struct nft_expr_info {
1816 const struct nft_expr_ops *ops;
1817 struct nlattr *tb[NFT_EXPR_MAXATTR + 1];
1820 static int nf_tables_expr_parse(const struct nft_ctx *ctx,
1821 const struct nlattr *nla,
1822 struct nft_expr_info *info)
1824 const struct nft_expr_type *type;
1825 const struct nft_expr_ops *ops;
1826 struct nlattr *tb[NFTA_EXPR_MAX + 1];
1829 err = nla_parse_nested(tb, NFTA_EXPR_MAX, nla, nft_expr_policy, NULL);
1833 type = nft_expr_type_get(ctx->family, tb[NFTA_EXPR_NAME]);
1835 return PTR_ERR(type);
1837 if (tb[NFTA_EXPR_DATA]) {
1838 err = nla_parse_nested(info->tb, type->maxattr,
1839 tb[NFTA_EXPR_DATA], type->policy, NULL);
1843 memset(info->tb, 0, sizeof(info->tb[0]) * (type->maxattr + 1));
1845 if (type->select_ops != NULL) {
1846 ops = type->select_ops(ctx,
1847 (const struct nlattr * const *)info->tb);
1859 module_put(type->owner);
1863 static int nf_tables_newexpr(const struct nft_ctx *ctx,
1864 const struct nft_expr_info *info,
1865 struct nft_expr *expr)
1867 const struct nft_expr_ops *ops = info->ops;
1872 err = ops->init(ctx, expr, (const struct nlattr **)info->tb);
1877 if (ops->validate) {
1878 const struct nft_data *data = NULL;
1880 err = ops->validate(ctx, expr, &data);
1889 ops->destroy(ctx, expr);
1895 static void nf_tables_expr_destroy(const struct nft_ctx *ctx,
1896 struct nft_expr *expr)
1898 if (expr->ops->destroy)
1899 expr->ops->destroy(ctx, expr);
1900 module_put(expr->ops->type->owner);
1903 struct nft_expr *nft_expr_init(const struct nft_ctx *ctx,
1904 const struct nlattr *nla)
1906 struct nft_expr_info info;
1907 struct nft_expr *expr;
1910 err = nf_tables_expr_parse(ctx, nla, &info);
1915 expr = kzalloc(info.ops->size, GFP_KERNEL);
1919 err = nf_tables_newexpr(ctx, &info, expr);
1927 module_put(info.ops->type->owner);
1929 return ERR_PTR(err);
1932 void nft_expr_destroy(const struct nft_ctx *ctx, struct nft_expr *expr)
1934 nf_tables_expr_destroy(ctx, expr);
1942 static struct nft_rule *__nf_tables_rule_lookup(const struct nft_chain *chain,
1945 struct nft_rule *rule;
1947 // FIXME: this sucks
1948 list_for_each_entry(rule, &chain->rules, list) {
1949 if (handle == rule->handle)
1953 return ERR_PTR(-ENOENT);
1956 static struct nft_rule *nf_tables_rule_lookup(const struct nft_chain *chain,
1957 const struct nlattr *nla)
1960 return ERR_PTR(-EINVAL);
1962 return __nf_tables_rule_lookup(chain, be64_to_cpu(nla_get_be64(nla)));
1965 static const struct nla_policy nft_rule_policy[NFTA_RULE_MAX + 1] = {
1966 [NFTA_RULE_TABLE] = { .type = NLA_STRING,
1967 .len = NFT_TABLE_MAXNAMELEN - 1 },
1968 [NFTA_RULE_CHAIN] = { .type = NLA_STRING,
1969 .len = NFT_CHAIN_MAXNAMELEN - 1 },
1970 [NFTA_RULE_HANDLE] = { .type = NLA_U64 },
1971 [NFTA_RULE_EXPRESSIONS] = { .type = NLA_NESTED },
1972 [NFTA_RULE_COMPAT] = { .type = NLA_NESTED },
1973 [NFTA_RULE_POSITION] = { .type = NLA_U64 },
1974 [NFTA_RULE_USERDATA] = { .type = NLA_BINARY,
1975 .len = NFT_USERDATA_MAXLEN },
1976 [NFTA_RULE_ID] = { .type = NLA_U32 },
1979 static int nf_tables_fill_rule_info(struct sk_buff *skb, struct net *net,
1980 u32 portid, u32 seq, int event,
1981 u32 flags, int family,
1982 const struct nft_table *table,
1983 const struct nft_chain *chain,
1984 const struct nft_rule *rule)
1986 struct nlmsghdr *nlh;
1987 struct nfgenmsg *nfmsg;
1988 const struct nft_expr *expr, *next;
1989 struct nlattr *list;
1990 const struct nft_rule *prule;
1991 u16 type = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, event);
1993 nlh = nlmsg_put(skb, portid, seq, type, sizeof(struct nfgenmsg), flags);
1995 goto nla_put_failure;
1997 nfmsg = nlmsg_data(nlh);
1998 nfmsg->nfgen_family = family;
1999 nfmsg->version = NFNETLINK_V0;
2000 nfmsg->res_id = htons(net->nft.base_seq & 0xffff);
2002 if (nla_put_string(skb, NFTA_RULE_TABLE, table->name))
2003 goto nla_put_failure;
2004 if (nla_put_string(skb, NFTA_RULE_CHAIN, chain->name))
2005 goto nla_put_failure;
2006 if (nla_put_be64(skb, NFTA_RULE_HANDLE, cpu_to_be64(rule->handle),
2008 goto nla_put_failure;
2010 if ((event != NFT_MSG_DELRULE) && (rule->list.prev != &chain->rules)) {
2011 prule = list_prev_entry(rule, list);
2012 if (nla_put_be64(skb, NFTA_RULE_POSITION,
2013 cpu_to_be64(prule->handle),
2015 goto nla_put_failure;
2018 list = nla_nest_start(skb, NFTA_RULE_EXPRESSIONS);
2020 goto nla_put_failure;
2021 nft_rule_for_each_expr(expr, next, rule) {
2022 if (nft_expr_dump(skb, NFTA_LIST_ELEM, expr) < 0)
2023 goto nla_put_failure;
2025 nla_nest_end(skb, list);
2028 struct nft_userdata *udata = nft_userdata(rule);
2029 if (nla_put(skb, NFTA_RULE_USERDATA, udata->len + 1,
2031 goto nla_put_failure;
2034 nlmsg_end(skb, nlh);
2038 nlmsg_trim(skb, nlh);
2042 static void nf_tables_rule_notify(const struct nft_ctx *ctx,
2043 const struct nft_rule *rule, int event)
2045 struct sk_buff *skb;
2049 !nfnetlink_has_listeners(ctx->net, NFNLGRP_NFTABLES))
2052 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
2056 err = nf_tables_fill_rule_info(skb, ctx->net, ctx->portid, ctx->seq,
2057 event, 0, ctx->family, ctx->table,
2064 nfnetlink_send(skb, ctx->net, ctx->portid, NFNLGRP_NFTABLES,
2065 ctx->report, GFP_KERNEL);
2068 nfnetlink_set_err(ctx->net, ctx->portid, NFNLGRP_NFTABLES, -ENOBUFS);
2071 struct nft_rule_dump_ctx {
2076 static int nf_tables_dump_rules(struct sk_buff *skb,
2077 struct netlink_callback *cb)
2079 const struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
2080 const struct nft_rule_dump_ctx *ctx = cb->data;
2081 const struct nft_table *table;
2082 const struct nft_chain *chain;
2083 const struct nft_rule *rule;
2084 unsigned int idx = 0, s_idx = cb->args[0];
2085 struct net *net = sock_net(skb->sk);
2086 int family = nfmsg->nfgen_family;
2089 cb->seq = net->nft.base_seq;
2091 list_for_each_entry_rcu(table, &net->nft.tables, list) {
2092 if (family != NFPROTO_UNSPEC && family != table->family)
2095 if (ctx && ctx->table && strcmp(ctx->table, table->name) != 0)
2098 list_for_each_entry_rcu(chain, &table->chains, list) {
2099 if (ctx && ctx->chain &&
2100 strcmp(ctx->chain, chain->name) != 0)
2103 list_for_each_entry_rcu(rule, &chain->rules, list) {
2104 if (!nft_is_active(net, rule))
2109 memset(&cb->args[1], 0,
2110 sizeof(cb->args) - sizeof(cb->args[0]));
2111 if (nf_tables_fill_rule_info(skb, net, NETLINK_CB(cb->skb).portid,
2114 NLM_F_MULTI | NLM_F_APPEND,
2116 table, chain, rule) < 0)
2119 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2132 static int nf_tables_dump_rules_done(struct netlink_callback *cb)
2134 struct nft_rule_dump_ctx *ctx = cb->data;
2144 static int nf_tables_getrule(struct net *net, struct sock *nlsk,
2145 struct sk_buff *skb, const struct nlmsghdr *nlh,
2146 const struct nlattr * const nla[],
2147 struct netlink_ext_ack *extack)
2149 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
2150 u8 genmask = nft_genmask_cur(net);
2151 const struct nft_table *table;
2152 const struct nft_chain *chain;
2153 const struct nft_rule *rule;
2154 struct sk_buff *skb2;
2155 int family = nfmsg->nfgen_family;
2158 if (nlh->nlmsg_flags & NLM_F_DUMP) {
2159 struct netlink_dump_control c = {
2160 .dump = nf_tables_dump_rules,
2161 .done = nf_tables_dump_rules_done,
2164 if (nla[NFTA_RULE_TABLE] || nla[NFTA_RULE_CHAIN]) {
2165 struct nft_rule_dump_ctx *ctx;
2167 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
2171 if (nla[NFTA_RULE_TABLE]) {
2172 ctx->table = nla_strdup(nla[NFTA_RULE_TABLE],
2179 if (nla[NFTA_RULE_CHAIN]) {
2180 ctx->chain = nla_strdup(nla[NFTA_RULE_CHAIN],
2191 return netlink_dump_start(nlsk, skb, nlh, &c);
2194 table = nf_tables_table_lookup(net, nla[NFTA_RULE_TABLE], family,
2197 return PTR_ERR(table);
2199 chain = nf_tables_chain_lookup(table, nla[NFTA_RULE_CHAIN], genmask);
2201 return PTR_ERR(chain);
2203 rule = nf_tables_rule_lookup(chain, nla[NFTA_RULE_HANDLE]);
2205 return PTR_ERR(rule);
2207 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2211 err = nf_tables_fill_rule_info(skb2, net, NETLINK_CB(skb).portid,
2212 nlh->nlmsg_seq, NFT_MSG_NEWRULE, 0,
2213 family, table, chain, rule);
2217 return nlmsg_unicast(nlsk, skb2, NETLINK_CB(skb).portid);
2224 static void nf_tables_rule_destroy(const struct nft_ctx *ctx,
2225 struct nft_rule *rule)
2227 struct nft_expr *expr;
2230 * Careful: some expressions might not be initialized in case this
2231 * is called on error from nf_tables_newrule().
2233 expr = nft_expr_first(rule);
2234 while (expr != nft_expr_last(rule) && expr->ops) {
2235 nf_tables_expr_destroy(ctx, expr);
2236 expr = nft_expr_next(expr);
2241 #define NFT_RULE_MAXEXPRS 128
2243 static struct nft_expr_info *info;
2245 static int nf_tables_newrule(struct net *net, struct sock *nlsk,
2246 struct sk_buff *skb, const struct nlmsghdr *nlh,
2247 const struct nlattr * const nla[],
2248 struct netlink_ext_ack *extack)
2250 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
2251 u8 genmask = nft_genmask_next(net);
2252 int family = nfmsg->nfgen_family;
2253 struct nft_table *table;
2254 struct nft_chain *chain;
2255 struct nft_rule *rule, *old_rule = NULL;
2256 struct nft_userdata *udata;
2257 struct nft_trans *trans = NULL;
2258 struct nft_expr *expr;
2261 unsigned int size, i, n, ulen = 0, usize = 0;
2264 u64 handle, pos_handle;
2266 create = nlh->nlmsg_flags & NLM_F_CREATE ? true : false;
2268 table = nf_tables_table_lookup(net, nla[NFTA_RULE_TABLE], family,
2271 return PTR_ERR(table);
2273 chain = nf_tables_chain_lookup(table, nla[NFTA_RULE_CHAIN], genmask);
2275 return PTR_ERR(chain);
2277 if (nla[NFTA_RULE_HANDLE]) {
2278 handle = be64_to_cpu(nla_get_be64(nla[NFTA_RULE_HANDLE]));
2279 rule = __nf_tables_rule_lookup(chain, handle);
2281 return PTR_ERR(rule);
2283 if (nlh->nlmsg_flags & NLM_F_EXCL)
2285 if (nlh->nlmsg_flags & NLM_F_REPLACE)
2290 if (!create || nlh->nlmsg_flags & NLM_F_REPLACE)
2292 handle = nf_tables_alloc_handle(table);
2294 if (chain->use == UINT_MAX)
2298 if (nla[NFTA_RULE_POSITION]) {
2299 if (!(nlh->nlmsg_flags & NLM_F_CREATE))
2302 pos_handle = be64_to_cpu(nla_get_be64(nla[NFTA_RULE_POSITION]));
2303 old_rule = __nf_tables_rule_lookup(chain, pos_handle);
2304 if (IS_ERR(old_rule))
2305 return PTR_ERR(old_rule);
2308 nft_ctx_init(&ctx, net, skb, nlh, family, table, chain, nla);
2312 if (nla[NFTA_RULE_EXPRESSIONS]) {
2313 nla_for_each_nested(tmp, nla[NFTA_RULE_EXPRESSIONS], rem) {
2315 if (nla_type(tmp) != NFTA_LIST_ELEM)
2317 if (n == NFT_RULE_MAXEXPRS)
2319 err = nf_tables_expr_parse(&ctx, tmp, &info[n]);
2322 size += info[n].ops->size;
2326 /* Check for overflow of dlen field */
2328 if (size >= 1 << 12)
2331 if (nla[NFTA_RULE_USERDATA]) {
2332 ulen = nla_len(nla[NFTA_RULE_USERDATA]);
2334 usize = sizeof(struct nft_userdata) + ulen;
2338 rule = kzalloc(sizeof(*rule) + size + usize, GFP_KERNEL);
2342 nft_activate_next(net, rule);
2344 rule->handle = handle;
2346 rule->udata = ulen ? 1 : 0;
2349 udata = nft_userdata(rule);
2350 udata->len = ulen - 1;
2351 nla_memcpy(udata->data, nla[NFTA_RULE_USERDATA], ulen);
2354 expr = nft_expr_first(rule);
2355 for (i = 0; i < n; i++) {
2356 err = nf_tables_newexpr(&ctx, &info[i], expr);
2360 expr = nft_expr_next(expr);
2363 if (nlh->nlmsg_flags & NLM_F_REPLACE) {
2364 if (nft_is_active_next(net, old_rule)) {
2365 trans = nft_trans_rule_add(&ctx, NFT_MSG_DELRULE,
2367 if (trans == NULL) {
2371 nft_deactivate_next(net, old_rule);
2373 list_add_tail_rcu(&rule->list, &old_rule->list);
2378 } else if (nlh->nlmsg_flags & NLM_F_APPEND)
2380 list_add_rcu(&rule->list, &old_rule->list);
2382 list_add_tail_rcu(&rule->list, &chain->rules);
2385 list_add_tail_rcu(&rule->list, &old_rule->list);
2387 list_add_rcu(&rule->list, &chain->rules);
2390 if (nft_trans_rule_add(&ctx, NFT_MSG_NEWRULE, rule) == NULL) {
2398 list_del_rcu(&rule->list);
2400 nf_tables_rule_destroy(&ctx, rule);
2402 for (i = 0; i < n; i++) {
2403 if (info[i].ops != NULL)
2404 module_put(info[i].ops->type->owner);
2409 static struct nft_rule *nft_rule_lookup_byid(const struct net *net,
2410 const struct nlattr *nla)
2412 u32 id = ntohl(nla_get_be32(nla));
2413 struct nft_trans *trans;
2415 list_for_each_entry(trans, &net->nft.commit_list, list) {
2416 struct nft_rule *rule = nft_trans_rule(trans);
2418 if (trans->msg_type == NFT_MSG_NEWRULE &&
2419 id == nft_trans_rule_id(trans))
2422 return ERR_PTR(-ENOENT);
2425 static int nf_tables_delrule(struct net *net, struct sock *nlsk,
2426 struct sk_buff *skb, const struct nlmsghdr *nlh,
2427 const struct nlattr * const nla[],
2428 struct netlink_ext_ack *extack)
2430 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
2431 u8 genmask = nft_genmask_next(net);
2432 struct nft_table *table;
2433 struct nft_chain *chain = NULL;
2434 struct nft_rule *rule;
2435 int family = nfmsg->nfgen_family, err = 0;
2438 table = nf_tables_table_lookup(net, nla[NFTA_RULE_TABLE], family,
2441 return PTR_ERR(table);
2443 if (nla[NFTA_RULE_CHAIN]) {
2444 chain = nf_tables_chain_lookup(table, nla[NFTA_RULE_CHAIN],
2447 return PTR_ERR(chain);
2450 nft_ctx_init(&ctx, net, skb, nlh, family, table, chain, nla);
2453 if (nla[NFTA_RULE_HANDLE]) {
2454 rule = nf_tables_rule_lookup(chain,
2455 nla[NFTA_RULE_HANDLE]);
2457 return PTR_ERR(rule);
2459 err = nft_delrule(&ctx, rule);
2460 } else if (nla[NFTA_RULE_ID]) {
2461 rule = nft_rule_lookup_byid(net, nla[NFTA_RULE_ID]);
2463 return PTR_ERR(rule);
2465 err = nft_delrule(&ctx, rule);
2467 err = nft_delrule_by_chain(&ctx);
2470 list_for_each_entry(chain, &table->chains, list) {
2471 if (!nft_is_active_next(net, chain))
2475 err = nft_delrule_by_chain(&ctx);
2488 static LIST_HEAD(nf_tables_set_types);
2490 int nft_register_set(struct nft_set_type *type)
2492 nfnl_lock(NFNL_SUBSYS_NFTABLES);
2493 list_add_tail_rcu(&type->list, &nf_tables_set_types);
2494 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
2497 EXPORT_SYMBOL_GPL(nft_register_set);
2499 void nft_unregister_set(struct nft_set_type *type)
2501 nfnl_lock(NFNL_SUBSYS_NFTABLES);
2502 list_del_rcu(&type->list);
2503 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
2505 EXPORT_SYMBOL_GPL(nft_unregister_set);
2507 #define NFT_SET_FEATURES (NFT_SET_INTERVAL | NFT_SET_MAP | \
2508 NFT_SET_TIMEOUT | NFT_SET_OBJECT)
2510 static bool nft_set_ops_candidate(const struct nft_set_ops *ops, u32 flags)
2512 if ((flags & NFT_SET_EVAL) && !ops->update)
2515 return (flags & ops->features) == (flags & NFT_SET_FEATURES);
2519 * Select a set implementation based on the data characteristics and the
2520 * given policy. The total memory use might not be known if no size is
2521 * given, in that case the amount of memory per element is used.
2523 static const struct nft_set_ops *
2524 nft_select_set_ops(const struct nft_ctx *ctx,
2525 const struct nlattr * const nla[],
2526 const struct nft_set_desc *desc,
2527 enum nft_set_policies policy)
2529 const struct nft_set_ops *ops, *bops;
2530 struct nft_set_estimate est, best;
2531 const struct nft_set_type *type;
2534 #ifdef CONFIG_MODULES
2535 if (list_empty(&nf_tables_set_types)) {
2536 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
2537 request_module("nft-set");
2538 nfnl_lock(NFNL_SUBSYS_NFTABLES);
2539 if (!list_empty(&nf_tables_set_types))
2540 return ERR_PTR(-EAGAIN);
2543 if (nla[NFTA_SET_FLAGS] != NULL)
2544 flags = ntohl(nla_get_be32(nla[NFTA_SET_FLAGS]));
2551 list_for_each_entry(type, &nf_tables_set_types, list) {
2552 if (!type->select_ops)
2555 ops = type->select_ops(ctx, desc, flags);
2559 if (!nft_set_ops_candidate(ops, flags))
2561 if (!ops->estimate(desc, flags, &est))
2565 case NFT_SET_POL_PERFORMANCE:
2566 if (est.lookup < best.lookup)
2568 if (est.lookup == best.lookup &&
2569 est.space < best.space)
2572 case NFT_SET_POL_MEMORY:
2574 if (est.space < best.space)
2576 if (est.space == best.space &&
2577 est.lookup < best.lookup)
2579 } else if (est.size < best.size || !bops) {
2587 if (!try_module_get(type->owner))
2590 module_put(bops->type->owner);
2599 return ERR_PTR(-EOPNOTSUPP);
2602 static const struct nla_policy nft_set_policy[NFTA_SET_MAX + 1] = {
2603 [NFTA_SET_TABLE] = { .type = NLA_STRING,
2604 .len = NFT_TABLE_MAXNAMELEN - 1 },
2605 [NFTA_SET_NAME] = { .type = NLA_STRING,
2606 .len = NFT_SET_MAXNAMELEN - 1 },
2607 [NFTA_SET_FLAGS] = { .type = NLA_U32 },
2608 [NFTA_SET_KEY_TYPE] = { .type = NLA_U32 },
2609 [NFTA_SET_KEY_LEN] = { .type = NLA_U32 },
2610 [NFTA_SET_DATA_TYPE] = { .type = NLA_U32 },
2611 [NFTA_SET_DATA_LEN] = { .type = NLA_U32 },
2612 [NFTA_SET_POLICY] = { .type = NLA_U32 },
2613 [NFTA_SET_DESC] = { .type = NLA_NESTED },
2614 [NFTA_SET_ID] = { .type = NLA_U32 },
2615 [NFTA_SET_TIMEOUT] = { .type = NLA_U64 },
2616 [NFTA_SET_GC_INTERVAL] = { .type = NLA_U32 },
2617 [NFTA_SET_USERDATA] = { .type = NLA_BINARY,
2618 .len = NFT_USERDATA_MAXLEN },
2619 [NFTA_SET_OBJ_TYPE] = { .type = NLA_U32 },
2620 [NFTA_SET_HANDLE] = { .type = NLA_U64 },
2623 static const struct nla_policy nft_set_desc_policy[NFTA_SET_DESC_MAX + 1] = {
2624 [NFTA_SET_DESC_SIZE] = { .type = NLA_U32 },
2627 static int nft_ctx_init_from_setattr(struct nft_ctx *ctx, struct net *net,
2628 const struct sk_buff *skb,
2629 const struct nlmsghdr *nlh,
2630 const struct nlattr * const nla[],
2633 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
2634 int family = nfmsg->nfgen_family;
2635 struct nft_table *table = NULL;
2637 if (nla[NFTA_SET_TABLE] != NULL) {
2638 table = nf_tables_table_lookup(net, nla[NFTA_SET_TABLE],
2641 return PTR_ERR(table);
2644 nft_ctx_init(ctx, net, skb, nlh, family, table, NULL, nla);
2648 static struct nft_set *nf_tables_set_lookup(const struct nft_table *table,
2649 const struct nlattr *nla, u8 genmask)
2651 struct nft_set *set;
2654 return ERR_PTR(-EINVAL);
2656 list_for_each_entry(set, &table->sets, list) {
2657 if (!nla_strcmp(nla, set->name) &&
2658 nft_active_genmask(set, genmask))
2661 return ERR_PTR(-ENOENT);
2664 static struct nft_set *nf_tables_set_lookup_byhandle(const struct nft_table *table,
2665 const struct nlattr *nla, u8 genmask)
2667 struct nft_set *set;
2670 return ERR_PTR(-EINVAL);
2672 list_for_each_entry(set, &table->sets, list) {
2673 if (be64_to_cpu(nla_get_be64(nla)) == set->handle &&
2674 nft_active_genmask(set, genmask))
2677 return ERR_PTR(-ENOENT);
2680 static struct nft_set *nf_tables_set_lookup_byid(const struct net *net,
2681 const struct nlattr *nla,
2684 struct nft_trans *trans;
2685 u32 id = ntohl(nla_get_be32(nla));
2687 list_for_each_entry(trans, &net->nft.commit_list, list) {
2688 struct nft_set *set = nft_trans_set(trans);
2690 if (trans->msg_type == NFT_MSG_NEWSET &&
2691 id == nft_trans_set_id(trans) &&
2692 nft_active_genmask(set, genmask))
2695 return ERR_PTR(-ENOENT);
2698 struct nft_set *nft_set_lookup_global(const struct net *net,
2699 const struct nft_table *table,
2700 const struct nlattr *nla_set_name,
2701 const struct nlattr *nla_set_id,
2704 struct nft_set *set;
2706 set = nf_tables_set_lookup(table, nla_set_name, genmask);
2711 set = nf_tables_set_lookup_byid(net, nla_set_id, genmask);
2715 EXPORT_SYMBOL_GPL(nft_set_lookup_global);
2717 static int nf_tables_set_alloc_name(struct nft_ctx *ctx, struct nft_set *set,
2720 const struct nft_set *i;
2722 unsigned long *inuse;
2723 unsigned int n = 0, min = 0;
2725 p = strchr(name, '%');
2727 if (p[1] != 'd' || strchr(p + 2, '%'))
2730 inuse = (unsigned long *)get_zeroed_page(GFP_KERNEL);
2734 list_for_each_entry(i, &ctx->table->sets, list) {
2737 if (!nft_is_active_next(ctx->net, set))
2739 if (!sscanf(i->name, name, &tmp))
2741 if (tmp < min || tmp >= min + BITS_PER_BYTE * PAGE_SIZE)
2744 set_bit(tmp - min, inuse);
2747 n = find_first_zero_bit(inuse, BITS_PER_BYTE * PAGE_SIZE);
2748 if (n >= BITS_PER_BYTE * PAGE_SIZE) {
2749 min += BITS_PER_BYTE * PAGE_SIZE;
2750 memset(inuse, 0, PAGE_SIZE);
2753 free_page((unsigned long)inuse);
2756 set->name = kasprintf(GFP_KERNEL, name, min + n);
2760 list_for_each_entry(i, &ctx->table->sets, list) {
2761 if (!nft_is_active_next(ctx->net, i))
2763 if (!strcmp(set->name, i->name)) {
2771 static int nf_tables_fill_set(struct sk_buff *skb, const struct nft_ctx *ctx,
2772 const struct nft_set *set, u16 event, u16 flags)
2774 struct nfgenmsg *nfmsg;
2775 struct nlmsghdr *nlh;
2776 struct nlattr *desc;
2777 u32 portid = ctx->portid;
2780 event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, event);
2781 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct nfgenmsg),
2784 goto nla_put_failure;
2786 nfmsg = nlmsg_data(nlh);
2787 nfmsg->nfgen_family = ctx->family;
2788 nfmsg->version = NFNETLINK_V0;
2789 nfmsg->res_id = htons(ctx->net->nft.base_seq & 0xffff);
2791 if (nla_put_string(skb, NFTA_SET_TABLE, ctx->table->name))
2792 goto nla_put_failure;
2793 if (nla_put_string(skb, NFTA_SET_NAME, set->name))
2794 goto nla_put_failure;
2795 if (nla_put_be64(skb, NFTA_SET_HANDLE, cpu_to_be64(set->handle),
2797 goto nla_put_failure;
2798 if (set->flags != 0)
2799 if (nla_put_be32(skb, NFTA_SET_FLAGS, htonl(set->flags)))
2800 goto nla_put_failure;
2802 if (nla_put_be32(skb, NFTA_SET_KEY_TYPE, htonl(set->ktype)))
2803 goto nla_put_failure;
2804 if (nla_put_be32(skb, NFTA_SET_KEY_LEN, htonl(set->klen)))
2805 goto nla_put_failure;
2806 if (set->flags & NFT_SET_MAP) {
2807 if (nla_put_be32(skb, NFTA_SET_DATA_TYPE, htonl(set->dtype)))
2808 goto nla_put_failure;
2809 if (nla_put_be32(skb, NFTA_SET_DATA_LEN, htonl(set->dlen)))
2810 goto nla_put_failure;
2812 if (set->flags & NFT_SET_OBJECT &&
2813 nla_put_be32(skb, NFTA_SET_OBJ_TYPE, htonl(set->objtype)))
2814 goto nla_put_failure;
2817 nla_put_be64(skb, NFTA_SET_TIMEOUT,
2818 cpu_to_be64(jiffies_to_msecs(set->timeout)),
2820 goto nla_put_failure;
2822 nla_put_be32(skb, NFTA_SET_GC_INTERVAL, htonl(set->gc_int)))
2823 goto nla_put_failure;
2825 if (set->policy != NFT_SET_POL_PERFORMANCE) {
2826 if (nla_put_be32(skb, NFTA_SET_POLICY, htonl(set->policy)))
2827 goto nla_put_failure;
2830 if (nla_put(skb, NFTA_SET_USERDATA, set->udlen, set->udata))
2831 goto nla_put_failure;
2833 desc = nla_nest_start(skb, NFTA_SET_DESC);
2835 goto nla_put_failure;
2837 nla_put_be32(skb, NFTA_SET_DESC_SIZE, htonl(set->size)))
2838 goto nla_put_failure;
2839 nla_nest_end(skb, desc);
2841 nlmsg_end(skb, nlh);
2845 nlmsg_trim(skb, nlh);
2849 static void nf_tables_set_notify(const struct nft_ctx *ctx,
2850 const struct nft_set *set, int event,
2853 struct sk_buff *skb;
2854 u32 portid = ctx->portid;
2858 !nfnetlink_has_listeners(ctx->net, NFNLGRP_NFTABLES))
2861 skb = nlmsg_new(NLMSG_GOODSIZE, gfp_flags);
2865 err = nf_tables_fill_set(skb, ctx, set, event, 0);
2871 nfnetlink_send(skb, ctx->net, portid, NFNLGRP_NFTABLES, ctx->report,
2875 nfnetlink_set_err(ctx->net, portid, NFNLGRP_NFTABLES, -ENOBUFS);
2878 static int nf_tables_dump_sets(struct sk_buff *skb, struct netlink_callback *cb)
2880 const struct nft_set *set;
2881 unsigned int idx, s_idx = cb->args[0];
2882 struct nft_table *table, *cur_table = (struct nft_table *)cb->args[2];
2883 struct net *net = sock_net(skb->sk);
2884 struct nft_ctx *ctx = cb->data, ctx_set;
2890 cb->seq = net->nft.base_seq;
2892 list_for_each_entry_rcu(table, &net->nft.tables, list) {
2893 if (ctx->family != NFPROTO_UNSPEC &&
2894 ctx->family != table->family)
2897 if (ctx->table && ctx->table != table)
2901 if (cur_table != table)
2907 list_for_each_entry_rcu(set, &table->sets, list) {
2910 if (!nft_is_active(net, set))
2914 ctx_set.table = table;
2915 ctx_set.family = table->family;
2917 if (nf_tables_fill_set(skb, &ctx_set, set,
2921 cb->args[2] = (unsigned long) table;
2924 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2937 static int nf_tables_dump_sets_done(struct netlink_callback *cb)
2943 static int nf_tables_getset(struct net *net, struct sock *nlsk,
2944 struct sk_buff *skb, const struct nlmsghdr *nlh,
2945 const struct nlattr * const nla[],
2946 struct netlink_ext_ack *extack)
2948 u8 genmask = nft_genmask_cur(net);
2949 const struct nft_set *set;
2951 struct sk_buff *skb2;
2952 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
2955 /* Verify existence before starting dump */
2956 err = nft_ctx_init_from_setattr(&ctx, net, skb, nlh, nla, genmask);
2960 if (nlh->nlmsg_flags & NLM_F_DUMP) {
2961 struct netlink_dump_control c = {
2962 .dump = nf_tables_dump_sets,
2963 .done = nf_tables_dump_sets_done,
2965 struct nft_ctx *ctx_dump;
2967 ctx_dump = kmalloc(sizeof(*ctx_dump), GFP_KERNEL);
2968 if (ctx_dump == NULL)
2974 return netlink_dump_start(nlsk, skb, nlh, &c);
2977 /* Only accept unspec with dump */
2978 if (nfmsg->nfgen_family == NFPROTO_UNSPEC)
2979 return -EAFNOSUPPORT;
2980 if (!nla[NFTA_SET_TABLE])
2983 set = nf_tables_set_lookup(ctx.table, nla[NFTA_SET_NAME], genmask);
2985 return PTR_ERR(set);
2987 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2991 err = nf_tables_fill_set(skb2, &ctx, set, NFT_MSG_NEWSET, 0);
2995 return nlmsg_unicast(nlsk, skb2, NETLINK_CB(skb).portid);
3002 static int nf_tables_set_desc_parse(const struct nft_ctx *ctx,
3003 struct nft_set_desc *desc,
3004 const struct nlattr *nla)
3006 struct nlattr *da[NFTA_SET_DESC_MAX + 1];
3009 err = nla_parse_nested(da, NFTA_SET_DESC_MAX, nla,
3010 nft_set_desc_policy, NULL);
3014 if (da[NFTA_SET_DESC_SIZE] != NULL)
3015 desc->size = ntohl(nla_get_be32(da[NFTA_SET_DESC_SIZE]));
3020 static int nf_tables_newset(struct net *net, struct sock *nlsk,
3021 struct sk_buff *skb, const struct nlmsghdr *nlh,
3022 const struct nlattr * const nla[],
3023 struct netlink_ext_ack *extack)
3025 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
3026 u8 genmask = nft_genmask_next(net);
3027 int family = nfmsg->nfgen_family;
3028 const struct nft_set_ops *ops;
3029 struct nft_table *table;
3030 struct nft_set *set;
3036 u32 ktype, dtype, flags, policy, gc_int, objtype;
3037 struct nft_set_desc desc;
3038 unsigned char *udata;
3042 if (nla[NFTA_SET_TABLE] == NULL ||
3043 nla[NFTA_SET_NAME] == NULL ||
3044 nla[NFTA_SET_KEY_LEN] == NULL ||
3045 nla[NFTA_SET_ID] == NULL)
3048 memset(&desc, 0, sizeof(desc));
3050 ktype = NFT_DATA_VALUE;
3051 if (nla[NFTA_SET_KEY_TYPE] != NULL) {
3052 ktype = ntohl(nla_get_be32(nla[NFTA_SET_KEY_TYPE]));
3053 if ((ktype & NFT_DATA_RESERVED_MASK) == NFT_DATA_RESERVED_MASK)
3057 desc.klen = ntohl(nla_get_be32(nla[NFTA_SET_KEY_LEN]));
3058 if (desc.klen == 0 || desc.klen > NFT_DATA_VALUE_MAXLEN)
3062 if (nla[NFTA_SET_FLAGS] != NULL) {
3063 flags = ntohl(nla_get_be32(nla[NFTA_SET_FLAGS]));
3064 if (flags & ~(NFT_SET_ANONYMOUS | NFT_SET_CONSTANT |
3065 NFT_SET_INTERVAL | NFT_SET_TIMEOUT |
3066 NFT_SET_MAP | NFT_SET_EVAL |
3069 /* Only one of these operations is supported */
3070 if ((flags & (NFT_SET_MAP | NFT_SET_EVAL | NFT_SET_OBJECT)) ==
3071 (NFT_SET_MAP | NFT_SET_EVAL | NFT_SET_OBJECT))
3076 if (nla[NFTA_SET_DATA_TYPE] != NULL) {
3077 if (!(flags & NFT_SET_MAP))
3080 dtype = ntohl(nla_get_be32(nla[NFTA_SET_DATA_TYPE]));
3081 if ((dtype & NFT_DATA_RESERVED_MASK) == NFT_DATA_RESERVED_MASK &&
3082 dtype != NFT_DATA_VERDICT)
3085 if (dtype != NFT_DATA_VERDICT) {
3086 if (nla[NFTA_SET_DATA_LEN] == NULL)
3088 desc.dlen = ntohl(nla_get_be32(nla[NFTA_SET_DATA_LEN]));
3089 if (desc.dlen == 0 || desc.dlen > NFT_DATA_VALUE_MAXLEN)
3092 desc.dlen = sizeof(struct nft_verdict);
3093 } else if (flags & NFT_SET_MAP)
3096 if (nla[NFTA_SET_OBJ_TYPE] != NULL) {
3097 if (!(flags & NFT_SET_OBJECT))
3100 objtype = ntohl(nla_get_be32(nla[NFTA_SET_OBJ_TYPE]));
3101 if (objtype == NFT_OBJECT_UNSPEC ||
3102 objtype > NFT_OBJECT_MAX)
3104 } else if (flags & NFT_SET_OBJECT)
3107 objtype = NFT_OBJECT_UNSPEC;
3110 if (nla[NFTA_SET_TIMEOUT] != NULL) {
3111 if (!(flags & NFT_SET_TIMEOUT))
3113 timeout = msecs_to_jiffies(be64_to_cpu(nla_get_be64(
3114 nla[NFTA_SET_TIMEOUT])));
3117 if (nla[NFTA_SET_GC_INTERVAL] != NULL) {
3118 if (!(flags & NFT_SET_TIMEOUT))
3120 gc_int = ntohl(nla_get_be32(nla[NFTA_SET_GC_INTERVAL]));
3123 policy = NFT_SET_POL_PERFORMANCE;
3124 if (nla[NFTA_SET_POLICY] != NULL)
3125 policy = ntohl(nla_get_be32(nla[NFTA_SET_POLICY]));
3127 if (nla[NFTA_SET_DESC] != NULL) {
3128 err = nf_tables_set_desc_parse(&ctx, &desc, nla[NFTA_SET_DESC]);
3133 create = nlh->nlmsg_flags & NLM_F_CREATE ? true : false;
3135 table = nf_tables_table_lookup(net, nla[NFTA_SET_TABLE], family,
3138 return PTR_ERR(table);
3140 nft_ctx_init(&ctx, net, skb, nlh, family, table, NULL, nla);
3142 set = nf_tables_set_lookup(table, nla[NFTA_SET_NAME], genmask);
3144 if (PTR_ERR(set) != -ENOENT)
3145 return PTR_ERR(set);
3147 if (nlh->nlmsg_flags & NLM_F_EXCL)
3149 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3154 if (!(nlh->nlmsg_flags & NLM_F_CREATE))
3157 ops = nft_select_set_ops(&ctx, nla, &desc, policy);
3159 return PTR_ERR(ops);
3162 if (nla[NFTA_SET_USERDATA])
3163 udlen = nla_len(nla[NFTA_SET_USERDATA]);
3166 if (ops->privsize != NULL)
3167 size = ops->privsize(nla, &desc);
3169 set = kvzalloc(sizeof(*set) + size + udlen, GFP_KERNEL);
3175 name = nla_strdup(nla[NFTA_SET_NAME], GFP_KERNEL);
3181 err = nf_tables_set_alloc_name(&ctx, set, name);
3188 udata = set->data + size;
3189 nla_memcpy(udata, nla[NFTA_SET_USERDATA], udlen);
3192 INIT_LIST_HEAD(&set->bindings);
3195 set->klen = desc.klen;
3197 set->objtype = objtype;
3198 set->dlen = desc.dlen;
3200 set->size = desc.size;
3201 set->policy = policy;
3204 set->timeout = timeout;
3205 set->gc_int = gc_int;
3206 set->handle = nf_tables_alloc_handle(table);
3208 err = ops->init(set, &desc, nla);
3212 err = nft_trans_set_add(&ctx, NFT_MSG_NEWSET, set);
3216 list_add_tail_rcu(&set->list, &table->sets);
3225 module_put(ops->type->owner);
3229 static void nft_set_destroy(struct nft_set *set)
3231 set->ops->destroy(set);
3232 module_put(set->ops->type->owner);
3237 static void nf_tables_set_destroy(const struct nft_ctx *ctx, struct nft_set *set)
3239 list_del_rcu(&set->list);
3240 nf_tables_set_notify(ctx, set, NFT_MSG_DELSET, GFP_ATOMIC);
3241 nft_set_destroy(set);
3244 static int nf_tables_delset(struct net *net, struct sock *nlsk,
3245 struct sk_buff *skb, const struct nlmsghdr *nlh,
3246 const struct nlattr * const nla[],
3247 struct netlink_ext_ack *extack)
3249 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
3250 u8 genmask = nft_genmask_next(net);
3251 struct nft_set *set;
3255 if (nfmsg->nfgen_family == NFPROTO_UNSPEC)
3256 return -EAFNOSUPPORT;
3257 if (nla[NFTA_SET_TABLE] == NULL)
3260 err = nft_ctx_init_from_setattr(&ctx, net, skb, nlh, nla, genmask);
3264 if (nla[NFTA_SET_HANDLE])
3265 set = nf_tables_set_lookup_byhandle(ctx.table, nla[NFTA_SET_HANDLE], genmask);
3267 set = nf_tables_set_lookup(ctx.table, nla[NFTA_SET_NAME], genmask);
3269 return PTR_ERR(set);
3271 if (!list_empty(&set->bindings) ||
3272 (nlh->nlmsg_flags & NLM_F_NONREC && atomic_read(&set->nelems) > 0))
3275 return nft_delset(&ctx, set);
3278 static int nf_tables_bind_check_setelem(const struct nft_ctx *ctx,
3279 struct nft_set *set,
3280 const struct nft_set_iter *iter,
3281 struct nft_set_elem *elem)
3283 const struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv);
3284 enum nft_registers dreg;
3286 dreg = nft_type_to_reg(set->dtype);
3287 return nft_validate_register_store(ctx, dreg, nft_set_ext_data(ext),
3288 set->dtype == NFT_DATA_VERDICT ?
3289 NFT_DATA_VERDICT : NFT_DATA_VALUE,
3293 int nf_tables_bind_set(const struct nft_ctx *ctx, struct nft_set *set,
3294 struct nft_set_binding *binding)
3296 struct nft_set_binding *i;
3297 struct nft_set_iter iter;
3299 if (!list_empty(&set->bindings) && nft_set_is_anonymous(set))
3302 if (binding->flags & NFT_SET_MAP) {
3303 /* If the set is already bound to the same chain all
3304 * jumps are already validated for that chain.
3306 list_for_each_entry(i, &set->bindings, list) {
3307 if (i->flags & NFT_SET_MAP &&
3308 i->chain == binding->chain)
3312 iter.genmask = nft_genmask_next(ctx->net);
3316 iter.fn = nf_tables_bind_check_setelem;
3318 set->ops->walk(ctx, set, &iter);
3323 binding->chain = ctx->chain;
3324 list_add_tail_rcu(&binding->list, &set->bindings);
3327 EXPORT_SYMBOL_GPL(nf_tables_bind_set);
3329 void nf_tables_unbind_set(const struct nft_ctx *ctx, struct nft_set *set,
3330 struct nft_set_binding *binding)
3332 list_del_rcu(&binding->list);
3334 if (list_empty(&set->bindings) && nft_set_is_anonymous(set) &&
3335 nft_is_active(ctx->net, set))
3336 nf_tables_set_destroy(ctx, set);
3338 EXPORT_SYMBOL_GPL(nf_tables_unbind_set);
3340 const struct nft_set_ext_type nft_set_ext_types[] = {
3341 [NFT_SET_EXT_KEY] = {
3342 .align = __alignof__(u32),
3344 [NFT_SET_EXT_DATA] = {
3345 .align = __alignof__(u32),
3347 [NFT_SET_EXT_EXPR] = {
3348 .align = __alignof__(struct nft_expr),
3350 [NFT_SET_EXT_OBJREF] = {
3351 .len = sizeof(struct nft_object *),
3352 .align = __alignof__(struct nft_object *),
3354 [NFT_SET_EXT_FLAGS] = {
3356 .align = __alignof__(u8),
3358 [NFT_SET_EXT_TIMEOUT] = {
3360 .align = __alignof__(u64),
3362 [NFT_SET_EXT_EXPIRATION] = {
3363 .len = sizeof(unsigned long),
3364 .align = __alignof__(unsigned long),
3366 [NFT_SET_EXT_USERDATA] = {
3367 .len = sizeof(struct nft_userdata),
3368 .align = __alignof__(struct nft_userdata),
3371 EXPORT_SYMBOL_GPL(nft_set_ext_types);
3377 static const struct nla_policy nft_set_elem_policy[NFTA_SET_ELEM_MAX + 1] = {
3378 [NFTA_SET_ELEM_KEY] = { .type = NLA_NESTED },
3379 [NFTA_SET_ELEM_DATA] = { .type = NLA_NESTED },
3380 [NFTA_SET_ELEM_FLAGS] = { .type = NLA_U32 },
3381 [NFTA_SET_ELEM_TIMEOUT] = { .type = NLA_U64 },
3382 [NFTA_SET_ELEM_USERDATA] = { .type = NLA_BINARY,
3383 .len = NFT_USERDATA_MAXLEN },
3384 [NFTA_SET_ELEM_EXPR] = { .type = NLA_NESTED },
3385 [NFTA_SET_ELEM_OBJREF] = { .type = NLA_STRING },
3388 static const struct nla_policy nft_set_elem_list_policy[NFTA_SET_ELEM_LIST_MAX + 1] = {
3389 [NFTA_SET_ELEM_LIST_TABLE] = { .type = NLA_STRING,
3390 .len = NFT_TABLE_MAXNAMELEN - 1 },
3391 [NFTA_SET_ELEM_LIST_SET] = { .type = NLA_STRING,
3392 .len = NFT_SET_MAXNAMELEN - 1 },
3393 [NFTA_SET_ELEM_LIST_ELEMENTS] = { .type = NLA_NESTED },
3394 [NFTA_SET_ELEM_LIST_SET_ID] = { .type = NLA_U32 },
3397 static int nft_ctx_init_from_elemattr(struct nft_ctx *ctx, struct net *net,
3398 const struct sk_buff *skb,
3399 const struct nlmsghdr *nlh,
3400 const struct nlattr * const nla[],
3403 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
3404 int family = nfmsg->nfgen_family;
3405 struct nft_table *table;
3407 table = nf_tables_table_lookup(net, nla[NFTA_SET_ELEM_LIST_TABLE],
3410 return PTR_ERR(table);
3412 nft_ctx_init(ctx, net, skb, nlh, family, table, NULL, nla);
3416 static int nf_tables_fill_setelem(struct sk_buff *skb,
3417 const struct nft_set *set,
3418 const struct nft_set_elem *elem)
3420 const struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv);
3421 unsigned char *b = skb_tail_pointer(skb);
3422 struct nlattr *nest;
3424 nest = nla_nest_start(skb, NFTA_LIST_ELEM);
3426 goto nla_put_failure;
3428 if (nft_data_dump(skb, NFTA_SET_ELEM_KEY, nft_set_ext_key(ext),
3429 NFT_DATA_VALUE, set->klen) < 0)
3430 goto nla_put_failure;
3432 if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA) &&
3433 nft_data_dump(skb, NFTA_SET_ELEM_DATA, nft_set_ext_data(ext),
3434 set->dtype == NFT_DATA_VERDICT ? NFT_DATA_VERDICT : NFT_DATA_VALUE,
3436 goto nla_put_failure;
3438 if (nft_set_ext_exists(ext, NFT_SET_EXT_EXPR) &&
3439 nft_expr_dump(skb, NFTA_SET_ELEM_EXPR, nft_set_ext_expr(ext)) < 0)
3440 goto nla_put_failure;
3442 if (nft_set_ext_exists(ext, NFT_SET_EXT_OBJREF) &&
3443 nla_put_string(skb, NFTA_SET_ELEM_OBJREF,
3444 (*nft_set_ext_obj(ext))->name) < 0)
3445 goto nla_put_failure;
3447 if (nft_set_ext_exists(ext, NFT_SET_EXT_FLAGS) &&
3448 nla_put_be32(skb, NFTA_SET_ELEM_FLAGS,
3449 htonl(*nft_set_ext_flags(ext))))
3450 goto nla_put_failure;
3452 if (nft_set_ext_exists(ext, NFT_SET_EXT_TIMEOUT) &&
3453 nla_put_be64(skb, NFTA_SET_ELEM_TIMEOUT,
3454 cpu_to_be64(jiffies_to_msecs(
3455 *nft_set_ext_timeout(ext))),
3457 goto nla_put_failure;
3459 if (nft_set_ext_exists(ext, NFT_SET_EXT_EXPIRATION)) {
3460 unsigned long expires, now = jiffies;
3462 expires = *nft_set_ext_expiration(ext);
3463 if (time_before(now, expires))
3468 if (nla_put_be64(skb, NFTA_SET_ELEM_EXPIRATION,
3469 cpu_to_be64(jiffies_to_msecs(expires)),
3471 goto nla_put_failure;
3474 if (nft_set_ext_exists(ext, NFT_SET_EXT_USERDATA)) {
3475 struct nft_userdata *udata;
3477 udata = nft_set_ext_userdata(ext);
3478 if (nla_put(skb, NFTA_SET_ELEM_USERDATA,
3479 udata->len + 1, udata->data))
3480 goto nla_put_failure;
3483 nla_nest_end(skb, nest);
3491 struct nft_set_dump_args {
3492 const struct netlink_callback *cb;
3493 struct nft_set_iter iter;
3494 struct sk_buff *skb;
3497 static int nf_tables_dump_setelem(const struct nft_ctx *ctx,
3498 struct nft_set *set,
3499 const struct nft_set_iter *iter,
3500 struct nft_set_elem *elem)
3502 struct nft_set_dump_args *args;
3504 args = container_of(iter, struct nft_set_dump_args, iter);
3505 return nf_tables_fill_setelem(args->skb, set, elem);
3508 struct nft_set_dump_ctx {
3509 const struct nft_set *set;
3513 static int nf_tables_dump_set(struct sk_buff *skb, struct netlink_callback *cb)
3515 struct nft_set_dump_ctx *dump_ctx = cb->data;
3516 struct net *net = sock_net(skb->sk);
3517 struct nft_table *table;
3518 struct nft_set *set;
3519 struct nft_set_dump_args args;
3520 bool set_found = false;
3521 struct nfgenmsg *nfmsg;
3522 struct nlmsghdr *nlh;
3523 struct nlattr *nest;
3528 list_for_each_entry_rcu(table, &net->nft.tables, list) {
3529 if (dump_ctx->ctx.family != NFPROTO_UNSPEC &&
3530 dump_ctx->ctx.family != table->family)
3533 if (table != dump_ctx->ctx.table)
3536 list_for_each_entry_rcu(set, &table->sets, list) {
3537 if (set == dump_ctx->set) {
3550 event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, NFT_MSG_NEWSETELEM);
3551 portid = NETLINK_CB(cb->skb).portid;
3552 seq = cb->nlh->nlmsg_seq;
3554 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct nfgenmsg),
3557 goto nla_put_failure;
3559 nfmsg = nlmsg_data(nlh);
3560 nfmsg->nfgen_family = table->family;
3561 nfmsg->version = NFNETLINK_V0;
3562 nfmsg->res_id = htons(net->nft.base_seq & 0xffff);
3564 if (nla_put_string(skb, NFTA_SET_ELEM_LIST_TABLE, table->name))
3565 goto nla_put_failure;
3566 if (nla_put_string(skb, NFTA_SET_ELEM_LIST_SET, set->name))
3567 goto nla_put_failure;
3569 nest = nla_nest_start(skb, NFTA_SET_ELEM_LIST_ELEMENTS);
3571 goto nla_put_failure;
3575 args.iter.genmask = nft_genmask_cur(net);
3576 args.iter.skip = cb->args[0];
3577 args.iter.count = 0;
3579 args.iter.fn = nf_tables_dump_setelem;
3580 set->ops->walk(&dump_ctx->ctx, set, &args.iter);
3583 nla_nest_end(skb, nest);
3584 nlmsg_end(skb, nlh);
3586 if (args.iter.err && args.iter.err != -EMSGSIZE)
3587 return args.iter.err;
3588 if (args.iter.count == cb->args[0])
3591 cb->args[0] = args.iter.count;
3599 static int nf_tables_dump_set_done(struct netlink_callback *cb)
3605 static int nf_tables_fill_setelem_info(struct sk_buff *skb,
3606 const struct nft_ctx *ctx, u32 seq,
3607 u32 portid, int event, u16 flags,
3608 const struct nft_set *set,
3609 const struct nft_set_elem *elem)
3611 struct nfgenmsg *nfmsg;
3612 struct nlmsghdr *nlh;
3613 struct nlattr *nest;
3616 event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, event);
3617 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct nfgenmsg),
3620 goto nla_put_failure;
3622 nfmsg = nlmsg_data(nlh);
3623 nfmsg->nfgen_family = ctx->family;
3624 nfmsg->version = NFNETLINK_V0;
3625 nfmsg->res_id = htons(ctx->net->nft.base_seq & 0xffff);
3627 if (nla_put_string(skb, NFTA_SET_TABLE, ctx->table->name))
3628 goto nla_put_failure;
3629 if (nla_put_string(skb, NFTA_SET_NAME, set->name))
3630 goto nla_put_failure;
3632 nest = nla_nest_start(skb, NFTA_SET_ELEM_LIST_ELEMENTS);
3634 goto nla_put_failure;
3636 err = nf_tables_fill_setelem(skb, set, elem);
3638 goto nla_put_failure;
3640 nla_nest_end(skb, nest);
3642 nlmsg_end(skb, nlh);
3646 nlmsg_trim(skb, nlh);
3650 static int nft_setelem_parse_flags(const struct nft_set *set,
3651 const struct nlattr *attr, u32 *flags)
3656 *flags = ntohl(nla_get_be32(attr));
3657 if (*flags & ~NFT_SET_ELEM_INTERVAL_END)
3659 if (!(set->flags & NFT_SET_INTERVAL) &&
3660 *flags & NFT_SET_ELEM_INTERVAL_END)
3666 static int nft_get_set_elem(struct nft_ctx *ctx, struct nft_set *set,
3667 const struct nlattr *attr)
3669 struct nlattr *nla[NFTA_SET_ELEM_MAX + 1];
3670 const struct nft_set_ext *ext;
3671 struct nft_data_desc desc;
3672 struct nft_set_elem elem;
3673 struct sk_buff *skb;
3678 err = nla_parse_nested(nla, NFTA_SET_ELEM_MAX, attr,
3679 nft_set_elem_policy, NULL);
3683 if (!nla[NFTA_SET_ELEM_KEY])
3686 err = nft_setelem_parse_flags(set, nla[NFTA_SET_ELEM_FLAGS], &flags);
3690 err = nft_data_init(ctx, &elem.key.val, sizeof(elem.key), &desc,
3691 nla[NFTA_SET_ELEM_KEY]);
3696 if (desc.type != NFT_DATA_VALUE || desc.len != set->klen)
3699 priv = set->ops->get(ctx->net, set, &elem, flags);
3701 return PTR_ERR(priv);
3704 ext = nft_set_elem_ext(set, &elem);
3707 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
3711 err = nf_tables_fill_setelem_info(skb, ctx, ctx->seq, ctx->portid,
3712 NFT_MSG_NEWSETELEM, 0, set, &elem);
3716 err = nfnetlink_unicast(skb, ctx->net, ctx->portid, MSG_DONTWAIT);
3717 /* This avoids a loop in nfnetlink. */
3725 /* this avoids a loop in nfnetlink. */
3726 return err == -EAGAIN ? -ENOBUFS : err;
3729 static int nf_tables_getsetelem(struct net *net, struct sock *nlsk,
3730 struct sk_buff *skb, const struct nlmsghdr *nlh,
3731 const struct nlattr * const nla[],
3732 struct netlink_ext_ack *extack)
3734 u8 genmask = nft_genmask_cur(net);
3735 struct nft_set *set;
3736 struct nlattr *attr;
3740 err = nft_ctx_init_from_elemattr(&ctx, net, skb, nlh, nla, genmask);
3744 set = nf_tables_set_lookup(ctx.table, nla[NFTA_SET_ELEM_LIST_SET],
3747 return PTR_ERR(set);
3749 if (nlh->nlmsg_flags & NLM_F_DUMP) {
3750 struct netlink_dump_control c = {
3751 .dump = nf_tables_dump_set,
3752 .done = nf_tables_dump_set_done,
3754 struct nft_set_dump_ctx *dump_ctx;
3756 dump_ctx = kmalloc(sizeof(*dump_ctx), GFP_KERNEL);
3760 dump_ctx->set = set;
3761 dump_ctx->ctx = ctx;
3764 return netlink_dump_start(nlsk, skb, nlh, &c);
3767 if (!nla[NFTA_SET_ELEM_LIST_ELEMENTS])
3770 nla_for_each_nested(attr, nla[NFTA_SET_ELEM_LIST_ELEMENTS], rem) {
3771 err = nft_get_set_elem(&ctx, set, attr);
3779 static void nf_tables_setelem_notify(const struct nft_ctx *ctx,
3780 const struct nft_set *set,
3781 const struct nft_set_elem *elem,
3782 int event, u16 flags)
3784 struct net *net = ctx->net;
3785 u32 portid = ctx->portid;
3786 struct sk_buff *skb;
3789 if (!ctx->report && !nfnetlink_has_listeners(net, NFNLGRP_NFTABLES))
3792 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
3796 err = nf_tables_fill_setelem_info(skb, ctx, 0, portid, event, flags,
3803 nfnetlink_send(skb, net, portid, NFNLGRP_NFTABLES, ctx->report,
3807 nfnetlink_set_err(net, portid, NFNLGRP_NFTABLES, -ENOBUFS);
3810 static struct nft_trans *nft_trans_elem_alloc(struct nft_ctx *ctx,
3812 struct nft_set *set)
3814 struct nft_trans *trans;
3816 trans = nft_trans_alloc(ctx, msg_type, sizeof(struct nft_trans_elem));
3820 nft_trans_elem_set(trans) = set;
3824 void *nft_set_elem_init(const struct nft_set *set,
3825 const struct nft_set_ext_tmpl *tmpl,
3826 const u32 *key, const u32 *data,
3827 u64 timeout, gfp_t gfp)
3829 struct nft_set_ext *ext;
3832 elem = kzalloc(set->ops->elemsize + tmpl->len, gfp);
3836 ext = nft_set_elem_ext(set, elem);
3837 nft_set_ext_init(ext, tmpl);
3839 memcpy(nft_set_ext_key(ext), key, set->klen);
3840 if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA))
3841 memcpy(nft_set_ext_data(ext), data, set->dlen);
3842 if (nft_set_ext_exists(ext, NFT_SET_EXT_EXPIRATION))
3843 *nft_set_ext_expiration(ext) =
3845 if (nft_set_ext_exists(ext, NFT_SET_EXT_TIMEOUT))
3846 *nft_set_ext_timeout(ext) = timeout;
3851 void nft_set_elem_destroy(const struct nft_set *set, void *elem,
3854 struct nft_set_ext *ext = nft_set_elem_ext(set, elem);
3856 nft_data_release(nft_set_ext_key(ext), NFT_DATA_VALUE);
3857 if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA))
3858 nft_data_release(nft_set_ext_data(ext), set->dtype);
3859 if (destroy_expr && nft_set_ext_exists(ext, NFT_SET_EXT_EXPR))
3860 nf_tables_expr_destroy(NULL, nft_set_ext_expr(ext));
3861 if (nft_set_ext_exists(ext, NFT_SET_EXT_OBJREF))
3862 (*nft_set_ext_obj(ext))->use--;
3865 EXPORT_SYMBOL_GPL(nft_set_elem_destroy);
3867 /* Only called from commit path, nft_set_elem_deactivate() already deals with
3868 * the refcounting from the preparation phase.
3870 static void nf_tables_set_elem_destroy(const struct nft_set *set, void *elem)
3872 struct nft_set_ext *ext = nft_set_elem_ext(set, elem);
3874 if (nft_set_ext_exists(ext, NFT_SET_EXT_EXPR))
3875 nf_tables_expr_destroy(NULL, nft_set_ext_expr(ext));
3879 static int nft_add_set_elem(struct nft_ctx *ctx, struct nft_set *set,
3880 const struct nlattr *attr, u32 nlmsg_flags)
3882 struct nlattr *nla[NFTA_SET_ELEM_MAX + 1];
3883 u8 genmask = nft_genmask_next(ctx->net);
3884 struct nft_data_desc d1, d2;
3885 struct nft_set_ext_tmpl tmpl;
3886 struct nft_set_ext *ext, *ext2;
3887 struct nft_set_elem elem;
3888 struct nft_set_binding *binding;
3889 struct nft_object *obj = NULL;
3890 struct nft_userdata *udata;
3891 struct nft_data data;
3892 enum nft_registers dreg;
3893 struct nft_trans *trans;
3899 err = nla_parse_nested(nla, NFTA_SET_ELEM_MAX, attr,
3900 nft_set_elem_policy, NULL);
3904 if (nla[NFTA_SET_ELEM_KEY] == NULL)
3907 nft_set_ext_prepare(&tmpl);
3909 err = nft_setelem_parse_flags(set, nla[NFTA_SET_ELEM_FLAGS], &flags);
3913 nft_set_ext_add(&tmpl, NFT_SET_EXT_FLAGS);
3915 if (set->flags & NFT_SET_MAP) {
3916 if (nla[NFTA_SET_ELEM_DATA] == NULL &&
3917 !(flags & NFT_SET_ELEM_INTERVAL_END))
3919 if (nla[NFTA_SET_ELEM_DATA] != NULL &&
3920 flags & NFT_SET_ELEM_INTERVAL_END)
3923 if (nla[NFTA_SET_ELEM_DATA] != NULL)
3928 if (nla[NFTA_SET_ELEM_TIMEOUT] != NULL) {
3929 if (!(set->flags & NFT_SET_TIMEOUT))
3931 timeout = msecs_to_jiffies(be64_to_cpu(nla_get_be64(
3932 nla[NFTA_SET_ELEM_TIMEOUT])));
3933 } else if (set->flags & NFT_SET_TIMEOUT) {
3934 timeout = set->timeout;
3937 err = nft_data_init(ctx, &elem.key.val, sizeof(elem.key), &d1,
3938 nla[NFTA_SET_ELEM_KEY]);
3942 if (d1.type != NFT_DATA_VALUE || d1.len != set->klen)
3945 nft_set_ext_add_length(&tmpl, NFT_SET_EXT_KEY, d1.len);
3947 nft_set_ext_add(&tmpl, NFT_SET_EXT_EXPIRATION);
3948 if (timeout != set->timeout)
3949 nft_set_ext_add(&tmpl, NFT_SET_EXT_TIMEOUT);
3952 if (nla[NFTA_SET_ELEM_OBJREF] != NULL) {
3953 if (!(set->flags & NFT_SET_OBJECT)) {
3957 obj = nf_tables_obj_lookup(ctx->table, nla[NFTA_SET_ELEM_OBJREF],
3958 set->objtype, genmask);
3963 nft_set_ext_add(&tmpl, NFT_SET_EXT_OBJREF);
3966 if (nla[NFTA_SET_ELEM_DATA] != NULL) {
3967 err = nft_data_init(ctx, &data, sizeof(data), &d2,
3968 nla[NFTA_SET_ELEM_DATA]);
3973 if (set->dtype != NFT_DATA_VERDICT && d2.len != set->dlen)
3976 dreg = nft_type_to_reg(set->dtype);
3977 list_for_each_entry(binding, &set->bindings, list) {
3978 struct nft_ctx bind_ctx = {
3980 .family = ctx->family,
3981 .table = ctx->table,
3982 .chain = (struct nft_chain *)binding->chain,
3985 if (!(binding->flags & NFT_SET_MAP))
3988 err = nft_validate_register_store(&bind_ctx, dreg,
3995 nft_set_ext_add_length(&tmpl, NFT_SET_EXT_DATA, d2.len);
3998 /* The full maximum length of userdata can exceed the maximum
3999 * offset value (U8_MAX) for following extensions, therefor it
4000 * must be the last extension added.
4003 if (nla[NFTA_SET_ELEM_USERDATA] != NULL) {
4004 ulen = nla_len(nla[NFTA_SET_ELEM_USERDATA]);
4006 nft_set_ext_add_length(&tmpl, NFT_SET_EXT_USERDATA,
4011 elem.priv = nft_set_elem_init(set, &tmpl, elem.key.val.data, data.data,
4012 timeout, GFP_KERNEL);
4013 if (elem.priv == NULL)
4016 ext = nft_set_elem_ext(set, elem.priv);
4018 *nft_set_ext_flags(ext) = flags;
4020 udata = nft_set_ext_userdata(ext);
4021 udata->len = ulen - 1;
4022 nla_memcpy(&udata->data, nla[NFTA_SET_ELEM_USERDATA], ulen);
4025 *nft_set_ext_obj(ext) = obj;
4029 trans = nft_trans_elem_alloc(ctx, NFT_MSG_NEWSETELEM, set);
4033 ext->genmask = nft_genmask_cur(ctx->net) | NFT_SET_ELEM_BUSY_MASK;
4034 err = set->ops->insert(ctx->net, set, &elem, &ext2);
4036 if (err == -EEXIST) {
4037 if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA) ^
4038 nft_set_ext_exists(ext2, NFT_SET_EXT_DATA) ||
4039 nft_set_ext_exists(ext, NFT_SET_EXT_OBJREF) ^
4040 nft_set_ext_exists(ext2, NFT_SET_EXT_OBJREF))
4042 if ((nft_set_ext_exists(ext, NFT_SET_EXT_DATA) &&
4043 nft_set_ext_exists(ext2, NFT_SET_EXT_DATA) &&
4044 memcmp(nft_set_ext_data(ext),
4045 nft_set_ext_data(ext2), set->dlen) != 0) ||
4046 (nft_set_ext_exists(ext, NFT_SET_EXT_OBJREF) &&
4047 nft_set_ext_exists(ext2, NFT_SET_EXT_OBJREF) &&
4048 *nft_set_ext_obj(ext) != *nft_set_ext_obj(ext2)))
4050 else if (!(nlmsg_flags & NLM_F_EXCL))
4057 !atomic_add_unless(&set->nelems, 1, set->size + set->ndeact)) {
4062 nft_trans_elem(trans) = elem;
4063 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
4067 set->ops->remove(ctx->net, set, &elem);
4073 if (nla[NFTA_SET_ELEM_DATA] != NULL)
4074 nft_data_release(&data, d2.type);
4076 nft_data_release(&elem.key.val, d1.type);
4081 static int nf_tables_newsetelem(struct net *net, struct sock *nlsk,
4082 struct sk_buff *skb, const struct nlmsghdr *nlh,
4083 const struct nlattr * const nla[],
4084 struct netlink_ext_ack *extack)
4086 u8 genmask = nft_genmask_next(net);
4087 const struct nlattr *attr;
4088 struct nft_set *set;
4092 if (nla[NFTA_SET_ELEM_LIST_ELEMENTS] == NULL)
4095 err = nft_ctx_init_from_elemattr(&ctx, net, skb, nlh, nla, genmask);
4099 set = nft_set_lookup_global(net, ctx.table, nla[NFTA_SET_ELEM_LIST_SET],
4100 nla[NFTA_SET_ELEM_LIST_SET_ID], genmask);
4102 return PTR_ERR(set);
4104 if (!list_empty(&set->bindings) && set->flags & NFT_SET_CONSTANT)
4107 nla_for_each_nested(attr, nla[NFTA_SET_ELEM_LIST_ELEMENTS], rem) {
4108 err = nft_add_set_elem(&ctx, set, attr, nlh->nlmsg_flags);
4116 * nft_data_hold - hold a nft_data item
4118 * @data: struct nft_data to release
4119 * @type: type of data
4121 * Hold a nft_data item. NFT_DATA_VALUE types can be silently discarded,
4122 * NFT_DATA_VERDICT bumps the reference to chains in case of NFT_JUMP and
4123 * NFT_GOTO verdicts. This function must be called on active data objects
4124 * from the second phase of the commit protocol.
4126 static void nft_data_hold(const struct nft_data *data, enum nft_data_types type)
4128 if (type == NFT_DATA_VERDICT) {
4129 switch (data->verdict.code) {
4132 data->verdict.chain->use++;
4138 static void nft_set_elem_activate(const struct net *net,
4139 const struct nft_set *set,
4140 struct nft_set_elem *elem)
4142 const struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv);
4144 if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA))
4145 nft_data_hold(nft_set_ext_data(ext), set->dtype);
4146 if (nft_set_ext_exists(ext, NFT_SET_EXT_OBJREF))
4147 (*nft_set_ext_obj(ext))->use++;
4150 static void nft_set_elem_deactivate(const struct net *net,
4151 const struct nft_set *set,
4152 struct nft_set_elem *elem)
4154 const struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv);
4156 if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA))
4157 nft_data_release(nft_set_ext_data(ext), set->dtype);
4158 if (nft_set_ext_exists(ext, NFT_SET_EXT_OBJREF))
4159 (*nft_set_ext_obj(ext))->use--;
4162 static int nft_del_setelem(struct nft_ctx *ctx, struct nft_set *set,
4163 const struct nlattr *attr)
4165 struct nlattr *nla[NFTA_SET_ELEM_MAX + 1];
4166 struct nft_set_ext_tmpl tmpl;
4167 struct nft_data_desc desc;
4168 struct nft_set_elem elem;
4169 struct nft_set_ext *ext;
4170 struct nft_trans *trans;
4175 err = nla_parse_nested(nla, NFTA_SET_ELEM_MAX, attr,
4176 nft_set_elem_policy, NULL);
4181 if (nla[NFTA_SET_ELEM_KEY] == NULL)
4184 nft_set_ext_prepare(&tmpl);
4186 err = nft_setelem_parse_flags(set, nla[NFTA_SET_ELEM_FLAGS], &flags);
4190 nft_set_ext_add(&tmpl, NFT_SET_EXT_FLAGS);
4192 err = nft_data_init(ctx, &elem.key.val, sizeof(elem.key), &desc,
4193 nla[NFTA_SET_ELEM_KEY]);
4198 if (desc.type != NFT_DATA_VALUE || desc.len != set->klen)
4201 nft_set_ext_add_length(&tmpl, NFT_SET_EXT_KEY, desc.len);
4204 elem.priv = nft_set_elem_init(set, &tmpl, elem.key.val.data, NULL, 0,
4206 if (elem.priv == NULL)
4209 ext = nft_set_elem_ext(set, elem.priv);
4211 *nft_set_ext_flags(ext) = flags;
4213 trans = nft_trans_elem_alloc(ctx, NFT_MSG_DELSETELEM, set);
4214 if (trans == NULL) {
4219 priv = set->ops->deactivate(ctx->net, set, &elem);
4227 nft_set_elem_deactivate(ctx->net, set, &elem);
4229 nft_trans_elem(trans) = elem;
4230 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
4238 nft_data_release(&elem.key.val, desc.type);
4243 static int nft_flush_set(const struct nft_ctx *ctx,
4244 struct nft_set *set,
4245 const struct nft_set_iter *iter,
4246 struct nft_set_elem *elem)
4248 struct nft_trans *trans;
4251 trans = nft_trans_alloc_gfp(ctx, NFT_MSG_DELSETELEM,
4252 sizeof(struct nft_trans_elem), GFP_ATOMIC);
4256 if (!set->ops->flush(ctx->net, set, elem->priv)) {
4262 nft_trans_elem_set(trans) = set;
4263 nft_trans_elem(trans) = *elem;
4264 list_add_tail(&trans->list, &ctx->net->nft.commit_list);
4272 static int nf_tables_delsetelem(struct net *net, struct sock *nlsk,
4273 struct sk_buff *skb, const struct nlmsghdr *nlh,
4274 const struct nlattr * const nla[],
4275 struct netlink_ext_ack *extack)
4277 u8 genmask = nft_genmask_next(net);
4278 const struct nlattr *attr;
4279 struct nft_set *set;
4283 err = nft_ctx_init_from_elemattr(&ctx, net, skb, nlh, nla, genmask);
4287 set = nf_tables_set_lookup(ctx.table, nla[NFTA_SET_ELEM_LIST_SET],
4290 return PTR_ERR(set);
4291 if (!list_empty(&set->bindings) && set->flags & NFT_SET_CONSTANT)
4294 if (nla[NFTA_SET_ELEM_LIST_ELEMENTS] == NULL) {
4295 struct nft_set_iter iter = {
4297 .fn = nft_flush_set,
4299 set->ops->walk(&ctx, set, &iter);
4304 nla_for_each_nested(attr, nla[NFTA_SET_ELEM_LIST_ELEMENTS], rem) {
4305 err = nft_del_setelem(&ctx, set, attr);
4314 void nft_set_gc_batch_release(struct rcu_head *rcu)
4316 struct nft_set_gc_batch *gcb;
4319 gcb = container_of(rcu, struct nft_set_gc_batch, head.rcu);
4320 for (i = 0; i < gcb->head.cnt; i++)
4321 nft_set_elem_destroy(gcb->head.set, gcb->elems[i], true);
4324 EXPORT_SYMBOL_GPL(nft_set_gc_batch_release);
4326 struct nft_set_gc_batch *nft_set_gc_batch_alloc(const struct nft_set *set,
4329 struct nft_set_gc_batch *gcb;
4331 gcb = kzalloc(sizeof(*gcb), gfp);
4334 gcb->head.set = set;
4337 EXPORT_SYMBOL_GPL(nft_set_gc_batch_alloc);
4344 * nft_register_obj- register nf_tables stateful object type
4347 * Registers the object type for use with nf_tables. Returns zero on
4348 * success or a negative errno code otherwise.
4350 int nft_register_obj(struct nft_object_type *obj_type)
4352 if (obj_type->type == NFT_OBJECT_UNSPEC)
4355 nfnl_lock(NFNL_SUBSYS_NFTABLES);
4356 list_add_rcu(&obj_type->list, &nf_tables_objects);
4357 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
4360 EXPORT_SYMBOL_GPL(nft_register_obj);
4363 * nft_unregister_obj - unregister nf_tables object type
4366 * Unregisters the object type for use with nf_tables.
4368 void nft_unregister_obj(struct nft_object_type *obj_type)
4370 nfnl_lock(NFNL_SUBSYS_NFTABLES);
4371 list_del_rcu(&obj_type->list);
4372 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
4374 EXPORT_SYMBOL_GPL(nft_unregister_obj);
4376 struct nft_object *nf_tables_obj_lookup(const struct nft_table *table,
4377 const struct nlattr *nla,
4378 u32 objtype, u8 genmask)
4380 struct nft_object *obj;
4382 list_for_each_entry(obj, &table->objects, list) {
4383 if (!nla_strcmp(nla, obj->name) &&
4384 objtype == obj->ops->type->type &&
4385 nft_active_genmask(obj, genmask))
4388 return ERR_PTR(-ENOENT);
4390 EXPORT_SYMBOL_GPL(nf_tables_obj_lookup);
4392 static struct nft_object *nf_tables_obj_lookup_byhandle(const struct nft_table *table,
4393 const struct nlattr *nla,
4394 u32 objtype, u8 genmask)
4396 struct nft_object *obj;
4398 list_for_each_entry(obj, &table->objects, list) {
4399 if (be64_to_cpu(nla_get_be64(nla)) == obj->handle &&
4400 objtype == obj->ops->type->type &&
4401 nft_active_genmask(obj, genmask))
4404 return ERR_PTR(-ENOENT);
4407 static const struct nla_policy nft_obj_policy[NFTA_OBJ_MAX + 1] = {
4408 [NFTA_OBJ_TABLE] = { .type = NLA_STRING,
4409 .len = NFT_TABLE_MAXNAMELEN - 1 },
4410 [NFTA_OBJ_NAME] = { .type = NLA_STRING,
4411 .len = NFT_OBJ_MAXNAMELEN - 1 },
4412 [NFTA_OBJ_TYPE] = { .type = NLA_U32 },
4413 [NFTA_OBJ_DATA] = { .type = NLA_NESTED },
4414 [NFTA_OBJ_HANDLE] = { .type = NLA_U64},
4417 static struct nft_object *nft_obj_init(const struct nft_ctx *ctx,
4418 const struct nft_object_type *type,
4419 const struct nlattr *attr)
4422 const struct nft_object_ops *ops;
4423 struct nft_object *obj;
4426 tb = kmalloc_array(type->maxattr + 1, sizeof(*tb), GFP_KERNEL);
4431 err = nla_parse_nested(tb, type->maxattr, attr, type->policy,
4436 memset(tb, 0, sizeof(tb[0]) * (type->maxattr + 1));
4439 if (type->select_ops) {
4440 ops = type->select_ops(ctx, (const struct nlattr * const *)tb);
4450 obj = kzalloc(sizeof(*obj) + ops->size, GFP_KERNEL);
4454 err = ops->init(ctx, (const struct nlattr * const *)tb, obj);
4467 return ERR_PTR(err);
4470 static int nft_object_dump(struct sk_buff *skb, unsigned int attr,
4471 struct nft_object *obj, bool reset)
4473 struct nlattr *nest;
4475 nest = nla_nest_start(skb, attr);
4477 goto nla_put_failure;
4478 if (obj->ops->dump(skb, obj, reset) < 0)
4479 goto nla_put_failure;
4480 nla_nest_end(skb, nest);
4487 static const struct nft_object_type *__nft_obj_type_get(u32 objtype)
4489 const struct nft_object_type *type;
4491 list_for_each_entry(type, &nf_tables_objects, list) {
4492 if (objtype == type->type)
4498 static const struct nft_object_type *nft_obj_type_get(u32 objtype)
4500 const struct nft_object_type *type;
4502 type = __nft_obj_type_get(objtype);
4503 if (type != NULL && try_module_get(type->owner))
4506 #ifdef CONFIG_MODULES
4508 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
4509 request_module("nft-obj-%u", objtype);
4510 nfnl_lock(NFNL_SUBSYS_NFTABLES);
4511 if (__nft_obj_type_get(objtype))
4512 return ERR_PTR(-EAGAIN);
4515 return ERR_PTR(-ENOENT);
4518 static int nf_tables_newobj(struct net *net, struct sock *nlsk,
4519 struct sk_buff *skb, const struct nlmsghdr *nlh,
4520 const struct nlattr * const nla[],
4521 struct netlink_ext_ack *extack)
4523 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
4524 const struct nft_object_type *type;
4525 u8 genmask = nft_genmask_next(net);
4526 int family = nfmsg->nfgen_family;
4527 struct nft_table *table;
4528 struct nft_object *obj;
4533 if (!nla[NFTA_OBJ_TYPE] ||
4534 !nla[NFTA_OBJ_NAME] ||
4535 !nla[NFTA_OBJ_DATA])
4538 table = nf_tables_table_lookup(net, nla[NFTA_OBJ_TABLE], family,
4541 return PTR_ERR(table);
4543 objtype = ntohl(nla_get_be32(nla[NFTA_OBJ_TYPE]));
4544 obj = nf_tables_obj_lookup(table, nla[NFTA_OBJ_NAME], objtype, genmask);
4551 if (nlh->nlmsg_flags & NLM_F_EXCL)
4557 nft_ctx_init(&ctx, net, skb, nlh, family, table, NULL, nla);
4559 type = nft_obj_type_get(objtype);
4561 return PTR_ERR(type);
4563 obj = nft_obj_init(&ctx, type, nla[NFTA_OBJ_DATA]);
4569 obj->handle = nf_tables_alloc_handle(table);
4571 obj->name = nla_strdup(nla[NFTA_OBJ_NAME], GFP_KERNEL);
4577 err = nft_trans_obj_add(&ctx, NFT_MSG_NEWOBJ, obj);
4581 list_add_tail_rcu(&obj->list, &table->objects);
4587 if (obj->ops->destroy)
4588 obj->ops->destroy(obj);
4591 module_put(type->owner);
4595 static int nf_tables_fill_obj_info(struct sk_buff *skb, struct net *net,
4596 u32 portid, u32 seq, int event, u32 flags,
4597 int family, const struct nft_table *table,
4598 struct nft_object *obj, bool reset)
4600 struct nfgenmsg *nfmsg;
4601 struct nlmsghdr *nlh;
4603 event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, event);
4604 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct nfgenmsg), flags);
4606 goto nla_put_failure;
4608 nfmsg = nlmsg_data(nlh);
4609 nfmsg->nfgen_family = family;
4610 nfmsg->version = NFNETLINK_V0;
4611 nfmsg->res_id = htons(net->nft.base_seq & 0xffff);
4613 if (nla_put_string(skb, NFTA_OBJ_TABLE, table->name) ||
4614 nla_put_string(skb, NFTA_OBJ_NAME, obj->name) ||
4615 nla_put_be32(skb, NFTA_OBJ_TYPE, htonl(obj->ops->type->type)) ||
4616 nla_put_be32(skb, NFTA_OBJ_USE, htonl(obj->use)) ||
4617 nft_object_dump(skb, NFTA_OBJ_DATA, obj, reset) ||
4618 nla_put_be64(skb, NFTA_OBJ_HANDLE, cpu_to_be64(obj->handle),
4620 goto nla_put_failure;
4622 nlmsg_end(skb, nlh);
4626 nlmsg_trim(skb, nlh);
4630 struct nft_obj_filter {
4635 static int nf_tables_dump_obj(struct sk_buff *skb, struct netlink_callback *cb)
4637 const struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
4638 const struct nft_table *table;
4639 unsigned int idx = 0, s_idx = cb->args[0];
4640 struct nft_obj_filter *filter = cb->data;
4641 struct net *net = sock_net(skb->sk);
4642 int family = nfmsg->nfgen_family;
4643 struct nft_object *obj;
4646 if (NFNL_MSG_TYPE(cb->nlh->nlmsg_type) == NFT_MSG_GETOBJ_RESET)
4650 cb->seq = net->nft.base_seq;
4652 list_for_each_entry_rcu(table, &net->nft.tables, list) {
4653 if (family != NFPROTO_UNSPEC && family != table->family)
4656 list_for_each_entry_rcu(obj, &table->objects, list) {
4657 if (!nft_is_active(net, obj))
4662 memset(&cb->args[1], 0,
4663 sizeof(cb->args) - sizeof(cb->args[0]));
4664 if (filter && filter->table[0] &&
4665 strcmp(filter->table, table->name))
4668 filter->type != NFT_OBJECT_UNSPEC &&
4669 obj->ops->type->type != filter->type)
4672 if (nf_tables_fill_obj_info(skb, net, NETLINK_CB(cb->skb).portid,
4675 NLM_F_MULTI | NLM_F_APPEND,
4676 table->family, table,
4680 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4692 static int nf_tables_dump_obj_done(struct netlink_callback *cb)
4694 struct nft_obj_filter *filter = cb->data;
4697 kfree(filter->table);
4704 static struct nft_obj_filter *
4705 nft_obj_filter_alloc(const struct nlattr * const nla[])
4707 struct nft_obj_filter *filter;
4709 filter = kzalloc(sizeof(*filter), GFP_KERNEL);
4711 return ERR_PTR(-ENOMEM);
4713 if (nla[NFTA_OBJ_TABLE]) {
4714 filter->table = nla_strdup(nla[NFTA_OBJ_TABLE], GFP_KERNEL);
4715 if (!filter->table) {
4717 return ERR_PTR(-ENOMEM);
4720 if (nla[NFTA_OBJ_TYPE])
4721 filter->type = ntohl(nla_get_be32(nla[NFTA_OBJ_TYPE]));
4726 static int nf_tables_getobj(struct net *net, struct sock *nlsk,
4727 struct sk_buff *skb, const struct nlmsghdr *nlh,
4728 const struct nlattr * const nla[],
4729 struct netlink_ext_ack *extack)
4731 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
4732 u8 genmask = nft_genmask_cur(net);
4733 int family = nfmsg->nfgen_family;
4734 const struct nft_table *table;
4735 struct nft_object *obj;
4736 struct sk_buff *skb2;
4741 if (nlh->nlmsg_flags & NLM_F_DUMP) {
4742 struct netlink_dump_control c = {
4743 .dump = nf_tables_dump_obj,
4744 .done = nf_tables_dump_obj_done,
4747 if (nla[NFTA_OBJ_TABLE] ||
4748 nla[NFTA_OBJ_TYPE]) {
4749 struct nft_obj_filter *filter;
4751 filter = nft_obj_filter_alloc(nla);
4757 return netlink_dump_start(nlsk, skb, nlh, &c);
4760 if (!nla[NFTA_OBJ_NAME] ||
4761 !nla[NFTA_OBJ_TYPE])
4764 table = nf_tables_table_lookup(net, nla[NFTA_OBJ_TABLE], family,
4767 return PTR_ERR(table);
4769 objtype = ntohl(nla_get_be32(nla[NFTA_OBJ_TYPE]));
4770 obj = nf_tables_obj_lookup(table, nla[NFTA_OBJ_NAME], objtype, genmask);
4772 return PTR_ERR(obj);
4774 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
4778 if (NFNL_MSG_TYPE(nlh->nlmsg_type) == NFT_MSG_GETOBJ_RESET)
4781 err = nf_tables_fill_obj_info(skb2, net, NETLINK_CB(skb).portid,
4782 nlh->nlmsg_seq, NFT_MSG_NEWOBJ, 0,
4783 family, table, obj, reset);
4787 return nlmsg_unicast(nlsk, skb2, NETLINK_CB(skb).portid);
4793 static void nft_obj_destroy(struct nft_object *obj)
4795 if (obj->ops->destroy)
4796 obj->ops->destroy(obj);
4798 module_put(obj->ops->type->owner);
4803 static int nf_tables_delobj(struct net *net, struct sock *nlsk,
4804 struct sk_buff *skb, const struct nlmsghdr *nlh,
4805 const struct nlattr * const nla[],
4806 struct netlink_ext_ack *extack)
4808 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
4809 u8 genmask = nft_genmask_next(net);
4810 int family = nfmsg->nfgen_family;
4811 struct nft_table *table;
4812 struct nft_object *obj;
4816 if (!nla[NFTA_OBJ_TYPE] ||
4817 (!nla[NFTA_OBJ_NAME] && !nla[NFTA_OBJ_HANDLE]))
4820 table = nf_tables_table_lookup(net, nla[NFTA_OBJ_TABLE], family,
4823 return PTR_ERR(table);
4825 objtype = ntohl(nla_get_be32(nla[NFTA_OBJ_TYPE]));
4826 if (nla[NFTA_OBJ_HANDLE])
4827 obj = nf_tables_obj_lookup_byhandle(table, nla[NFTA_OBJ_HANDLE],
4830 obj = nf_tables_obj_lookup(table, nla[NFTA_OBJ_NAME],
4833 return PTR_ERR(obj);
4837 nft_ctx_init(&ctx, net, skb, nlh, family, table, NULL, nla);
4839 return nft_delobj(&ctx, obj);
4842 void nft_obj_notify(struct net *net, struct nft_table *table,
4843 struct nft_object *obj, u32 portid, u32 seq, int event,
4844 int family, int report, gfp_t gfp)
4846 struct sk_buff *skb;
4850 !nfnetlink_has_listeners(net, NFNLGRP_NFTABLES))
4853 skb = nlmsg_new(NLMSG_GOODSIZE, gfp);
4857 err = nf_tables_fill_obj_info(skb, net, portid, seq, event, 0, family,
4864 nfnetlink_send(skb, net, portid, NFNLGRP_NFTABLES, report, gfp);
4867 nfnetlink_set_err(net, portid, NFNLGRP_NFTABLES, -ENOBUFS);
4869 EXPORT_SYMBOL_GPL(nft_obj_notify);
4871 static void nf_tables_obj_notify(const struct nft_ctx *ctx,
4872 struct nft_object *obj, int event)
4874 nft_obj_notify(ctx->net, ctx->table, obj, ctx->portid, ctx->seq, event,
4875 ctx->family, ctx->report, GFP_KERNEL);
4881 void nft_register_flowtable_type(struct nf_flowtable_type *type)
4883 nfnl_lock(NFNL_SUBSYS_NFTABLES);
4884 list_add_tail_rcu(&type->list, &nf_tables_flowtables);
4885 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
4887 EXPORT_SYMBOL_GPL(nft_register_flowtable_type);
4889 void nft_unregister_flowtable_type(struct nf_flowtable_type *type)
4891 nfnl_lock(NFNL_SUBSYS_NFTABLES);
4892 list_del_rcu(&type->list);
4893 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
4895 EXPORT_SYMBOL_GPL(nft_unregister_flowtable_type);
4897 static const struct nla_policy nft_flowtable_policy[NFTA_FLOWTABLE_MAX + 1] = {
4898 [NFTA_FLOWTABLE_TABLE] = { .type = NLA_STRING,
4899 .len = NFT_NAME_MAXLEN - 1 },
4900 [NFTA_FLOWTABLE_NAME] = { .type = NLA_STRING,
4901 .len = NFT_NAME_MAXLEN - 1 },
4902 [NFTA_FLOWTABLE_HOOK] = { .type = NLA_NESTED },
4903 [NFTA_FLOWTABLE_HANDLE] = { .type = NLA_U64 },
4906 struct nft_flowtable *nf_tables_flowtable_lookup(const struct nft_table *table,
4907 const struct nlattr *nla,
4910 struct nft_flowtable *flowtable;
4912 list_for_each_entry(flowtable, &table->flowtables, list) {
4913 if (!nla_strcmp(nla, flowtable->name) &&
4914 nft_active_genmask(flowtable, genmask))
4917 return ERR_PTR(-ENOENT);
4919 EXPORT_SYMBOL_GPL(nf_tables_flowtable_lookup);
4921 static struct nft_flowtable *
4922 nf_tables_flowtable_lookup_byhandle(const struct nft_table *table,
4923 const struct nlattr *nla, u8 genmask)
4925 struct nft_flowtable *flowtable;
4927 list_for_each_entry(flowtable, &table->flowtables, list) {
4928 if (be64_to_cpu(nla_get_be64(nla)) == flowtable->handle &&
4929 nft_active_genmask(flowtable, genmask))
4932 return ERR_PTR(-ENOENT);
4935 static int nf_tables_parse_devices(const struct nft_ctx *ctx,
4936 const struct nlattr *attr,
4937 struct net_device *dev_array[], int *len)
4939 const struct nlattr *tmp;
4940 struct net_device *dev;
4941 char ifname[IFNAMSIZ];
4942 int rem, n = 0, err;
4944 nla_for_each_nested(tmp, attr, rem) {
4945 if (nla_type(tmp) != NFTA_DEVICE_NAME) {
4950 nla_strlcpy(ifname, tmp, IFNAMSIZ);
4951 dev = __dev_get_by_name(ctx->net, ifname);
4957 dev_array[n++] = dev;
4958 if (n == NFT_FLOWTABLE_DEVICE_MAX) {
4972 static const struct nla_policy nft_flowtable_hook_policy[NFTA_FLOWTABLE_HOOK_MAX + 1] = {
4973 [NFTA_FLOWTABLE_HOOK_NUM] = { .type = NLA_U32 },
4974 [NFTA_FLOWTABLE_HOOK_PRIORITY] = { .type = NLA_U32 },
4975 [NFTA_FLOWTABLE_HOOK_DEVS] = { .type = NLA_NESTED },
4978 static int nf_tables_flowtable_parse_hook(const struct nft_ctx *ctx,
4979 const struct nlattr *attr,
4980 struct nft_flowtable *flowtable)
4982 struct net_device *dev_array[NFT_FLOWTABLE_DEVICE_MAX];
4983 struct nlattr *tb[NFTA_FLOWTABLE_HOOK_MAX + 1];
4984 struct nf_hook_ops *ops;
4985 int hooknum, priority;
4988 err = nla_parse_nested(tb, NFTA_FLOWTABLE_HOOK_MAX, attr,
4989 nft_flowtable_hook_policy, NULL);
4993 if (!tb[NFTA_FLOWTABLE_HOOK_NUM] ||
4994 !tb[NFTA_FLOWTABLE_HOOK_PRIORITY] ||
4995 !tb[NFTA_FLOWTABLE_HOOK_DEVS])
4998 hooknum = ntohl(nla_get_be32(tb[NFTA_FLOWTABLE_HOOK_NUM]));
4999 if (hooknum != NF_NETDEV_INGRESS)
5002 priority = ntohl(nla_get_be32(tb[NFTA_FLOWTABLE_HOOK_PRIORITY]));
5004 err = nf_tables_parse_devices(ctx, tb[NFTA_FLOWTABLE_HOOK_DEVS],
5009 ops = kzalloc(sizeof(struct nf_hook_ops) * n, GFP_KERNEL);
5013 flowtable->hooknum = hooknum;
5014 flowtable->priority = priority;
5015 flowtable->ops = ops;
5016 flowtable->ops_len = n;
5018 for (i = 0; i < n; i++) {
5019 flowtable->ops[i].pf = NFPROTO_NETDEV;
5020 flowtable->ops[i].hooknum = hooknum;
5021 flowtable->ops[i].priority = priority;
5022 flowtable->ops[i].priv = &flowtable->data.rhashtable;
5023 flowtable->ops[i].hook = flowtable->data.type->hook;
5024 flowtable->ops[i].dev = dev_array[i];
5025 flowtable->dev_name[i] = kstrdup(dev_array[i]->name,
5032 static const struct nf_flowtable_type *__nft_flowtable_type_get(u8 family)
5034 const struct nf_flowtable_type *type;
5036 list_for_each_entry(type, &nf_tables_flowtables, list) {
5037 if (family == type->family)
5043 static const struct nf_flowtable_type *nft_flowtable_type_get(u8 family)
5045 const struct nf_flowtable_type *type;
5047 type = __nft_flowtable_type_get(family);
5048 if (type != NULL && try_module_get(type->owner))
5051 #ifdef CONFIG_MODULES
5053 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
5054 request_module("nf-flowtable-%u", family);
5055 nfnl_lock(NFNL_SUBSYS_NFTABLES);
5056 if (__nft_flowtable_type_get(family))
5057 return ERR_PTR(-EAGAIN);
5060 return ERR_PTR(-ENOENT);
5063 void nft_flow_table_iterate(struct net *net,
5064 void (*iter)(struct nf_flowtable *flowtable, void *data),
5067 struct nft_flowtable *flowtable;
5068 const struct nft_table *table;
5070 nfnl_lock(NFNL_SUBSYS_NFTABLES);
5071 list_for_each_entry(table, &net->nft.tables, list) {
5072 list_for_each_entry(flowtable, &table->flowtables, list) {
5073 iter(&flowtable->data, data);
5076 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
5078 EXPORT_SYMBOL_GPL(nft_flow_table_iterate);
5080 static void nft_unregister_flowtable_net_hooks(struct net *net,
5081 struct nft_flowtable *flowtable)
5085 for (i = 0; i < flowtable->ops_len; i++) {
5086 if (!flowtable->ops[i].dev)
5089 nf_unregister_net_hook(net, &flowtable->ops[i]);
5093 static int nf_tables_newflowtable(struct net *net, struct sock *nlsk,
5094 struct sk_buff *skb,
5095 const struct nlmsghdr *nlh,
5096 const struct nlattr * const nla[],
5097 struct netlink_ext_ack *extack)
5099 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
5100 const struct nf_flowtable_type *type;
5101 struct nft_flowtable *flowtable, *ft;
5102 u8 genmask = nft_genmask_next(net);
5103 int family = nfmsg->nfgen_family;
5104 struct nft_table *table;
5108 if (!nla[NFTA_FLOWTABLE_TABLE] ||
5109 !nla[NFTA_FLOWTABLE_NAME] ||
5110 !nla[NFTA_FLOWTABLE_HOOK])
5113 table = nf_tables_table_lookup(net, nla[NFTA_FLOWTABLE_TABLE],
5116 return PTR_ERR(table);
5118 flowtable = nf_tables_flowtable_lookup(table, nla[NFTA_FLOWTABLE_NAME],
5120 if (IS_ERR(flowtable)) {
5121 err = PTR_ERR(flowtable);
5125 if (nlh->nlmsg_flags & NLM_F_EXCL)
5131 nft_ctx_init(&ctx, net, skb, nlh, family, table, NULL, nla);
5133 flowtable = kzalloc(sizeof(*flowtable), GFP_KERNEL);
5137 flowtable->table = table;
5138 flowtable->handle = nf_tables_alloc_handle(table);
5140 flowtable->name = nla_strdup(nla[NFTA_FLOWTABLE_NAME], GFP_KERNEL);
5141 if (!flowtable->name) {
5146 type = nft_flowtable_type_get(family);
5148 err = PTR_ERR(type);
5152 flowtable->data.type = type;
5153 err = rhashtable_init(&flowtable->data.rhashtable, type->params);
5157 err = nf_tables_flowtable_parse_hook(&ctx, nla[NFTA_FLOWTABLE_HOOK],
5162 for (i = 0; i < flowtable->ops_len; i++) {
5163 if (!flowtable->ops[i].dev)
5166 list_for_each_entry(ft, &table->flowtables, list) {
5167 for (k = 0; k < ft->ops_len; k++) {
5168 if (!ft->ops[k].dev)
5171 if (flowtable->ops[i].dev == ft->ops[k].dev &&
5172 flowtable->ops[i].pf == ft->ops[k].pf) {
5179 err = nf_register_net_hook(net, &flowtable->ops[i]);
5184 err = nft_trans_flowtable_add(&ctx, NFT_MSG_NEWFLOWTABLE, flowtable);
5188 INIT_DEFERRABLE_WORK(&flowtable->data.gc_work, type->gc);
5189 queue_delayed_work(system_power_efficient_wq,
5190 &flowtable->data.gc_work, HZ);
5192 list_add_tail_rcu(&flowtable->list, &table->flowtables);
5197 i = flowtable->ops_len;
5199 for (k = i - 1; k >= 0; k--) {
5200 kfree(flowtable->dev_name[k]);
5201 nf_unregister_net_hook(net, &flowtable->ops[k]);
5204 kfree(flowtable->ops);
5206 module_put(type->owner);
5208 kfree(flowtable->name);
5214 static int nf_tables_delflowtable(struct net *net, struct sock *nlsk,
5215 struct sk_buff *skb,
5216 const struct nlmsghdr *nlh,
5217 const struct nlattr * const nla[],
5218 struct netlink_ext_ack *extack)
5220 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
5221 u8 genmask = nft_genmask_next(net);
5222 int family = nfmsg->nfgen_family;
5223 struct nft_flowtable *flowtable;
5224 struct nft_table *table;
5227 if (!nla[NFTA_FLOWTABLE_TABLE] ||
5228 (!nla[NFTA_FLOWTABLE_NAME] &&
5229 !nla[NFTA_FLOWTABLE_HANDLE]))
5232 table = nf_tables_table_lookup(net, nla[NFTA_FLOWTABLE_TABLE],
5235 return PTR_ERR(table);
5237 if (nla[NFTA_FLOWTABLE_HANDLE])
5238 flowtable = nf_tables_flowtable_lookup_byhandle(table,
5239 nla[NFTA_FLOWTABLE_HANDLE],
5242 flowtable = nf_tables_flowtable_lookup(table,
5243 nla[NFTA_FLOWTABLE_NAME],
5245 if (IS_ERR(flowtable))
5246 return PTR_ERR(flowtable);
5247 if (flowtable->use > 0)
5250 nft_ctx_init(&ctx, net, skb, nlh, family, table, NULL, nla);
5252 return nft_delflowtable(&ctx, flowtable);
5255 static int nf_tables_fill_flowtable_info(struct sk_buff *skb, struct net *net,
5256 u32 portid, u32 seq, int event,
5257 u32 flags, int family,
5258 struct nft_flowtable *flowtable)
5260 struct nlattr *nest, *nest_devs;
5261 struct nfgenmsg *nfmsg;
5262 struct nlmsghdr *nlh;
5265 event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, event);
5266 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct nfgenmsg), flags);
5268 goto nla_put_failure;
5270 nfmsg = nlmsg_data(nlh);
5271 nfmsg->nfgen_family = family;
5272 nfmsg->version = NFNETLINK_V0;
5273 nfmsg->res_id = htons(net->nft.base_seq & 0xffff);
5275 if (nla_put_string(skb, NFTA_FLOWTABLE_TABLE, flowtable->table->name) ||
5276 nla_put_string(skb, NFTA_FLOWTABLE_NAME, flowtable->name) ||
5277 nla_put_be32(skb, NFTA_FLOWTABLE_USE, htonl(flowtable->use)) ||
5278 nla_put_be64(skb, NFTA_FLOWTABLE_HANDLE, cpu_to_be64(flowtable->handle),
5279 NFTA_FLOWTABLE_PAD))
5280 goto nla_put_failure;
5282 nest = nla_nest_start(skb, NFTA_FLOWTABLE_HOOK);
5283 if (nla_put_be32(skb, NFTA_FLOWTABLE_HOOK_NUM, htonl(flowtable->hooknum)) ||
5284 nla_put_be32(skb, NFTA_FLOWTABLE_HOOK_PRIORITY, htonl(flowtable->priority)))
5285 goto nla_put_failure;
5287 nest_devs = nla_nest_start(skb, NFTA_FLOWTABLE_HOOK_DEVS);
5289 goto nla_put_failure;
5291 for (i = 0; i < flowtable->ops_len; i++) {
5292 if (flowtable->dev_name[i][0] &&
5293 nla_put_string(skb, NFTA_DEVICE_NAME,
5294 flowtable->dev_name[i]))
5295 goto nla_put_failure;
5297 nla_nest_end(skb, nest_devs);
5298 nla_nest_end(skb, nest);
5300 nlmsg_end(skb, nlh);
5304 nlmsg_trim(skb, nlh);
5308 struct nft_flowtable_filter {
5312 static int nf_tables_dump_flowtable(struct sk_buff *skb,
5313 struct netlink_callback *cb)
5315 const struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
5316 struct nft_flowtable_filter *filter = cb->data;
5317 unsigned int idx = 0, s_idx = cb->args[0];
5318 struct net *net = sock_net(skb->sk);
5319 int family = nfmsg->nfgen_family;
5320 struct nft_flowtable *flowtable;
5321 const struct nft_table *table;
5324 cb->seq = net->nft.base_seq;
5326 list_for_each_entry_rcu(table, &net->nft.tables, list) {
5327 if (family != NFPROTO_UNSPEC && family != table->family)
5330 list_for_each_entry_rcu(flowtable, &table->flowtables, list) {
5331 if (!nft_is_active(net, flowtable))
5336 memset(&cb->args[1], 0,
5337 sizeof(cb->args) - sizeof(cb->args[0]));
5338 if (filter && filter->table[0] &&
5339 strcmp(filter->table, table->name))
5342 if (nf_tables_fill_flowtable_info(skb, net, NETLINK_CB(cb->skb).portid,
5344 NFT_MSG_NEWFLOWTABLE,
5345 NLM_F_MULTI | NLM_F_APPEND,
5346 table->family, flowtable) < 0)
5349 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5361 static int nf_tables_dump_flowtable_done(struct netlink_callback *cb)
5363 struct nft_flowtable_filter *filter = cb->data;
5368 kfree(filter->table);
5374 static struct nft_flowtable_filter *
5375 nft_flowtable_filter_alloc(const struct nlattr * const nla[])
5377 struct nft_flowtable_filter *filter;
5379 filter = kzalloc(sizeof(*filter), GFP_KERNEL);
5381 return ERR_PTR(-ENOMEM);
5383 if (nla[NFTA_FLOWTABLE_TABLE]) {
5384 filter->table = nla_strdup(nla[NFTA_FLOWTABLE_TABLE],
5386 if (!filter->table) {
5388 return ERR_PTR(-ENOMEM);
5394 static int nf_tables_getflowtable(struct net *net, struct sock *nlsk,
5395 struct sk_buff *skb,
5396 const struct nlmsghdr *nlh,
5397 const struct nlattr * const nla[],
5398 struct netlink_ext_ack *extack)
5400 const struct nfgenmsg *nfmsg = nlmsg_data(nlh);
5401 u8 genmask = nft_genmask_cur(net);
5402 int family = nfmsg->nfgen_family;
5403 struct nft_flowtable *flowtable;
5404 const struct nft_table *table;
5405 struct sk_buff *skb2;
5408 if (nlh->nlmsg_flags & NLM_F_DUMP) {
5409 struct netlink_dump_control c = {
5410 .dump = nf_tables_dump_flowtable,
5411 .done = nf_tables_dump_flowtable_done,
5414 if (nla[NFTA_FLOWTABLE_TABLE]) {
5415 struct nft_flowtable_filter *filter;
5417 filter = nft_flowtable_filter_alloc(nla);
5423 return netlink_dump_start(nlsk, skb, nlh, &c);
5426 if (!nla[NFTA_FLOWTABLE_NAME])
5429 table = nf_tables_table_lookup(net, nla[NFTA_FLOWTABLE_TABLE],
5432 return PTR_ERR(table);
5434 flowtable = nf_tables_flowtable_lookup(table, nla[NFTA_FLOWTABLE_NAME],
5436 if (IS_ERR(flowtable))
5437 return PTR_ERR(flowtable);
5439 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
5443 err = nf_tables_fill_flowtable_info(skb2, net, NETLINK_CB(skb).portid,
5445 NFT_MSG_NEWFLOWTABLE, 0, family,
5450 return nlmsg_unicast(nlsk, skb2, NETLINK_CB(skb).portid);
5456 static void nf_tables_flowtable_notify(struct nft_ctx *ctx,
5457 struct nft_flowtable *flowtable,
5460 struct sk_buff *skb;
5464 !nfnetlink_has_listeners(ctx->net, NFNLGRP_NFTABLES))
5467 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
5471 err = nf_tables_fill_flowtable_info(skb, ctx->net, ctx->portid,
5473 ctx->family, flowtable);
5479 nfnetlink_send(skb, ctx->net, ctx->portid, NFNLGRP_NFTABLES,
5480 ctx->report, GFP_KERNEL);
5483 nfnetlink_set_err(ctx->net, ctx->portid, NFNLGRP_NFTABLES, -ENOBUFS);
5486 static void nf_tables_flowtable_destroy(struct nft_flowtable *flowtable)
5488 cancel_delayed_work_sync(&flowtable->data.gc_work);
5489 kfree(flowtable->ops);
5490 kfree(flowtable->name);
5491 flowtable->data.type->free(&flowtable->data);
5492 rhashtable_destroy(&flowtable->data.rhashtable);
5493 module_put(flowtable->data.type->owner);
5496 static int nf_tables_fill_gen_info(struct sk_buff *skb, struct net *net,
5497 u32 portid, u32 seq)
5499 struct nlmsghdr *nlh;
5500 struct nfgenmsg *nfmsg;
5501 char buf[TASK_COMM_LEN];
5502 int event = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, NFT_MSG_NEWGEN);
5504 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct nfgenmsg), 0);
5506 goto nla_put_failure;
5508 nfmsg = nlmsg_data(nlh);
5509 nfmsg->nfgen_family = AF_UNSPEC;
5510 nfmsg->version = NFNETLINK_V0;
5511 nfmsg->res_id = htons(net->nft.base_seq & 0xffff);
5513 if (nla_put_be32(skb, NFTA_GEN_ID, htonl(net->nft.base_seq)) ||
5514 nla_put_be32(skb, NFTA_GEN_PROC_PID, htonl(task_pid_nr(current))) ||
5515 nla_put_string(skb, NFTA_GEN_PROC_NAME, get_task_comm(buf, current)))
5516 goto nla_put_failure;
5518 nlmsg_end(skb, nlh);
5522 nlmsg_trim(skb, nlh);
5526 static void nft_flowtable_event(unsigned long event, struct net_device *dev,
5527 struct nft_flowtable *flowtable)
5531 for (i = 0; i < flowtable->ops_len; i++) {
5532 if (flowtable->ops[i].dev != dev)
5535 nf_unregister_net_hook(dev_net(dev), &flowtable->ops[i]);
5536 flowtable->dev_name[i][0] = '\0';
5537 flowtable->ops[i].dev = NULL;
5542 static int nf_tables_flowtable_event(struct notifier_block *this,
5543 unsigned long event, void *ptr)
5545 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5546 struct nft_flowtable *flowtable;
5547 struct nft_table *table;
5549 if (event != NETDEV_UNREGISTER)
5552 nfnl_lock(NFNL_SUBSYS_NFTABLES);
5553 list_for_each_entry(table, &dev_net(dev)->nft.tables, list) {
5554 list_for_each_entry(flowtable, &table->flowtables, list) {
5555 nft_flowtable_event(event, dev, flowtable);
5558 nfnl_unlock(NFNL_SUBSYS_NFTABLES);
5563 static struct notifier_block nf_tables_flowtable_notifier = {
5564 .notifier_call = nf_tables_flowtable_event,
5567 static void nf_tables_gen_notify(struct net *net, struct sk_buff *skb,
5570 struct nlmsghdr *nlh = nlmsg_hdr(skb);
5571 struct sk_buff *skb2;
5574 if (nlmsg_report(nlh) &&
5575 !nfnetlink_has_listeners(net, NFNLGRP_NFTABLES))
5578 skb2 = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
5582 err = nf_tables_fill_gen_info(skb2, net, NETLINK_CB(skb).portid,
5589 nfnetlink_send(skb2, net, NETLINK_CB(skb).portid, NFNLGRP_NFTABLES,
5590 nlmsg_report(nlh), GFP_KERNEL);
5593 nfnetlink_set_err(net, NETLINK_CB(skb).portid, NFNLGRP_NFTABLES,
5597 static int nf_tables_getgen(struct net *net, struct sock *nlsk,
5598 struct sk_buff *skb, const struct nlmsghdr *nlh,
5599 const struct nlattr * const nla[],
5600 struct netlink_ext_ack *extack)
5602 struct sk_buff *skb2;
5605 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
5609 err = nf_tables_fill_gen_info(skb2, net, NETLINK_CB(skb).portid,
5614 return nlmsg_unicast(nlsk, skb2, NETLINK_CB(skb).portid);
5620 static const struct nfnl_callback nf_tables_cb[NFT_MSG_MAX] = {
5621 [NFT_MSG_NEWTABLE] = {
5622 .call_batch = nf_tables_newtable,
5623 .attr_count = NFTA_TABLE_MAX,
5624 .policy = nft_table_policy,
5626 [NFT_MSG_GETTABLE] = {
5627 .call = nf_tables_gettable,
5628 .attr_count = NFTA_TABLE_MAX,
5629 .policy = nft_table_policy,
5631 [NFT_MSG_DELTABLE] = {
5632 .call_batch = nf_tables_deltable,
5633 .attr_count = NFTA_TABLE_MAX,
5634 .policy = nft_table_policy,
5636 [NFT_MSG_NEWCHAIN] = {
5637 .call_batch = nf_tables_newchain,
5638 .attr_count = NFTA_CHAIN_MAX,
5639 .policy = nft_chain_policy,
5641 [NFT_MSG_GETCHAIN] = {
5642 .call = nf_tables_getchain,
5643 .attr_count = NFTA_CHAIN_MAX,
5644 .policy = nft_chain_policy,
5646 [NFT_MSG_DELCHAIN] = {
5647 .call_batch = nf_tables_delchain,
5648 .attr_count = NFTA_CHAIN_MAX,
5649 .policy = nft_chain_policy,
5651 [NFT_MSG_NEWRULE] = {
5652 .call_batch = nf_tables_newrule,
5653 .attr_count = NFTA_RULE_MAX,
5654 .policy = nft_rule_policy,
5656 [NFT_MSG_GETRULE] = {
5657 .call = nf_tables_getrule,
5658 .attr_count = NFTA_RULE_MAX,
5659 .policy = nft_rule_policy,
5661 [NFT_MSG_DELRULE] = {
5662 .call_batch = nf_tables_delrule,
5663 .attr_count = NFTA_RULE_MAX,
5664 .policy = nft_rule_policy,
5666 [NFT_MSG_NEWSET] = {
5667 .call_batch = nf_tables_newset,
5668 .attr_count = NFTA_SET_MAX,
5669 .policy = nft_set_policy,
5671 [NFT_MSG_GETSET] = {
5672 .call = nf_tables_getset,
5673 .attr_count = NFTA_SET_MAX,
5674 .policy = nft_set_policy,
5676 [NFT_MSG_DELSET] = {
5677 .call_batch = nf_tables_delset,
5678 .attr_count = NFTA_SET_MAX,
5679 .policy = nft_set_policy,
5681 [NFT_MSG_NEWSETELEM] = {
5682 .call_batch = nf_tables_newsetelem,
5683 .attr_count = NFTA_SET_ELEM_LIST_MAX,
5684 .policy = nft_set_elem_list_policy,
5686 [NFT_MSG_GETSETELEM] = {
5687 .call = nf_tables_getsetelem,
5688 .attr_count = NFTA_SET_ELEM_LIST_MAX,
5689 .policy = nft_set_elem_list_policy,
5691 [NFT_MSG_DELSETELEM] = {
5692 .call_batch = nf_tables_delsetelem,
5693 .attr_count = NFTA_SET_ELEM_LIST_MAX,
5694 .policy = nft_set_elem_list_policy,
5696 [NFT_MSG_GETGEN] = {
5697 .call = nf_tables_getgen,
5699 [NFT_MSG_NEWOBJ] = {
5700 .call_batch = nf_tables_newobj,
5701 .attr_count = NFTA_OBJ_MAX,
5702 .policy = nft_obj_policy,
5704 [NFT_MSG_GETOBJ] = {
5705 .call = nf_tables_getobj,
5706 .attr_count = NFTA_OBJ_MAX,
5707 .policy = nft_obj_policy,
5709 [NFT_MSG_DELOBJ] = {
5710 .call_batch = nf_tables_delobj,
5711 .attr_count = NFTA_OBJ_MAX,
5712 .policy = nft_obj_policy,
5714 [NFT_MSG_GETOBJ_RESET] = {
5715 .call = nf_tables_getobj,
5716 .attr_count = NFTA_OBJ_MAX,
5717 .policy = nft_obj_policy,
5719 [NFT_MSG_NEWFLOWTABLE] = {
5720 .call_batch = nf_tables_newflowtable,
5721 .attr_count = NFTA_FLOWTABLE_MAX,
5722 .policy = nft_flowtable_policy,
5724 [NFT_MSG_GETFLOWTABLE] = {
5725 .call = nf_tables_getflowtable,
5726 .attr_count = NFTA_FLOWTABLE_MAX,
5727 .policy = nft_flowtable_policy,
5729 [NFT_MSG_DELFLOWTABLE] = {
5730 .call_batch = nf_tables_delflowtable,
5731 .attr_count = NFTA_FLOWTABLE_MAX,
5732 .policy = nft_flowtable_policy,
5736 static void nft_chain_commit_update(struct nft_trans *trans)
5738 struct nft_base_chain *basechain;
5740 if (nft_trans_chain_name(trans))
5741 strcpy(trans->ctx.chain->name, nft_trans_chain_name(trans));
5743 if (!nft_is_base_chain(trans->ctx.chain))
5746 basechain = nft_base_chain(trans->ctx.chain);
5747 nft_chain_stats_replace(basechain, nft_trans_chain_stats(trans));
5749 switch (nft_trans_chain_policy(trans)) {
5752 basechain->policy = nft_trans_chain_policy(trans);
5757 static void nf_tables_commit_release(struct nft_trans *trans)
5759 switch (trans->msg_type) {
5760 case NFT_MSG_DELTABLE:
5761 nf_tables_table_destroy(&trans->ctx);
5763 case NFT_MSG_DELCHAIN:
5764 nf_tables_chain_destroy(&trans->ctx);
5766 case NFT_MSG_DELRULE:
5767 nf_tables_rule_destroy(&trans->ctx, nft_trans_rule(trans));
5769 case NFT_MSG_DELSET:
5770 nft_set_destroy(nft_trans_set(trans));
5772 case NFT_MSG_DELSETELEM:
5773 nf_tables_set_elem_destroy(nft_trans_elem_set(trans),
5774 nft_trans_elem(trans).priv);
5776 case NFT_MSG_DELOBJ:
5777 nft_obj_destroy(nft_trans_obj(trans));
5779 case NFT_MSG_DELFLOWTABLE:
5780 nf_tables_flowtable_destroy(nft_trans_flowtable(trans));
5786 static int nf_tables_commit(struct net *net, struct sk_buff *skb)
5788 struct nft_trans *trans, *next;
5789 struct nft_trans_elem *te;
5791 /* Bump generation counter, invalidate any dump in progress */
5792 while (++net->nft.base_seq == 0);
5794 /* A new generation has just started */
5795 net->nft.gencursor = nft_gencursor_next(net);
5797 /* Make sure all packets have left the previous generation before
5798 * purging old rules.
5802 list_for_each_entry_safe(trans, next, &net->nft.commit_list, list) {
5803 switch (trans->msg_type) {
5804 case NFT_MSG_NEWTABLE:
5805 if (nft_trans_table_update(trans)) {
5806 if (!nft_trans_table_enable(trans)) {
5807 nf_tables_table_disable(net,
5809 trans->ctx.table->flags |= NFT_TABLE_F_DORMANT;
5812 nft_clear(net, trans->ctx.table);
5814 nf_tables_table_notify(&trans->ctx, NFT_MSG_NEWTABLE);
5815 nft_trans_destroy(trans);
5817 case NFT_MSG_DELTABLE:
5818 list_del_rcu(&trans->ctx.table->list);
5819 nf_tables_table_notify(&trans->ctx, NFT_MSG_DELTABLE);
5821 case NFT_MSG_NEWCHAIN:
5822 if (nft_trans_chain_update(trans))
5823 nft_chain_commit_update(trans);
5825 nft_clear(net, trans->ctx.chain);
5827 nf_tables_chain_notify(&trans->ctx, NFT_MSG_NEWCHAIN);
5828 nft_trans_destroy(trans);
5830 case NFT_MSG_DELCHAIN:
5831 list_del_rcu(&trans->ctx.chain->list);
5832 nf_tables_chain_notify(&trans->ctx, NFT_MSG_DELCHAIN);
5833 nf_tables_unregister_hook(trans->ctx.net,
5837 case NFT_MSG_NEWRULE:
5838 nft_clear(trans->ctx.net, nft_trans_rule(trans));
5839 nf_tables_rule_notify(&trans->ctx,
5840 nft_trans_rule(trans),
5842 nft_trans_destroy(trans);
5844 case NFT_MSG_DELRULE:
5845 list_del_rcu(&nft_trans_rule(trans)->list);
5846 nf_tables_rule_notify(&trans->ctx,
5847 nft_trans_rule(trans),
5850 case NFT_MSG_NEWSET:
5851 nft_clear(net, nft_trans_set(trans));
5852 /* This avoids hitting -EBUSY when deleting the table
5853 * from the transaction.
5855 if (nft_set_is_anonymous(nft_trans_set(trans)) &&
5856 !list_empty(&nft_trans_set(trans)->bindings))
5857 trans->ctx.table->use--;
5859 nf_tables_set_notify(&trans->ctx, nft_trans_set(trans),
5860 NFT_MSG_NEWSET, GFP_KERNEL);
5861 nft_trans_destroy(trans);
5863 case NFT_MSG_DELSET:
5864 list_del_rcu(&nft_trans_set(trans)->list);
5865 nf_tables_set_notify(&trans->ctx, nft_trans_set(trans),
5866 NFT_MSG_DELSET, GFP_KERNEL);
5868 case NFT_MSG_NEWSETELEM:
5869 te = (struct nft_trans_elem *)trans->data;
5871 te->set->ops->activate(net, te->set, &te->elem);
5872 nf_tables_setelem_notify(&trans->ctx, te->set,
5874 NFT_MSG_NEWSETELEM, 0);
5875 nft_trans_destroy(trans);
5877 case NFT_MSG_DELSETELEM:
5878 te = (struct nft_trans_elem *)trans->data;
5880 nf_tables_setelem_notify(&trans->ctx, te->set,
5882 NFT_MSG_DELSETELEM, 0);
5883 te->set->ops->remove(net, te->set, &te->elem);
5884 atomic_dec(&te->set->nelems);
5887 case NFT_MSG_NEWOBJ:
5888 nft_clear(net, nft_trans_obj(trans));
5889 nf_tables_obj_notify(&trans->ctx, nft_trans_obj(trans),
5891 nft_trans_destroy(trans);
5893 case NFT_MSG_DELOBJ:
5894 list_del_rcu(&nft_trans_obj(trans)->list);
5895 nf_tables_obj_notify(&trans->ctx, nft_trans_obj(trans),
5898 case NFT_MSG_NEWFLOWTABLE:
5899 nft_clear(net, nft_trans_flowtable(trans));
5900 nf_tables_flowtable_notify(&trans->ctx,
5901 nft_trans_flowtable(trans),
5902 NFT_MSG_NEWFLOWTABLE);
5903 nft_trans_destroy(trans);
5905 case NFT_MSG_DELFLOWTABLE:
5906 list_del_rcu(&nft_trans_flowtable(trans)->list);
5907 nf_tables_flowtable_notify(&trans->ctx,
5908 nft_trans_flowtable(trans),
5909 NFT_MSG_DELFLOWTABLE);
5910 nft_unregister_flowtable_net_hooks(net,
5911 nft_trans_flowtable(trans));
5918 list_for_each_entry_safe(trans, next, &net->nft.commit_list, list) {
5919 list_del(&trans->list);
5920 nf_tables_commit_release(trans);
5923 nf_tables_gen_notify(net, skb, NFT_MSG_NEWGEN);
5928 static void nf_tables_abort_release(struct nft_trans *trans)
5930 switch (trans->msg_type) {
5931 case NFT_MSG_NEWTABLE:
5932 nf_tables_table_destroy(&trans->ctx);
5934 case NFT_MSG_NEWCHAIN:
5935 nf_tables_chain_destroy(&trans->ctx);
5937 case NFT_MSG_NEWRULE:
5938 nf_tables_rule_destroy(&trans->ctx, nft_trans_rule(trans));
5940 case NFT_MSG_NEWSET:
5941 nft_set_destroy(nft_trans_set(trans));
5943 case NFT_MSG_NEWSETELEM:
5944 nft_set_elem_destroy(nft_trans_elem_set(trans),
5945 nft_trans_elem(trans).priv, true);
5947 case NFT_MSG_NEWOBJ:
5948 nft_obj_destroy(nft_trans_obj(trans));
5950 case NFT_MSG_NEWFLOWTABLE:
5951 nf_tables_flowtable_destroy(nft_trans_flowtable(trans));
5957 static int nf_tables_abort(struct net *net, struct sk_buff *skb)
5959 struct nft_trans *trans, *next;
5960 struct nft_trans_elem *te;
5962 list_for_each_entry_safe_reverse(trans, next, &net->nft.commit_list,
5964 switch (trans->msg_type) {
5965 case NFT_MSG_NEWTABLE:
5966 if (nft_trans_table_update(trans)) {
5967 if (nft_trans_table_enable(trans)) {
5968 nf_tables_table_disable(net,
5970 trans->ctx.table->flags |= NFT_TABLE_F_DORMANT;
5972 nft_trans_destroy(trans);
5974 list_del_rcu(&trans->ctx.table->list);
5977 case NFT_MSG_DELTABLE:
5978 nft_clear(trans->ctx.net, trans->ctx.table);
5979 nft_trans_destroy(trans);
5981 case NFT_MSG_NEWCHAIN:
5982 if (nft_trans_chain_update(trans)) {
5983 free_percpu(nft_trans_chain_stats(trans));
5985 nft_trans_destroy(trans);
5987 trans->ctx.table->use--;
5988 list_del_rcu(&trans->ctx.chain->list);
5989 nf_tables_unregister_hook(trans->ctx.net,
5994 case NFT_MSG_DELCHAIN:
5995 trans->ctx.table->use++;
5996 nft_clear(trans->ctx.net, trans->ctx.chain);
5997 nft_trans_destroy(trans);
5999 case NFT_MSG_NEWRULE:
6000 trans->ctx.chain->use--;
6001 list_del_rcu(&nft_trans_rule(trans)->list);
6003 case NFT_MSG_DELRULE:
6004 trans->ctx.chain->use++;
6005 nft_clear(trans->ctx.net, nft_trans_rule(trans));
6006 nft_trans_destroy(trans);
6008 case NFT_MSG_NEWSET:
6009 trans->ctx.table->use--;
6010 list_del_rcu(&nft_trans_set(trans)->list);
6012 case NFT_MSG_DELSET:
6013 trans->ctx.table->use++;
6014 nft_clear(trans->ctx.net, nft_trans_set(trans));
6015 nft_trans_destroy(trans);
6017 case NFT_MSG_NEWSETELEM:
6018 te = (struct nft_trans_elem *)trans->data;
6020 te->set->ops->remove(net, te->set, &te->elem);
6021 atomic_dec(&te->set->nelems);
6023 case NFT_MSG_DELSETELEM:
6024 te = (struct nft_trans_elem *)trans->data;
6026 nft_set_elem_activate(net, te->set, &te->elem);
6027 te->set->ops->activate(net, te->set, &te->elem);
6030 nft_trans_destroy(trans);
6032 case NFT_MSG_NEWOBJ:
6033 trans->ctx.table->use--;
6034 list_del_rcu(&nft_trans_obj(trans)->list);
6036 case NFT_MSG_DELOBJ:
6037 trans->ctx.table->use++;
6038 nft_clear(trans->ctx.net, nft_trans_obj(trans));
6039 nft_trans_destroy(trans);
6041 case NFT_MSG_NEWFLOWTABLE:
6042 trans->ctx.table->use--;
6043 list_del_rcu(&nft_trans_flowtable(trans)->list);
6044 nft_unregister_flowtable_net_hooks(net,
6045 nft_trans_flowtable(trans));
6047 case NFT_MSG_DELFLOWTABLE:
6048 trans->ctx.table->use++;
6049 nft_clear(trans->ctx.net, nft_trans_flowtable(trans));
6050 nft_trans_destroy(trans);
6057 list_for_each_entry_safe_reverse(trans, next,
6058 &net->nft.commit_list, list) {
6059 list_del(&trans->list);
6060 nf_tables_abort_release(trans);
6066 static bool nf_tables_valid_genid(struct net *net, u32 genid)
6068 return net->nft.base_seq == genid;
6071 static const struct nfnetlink_subsystem nf_tables_subsys = {
6072 .name = "nf_tables",
6073 .subsys_id = NFNL_SUBSYS_NFTABLES,
6074 .cb_count = NFT_MSG_MAX,
6076 .commit = nf_tables_commit,
6077 .abort = nf_tables_abort,
6078 .valid_genid = nf_tables_valid_genid,
6081 int nft_chain_validate_dependency(const struct nft_chain *chain,
6082 enum nft_chain_types type)
6084 const struct nft_base_chain *basechain;
6086 if (nft_is_base_chain(chain)) {
6087 basechain = nft_base_chain(chain);
6088 if (basechain->type->type != type)
6093 EXPORT_SYMBOL_GPL(nft_chain_validate_dependency);
6095 int nft_chain_validate_hooks(const struct nft_chain *chain,
6096 unsigned int hook_flags)
6098 struct nft_base_chain *basechain;
6100 if (nft_is_base_chain(chain)) {
6101 basechain = nft_base_chain(chain);
6103 if ((1 << basechain->ops.hooknum) & hook_flags)
6111 EXPORT_SYMBOL_GPL(nft_chain_validate_hooks);
6114 * Loop detection - walk through the ruleset beginning at the destination chain
6115 * of a new jump until either the source chain is reached (loop) or all
6116 * reachable chains have been traversed.
6118 * The loop check is performed whenever a new jump verdict is added to an
6119 * expression or verdict map or a verdict map is bound to a new chain.
6122 static int nf_tables_check_loops(const struct nft_ctx *ctx,
6123 const struct nft_chain *chain);
6125 static int nf_tables_loop_check_setelem(const struct nft_ctx *ctx,
6126 struct nft_set *set,
6127 const struct nft_set_iter *iter,
6128 struct nft_set_elem *elem)
6130 const struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv);
6131 const struct nft_data *data;
6133 if (nft_set_ext_exists(ext, NFT_SET_EXT_FLAGS) &&
6134 *nft_set_ext_flags(ext) & NFT_SET_ELEM_INTERVAL_END)
6137 data = nft_set_ext_data(ext);
6138 switch (data->verdict.code) {
6141 return nf_tables_check_loops(ctx, data->verdict.chain);
6147 static int nf_tables_check_loops(const struct nft_ctx *ctx,
6148 const struct nft_chain *chain)
6150 const struct nft_rule *rule;
6151 const struct nft_expr *expr, *last;
6152 struct nft_set *set;
6153 struct nft_set_binding *binding;
6154 struct nft_set_iter iter;
6156 if (ctx->chain == chain)
6159 list_for_each_entry(rule, &chain->rules, list) {
6160 nft_rule_for_each_expr(expr, last, rule) {
6161 const struct nft_data *data = NULL;
6164 if (!expr->ops->validate)
6167 err = expr->ops->validate(ctx, expr, &data);
6174 switch (data->verdict.code) {
6177 err = nf_tables_check_loops(ctx,
6178 data->verdict.chain);
6187 list_for_each_entry(set, &ctx->table->sets, list) {
6188 if (!nft_is_active_next(ctx->net, set))
6190 if (!(set->flags & NFT_SET_MAP) ||
6191 set->dtype != NFT_DATA_VERDICT)
6194 list_for_each_entry(binding, &set->bindings, list) {
6195 if (!(binding->flags & NFT_SET_MAP) ||
6196 binding->chain != chain)
6199 iter.genmask = nft_genmask_next(ctx->net);
6203 iter.fn = nf_tables_loop_check_setelem;
6205 set->ops->walk(ctx, set, &iter);
6215 * nft_parse_u32_check - fetch u32 attribute and check for maximum value
6217 * @attr: netlink attribute to fetch value from
6218 * @max: maximum value to be stored in dest
6219 * @dest: pointer to the variable
6221 * Parse, check and store a given u32 netlink attribute into variable.
6222 * This function returns -ERANGE if the value goes over maximum value.
6223 * Otherwise a 0 is returned and the attribute value is stored in the
6224 * destination variable.
6226 int nft_parse_u32_check(const struct nlattr *attr, int max, u32 *dest)
6230 val = ntohl(nla_get_be32(attr));
6237 EXPORT_SYMBOL_GPL(nft_parse_u32_check);
6240 * nft_parse_register - parse a register value from a netlink attribute
6242 * @attr: netlink attribute
6244 * Parse and translate a register value from a netlink attribute.
6245 * Registers used to be 128 bit wide, these register numbers will be
6246 * mapped to the corresponding 32 bit register numbers.
6248 unsigned int nft_parse_register(const struct nlattr *attr)
6252 reg = ntohl(nla_get_be32(attr));
6254 case NFT_REG_VERDICT...NFT_REG_4:
6255 return reg * NFT_REG_SIZE / NFT_REG32_SIZE;
6257 return reg + NFT_REG_SIZE / NFT_REG32_SIZE - NFT_REG32_00;
6260 EXPORT_SYMBOL_GPL(nft_parse_register);
6263 * nft_dump_register - dump a register value to a netlink attribute
6265 * @skb: socket buffer
6266 * @attr: attribute number
6267 * @reg: register number
6269 * Construct a netlink attribute containing the register number. For
6270 * compatibility reasons, register numbers being a multiple of 4 are
6271 * translated to the corresponding 128 bit register numbers.
6273 int nft_dump_register(struct sk_buff *skb, unsigned int attr, unsigned int reg)
6275 if (reg % (NFT_REG_SIZE / NFT_REG32_SIZE) == 0)
6276 reg = reg / (NFT_REG_SIZE / NFT_REG32_SIZE);
6278 reg = reg - NFT_REG_SIZE / NFT_REG32_SIZE + NFT_REG32_00;
6280 return nla_put_be32(skb, attr, htonl(reg));
6282 EXPORT_SYMBOL_GPL(nft_dump_register);
6285 * nft_validate_register_load - validate a load from a register
6287 * @reg: the register number
6288 * @len: the length of the data
6290 * Validate that the input register is one of the general purpose
6291 * registers and that the length of the load is within the bounds.
6293 int nft_validate_register_load(enum nft_registers reg, unsigned int len)
6295 if (reg < NFT_REG_1 * NFT_REG_SIZE / NFT_REG32_SIZE)
6299 if (reg * NFT_REG32_SIZE + len > FIELD_SIZEOF(struct nft_regs, data))
6304 EXPORT_SYMBOL_GPL(nft_validate_register_load);
6307 * nft_validate_register_store - validate an expressions' register store
6309 * @ctx: context of the expression performing the load
6310 * @reg: the destination register number
6311 * @data: the data to load
6312 * @type: the data type
6313 * @len: the length of the data
6315 * Validate that a data load uses the appropriate data type for
6316 * the destination register and the length is within the bounds.
6317 * A value of NULL for the data means that its runtime gathered
6320 int nft_validate_register_store(const struct nft_ctx *ctx,
6321 enum nft_registers reg,
6322 const struct nft_data *data,
6323 enum nft_data_types type, unsigned int len)
6328 case NFT_REG_VERDICT:
6329 if (type != NFT_DATA_VERDICT)
6333 (data->verdict.code == NFT_GOTO ||
6334 data->verdict.code == NFT_JUMP)) {
6335 err = nf_tables_check_loops(ctx, data->verdict.chain);
6339 if (ctx->chain->level + 1 >
6340 data->verdict.chain->level) {
6341 if (ctx->chain->level + 1 == NFT_JUMP_STACK_SIZE)
6343 data->verdict.chain->level = ctx->chain->level + 1;
6349 if (reg < NFT_REG_1 * NFT_REG_SIZE / NFT_REG32_SIZE)
6353 if (reg * NFT_REG32_SIZE + len >
6354 FIELD_SIZEOF(struct nft_regs, data))
6357 if (data != NULL && type != NFT_DATA_VALUE)
6362 EXPORT_SYMBOL_GPL(nft_validate_register_store);
6364 static const struct nla_policy nft_verdict_policy[NFTA_VERDICT_MAX + 1] = {
6365 [NFTA_VERDICT_CODE] = { .type = NLA_U32 },
6366 [NFTA_VERDICT_CHAIN] = { .type = NLA_STRING,
6367 .len = NFT_CHAIN_MAXNAMELEN - 1 },
6370 static int nft_verdict_init(const struct nft_ctx *ctx, struct nft_data *data,
6371 struct nft_data_desc *desc, const struct nlattr *nla)
6373 u8 genmask = nft_genmask_next(ctx->net);
6374 struct nlattr *tb[NFTA_VERDICT_MAX + 1];
6375 struct nft_chain *chain;
6378 err = nla_parse_nested(tb, NFTA_VERDICT_MAX, nla, nft_verdict_policy,
6383 if (!tb[NFTA_VERDICT_CODE])
6385 data->verdict.code = ntohl(nla_get_be32(tb[NFTA_VERDICT_CODE]));
6387 switch (data->verdict.code) {
6389 switch (data->verdict.code & NF_VERDICT_MASK) {
6404 if (!tb[NFTA_VERDICT_CHAIN])
6406 chain = nf_tables_chain_lookup(ctx->table,
6407 tb[NFTA_VERDICT_CHAIN], genmask);
6409 return PTR_ERR(chain);
6410 if (nft_is_base_chain(chain))
6414 data->verdict.chain = chain;
6418 desc->len = sizeof(data->verdict);
6419 desc->type = NFT_DATA_VERDICT;
6423 static void nft_verdict_uninit(const struct nft_data *data)
6425 switch (data->verdict.code) {
6428 data->verdict.chain->use--;
6433 int nft_verdict_dump(struct sk_buff *skb, int type, const struct nft_verdict *v)
6435 struct nlattr *nest;
6437 nest = nla_nest_start(skb, type);
6439 goto nla_put_failure;
6441 if (nla_put_be32(skb, NFTA_VERDICT_CODE, htonl(v->code)))
6442 goto nla_put_failure;
6447 if (nla_put_string(skb, NFTA_VERDICT_CHAIN,
6449 goto nla_put_failure;
6451 nla_nest_end(skb, nest);
6458 static int nft_value_init(const struct nft_ctx *ctx,
6459 struct nft_data *data, unsigned int size,
6460 struct nft_data_desc *desc, const struct nlattr *nla)
6470 nla_memcpy(data->data, nla, len);
6471 desc->type = NFT_DATA_VALUE;
6476 static int nft_value_dump(struct sk_buff *skb, const struct nft_data *data,
6479 return nla_put(skb, NFTA_DATA_VALUE, len, data->data);
6482 static const struct nla_policy nft_data_policy[NFTA_DATA_MAX + 1] = {
6483 [NFTA_DATA_VALUE] = { .type = NLA_BINARY },
6484 [NFTA_DATA_VERDICT] = { .type = NLA_NESTED },
6488 * nft_data_init - parse nf_tables data netlink attributes
6490 * @ctx: context of the expression using the data
6491 * @data: destination struct nft_data
6492 * @size: maximum data length
6493 * @desc: data description
6494 * @nla: netlink attribute containing data
6496 * Parse the netlink data attributes and initialize a struct nft_data.
6497 * The type and length of data are returned in the data description.
6499 * The caller can indicate that it only wants to accept data of type
6500 * NFT_DATA_VALUE by passing NULL for the ctx argument.
6502 int nft_data_init(const struct nft_ctx *ctx,
6503 struct nft_data *data, unsigned int size,
6504 struct nft_data_desc *desc, const struct nlattr *nla)
6506 struct nlattr *tb[NFTA_DATA_MAX + 1];
6509 err = nla_parse_nested(tb, NFTA_DATA_MAX, nla, nft_data_policy, NULL);
6513 if (tb[NFTA_DATA_VALUE])
6514 return nft_value_init(ctx, data, size, desc,
6515 tb[NFTA_DATA_VALUE]);
6516 if (tb[NFTA_DATA_VERDICT] && ctx != NULL)
6517 return nft_verdict_init(ctx, data, desc, tb[NFTA_DATA_VERDICT]);
6520 EXPORT_SYMBOL_GPL(nft_data_init);
6523 * nft_data_release - release a nft_data item
6525 * @data: struct nft_data to release
6526 * @type: type of data
6528 * Release a nft_data item. NFT_DATA_VALUE types can be silently discarded,
6529 * all others need to be released by calling this function.
6531 void nft_data_release(const struct nft_data *data, enum nft_data_types type)
6533 if (type < NFT_DATA_VERDICT)
6536 case NFT_DATA_VERDICT:
6537 return nft_verdict_uninit(data);
6542 EXPORT_SYMBOL_GPL(nft_data_release);
6544 int nft_data_dump(struct sk_buff *skb, int attr, const struct nft_data *data,
6545 enum nft_data_types type, unsigned int len)
6547 struct nlattr *nest;
6550 nest = nla_nest_start(skb, attr);
6555 case NFT_DATA_VALUE:
6556 err = nft_value_dump(skb, data, len);
6558 case NFT_DATA_VERDICT:
6559 err = nft_verdict_dump(skb, NFTA_DATA_VERDICT, &data->verdict);
6566 nla_nest_end(skb, nest);
6569 EXPORT_SYMBOL_GPL(nft_data_dump);
6571 int __nft_release_basechain(struct nft_ctx *ctx)
6573 struct nft_rule *rule, *nr;
6575 BUG_ON(!nft_is_base_chain(ctx->chain));
6577 nf_tables_unregister_hook(ctx->net, ctx->chain->table, ctx->chain);
6578 list_for_each_entry_safe(rule, nr, &ctx->chain->rules, list) {
6579 list_del(&rule->list);
6581 nf_tables_rule_destroy(ctx, rule);
6583 list_del(&ctx->chain->list);
6585 nf_tables_chain_destroy(ctx);
6589 EXPORT_SYMBOL_GPL(__nft_release_basechain);
6591 static void __nft_release_tables(struct net *net)
6593 struct nft_flowtable *flowtable, *nf;
6594 struct nft_table *table, *nt;
6595 struct nft_chain *chain, *nc;
6596 struct nft_object *obj, *ne;
6597 struct nft_rule *rule, *nr;
6598 struct nft_set *set, *ns;
6599 struct nft_ctx ctx = {
6601 .family = NFPROTO_NETDEV,
6604 list_for_each_entry_safe(table, nt, &net->nft.tables, list) {
6605 ctx.family = table->family;
6607 list_for_each_entry(chain, &table->chains, list)
6608 nf_tables_unregister_hook(net, table, chain);
6609 list_for_each_entry(flowtable, &table->flowtables, list)
6610 nf_unregister_net_hooks(net, flowtable->ops,
6611 flowtable->ops_len);
6612 /* No packets are walking on these chains anymore. */
6614 list_for_each_entry(chain, &table->chains, list) {
6616 list_for_each_entry_safe(rule, nr, &chain->rules, list) {
6617 list_del(&rule->list);
6619 nf_tables_rule_destroy(&ctx, rule);
6622 list_for_each_entry_safe(flowtable, nf, &table->flowtables, list) {
6623 list_del(&flowtable->list);
6625 nf_tables_flowtable_destroy(flowtable);
6627 list_for_each_entry_safe(set, ns, &table->sets, list) {
6628 list_del(&set->list);
6630 nft_set_destroy(set);
6632 list_for_each_entry_safe(obj, ne, &table->objects, list) {
6633 list_del(&obj->list);
6635 nft_obj_destroy(obj);
6637 list_for_each_entry_safe(chain, nc, &table->chains, list) {
6639 list_del(&chain->list);
6641 nf_tables_chain_destroy(&ctx);
6643 list_del(&table->list);
6644 nf_tables_table_destroy(&ctx);
6648 static int __net_init nf_tables_init_net(struct net *net)
6650 INIT_LIST_HEAD(&net->nft.tables);
6651 INIT_LIST_HEAD(&net->nft.commit_list);
6652 net->nft.base_seq = 1;
6656 static void __net_exit nf_tables_exit_net(struct net *net)
6658 __nft_release_tables(net);
6659 WARN_ON_ONCE(!list_empty(&net->nft.tables));
6660 WARN_ON_ONCE(!list_empty(&net->nft.commit_list));
6663 static struct pernet_operations nf_tables_net_ops = {
6664 .init = nf_tables_init_net,
6665 .exit = nf_tables_exit_net,
6668 static int __init nf_tables_module_init(void)
6672 nft_chain_filter_init();
6674 info = kmalloc(sizeof(struct nft_expr_info) * NFT_RULE_MAXEXPRS,
6681 err = nf_tables_core_module_init();
6685 err = nfnetlink_subsys_register(&nf_tables_subsys);
6689 register_netdevice_notifier(&nf_tables_flowtable_notifier);
6691 return register_pernet_subsys(&nf_tables_net_ops);
6693 nf_tables_core_module_exit();
6700 static void __exit nf_tables_module_exit(void)
6702 unregister_pernet_subsys(&nf_tables_net_ops);
6703 nfnetlink_subsys_unregister(&nf_tables_subsys);
6704 unregister_netdevice_notifier(&nf_tables_flowtable_notifier);
6706 nf_tables_core_module_exit();
6708 nft_chain_filter_fini();
6711 module_init(nf_tables_module_init);
6712 module_exit(nf_tables_module_exit);
6714 MODULE_LICENSE("GPL");
6715 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
6716 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_NFTABLES);