Back to home page

OSCL-LXR
 
 

     

0001 // SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
0002 /* -
0003  * net/sched/act_ct.c  Connection Tracking action
0004  *
0005  * Authors:   Paul Blakey <paulb@mellanox.com>
0006  *            Yossi Kuperman <yossiku@mellanox.com>
0007  *            Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
0008  */
0009 
0010 #include <linux/module.h>
0011 #include <linux/init.h>
0012 #include <linux/kernel.h>
0013 #include <linux/skbuff.h>
0014 #include <linux/rtnetlink.h>
0015 #include <linux/pkt_cls.h>
0016 #include <linux/ip.h>
0017 #include <linux/ipv6.h>
0018 #include <linux/rhashtable.h>
0019 #include <net/netlink.h>
0020 #include <net/pkt_sched.h>
0021 #include <net/pkt_cls.h>
0022 #include <net/act_api.h>
0023 #include <net/ip.h>
0024 #include <net/ipv6_frag.h>
0025 #include <uapi/linux/tc_act/tc_ct.h>
0026 #include <net/tc_act/tc_ct.h>
0027 
0028 #include <net/netfilter/nf_flow_table.h>
0029 #include <net/netfilter/nf_conntrack.h>
0030 #include <net/netfilter/nf_conntrack_core.h>
0031 #include <net/netfilter/nf_conntrack_zones.h>
0032 #include <net/netfilter/nf_conntrack_helper.h>
0033 #include <net/netfilter/nf_conntrack_acct.h>
0034 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
0035 #include <net/netfilter/nf_conntrack_act_ct.h>
0036 #include <uapi/linux/netfilter/nf_nat.h>
0037 
0038 static struct workqueue_struct *act_ct_wq;
0039 static struct rhashtable zones_ht;
0040 static DEFINE_MUTEX(zones_mutex);
0041 
0042 struct tcf_ct_flow_table {
0043     struct rhash_head node; /* In zones tables */
0044 
0045     struct rcu_work rwork;
0046     struct nf_flowtable nf_ft;
0047     refcount_t ref;
0048     u16 zone;
0049 
0050     bool dying;
0051 };
0052 
0053 static const struct rhashtable_params zones_params = {
0054     .head_offset = offsetof(struct tcf_ct_flow_table, node),
0055     .key_offset = offsetof(struct tcf_ct_flow_table, zone),
0056     .key_len = sizeof_field(struct tcf_ct_flow_table, zone),
0057     .automatic_shrinking = true,
0058 };
0059 
0060 static struct flow_action_entry *
0061 tcf_ct_flow_table_flow_action_get_next(struct flow_action *flow_action)
0062 {
0063     int i = flow_action->num_entries++;
0064 
0065     return &flow_action->entries[i];
0066 }
0067 
0068 static void tcf_ct_add_mangle_action(struct flow_action *action,
0069                      enum flow_action_mangle_base htype,
0070                      u32 offset,
0071                      u32 mask,
0072                      u32 val)
0073 {
0074     struct flow_action_entry *entry;
0075 
0076     entry = tcf_ct_flow_table_flow_action_get_next(action);
0077     entry->id = FLOW_ACTION_MANGLE;
0078     entry->mangle.htype = htype;
0079     entry->mangle.mask = ~mask;
0080     entry->mangle.offset = offset;
0081     entry->mangle.val = val;
0082 }
0083 
0084 /* The following nat helper functions check if the inverted reverse tuple
0085  * (target) is different then the current dir tuple - meaning nat for ports
0086  * and/or ip is needed, and add the relevant mangle actions.
0087  */
0088 static void
0089 tcf_ct_flow_table_add_action_nat_ipv4(const struct nf_conntrack_tuple *tuple,
0090                       struct nf_conntrack_tuple target,
0091                       struct flow_action *action)
0092 {
0093     if (memcmp(&target.src.u3, &tuple->src.u3, sizeof(target.src.u3)))
0094         tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_IP4,
0095                      offsetof(struct iphdr, saddr),
0096                      0xFFFFFFFF,
0097                      be32_to_cpu(target.src.u3.ip));
0098     if (memcmp(&target.dst.u3, &tuple->dst.u3, sizeof(target.dst.u3)))
0099         tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_IP4,
0100                      offsetof(struct iphdr, daddr),
0101                      0xFFFFFFFF,
0102                      be32_to_cpu(target.dst.u3.ip));
0103 }
0104 
0105 static void
0106 tcf_ct_add_ipv6_addr_mangle_action(struct flow_action *action,
0107                    union nf_inet_addr *addr,
0108                    u32 offset)
0109 {
0110     int i;
0111 
0112     for (i = 0; i < sizeof(struct in6_addr) / sizeof(u32); i++)
0113         tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_IP6,
0114                      i * sizeof(u32) + offset,
0115                      0xFFFFFFFF, be32_to_cpu(addr->ip6[i]));
0116 }
0117 
0118 static void
0119 tcf_ct_flow_table_add_action_nat_ipv6(const struct nf_conntrack_tuple *tuple,
0120                       struct nf_conntrack_tuple target,
0121                       struct flow_action *action)
0122 {
0123     if (memcmp(&target.src.u3, &tuple->src.u3, sizeof(target.src.u3)))
0124         tcf_ct_add_ipv6_addr_mangle_action(action, &target.src.u3,
0125                            offsetof(struct ipv6hdr,
0126                                 saddr));
0127     if (memcmp(&target.dst.u3, &tuple->dst.u3, sizeof(target.dst.u3)))
0128         tcf_ct_add_ipv6_addr_mangle_action(action, &target.dst.u3,
0129                            offsetof(struct ipv6hdr,
0130                                 daddr));
0131 }
0132 
0133 static void
0134 tcf_ct_flow_table_add_action_nat_tcp(const struct nf_conntrack_tuple *tuple,
0135                      struct nf_conntrack_tuple target,
0136                      struct flow_action *action)
0137 {
0138     __be16 target_src = target.src.u.tcp.port;
0139     __be16 target_dst = target.dst.u.tcp.port;
0140 
0141     if (target_src != tuple->src.u.tcp.port)
0142         tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_TCP,
0143                      offsetof(struct tcphdr, source),
0144                      0xFFFF, be16_to_cpu(target_src));
0145     if (target_dst != tuple->dst.u.tcp.port)
0146         tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_TCP,
0147                      offsetof(struct tcphdr, dest),
0148                      0xFFFF, be16_to_cpu(target_dst));
0149 }
0150 
0151 static void
0152 tcf_ct_flow_table_add_action_nat_udp(const struct nf_conntrack_tuple *tuple,
0153                      struct nf_conntrack_tuple target,
0154                      struct flow_action *action)
0155 {
0156     __be16 target_src = target.src.u.udp.port;
0157     __be16 target_dst = target.dst.u.udp.port;
0158 
0159     if (target_src != tuple->src.u.udp.port)
0160         tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_UDP,
0161                      offsetof(struct udphdr, source),
0162                      0xFFFF, be16_to_cpu(target_src));
0163     if (target_dst != tuple->dst.u.udp.port)
0164         tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_UDP,
0165                      offsetof(struct udphdr, dest),
0166                      0xFFFF, be16_to_cpu(target_dst));
0167 }
0168 
0169 static void tcf_ct_flow_table_add_action_meta(struct nf_conn *ct,
0170                           enum ip_conntrack_dir dir,
0171                           struct flow_action *action)
0172 {
0173     struct nf_conn_labels *ct_labels;
0174     struct flow_action_entry *entry;
0175     enum ip_conntrack_info ctinfo;
0176     u32 *act_ct_labels;
0177 
0178     entry = tcf_ct_flow_table_flow_action_get_next(action);
0179     entry->id = FLOW_ACTION_CT_METADATA;
0180 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
0181     entry->ct_metadata.mark = ct->mark;
0182 #endif
0183     ctinfo = dir == IP_CT_DIR_ORIGINAL ? IP_CT_ESTABLISHED :
0184                          IP_CT_ESTABLISHED_REPLY;
0185     /* aligns with the CT reference on the SKB nf_ct_set */
0186     entry->ct_metadata.cookie = (unsigned long)ct | ctinfo;
0187     entry->ct_metadata.orig_dir = dir == IP_CT_DIR_ORIGINAL;
0188 
0189     act_ct_labels = entry->ct_metadata.labels;
0190     ct_labels = nf_ct_labels_find(ct);
0191     if (ct_labels)
0192         memcpy(act_ct_labels, ct_labels->bits, NF_CT_LABELS_MAX_SIZE);
0193     else
0194         memset(act_ct_labels, 0, NF_CT_LABELS_MAX_SIZE);
0195 }
0196 
0197 static int tcf_ct_flow_table_add_action_nat(struct net *net,
0198                         struct nf_conn *ct,
0199                         enum ip_conntrack_dir dir,
0200                         struct flow_action *action)
0201 {
0202     const struct nf_conntrack_tuple *tuple = &ct->tuplehash[dir].tuple;
0203     struct nf_conntrack_tuple target;
0204 
0205     if (!(ct->status & IPS_NAT_MASK))
0206         return 0;
0207 
0208     nf_ct_invert_tuple(&target, &ct->tuplehash[!dir].tuple);
0209 
0210     switch (tuple->src.l3num) {
0211     case NFPROTO_IPV4:
0212         tcf_ct_flow_table_add_action_nat_ipv4(tuple, target,
0213                               action);
0214         break;
0215     case NFPROTO_IPV6:
0216         tcf_ct_flow_table_add_action_nat_ipv6(tuple, target,
0217                               action);
0218         break;
0219     default:
0220         return -EOPNOTSUPP;
0221     }
0222 
0223     switch (nf_ct_protonum(ct)) {
0224     case IPPROTO_TCP:
0225         tcf_ct_flow_table_add_action_nat_tcp(tuple, target, action);
0226         break;
0227     case IPPROTO_UDP:
0228         tcf_ct_flow_table_add_action_nat_udp(tuple, target, action);
0229         break;
0230     default:
0231         return -EOPNOTSUPP;
0232     }
0233 
0234     return 0;
0235 }
0236 
0237 static int tcf_ct_flow_table_fill_actions(struct net *net,
0238                       const struct flow_offload *flow,
0239                       enum flow_offload_tuple_dir tdir,
0240                       struct nf_flow_rule *flow_rule)
0241 {
0242     struct flow_action *action = &flow_rule->rule->action;
0243     int num_entries = action->num_entries;
0244     struct nf_conn *ct = flow->ct;
0245     enum ip_conntrack_dir dir;
0246     int i, err;
0247 
0248     switch (tdir) {
0249     case FLOW_OFFLOAD_DIR_ORIGINAL:
0250         dir = IP_CT_DIR_ORIGINAL;
0251         break;
0252     case FLOW_OFFLOAD_DIR_REPLY:
0253         dir = IP_CT_DIR_REPLY;
0254         break;
0255     default:
0256         return -EOPNOTSUPP;
0257     }
0258 
0259     err = tcf_ct_flow_table_add_action_nat(net, ct, dir, action);
0260     if (err)
0261         goto err_nat;
0262 
0263     tcf_ct_flow_table_add_action_meta(ct, dir, action);
0264     return 0;
0265 
0266 err_nat:
0267     /* Clear filled actions */
0268     for (i = num_entries; i < action->num_entries; i++)
0269         memset(&action->entries[i], 0, sizeof(action->entries[i]));
0270     action->num_entries = num_entries;
0271 
0272     return err;
0273 }
0274 
0275 static struct nf_flowtable_type flowtable_ct = {
0276     .action     = tcf_ct_flow_table_fill_actions,
0277     .owner      = THIS_MODULE,
0278 };
0279 
0280 static int tcf_ct_flow_table_get(struct net *net, struct tcf_ct_params *params)
0281 {
0282     struct tcf_ct_flow_table *ct_ft;
0283     int err = -ENOMEM;
0284 
0285     mutex_lock(&zones_mutex);
0286     ct_ft = rhashtable_lookup_fast(&zones_ht, &params->zone, zones_params);
0287     if (ct_ft && refcount_inc_not_zero(&ct_ft->ref))
0288         goto out_unlock;
0289 
0290     ct_ft = kzalloc(sizeof(*ct_ft), GFP_KERNEL);
0291     if (!ct_ft)
0292         goto err_alloc;
0293     refcount_set(&ct_ft->ref, 1);
0294 
0295     ct_ft->zone = params->zone;
0296     err = rhashtable_insert_fast(&zones_ht, &ct_ft->node, zones_params);
0297     if (err)
0298         goto err_insert;
0299 
0300     ct_ft->nf_ft.type = &flowtable_ct;
0301     ct_ft->nf_ft.flags |= NF_FLOWTABLE_HW_OFFLOAD |
0302                   NF_FLOWTABLE_COUNTER;
0303     err = nf_flow_table_init(&ct_ft->nf_ft);
0304     if (err)
0305         goto err_init;
0306     write_pnet(&ct_ft->nf_ft.net, net);
0307 
0308     __module_get(THIS_MODULE);
0309 out_unlock:
0310     params->ct_ft = ct_ft;
0311     params->nf_ft = &ct_ft->nf_ft;
0312     mutex_unlock(&zones_mutex);
0313 
0314     return 0;
0315 
0316 err_init:
0317     rhashtable_remove_fast(&zones_ht, &ct_ft->node, zones_params);
0318 err_insert:
0319     kfree(ct_ft);
0320 err_alloc:
0321     mutex_unlock(&zones_mutex);
0322     return err;
0323 }
0324 
0325 static void tcf_ct_flow_table_cleanup_work(struct work_struct *work)
0326 {
0327     struct flow_block_cb *block_cb, *tmp_cb;
0328     struct tcf_ct_flow_table *ct_ft;
0329     struct flow_block *block;
0330 
0331     ct_ft = container_of(to_rcu_work(work), struct tcf_ct_flow_table,
0332                  rwork);
0333     nf_flow_table_free(&ct_ft->nf_ft);
0334 
0335     /* Remove any remaining callbacks before cleanup */
0336     block = &ct_ft->nf_ft.flow_block;
0337     down_write(&ct_ft->nf_ft.flow_block_lock);
0338     list_for_each_entry_safe(block_cb, tmp_cb, &block->cb_list, list) {
0339         list_del(&block_cb->list);
0340         flow_block_cb_free(block_cb);
0341     }
0342     up_write(&ct_ft->nf_ft.flow_block_lock);
0343     kfree(ct_ft);
0344 
0345     module_put(THIS_MODULE);
0346 }
0347 
0348 static void tcf_ct_flow_table_put(struct tcf_ct_params *params)
0349 {
0350     struct tcf_ct_flow_table *ct_ft = params->ct_ft;
0351 
0352     if (refcount_dec_and_test(&params->ct_ft->ref)) {
0353         rhashtable_remove_fast(&zones_ht, &ct_ft->node, zones_params);
0354         INIT_RCU_WORK(&ct_ft->rwork, tcf_ct_flow_table_cleanup_work);
0355         queue_rcu_work(act_ct_wq, &ct_ft->rwork);
0356     }
0357 }
0358 
0359 static void tcf_ct_flow_tc_ifidx(struct flow_offload *entry,
0360                  struct nf_conn_act_ct_ext *act_ct_ext, u8 dir)
0361 {
0362     entry->tuplehash[dir].tuple.xmit_type = FLOW_OFFLOAD_XMIT_TC;
0363     entry->tuplehash[dir].tuple.tc.iifidx = act_ct_ext->ifindex[dir];
0364 }
0365 
0366 static void tcf_ct_flow_table_add(struct tcf_ct_flow_table *ct_ft,
0367                   struct nf_conn *ct,
0368                   bool tcp)
0369 {
0370     struct nf_conn_act_ct_ext *act_ct_ext;
0371     struct flow_offload *entry;
0372     int err;
0373 
0374     if (test_and_set_bit(IPS_OFFLOAD_BIT, &ct->status))
0375         return;
0376 
0377     entry = flow_offload_alloc(ct);
0378     if (!entry) {
0379         WARN_ON_ONCE(1);
0380         goto err_alloc;
0381     }
0382 
0383     if (tcp) {
0384         ct->proto.tcp.seen[0].flags |= IP_CT_TCP_FLAG_BE_LIBERAL;
0385         ct->proto.tcp.seen[1].flags |= IP_CT_TCP_FLAG_BE_LIBERAL;
0386     }
0387 
0388     act_ct_ext = nf_conn_act_ct_ext_find(ct);
0389     if (act_ct_ext) {
0390         tcf_ct_flow_tc_ifidx(entry, act_ct_ext, FLOW_OFFLOAD_DIR_ORIGINAL);
0391         tcf_ct_flow_tc_ifidx(entry, act_ct_ext, FLOW_OFFLOAD_DIR_REPLY);
0392     }
0393 
0394     err = flow_offload_add(&ct_ft->nf_ft, entry);
0395     if (err)
0396         goto err_add;
0397 
0398     return;
0399 
0400 err_add:
0401     flow_offload_free(entry);
0402 err_alloc:
0403     clear_bit(IPS_OFFLOAD_BIT, &ct->status);
0404 }
0405 
0406 static void tcf_ct_flow_table_process_conn(struct tcf_ct_flow_table *ct_ft,
0407                        struct nf_conn *ct,
0408                        enum ip_conntrack_info ctinfo)
0409 {
0410     bool tcp = false;
0411 
0412     if ((ctinfo != IP_CT_ESTABLISHED && ctinfo != IP_CT_ESTABLISHED_REPLY) ||
0413         !test_bit(IPS_ASSURED_BIT, &ct->status))
0414         return;
0415 
0416     switch (nf_ct_protonum(ct)) {
0417     case IPPROTO_TCP:
0418         tcp = true;
0419         if (ct->proto.tcp.state != TCP_CONNTRACK_ESTABLISHED)
0420             return;
0421         break;
0422     case IPPROTO_UDP:
0423         break;
0424 #ifdef CONFIG_NF_CT_PROTO_GRE
0425     case IPPROTO_GRE: {
0426         struct nf_conntrack_tuple *tuple;
0427 
0428         if (ct->status & IPS_NAT_MASK)
0429             return;
0430         tuple = &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple;
0431         /* No support for GRE v1 */
0432         if (tuple->src.u.gre.key || tuple->dst.u.gre.key)
0433             return;
0434         break;
0435     }
0436 #endif
0437     default:
0438         return;
0439     }
0440 
0441     if (nf_ct_ext_exist(ct, NF_CT_EXT_HELPER) ||
0442         ct->status & IPS_SEQ_ADJUST)
0443         return;
0444 
0445     tcf_ct_flow_table_add(ct_ft, ct, tcp);
0446 }
0447 
0448 static bool
0449 tcf_ct_flow_table_fill_tuple_ipv4(struct sk_buff *skb,
0450                   struct flow_offload_tuple *tuple,
0451                   struct tcphdr **tcph)
0452 {
0453     struct flow_ports *ports;
0454     unsigned int thoff;
0455     struct iphdr *iph;
0456     size_t hdrsize;
0457     u8 ipproto;
0458 
0459     if (!pskb_network_may_pull(skb, sizeof(*iph)))
0460         return false;
0461 
0462     iph = ip_hdr(skb);
0463     thoff = iph->ihl * 4;
0464 
0465     if (ip_is_fragment(iph) ||
0466         unlikely(thoff != sizeof(struct iphdr)))
0467         return false;
0468 
0469     ipproto = iph->protocol;
0470     switch (ipproto) {
0471     case IPPROTO_TCP:
0472         hdrsize = sizeof(struct tcphdr);
0473         break;
0474     case IPPROTO_UDP:
0475         hdrsize = sizeof(*ports);
0476         break;
0477 #ifdef CONFIG_NF_CT_PROTO_GRE
0478     case IPPROTO_GRE:
0479         hdrsize = sizeof(struct gre_base_hdr);
0480         break;
0481 #endif
0482     default:
0483         return false;
0484     }
0485 
0486     if (iph->ttl <= 1)
0487         return false;
0488 
0489     if (!pskb_network_may_pull(skb, thoff + hdrsize))
0490         return false;
0491 
0492     switch (ipproto) {
0493     case IPPROTO_TCP:
0494         *tcph = (void *)(skb_network_header(skb) + thoff);
0495         fallthrough;
0496     case IPPROTO_UDP:
0497         ports = (struct flow_ports *)(skb_network_header(skb) + thoff);
0498         tuple->src_port = ports->source;
0499         tuple->dst_port = ports->dest;
0500         break;
0501     case IPPROTO_GRE: {
0502         struct gre_base_hdr *greh;
0503 
0504         greh = (struct gre_base_hdr *)(skb_network_header(skb) + thoff);
0505         if ((greh->flags & GRE_VERSION) != GRE_VERSION_0)
0506             return false;
0507         break;
0508     }
0509     }
0510 
0511     iph = ip_hdr(skb);
0512 
0513     tuple->src_v4.s_addr = iph->saddr;
0514     tuple->dst_v4.s_addr = iph->daddr;
0515     tuple->l3proto = AF_INET;
0516     tuple->l4proto = ipproto;
0517 
0518     return true;
0519 }
0520 
0521 static bool
0522 tcf_ct_flow_table_fill_tuple_ipv6(struct sk_buff *skb,
0523                   struct flow_offload_tuple *tuple,
0524                   struct tcphdr **tcph)
0525 {
0526     struct flow_ports *ports;
0527     struct ipv6hdr *ip6h;
0528     unsigned int thoff;
0529     size_t hdrsize;
0530     u8 nexthdr;
0531 
0532     if (!pskb_network_may_pull(skb, sizeof(*ip6h)))
0533         return false;
0534 
0535     ip6h = ipv6_hdr(skb);
0536     thoff = sizeof(*ip6h);
0537 
0538     nexthdr = ip6h->nexthdr;
0539     switch (nexthdr) {
0540     case IPPROTO_TCP:
0541         hdrsize = sizeof(struct tcphdr);
0542         break;
0543     case IPPROTO_UDP:
0544         hdrsize = sizeof(*ports);
0545         break;
0546 #ifdef CONFIG_NF_CT_PROTO_GRE
0547     case IPPROTO_GRE:
0548         hdrsize = sizeof(struct gre_base_hdr);
0549         break;
0550 #endif
0551     default:
0552         return false;
0553     }
0554 
0555     if (ip6h->hop_limit <= 1)
0556         return false;
0557 
0558     if (!pskb_network_may_pull(skb, thoff + hdrsize))
0559         return false;
0560 
0561     switch (nexthdr) {
0562     case IPPROTO_TCP:
0563         *tcph = (void *)(skb_network_header(skb) + thoff);
0564         fallthrough;
0565     case IPPROTO_UDP:
0566         ports = (struct flow_ports *)(skb_network_header(skb) + thoff);
0567         tuple->src_port = ports->source;
0568         tuple->dst_port = ports->dest;
0569         break;
0570     case IPPROTO_GRE: {
0571         struct gre_base_hdr *greh;
0572 
0573         greh = (struct gre_base_hdr *)(skb_network_header(skb) + thoff);
0574         if ((greh->flags & GRE_VERSION) != GRE_VERSION_0)
0575             return false;
0576         break;
0577     }
0578     }
0579 
0580     ip6h = ipv6_hdr(skb);
0581 
0582     tuple->src_v6 = ip6h->saddr;
0583     tuple->dst_v6 = ip6h->daddr;
0584     tuple->l3proto = AF_INET6;
0585     tuple->l4proto = nexthdr;
0586 
0587     return true;
0588 }
0589 
0590 static bool tcf_ct_flow_table_lookup(struct tcf_ct_params *p,
0591                      struct sk_buff *skb,
0592                      u8 family)
0593 {
0594     struct nf_flowtable *nf_ft = &p->ct_ft->nf_ft;
0595     struct flow_offload_tuple_rhash *tuplehash;
0596     struct flow_offload_tuple tuple = {};
0597     enum ip_conntrack_info ctinfo;
0598     struct tcphdr *tcph = NULL;
0599     struct flow_offload *flow;
0600     struct nf_conn *ct;
0601     u8 dir;
0602 
0603     switch (family) {
0604     case NFPROTO_IPV4:
0605         if (!tcf_ct_flow_table_fill_tuple_ipv4(skb, &tuple, &tcph))
0606             return false;
0607         break;
0608     case NFPROTO_IPV6:
0609         if (!tcf_ct_flow_table_fill_tuple_ipv6(skb, &tuple, &tcph))
0610             return false;
0611         break;
0612     default:
0613         return false;
0614     }
0615 
0616     tuplehash = flow_offload_lookup(nf_ft, &tuple);
0617     if (!tuplehash)
0618         return false;
0619 
0620     dir = tuplehash->tuple.dir;
0621     flow = container_of(tuplehash, struct flow_offload, tuplehash[dir]);
0622     ct = flow->ct;
0623 
0624     if (tcph && (unlikely(tcph->fin || tcph->rst))) {
0625         flow_offload_teardown(flow);
0626         return false;
0627     }
0628 
0629     ctinfo = dir == FLOW_OFFLOAD_DIR_ORIGINAL ? IP_CT_ESTABLISHED :
0630                             IP_CT_ESTABLISHED_REPLY;
0631 
0632     flow_offload_refresh(nf_ft, flow);
0633     nf_conntrack_get(&ct->ct_general);
0634     nf_ct_set(skb, ct, ctinfo);
0635     if (nf_ft->flags & NF_FLOWTABLE_COUNTER)
0636         nf_ct_acct_update(ct, dir, skb->len);
0637 
0638     return true;
0639 }
0640 
0641 static int tcf_ct_flow_tables_init(void)
0642 {
0643     return rhashtable_init(&zones_ht, &zones_params);
0644 }
0645 
0646 static void tcf_ct_flow_tables_uninit(void)
0647 {
0648     rhashtable_destroy(&zones_ht);
0649 }
0650 
0651 static struct tc_action_ops act_ct_ops;
0652 static unsigned int ct_net_id;
0653 
0654 struct tc_ct_action_net {
0655     struct tc_action_net tn; /* Must be first */
0656     bool labels;
0657 };
0658 
0659 /* Determine whether skb->_nfct is equal to the result of conntrack lookup. */
0660 static bool tcf_ct_skb_nfct_cached(struct net *net, struct sk_buff *skb,
0661                    u16 zone_id, bool force)
0662 {
0663     enum ip_conntrack_info ctinfo;
0664     struct nf_conn *ct;
0665 
0666     ct = nf_ct_get(skb, &ctinfo);
0667     if (!ct)
0668         return false;
0669     if (!net_eq(net, read_pnet(&ct->ct_net)))
0670         goto drop_ct;
0671     if (nf_ct_zone(ct)->id != zone_id)
0672         goto drop_ct;
0673 
0674     /* Force conntrack entry direction. */
0675     if (force && CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) {
0676         if (nf_ct_is_confirmed(ct))
0677             nf_ct_kill(ct);
0678 
0679         goto drop_ct;
0680     }
0681 
0682     return true;
0683 
0684 drop_ct:
0685     nf_ct_put(ct);
0686     nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
0687 
0688     return false;
0689 }
0690 
0691 /* Trim the skb to the length specified by the IP/IPv6 header,
0692  * removing any trailing lower-layer padding. This prepares the skb
0693  * for higher-layer processing that assumes skb->len excludes padding
0694  * (such as nf_ip_checksum). The caller needs to pull the skb to the
0695  * network header, and ensure ip_hdr/ipv6_hdr points to valid data.
0696  */
0697 static int tcf_ct_skb_network_trim(struct sk_buff *skb, int family)
0698 {
0699     unsigned int len;
0700     int err;
0701 
0702     switch (family) {
0703     case NFPROTO_IPV4:
0704         len = ntohs(ip_hdr(skb)->tot_len);
0705         break;
0706     case NFPROTO_IPV6:
0707         len = sizeof(struct ipv6hdr)
0708             + ntohs(ipv6_hdr(skb)->payload_len);
0709         break;
0710     default:
0711         len = skb->len;
0712     }
0713 
0714     err = pskb_trim_rcsum(skb, len);
0715 
0716     return err;
0717 }
0718 
0719 static u8 tcf_ct_skb_nf_family(struct sk_buff *skb)
0720 {
0721     u8 family = NFPROTO_UNSPEC;
0722 
0723     switch (skb_protocol(skb, true)) {
0724     case htons(ETH_P_IP):
0725         family = NFPROTO_IPV4;
0726         break;
0727     case htons(ETH_P_IPV6):
0728         family = NFPROTO_IPV6;
0729         break;
0730     default:
0731         break;
0732     }
0733 
0734     return family;
0735 }
0736 
0737 static int tcf_ct_ipv4_is_fragment(struct sk_buff *skb, bool *frag)
0738 {
0739     unsigned int len;
0740 
0741     len =  skb_network_offset(skb) + sizeof(struct iphdr);
0742     if (unlikely(skb->len < len))
0743         return -EINVAL;
0744     if (unlikely(!pskb_may_pull(skb, len)))
0745         return -ENOMEM;
0746 
0747     *frag = ip_is_fragment(ip_hdr(skb));
0748     return 0;
0749 }
0750 
0751 static int tcf_ct_ipv6_is_fragment(struct sk_buff *skb, bool *frag)
0752 {
0753     unsigned int flags = 0, len, payload_ofs = 0;
0754     unsigned short frag_off;
0755     int nexthdr;
0756 
0757     len =  skb_network_offset(skb) + sizeof(struct ipv6hdr);
0758     if (unlikely(skb->len < len))
0759         return -EINVAL;
0760     if (unlikely(!pskb_may_pull(skb, len)))
0761         return -ENOMEM;
0762 
0763     nexthdr = ipv6_find_hdr(skb, &payload_ofs, -1, &frag_off, &flags);
0764     if (unlikely(nexthdr < 0))
0765         return -EPROTO;
0766 
0767     *frag = flags & IP6_FH_F_FRAG;
0768     return 0;
0769 }
0770 
0771 static int tcf_ct_handle_fragments(struct net *net, struct sk_buff *skb,
0772                    u8 family, u16 zone, bool *defrag)
0773 {
0774     enum ip_conntrack_info ctinfo;
0775     struct nf_conn *ct;
0776     int err = 0;
0777     bool frag;
0778     u16 mru;
0779 
0780     /* Previously seen (loopback)? Ignore. */
0781     ct = nf_ct_get(skb, &ctinfo);
0782     if ((ct && !nf_ct_is_template(ct)) || ctinfo == IP_CT_UNTRACKED)
0783         return 0;
0784 
0785     if (family == NFPROTO_IPV4)
0786         err = tcf_ct_ipv4_is_fragment(skb, &frag);
0787     else
0788         err = tcf_ct_ipv6_is_fragment(skb, &frag);
0789     if (err || !frag)
0790         return err;
0791 
0792     skb_get(skb);
0793     mru = tc_skb_cb(skb)->mru;
0794 
0795     if (family == NFPROTO_IPV4) {
0796         enum ip_defrag_users user = IP_DEFRAG_CONNTRACK_IN + zone;
0797 
0798         memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
0799         local_bh_disable();
0800         err = ip_defrag(net, skb, user);
0801         local_bh_enable();
0802         if (err && err != -EINPROGRESS)
0803             return err;
0804 
0805         if (!err) {
0806             *defrag = true;
0807             mru = IPCB(skb)->frag_max_size;
0808         }
0809     } else { /* NFPROTO_IPV6 */
0810 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
0811         enum ip6_defrag_users user = IP6_DEFRAG_CONNTRACK_IN + zone;
0812 
0813         memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
0814         err = nf_ct_frag6_gather(net, skb, user);
0815         if (err && err != -EINPROGRESS)
0816             goto out_free;
0817 
0818         if (!err) {
0819             *defrag = true;
0820             mru = IP6CB(skb)->frag_max_size;
0821         }
0822 #else
0823         err = -EOPNOTSUPP;
0824         goto out_free;
0825 #endif
0826     }
0827 
0828     if (err != -EINPROGRESS)
0829         tc_skb_cb(skb)->mru = mru;
0830     skb_clear_hash(skb);
0831     skb->ignore_df = 1;
0832     return err;
0833 
0834 out_free:
0835     kfree_skb(skb);
0836     return err;
0837 }
0838 
0839 static void tcf_ct_params_free(struct rcu_head *head)
0840 {
0841     struct tcf_ct_params *params = container_of(head,
0842                             struct tcf_ct_params, rcu);
0843 
0844     tcf_ct_flow_table_put(params);
0845 
0846     if (params->tmpl)
0847         nf_ct_put(params->tmpl);
0848     kfree(params);
0849 }
0850 
0851 #if IS_ENABLED(CONFIG_NF_NAT)
0852 /* Modelled after nf_nat_ipv[46]_fn().
0853  * range is only used for new, uninitialized NAT state.
0854  * Returns either NF_ACCEPT or NF_DROP.
0855  */
0856 static int ct_nat_execute(struct sk_buff *skb, struct nf_conn *ct,
0857               enum ip_conntrack_info ctinfo,
0858               const struct nf_nat_range2 *range,
0859               enum nf_nat_manip_type maniptype)
0860 {
0861     __be16 proto = skb_protocol(skb, true);
0862     int hooknum, err = NF_ACCEPT;
0863 
0864     /* See HOOK2MANIP(). */
0865     if (maniptype == NF_NAT_MANIP_SRC)
0866         hooknum = NF_INET_LOCAL_IN; /* Source NAT */
0867     else
0868         hooknum = NF_INET_LOCAL_OUT; /* Destination NAT */
0869 
0870     switch (ctinfo) {
0871     case IP_CT_RELATED:
0872     case IP_CT_RELATED_REPLY:
0873         if (proto == htons(ETH_P_IP) &&
0874             ip_hdr(skb)->protocol == IPPROTO_ICMP) {
0875             if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo,
0876                                hooknum))
0877                 err = NF_DROP;
0878             goto out;
0879         } else if (IS_ENABLED(CONFIG_IPV6) && proto == htons(ETH_P_IPV6)) {
0880             __be16 frag_off;
0881             u8 nexthdr = ipv6_hdr(skb)->nexthdr;
0882             int hdrlen = ipv6_skip_exthdr(skb,
0883                               sizeof(struct ipv6hdr),
0884                               &nexthdr, &frag_off);
0885 
0886             if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) {
0887                 if (!nf_nat_icmpv6_reply_translation(skb, ct,
0888                                      ctinfo,
0889                                      hooknum,
0890                                      hdrlen))
0891                     err = NF_DROP;
0892                 goto out;
0893             }
0894         }
0895         /* Non-ICMP, fall thru to initialize if needed. */
0896         fallthrough;
0897     case IP_CT_NEW:
0898         /* Seen it before?  This can happen for loopback, retrans,
0899          * or local packets.
0900          */
0901         if (!nf_nat_initialized(ct, maniptype)) {
0902             /* Initialize according to the NAT action. */
0903             err = (range && range->flags & NF_NAT_RANGE_MAP_IPS)
0904                 /* Action is set up to establish a new
0905                  * mapping.
0906                  */
0907                 ? nf_nat_setup_info(ct, range, maniptype)
0908                 : nf_nat_alloc_null_binding(ct, hooknum);
0909             if (err != NF_ACCEPT)
0910                 goto out;
0911         }
0912         break;
0913 
0914     case IP_CT_ESTABLISHED:
0915     case IP_CT_ESTABLISHED_REPLY:
0916         break;
0917 
0918     default:
0919         err = NF_DROP;
0920         goto out;
0921     }
0922 
0923     err = nf_nat_packet(ct, ctinfo, hooknum, skb);
0924     if (err == NF_ACCEPT) {
0925         if (maniptype == NF_NAT_MANIP_SRC)
0926             tc_skb_cb(skb)->post_ct_snat = 1;
0927         if (maniptype == NF_NAT_MANIP_DST)
0928             tc_skb_cb(skb)->post_ct_dnat = 1;
0929     }
0930 out:
0931     return err;
0932 }
0933 #endif /* CONFIG_NF_NAT */
0934 
0935 static void tcf_ct_act_set_mark(struct nf_conn *ct, u32 mark, u32 mask)
0936 {
0937 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
0938     u32 new_mark;
0939 
0940     if (!mask)
0941         return;
0942 
0943     new_mark = mark | (ct->mark & ~(mask));
0944     if (ct->mark != new_mark) {
0945         ct->mark = new_mark;
0946         if (nf_ct_is_confirmed(ct))
0947             nf_conntrack_event_cache(IPCT_MARK, ct);
0948     }
0949 #endif
0950 }
0951 
0952 static void tcf_ct_act_set_labels(struct nf_conn *ct,
0953                   u32 *labels,
0954                   u32 *labels_m)
0955 {
0956 #if IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS)
0957     size_t labels_sz = sizeof_field(struct tcf_ct_params, labels);
0958 
0959     if (!memchr_inv(labels_m, 0, labels_sz))
0960         return;
0961 
0962     nf_connlabels_replace(ct, labels, labels_m, 4);
0963 #endif
0964 }
0965 
0966 static int tcf_ct_act_nat(struct sk_buff *skb,
0967               struct nf_conn *ct,
0968               enum ip_conntrack_info ctinfo,
0969               int ct_action,
0970               struct nf_nat_range2 *range,
0971               bool commit)
0972 {
0973 #if IS_ENABLED(CONFIG_NF_NAT)
0974     int err;
0975     enum nf_nat_manip_type maniptype;
0976 
0977     if (!(ct_action & TCA_CT_ACT_NAT))
0978         return NF_ACCEPT;
0979 
0980     /* Add NAT extension if not confirmed yet. */
0981     if (!nf_ct_is_confirmed(ct) && !nf_ct_nat_ext_add(ct))
0982         return NF_DROP;   /* Can't NAT. */
0983 
0984     if (ctinfo != IP_CT_NEW && (ct->status & IPS_NAT_MASK) &&
0985         (ctinfo != IP_CT_RELATED || commit)) {
0986         /* NAT an established or related connection like before. */
0987         if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY)
0988             /* This is the REPLY direction for a connection
0989              * for which NAT was applied in the forward
0990              * direction.  Do the reverse NAT.
0991              */
0992             maniptype = ct->status & IPS_SRC_NAT
0993                 ? NF_NAT_MANIP_DST : NF_NAT_MANIP_SRC;
0994         else
0995             maniptype = ct->status & IPS_SRC_NAT
0996                 ? NF_NAT_MANIP_SRC : NF_NAT_MANIP_DST;
0997     } else if (ct_action & TCA_CT_ACT_NAT_SRC) {
0998         maniptype = NF_NAT_MANIP_SRC;
0999     } else if (ct_action & TCA_CT_ACT_NAT_DST) {
1000         maniptype = NF_NAT_MANIP_DST;
1001     } else {
1002         return NF_ACCEPT;
1003     }
1004 
1005     err = ct_nat_execute(skb, ct, ctinfo, range, maniptype);
1006     if (err == NF_ACCEPT && ct->status & IPS_DST_NAT) {
1007         if (ct->status & IPS_SRC_NAT) {
1008             if (maniptype == NF_NAT_MANIP_SRC)
1009                 maniptype = NF_NAT_MANIP_DST;
1010             else
1011                 maniptype = NF_NAT_MANIP_SRC;
1012 
1013             err = ct_nat_execute(skb, ct, ctinfo, range,
1014                          maniptype);
1015         } else if (CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) {
1016             err = ct_nat_execute(skb, ct, ctinfo, NULL,
1017                          NF_NAT_MANIP_SRC);
1018         }
1019     }
1020     return err;
1021 #else
1022     return NF_ACCEPT;
1023 #endif
1024 }
1025 
1026 static int tcf_ct_act(struct sk_buff *skb, const struct tc_action *a,
1027               struct tcf_result *res)
1028 {
1029     struct net *net = dev_net(skb->dev);
1030     bool cached, commit, clear, force;
1031     enum ip_conntrack_info ctinfo;
1032     struct tcf_ct *c = to_ct(a);
1033     struct nf_conn *tmpl = NULL;
1034     struct nf_hook_state state;
1035     int nh_ofs, err, retval;
1036     struct tcf_ct_params *p;
1037     bool skip_add = false;
1038     bool defrag = false;
1039     struct nf_conn *ct;
1040     u8 family;
1041 
1042     p = rcu_dereference_bh(c->params);
1043 
1044     retval = READ_ONCE(c->tcf_action);
1045     commit = p->ct_action & TCA_CT_ACT_COMMIT;
1046     clear = p->ct_action & TCA_CT_ACT_CLEAR;
1047     force = p->ct_action & TCA_CT_ACT_FORCE;
1048     tmpl = p->tmpl;
1049 
1050     tcf_lastuse_update(&c->tcf_tm);
1051     tcf_action_update_bstats(&c->common, skb);
1052 
1053     if (clear) {
1054         tc_skb_cb(skb)->post_ct = false;
1055         ct = nf_ct_get(skb, &ctinfo);
1056         if (ct) {
1057             nf_ct_put(ct);
1058             nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
1059         }
1060 
1061         goto out_clear;
1062     }
1063 
1064     family = tcf_ct_skb_nf_family(skb);
1065     if (family == NFPROTO_UNSPEC)
1066         goto drop;
1067 
1068     /* The conntrack module expects to be working at L3.
1069      * We also try to pull the IPv4/6 header to linear area
1070      */
1071     nh_ofs = skb_network_offset(skb);
1072     skb_pull_rcsum(skb, nh_ofs);
1073     err = tcf_ct_handle_fragments(net, skb, family, p->zone, &defrag);
1074     if (err == -EINPROGRESS) {
1075         retval = TC_ACT_STOLEN;
1076         goto out_clear;
1077     }
1078     if (err)
1079         goto drop;
1080 
1081     err = tcf_ct_skb_network_trim(skb, family);
1082     if (err)
1083         goto drop;
1084 
1085     /* If we are recirculating packets to match on ct fields and
1086      * committing with a separate ct action, then we don't need to
1087      * actually run the packet through conntrack twice unless it's for a
1088      * different zone.
1089      */
1090     cached = tcf_ct_skb_nfct_cached(net, skb, p->zone, force);
1091     if (!cached) {
1092         if (tcf_ct_flow_table_lookup(p, skb, family)) {
1093             skip_add = true;
1094             goto do_nat;
1095         }
1096 
1097         /* Associate skb with specified zone. */
1098         if (tmpl) {
1099             nf_conntrack_put(skb_nfct(skb));
1100             nf_conntrack_get(&tmpl->ct_general);
1101             nf_ct_set(skb, tmpl, IP_CT_NEW);
1102         }
1103 
1104         state.hook = NF_INET_PRE_ROUTING;
1105         state.net = net;
1106         state.pf = family;
1107         err = nf_conntrack_in(skb, &state);
1108         if (err != NF_ACCEPT)
1109             goto out_push;
1110     }
1111 
1112 do_nat:
1113     ct = nf_ct_get(skb, &ctinfo);
1114     if (!ct)
1115         goto out_push;
1116     nf_ct_deliver_cached_events(ct);
1117     nf_conn_act_ct_ext_fill(skb, ct, ctinfo);
1118 
1119     err = tcf_ct_act_nat(skb, ct, ctinfo, p->ct_action, &p->range, commit);
1120     if (err != NF_ACCEPT)
1121         goto drop;
1122 
1123     if (commit) {
1124         tcf_ct_act_set_mark(ct, p->mark, p->mark_mask);
1125         tcf_ct_act_set_labels(ct, p->labels, p->labels_mask);
1126 
1127         if (!nf_ct_is_confirmed(ct))
1128             nf_conn_act_ct_ext_add(ct);
1129 
1130         /* This will take care of sending queued events
1131          * even if the connection is already confirmed.
1132          */
1133         if (nf_conntrack_confirm(skb) != NF_ACCEPT)
1134             goto drop;
1135     }
1136 
1137     if (!skip_add)
1138         tcf_ct_flow_table_process_conn(p->ct_ft, ct, ctinfo);
1139 
1140 out_push:
1141     skb_push_rcsum(skb, nh_ofs);
1142 
1143     tc_skb_cb(skb)->post_ct = true;
1144     tc_skb_cb(skb)->zone = p->zone;
1145 out_clear:
1146     if (defrag)
1147         qdisc_skb_cb(skb)->pkt_len = skb->len;
1148     return retval;
1149 
1150 drop:
1151     tcf_action_inc_drop_qstats(&c->common);
1152     return TC_ACT_SHOT;
1153 }
1154 
1155 static const struct nla_policy ct_policy[TCA_CT_MAX + 1] = {
1156     [TCA_CT_ACTION] = { .type = NLA_U16 },
1157     [TCA_CT_PARMS] = NLA_POLICY_EXACT_LEN(sizeof(struct tc_ct)),
1158     [TCA_CT_ZONE] = { .type = NLA_U16 },
1159     [TCA_CT_MARK] = { .type = NLA_U32 },
1160     [TCA_CT_MARK_MASK] = { .type = NLA_U32 },
1161     [TCA_CT_LABELS] = { .type = NLA_BINARY,
1162                 .len = 128 / BITS_PER_BYTE },
1163     [TCA_CT_LABELS_MASK] = { .type = NLA_BINARY,
1164                  .len = 128 / BITS_PER_BYTE },
1165     [TCA_CT_NAT_IPV4_MIN] = { .type = NLA_U32 },
1166     [TCA_CT_NAT_IPV4_MAX] = { .type = NLA_U32 },
1167     [TCA_CT_NAT_IPV6_MIN] = NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
1168     [TCA_CT_NAT_IPV6_MAX] = NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
1169     [TCA_CT_NAT_PORT_MIN] = { .type = NLA_U16 },
1170     [TCA_CT_NAT_PORT_MAX] = { .type = NLA_U16 },
1171 };
1172 
1173 static int tcf_ct_fill_params_nat(struct tcf_ct_params *p,
1174                   struct tc_ct *parm,
1175                   struct nlattr **tb,
1176                   struct netlink_ext_ack *extack)
1177 {
1178     struct nf_nat_range2 *range;
1179 
1180     if (!(p->ct_action & TCA_CT_ACT_NAT))
1181         return 0;
1182 
1183     if (!IS_ENABLED(CONFIG_NF_NAT)) {
1184         NL_SET_ERR_MSG_MOD(extack, "Netfilter nat isn't enabled in kernel");
1185         return -EOPNOTSUPP;
1186     }
1187 
1188     if (!(p->ct_action & (TCA_CT_ACT_NAT_SRC | TCA_CT_ACT_NAT_DST)))
1189         return 0;
1190 
1191     if ((p->ct_action & TCA_CT_ACT_NAT_SRC) &&
1192         (p->ct_action & TCA_CT_ACT_NAT_DST)) {
1193         NL_SET_ERR_MSG_MOD(extack, "dnat and snat can't be enabled at the same time");
1194         return -EOPNOTSUPP;
1195     }
1196 
1197     range = &p->range;
1198     if (tb[TCA_CT_NAT_IPV4_MIN]) {
1199         struct nlattr *max_attr = tb[TCA_CT_NAT_IPV4_MAX];
1200 
1201         p->ipv4_range = true;
1202         range->flags |= NF_NAT_RANGE_MAP_IPS;
1203         range->min_addr.ip =
1204             nla_get_in_addr(tb[TCA_CT_NAT_IPV4_MIN]);
1205 
1206         range->max_addr.ip = max_attr ?
1207                      nla_get_in_addr(max_attr) :
1208                      range->min_addr.ip;
1209     } else if (tb[TCA_CT_NAT_IPV6_MIN]) {
1210         struct nlattr *max_attr = tb[TCA_CT_NAT_IPV6_MAX];
1211 
1212         p->ipv4_range = false;
1213         range->flags |= NF_NAT_RANGE_MAP_IPS;
1214         range->min_addr.in6 =
1215             nla_get_in6_addr(tb[TCA_CT_NAT_IPV6_MIN]);
1216 
1217         range->max_addr.in6 = max_attr ?
1218                       nla_get_in6_addr(max_attr) :
1219                       range->min_addr.in6;
1220     }
1221 
1222     if (tb[TCA_CT_NAT_PORT_MIN]) {
1223         range->flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1224         range->min_proto.all = nla_get_be16(tb[TCA_CT_NAT_PORT_MIN]);
1225 
1226         range->max_proto.all = tb[TCA_CT_NAT_PORT_MAX] ?
1227                        nla_get_be16(tb[TCA_CT_NAT_PORT_MAX]) :
1228                        range->min_proto.all;
1229     }
1230 
1231     return 0;
1232 }
1233 
1234 static void tcf_ct_set_key_val(struct nlattr **tb,
1235                    void *val, int val_type,
1236                    void *mask, int mask_type,
1237                    int len)
1238 {
1239     if (!tb[val_type])
1240         return;
1241     nla_memcpy(val, tb[val_type], len);
1242 
1243     if (!mask)
1244         return;
1245 
1246     if (mask_type == TCA_CT_UNSPEC || !tb[mask_type])
1247         memset(mask, 0xff, len);
1248     else
1249         nla_memcpy(mask, tb[mask_type], len);
1250 }
1251 
1252 static int tcf_ct_fill_params(struct net *net,
1253                   struct tcf_ct_params *p,
1254                   struct tc_ct *parm,
1255                   struct nlattr **tb,
1256                   struct netlink_ext_ack *extack)
1257 {
1258     struct tc_ct_action_net *tn = net_generic(net, ct_net_id);
1259     struct nf_conntrack_zone zone;
1260     struct nf_conn *tmpl;
1261     int err;
1262 
1263     p->zone = NF_CT_DEFAULT_ZONE_ID;
1264 
1265     tcf_ct_set_key_val(tb,
1266                &p->ct_action, TCA_CT_ACTION,
1267                NULL, TCA_CT_UNSPEC,
1268                sizeof(p->ct_action));
1269 
1270     if (p->ct_action & TCA_CT_ACT_CLEAR)
1271         return 0;
1272 
1273     err = tcf_ct_fill_params_nat(p, parm, tb, extack);
1274     if (err)
1275         return err;
1276 
1277     if (tb[TCA_CT_MARK]) {
1278         if (!IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)) {
1279             NL_SET_ERR_MSG_MOD(extack, "Conntrack mark isn't enabled.");
1280             return -EOPNOTSUPP;
1281         }
1282         tcf_ct_set_key_val(tb,
1283                    &p->mark, TCA_CT_MARK,
1284                    &p->mark_mask, TCA_CT_MARK_MASK,
1285                    sizeof(p->mark));
1286     }
1287 
1288     if (tb[TCA_CT_LABELS]) {
1289         if (!IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS)) {
1290             NL_SET_ERR_MSG_MOD(extack, "Conntrack labels isn't enabled.");
1291             return -EOPNOTSUPP;
1292         }
1293 
1294         if (!tn->labels) {
1295             NL_SET_ERR_MSG_MOD(extack, "Failed to set connlabel length");
1296             return -EOPNOTSUPP;
1297         }
1298         tcf_ct_set_key_val(tb,
1299                    p->labels, TCA_CT_LABELS,
1300                    p->labels_mask, TCA_CT_LABELS_MASK,
1301                    sizeof(p->labels));
1302     }
1303 
1304     if (tb[TCA_CT_ZONE]) {
1305         if (!IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)) {
1306             NL_SET_ERR_MSG_MOD(extack, "Conntrack zones isn't enabled.");
1307             return -EOPNOTSUPP;
1308         }
1309 
1310         tcf_ct_set_key_val(tb,
1311                    &p->zone, TCA_CT_ZONE,
1312                    NULL, TCA_CT_UNSPEC,
1313                    sizeof(p->zone));
1314     }
1315 
1316     nf_ct_zone_init(&zone, p->zone, NF_CT_DEFAULT_ZONE_DIR, 0);
1317     tmpl = nf_ct_tmpl_alloc(net, &zone, GFP_KERNEL);
1318     if (!tmpl) {
1319         NL_SET_ERR_MSG_MOD(extack, "Failed to allocate conntrack template");
1320         return -ENOMEM;
1321     }
1322     __set_bit(IPS_CONFIRMED_BIT, &tmpl->status);
1323     p->tmpl = tmpl;
1324 
1325     return 0;
1326 }
1327 
1328 static int tcf_ct_init(struct net *net, struct nlattr *nla,
1329                struct nlattr *est, struct tc_action **a,
1330                struct tcf_proto *tp, u32 flags,
1331                struct netlink_ext_ack *extack)
1332 {
1333     struct tc_action_net *tn = net_generic(net, ct_net_id);
1334     bool bind = flags & TCA_ACT_FLAGS_BIND;
1335     struct tcf_ct_params *params = NULL;
1336     struct nlattr *tb[TCA_CT_MAX + 1];
1337     struct tcf_chain *goto_ch = NULL;
1338     struct tc_ct *parm;
1339     struct tcf_ct *c;
1340     int err, res = 0;
1341     u32 index;
1342 
1343     if (!nla) {
1344         NL_SET_ERR_MSG_MOD(extack, "Ct requires attributes to be passed");
1345         return -EINVAL;
1346     }
1347 
1348     err = nla_parse_nested(tb, TCA_CT_MAX, nla, ct_policy, extack);
1349     if (err < 0)
1350         return err;
1351 
1352     if (!tb[TCA_CT_PARMS]) {
1353         NL_SET_ERR_MSG_MOD(extack, "Missing required ct parameters");
1354         return -EINVAL;
1355     }
1356     parm = nla_data(tb[TCA_CT_PARMS]);
1357     index = parm->index;
1358     err = tcf_idr_check_alloc(tn, &index, a, bind);
1359     if (err < 0)
1360         return err;
1361 
1362     if (!err) {
1363         err = tcf_idr_create_from_flags(tn, index, est, a,
1364                         &act_ct_ops, bind, flags);
1365         if (err) {
1366             tcf_idr_cleanup(tn, index);
1367             return err;
1368         }
1369         res = ACT_P_CREATED;
1370     } else {
1371         if (bind)
1372             return 0;
1373 
1374         if (!(flags & TCA_ACT_FLAGS_REPLACE)) {
1375             tcf_idr_release(*a, bind);
1376             return -EEXIST;
1377         }
1378     }
1379     err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
1380     if (err < 0)
1381         goto cleanup;
1382 
1383     c = to_ct(*a);
1384 
1385     params = kzalloc(sizeof(*params), GFP_KERNEL);
1386     if (unlikely(!params)) {
1387         err = -ENOMEM;
1388         goto cleanup;
1389     }
1390 
1391     err = tcf_ct_fill_params(net, params, parm, tb, extack);
1392     if (err)
1393         goto cleanup;
1394 
1395     err = tcf_ct_flow_table_get(net, params);
1396     if (err)
1397         goto cleanup_params;
1398 
1399     spin_lock_bh(&c->tcf_lock);
1400     goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
1401     params = rcu_replace_pointer(c->params, params,
1402                      lockdep_is_held(&c->tcf_lock));
1403     spin_unlock_bh(&c->tcf_lock);
1404 
1405     if (goto_ch)
1406         tcf_chain_put_by_act(goto_ch);
1407     if (params)
1408         call_rcu(&params->rcu, tcf_ct_params_free);
1409 
1410     return res;
1411 
1412 cleanup_params:
1413     if (params->tmpl)
1414         nf_ct_put(params->tmpl);
1415 cleanup:
1416     if (goto_ch)
1417         tcf_chain_put_by_act(goto_ch);
1418     kfree(params);
1419     tcf_idr_release(*a, bind);
1420     return err;
1421 }
1422 
1423 static void tcf_ct_cleanup(struct tc_action *a)
1424 {
1425     struct tcf_ct_params *params;
1426     struct tcf_ct *c = to_ct(a);
1427 
1428     params = rcu_dereference_protected(c->params, 1);
1429     if (params)
1430         call_rcu(&params->rcu, tcf_ct_params_free);
1431 }
1432 
1433 static int tcf_ct_dump_key_val(struct sk_buff *skb,
1434                    void *val, int val_type,
1435                    void *mask, int mask_type,
1436                    int len)
1437 {
1438     int err;
1439 
1440     if (mask && !memchr_inv(mask, 0, len))
1441         return 0;
1442 
1443     err = nla_put(skb, val_type, len, val);
1444     if (err)
1445         return err;
1446 
1447     if (mask_type != TCA_CT_UNSPEC) {
1448         err = nla_put(skb, mask_type, len, mask);
1449         if (err)
1450             return err;
1451     }
1452 
1453     return 0;
1454 }
1455 
1456 static int tcf_ct_dump_nat(struct sk_buff *skb, struct tcf_ct_params *p)
1457 {
1458     struct nf_nat_range2 *range = &p->range;
1459 
1460     if (!(p->ct_action & TCA_CT_ACT_NAT))
1461         return 0;
1462 
1463     if (!(p->ct_action & (TCA_CT_ACT_NAT_SRC | TCA_CT_ACT_NAT_DST)))
1464         return 0;
1465 
1466     if (range->flags & NF_NAT_RANGE_MAP_IPS) {
1467         if (p->ipv4_range) {
1468             if (nla_put_in_addr(skb, TCA_CT_NAT_IPV4_MIN,
1469                         range->min_addr.ip))
1470                 return -1;
1471             if (nla_put_in_addr(skb, TCA_CT_NAT_IPV4_MAX,
1472                         range->max_addr.ip))
1473                 return -1;
1474         } else {
1475             if (nla_put_in6_addr(skb, TCA_CT_NAT_IPV6_MIN,
1476                          &range->min_addr.in6))
1477                 return -1;
1478             if (nla_put_in6_addr(skb, TCA_CT_NAT_IPV6_MAX,
1479                          &range->max_addr.in6))
1480                 return -1;
1481         }
1482     }
1483 
1484     if (range->flags & NF_NAT_RANGE_PROTO_SPECIFIED) {
1485         if (nla_put_be16(skb, TCA_CT_NAT_PORT_MIN,
1486                  range->min_proto.all))
1487             return -1;
1488         if (nla_put_be16(skb, TCA_CT_NAT_PORT_MAX,
1489                  range->max_proto.all))
1490             return -1;
1491     }
1492 
1493     return 0;
1494 }
1495 
1496 static inline int tcf_ct_dump(struct sk_buff *skb, struct tc_action *a,
1497                   int bind, int ref)
1498 {
1499     unsigned char *b = skb_tail_pointer(skb);
1500     struct tcf_ct *c = to_ct(a);
1501     struct tcf_ct_params *p;
1502 
1503     struct tc_ct opt = {
1504         .index   = c->tcf_index,
1505         .refcnt  = refcount_read(&c->tcf_refcnt) - ref,
1506         .bindcnt = atomic_read(&c->tcf_bindcnt) - bind,
1507     };
1508     struct tcf_t t;
1509 
1510     spin_lock_bh(&c->tcf_lock);
1511     p = rcu_dereference_protected(c->params,
1512                       lockdep_is_held(&c->tcf_lock));
1513     opt.action = c->tcf_action;
1514 
1515     if (tcf_ct_dump_key_val(skb,
1516                 &p->ct_action, TCA_CT_ACTION,
1517                 NULL, TCA_CT_UNSPEC,
1518                 sizeof(p->ct_action)))
1519         goto nla_put_failure;
1520 
1521     if (p->ct_action & TCA_CT_ACT_CLEAR)
1522         goto skip_dump;
1523 
1524     if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
1525         tcf_ct_dump_key_val(skb,
1526                 &p->mark, TCA_CT_MARK,
1527                 &p->mark_mask, TCA_CT_MARK_MASK,
1528                 sizeof(p->mark)))
1529         goto nla_put_failure;
1530 
1531     if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
1532         tcf_ct_dump_key_val(skb,
1533                 p->labels, TCA_CT_LABELS,
1534                 p->labels_mask, TCA_CT_LABELS_MASK,
1535                 sizeof(p->labels)))
1536         goto nla_put_failure;
1537 
1538     if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1539         tcf_ct_dump_key_val(skb,
1540                 &p->zone, TCA_CT_ZONE,
1541                 NULL, TCA_CT_UNSPEC,
1542                 sizeof(p->zone)))
1543         goto nla_put_failure;
1544 
1545     if (tcf_ct_dump_nat(skb, p))
1546         goto nla_put_failure;
1547 
1548 skip_dump:
1549     if (nla_put(skb, TCA_CT_PARMS, sizeof(opt), &opt))
1550         goto nla_put_failure;
1551 
1552     tcf_tm_dump(&t, &c->tcf_tm);
1553     if (nla_put_64bit(skb, TCA_CT_TM, sizeof(t), &t, TCA_CT_PAD))
1554         goto nla_put_failure;
1555     spin_unlock_bh(&c->tcf_lock);
1556 
1557     return skb->len;
1558 nla_put_failure:
1559     spin_unlock_bh(&c->tcf_lock);
1560     nlmsg_trim(skb, b);
1561     return -1;
1562 }
1563 
1564 static int tcf_ct_walker(struct net *net, struct sk_buff *skb,
1565              struct netlink_callback *cb, int type,
1566              const struct tc_action_ops *ops,
1567              struct netlink_ext_ack *extack)
1568 {
1569     struct tc_action_net *tn = net_generic(net, ct_net_id);
1570 
1571     return tcf_generic_walker(tn, skb, cb, type, ops, extack);
1572 }
1573 
1574 static int tcf_ct_search(struct net *net, struct tc_action **a, u32 index)
1575 {
1576     struct tc_action_net *tn = net_generic(net, ct_net_id);
1577 
1578     return tcf_idr_search(tn, a, index);
1579 }
1580 
1581 static void tcf_stats_update(struct tc_action *a, u64 bytes, u64 packets,
1582                  u64 drops, u64 lastuse, bool hw)
1583 {
1584     struct tcf_ct *c = to_ct(a);
1585 
1586     tcf_action_update_stats(a, bytes, packets, drops, hw);
1587     c->tcf_tm.lastuse = max_t(u64, c->tcf_tm.lastuse, lastuse);
1588 }
1589 
1590 static int tcf_ct_offload_act_setup(struct tc_action *act, void *entry_data,
1591                     u32 *index_inc, bool bind,
1592                     struct netlink_ext_ack *extack)
1593 {
1594     if (bind) {
1595         struct flow_action_entry *entry = entry_data;
1596 
1597         entry->id = FLOW_ACTION_CT;
1598         entry->ct.action = tcf_ct_action(act);
1599         entry->ct.zone = tcf_ct_zone(act);
1600         entry->ct.flow_table = tcf_ct_ft(act);
1601         *index_inc = 1;
1602     } else {
1603         struct flow_offload_action *fl_action = entry_data;
1604 
1605         fl_action->id = FLOW_ACTION_CT;
1606     }
1607 
1608     return 0;
1609 }
1610 
1611 static struct tc_action_ops act_ct_ops = {
1612     .kind       =   "ct",
1613     .id     =   TCA_ID_CT,
1614     .owner      =   THIS_MODULE,
1615     .act        =   tcf_ct_act,
1616     .dump       =   tcf_ct_dump,
1617     .init       =   tcf_ct_init,
1618     .cleanup    =   tcf_ct_cleanup,
1619     .walk       =   tcf_ct_walker,
1620     .lookup     =   tcf_ct_search,
1621     .stats_update   =   tcf_stats_update,
1622     .offload_act_setup =    tcf_ct_offload_act_setup,
1623     .size       =   sizeof(struct tcf_ct),
1624 };
1625 
1626 static __net_init int ct_init_net(struct net *net)
1627 {
1628     unsigned int n_bits = sizeof_field(struct tcf_ct_params, labels) * 8;
1629     struct tc_ct_action_net *tn = net_generic(net, ct_net_id);
1630 
1631     if (nf_connlabels_get(net, n_bits - 1)) {
1632         tn->labels = false;
1633         pr_err("act_ct: Failed to set connlabels length");
1634     } else {
1635         tn->labels = true;
1636     }
1637 
1638     return tc_action_net_init(net, &tn->tn, &act_ct_ops);
1639 }
1640 
1641 static void __net_exit ct_exit_net(struct list_head *net_list)
1642 {
1643     struct net *net;
1644 
1645     rtnl_lock();
1646     list_for_each_entry(net, net_list, exit_list) {
1647         struct tc_ct_action_net *tn = net_generic(net, ct_net_id);
1648 
1649         if (tn->labels)
1650             nf_connlabels_put(net);
1651     }
1652     rtnl_unlock();
1653 
1654     tc_action_net_exit(net_list, ct_net_id);
1655 }
1656 
1657 static struct pernet_operations ct_net_ops = {
1658     .init = ct_init_net,
1659     .exit_batch = ct_exit_net,
1660     .id   = &ct_net_id,
1661     .size = sizeof(struct tc_ct_action_net),
1662 };
1663 
1664 static int __init ct_init_module(void)
1665 {
1666     int err;
1667 
1668     act_ct_wq = alloc_ordered_workqueue("act_ct_workqueue", 0);
1669     if (!act_ct_wq)
1670         return -ENOMEM;
1671 
1672     err = tcf_ct_flow_tables_init();
1673     if (err)
1674         goto err_tbl_init;
1675 
1676     err = tcf_register_action(&act_ct_ops, &ct_net_ops);
1677     if (err)
1678         goto err_register;
1679 
1680     static_branch_inc(&tcf_frag_xmit_count);
1681 
1682     return 0;
1683 
1684 err_register:
1685     tcf_ct_flow_tables_uninit();
1686 err_tbl_init:
1687     destroy_workqueue(act_ct_wq);
1688     return err;
1689 }
1690 
1691 static void __exit ct_cleanup_module(void)
1692 {
1693     static_branch_dec(&tcf_frag_xmit_count);
1694     tcf_unregister_action(&act_ct_ops, &ct_net_ops);
1695     tcf_ct_flow_tables_uninit();
1696     destroy_workqueue(act_ct_wq);
1697 }
1698 
1699 module_init(ct_init_module);
1700 module_exit(ct_cleanup_module);
1701 MODULE_AUTHOR("Paul Blakey <paulb@mellanox.com>");
1702 MODULE_AUTHOR("Yossi Kuperman <yossiku@mellanox.com>");
1703 MODULE_AUTHOR("Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>");
1704 MODULE_DESCRIPTION("Connection tracking action");
1705 MODULE_LICENSE("GPL v2");
[ Source navigation ] [ Diff markup ] [ Identifier search ] [ general search ]

This page was automatically generated by the 2.1.0 LXR engine.

The LXR team
Valid CSS 2.1! Valid HTML 4.01!