1b57dc7c1SPaul Blakey // SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB 2b57dc7c1SPaul Blakey /* - 3b57dc7c1SPaul Blakey * net/sched/act_ct.c Connection Tracking action 4b57dc7c1SPaul Blakey * 5b57dc7c1SPaul Blakey * Authors: Paul Blakey <paulb@mellanox.com> 6b57dc7c1SPaul Blakey * Yossi Kuperman <yossiku@mellanox.com> 7b57dc7c1SPaul Blakey * Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> 8b57dc7c1SPaul Blakey */ 9b57dc7c1SPaul Blakey 10b57dc7c1SPaul Blakey #include <linux/module.h> 11b57dc7c1SPaul Blakey #include <linux/init.h> 12b57dc7c1SPaul Blakey #include <linux/kernel.h> 13b57dc7c1SPaul Blakey #include <linux/skbuff.h> 14b57dc7c1SPaul Blakey #include <linux/rtnetlink.h> 15b57dc7c1SPaul Blakey #include <linux/pkt_cls.h> 16b57dc7c1SPaul Blakey #include <linux/ip.h> 17b57dc7c1SPaul Blakey #include <linux/ipv6.h> 18c34b961aSPaul Blakey #include <linux/rhashtable.h> 19b57dc7c1SPaul Blakey #include <net/netlink.h> 20b57dc7c1SPaul Blakey #include <net/pkt_sched.h> 21b57dc7c1SPaul Blakey #include <net/pkt_cls.h> 22b57dc7c1SPaul Blakey #include <net/act_api.h> 23b57dc7c1SPaul Blakey #include <net/ip.h> 24b57dc7c1SPaul Blakey #include <net/ipv6_frag.h> 25b57dc7c1SPaul Blakey #include <uapi/linux/tc_act/tc_ct.h> 26b57dc7c1SPaul Blakey #include <net/tc_act/tc_ct.h> 27b57dc7c1SPaul Blakey 28c34b961aSPaul Blakey #include <net/netfilter/nf_flow_table.h> 29b57dc7c1SPaul Blakey #include <net/netfilter/nf_conntrack.h> 30b57dc7c1SPaul Blakey #include <net/netfilter/nf_conntrack_core.h> 31b57dc7c1SPaul Blakey #include <net/netfilter/nf_conntrack_zones.h> 32b57dc7c1SPaul Blakey #include <net/netfilter/nf_conntrack_helper.h> 33b57dc7c1SPaul Blakey #include <net/netfilter/ipv6/nf_defrag_ipv6.h> 3440d102cdSJeremy Sowden #include <uapi/linux/netfilter/nf_nat.h> 35b57dc7c1SPaul Blakey 36c34b961aSPaul Blakey static struct workqueue_struct *act_ct_wq; 37c34b961aSPaul Blakey static struct rhashtable zones_ht; 38138470a9SEric Dumazet static DEFINE_MUTEX(zones_mutex); 39c34b961aSPaul Blakey 40c34b961aSPaul Blakey struct tcf_ct_flow_table { 41c34b961aSPaul Blakey struct rhash_head node; /* In zones tables */ 42c34b961aSPaul Blakey 43c34b961aSPaul Blakey struct rcu_work rwork; 44c34b961aSPaul Blakey struct nf_flowtable nf_ft; 45138470a9SEric Dumazet refcount_t ref; 46c34b961aSPaul Blakey u16 zone; 47c34b961aSPaul Blakey 48c34b961aSPaul Blakey bool dying; 49c34b961aSPaul Blakey }; 50c34b961aSPaul Blakey 51c34b961aSPaul Blakey static const struct rhashtable_params zones_params = { 52c34b961aSPaul Blakey .head_offset = offsetof(struct tcf_ct_flow_table, node), 53c34b961aSPaul Blakey .key_offset = offsetof(struct tcf_ct_flow_table, zone), 54c34b961aSPaul Blakey .key_len = sizeof_field(struct tcf_ct_flow_table, zone), 55c34b961aSPaul Blakey .automatic_shrinking = true, 56c34b961aSPaul Blakey }; 57c34b961aSPaul Blakey 589c26ba9bSPaul Blakey static struct flow_action_entry * 599c26ba9bSPaul Blakey tcf_ct_flow_table_flow_action_get_next(struct flow_action *flow_action) 609c26ba9bSPaul Blakey { 619c26ba9bSPaul Blakey int i = flow_action->num_entries++; 629c26ba9bSPaul Blakey 639c26ba9bSPaul Blakey return &flow_action->entries[i]; 649c26ba9bSPaul Blakey } 659c26ba9bSPaul Blakey 669c26ba9bSPaul Blakey static void tcf_ct_add_mangle_action(struct flow_action *action, 679c26ba9bSPaul Blakey enum flow_action_mangle_base htype, 689c26ba9bSPaul Blakey u32 offset, 699c26ba9bSPaul Blakey u32 mask, 709c26ba9bSPaul Blakey u32 val) 719c26ba9bSPaul Blakey { 729c26ba9bSPaul Blakey struct flow_action_entry *entry; 739c26ba9bSPaul Blakey 749c26ba9bSPaul Blakey entry = tcf_ct_flow_table_flow_action_get_next(action); 759c26ba9bSPaul Blakey entry->id = FLOW_ACTION_MANGLE; 769c26ba9bSPaul Blakey entry->mangle.htype = htype; 779c26ba9bSPaul Blakey entry->mangle.mask = ~mask; 789c26ba9bSPaul Blakey entry->mangle.offset = offset; 799c26ba9bSPaul Blakey entry->mangle.val = val; 809c26ba9bSPaul Blakey } 819c26ba9bSPaul Blakey 829c26ba9bSPaul Blakey /* The following nat helper functions check if the inverted reverse tuple 839c26ba9bSPaul Blakey * (target) is different then the current dir tuple - meaning nat for ports 849c26ba9bSPaul Blakey * and/or ip is needed, and add the relevant mangle actions. 859c26ba9bSPaul Blakey */ 869c26ba9bSPaul Blakey static void 879c26ba9bSPaul Blakey tcf_ct_flow_table_add_action_nat_ipv4(const struct nf_conntrack_tuple *tuple, 889c26ba9bSPaul Blakey struct nf_conntrack_tuple target, 899c26ba9bSPaul Blakey struct flow_action *action) 909c26ba9bSPaul Blakey { 919c26ba9bSPaul Blakey if (memcmp(&target.src.u3, &tuple->src.u3, sizeof(target.src.u3))) 929c26ba9bSPaul Blakey tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_IP4, 939c26ba9bSPaul Blakey offsetof(struct iphdr, saddr), 949c26ba9bSPaul Blakey 0xFFFFFFFF, 959c26ba9bSPaul Blakey be32_to_cpu(target.src.u3.ip)); 969c26ba9bSPaul Blakey if (memcmp(&target.dst.u3, &tuple->dst.u3, sizeof(target.dst.u3))) 979c26ba9bSPaul Blakey tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_IP4, 989c26ba9bSPaul Blakey offsetof(struct iphdr, daddr), 999c26ba9bSPaul Blakey 0xFFFFFFFF, 1009c26ba9bSPaul Blakey be32_to_cpu(target.dst.u3.ip)); 1019c26ba9bSPaul Blakey } 1029c26ba9bSPaul Blakey 1039c26ba9bSPaul Blakey static void 1049c26ba9bSPaul Blakey tcf_ct_add_ipv6_addr_mangle_action(struct flow_action *action, 1059c26ba9bSPaul Blakey union nf_inet_addr *addr, 1069c26ba9bSPaul Blakey u32 offset) 1079c26ba9bSPaul Blakey { 1089c26ba9bSPaul Blakey int i; 1099c26ba9bSPaul Blakey 1109c26ba9bSPaul Blakey for (i = 0; i < sizeof(struct in6_addr) / sizeof(u32); i++) 1119c26ba9bSPaul Blakey tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_IP6, 1129c26ba9bSPaul Blakey i * sizeof(u32) + offset, 1139c26ba9bSPaul Blakey 0xFFFFFFFF, be32_to_cpu(addr->ip6[i])); 1149c26ba9bSPaul Blakey } 1159c26ba9bSPaul Blakey 1169c26ba9bSPaul Blakey static void 1179c26ba9bSPaul Blakey tcf_ct_flow_table_add_action_nat_ipv6(const struct nf_conntrack_tuple *tuple, 1189c26ba9bSPaul Blakey struct nf_conntrack_tuple target, 1199c26ba9bSPaul Blakey struct flow_action *action) 1209c26ba9bSPaul Blakey { 1219c26ba9bSPaul Blakey if (memcmp(&target.src.u3, &tuple->src.u3, sizeof(target.src.u3))) 1229c26ba9bSPaul Blakey tcf_ct_add_ipv6_addr_mangle_action(action, &target.src.u3, 1239c26ba9bSPaul Blakey offsetof(struct ipv6hdr, 1249c26ba9bSPaul Blakey saddr)); 1259c26ba9bSPaul Blakey if (memcmp(&target.dst.u3, &tuple->dst.u3, sizeof(target.dst.u3))) 1269c26ba9bSPaul Blakey tcf_ct_add_ipv6_addr_mangle_action(action, &target.dst.u3, 1279c26ba9bSPaul Blakey offsetof(struct ipv6hdr, 1289c26ba9bSPaul Blakey daddr)); 1299c26ba9bSPaul Blakey } 1309c26ba9bSPaul Blakey 1319c26ba9bSPaul Blakey static void 1329c26ba9bSPaul Blakey tcf_ct_flow_table_add_action_nat_tcp(const struct nf_conntrack_tuple *tuple, 1339c26ba9bSPaul Blakey struct nf_conntrack_tuple target, 1349c26ba9bSPaul Blakey struct flow_action *action) 1359c26ba9bSPaul Blakey { 1369c26ba9bSPaul Blakey __be16 target_src = target.src.u.tcp.port; 1379c26ba9bSPaul Blakey __be16 target_dst = target.dst.u.tcp.port; 1389c26ba9bSPaul Blakey 1399c26ba9bSPaul Blakey if (target_src != tuple->src.u.tcp.port) 1409c26ba9bSPaul Blakey tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_TCP, 1419c26ba9bSPaul Blakey offsetof(struct tcphdr, source), 1429c26ba9bSPaul Blakey 0xFFFF, be16_to_cpu(target_src)); 1439c26ba9bSPaul Blakey if (target_dst != tuple->dst.u.tcp.port) 1449c26ba9bSPaul Blakey tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_TCP, 1459c26ba9bSPaul Blakey offsetof(struct tcphdr, dest), 1469c26ba9bSPaul Blakey 0xFFFF, be16_to_cpu(target_dst)); 1479c26ba9bSPaul Blakey } 1489c26ba9bSPaul Blakey 1499c26ba9bSPaul Blakey static void 1509c26ba9bSPaul Blakey tcf_ct_flow_table_add_action_nat_udp(const struct nf_conntrack_tuple *tuple, 1519c26ba9bSPaul Blakey struct nf_conntrack_tuple target, 1529c26ba9bSPaul Blakey struct flow_action *action) 1539c26ba9bSPaul Blakey { 1549c26ba9bSPaul Blakey __be16 target_src = target.src.u.udp.port; 1559c26ba9bSPaul Blakey __be16 target_dst = target.dst.u.udp.port; 1569c26ba9bSPaul Blakey 1579c26ba9bSPaul Blakey if (target_src != tuple->src.u.udp.port) 1589c26ba9bSPaul Blakey tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_TCP, 1599c26ba9bSPaul Blakey offsetof(struct udphdr, source), 1609c26ba9bSPaul Blakey 0xFFFF, be16_to_cpu(target_src)); 1619c26ba9bSPaul Blakey if (target_dst != tuple->dst.u.udp.port) 1629c26ba9bSPaul Blakey tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_TCP, 1639c26ba9bSPaul Blakey offsetof(struct udphdr, dest), 1649c26ba9bSPaul Blakey 0xFFFF, be16_to_cpu(target_dst)); 1659c26ba9bSPaul Blakey } 1669c26ba9bSPaul Blakey 1679c26ba9bSPaul Blakey static void tcf_ct_flow_table_add_action_meta(struct nf_conn *ct, 1689c26ba9bSPaul Blakey enum ip_conntrack_dir dir, 1699c26ba9bSPaul Blakey struct flow_action *action) 1709c26ba9bSPaul Blakey { 1719c26ba9bSPaul Blakey struct nf_conn_labels *ct_labels; 1729c26ba9bSPaul Blakey struct flow_action_entry *entry; 17330b0cf90SPaul Blakey enum ip_conntrack_info ctinfo; 1749c26ba9bSPaul Blakey u32 *act_ct_labels; 1759c26ba9bSPaul Blakey 1769c26ba9bSPaul Blakey entry = tcf_ct_flow_table_flow_action_get_next(action); 1779c26ba9bSPaul Blakey entry->id = FLOW_ACTION_CT_METADATA; 1789c26ba9bSPaul Blakey #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) 1799c26ba9bSPaul Blakey entry->ct_metadata.mark = ct->mark; 1809c26ba9bSPaul Blakey #endif 18130b0cf90SPaul Blakey ctinfo = dir == IP_CT_DIR_ORIGINAL ? IP_CT_ESTABLISHED : 18230b0cf90SPaul Blakey IP_CT_ESTABLISHED_REPLY; 18330b0cf90SPaul Blakey /* aligns with the CT reference on the SKB nf_ct_set */ 18430b0cf90SPaul Blakey entry->ct_metadata.cookie = (unsigned long)ct | ctinfo; 1859c26ba9bSPaul Blakey 1869c26ba9bSPaul Blakey act_ct_labels = entry->ct_metadata.labels; 1879c26ba9bSPaul Blakey ct_labels = nf_ct_labels_find(ct); 1889c26ba9bSPaul Blakey if (ct_labels) 1899c26ba9bSPaul Blakey memcpy(act_ct_labels, ct_labels->bits, NF_CT_LABELS_MAX_SIZE); 1909c26ba9bSPaul Blakey else 1919c26ba9bSPaul Blakey memset(act_ct_labels, 0, NF_CT_LABELS_MAX_SIZE); 1929c26ba9bSPaul Blakey } 1939c26ba9bSPaul Blakey 1949c26ba9bSPaul Blakey static int tcf_ct_flow_table_add_action_nat(struct net *net, 1959c26ba9bSPaul Blakey struct nf_conn *ct, 1969c26ba9bSPaul Blakey enum ip_conntrack_dir dir, 1979c26ba9bSPaul Blakey struct flow_action *action) 1989c26ba9bSPaul Blakey { 1999c26ba9bSPaul Blakey const struct nf_conntrack_tuple *tuple = &ct->tuplehash[dir].tuple; 2009c26ba9bSPaul Blakey struct nf_conntrack_tuple target; 2019c26ba9bSPaul Blakey 20205aa69e5Swenxu if (!(ct->status & IPS_NAT_MASK)) 20305aa69e5Swenxu return 0; 20405aa69e5Swenxu 2059c26ba9bSPaul Blakey nf_ct_invert_tuple(&target, &ct->tuplehash[!dir].tuple); 2069c26ba9bSPaul Blakey 2079c26ba9bSPaul Blakey switch (tuple->src.l3num) { 2089c26ba9bSPaul Blakey case NFPROTO_IPV4: 2099c26ba9bSPaul Blakey tcf_ct_flow_table_add_action_nat_ipv4(tuple, target, 2109c26ba9bSPaul Blakey action); 2119c26ba9bSPaul Blakey break; 2129c26ba9bSPaul Blakey case NFPROTO_IPV6: 2139c26ba9bSPaul Blakey tcf_ct_flow_table_add_action_nat_ipv6(tuple, target, 2149c26ba9bSPaul Blakey action); 2159c26ba9bSPaul Blakey break; 2169c26ba9bSPaul Blakey default: 2179c26ba9bSPaul Blakey return -EOPNOTSUPP; 2189c26ba9bSPaul Blakey } 2199c26ba9bSPaul Blakey 2209c26ba9bSPaul Blakey switch (nf_ct_protonum(ct)) { 2219c26ba9bSPaul Blakey case IPPROTO_TCP: 2229c26ba9bSPaul Blakey tcf_ct_flow_table_add_action_nat_tcp(tuple, target, action); 2239c26ba9bSPaul Blakey break; 2249c26ba9bSPaul Blakey case IPPROTO_UDP: 2259c26ba9bSPaul Blakey tcf_ct_flow_table_add_action_nat_udp(tuple, target, action); 2269c26ba9bSPaul Blakey break; 2279c26ba9bSPaul Blakey default: 2289c26ba9bSPaul Blakey return -EOPNOTSUPP; 2299c26ba9bSPaul Blakey } 2309c26ba9bSPaul Blakey 2319c26ba9bSPaul Blakey return 0; 2329c26ba9bSPaul Blakey } 2339c26ba9bSPaul Blakey 2349c26ba9bSPaul Blakey static int tcf_ct_flow_table_fill_actions(struct net *net, 2359c26ba9bSPaul Blakey const struct flow_offload *flow, 2369c26ba9bSPaul Blakey enum flow_offload_tuple_dir tdir, 2379c26ba9bSPaul Blakey struct nf_flow_rule *flow_rule) 2389c26ba9bSPaul Blakey { 2399c26ba9bSPaul Blakey struct flow_action *action = &flow_rule->rule->action; 2409c26ba9bSPaul Blakey int num_entries = action->num_entries; 2419c26ba9bSPaul Blakey struct nf_conn *ct = flow->ct; 2429c26ba9bSPaul Blakey enum ip_conntrack_dir dir; 2439c26ba9bSPaul Blakey int i, err; 2449c26ba9bSPaul Blakey 2459c26ba9bSPaul Blakey switch (tdir) { 2469c26ba9bSPaul Blakey case FLOW_OFFLOAD_DIR_ORIGINAL: 2479c26ba9bSPaul Blakey dir = IP_CT_DIR_ORIGINAL; 2489c26ba9bSPaul Blakey break; 2499c26ba9bSPaul Blakey case FLOW_OFFLOAD_DIR_REPLY: 2509c26ba9bSPaul Blakey dir = IP_CT_DIR_REPLY; 2519c26ba9bSPaul Blakey break; 2529c26ba9bSPaul Blakey default: 2539c26ba9bSPaul Blakey return -EOPNOTSUPP; 2549c26ba9bSPaul Blakey } 2559c26ba9bSPaul Blakey 2569c26ba9bSPaul Blakey err = tcf_ct_flow_table_add_action_nat(net, ct, dir, action); 2579c26ba9bSPaul Blakey if (err) 2589c26ba9bSPaul Blakey goto err_nat; 2599c26ba9bSPaul Blakey 2609c26ba9bSPaul Blakey tcf_ct_flow_table_add_action_meta(ct, dir, action); 2619c26ba9bSPaul Blakey return 0; 2629c26ba9bSPaul Blakey 2639c26ba9bSPaul Blakey err_nat: 2649c26ba9bSPaul Blakey /* Clear filled actions */ 2659c26ba9bSPaul Blakey for (i = num_entries; i < action->num_entries; i++) 2669c26ba9bSPaul Blakey memset(&action->entries[i], 0, sizeof(action->entries[i])); 2679c26ba9bSPaul Blakey action->num_entries = num_entries; 2689c26ba9bSPaul Blakey 2699c26ba9bSPaul Blakey return err; 2709c26ba9bSPaul Blakey } 2719c26ba9bSPaul Blakey 272c34b961aSPaul Blakey static struct nf_flowtable_type flowtable_ct = { 2739c26ba9bSPaul Blakey .action = tcf_ct_flow_table_fill_actions, 274c34b961aSPaul Blakey .owner = THIS_MODULE, 275c34b961aSPaul Blakey }; 276c34b961aSPaul Blakey 277c34b961aSPaul Blakey static int tcf_ct_flow_table_get(struct tcf_ct_params *params) 278c34b961aSPaul Blakey { 279c34b961aSPaul Blakey struct tcf_ct_flow_table *ct_ft; 280c34b961aSPaul Blakey int err = -ENOMEM; 281c34b961aSPaul Blakey 282138470a9SEric Dumazet mutex_lock(&zones_mutex); 283c34b961aSPaul Blakey ct_ft = rhashtable_lookup_fast(&zones_ht, ¶ms->zone, zones_params); 284138470a9SEric Dumazet if (ct_ft && refcount_inc_not_zero(&ct_ft->ref)) 285138470a9SEric Dumazet goto out_unlock; 286c34b961aSPaul Blakey 287138470a9SEric Dumazet ct_ft = kzalloc(sizeof(*ct_ft), GFP_KERNEL); 288c34b961aSPaul Blakey if (!ct_ft) 289c34b961aSPaul Blakey goto err_alloc; 290138470a9SEric Dumazet refcount_set(&ct_ft->ref, 1); 291c34b961aSPaul Blakey 292c34b961aSPaul Blakey ct_ft->zone = params->zone; 293c34b961aSPaul Blakey err = rhashtable_insert_fast(&zones_ht, &ct_ft->node, zones_params); 294c34b961aSPaul Blakey if (err) 295c34b961aSPaul Blakey goto err_insert; 296c34b961aSPaul Blakey 297c34b961aSPaul Blakey ct_ft->nf_ft.type = &flowtable_ct; 298edd5861eSPaul Blakey ct_ft->nf_ft.flags |= NF_FLOWTABLE_HW_OFFLOAD; 299c34b961aSPaul Blakey err = nf_flow_table_init(&ct_ft->nf_ft); 300c34b961aSPaul Blakey if (err) 301c34b961aSPaul Blakey goto err_init; 302c34b961aSPaul Blakey 303c34b961aSPaul Blakey __module_get(THIS_MODULE); 304138470a9SEric Dumazet out_unlock: 305c34b961aSPaul Blakey params->ct_ft = ct_ft; 306edd5861eSPaul Blakey params->nf_ft = &ct_ft->nf_ft; 307138470a9SEric Dumazet mutex_unlock(&zones_mutex); 308c34b961aSPaul Blakey 309c34b961aSPaul Blakey return 0; 310c34b961aSPaul Blakey 311c34b961aSPaul Blakey err_init: 312c34b961aSPaul Blakey rhashtable_remove_fast(&zones_ht, &ct_ft->node, zones_params); 313c34b961aSPaul Blakey err_insert: 314c34b961aSPaul Blakey kfree(ct_ft); 315c34b961aSPaul Blakey err_alloc: 316138470a9SEric Dumazet mutex_unlock(&zones_mutex); 317c34b961aSPaul Blakey return err; 318c34b961aSPaul Blakey } 319c34b961aSPaul Blakey 320c34b961aSPaul Blakey static void tcf_ct_flow_table_cleanup_work(struct work_struct *work) 321c34b961aSPaul Blakey { 322c34b961aSPaul Blakey struct tcf_ct_flow_table *ct_ft; 323c34b961aSPaul Blakey 324c34b961aSPaul Blakey ct_ft = container_of(to_rcu_work(work), struct tcf_ct_flow_table, 325c34b961aSPaul Blakey rwork); 326c34b961aSPaul Blakey nf_flow_table_free(&ct_ft->nf_ft); 327c34b961aSPaul Blakey kfree(ct_ft); 328c34b961aSPaul Blakey 329c34b961aSPaul Blakey module_put(THIS_MODULE); 330c34b961aSPaul Blakey } 331c34b961aSPaul Blakey 332c34b961aSPaul Blakey static void tcf_ct_flow_table_put(struct tcf_ct_params *params) 333c34b961aSPaul Blakey { 334c34b961aSPaul Blakey struct tcf_ct_flow_table *ct_ft = params->ct_ft; 335c34b961aSPaul Blakey 336138470a9SEric Dumazet if (refcount_dec_and_test(¶ms->ct_ft->ref)) { 337c34b961aSPaul Blakey rhashtable_remove_fast(&zones_ht, &ct_ft->node, zones_params); 338c34b961aSPaul Blakey INIT_RCU_WORK(&ct_ft->rwork, tcf_ct_flow_table_cleanup_work); 339c34b961aSPaul Blakey queue_rcu_work(act_ct_wq, &ct_ft->rwork); 340c34b961aSPaul Blakey } 341c34b961aSPaul Blakey } 342c34b961aSPaul Blakey 34364ff70b8SPaul Blakey static void tcf_ct_flow_table_add(struct tcf_ct_flow_table *ct_ft, 34464ff70b8SPaul Blakey struct nf_conn *ct, 34564ff70b8SPaul Blakey bool tcp) 34664ff70b8SPaul Blakey { 34764ff70b8SPaul Blakey struct flow_offload *entry; 34864ff70b8SPaul Blakey int err; 34964ff70b8SPaul Blakey 35064ff70b8SPaul Blakey if (test_and_set_bit(IPS_OFFLOAD_BIT, &ct->status)) 35164ff70b8SPaul Blakey return; 35264ff70b8SPaul Blakey 35364ff70b8SPaul Blakey entry = flow_offload_alloc(ct); 35464ff70b8SPaul Blakey if (!entry) { 35564ff70b8SPaul Blakey WARN_ON_ONCE(1); 35664ff70b8SPaul Blakey goto err_alloc; 35764ff70b8SPaul Blakey } 35864ff70b8SPaul Blakey 35964ff70b8SPaul Blakey if (tcp) { 36064ff70b8SPaul Blakey ct->proto.tcp.seen[0].flags |= IP_CT_TCP_FLAG_BE_LIBERAL; 36164ff70b8SPaul Blakey ct->proto.tcp.seen[1].flags |= IP_CT_TCP_FLAG_BE_LIBERAL; 36264ff70b8SPaul Blakey } 36364ff70b8SPaul Blakey 36464ff70b8SPaul Blakey err = flow_offload_add(&ct_ft->nf_ft, entry); 36564ff70b8SPaul Blakey if (err) 36664ff70b8SPaul Blakey goto err_add; 36764ff70b8SPaul Blakey 36864ff70b8SPaul Blakey return; 36964ff70b8SPaul Blakey 37064ff70b8SPaul Blakey err_add: 37164ff70b8SPaul Blakey flow_offload_free(entry); 37264ff70b8SPaul Blakey err_alloc: 37364ff70b8SPaul Blakey clear_bit(IPS_OFFLOAD_BIT, &ct->status); 37464ff70b8SPaul Blakey } 37564ff70b8SPaul Blakey 37664ff70b8SPaul Blakey static void tcf_ct_flow_table_process_conn(struct tcf_ct_flow_table *ct_ft, 37764ff70b8SPaul Blakey struct nf_conn *ct, 37864ff70b8SPaul Blakey enum ip_conntrack_info ctinfo) 37964ff70b8SPaul Blakey { 38064ff70b8SPaul Blakey bool tcp = false; 38164ff70b8SPaul Blakey 38264ff70b8SPaul Blakey if (ctinfo != IP_CT_ESTABLISHED && ctinfo != IP_CT_ESTABLISHED_REPLY) 38364ff70b8SPaul Blakey return; 38464ff70b8SPaul Blakey 38564ff70b8SPaul Blakey switch (nf_ct_protonum(ct)) { 38664ff70b8SPaul Blakey case IPPROTO_TCP: 38764ff70b8SPaul Blakey tcp = true; 38864ff70b8SPaul Blakey if (ct->proto.tcp.state != TCP_CONNTRACK_ESTABLISHED) 38964ff70b8SPaul Blakey return; 39064ff70b8SPaul Blakey break; 39164ff70b8SPaul Blakey case IPPROTO_UDP: 39264ff70b8SPaul Blakey break; 39364ff70b8SPaul Blakey default: 39464ff70b8SPaul Blakey return; 39564ff70b8SPaul Blakey } 39664ff70b8SPaul Blakey 39764ff70b8SPaul Blakey if (nf_ct_ext_exist(ct, NF_CT_EXT_HELPER) || 39864ff70b8SPaul Blakey ct->status & IPS_SEQ_ADJUST) 39964ff70b8SPaul Blakey return; 40064ff70b8SPaul Blakey 40164ff70b8SPaul Blakey tcf_ct_flow_table_add(ct_ft, ct, tcp); 40264ff70b8SPaul Blakey } 40364ff70b8SPaul Blakey 40446475bb2SPaul Blakey static bool 40546475bb2SPaul Blakey tcf_ct_flow_table_fill_tuple_ipv4(struct sk_buff *skb, 40607ac9d16SPaul Blakey struct flow_offload_tuple *tuple, 40707ac9d16SPaul Blakey struct tcphdr **tcph) 40846475bb2SPaul Blakey { 40946475bb2SPaul Blakey struct flow_ports *ports; 41046475bb2SPaul Blakey unsigned int thoff; 41146475bb2SPaul Blakey struct iphdr *iph; 41246475bb2SPaul Blakey 4134cc5fdecSPaul Blakey if (!pskb_network_may_pull(skb, sizeof(*iph))) 41446475bb2SPaul Blakey return false; 41546475bb2SPaul Blakey 41646475bb2SPaul Blakey iph = ip_hdr(skb); 41746475bb2SPaul Blakey thoff = iph->ihl * 4; 41846475bb2SPaul Blakey 41946475bb2SPaul Blakey if (ip_is_fragment(iph) || 42046475bb2SPaul Blakey unlikely(thoff != sizeof(struct iphdr))) 42146475bb2SPaul Blakey return false; 42246475bb2SPaul Blakey 42346475bb2SPaul Blakey if (iph->protocol != IPPROTO_TCP && 42446475bb2SPaul Blakey iph->protocol != IPPROTO_UDP) 42546475bb2SPaul Blakey return false; 42646475bb2SPaul Blakey 42746475bb2SPaul Blakey if (iph->ttl <= 1) 42846475bb2SPaul Blakey return false; 42946475bb2SPaul Blakey 4304cc5fdecSPaul Blakey if (!pskb_network_may_pull(skb, iph->protocol == IPPROTO_TCP ? 43107ac9d16SPaul Blakey thoff + sizeof(struct tcphdr) : 43207ac9d16SPaul Blakey thoff + sizeof(*ports))) 43346475bb2SPaul Blakey return false; 43446475bb2SPaul Blakey 43507ac9d16SPaul Blakey iph = ip_hdr(skb); 43607ac9d16SPaul Blakey if (iph->protocol == IPPROTO_TCP) 43707ac9d16SPaul Blakey *tcph = (void *)(skb_network_header(skb) + thoff); 43846475bb2SPaul Blakey 43907ac9d16SPaul Blakey ports = (struct flow_ports *)(skb_network_header(skb) + thoff); 44046475bb2SPaul Blakey tuple->src_v4.s_addr = iph->saddr; 44146475bb2SPaul Blakey tuple->dst_v4.s_addr = iph->daddr; 44246475bb2SPaul Blakey tuple->src_port = ports->source; 44346475bb2SPaul Blakey tuple->dst_port = ports->dest; 44446475bb2SPaul Blakey tuple->l3proto = AF_INET; 44546475bb2SPaul Blakey tuple->l4proto = iph->protocol; 44646475bb2SPaul Blakey 44746475bb2SPaul Blakey return true; 44846475bb2SPaul Blakey } 44946475bb2SPaul Blakey 45046475bb2SPaul Blakey static bool 45146475bb2SPaul Blakey tcf_ct_flow_table_fill_tuple_ipv6(struct sk_buff *skb, 45207ac9d16SPaul Blakey struct flow_offload_tuple *tuple, 45307ac9d16SPaul Blakey struct tcphdr **tcph) 45446475bb2SPaul Blakey { 45546475bb2SPaul Blakey struct flow_ports *ports; 45646475bb2SPaul Blakey struct ipv6hdr *ip6h; 45746475bb2SPaul Blakey unsigned int thoff; 45846475bb2SPaul Blakey 4594cc5fdecSPaul Blakey if (!pskb_network_may_pull(skb, sizeof(*ip6h))) 46046475bb2SPaul Blakey return false; 46146475bb2SPaul Blakey 46246475bb2SPaul Blakey ip6h = ipv6_hdr(skb); 46346475bb2SPaul Blakey 46446475bb2SPaul Blakey if (ip6h->nexthdr != IPPROTO_TCP && 46546475bb2SPaul Blakey ip6h->nexthdr != IPPROTO_UDP) 46646475bb2SPaul Blakey return false; 46746475bb2SPaul Blakey 46846475bb2SPaul Blakey if (ip6h->hop_limit <= 1) 46946475bb2SPaul Blakey return false; 47046475bb2SPaul Blakey 47146475bb2SPaul Blakey thoff = sizeof(*ip6h); 4724cc5fdecSPaul Blakey if (!pskb_network_may_pull(skb, ip6h->nexthdr == IPPROTO_TCP ? 47307ac9d16SPaul Blakey thoff + sizeof(struct tcphdr) : 47407ac9d16SPaul Blakey thoff + sizeof(*ports))) 47546475bb2SPaul Blakey return false; 47646475bb2SPaul Blakey 47707ac9d16SPaul Blakey ip6h = ipv6_hdr(skb); 47807ac9d16SPaul Blakey if (ip6h->nexthdr == IPPROTO_TCP) 47907ac9d16SPaul Blakey *tcph = (void *)(skb_network_header(skb) + thoff); 48046475bb2SPaul Blakey 48107ac9d16SPaul Blakey ports = (struct flow_ports *)(skb_network_header(skb) + thoff); 48246475bb2SPaul Blakey tuple->src_v6 = ip6h->saddr; 48346475bb2SPaul Blakey tuple->dst_v6 = ip6h->daddr; 48446475bb2SPaul Blakey tuple->src_port = ports->source; 48546475bb2SPaul Blakey tuple->dst_port = ports->dest; 48646475bb2SPaul Blakey tuple->l3proto = AF_INET6; 48746475bb2SPaul Blakey tuple->l4proto = ip6h->nexthdr; 48846475bb2SPaul Blakey 48946475bb2SPaul Blakey return true; 49046475bb2SPaul Blakey } 49146475bb2SPaul Blakey 49246475bb2SPaul Blakey static bool tcf_ct_flow_table_lookup(struct tcf_ct_params *p, 49346475bb2SPaul Blakey struct sk_buff *skb, 49446475bb2SPaul Blakey u8 family) 49546475bb2SPaul Blakey { 49646475bb2SPaul Blakey struct nf_flowtable *nf_ft = &p->ct_ft->nf_ft; 49746475bb2SPaul Blakey struct flow_offload_tuple_rhash *tuplehash; 49846475bb2SPaul Blakey struct flow_offload_tuple tuple = {}; 49946475bb2SPaul Blakey enum ip_conntrack_info ctinfo; 50007ac9d16SPaul Blakey struct tcphdr *tcph = NULL; 50146475bb2SPaul Blakey struct flow_offload *flow; 50246475bb2SPaul Blakey struct nf_conn *ct; 50346475bb2SPaul Blakey u8 dir; 50446475bb2SPaul Blakey 50546475bb2SPaul Blakey /* Previously seen or loopback */ 50646475bb2SPaul Blakey ct = nf_ct_get(skb, &ctinfo); 50746475bb2SPaul Blakey if ((ct && !nf_ct_is_template(ct)) || ctinfo == IP_CT_UNTRACKED) 50846475bb2SPaul Blakey return false; 50946475bb2SPaul Blakey 51046475bb2SPaul Blakey switch (family) { 51146475bb2SPaul Blakey case NFPROTO_IPV4: 51207ac9d16SPaul Blakey if (!tcf_ct_flow_table_fill_tuple_ipv4(skb, &tuple, &tcph)) 51346475bb2SPaul Blakey return false; 51446475bb2SPaul Blakey break; 51546475bb2SPaul Blakey case NFPROTO_IPV6: 51607ac9d16SPaul Blakey if (!tcf_ct_flow_table_fill_tuple_ipv6(skb, &tuple, &tcph)) 51746475bb2SPaul Blakey return false; 51846475bb2SPaul Blakey break; 51946475bb2SPaul Blakey default: 52046475bb2SPaul Blakey return false; 52146475bb2SPaul Blakey } 52246475bb2SPaul Blakey 52346475bb2SPaul Blakey tuplehash = flow_offload_lookup(nf_ft, &tuple); 52446475bb2SPaul Blakey if (!tuplehash) 52546475bb2SPaul Blakey return false; 52646475bb2SPaul Blakey 52746475bb2SPaul Blakey dir = tuplehash->tuple.dir; 52846475bb2SPaul Blakey flow = container_of(tuplehash, struct flow_offload, tuplehash[dir]); 52946475bb2SPaul Blakey ct = flow->ct; 53046475bb2SPaul Blakey 53107ac9d16SPaul Blakey if (tcph && (unlikely(tcph->fin || tcph->rst))) { 53207ac9d16SPaul Blakey flow_offload_teardown(flow); 53307ac9d16SPaul Blakey return false; 53407ac9d16SPaul Blakey } 53507ac9d16SPaul Blakey 53646475bb2SPaul Blakey ctinfo = dir == FLOW_OFFLOAD_DIR_ORIGINAL ? IP_CT_ESTABLISHED : 53746475bb2SPaul Blakey IP_CT_ESTABLISHED_REPLY; 53846475bb2SPaul Blakey 5398b3646d6SPaul Blakey flow_offload_refresh(nf_ft, flow); 54046475bb2SPaul Blakey nf_conntrack_get(&ct->ct_general); 54146475bb2SPaul Blakey nf_ct_set(skb, ct, ctinfo); 54246475bb2SPaul Blakey 54346475bb2SPaul Blakey return true; 54446475bb2SPaul Blakey } 54546475bb2SPaul Blakey 546c34b961aSPaul Blakey static int tcf_ct_flow_tables_init(void) 547c34b961aSPaul Blakey { 548c34b961aSPaul Blakey return rhashtable_init(&zones_ht, &zones_params); 549c34b961aSPaul Blakey } 550c34b961aSPaul Blakey 551c34b961aSPaul Blakey static void tcf_ct_flow_tables_uninit(void) 552c34b961aSPaul Blakey { 553c34b961aSPaul Blakey rhashtable_destroy(&zones_ht); 554c34b961aSPaul Blakey } 555c34b961aSPaul Blakey 556b57dc7c1SPaul Blakey static struct tc_action_ops act_ct_ops; 557b57dc7c1SPaul Blakey static unsigned int ct_net_id; 558b57dc7c1SPaul Blakey 559b57dc7c1SPaul Blakey struct tc_ct_action_net { 560b57dc7c1SPaul Blakey struct tc_action_net tn; /* Must be first */ 561b57dc7c1SPaul Blakey bool labels; 562b57dc7c1SPaul Blakey }; 563b57dc7c1SPaul Blakey 564b57dc7c1SPaul Blakey /* Determine whether skb->_nfct is equal to the result of conntrack lookup. */ 565b57dc7c1SPaul Blakey static bool tcf_ct_skb_nfct_cached(struct net *net, struct sk_buff *skb, 566b57dc7c1SPaul Blakey u16 zone_id, bool force) 567b57dc7c1SPaul Blakey { 568b57dc7c1SPaul Blakey enum ip_conntrack_info ctinfo; 569b57dc7c1SPaul Blakey struct nf_conn *ct; 570b57dc7c1SPaul Blakey 571b57dc7c1SPaul Blakey ct = nf_ct_get(skb, &ctinfo); 572b57dc7c1SPaul Blakey if (!ct) 573b57dc7c1SPaul Blakey return false; 574b57dc7c1SPaul Blakey if (!net_eq(net, read_pnet(&ct->ct_net))) 575b57dc7c1SPaul Blakey return false; 576b57dc7c1SPaul Blakey if (nf_ct_zone(ct)->id != zone_id) 577b57dc7c1SPaul Blakey return false; 578b57dc7c1SPaul Blakey 579b57dc7c1SPaul Blakey /* Force conntrack entry direction. */ 580b57dc7c1SPaul Blakey if (force && CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) { 581b57dc7c1SPaul Blakey if (nf_ct_is_confirmed(ct)) 582b57dc7c1SPaul Blakey nf_ct_kill(ct); 583b57dc7c1SPaul Blakey 584b57dc7c1SPaul Blakey nf_conntrack_put(&ct->ct_general); 585b57dc7c1SPaul Blakey nf_ct_set(skb, NULL, IP_CT_UNTRACKED); 586b57dc7c1SPaul Blakey 587b57dc7c1SPaul Blakey return false; 588b57dc7c1SPaul Blakey } 589b57dc7c1SPaul Blakey 590b57dc7c1SPaul Blakey return true; 591b57dc7c1SPaul Blakey } 592b57dc7c1SPaul Blakey 593b57dc7c1SPaul Blakey /* Trim the skb to the length specified by the IP/IPv6 header, 594b57dc7c1SPaul Blakey * removing any trailing lower-layer padding. This prepares the skb 595b57dc7c1SPaul Blakey * for higher-layer processing that assumes skb->len excludes padding 596b57dc7c1SPaul Blakey * (such as nf_ip_checksum). The caller needs to pull the skb to the 597b57dc7c1SPaul Blakey * network header, and ensure ip_hdr/ipv6_hdr points to valid data. 598b57dc7c1SPaul Blakey */ 599b57dc7c1SPaul Blakey static int tcf_ct_skb_network_trim(struct sk_buff *skb, int family) 600b57dc7c1SPaul Blakey { 601b57dc7c1SPaul Blakey unsigned int len; 602b57dc7c1SPaul Blakey int err; 603b57dc7c1SPaul Blakey 604b57dc7c1SPaul Blakey switch (family) { 605b57dc7c1SPaul Blakey case NFPROTO_IPV4: 606b57dc7c1SPaul Blakey len = ntohs(ip_hdr(skb)->tot_len); 607b57dc7c1SPaul Blakey break; 608b57dc7c1SPaul Blakey case NFPROTO_IPV6: 609b57dc7c1SPaul Blakey len = sizeof(struct ipv6hdr) 610b57dc7c1SPaul Blakey + ntohs(ipv6_hdr(skb)->payload_len); 611b57dc7c1SPaul Blakey break; 612b57dc7c1SPaul Blakey default: 613b57dc7c1SPaul Blakey len = skb->len; 614b57dc7c1SPaul Blakey } 615b57dc7c1SPaul Blakey 616b57dc7c1SPaul Blakey err = pskb_trim_rcsum(skb, len); 617b57dc7c1SPaul Blakey 618b57dc7c1SPaul Blakey return err; 619b57dc7c1SPaul Blakey } 620b57dc7c1SPaul Blakey 621b57dc7c1SPaul Blakey static u8 tcf_ct_skb_nf_family(struct sk_buff *skb) 622b57dc7c1SPaul Blakey { 623b57dc7c1SPaul Blakey u8 family = NFPROTO_UNSPEC; 624b57dc7c1SPaul Blakey 625b57dc7c1SPaul Blakey switch (skb->protocol) { 626b57dc7c1SPaul Blakey case htons(ETH_P_IP): 627b57dc7c1SPaul Blakey family = NFPROTO_IPV4; 628b57dc7c1SPaul Blakey break; 629b57dc7c1SPaul Blakey case htons(ETH_P_IPV6): 630b57dc7c1SPaul Blakey family = NFPROTO_IPV6; 631b57dc7c1SPaul Blakey break; 632b57dc7c1SPaul Blakey default: 633b57dc7c1SPaul Blakey break; 634b57dc7c1SPaul Blakey } 635b57dc7c1SPaul Blakey 636b57dc7c1SPaul Blakey return family; 637b57dc7c1SPaul Blakey } 638b57dc7c1SPaul Blakey 639b57dc7c1SPaul Blakey static int tcf_ct_ipv4_is_fragment(struct sk_buff *skb, bool *frag) 640b57dc7c1SPaul Blakey { 641b57dc7c1SPaul Blakey unsigned int len; 642b57dc7c1SPaul Blakey 643b57dc7c1SPaul Blakey len = skb_network_offset(skb) + sizeof(struct iphdr); 644b57dc7c1SPaul Blakey if (unlikely(skb->len < len)) 645b57dc7c1SPaul Blakey return -EINVAL; 646b57dc7c1SPaul Blakey if (unlikely(!pskb_may_pull(skb, len))) 647b57dc7c1SPaul Blakey return -ENOMEM; 648b57dc7c1SPaul Blakey 649b57dc7c1SPaul Blakey *frag = ip_is_fragment(ip_hdr(skb)); 650b57dc7c1SPaul Blakey return 0; 651b57dc7c1SPaul Blakey } 652b57dc7c1SPaul Blakey 653b57dc7c1SPaul Blakey static int tcf_ct_ipv6_is_fragment(struct sk_buff *skb, bool *frag) 654b57dc7c1SPaul Blakey { 655b57dc7c1SPaul Blakey unsigned int flags = 0, len, payload_ofs = 0; 656b57dc7c1SPaul Blakey unsigned short frag_off; 657b57dc7c1SPaul Blakey int nexthdr; 658b57dc7c1SPaul Blakey 659b57dc7c1SPaul Blakey len = skb_network_offset(skb) + sizeof(struct ipv6hdr); 660b57dc7c1SPaul Blakey if (unlikely(skb->len < len)) 661b57dc7c1SPaul Blakey return -EINVAL; 662b57dc7c1SPaul Blakey if (unlikely(!pskb_may_pull(skb, len))) 663b57dc7c1SPaul Blakey return -ENOMEM; 664b57dc7c1SPaul Blakey 665b57dc7c1SPaul Blakey nexthdr = ipv6_find_hdr(skb, &payload_ofs, -1, &frag_off, &flags); 666b57dc7c1SPaul Blakey if (unlikely(nexthdr < 0)) 667b57dc7c1SPaul Blakey return -EPROTO; 668b57dc7c1SPaul Blakey 669b57dc7c1SPaul Blakey *frag = flags & IP6_FH_F_FRAG; 670b57dc7c1SPaul Blakey return 0; 671b57dc7c1SPaul Blakey } 672b57dc7c1SPaul Blakey 673b57dc7c1SPaul Blakey static int tcf_ct_handle_fragments(struct net *net, struct sk_buff *skb, 674b57dc7c1SPaul Blakey u8 family, u16 zone) 675b57dc7c1SPaul Blakey { 676b57dc7c1SPaul Blakey enum ip_conntrack_info ctinfo; 677b57dc7c1SPaul Blakey struct nf_conn *ct; 678b57dc7c1SPaul Blakey int err = 0; 679b57dc7c1SPaul Blakey bool frag; 680b57dc7c1SPaul Blakey 681b57dc7c1SPaul Blakey /* Previously seen (loopback)? Ignore. */ 682b57dc7c1SPaul Blakey ct = nf_ct_get(skb, &ctinfo); 683b57dc7c1SPaul Blakey if ((ct && !nf_ct_is_template(ct)) || ctinfo == IP_CT_UNTRACKED) 684b57dc7c1SPaul Blakey return 0; 685b57dc7c1SPaul Blakey 686b57dc7c1SPaul Blakey if (family == NFPROTO_IPV4) 687b57dc7c1SPaul Blakey err = tcf_ct_ipv4_is_fragment(skb, &frag); 688b57dc7c1SPaul Blakey else 689b57dc7c1SPaul Blakey err = tcf_ct_ipv6_is_fragment(skb, &frag); 690b57dc7c1SPaul Blakey if (err || !frag) 691b57dc7c1SPaul Blakey return err; 692b57dc7c1SPaul Blakey 693b57dc7c1SPaul Blakey skb_get(skb); 694b57dc7c1SPaul Blakey 695b57dc7c1SPaul Blakey if (family == NFPROTO_IPV4) { 696b57dc7c1SPaul Blakey enum ip_defrag_users user = IP_DEFRAG_CONNTRACK_IN + zone; 697b57dc7c1SPaul Blakey 698b57dc7c1SPaul Blakey memset(IPCB(skb), 0, sizeof(struct inet_skb_parm)); 699b57dc7c1SPaul Blakey local_bh_disable(); 700b57dc7c1SPaul Blakey err = ip_defrag(net, skb, user); 701b57dc7c1SPaul Blakey local_bh_enable(); 702b57dc7c1SPaul Blakey if (err && err != -EINPROGRESS) 703b57dc7c1SPaul Blakey goto out_free; 704b57dc7c1SPaul Blakey } else { /* NFPROTO_IPV6 */ 705b57dc7c1SPaul Blakey #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6) 706b57dc7c1SPaul Blakey enum ip6_defrag_users user = IP6_DEFRAG_CONNTRACK_IN + zone; 707b57dc7c1SPaul Blakey 708b57dc7c1SPaul Blakey memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm)); 709b57dc7c1SPaul Blakey err = nf_ct_frag6_gather(net, skb, user); 710b57dc7c1SPaul Blakey if (err && err != -EINPROGRESS) 711b57dc7c1SPaul Blakey goto out_free; 712b57dc7c1SPaul Blakey #else 713b57dc7c1SPaul Blakey err = -EOPNOTSUPP; 714b57dc7c1SPaul Blakey goto out_free; 715b57dc7c1SPaul Blakey #endif 716b57dc7c1SPaul Blakey } 717b57dc7c1SPaul Blakey 718b57dc7c1SPaul Blakey skb_clear_hash(skb); 719b57dc7c1SPaul Blakey skb->ignore_df = 1; 720b57dc7c1SPaul Blakey return err; 721b57dc7c1SPaul Blakey 722b57dc7c1SPaul Blakey out_free: 723b57dc7c1SPaul Blakey kfree_skb(skb); 724b57dc7c1SPaul Blakey return err; 725b57dc7c1SPaul Blakey } 726b57dc7c1SPaul Blakey 727b57dc7c1SPaul Blakey static void tcf_ct_params_free(struct rcu_head *head) 728b57dc7c1SPaul Blakey { 729b57dc7c1SPaul Blakey struct tcf_ct_params *params = container_of(head, 730b57dc7c1SPaul Blakey struct tcf_ct_params, rcu); 731b57dc7c1SPaul Blakey 732c34b961aSPaul Blakey tcf_ct_flow_table_put(params); 733c34b961aSPaul Blakey 734b57dc7c1SPaul Blakey if (params->tmpl) 735b57dc7c1SPaul Blakey nf_conntrack_put(¶ms->tmpl->ct_general); 736b57dc7c1SPaul Blakey kfree(params); 737b57dc7c1SPaul Blakey } 738b57dc7c1SPaul Blakey 739b57dc7c1SPaul Blakey #if IS_ENABLED(CONFIG_NF_NAT) 740b57dc7c1SPaul Blakey /* Modelled after nf_nat_ipv[46]_fn(). 741b57dc7c1SPaul Blakey * range is only used for new, uninitialized NAT state. 742b57dc7c1SPaul Blakey * Returns either NF_ACCEPT or NF_DROP. 743b57dc7c1SPaul Blakey */ 744b57dc7c1SPaul Blakey static int ct_nat_execute(struct sk_buff *skb, struct nf_conn *ct, 745b57dc7c1SPaul Blakey enum ip_conntrack_info ctinfo, 746b57dc7c1SPaul Blakey const struct nf_nat_range2 *range, 747b57dc7c1SPaul Blakey enum nf_nat_manip_type maniptype) 748b57dc7c1SPaul Blakey { 749b57dc7c1SPaul Blakey int hooknum, err = NF_ACCEPT; 750b57dc7c1SPaul Blakey 751b57dc7c1SPaul Blakey /* See HOOK2MANIP(). */ 752b57dc7c1SPaul Blakey if (maniptype == NF_NAT_MANIP_SRC) 753b57dc7c1SPaul Blakey hooknum = NF_INET_LOCAL_IN; /* Source NAT */ 754b57dc7c1SPaul Blakey else 755b57dc7c1SPaul Blakey hooknum = NF_INET_LOCAL_OUT; /* Destination NAT */ 756b57dc7c1SPaul Blakey 757b57dc7c1SPaul Blakey switch (ctinfo) { 758b57dc7c1SPaul Blakey case IP_CT_RELATED: 759b57dc7c1SPaul Blakey case IP_CT_RELATED_REPLY: 760b57dc7c1SPaul Blakey if (skb->protocol == htons(ETH_P_IP) && 761b57dc7c1SPaul Blakey ip_hdr(skb)->protocol == IPPROTO_ICMP) { 762b57dc7c1SPaul Blakey if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo, 763b57dc7c1SPaul Blakey hooknum)) 764b57dc7c1SPaul Blakey err = NF_DROP; 765b57dc7c1SPaul Blakey goto out; 766b57dc7c1SPaul Blakey } else if (IS_ENABLED(CONFIG_IPV6) && 767b57dc7c1SPaul Blakey skb->protocol == htons(ETH_P_IPV6)) { 768b57dc7c1SPaul Blakey __be16 frag_off; 769b57dc7c1SPaul Blakey u8 nexthdr = ipv6_hdr(skb)->nexthdr; 770b57dc7c1SPaul Blakey int hdrlen = ipv6_skip_exthdr(skb, 771b57dc7c1SPaul Blakey sizeof(struct ipv6hdr), 772b57dc7c1SPaul Blakey &nexthdr, &frag_off); 773b57dc7c1SPaul Blakey 774b57dc7c1SPaul Blakey if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) { 775b57dc7c1SPaul Blakey if (!nf_nat_icmpv6_reply_translation(skb, ct, 776b57dc7c1SPaul Blakey ctinfo, 777b57dc7c1SPaul Blakey hooknum, 778b57dc7c1SPaul Blakey hdrlen)) 779b57dc7c1SPaul Blakey err = NF_DROP; 780b57dc7c1SPaul Blakey goto out; 781b57dc7c1SPaul Blakey } 782b57dc7c1SPaul Blakey } 783b57dc7c1SPaul Blakey /* Non-ICMP, fall thru to initialize if needed. */ 784b57dc7c1SPaul Blakey /* fall through */ 785b57dc7c1SPaul Blakey case IP_CT_NEW: 786b57dc7c1SPaul Blakey /* Seen it before? This can happen for loopback, retrans, 787b57dc7c1SPaul Blakey * or local packets. 788b57dc7c1SPaul Blakey */ 789b57dc7c1SPaul Blakey if (!nf_nat_initialized(ct, maniptype)) { 790b57dc7c1SPaul Blakey /* Initialize according to the NAT action. */ 791b57dc7c1SPaul Blakey err = (range && range->flags & NF_NAT_RANGE_MAP_IPS) 792b57dc7c1SPaul Blakey /* Action is set up to establish a new 793b57dc7c1SPaul Blakey * mapping. 794b57dc7c1SPaul Blakey */ 795b57dc7c1SPaul Blakey ? nf_nat_setup_info(ct, range, maniptype) 796b57dc7c1SPaul Blakey : nf_nat_alloc_null_binding(ct, hooknum); 797b57dc7c1SPaul Blakey if (err != NF_ACCEPT) 798b57dc7c1SPaul Blakey goto out; 799b57dc7c1SPaul Blakey } 800b57dc7c1SPaul Blakey break; 801b57dc7c1SPaul Blakey 802b57dc7c1SPaul Blakey case IP_CT_ESTABLISHED: 803b57dc7c1SPaul Blakey case IP_CT_ESTABLISHED_REPLY: 804b57dc7c1SPaul Blakey break; 805b57dc7c1SPaul Blakey 806b57dc7c1SPaul Blakey default: 807b57dc7c1SPaul Blakey err = NF_DROP; 808b57dc7c1SPaul Blakey goto out; 809b57dc7c1SPaul Blakey } 810b57dc7c1SPaul Blakey 811b57dc7c1SPaul Blakey err = nf_nat_packet(ct, ctinfo, hooknum, skb); 812b57dc7c1SPaul Blakey out: 813b57dc7c1SPaul Blakey return err; 814b57dc7c1SPaul Blakey } 815b57dc7c1SPaul Blakey #endif /* CONFIG_NF_NAT */ 816b57dc7c1SPaul Blakey 817b57dc7c1SPaul Blakey static void tcf_ct_act_set_mark(struct nf_conn *ct, u32 mark, u32 mask) 818b57dc7c1SPaul Blakey { 819b57dc7c1SPaul Blakey #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) 820b57dc7c1SPaul Blakey u32 new_mark; 821b57dc7c1SPaul Blakey 822b57dc7c1SPaul Blakey if (!mask) 823b57dc7c1SPaul Blakey return; 824b57dc7c1SPaul Blakey 825b57dc7c1SPaul Blakey new_mark = mark | (ct->mark & ~(mask)); 826b57dc7c1SPaul Blakey if (ct->mark != new_mark) { 827b57dc7c1SPaul Blakey ct->mark = new_mark; 828b57dc7c1SPaul Blakey if (nf_ct_is_confirmed(ct)) 829b57dc7c1SPaul Blakey nf_conntrack_event_cache(IPCT_MARK, ct); 830b57dc7c1SPaul Blakey } 831b57dc7c1SPaul Blakey #endif 832b57dc7c1SPaul Blakey } 833b57dc7c1SPaul Blakey 834b57dc7c1SPaul Blakey static void tcf_ct_act_set_labels(struct nf_conn *ct, 835b57dc7c1SPaul Blakey u32 *labels, 836b57dc7c1SPaul Blakey u32 *labels_m) 837b57dc7c1SPaul Blakey { 838b57dc7c1SPaul Blakey #if IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) 839c593642cSPankaj Bharadiya size_t labels_sz = sizeof_field(struct tcf_ct_params, labels); 840b57dc7c1SPaul Blakey 841b57dc7c1SPaul Blakey if (!memchr_inv(labels_m, 0, labels_sz)) 842b57dc7c1SPaul Blakey return; 843b57dc7c1SPaul Blakey 844b57dc7c1SPaul Blakey nf_connlabels_replace(ct, labels, labels_m, 4); 845b57dc7c1SPaul Blakey #endif 846b57dc7c1SPaul Blakey } 847b57dc7c1SPaul Blakey 848b57dc7c1SPaul Blakey static int tcf_ct_act_nat(struct sk_buff *skb, 849b57dc7c1SPaul Blakey struct nf_conn *ct, 850b57dc7c1SPaul Blakey enum ip_conntrack_info ctinfo, 851b57dc7c1SPaul Blakey int ct_action, 852b57dc7c1SPaul Blakey struct nf_nat_range2 *range, 853b57dc7c1SPaul Blakey bool commit) 854b57dc7c1SPaul Blakey { 855b57dc7c1SPaul Blakey #if IS_ENABLED(CONFIG_NF_NAT) 85695219afbSAaron Conole int err; 857b57dc7c1SPaul Blakey enum nf_nat_manip_type maniptype; 858b57dc7c1SPaul Blakey 859b57dc7c1SPaul Blakey if (!(ct_action & TCA_CT_ACT_NAT)) 860b57dc7c1SPaul Blakey return NF_ACCEPT; 861b57dc7c1SPaul Blakey 862b57dc7c1SPaul Blakey /* Add NAT extension if not confirmed yet. */ 863b57dc7c1SPaul Blakey if (!nf_ct_is_confirmed(ct) && !nf_ct_nat_ext_add(ct)) 864b57dc7c1SPaul Blakey return NF_DROP; /* Can't NAT. */ 865b57dc7c1SPaul Blakey 866b57dc7c1SPaul Blakey if (ctinfo != IP_CT_NEW && (ct->status & IPS_NAT_MASK) && 867b57dc7c1SPaul Blakey (ctinfo != IP_CT_RELATED || commit)) { 868b57dc7c1SPaul Blakey /* NAT an established or related connection like before. */ 869b57dc7c1SPaul Blakey if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY) 870b57dc7c1SPaul Blakey /* This is the REPLY direction for a connection 871b57dc7c1SPaul Blakey * for which NAT was applied in the forward 872b57dc7c1SPaul Blakey * direction. Do the reverse NAT. 873b57dc7c1SPaul Blakey */ 874b57dc7c1SPaul Blakey maniptype = ct->status & IPS_SRC_NAT 875b57dc7c1SPaul Blakey ? NF_NAT_MANIP_DST : NF_NAT_MANIP_SRC; 876b57dc7c1SPaul Blakey else 877b57dc7c1SPaul Blakey maniptype = ct->status & IPS_SRC_NAT 878b57dc7c1SPaul Blakey ? NF_NAT_MANIP_SRC : NF_NAT_MANIP_DST; 879b57dc7c1SPaul Blakey } else if (ct_action & TCA_CT_ACT_NAT_SRC) { 880b57dc7c1SPaul Blakey maniptype = NF_NAT_MANIP_SRC; 881b57dc7c1SPaul Blakey } else if (ct_action & TCA_CT_ACT_NAT_DST) { 882b57dc7c1SPaul Blakey maniptype = NF_NAT_MANIP_DST; 883b57dc7c1SPaul Blakey } else { 884b57dc7c1SPaul Blakey return NF_ACCEPT; 885b57dc7c1SPaul Blakey } 886b57dc7c1SPaul Blakey 88795219afbSAaron Conole err = ct_nat_execute(skb, ct, ctinfo, range, maniptype); 88895219afbSAaron Conole if (err == NF_ACCEPT && 88995219afbSAaron Conole ct->status & IPS_SRC_NAT && ct->status & IPS_DST_NAT) { 89095219afbSAaron Conole if (maniptype == NF_NAT_MANIP_SRC) 89195219afbSAaron Conole maniptype = NF_NAT_MANIP_DST; 89295219afbSAaron Conole else 89395219afbSAaron Conole maniptype = NF_NAT_MANIP_SRC; 89495219afbSAaron Conole 89595219afbSAaron Conole err = ct_nat_execute(skb, ct, ctinfo, range, maniptype); 89695219afbSAaron Conole } 89795219afbSAaron Conole return err; 898b57dc7c1SPaul Blakey #else 899b57dc7c1SPaul Blakey return NF_ACCEPT; 900b57dc7c1SPaul Blakey #endif 901b57dc7c1SPaul Blakey } 902b57dc7c1SPaul Blakey 903b57dc7c1SPaul Blakey static int tcf_ct_act(struct sk_buff *skb, const struct tc_action *a, 904b57dc7c1SPaul Blakey struct tcf_result *res) 905b57dc7c1SPaul Blakey { 906b57dc7c1SPaul Blakey struct net *net = dev_net(skb->dev); 907b57dc7c1SPaul Blakey bool cached, commit, clear, force; 908b57dc7c1SPaul Blakey enum ip_conntrack_info ctinfo; 909b57dc7c1SPaul Blakey struct tcf_ct *c = to_ct(a); 910b57dc7c1SPaul Blakey struct nf_conn *tmpl = NULL; 911b57dc7c1SPaul Blakey struct nf_hook_state state; 912b57dc7c1SPaul Blakey int nh_ofs, err, retval; 913b57dc7c1SPaul Blakey struct tcf_ct_params *p; 91446475bb2SPaul Blakey bool skip_add = false; 915b57dc7c1SPaul Blakey struct nf_conn *ct; 916b57dc7c1SPaul Blakey u8 family; 917b57dc7c1SPaul Blakey 918b57dc7c1SPaul Blakey p = rcu_dereference_bh(c->params); 919b57dc7c1SPaul Blakey 920b57dc7c1SPaul Blakey retval = READ_ONCE(c->tcf_action); 921b57dc7c1SPaul Blakey commit = p->ct_action & TCA_CT_ACT_COMMIT; 922b57dc7c1SPaul Blakey clear = p->ct_action & TCA_CT_ACT_CLEAR; 923b57dc7c1SPaul Blakey force = p->ct_action & TCA_CT_ACT_FORCE; 924b57dc7c1SPaul Blakey tmpl = p->tmpl; 925b57dc7c1SPaul Blakey 926b57dc7c1SPaul Blakey if (clear) { 927b57dc7c1SPaul Blakey ct = nf_ct_get(skb, &ctinfo); 928b57dc7c1SPaul Blakey if (ct) { 929b57dc7c1SPaul Blakey nf_conntrack_put(&ct->ct_general); 930b57dc7c1SPaul Blakey nf_ct_set(skb, NULL, IP_CT_UNTRACKED); 931b57dc7c1SPaul Blakey } 932b57dc7c1SPaul Blakey 933b57dc7c1SPaul Blakey goto out; 934b57dc7c1SPaul Blakey } 935b57dc7c1SPaul Blakey 936b57dc7c1SPaul Blakey family = tcf_ct_skb_nf_family(skb); 937b57dc7c1SPaul Blakey if (family == NFPROTO_UNSPEC) 938b57dc7c1SPaul Blakey goto drop; 939b57dc7c1SPaul Blakey 940b57dc7c1SPaul Blakey /* The conntrack module expects to be working at L3. 941b57dc7c1SPaul Blakey * We also try to pull the IPv4/6 header to linear area 942b57dc7c1SPaul Blakey */ 943b57dc7c1SPaul Blakey nh_ofs = skb_network_offset(skb); 944b57dc7c1SPaul Blakey skb_pull_rcsum(skb, nh_ofs); 945b57dc7c1SPaul Blakey err = tcf_ct_handle_fragments(net, skb, family, p->zone); 946b57dc7c1SPaul Blakey if (err == -EINPROGRESS) { 947b57dc7c1SPaul Blakey retval = TC_ACT_STOLEN; 948b57dc7c1SPaul Blakey goto out; 949b57dc7c1SPaul Blakey } 950b57dc7c1SPaul Blakey if (err) 951b57dc7c1SPaul Blakey goto drop; 952b57dc7c1SPaul Blakey 953b57dc7c1SPaul Blakey err = tcf_ct_skb_network_trim(skb, family); 954b57dc7c1SPaul Blakey if (err) 955b57dc7c1SPaul Blakey goto drop; 956b57dc7c1SPaul Blakey 957b57dc7c1SPaul Blakey /* If we are recirculating packets to match on ct fields and 958b57dc7c1SPaul Blakey * committing with a separate ct action, then we don't need to 959b57dc7c1SPaul Blakey * actually run the packet through conntrack twice unless it's for a 960b57dc7c1SPaul Blakey * different zone. 961b57dc7c1SPaul Blakey */ 962b57dc7c1SPaul Blakey cached = tcf_ct_skb_nfct_cached(net, skb, p->zone, force); 963b57dc7c1SPaul Blakey if (!cached) { 96446475bb2SPaul Blakey if (!commit && tcf_ct_flow_table_lookup(p, skb, family)) { 96546475bb2SPaul Blakey skip_add = true; 96646475bb2SPaul Blakey goto do_nat; 96746475bb2SPaul Blakey } 96846475bb2SPaul Blakey 969b57dc7c1SPaul Blakey /* Associate skb with specified zone. */ 970b57dc7c1SPaul Blakey if (tmpl) { 971b57dc7c1SPaul Blakey ct = nf_ct_get(skb, &ctinfo); 972b57dc7c1SPaul Blakey if (skb_nfct(skb)) 973b57dc7c1SPaul Blakey nf_conntrack_put(skb_nfct(skb)); 974b57dc7c1SPaul Blakey nf_conntrack_get(&tmpl->ct_general); 975b57dc7c1SPaul Blakey nf_ct_set(skb, tmpl, IP_CT_NEW); 976b57dc7c1SPaul Blakey } 977b57dc7c1SPaul Blakey 978b57dc7c1SPaul Blakey state.hook = NF_INET_PRE_ROUTING; 979b57dc7c1SPaul Blakey state.net = net; 980b57dc7c1SPaul Blakey state.pf = family; 981b57dc7c1SPaul Blakey err = nf_conntrack_in(skb, &state); 982b57dc7c1SPaul Blakey if (err != NF_ACCEPT) 983b57dc7c1SPaul Blakey goto out_push; 984b57dc7c1SPaul Blakey } 985b57dc7c1SPaul Blakey 98646475bb2SPaul Blakey do_nat: 987b57dc7c1SPaul Blakey ct = nf_ct_get(skb, &ctinfo); 988b57dc7c1SPaul Blakey if (!ct) 989b57dc7c1SPaul Blakey goto out_push; 990b57dc7c1SPaul Blakey nf_ct_deliver_cached_events(ct); 991b57dc7c1SPaul Blakey 992b57dc7c1SPaul Blakey err = tcf_ct_act_nat(skb, ct, ctinfo, p->ct_action, &p->range, commit); 993b57dc7c1SPaul Blakey if (err != NF_ACCEPT) 994b57dc7c1SPaul Blakey goto drop; 995b57dc7c1SPaul Blakey 996b57dc7c1SPaul Blakey if (commit) { 997b57dc7c1SPaul Blakey tcf_ct_act_set_mark(ct, p->mark, p->mark_mask); 998b57dc7c1SPaul Blakey tcf_ct_act_set_labels(ct, p->labels, p->labels_mask); 999b57dc7c1SPaul Blakey 1000b57dc7c1SPaul Blakey /* This will take care of sending queued events 1001b57dc7c1SPaul Blakey * even if the connection is already confirmed. 1002b57dc7c1SPaul Blakey */ 1003b57dc7c1SPaul Blakey nf_conntrack_confirm(skb); 100446475bb2SPaul Blakey } else if (!skip_add) { 100564ff70b8SPaul Blakey tcf_ct_flow_table_process_conn(p->ct_ft, ct, ctinfo); 100646475bb2SPaul Blakey } 100764ff70b8SPaul Blakey 1008b57dc7c1SPaul Blakey out_push: 1009b57dc7c1SPaul Blakey skb_push_rcsum(skb, nh_ofs); 1010b57dc7c1SPaul Blakey 1011b57dc7c1SPaul Blakey out: 10125e1ad95bSVlad Buslov tcf_action_update_bstats(&c->common, skb); 1013b57dc7c1SPaul Blakey return retval; 1014b57dc7c1SPaul Blakey 1015b57dc7c1SPaul Blakey drop: 101626b537a8SVlad Buslov tcf_action_inc_drop_qstats(&c->common); 1017b57dc7c1SPaul Blakey return TC_ACT_SHOT; 1018b57dc7c1SPaul Blakey } 1019b57dc7c1SPaul Blakey 1020b57dc7c1SPaul Blakey static const struct nla_policy ct_policy[TCA_CT_MAX + 1] = { 1021b57dc7c1SPaul Blakey [TCA_CT_ACTION] = { .type = NLA_U16 }, 1022b57dc7c1SPaul Blakey [TCA_CT_PARMS] = { .type = NLA_EXACT_LEN, .len = sizeof(struct tc_ct) }, 1023b57dc7c1SPaul Blakey [TCA_CT_ZONE] = { .type = NLA_U16 }, 1024b57dc7c1SPaul Blakey [TCA_CT_MARK] = { .type = NLA_U32 }, 1025b57dc7c1SPaul Blakey [TCA_CT_MARK_MASK] = { .type = NLA_U32 }, 1026b57dc7c1SPaul Blakey [TCA_CT_LABELS] = { .type = NLA_BINARY, 1027b57dc7c1SPaul Blakey .len = 128 / BITS_PER_BYTE }, 1028b57dc7c1SPaul Blakey [TCA_CT_LABELS_MASK] = { .type = NLA_BINARY, 1029b57dc7c1SPaul Blakey .len = 128 / BITS_PER_BYTE }, 1030b57dc7c1SPaul Blakey [TCA_CT_NAT_IPV4_MIN] = { .type = NLA_U32 }, 1031b57dc7c1SPaul Blakey [TCA_CT_NAT_IPV4_MAX] = { .type = NLA_U32 }, 1032b57dc7c1SPaul Blakey [TCA_CT_NAT_IPV6_MIN] = { .type = NLA_EXACT_LEN, 1033b57dc7c1SPaul Blakey .len = sizeof(struct in6_addr) }, 1034b57dc7c1SPaul Blakey [TCA_CT_NAT_IPV6_MAX] = { .type = NLA_EXACT_LEN, 1035b57dc7c1SPaul Blakey .len = sizeof(struct in6_addr) }, 1036b57dc7c1SPaul Blakey [TCA_CT_NAT_PORT_MIN] = { .type = NLA_U16 }, 1037b57dc7c1SPaul Blakey [TCA_CT_NAT_PORT_MAX] = { .type = NLA_U16 }, 1038b57dc7c1SPaul Blakey }; 1039b57dc7c1SPaul Blakey 1040b57dc7c1SPaul Blakey static int tcf_ct_fill_params_nat(struct tcf_ct_params *p, 1041b57dc7c1SPaul Blakey struct tc_ct *parm, 1042b57dc7c1SPaul Blakey struct nlattr **tb, 1043b57dc7c1SPaul Blakey struct netlink_ext_ack *extack) 1044b57dc7c1SPaul Blakey { 1045b57dc7c1SPaul Blakey struct nf_nat_range2 *range; 1046b57dc7c1SPaul Blakey 1047b57dc7c1SPaul Blakey if (!(p->ct_action & TCA_CT_ACT_NAT)) 1048b57dc7c1SPaul Blakey return 0; 1049b57dc7c1SPaul Blakey 1050b57dc7c1SPaul Blakey if (!IS_ENABLED(CONFIG_NF_NAT)) { 1051b57dc7c1SPaul Blakey NL_SET_ERR_MSG_MOD(extack, "Netfilter nat isn't enabled in kernel"); 1052b57dc7c1SPaul Blakey return -EOPNOTSUPP; 1053b57dc7c1SPaul Blakey } 1054b57dc7c1SPaul Blakey 1055b57dc7c1SPaul Blakey if (!(p->ct_action & (TCA_CT_ACT_NAT_SRC | TCA_CT_ACT_NAT_DST))) 1056b57dc7c1SPaul Blakey return 0; 1057b57dc7c1SPaul Blakey 1058b57dc7c1SPaul Blakey if ((p->ct_action & TCA_CT_ACT_NAT_SRC) && 1059b57dc7c1SPaul Blakey (p->ct_action & TCA_CT_ACT_NAT_DST)) { 1060b57dc7c1SPaul Blakey NL_SET_ERR_MSG_MOD(extack, "dnat and snat can't be enabled at the same time"); 1061b57dc7c1SPaul Blakey return -EOPNOTSUPP; 1062b57dc7c1SPaul Blakey } 1063b57dc7c1SPaul Blakey 1064b57dc7c1SPaul Blakey range = &p->range; 1065b57dc7c1SPaul Blakey if (tb[TCA_CT_NAT_IPV4_MIN]) { 1066b57dc7c1SPaul Blakey struct nlattr *max_attr = tb[TCA_CT_NAT_IPV4_MAX]; 1067b57dc7c1SPaul Blakey 1068b57dc7c1SPaul Blakey p->ipv4_range = true; 1069b57dc7c1SPaul Blakey range->flags |= NF_NAT_RANGE_MAP_IPS; 1070b57dc7c1SPaul Blakey range->min_addr.ip = 1071b57dc7c1SPaul Blakey nla_get_in_addr(tb[TCA_CT_NAT_IPV4_MIN]); 1072b57dc7c1SPaul Blakey 1073b57dc7c1SPaul Blakey range->max_addr.ip = max_attr ? 1074b57dc7c1SPaul Blakey nla_get_in_addr(max_attr) : 1075b57dc7c1SPaul Blakey range->min_addr.ip; 1076b57dc7c1SPaul Blakey } else if (tb[TCA_CT_NAT_IPV6_MIN]) { 1077b57dc7c1SPaul Blakey struct nlattr *max_attr = tb[TCA_CT_NAT_IPV6_MAX]; 1078b57dc7c1SPaul Blakey 1079b57dc7c1SPaul Blakey p->ipv4_range = false; 1080b57dc7c1SPaul Blakey range->flags |= NF_NAT_RANGE_MAP_IPS; 1081b57dc7c1SPaul Blakey range->min_addr.in6 = 1082b57dc7c1SPaul Blakey nla_get_in6_addr(tb[TCA_CT_NAT_IPV6_MIN]); 1083b57dc7c1SPaul Blakey 1084b57dc7c1SPaul Blakey range->max_addr.in6 = max_attr ? 1085b57dc7c1SPaul Blakey nla_get_in6_addr(max_attr) : 1086b57dc7c1SPaul Blakey range->min_addr.in6; 1087b57dc7c1SPaul Blakey } 1088b57dc7c1SPaul Blakey 1089b57dc7c1SPaul Blakey if (tb[TCA_CT_NAT_PORT_MIN]) { 1090b57dc7c1SPaul Blakey range->flags |= NF_NAT_RANGE_PROTO_SPECIFIED; 1091b57dc7c1SPaul Blakey range->min_proto.all = nla_get_be16(tb[TCA_CT_NAT_PORT_MIN]); 1092b57dc7c1SPaul Blakey 1093b57dc7c1SPaul Blakey range->max_proto.all = tb[TCA_CT_NAT_PORT_MAX] ? 1094b57dc7c1SPaul Blakey nla_get_be16(tb[TCA_CT_NAT_PORT_MAX]) : 1095b57dc7c1SPaul Blakey range->min_proto.all; 1096b57dc7c1SPaul Blakey } 1097b57dc7c1SPaul Blakey 1098b57dc7c1SPaul Blakey return 0; 1099b57dc7c1SPaul Blakey } 1100b57dc7c1SPaul Blakey 1101b57dc7c1SPaul Blakey static void tcf_ct_set_key_val(struct nlattr **tb, 1102b57dc7c1SPaul Blakey void *val, int val_type, 1103b57dc7c1SPaul Blakey void *mask, int mask_type, 1104b57dc7c1SPaul Blakey int len) 1105b57dc7c1SPaul Blakey { 1106b57dc7c1SPaul Blakey if (!tb[val_type]) 1107b57dc7c1SPaul Blakey return; 1108b57dc7c1SPaul Blakey nla_memcpy(val, tb[val_type], len); 1109b57dc7c1SPaul Blakey 1110b57dc7c1SPaul Blakey if (!mask) 1111b57dc7c1SPaul Blakey return; 1112b57dc7c1SPaul Blakey 1113b57dc7c1SPaul Blakey if (mask_type == TCA_CT_UNSPEC || !tb[mask_type]) 1114b57dc7c1SPaul Blakey memset(mask, 0xff, len); 1115b57dc7c1SPaul Blakey else 1116b57dc7c1SPaul Blakey nla_memcpy(mask, tb[mask_type], len); 1117b57dc7c1SPaul Blakey } 1118b57dc7c1SPaul Blakey 1119b57dc7c1SPaul Blakey static int tcf_ct_fill_params(struct net *net, 1120b57dc7c1SPaul Blakey struct tcf_ct_params *p, 1121b57dc7c1SPaul Blakey struct tc_ct *parm, 1122b57dc7c1SPaul Blakey struct nlattr **tb, 1123b57dc7c1SPaul Blakey struct netlink_ext_ack *extack) 1124b57dc7c1SPaul Blakey { 1125b57dc7c1SPaul Blakey struct tc_ct_action_net *tn = net_generic(net, ct_net_id); 1126b57dc7c1SPaul Blakey struct nf_conntrack_zone zone; 1127b57dc7c1SPaul Blakey struct nf_conn *tmpl; 1128b57dc7c1SPaul Blakey int err; 1129b57dc7c1SPaul Blakey 1130b57dc7c1SPaul Blakey p->zone = NF_CT_DEFAULT_ZONE_ID; 1131b57dc7c1SPaul Blakey 1132b57dc7c1SPaul Blakey tcf_ct_set_key_val(tb, 1133b57dc7c1SPaul Blakey &p->ct_action, TCA_CT_ACTION, 1134b57dc7c1SPaul Blakey NULL, TCA_CT_UNSPEC, 1135b57dc7c1SPaul Blakey sizeof(p->ct_action)); 1136b57dc7c1SPaul Blakey 1137b57dc7c1SPaul Blakey if (p->ct_action & TCA_CT_ACT_CLEAR) 1138b57dc7c1SPaul Blakey return 0; 1139b57dc7c1SPaul Blakey 1140b57dc7c1SPaul Blakey err = tcf_ct_fill_params_nat(p, parm, tb, extack); 1141b57dc7c1SPaul Blakey if (err) 1142b57dc7c1SPaul Blakey return err; 1143b57dc7c1SPaul Blakey 1144b57dc7c1SPaul Blakey if (tb[TCA_CT_MARK]) { 1145b57dc7c1SPaul Blakey if (!IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)) { 1146b57dc7c1SPaul Blakey NL_SET_ERR_MSG_MOD(extack, "Conntrack mark isn't enabled."); 1147b57dc7c1SPaul Blakey return -EOPNOTSUPP; 1148b57dc7c1SPaul Blakey } 1149b57dc7c1SPaul Blakey tcf_ct_set_key_val(tb, 1150b57dc7c1SPaul Blakey &p->mark, TCA_CT_MARK, 1151b57dc7c1SPaul Blakey &p->mark_mask, TCA_CT_MARK_MASK, 1152b57dc7c1SPaul Blakey sizeof(p->mark)); 1153b57dc7c1SPaul Blakey } 1154b57dc7c1SPaul Blakey 1155b57dc7c1SPaul Blakey if (tb[TCA_CT_LABELS]) { 1156b57dc7c1SPaul Blakey if (!IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS)) { 1157b57dc7c1SPaul Blakey NL_SET_ERR_MSG_MOD(extack, "Conntrack labels isn't enabled."); 1158b57dc7c1SPaul Blakey return -EOPNOTSUPP; 1159b57dc7c1SPaul Blakey } 1160b57dc7c1SPaul Blakey 1161b57dc7c1SPaul Blakey if (!tn->labels) { 1162b57dc7c1SPaul Blakey NL_SET_ERR_MSG_MOD(extack, "Failed to set connlabel length"); 1163b57dc7c1SPaul Blakey return -EOPNOTSUPP; 1164b57dc7c1SPaul Blakey } 1165b57dc7c1SPaul Blakey tcf_ct_set_key_val(tb, 1166b57dc7c1SPaul Blakey p->labels, TCA_CT_LABELS, 1167b57dc7c1SPaul Blakey p->labels_mask, TCA_CT_LABELS_MASK, 1168b57dc7c1SPaul Blakey sizeof(p->labels)); 1169b57dc7c1SPaul Blakey } 1170b57dc7c1SPaul Blakey 1171b57dc7c1SPaul Blakey if (tb[TCA_CT_ZONE]) { 1172b57dc7c1SPaul Blakey if (!IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)) { 1173b57dc7c1SPaul Blakey NL_SET_ERR_MSG_MOD(extack, "Conntrack zones isn't enabled."); 1174b57dc7c1SPaul Blakey return -EOPNOTSUPP; 1175b57dc7c1SPaul Blakey } 1176b57dc7c1SPaul Blakey 1177b57dc7c1SPaul Blakey tcf_ct_set_key_val(tb, 1178b57dc7c1SPaul Blakey &p->zone, TCA_CT_ZONE, 1179b57dc7c1SPaul Blakey NULL, TCA_CT_UNSPEC, 1180b57dc7c1SPaul Blakey sizeof(p->zone)); 1181b57dc7c1SPaul Blakey } 1182b57dc7c1SPaul Blakey 1183b57dc7c1SPaul Blakey if (p->zone == NF_CT_DEFAULT_ZONE_ID) 1184b57dc7c1SPaul Blakey return 0; 1185b57dc7c1SPaul Blakey 1186b57dc7c1SPaul Blakey nf_ct_zone_init(&zone, p->zone, NF_CT_DEFAULT_ZONE_DIR, 0); 1187b57dc7c1SPaul Blakey tmpl = nf_ct_tmpl_alloc(net, &zone, GFP_KERNEL); 1188b57dc7c1SPaul Blakey if (!tmpl) { 1189b57dc7c1SPaul Blakey NL_SET_ERR_MSG_MOD(extack, "Failed to allocate conntrack template"); 1190b57dc7c1SPaul Blakey return -ENOMEM; 1191b57dc7c1SPaul Blakey } 1192b57dc7c1SPaul Blakey __set_bit(IPS_CONFIRMED_BIT, &tmpl->status); 1193b57dc7c1SPaul Blakey nf_conntrack_get(&tmpl->ct_general); 1194b57dc7c1SPaul Blakey p->tmpl = tmpl; 1195b57dc7c1SPaul Blakey 1196b57dc7c1SPaul Blakey return 0; 1197b57dc7c1SPaul Blakey } 1198b57dc7c1SPaul Blakey 1199b57dc7c1SPaul Blakey static int tcf_ct_init(struct net *net, struct nlattr *nla, 1200b57dc7c1SPaul Blakey struct nlattr *est, struct tc_action **a, 1201b57dc7c1SPaul Blakey int replace, int bind, bool rtnl_held, 1202abbb0d33SVlad Buslov struct tcf_proto *tp, u32 flags, 1203b57dc7c1SPaul Blakey struct netlink_ext_ack *extack) 1204b57dc7c1SPaul Blakey { 1205b57dc7c1SPaul Blakey struct tc_action_net *tn = net_generic(net, ct_net_id); 1206b57dc7c1SPaul Blakey struct tcf_ct_params *params = NULL; 1207b57dc7c1SPaul Blakey struct nlattr *tb[TCA_CT_MAX + 1]; 1208b57dc7c1SPaul Blakey struct tcf_chain *goto_ch = NULL; 1209b57dc7c1SPaul Blakey struct tc_ct *parm; 1210b57dc7c1SPaul Blakey struct tcf_ct *c; 1211b57dc7c1SPaul Blakey int err, res = 0; 12127be8ef2cSDmytro Linkin u32 index; 1213b57dc7c1SPaul Blakey 1214b57dc7c1SPaul Blakey if (!nla) { 1215b57dc7c1SPaul Blakey NL_SET_ERR_MSG_MOD(extack, "Ct requires attributes to be passed"); 1216b57dc7c1SPaul Blakey return -EINVAL; 1217b57dc7c1SPaul Blakey } 1218b57dc7c1SPaul Blakey 1219b57dc7c1SPaul Blakey err = nla_parse_nested(tb, TCA_CT_MAX, nla, ct_policy, extack); 1220b57dc7c1SPaul Blakey if (err < 0) 1221b57dc7c1SPaul Blakey return err; 1222b57dc7c1SPaul Blakey 1223b57dc7c1SPaul Blakey if (!tb[TCA_CT_PARMS]) { 1224b57dc7c1SPaul Blakey NL_SET_ERR_MSG_MOD(extack, "Missing required ct parameters"); 1225b57dc7c1SPaul Blakey return -EINVAL; 1226b57dc7c1SPaul Blakey } 1227b57dc7c1SPaul Blakey parm = nla_data(tb[TCA_CT_PARMS]); 12287be8ef2cSDmytro Linkin index = parm->index; 12297be8ef2cSDmytro Linkin err = tcf_idr_check_alloc(tn, &index, a, bind); 1230b57dc7c1SPaul Blakey if (err < 0) 1231b57dc7c1SPaul Blakey return err; 1232b57dc7c1SPaul Blakey 1233b57dc7c1SPaul Blakey if (!err) { 1234e3822678SVlad Buslov err = tcf_idr_create_from_flags(tn, index, est, a, 1235e3822678SVlad Buslov &act_ct_ops, bind, flags); 1236b57dc7c1SPaul Blakey if (err) { 12377be8ef2cSDmytro Linkin tcf_idr_cleanup(tn, index); 1238b57dc7c1SPaul Blakey return err; 1239b57dc7c1SPaul Blakey } 1240b57dc7c1SPaul Blakey res = ACT_P_CREATED; 1241b57dc7c1SPaul Blakey } else { 1242b57dc7c1SPaul Blakey if (bind) 1243b57dc7c1SPaul Blakey return 0; 1244b57dc7c1SPaul Blakey 1245b57dc7c1SPaul Blakey if (!replace) { 1246b57dc7c1SPaul Blakey tcf_idr_release(*a, bind); 1247b57dc7c1SPaul Blakey return -EEXIST; 1248b57dc7c1SPaul Blakey } 1249b57dc7c1SPaul Blakey } 1250b57dc7c1SPaul Blakey err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack); 1251b57dc7c1SPaul Blakey if (err < 0) 1252b57dc7c1SPaul Blakey goto cleanup; 1253b57dc7c1SPaul Blakey 1254b57dc7c1SPaul Blakey c = to_ct(*a); 1255b57dc7c1SPaul Blakey 1256b57dc7c1SPaul Blakey params = kzalloc(sizeof(*params), GFP_KERNEL); 1257b57dc7c1SPaul Blakey if (unlikely(!params)) { 1258b57dc7c1SPaul Blakey err = -ENOMEM; 1259b57dc7c1SPaul Blakey goto cleanup; 1260b57dc7c1SPaul Blakey } 1261b57dc7c1SPaul Blakey 1262b57dc7c1SPaul Blakey err = tcf_ct_fill_params(net, params, parm, tb, extack); 1263b57dc7c1SPaul Blakey if (err) 1264b57dc7c1SPaul Blakey goto cleanup; 1265b57dc7c1SPaul Blakey 1266c34b961aSPaul Blakey err = tcf_ct_flow_table_get(params); 1267c34b961aSPaul Blakey if (err) 1268c34b961aSPaul Blakey goto cleanup; 1269c34b961aSPaul Blakey 1270b57dc7c1SPaul Blakey spin_lock_bh(&c->tcf_lock); 1271b57dc7c1SPaul Blakey goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch); 1272445d3749SPaul E. McKenney params = rcu_replace_pointer(c->params, params, 1273445d3749SPaul E. McKenney lockdep_is_held(&c->tcf_lock)); 1274b57dc7c1SPaul Blakey spin_unlock_bh(&c->tcf_lock); 1275b57dc7c1SPaul Blakey 1276b57dc7c1SPaul Blakey if (goto_ch) 1277b57dc7c1SPaul Blakey tcf_chain_put_by_act(goto_ch); 1278b57dc7c1SPaul Blakey if (params) 1279dd2af104SPaul Blakey call_rcu(¶ms->rcu, tcf_ct_params_free); 1280b57dc7c1SPaul Blakey if (res == ACT_P_CREATED) 1281b57dc7c1SPaul Blakey tcf_idr_insert(tn, *a); 1282b57dc7c1SPaul Blakey 1283b57dc7c1SPaul Blakey return res; 1284b57dc7c1SPaul Blakey 1285b57dc7c1SPaul Blakey cleanup: 1286b57dc7c1SPaul Blakey if (goto_ch) 1287b57dc7c1SPaul Blakey tcf_chain_put_by_act(goto_ch); 1288b57dc7c1SPaul Blakey kfree(params); 1289b57dc7c1SPaul Blakey tcf_idr_release(*a, bind); 1290b57dc7c1SPaul Blakey return err; 1291b57dc7c1SPaul Blakey } 1292b57dc7c1SPaul Blakey 1293b57dc7c1SPaul Blakey static void tcf_ct_cleanup(struct tc_action *a) 1294b57dc7c1SPaul Blakey { 1295b57dc7c1SPaul Blakey struct tcf_ct_params *params; 1296b57dc7c1SPaul Blakey struct tcf_ct *c = to_ct(a); 1297b57dc7c1SPaul Blakey 1298b57dc7c1SPaul Blakey params = rcu_dereference_protected(c->params, 1); 1299b57dc7c1SPaul Blakey if (params) 1300b57dc7c1SPaul Blakey call_rcu(¶ms->rcu, tcf_ct_params_free); 1301b57dc7c1SPaul Blakey } 1302b57dc7c1SPaul Blakey 1303b57dc7c1SPaul Blakey static int tcf_ct_dump_key_val(struct sk_buff *skb, 1304b57dc7c1SPaul Blakey void *val, int val_type, 1305b57dc7c1SPaul Blakey void *mask, int mask_type, 1306b57dc7c1SPaul Blakey int len) 1307b57dc7c1SPaul Blakey { 1308b57dc7c1SPaul Blakey int err; 1309b57dc7c1SPaul Blakey 1310b57dc7c1SPaul Blakey if (mask && !memchr_inv(mask, 0, len)) 1311b57dc7c1SPaul Blakey return 0; 1312b57dc7c1SPaul Blakey 1313b57dc7c1SPaul Blakey err = nla_put(skb, val_type, len, val); 1314b57dc7c1SPaul Blakey if (err) 1315b57dc7c1SPaul Blakey return err; 1316b57dc7c1SPaul Blakey 1317b57dc7c1SPaul Blakey if (mask_type != TCA_CT_UNSPEC) { 1318b57dc7c1SPaul Blakey err = nla_put(skb, mask_type, len, mask); 1319b57dc7c1SPaul Blakey if (err) 1320b57dc7c1SPaul Blakey return err; 1321b57dc7c1SPaul Blakey } 1322b57dc7c1SPaul Blakey 1323b57dc7c1SPaul Blakey return 0; 1324b57dc7c1SPaul Blakey } 1325b57dc7c1SPaul Blakey 1326b57dc7c1SPaul Blakey static int tcf_ct_dump_nat(struct sk_buff *skb, struct tcf_ct_params *p) 1327b57dc7c1SPaul Blakey { 1328b57dc7c1SPaul Blakey struct nf_nat_range2 *range = &p->range; 1329b57dc7c1SPaul Blakey 1330b57dc7c1SPaul Blakey if (!(p->ct_action & TCA_CT_ACT_NAT)) 1331b57dc7c1SPaul Blakey return 0; 1332b57dc7c1SPaul Blakey 1333b57dc7c1SPaul Blakey if (!(p->ct_action & (TCA_CT_ACT_NAT_SRC | TCA_CT_ACT_NAT_DST))) 1334b57dc7c1SPaul Blakey return 0; 1335b57dc7c1SPaul Blakey 1336b57dc7c1SPaul Blakey if (range->flags & NF_NAT_RANGE_MAP_IPS) { 1337b57dc7c1SPaul Blakey if (p->ipv4_range) { 1338b57dc7c1SPaul Blakey if (nla_put_in_addr(skb, TCA_CT_NAT_IPV4_MIN, 1339b57dc7c1SPaul Blakey range->min_addr.ip)) 1340b57dc7c1SPaul Blakey return -1; 1341b57dc7c1SPaul Blakey if (nla_put_in_addr(skb, TCA_CT_NAT_IPV4_MAX, 1342b57dc7c1SPaul Blakey range->max_addr.ip)) 1343b57dc7c1SPaul Blakey return -1; 1344b57dc7c1SPaul Blakey } else { 1345b57dc7c1SPaul Blakey if (nla_put_in6_addr(skb, TCA_CT_NAT_IPV6_MIN, 1346b57dc7c1SPaul Blakey &range->min_addr.in6)) 1347b57dc7c1SPaul Blakey return -1; 1348b57dc7c1SPaul Blakey if (nla_put_in6_addr(skb, TCA_CT_NAT_IPV6_MAX, 1349b57dc7c1SPaul Blakey &range->max_addr.in6)) 1350b57dc7c1SPaul Blakey return -1; 1351b57dc7c1SPaul Blakey } 1352b57dc7c1SPaul Blakey } 1353b57dc7c1SPaul Blakey 1354b57dc7c1SPaul Blakey if (range->flags & NF_NAT_RANGE_PROTO_SPECIFIED) { 1355b57dc7c1SPaul Blakey if (nla_put_be16(skb, TCA_CT_NAT_PORT_MIN, 1356b57dc7c1SPaul Blakey range->min_proto.all)) 1357b57dc7c1SPaul Blakey return -1; 1358b57dc7c1SPaul Blakey if (nla_put_be16(skb, TCA_CT_NAT_PORT_MAX, 1359b57dc7c1SPaul Blakey range->max_proto.all)) 1360b57dc7c1SPaul Blakey return -1; 1361b57dc7c1SPaul Blakey } 1362b57dc7c1SPaul Blakey 1363b57dc7c1SPaul Blakey return 0; 1364b57dc7c1SPaul Blakey } 1365b57dc7c1SPaul Blakey 1366b57dc7c1SPaul Blakey static inline int tcf_ct_dump(struct sk_buff *skb, struct tc_action *a, 1367b57dc7c1SPaul Blakey int bind, int ref) 1368b57dc7c1SPaul Blakey { 1369b57dc7c1SPaul Blakey unsigned char *b = skb_tail_pointer(skb); 1370b57dc7c1SPaul Blakey struct tcf_ct *c = to_ct(a); 1371b57dc7c1SPaul Blakey struct tcf_ct_params *p; 1372b57dc7c1SPaul Blakey 1373b57dc7c1SPaul Blakey struct tc_ct opt = { 1374b57dc7c1SPaul Blakey .index = c->tcf_index, 1375b57dc7c1SPaul Blakey .refcnt = refcount_read(&c->tcf_refcnt) - ref, 1376b57dc7c1SPaul Blakey .bindcnt = atomic_read(&c->tcf_bindcnt) - bind, 1377b57dc7c1SPaul Blakey }; 1378b57dc7c1SPaul Blakey struct tcf_t t; 1379b57dc7c1SPaul Blakey 1380b57dc7c1SPaul Blakey spin_lock_bh(&c->tcf_lock); 1381b57dc7c1SPaul Blakey p = rcu_dereference_protected(c->params, 1382b57dc7c1SPaul Blakey lockdep_is_held(&c->tcf_lock)); 1383b57dc7c1SPaul Blakey opt.action = c->tcf_action; 1384b57dc7c1SPaul Blakey 1385b57dc7c1SPaul Blakey if (tcf_ct_dump_key_val(skb, 1386b57dc7c1SPaul Blakey &p->ct_action, TCA_CT_ACTION, 1387b57dc7c1SPaul Blakey NULL, TCA_CT_UNSPEC, 1388b57dc7c1SPaul Blakey sizeof(p->ct_action))) 1389b57dc7c1SPaul Blakey goto nla_put_failure; 1390b57dc7c1SPaul Blakey 1391b57dc7c1SPaul Blakey if (p->ct_action & TCA_CT_ACT_CLEAR) 1392b57dc7c1SPaul Blakey goto skip_dump; 1393b57dc7c1SPaul Blakey 1394b57dc7c1SPaul Blakey if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && 1395b57dc7c1SPaul Blakey tcf_ct_dump_key_val(skb, 1396b57dc7c1SPaul Blakey &p->mark, TCA_CT_MARK, 1397b57dc7c1SPaul Blakey &p->mark_mask, TCA_CT_MARK_MASK, 1398b57dc7c1SPaul Blakey sizeof(p->mark))) 1399b57dc7c1SPaul Blakey goto nla_put_failure; 1400b57dc7c1SPaul Blakey 1401b57dc7c1SPaul Blakey if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) && 1402b57dc7c1SPaul Blakey tcf_ct_dump_key_val(skb, 1403b57dc7c1SPaul Blakey p->labels, TCA_CT_LABELS, 1404b57dc7c1SPaul Blakey p->labels_mask, TCA_CT_LABELS_MASK, 1405b57dc7c1SPaul Blakey sizeof(p->labels))) 1406b57dc7c1SPaul Blakey goto nla_put_failure; 1407b57dc7c1SPaul Blakey 1408b57dc7c1SPaul Blakey if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) && 1409b57dc7c1SPaul Blakey tcf_ct_dump_key_val(skb, 1410b57dc7c1SPaul Blakey &p->zone, TCA_CT_ZONE, 1411b57dc7c1SPaul Blakey NULL, TCA_CT_UNSPEC, 1412b57dc7c1SPaul Blakey sizeof(p->zone))) 1413b57dc7c1SPaul Blakey goto nla_put_failure; 1414b57dc7c1SPaul Blakey 1415b57dc7c1SPaul Blakey if (tcf_ct_dump_nat(skb, p)) 1416b57dc7c1SPaul Blakey goto nla_put_failure; 1417b57dc7c1SPaul Blakey 1418b57dc7c1SPaul Blakey skip_dump: 1419b57dc7c1SPaul Blakey if (nla_put(skb, TCA_CT_PARMS, sizeof(opt), &opt)) 1420b57dc7c1SPaul Blakey goto nla_put_failure; 1421b57dc7c1SPaul Blakey 1422b57dc7c1SPaul Blakey tcf_tm_dump(&t, &c->tcf_tm); 1423b57dc7c1SPaul Blakey if (nla_put_64bit(skb, TCA_CT_TM, sizeof(t), &t, TCA_CT_PAD)) 1424b57dc7c1SPaul Blakey goto nla_put_failure; 1425b57dc7c1SPaul Blakey spin_unlock_bh(&c->tcf_lock); 1426b57dc7c1SPaul Blakey 1427b57dc7c1SPaul Blakey return skb->len; 1428b57dc7c1SPaul Blakey nla_put_failure: 1429b57dc7c1SPaul Blakey spin_unlock_bh(&c->tcf_lock); 1430b57dc7c1SPaul Blakey nlmsg_trim(skb, b); 1431b57dc7c1SPaul Blakey return -1; 1432b57dc7c1SPaul Blakey } 1433b57dc7c1SPaul Blakey 1434b57dc7c1SPaul Blakey static int tcf_ct_walker(struct net *net, struct sk_buff *skb, 1435b57dc7c1SPaul Blakey struct netlink_callback *cb, int type, 1436b57dc7c1SPaul Blakey const struct tc_action_ops *ops, 1437b57dc7c1SPaul Blakey struct netlink_ext_ack *extack) 1438b57dc7c1SPaul Blakey { 1439b57dc7c1SPaul Blakey struct tc_action_net *tn = net_generic(net, ct_net_id); 1440b57dc7c1SPaul Blakey 1441b57dc7c1SPaul Blakey return tcf_generic_walker(tn, skb, cb, type, ops, extack); 1442b57dc7c1SPaul Blakey } 1443b57dc7c1SPaul Blakey 1444b57dc7c1SPaul Blakey static int tcf_ct_search(struct net *net, struct tc_action **a, u32 index) 1445b57dc7c1SPaul Blakey { 1446b57dc7c1SPaul Blakey struct tc_action_net *tn = net_generic(net, ct_net_id); 1447b57dc7c1SPaul Blakey 1448b57dc7c1SPaul Blakey return tcf_idr_search(tn, a, index); 1449b57dc7c1SPaul Blakey } 1450b57dc7c1SPaul Blakey 1451b57dc7c1SPaul Blakey static void tcf_stats_update(struct tc_action *a, u64 bytes, u32 packets, 1452b57dc7c1SPaul Blakey u64 lastuse, bool hw) 1453b57dc7c1SPaul Blakey { 1454b57dc7c1SPaul Blakey struct tcf_ct *c = to_ct(a); 1455b57dc7c1SPaul Blakey 1456c8ecebd0SVlad Buslov tcf_action_update_stats(a, bytes, packets, false, hw); 1457b57dc7c1SPaul Blakey c->tcf_tm.lastuse = max_t(u64, c->tcf_tm.lastuse, lastuse); 1458b57dc7c1SPaul Blakey } 1459b57dc7c1SPaul Blakey 1460b57dc7c1SPaul Blakey static struct tc_action_ops act_ct_ops = { 1461b57dc7c1SPaul Blakey .kind = "ct", 1462b57dc7c1SPaul Blakey .id = TCA_ID_CT, 1463b57dc7c1SPaul Blakey .owner = THIS_MODULE, 1464b57dc7c1SPaul Blakey .act = tcf_ct_act, 1465b57dc7c1SPaul Blakey .dump = tcf_ct_dump, 1466b57dc7c1SPaul Blakey .init = tcf_ct_init, 1467b57dc7c1SPaul Blakey .cleanup = tcf_ct_cleanup, 1468b57dc7c1SPaul Blakey .walk = tcf_ct_walker, 1469b57dc7c1SPaul Blakey .lookup = tcf_ct_search, 1470b57dc7c1SPaul Blakey .stats_update = tcf_stats_update, 1471b57dc7c1SPaul Blakey .size = sizeof(struct tcf_ct), 1472b57dc7c1SPaul Blakey }; 1473b57dc7c1SPaul Blakey 1474b57dc7c1SPaul Blakey static __net_init int ct_init_net(struct net *net) 1475b57dc7c1SPaul Blakey { 1476c593642cSPankaj Bharadiya unsigned int n_bits = sizeof_field(struct tcf_ct_params, labels) * 8; 1477b57dc7c1SPaul Blakey struct tc_ct_action_net *tn = net_generic(net, ct_net_id); 1478b57dc7c1SPaul Blakey 1479b57dc7c1SPaul Blakey if (nf_connlabels_get(net, n_bits - 1)) { 1480b57dc7c1SPaul Blakey tn->labels = false; 1481b57dc7c1SPaul Blakey pr_err("act_ct: Failed to set connlabels length"); 1482b57dc7c1SPaul Blakey } else { 1483b57dc7c1SPaul Blakey tn->labels = true; 1484b57dc7c1SPaul Blakey } 1485b57dc7c1SPaul Blakey 1486981471bdSCong Wang return tc_action_net_init(net, &tn->tn, &act_ct_ops); 1487b57dc7c1SPaul Blakey } 1488b57dc7c1SPaul Blakey 1489b57dc7c1SPaul Blakey static void __net_exit ct_exit_net(struct list_head *net_list) 1490b57dc7c1SPaul Blakey { 1491b57dc7c1SPaul Blakey struct net *net; 1492b57dc7c1SPaul Blakey 1493b57dc7c1SPaul Blakey rtnl_lock(); 1494b57dc7c1SPaul Blakey list_for_each_entry(net, net_list, exit_list) { 1495b57dc7c1SPaul Blakey struct tc_ct_action_net *tn = net_generic(net, ct_net_id); 1496b57dc7c1SPaul Blakey 1497b57dc7c1SPaul Blakey if (tn->labels) 1498b57dc7c1SPaul Blakey nf_connlabels_put(net); 1499b57dc7c1SPaul Blakey } 1500b57dc7c1SPaul Blakey rtnl_unlock(); 1501b57dc7c1SPaul Blakey 1502b57dc7c1SPaul Blakey tc_action_net_exit(net_list, ct_net_id); 1503b57dc7c1SPaul Blakey } 1504b57dc7c1SPaul Blakey 1505b57dc7c1SPaul Blakey static struct pernet_operations ct_net_ops = { 1506b57dc7c1SPaul Blakey .init = ct_init_net, 1507b57dc7c1SPaul Blakey .exit_batch = ct_exit_net, 1508b57dc7c1SPaul Blakey .id = &ct_net_id, 1509b57dc7c1SPaul Blakey .size = sizeof(struct tc_ct_action_net), 1510b57dc7c1SPaul Blakey }; 1511b57dc7c1SPaul Blakey 1512b57dc7c1SPaul Blakey static int __init ct_init_module(void) 1513b57dc7c1SPaul Blakey { 1514c34b961aSPaul Blakey int err; 1515c34b961aSPaul Blakey 1516c34b961aSPaul Blakey act_ct_wq = alloc_ordered_workqueue("act_ct_workqueue", 0); 1517c34b961aSPaul Blakey if (!act_ct_wq) 1518c34b961aSPaul Blakey return -ENOMEM; 1519c34b961aSPaul Blakey 1520c34b961aSPaul Blakey err = tcf_ct_flow_tables_init(); 1521c34b961aSPaul Blakey if (err) 1522c34b961aSPaul Blakey goto err_tbl_init; 1523c34b961aSPaul Blakey 1524c34b961aSPaul Blakey err = tcf_register_action(&act_ct_ops, &ct_net_ops); 1525c34b961aSPaul Blakey if (err) 1526c34b961aSPaul Blakey goto err_register; 1527c34b961aSPaul Blakey 1528c34b961aSPaul Blakey return 0; 1529c34b961aSPaul Blakey 1530c34b961aSPaul Blakey err_tbl_init: 1531c34b961aSPaul Blakey destroy_workqueue(act_ct_wq); 1532c34b961aSPaul Blakey err_register: 1533c34b961aSPaul Blakey tcf_ct_flow_tables_uninit(); 1534c34b961aSPaul Blakey return err; 1535b57dc7c1SPaul Blakey } 1536b57dc7c1SPaul Blakey 1537b57dc7c1SPaul Blakey static void __exit ct_cleanup_module(void) 1538b57dc7c1SPaul Blakey { 1539b57dc7c1SPaul Blakey tcf_unregister_action(&act_ct_ops, &ct_net_ops); 1540c34b961aSPaul Blakey tcf_ct_flow_tables_uninit(); 1541c34b961aSPaul Blakey destroy_workqueue(act_ct_wq); 1542b57dc7c1SPaul Blakey } 1543b57dc7c1SPaul Blakey 154430b0cf90SPaul Blakey void tcf_ct_flow_table_restore_skb(struct sk_buff *skb, unsigned long cookie) 154530b0cf90SPaul Blakey { 154630b0cf90SPaul Blakey enum ip_conntrack_info ctinfo = cookie & NFCT_INFOMASK; 154730b0cf90SPaul Blakey struct nf_conn *ct; 154830b0cf90SPaul Blakey 154930b0cf90SPaul Blakey ct = (struct nf_conn *)(cookie & NFCT_PTRMASK); 155030b0cf90SPaul Blakey nf_conntrack_get(&ct->ct_general); 155130b0cf90SPaul Blakey nf_ct_set(skb, ct, ctinfo); 155230b0cf90SPaul Blakey } 155330b0cf90SPaul Blakey EXPORT_SYMBOL_GPL(tcf_ct_flow_table_restore_skb); 155430b0cf90SPaul Blakey 1555b57dc7c1SPaul Blakey module_init(ct_init_module); 1556b57dc7c1SPaul Blakey module_exit(ct_cleanup_module); 1557b57dc7c1SPaul Blakey MODULE_AUTHOR("Paul Blakey <paulb@mellanox.com>"); 1558b57dc7c1SPaul Blakey MODULE_AUTHOR("Yossi Kuperman <yossiku@mellanox.com>"); 1559b57dc7c1SPaul Blakey MODULE_AUTHOR("Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>"); 1560b57dc7c1SPaul Blakey MODULE_DESCRIPTION("Connection tracking action"); 1561b57dc7c1SPaul Blakey MODULE_LICENSE("GPL v2"); 1562b57dc7c1SPaul Blakey 1563