1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* Copyright (c) 2021 Taehee Yoo <ap420073@gmail.com> */
4 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
6 #include <linux/module.h>
7 #include <linux/skbuff.h>
9 #include <linux/jhash.h>
10 #include <linux/if_tunnel.h>
11 #include <linux/net.h>
12 #include <linux/igmp.h>
13 #include <linux/workqueue.h>
14 #include <net/sch_generic.h>
15 #include <net/net_namespace.h>
18 #include <net/udp_tunnel.h>
22 #include <uapi/linux/amt.h>
23 #include <linux/security.h>
24 #include <net/gro_cells.h>
26 #include <net/if_inet6.h>
27 #include <net/ndisc.h>
28 #include <net/addrconf.h>
29 #include <net/ip6_route.h>
30 #include <net/inet_common.h>
31 #include <net/ip6_checksum.h>
33 static struct workqueue_struct *amt_wq;
35 static HLIST_HEAD(source_gc_list);
36 /* Lock for source_gc_list */
37 static spinlock_t source_gc_lock;
38 static struct delayed_work source_gc_wq;
39 static char *status_str[] = {
41 "AMT_STATUS_SENT_DISCOVERY",
42 "AMT_STATUS_RECEIVED_DISCOVERY",
43 "AMT_STATUS_SENT_ADVERTISEMENT",
44 "AMT_STATUS_RECEIVED_ADVERTISEMENT",
45 "AMT_STATUS_SENT_REQUEST",
46 "AMT_STATUS_RECEIVED_REQUEST",
47 "AMT_STATUS_SENT_QUERY",
48 "AMT_STATUS_RECEIVED_QUERY",
49 "AMT_STATUS_SENT_UPDATE",
50 "AMT_STATUS_RECEIVED_UPDATE",
53 static char *type_str[] = {
54 "", /* Type 0 is not defined */
56 "AMT_MSG_ADVERTISEMENT",
58 "AMT_MSG_MEMBERSHIP_QUERY",
59 "AMT_MSG_MEMBERSHIP_UPDATE",
60 "AMT_MSG_MULTICAST_DATA",
64 static char *action_str[] = {
68 "AMT_ACT_STATUS_FWD_NEW",
69 "AMT_ACT_STATUS_D_FWD_NEW",
70 "AMT_ACT_STATUS_NONE_NEW",
73 static struct igmpv3_grec igmpv3_zero_grec;
75 #if IS_ENABLED(CONFIG_IPV6)
76 #define MLD2_ALL_NODE_INIT { { { 0xff, 0x02, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0x01 } } }
77 static struct in6_addr mld2_all_node = MLD2_ALL_NODE_INIT;
78 static struct mld2_grec mldv2_zero_grec;
81 static struct amt_skb_cb *amt_skb_cb(struct sk_buff *skb)
83 BUILD_BUG_ON(sizeof(struct amt_skb_cb) + sizeof(struct qdisc_skb_cb) >
84 sizeof_field(struct sk_buff, cb));
86 return (struct amt_skb_cb *)((void *)skb->cb +
87 sizeof(struct qdisc_skb_cb));
90 static void __amt_source_gc_work(void)
92 struct amt_source_node *snode;
93 struct hlist_head gc_list;
96 spin_lock_bh(&source_gc_lock);
97 hlist_move_list(&source_gc_list, &gc_list);
98 spin_unlock_bh(&source_gc_lock);
100 hlist_for_each_entry_safe(snode, t, &gc_list, node) {
101 hlist_del_rcu(&snode->node);
102 kfree_rcu(snode, rcu);
106 static void amt_source_gc_work(struct work_struct *work)
108 __amt_source_gc_work();
110 spin_lock_bh(&source_gc_lock);
111 mod_delayed_work(amt_wq, &source_gc_wq,
112 msecs_to_jiffies(AMT_GC_INTERVAL));
113 spin_unlock_bh(&source_gc_lock);
116 static bool amt_addr_equal(union amt_addr *a, union amt_addr *b)
118 return !memcmp(a, b, sizeof(union amt_addr));
121 static u32 amt_source_hash(struct amt_tunnel_list *tunnel, union amt_addr *src)
123 u32 hash = jhash(src, sizeof(*src), tunnel->amt->hash_seed);
125 return reciprocal_scale(hash, tunnel->amt->hash_buckets);
128 static bool amt_status_filter(struct amt_source_node *snode,
129 enum amt_filter filter)
135 if (snode->status == AMT_SOURCE_STATUS_FWD &&
136 snode->flags == AMT_SOURCE_OLD)
139 case AMT_FILTER_D_FWD:
140 if (snode->status == AMT_SOURCE_STATUS_D_FWD &&
141 snode->flags == AMT_SOURCE_OLD)
144 case AMT_FILTER_FWD_NEW:
145 if (snode->status == AMT_SOURCE_STATUS_FWD &&
146 snode->flags == AMT_SOURCE_NEW)
149 case AMT_FILTER_D_FWD_NEW:
150 if (snode->status == AMT_SOURCE_STATUS_D_FWD &&
151 snode->flags == AMT_SOURCE_NEW)
157 case AMT_FILTER_NONE_NEW:
158 if (snode->status == AMT_SOURCE_STATUS_NONE &&
159 snode->flags == AMT_SOURCE_NEW)
162 case AMT_FILTER_BOTH:
163 if ((snode->status == AMT_SOURCE_STATUS_D_FWD ||
164 snode->status == AMT_SOURCE_STATUS_FWD) &&
165 snode->flags == AMT_SOURCE_OLD)
168 case AMT_FILTER_BOTH_NEW:
169 if ((snode->status == AMT_SOURCE_STATUS_D_FWD ||
170 snode->status == AMT_SOURCE_STATUS_FWD) &&
171 snode->flags == AMT_SOURCE_NEW)
182 static struct amt_source_node *amt_lookup_src(struct amt_tunnel_list *tunnel,
183 struct amt_group_node *gnode,
184 enum amt_filter filter,
187 u32 hash = amt_source_hash(tunnel, src);
188 struct amt_source_node *snode;
190 hlist_for_each_entry_rcu(snode, &gnode->sources[hash], node)
191 if (amt_status_filter(snode, filter) &&
192 amt_addr_equal(&snode->source_addr, src))
198 static u32 amt_group_hash(struct amt_tunnel_list *tunnel, union amt_addr *group)
200 u32 hash = jhash(group, sizeof(*group), tunnel->amt->hash_seed);
202 return reciprocal_scale(hash, tunnel->amt->hash_buckets);
205 static struct amt_group_node *amt_lookup_group(struct amt_tunnel_list *tunnel,
206 union amt_addr *group,
207 union amt_addr *host,
210 u32 hash = amt_group_hash(tunnel, group);
211 struct amt_group_node *gnode;
213 hlist_for_each_entry_rcu(gnode, &tunnel->groups[hash], node) {
214 if (amt_addr_equal(&gnode->group_addr, group) &&
215 amt_addr_equal(&gnode->host_addr, host) &&
223 static void amt_destroy_source(struct amt_source_node *snode)
225 struct amt_group_node *gnode = snode->gnode;
226 struct amt_tunnel_list *tunnel;
228 tunnel = gnode->tunnel_list;
231 netdev_dbg(snode->gnode->amt->dev,
232 "Delete source %pI4 from %pI4\n",
233 &snode->source_addr.ip4,
234 &gnode->group_addr.ip4);
235 #if IS_ENABLED(CONFIG_IPV6)
237 netdev_dbg(snode->gnode->amt->dev,
238 "Delete source %pI6 from %pI6\n",
239 &snode->source_addr.ip6,
240 &gnode->group_addr.ip6);
244 cancel_delayed_work(&snode->source_timer);
245 hlist_del_init_rcu(&snode->node);
246 tunnel->nr_sources--;
248 spin_lock_bh(&source_gc_lock);
249 hlist_add_head_rcu(&snode->node, &source_gc_list);
250 spin_unlock_bh(&source_gc_lock);
253 static void amt_del_group(struct amt_dev *amt, struct amt_group_node *gnode)
255 struct amt_source_node *snode;
256 struct hlist_node *t;
259 if (cancel_delayed_work(&gnode->group_timer))
261 hlist_del_rcu(&gnode->node);
262 gnode->tunnel_list->nr_groups--;
265 netdev_dbg(amt->dev, "Leave group %pI4\n",
266 &gnode->group_addr.ip4);
267 #if IS_ENABLED(CONFIG_IPV6)
269 netdev_dbg(amt->dev, "Leave group %pI6\n",
270 &gnode->group_addr.ip6);
272 for (i = 0; i < amt->hash_buckets; i++)
273 hlist_for_each_entry_safe(snode, t, &gnode->sources[i], node)
274 amt_destroy_source(snode);
276 /* tunnel->lock was acquired outside of amt_del_group()
277 * But rcu_read_lock() was acquired too so It's safe.
279 kfree_rcu(gnode, rcu);
282 /* If a source timer expires with a router filter-mode for the group of
283 * INCLUDE, the router concludes that traffic from this particular
284 * source is no longer desired on the attached network, and deletes the
285 * associated source record.
287 static void amt_source_work(struct work_struct *work)
289 struct amt_source_node *snode = container_of(to_delayed_work(work),
290 struct amt_source_node,
292 struct amt_group_node *gnode = snode->gnode;
293 struct amt_dev *amt = gnode->amt;
294 struct amt_tunnel_list *tunnel;
296 tunnel = gnode->tunnel_list;
297 spin_lock_bh(&tunnel->lock);
299 if (gnode->filter_mode == MCAST_INCLUDE) {
300 amt_destroy_source(snode);
301 if (!gnode->nr_sources)
302 amt_del_group(amt, gnode);
304 /* When a router filter-mode for a group is EXCLUDE,
305 * source records are only deleted when the group timer expires
307 snode->status = AMT_SOURCE_STATUS_D_FWD;
310 spin_unlock_bh(&tunnel->lock);
313 static void amt_act_src(struct amt_tunnel_list *tunnel,
314 struct amt_group_node *gnode,
315 struct amt_source_node *snode,
318 struct amt_dev *amt = tunnel->amt;
322 mod_delayed_work(amt_wq, &snode->source_timer,
323 msecs_to_jiffies(amt_gmi(amt)));
325 case AMT_ACT_GMI_ZERO:
326 cancel_delayed_work(&snode->source_timer);
329 mod_delayed_work(amt_wq, &snode->source_timer,
330 gnode->group_timer.timer.expires);
332 case AMT_ACT_STATUS_FWD_NEW:
333 snode->status = AMT_SOURCE_STATUS_FWD;
334 snode->flags = AMT_SOURCE_NEW;
336 case AMT_ACT_STATUS_D_FWD_NEW:
337 snode->status = AMT_SOURCE_STATUS_D_FWD;
338 snode->flags = AMT_SOURCE_NEW;
340 case AMT_ACT_STATUS_NONE_NEW:
341 cancel_delayed_work(&snode->source_timer);
342 snode->status = AMT_SOURCE_STATUS_NONE;
343 snode->flags = AMT_SOURCE_NEW;
351 netdev_dbg(amt->dev, "Source %pI4 from %pI4 Acted %s\n",
352 &snode->source_addr.ip4,
353 &gnode->group_addr.ip4,
355 #if IS_ENABLED(CONFIG_IPV6)
357 netdev_dbg(amt->dev, "Source %pI6 from %pI6 Acted %s\n",
358 &snode->source_addr.ip6,
359 &gnode->group_addr.ip6,
364 static struct amt_source_node *amt_alloc_snode(struct amt_group_node *gnode,
367 struct amt_source_node *snode;
369 snode = kzalloc(sizeof(*snode), GFP_ATOMIC);
373 memcpy(&snode->source_addr, src, sizeof(union amt_addr));
374 snode->gnode = gnode;
375 snode->status = AMT_SOURCE_STATUS_NONE;
376 snode->flags = AMT_SOURCE_NEW;
377 INIT_HLIST_NODE(&snode->node);
378 INIT_DELAYED_WORK(&snode->source_timer, amt_source_work);
383 /* RFC 3810 - 7.2.2. Definition of Filter Timers
385 * Router Mode Filter Timer Actions/Comments
386 * ----------- ----------------- ----------------
388 * INCLUDE Not Used All listeners in
391 * EXCLUDE Timer > 0 At least one listener
394 * EXCLUDE Timer == 0 No more listeners in
395 * EXCLUDE mode for the
397 * If the Requested List
400 * Record. If not, switch
401 * to INCLUDE filter mode;
404 * moved to the Include
405 * List, and the Exclude
408 static void amt_group_work(struct work_struct *work)
410 struct amt_group_node *gnode = container_of(to_delayed_work(work),
411 struct amt_group_node,
413 struct amt_tunnel_list *tunnel = gnode->tunnel_list;
414 struct amt_dev *amt = gnode->amt;
415 struct amt_source_node *snode;
416 bool delete_group = true;
417 struct hlist_node *t;
420 buckets = amt->hash_buckets;
422 spin_lock_bh(&tunnel->lock);
423 if (gnode->filter_mode == MCAST_INCLUDE) {
425 spin_unlock_bh(&tunnel->lock);
430 for (i = 0; i < buckets; i++) {
431 hlist_for_each_entry_safe(snode, t,
432 &gnode->sources[i], node) {
433 if (!delayed_work_pending(&snode->source_timer) ||
434 snode->status == AMT_SOURCE_STATUS_D_FWD) {
435 amt_destroy_source(snode);
437 delete_group = false;
438 snode->status = AMT_SOURCE_STATUS_FWD;
443 amt_del_group(amt, gnode);
445 gnode->filter_mode = MCAST_INCLUDE;
447 spin_unlock_bh(&tunnel->lock);
452 /* Non-existent group is created as INCLUDE {empty}:
454 * RFC 3376 - 5.1. Action on Change of Interface State
456 * If no interface state existed for that multicast address before
457 * the change (i.e., the change consisted of creating a new
458 * per-interface record), or if no state exists after the change
459 * (i.e., the change consisted of deleting a per-interface record),
460 * then the "non-existent" state is considered to have a filter mode
461 * of INCLUDE and an empty source list.
463 static struct amt_group_node *amt_add_group(struct amt_dev *amt,
464 struct amt_tunnel_list *tunnel,
465 union amt_addr *group,
466 union amt_addr *host,
469 struct amt_group_node *gnode;
473 if (tunnel->nr_groups >= amt->max_groups)
474 return ERR_PTR(-ENOSPC);
476 gnode = kzalloc(sizeof(*gnode) +
477 (sizeof(struct hlist_head) * amt->hash_buckets),
479 if (unlikely(!gnode))
480 return ERR_PTR(-ENOMEM);
483 gnode->group_addr = *group;
484 gnode->host_addr = *host;
486 gnode->tunnel_list = tunnel;
487 gnode->filter_mode = MCAST_INCLUDE;
488 INIT_HLIST_NODE(&gnode->node);
489 INIT_DELAYED_WORK(&gnode->group_timer, amt_group_work);
490 for (i = 0; i < amt->hash_buckets; i++)
491 INIT_HLIST_HEAD(&gnode->sources[i]);
493 hash = amt_group_hash(tunnel, group);
494 hlist_add_head_rcu(&gnode->node, &tunnel->groups[hash]);
498 netdev_dbg(amt->dev, "Join group %pI4\n",
499 &gnode->group_addr.ip4);
500 #if IS_ENABLED(CONFIG_IPV6)
502 netdev_dbg(amt->dev, "Join group %pI6\n",
503 &gnode->group_addr.ip6);
509 static struct sk_buff *amt_build_igmp_gq(struct amt_dev *amt)
511 u8 ra[AMT_IPHDR_OPTS] = { IPOPT_RA, 4, 0, 0 };
512 int hlen = LL_RESERVED_SPACE(amt->dev);
513 int tlen = amt->dev->needed_tailroom;
514 struct igmpv3_query *ihv3;
515 void *csum_start = NULL;
516 __sum16 *csum = NULL;
523 len = hlen + tlen + sizeof(*iph) + AMT_IPHDR_OPTS + sizeof(*ihv3);
524 skb = netdev_alloc_skb_ip_align(amt->dev, len);
528 skb_reserve(skb, hlen);
529 skb_push(skb, sizeof(*eth));
530 skb->protocol = htons(ETH_P_IP);
531 skb_reset_mac_header(skb);
532 skb->priority = TC_PRIO_CONTROL;
533 skb_put(skb, sizeof(*iph));
534 skb_put_data(skb, ra, sizeof(ra));
535 skb_put(skb, sizeof(*ihv3));
536 skb_pull(skb, sizeof(*eth));
537 skb_reset_network_header(skb);
541 iph->ihl = (sizeof(struct iphdr) + AMT_IPHDR_OPTS) >> 2;
543 iph->tot_len = htons(sizeof(*iph) + AMT_IPHDR_OPTS + sizeof(*ihv3));
544 iph->frag_off = htons(IP_DF);
547 iph->protocol = IPPROTO_IGMP;
548 iph->daddr = htonl(INADDR_ALLHOSTS_GROUP);
549 iph->saddr = htonl(INADDR_ANY);
553 ether_addr_copy(eth->h_source, amt->dev->dev_addr);
554 ip_eth_mc_map(htonl(INADDR_ALLHOSTS_GROUP), eth->h_dest);
555 eth->h_proto = htons(ETH_P_IP);
557 ihv3 = skb_pull(skb, sizeof(*iph) + AMT_IPHDR_OPTS);
558 skb_reset_transport_header(skb);
559 ihv3->type = IGMP_HOST_MEMBERSHIP_QUERY;
562 ihv3->qqic = amt->qi;
565 ihv3->suppress = false;
566 ihv3->qrv = READ_ONCE(amt->net->ipv4.sysctl_igmp_qrv);
569 csum_start = (void *)ihv3;
570 *csum = ip_compute_csum(csum_start, sizeof(*ihv3));
571 offset = skb_transport_offset(skb);
572 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
573 skb->ip_summed = CHECKSUM_NONE;
575 skb_push(skb, sizeof(*eth) + sizeof(*iph) + AMT_IPHDR_OPTS);
580 static void amt_update_gw_status(struct amt_dev *amt, enum amt_status status,
583 if (validate && amt->status >= status)
585 netdev_dbg(amt->dev, "Update GW status %s -> %s",
586 status_str[amt->status], status_str[status]);
587 WRITE_ONCE(amt->status, status);
590 static void __amt_update_relay_status(struct amt_tunnel_list *tunnel,
591 enum amt_status status,
594 if (validate && tunnel->status >= status)
596 netdev_dbg(tunnel->amt->dev,
597 "Update Tunnel(IP = %pI4, PORT = %u) status %s -> %s",
598 &tunnel->ip4, ntohs(tunnel->source_port),
599 status_str[tunnel->status], status_str[status]);
600 tunnel->status = status;
603 static void amt_update_relay_status(struct amt_tunnel_list *tunnel,
604 enum amt_status status, bool validate)
606 spin_lock_bh(&tunnel->lock);
607 __amt_update_relay_status(tunnel, status, validate);
608 spin_unlock_bh(&tunnel->lock);
611 static void amt_send_discovery(struct amt_dev *amt)
613 struct amt_header_discovery *amtd;
614 int hlen, tlen, offset;
625 sock = rcu_dereference(amt->sock);
629 if (!netif_running(amt->stream_dev) || !netif_running(amt->dev))
632 rt = ip_route_output_ports(amt->net, &fl4, sock->sk,
633 amt->discovery_ip, amt->local_ip,
634 amt->gw_port, amt->relay_port,
636 amt->stream_dev->ifindex);
638 amt->dev->stats.tx_errors++;
642 hlen = LL_RESERVED_SPACE(amt->dev);
643 tlen = amt->dev->needed_tailroom;
644 len = hlen + tlen + sizeof(*iph) + sizeof(*udph) + sizeof(*amtd);
645 skb = netdev_alloc_skb_ip_align(amt->dev, len);
648 amt->dev->stats.tx_errors++;
652 skb->priority = TC_PRIO_CONTROL;
653 skb_dst_set(skb, &rt->dst);
655 len = sizeof(*iph) + sizeof(*udph) + sizeof(*amtd);
656 skb_reset_network_header(skb);
658 amtd = skb_pull(skb, sizeof(*iph) + sizeof(*udph));
660 amtd->type = AMT_MSG_DISCOVERY;
662 amtd->nonce = amt->nonce;
663 skb_push(skb, sizeof(*udph));
664 skb_reset_transport_header(skb);
666 udph->source = amt->gw_port;
667 udph->dest = amt->relay_port;
668 udph->len = htons(sizeof(*udph) + sizeof(*amtd));
670 offset = skb_transport_offset(skb);
671 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
672 udph->check = csum_tcpudp_magic(amt->local_ip, amt->discovery_ip,
673 sizeof(*udph) + sizeof(*amtd),
674 IPPROTO_UDP, skb->csum);
676 skb_push(skb, sizeof(*iph));
679 iph->ihl = (sizeof(struct iphdr)) >> 2;
682 iph->ttl = ip4_dst_hoplimit(&rt->dst);
683 iph->daddr = amt->discovery_ip;
684 iph->saddr = amt->local_ip;
685 iph->protocol = IPPROTO_UDP;
686 iph->tot_len = htons(len);
688 skb->ip_summed = CHECKSUM_NONE;
689 ip_select_ident(amt->net, skb, NULL);
691 err = ip_local_out(amt->net, sock->sk, skb);
692 if (unlikely(net_xmit_eval(err)))
693 amt->dev->stats.tx_errors++;
695 amt_update_gw_status(amt, AMT_STATUS_SENT_DISCOVERY, true);
700 static void amt_send_request(struct amt_dev *amt, bool v6)
702 struct amt_header_request *amtrh;
703 int hlen, tlen, offset;
714 sock = rcu_dereference(amt->sock);
718 if (!netif_running(amt->stream_dev) || !netif_running(amt->dev))
721 rt = ip_route_output_ports(amt->net, &fl4, sock->sk,
722 amt->remote_ip, amt->local_ip,
723 amt->gw_port, amt->relay_port,
725 amt->stream_dev->ifindex);
727 amt->dev->stats.tx_errors++;
731 hlen = LL_RESERVED_SPACE(amt->dev);
732 tlen = amt->dev->needed_tailroom;
733 len = hlen + tlen + sizeof(*iph) + sizeof(*udph) + sizeof(*amtrh);
734 skb = netdev_alloc_skb_ip_align(amt->dev, len);
737 amt->dev->stats.tx_errors++;
741 skb->priority = TC_PRIO_CONTROL;
742 skb_dst_set(skb, &rt->dst);
744 len = sizeof(*iph) + sizeof(*udph) + sizeof(*amtrh);
745 skb_reset_network_header(skb);
747 amtrh = skb_pull(skb, sizeof(*iph) + sizeof(*udph));
749 amtrh->type = AMT_MSG_REQUEST;
750 amtrh->reserved1 = 0;
752 amtrh->reserved2 = 0;
753 amtrh->nonce = amt->nonce;
754 skb_push(skb, sizeof(*udph));
755 skb_reset_transport_header(skb);
757 udph->source = amt->gw_port;
758 udph->dest = amt->relay_port;
759 udph->len = htons(sizeof(*amtrh) + sizeof(*udph));
761 offset = skb_transport_offset(skb);
762 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
763 udph->check = csum_tcpudp_magic(amt->local_ip, amt->remote_ip,
764 sizeof(*udph) + sizeof(*amtrh),
765 IPPROTO_UDP, skb->csum);
767 skb_push(skb, sizeof(*iph));
770 iph->ihl = (sizeof(struct iphdr)) >> 2;
773 iph->ttl = ip4_dst_hoplimit(&rt->dst);
774 iph->daddr = amt->remote_ip;
775 iph->saddr = amt->local_ip;
776 iph->protocol = IPPROTO_UDP;
777 iph->tot_len = htons(len);
779 skb->ip_summed = CHECKSUM_NONE;
780 ip_select_ident(amt->net, skb, NULL);
782 err = ip_local_out(amt->net, sock->sk, skb);
783 if (unlikely(net_xmit_eval(err)))
784 amt->dev->stats.tx_errors++;
790 static void amt_send_igmp_gq(struct amt_dev *amt,
791 struct amt_tunnel_list *tunnel)
795 skb = amt_build_igmp_gq(amt);
799 amt_skb_cb(skb)->tunnel = tunnel;
803 #if IS_ENABLED(CONFIG_IPV6)
804 static struct sk_buff *amt_build_mld_gq(struct amt_dev *amt)
806 u8 ra[AMT_IP6HDR_OPTS] = { IPPROTO_ICMPV6, 0, IPV6_TLV_ROUTERALERT,
807 2, 0, 0, IPV6_TLV_PAD1, IPV6_TLV_PAD1 };
808 int hlen = LL_RESERVED_SPACE(amt->dev);
809 int tlen = amt->dev->needed_tailroom;
810 struct mld2_query *mld2q;
811 void *csum_start = NULL;
812 struct ipv6hdr *ip6h;
817 len = hlen + tlen + sizeof(*ip6h) + sizeof(ra) + sizeof(*mld2q);
818 skb = netdev_alloc_skb_ip_align(amt->dev, len);
822 skb_reserve(skb, hlen);
823 skb_push(skb, sizeof(*eth));
824 skb_reset_mac_header(skb);
826 skb->priority = TC_PRIO_CONTROL;
827 skb->protocol = htons(ETH_P_IPV6);
828 skb_put_zero(skb, sizeof(*ip6h));
829 skb_put_data(skb, ra, sizeof(ra));
830 skb_put_zero(skb, sizeof(*mld2q));
831 skb_pull(skb, sizeof(*eth));
832 skb_reset_network_header(skb);
833 ip6h = ipv6_hdr(skb);
834 ip6h->payload_len = htons(sizeof(ra) + sizeof(*mld2q));
835 ip6h->nexthdr = NEXTHDR_HOP;
837 ip6h->daddr = mld2_all_node;
838 ip6_flow_hdr(ip6h, 0, 0);
840 if (ipv6_dev_get_saddr(amt->net, amt->dev, &ip6h->daddr, 0,
842 amt->dev->stats.tx_errors++;
847 eth->h_proto = htons(ETH_P_IPV6);
848 ether_addr_copy(eth->h_source, amt->dev->dev_addr);
849 ipv6_eth_mc_map(&mld2_all_node, eth->h_dest);
851 skb_pull(skb, sizeof(*ip6h) + sizeof(ra));
852 skb_reset_transport_header(skb);
853 mld2q = (struct mld2_query *)icmp6_hdr(skb);
854 mld2q->mld2q_mrc = htons(1);
855 mld2q->mld2q_type = ICMPV6_MGM_QUERY;
856 mld2q->mld2q_code = 0;
857 mld2q->mld2q_cksum = 0;
858 mld2q->mld2q_resv1 = 0;
859 mld2q->mld2q_resv2 = 0;
860 mld2q->mld2q_suppress = 0;
861 mld2q->mld2q_qrv = amt->qrv;
862 mld2q->mld2q_nsrcs = 0;
863 mld2q->mld2q_qqic = amt->qi;
864 csum_start = (void *)mld2q;
865 mld2q->mld2q_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
868 csum_partial(csum_start,
871 skb->ip_summed = CHECKSUM_NONE;
872 skb_push(skb, sizeof(*eth) + sizeof(*ip6h) + sizeof(ra));
876 static void amt_send_mld_gq(struct amt_dev *amt, struct amt_tunnel_list *tunnel)
880 skb = amt_build_mld_gq(amt);
884 amt_skb_cb(skb)->tunnel = tunnel;
888 static void amt_send_mld_gq(struct amt_dev *amt, struct amt_tunnel_list *tunnel)
893 static bool amt_queue_event(struct amt_dev *amt, enum amt_event event,
898 spin_lock_bh(&amt->lock);
899 if (amt->nr_events >= AMT_MAX_EVENTS) {
900 spin_unlock_bh(&amt->lock);
904 index = (amt->event_idx + amt->nr_events) % AMT_MAX_EVENTS;
905 amt->events[index].event = event;
906 amt->events[index].skb = skb;
908 amt->event_idx %= AMT_MAX_EVENTS;
909 queue_work(amt_wq, &amt->event_wq);
910 spin_unlock_bh(&amt->lock);
915 static void amt_secret_work(struct work_struct *work)
917 struct amt_dev *amt = container_of(to_delayed_work(work),
921 spin_lock_bh(&amt->lock);
922 get_random_bytes(&amt->key, sizeof(siphash_key_t));
923 spin_unlock_bh(&amt->lock);
924 mod_delayed_work(amt_wq, &amt->secret_wq,
925 msecs_to_jiffies(AMT_SECRET_TIMEOUT));
928 static void amt_event_send_discovery(struct amt_dev *amt)
930 if (amt->status > AMT_STATUS_SENT_DISCOVERY)
932 get_random_bytes(&amt->nonce, sizeof(__be32));
934 amt_send_discovery(amt);
936 mod_delayed_work(amt_wq, &amt->discovery_wq,
937 msecs_to_jiffies(AMT_DISCOVERY_TIMEOUT));
940 static void amt_discovery_work(struct work_struct *work)
942 struct amt_dev *amt = container_of(to_delayed_work(work),
946 if (amt_queue_event(amt, AMT_EVENT_SEND_DISCOVERY, NULL))
947 mod_delayed_work(amt_wq, &amt->discovery_wq,
948 msecs_to_jiffies(AMT_DISCOVERY_TIMEOUT));
951 static void amt_event_send_request(struct amt_dev *amt)
955 if (amt->status < AMT_STATUS_RECEIVED_ADVERTISEMENT)
958 if (amt->req_cnt > AMT_MAX_REQ_COUNT) {
959 netdev_dbg(amt->dev, "Gateway is not ready");
960 amt->qi = AMT_INIT_REQ_TIMEOUT;
961 WRITE_ONCE(amt->ready4, false);
962 WRITE_ONCE(amt->ready6, false);
964 amt_update_gw_status(amt, AMT_STATUS_INIT, false);
971 WRITE_ONCE(amt->ready4, false);
972 WRITE_ONCE(amt->ready6, false);
973 get_random_bytes(&amt->nonce, sizeof(__be32));
976 amt_send_request(amt, false);
977 amt_send_request(amt, true);
978 amt_update_gw_status(amt, AMT_STATUS_SENT_REQUEST, true);
981 exp = min_t(u32, (1 * (1 << amt->req_cnt)), AMT_MAX_REQ_TIMEOUT);
982 mod_delayed_work(amt_wq, &amt->req_wq, msecs_to_jiffies(exp * 1000));
985 static void amt_req_work(struct work_struct *work)
987 struct amt_dev *amt = container_of(to_delayed_work(work),
991 if (amt_queue_event(amt, AMT_EVENT_SEND_REQUEST, NULL))
992 mod_delayed_work(amt_wq, &amt->req_wq,
993 msecs_to_jiffies(100));
996 static bool amt_send_membership_update(struct amt_dev *amt,
1000 struct amt_header_membership_update *amtmu;
1001 struct socket *sock;
1007 sock = rcu_dereference_bh(amt->sock);
1011 err = skb_cow_head(skb, LL_RESERVED_SPACE(amt->dev) + sizeof(*amtmu) +
1012 sizeof(*iph) + sizeof(struct udphdr));
1016 skb_reset_inner_headers(skb);
1017 memset(&fl4, 0, sizeof(struct flowi4));
1018 fl4.flowi4_oif = amt->stream_dev->ifindex;
1019 fl4.daddr = amt->remote_ip;
1020 fl4.saddr = amt->local_ip;
1021 fl4.flowi4_tos = AMT_TOS;
1022 fl4.flowi4_proto = IPPROTO_UDP;
1023 rt = ip_route_output_key(amt->net, &fl4);
1025 netdev_dbg(amt->dev, "no route to %pI4\n", &amt->remote_ip);
1029 amtmu = skb_push(skb, sizeof(*amtmu));
1031 amtmu->type = AMT_MSG_MEMBERSHIP_UPDATE;
1032 amtmu->reserved = 0;
1033 amtmu->nonce = amt->nonce;
1034 amtmu->response_mac = amt->mac;
1037 skb_set_inner_protocol(skb, htons(ETH_P_IP));
1039 skb_set_inner_protocol(skb, htons(ETH_P_IPV6));
1040 udp_tunnel_xmit_skb(rt, sock->sk, skb,
1044 ip4_dst_hoplimit(&rt->dst),
1050 amt_update_gw_status(amt, AMT_STATUS_SENT_UPDATE, true);
1054 static void amt_send_multicast_data(struct amt_dev *amt,
1055 const struct sk_buff *oskb,
1056 struct amt_tunnel_list *tunnel,
1059 struct amt_header_mcast_data *amtmd;
1060 struct socket *sock;
1061 struct sk_buff *skb;
1066 sock = rcu_dereference_bh(amt->sock);
1070 skb = skb_copy_expand(oskb, sizeof(*amtmd) + sizeof(*iph) +
1071 sizeof(struct udphdr), 0, GFP_ATOMIC);
1075 skb_reset_inner_headers(skb);
1076 memset(&fl4, 0, sizeof(struct flowi4));
1077 fl4.flowi4_oif = amt->stream_dev->ifindex;
1078 fl4.daddr = tunnel->ip4;
1079 fl4.saddr = amt->local_ip;
1080 fl4.flowi4_proto = IPPROTO_UDP;
1081 rt = ip_route_output_key(amt->net, &fl4);
1083 netdev_dbg(amt->dev, "no route to %pI4\n", &tunnel->ip4);
1088 amtmd = skb_push(skb, sizeof(*amtmd));
1090 amtmd->reserved = 0;
1091 amtmd->type = AMT_MSG_MULTICAST_DATA;
1094 skb_set_inner_protocol(skb, htons(ETH_P_IP));
1096 skb_set_inner_protocol(skb, htons(ETH_P_IPV6));
1097 udp_tunnel_xmit_skb(rt, sock->sk, skb,
1101 ip4_dst_hoplimit(&rt->dst),
1104 tunnel->source_port,
1109 static bool amt_send_membership_query(struct amt_dev *amt,
1110 struct sk_buff *skb,
1111 struct amt_tunnel_list *tunnel,
1114 struct amt_header_membership_query *amtmq;
1115 struct socket *sock;
1120 sock = rcu_dereference_bh(amt->sock);
1124 err = skb_cow_head(skb, LL_RESERVED_SPACE(amt->dev) + sizeof(*amtmq) +
1125 sizeof(struct iphdr) + sizeof(struct udphdr));
1129 skb_reset_inner_headers(skb);
1130 memset(&fl4, 0, sizeof(struct flowi4));
1131 fl4.flowi4_oif = amt->stream_dev->ifindex;
1132 fl4.daddr = tunnel->ip4;
1133 fl4.saddr = amt->local_ip;
1134 fl4.flowi4_tos = AMT_TOS;
1135 fl4.flowi4_proto = IPPROTO_UDP;
1136 rt = ip_route_output_key(amt->net, &fl4);
1138 netdev_dbg(amt->dev, "no route to %pI4\n", &tunnel->ip4);
1142 amtmq = skb_push(skb, sizeof(*amtmq));
1144 amtmq->type = AMT_MSG_MEMBERSHIP_QUERY;
1145 amtmq->reserved = 0;
1148 amtmq->nonce = tunnel->nonce;
1149 amtmq->response_mac = tunnel->mac;
1152 skb_set_inner_protocol(skb, htons(ETH_P_IP));
1154 skb_set_inner_protocol(skb, htons(ETH_P_IPV6));
1155 udp_tunnel_xmit_skb(rt, sock->sk, skb,
1159 ip4_dst_hoplimit(&rt->dst),
1162 tunnel->source_port,
1165 amt_update_relay_status(tunnel, AMT_STATUS_SENT_QUERY, true);
1169 static netdev_tx_t amt_dev_xmit(struct sk_buff *skb, struct net_device *dev)
1171 struct amt_dev *amt = netdev_priv(dev);
1172 struct amt_tunnel_list *tunnel;
1173 struct amt_group_node *gnode;
1174 union amt_addr group = {0,};
1175 #if IS_ENABLED(CONFIG_IPV6)
1176 struct ipv6hdr *ip6h;
1177 struct mld_msg *mld;
1179 bool report = false;
1188 if (iph->version == 4) {
1189 if (!ipv4_is_multicast(iph->daddr))
1192 if (!ip_mc_check_igmp(skb)) {
1195 case IGMPV3_HOST_MEMBERSHIP_REPORT:
1196 case IGMP_HOST_MEMBERSHIP_REPORT:
1199 case IGMP_HOST_MEMBERSHIP_QUERY:
1209 group.ip4 = iph->daddr;
1210 #if IS_ENABLED(CONFIG_IPV6)
1211 } else if (iph->version == 6) {
1212 ip6h = ipv6_hdr(skb);
1213 if (!ipv6_addr_is_multicast(&ip6h->daddr))
1216 if (!ipv6_mc_check_mld(skb)) {
1217 mld = (struct mld_msg *)skb_transport_header(skb);
1218 switch (mld->mld_type) {
1219 case ICMPV6_MGM_REPORT:
1220 case ICMPV6_MLD2_REPORT:
1223 case ICMPV6_MGM_QUERY:
1233 group.ip6 = ip6h->daddr;
1236 dev->stats.tx_errors++;
1240 if (!pskb_may_pull(skb, sizeof(struct ethhdr)))
1243 skb_pull(skb, sizeof(struct ethhdr));
1245 if (amt->mode == AMT_MODE_GATEWAY) {
1246 /* Gateway only passes IGMP/MLD packets */
1249 if ((!v6 && !READ_ONCE(amt->ready4)) ||
1250 (v6 && !READ_ONCE(amt->ready6)))
1252 if (amt_send_membership_update(amt, skb, v6))
1255 } else if (amt->mode == AMT_MODE_RELAY) {
1257 tunnel = amt_skb_cb(skb)->tunnel;
1263 /* Do not forward unexpected query */
1264 if (amt_send_membership_query(amt, skb, tunnel, v6))
1271 list_for_each_entry_rcu(tunnel, &amt->tunnel_list, list) {
1272 hash = amt_group_hash(tunnel, &group);
1273 hlist_for_each_entry_rcu(gnode, &tunnel->groups[hash],
1276 if (gnode->group_addr.ip4 == iph->daddr)
1278 #if IS_ENABLED(CONFIG_IPV6)
1280 if (ipv6_addr_equal(&gnode->group_addr.ip6,
1288 amt_send_multicast_data(amt, skb, tunnel, v6);
1293 return NETDEV_TX_OK;
1297 dev->stats.tx_dropped++;
1298 return NETDEV_TX_OK;
1301 static int amt_parse_type(struct sk_buff *skb)
1303 struct amt_header *amth;
1305 if (!pskb_may_pull(skb, sizeof(struct udphdr) +
1306 sizeof(struct amt_header)))
1309 amth = (struct amt_header *)(udp_hdr(skb) + 1);
1311 if (amth->version != 0)
1314 if (amth->type >= __AMT_MSG_MAX || !amth->type)
1319 static void amt_clear_groups(struct amt_tunnel_list *tunnel)
1321 struct amt_dev *amt = tunnel->amt;
1322 struct amt_group_node *gnode;
1323 struct hlist_node *t;
1326 spin_lock_bh(&tunnel->lock);
1328 for (i = 0; i < amt->hash_buckets; i++)
1329 hlist_for_each_entry_safe(gnode, t, &tunnel->groups[i], node)
1330 amt_del_group(amt, gnode);
1332 spin_unlock_bh(&tunnel->lock);
1335 static void amt_tunnel_expire(struct work_struct *work)
1337 struct amt_tunnel_list *tunnel = container_of(to_delayed_work(work),
1338 struct amt_tunnel_list,
1340 struct amt_dev *amt = tunnel->amt;
1342 spin_lock_bh(&amt->lock);
1344 list_del_rcu(&tunnel->list);
1346 amt_clear_groups(tunnel);
1348 spin_unlock_bh(&amt->lock);
1349 kfree_rcu(tunnel, rcu);
1352 static void amt_cleanup_srcs(struct amt_dev *amt,
1353 struct amt_tunnel_list *tunnel,
1354 struct amt_group_node *gnode)
1356 struct amt_source_node *snode;
1357 struct hlist_node *t;
1360 /* Delete old sources */
1361 for (i = 0; i < amt->hash_buckets; i++) {
1362 hlist_for_each_entry_safe(snode, t, &gnode->sources[i], node) {
1363 if (snode->flags == AMT_SOURCE_OLD)
1364 amt_destroy_source(snode);
1368 /* switch from new to old */
1369 for (i = 0; i < amt->hash_buckets; i++) {
1370 hlist_for_each_entry_rcu(snode, &gnode->sources[i], node) {
1371 snode->flags = AMT_SOURCE_OLD;
1373 netdev_dbg(snode->gnode->amt->dev,
1374 "Add source as OLD %pI4 from %pI4\n",
1375 &snode->source_addr.ip4,
1376 &gnode->group_addr.ip4);
1377 #if IS_ENABLED(CONFIG_IPV6)
1379 netdev_dbg(snode->gnode->amt->dev,
1380 "Add source as OLD %pI6 from %pI6\n",
1381 &snode->source_addr.ip6,
1382 &gnode->group_addr.ip6);
1388 static void amt_add_srcs(struct amt_dev *amt, struct amt_tunnel_list *tunnel,
1389 struct amt_group_node *gnode, void *grec,
1392 struct igmpv3_grec *igmp_grec;
1393 struct amt_source_node *snode;
1394 #if IS_ENABLED(CONFIG_IPV6)
1395 struct mld2_grec *mld_grec;
1397 union amt_addr src = {0,};
1404 nsrcs = ntohs(igmp_grec->grec_nsrcs);
1406 #if IS_ENABLED(CONFIG_IPV6)
1408 nsrcs = ntohs(mld_grec->grec_nsrcs);
1413 for (i = 0; i < nsrcs; i++) {
1414 if (tunnel->nr_sources >= amt->max_sources)
1417 src.ip4 = igmp_grec->grec_src[i];
1418 #if IS_ENABLED(CONFIG_IPV6)
1420 memcpy(&src.ip6, &mld_grec->grec_src[i],
1421 sizeof(struct in6_addr));
1423 if (amt_lookup_src(tunnel, gnode, AMT_FILTER_ALL, &src))
1426 snode = amt_alloc_snode(gnode, &src);
1428 hash = amt_source_hash(tunnel, &snode->source_addr);
1429 hlist_add_head_rcu(&snode->node, &gnode->sources[hash]);
1430 tunnel->nr_sources++;
1431 gnode->nr_sources++;
1434 netdev_dbg(snode->gnode->amt->dev,
1435 "Add source as NEW %pI4 from %pI4\n",
1436 &snode->source_addr.ip4,
1437 &gnode->group_addr.ip4);
1438 #if IS_ENABLED(CONFIG_IPV6)
1440 netdev_dbg(snode->gnode->amt->dev,
1441 "Add source as NEW %pI6 from %pI6\n",
1442 &snode->source_addr.ip6,
1443 &gnode->group_addr.ip6);
1449 /* Router State Report Rec'd New Router State
1450 * ------------ ------------ ----------------
1451 * EXCLUDE (X,Y) IS_IN (A) EXCLUDE (X+A,Y-A)
1453 * -----------+-----------+-----------+
1455 * -----------+-----------+-----------+
1457 * -----------+-----------+-----------+
1459 * -----------+-----------+-----------+
1461 * -----------+-----------+-----------+
1463 * a) Received sources are NONE/NEW
1464 * b) All NONE will be deleted by amt_cleanup_srcs().
1465 * c) All OLD will be deleted by amt_cleanup_srcs().
1466 * d) After delete, NEW source will be switched to OLD.
1468 static void amt_lookup_act_srcs(struct amt_tunnel_list *tunnel,
1469 struct amt_group_node *gnode,
1472 enum amt_filter filter,
1476 struct amt_dev *amt = tunnel->amt;
1477 struct amt_source_node *snode;
1478 struct igmpv3_grec *igmp_grec;
1479 #if IS_ENABLED(CONFIG_IPV6)
1480 struct mld2_grec *mld_grec;
1482 union amt_addr src = {0,};
1483 struct hlist_node *t;
1489 nsrcs = ntohs(igmp_grec->grec_nsrcs);
1491 #if IS_ENABLED(CONFIG_IPV6)
1493 nsrcs = ntohs(mld_grec->grec_nsrcs);
1499 memset(&src, 0, sizeof(union amt_addr));
1503 for (i = 0; i < nsrcs; i++) {
1505 src.ip4 = igmp_grec->grec_src[i];
1506 #if IS_ENABLED(CONFIG_IPV6)
1508 memcpy(&src.ip6, &mld_grec->grec_src[i],
1509 sizeof(struct in6_addr));
1511 snode = amt_lookup_src(tunnel, gnode, filter, &src);
1514 amt_act_src(tunnel, gnode, snode, act);
1519 for (i = 0; i < amt->hash_buckets; i++) {
1520 hlist_for_each_entry_safe(snode, t, &gnode->sources[i],
1522 if (amt_status_filter(snode, filter))
1523 amt_act_src(tunnel, gnode, snode, act);
1526 for (i = 0; i < nsrcs; i++) {
1528 src.ip4 = igmp_grec->grec_src[i];
1529 #if IS_ENABLED(CONFIG_IPV6)
1531 memcpy(&src.ip6, &mld_grec->grec_src[i],
1532 sizeof(struct in6_addr));
1534 snode = amt_lookup_src(tunnel, gnode, filter, &src);
1537 amt_act_src(tunnel, gnode, snode, act);
1542 for (i = 0; i < amt->hash_buckets; i++) {
1543 hlist_for_each_entry_safe(snode, t, &gnode->sources[i],
1545 if (!amt_status_filter(snode, filter))
1547 for (j = 0; j < nsrcs; j++) {
1549 src.ip4 = igmp_grec->grec_src[j];
1550 #if IS_ENABLED(CONFIG_IPV6)
1553 &mld_grec->grec_src[j],
1554 sizeof(struct in6_addr));
1556 if (amt_addr_equal(&snode->source_addr,
1560 amt_act_src(tunnel, gnode, snode, act);
1566 case AMT_OPS_SUB_REV:
1568 for (i = 0; i < nsrcs; i++) {
1570 src.ip4 = igmp_grec->grec_src[i];
1571 #if IS_ENABLED(CONFIG_IPV6)
1573 memcpy(&src.ip6, &mld_grec->grec_src[i],
1574 sizeof(struct in6_addr));
1576 snode = amt_lookup_src(tunnel, gnode, AMT_FILTER_ALL,
1579 snode = amt_lookup_src(tunnel, gnode,
1582 amt_act_src(tunnel, gnode, snode, act);
1587 netdev_dbg(amt->dev, "Invalid type\n");
1592 static void amt_mcast_is_in_handler(struct amt_dev *amt,
1593 struct amt_tunnel_list *tunnel,
1594 struct amt_group_node *gnode,
1595 void *grec, void *zero_grec, bool v6)
1597 if (gnode->filter_mode == MCAST_INCLUDE) {
1598 /* Router State Report Rec'd New Router State Actions
1599 * ------------ ------------ ---------------- -------
1600 * INCLUDE (A) IS_IN (B) INCLUDE (A+B) (B)=GMI
1602 /* Update IS_IN (B) as FWD/NEW */
1603 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1604 AMT_FILTER_NONE_NEW,
1605 AMT_ACT_STATUS_FWD_NEW,
1607 /* Update INCLUDE (A) as NEW */
1608 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1610 AMT_ACT_STATUS_FWD_NEW,
1613 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1619 * ------------ ------------ ---------------- -------
1620 * EXCLUDE (X,Y) IS_IN (A) EXCLUDE (X+A,Y-A) (A)=GMI
1622 /* Update (A) in (X, Y) as NONE/NEW */
1623 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1625 AMT_ACT_STATUS_NONE_NEW,
1627 /* Update FWD/OLD as FWD/NEW */
1628 amt_lookup_act_srcs(tunnel, gnode, zero_grec, AMT_OPS_UNI,
1630 AMT_ACT_STATUS_FWD_NEW,
1632 /* Update IS_IN (A) as FWD/NEW */
1633 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1634 AMT_FILTER_NONE_NEW,
1635 AMT_ACT_STATUS_FWD_NEW,
1637 /* Update EXCLUDE (, Y-A) as D_FWD_NEW */
1638 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB,
1640 AMT_ACT_STATUS_D_FWD_NEW,
1645 static void amt_mcast_is_ex_handler(struct amt_dev *amt,
1646 struct amt_tunnel_list *tunnel,
1647 struct amt_group_node *gnode,
1648 void *grec, void *zero_grec, bool v6)
1650 if (gnode->filter_mode == MCAST_INCLUDE) {
1651 /* Router State Report Rec'd New Router State Actions
1652 * ------------ ------------ ---------------- -------
1653 * INCLUDE (A) IS_EX (B) EXCLUDE (A*B,B-A) (B-A)=0
1657 /* EXCLUDE(A*B, ) */
1658 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1660 AMT_ACT_STATUS_FWD_NEW,
1662 /* EXCLUDE(, B-A) */
1663 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB_REV,
1665 AMT_ACT_STATUS_D_FWD_NEW,
1668 amt_lookup_act_srcs(tunnel, gnode, zero_grec, AMT_OPS_UNI,
1669 AMT_FILTER_D_FWD_NEW,
1672 /* Group Timer=GMI */
1673 if (!mod_delayed_work(amt_wq, &gnode->group_timer,
1674 msecs_to_jiffies(amt_gmi(amt))))
1676 gnode->filter_mode = MCAST_EXCLUDE;
1677 /* Delete (A-B) will be worked by amt_cleanup_srcs(). */
1679 /* Router State Report Rec'd New Router State Actions
1680 * ------------ ------------ ---------------- -------
1681 * EXCLUDE (X,Y) IS_EX (A) EXCLUDE (A-Y,Y*A) (A-X-Y)=GMI
1686 /* EXCLUDE (A-Y, ) */
1687 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB_REV,
1689 AMT_ACT_STATUS_FWD_NEW,
1691 /* EXCLUDE (, Y*A ) */
1692 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1694 AMT_ACT_STATUS_D_FWD_NEW,
1697 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB_REV,
1698 AMT_FILTER_BOTH_NEW,
1701 /* Group Timer=GMI */
1702 if (!mod_delayed_work(amt_wq, &gnode->group_timer,
1703 msecs_to_jiffies(amt_gmi(amt))))
1705 /* Delete (X-A), (Y-A) will be worked by amt_cleanup_srcs(). */
1709 static void amt_mcast_to_in_handler(struct amt_dev *amt,
1710 struct amt_tunnel_list *tunnel,
1711 struct amt_group_node *gnode,
1712 void *grec, void *zero_grec, bool v6)
1714 if (gnode->filter_mode == MCAST_INCLUDE) {
1715 /* Router State Report Rec'd New Router State Actions
1716 * ------------ ------------ ---------------- -------
1717 * INCLUDE (A) TO_IN (B) INCLUDE (A+B) (B)=GMI
1720 /* Update TO_IN (B) sources as FWD/NEW */
1721 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1722 AMT_FILTER_NONE_NEW,
1723 AMT_ACT_STATUS_FWD_NEW,
1725 /* Update INCLUDE (A) sources as NEW */
1726 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1728 AMT_ACT_STATUS_FWD_NEW,
1731 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1736 /* Router State Report Rec'd New Router State Actions
1737 * ------------ ------------ ---------------- -------
1738 * EXCLUDE (X,Y) TO_IN (A) EXCLUDE (X+A,Y-A) (A)=GMI
1742 /* Update TO_IN (A) sources as FWD/NEW */
1743 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1744 AMT_FILTER_NONE_NEW,
1745 AMT_ACT_STATUS_FWD_NEW,
1747 /* Update EXCLUDE(X,) sources as FWD/NEW */
1748 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1750 AMT_ACT_STATUS_FWD_NEW,
1753 * (A) are already switched to FWD_NEW.
1754 * So, D_FWD/OLD -> D_FWD/NEW is okay.
1756 amt_lookup_act_srcs(tunnel, gnode, zero_grec, AMT_OPS_UNI,
1758 AMT_ACT_STATUS_D_FWD_NEW,
1761 * Only FWD_NEW will have (A) sources.
1763 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1770 static void amt_mcast_to_ex_handler(struct amt_dev *amt,
1771 struct amt_tunnel_list *tunnel,
1772 struct amt_group_node *gnode,
1773 void *grec, void *zero_grec, bool v6)
1775 if (gnode->filter_mode == MCAST_INCLUDE) {
1776 /* Router State Report Rec'd New Router State Actions
1777 * ------------ ------------ ---------------- -------
1778 * INCLUDE (A) TO_EX (B) EXCLUDE (A*B,B-A) (B-A)=0
1783 /* EXCLUDE (A*B, ) */
1784 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1786 AMT_ACT_STATUS_FWD_NEW,
1788 /* EXCLUDE (, B-A) */
1789 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB_REV,
1791 AMT_ACT_STATUS_D_FWD_NEW,
1794 amt_lookup_act_srcs(tunnel, gnode, zero_grec, AMT_OPS_UNI,
1795 AMT_FILTER_D_FWD_NEW,
1798 /* Group Timer=GMI */
1799 if (!mod_delayed_work(amt_wq, &gnode->group_timer,
1800 msecs_to_jiffies(amt_gmi(amt))))
1802 gnode->filter_mode = MCAST_EXCLUDE;
1803 /* Delete (A-B) will be worked by amt_cleanup_srcs(). */
1805 /* Router State Report Rec'd New Router State Actions
1806 * ------------ ------------ ---------------- -------
1807 * EXCLUDE (X,Y) TO_EX (A) EXCLUDE (A-Y,Y*A) (A-X-Y)=Group Timer
1813 /* Update (A-X-Y) as NONE/OLD */
1814 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB_REV,
1818 /* EXCLUDE (A-Y, ) */
1819 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB_REV,
1821 AMT_ACT_STATUS_FWD_NEW,
1823 /* EXCLUDE (, Y*A) */
1824 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1826 AMT_ACT_STATUS_D_FWD_NEW,
1828 /* Group Timer=GMI */
1829 if (!mod_delayed_work(amt_wq, &gnode->group_timer,
1830 msecs_to_jiffies(amt_gmi(amt))))
1832 /* Delete (X-A), (Y-A) will be worked by amt_cleanup_srcs(). */
1836 static void amt_mcast_allow_handler(struct amt_dev *amt,
1837 struct amt_tunnel_list *tunnel,
1838 struct amt_group_node *gnode,
1839 void *grec, void *zero_grec, bool v6)
1841 if (gnode->filter_mode == MCAST_INCLUDE) {
1842 /* Router State Report Rec'd New Router State Actions
1843 * ------------ ------------ ---------------- -------
1844 * INCLUDE (A) ALLOW (B) INCLUDE (A+B) (B)=GMI
1847 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1849 AMT_ACT_STATUS_FWD_NEW,
1852 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1857 /* Router State Report Rec'd New Router State Actions
1858 * ------------ ------------ ---------------- -------
1859 * EXCLUDE (X,Y) ALLOW (A) EXCLUDE (X+A,Y-A) (A)=GMI
1861 /* EXCLUDE (X+A, ) */
1862 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1864 AMT_ACT_STATUS_FWD_NEW,
1866 /* EXCLUDE (, Y-A) */
1867 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB,
1869 AMT_ACT_STATUS_D_FWD_NEW,
1872 * All (A) source are now FWD/NEW status.
1874 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1881 static void amt_mcast_block_handler(struct amt_dev *amt,
1882 struct amt_tunnel_list *tunnel,
1883 struct amt_group_node *gnode,
1884 void *grec, void *zero_grec, bool v6)
1886 if (gnode->filter_mode == MCAST_INCLUDE) {
1887 /* Router State Report Rec'd New Router State Actions
1888 * ------------ ------------ ---------------- -------
1889 * INCLUDE (A) BLOCK (B) INCLUDE (A) Send Q(G,A*B)
1892 amt_lookup_act_srcs(tunnel, gnode, zero_grec, AMT_OPS_UNI,
1894 AMT_ACT_STATUS_FWD_NEW,
1897 /* Router State Report Rec'd New Router State Actions
1898 * ------------ ------------ ---------------- -------
1899 * EXCLUDE (X,Y) BLOCK (A) EXCLUDE (X+(A-Y),Y) (A-X-Y)=Group Timer
1902 /* (A-X-Y)=Group Timer */
1903 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB_REV,
1908 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1910 AMT_ACT_STATUS_FWD_NEW,
1912 /* EXCLUDE (X+(A-Y) */
1913 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB_REV,
1915 AMT_ACT_STATUS_FWD_NEW,
1918 amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1920 AMT_ACT_STATUS_D_FWD_NEW,
1926 * 7.3.2. In the Presence of Older Version Group Members
1928 * When Group Compatibility Mode is IGMPv2, a router internally
1929 * translates the following IGMPv2 messages for that group to their
1930 * IGMPv3 equivalents:
1932 * IGMPv2 Message IGMPv3 Equivalent
1933 * -------------- -----------------
1934 * Report IS_EX( {} )
1937 static void amt_igmpv2_report_handler(struct amt_dev *amt, struct sk_buff *skb,
1938 struct amt_tunnel_list *tunnel)
1940 struct igmphdr *ih = igmp_hdr(skb);
1941 struct iphdr *iph = ip_hdr(skb);
1942 struct amt_group_node *gnode;
1943 union amt_addr group, host;
1945 memset(&group, 0, sizeof(union amt_addr));
1946 group.ip4 = ih->group;
1947 memset(&host, 0, sizeof(union amt_addr));
1948 host.ip4 = iph->saddr;
1950 gnode = amt_lookup_group(tunnel, &group, &host, false);
1952 gnode = amt_add_group(amt, tunnel, &group, &host, false);
1953 if (!IS_ERR(gnode)) {
1954 gnode->filter_mode = MCAST_EXCLUDE;
1955 if (!mod_delayed_work(amt_wq, &gnode->group_timer,
1956 msecs_to_jiffies(amt_gmi(amt))))
1963 * 7.3.2. In the Presence of Older Version Group Members
1965 * When Group Compatibility Mode is IGMPv2, a router internally
1966 * translates the following IGMPv2 messages for that group to their
1967 * IGMPv3 equivalents:
1969 * IGMPv2 Message IGMPv3 Equivalent
1970 * -------------- -----------------
1971 * Report IS_EX( {} )
1974 static void amt_igmpv2_leave_handler(struct amt_dev *amt, struct sk_buff *skb,
1975 struct amt_tunnel_list *tunnel)
1977 struct igmphdr *ih = igmp_hdr(skb);
1978 struct iphdr *iph = ip_hdr(skb);
1979 struct amt_group_node *gnode;
1980 union amt_addr group, host;
1982 memset(&group, 0, sizeof(union amt_addr));
1983 group.ip4 = ih->group;
1984 memset(&host, 0, sizeof(union amt_addr));
1985 host.ip4 = iph->saddr;
1987 gnode = amt_lookup_group(tunnel, &group, &host, false);
1989 amt_del_group(amt, gnode);
1992 static void amt_igmpv3_report_handler(struct amt_dev *amt, struct sk_buff *skb,
1993 struct amt_tunnel_list *tunnel)
1995 struct igmpv3_report *ihrv3 = igmpv3_report_hdr(skb);
1996 int len = skb_transport_offset(skb) + sizeof(*ihrv3);
1997 void *zero_grec = (void *)&igmpv3_zero_grec;
1998 struct iphdr *iph = ip_hdr(skb);
1999 struct amt_group_node *gnode;
2000 union amt_addr group, host;
2001 struct igmpv3_grec *grec;
2005 for (i = 0; i < ntohs(ihrv3->ngrec); i++) {
2006 len += sizeof(*grec);
2007 if (!ip_mc_may_pull(skb, len))
2010 grec = (void *)(skb->data + len - sizeof(*grec));
2011 nsrcs = ntohs(grec->grec_nsrcs);
2013 len += nsrcs * sizeof(__be32);
2014 if (!ip_mc_may_pull(skb, len))
2017 memset(&group, 0, sizeof(union amt_addr));
2018 group.ip4 = grec->grec_mca;
2019 memset(&host, 0, sizeof(union amt_addr));
2020 host.ip4 = iph->saddr;
2021 gnode = amt_lookup_group(tunnel, &group, &host, false);
2023 gnode = amt_add_group(amt, tunnel, &group, &host,
2029 amt_add_srcs(amt, tunnel, gnode, grec, false);
2030 switch (grec->grec_type) {
2031 case IGMPV3_MODE_IS_INCLUDE:
2032 amt_mcast_is_in_handler(amt, tunnel, gnode, grec,
2035 case IGMPV3_MODE_IS_EXCLUDE:
2036 amt_mcast_is_ex_handler(amt, tunnel, gnode, grec,
2039 case IGMPV3_CHANGE_TO_INCLUDE:
2040 amt_mcast_to_in_handler(amt, tunnel, gnode, grec,
2043 case IGMPV3_CHANGE_TO_EXCLUDE:
2044 amt_mcast_to_ex_handler(amt, tunnel, gnode, grec,
2047 case IGMPV3_ALLOW_NEW_SOURCES:
2048 amt_mcast_allow_handler(amt, tunnel, gnode, grec,
2051 case IGMPV3_BLOCK_OLD_SOURCES:
2052 amt_mcast_block_handler(amt, tunnel, gnode, grec,
2058 amt_cleanup_srcs(amt, tunnel, gnode);
2062 /* caller held tunnel->lock */
2063 static void amt_igmp_report_handler(struct amt_dev *amt, struct sk_buff *skb,
2064 struct amt_tunnel_list *tunnel)
2066 struct igmphdr *ih = igmp_hdr(skb);
2069 case IGMPV3_HOST_MEMBERSHIP_REPORT:
2070 amt_igmpv3_report_handler(amt, skb, tunnel);
2072 case IGMPV2_HOST_MEMBERSHIP_REPORT:
2073 amt_igmpv2_report_handler(amt, skb, tunnel);
2075 case IGMP_HOST_LEAVE_MESSAGE:
2076 amt_igmpv2_leave_handler(amt, skb, tunnel);
2083 #if IS_ENABLED(CONFIG_IPV6)
2085 * 8.3.2. In the Presence of MLDv1 Multicast Address Listeners
2087 * When Multicast Address Compatibility Mode is MLDv2, a router acts
2088 * using the MLDv2 protocol for that multicast address. When Multicast
2089 * Address Compatibility Mode is MLDv1, a router internally translates
2090 * the following MLDv1 messages for that multicast address to their
2091 * MLDv2 equivalents:
2093 * MLDv1 Message MLDv2 Equivalent
2094 * -------------- -----------------
2095 * Report IS_EX( {} )
2098 static void amt_mldv1_report_handler(struct amt_dev *amt, struct sk_buff *skb,
2099 struct amt_tunnel_list *tunnel)
2101 struct mld_msg *mld = (struct mld_msg *)icmp6_hdr(skb);
2102 struct ipv6hdr *ip6h = ipv6_hdr(skb);
2103 struct amt_group_node *gnode;
2104 union amt_addr group, host;
2106 memcpy(&group.ip6, &mld->mld_mca, sizeof(struct in6_addr));
2107 memcpy(&host.ip6, &ip6h->saddr, sizeof(struct in6_addr));
2109 gnode = amt_lookup_group(tunnel, &group, &host, true);
2111 gnode = amt_add_group(amt, tunnel, &group, &host, true);
2112 if (!IS_ERR(gnode)) {
2113 gnode->filter_mode = MCAST_EXCLUDE;
2114 if (!mod_delayed_work(amt_wq, &gnode->group_timer,
2115 msecs_to_jiffies(amt_gmi(amt))))
2122 * 8.3.2. In the Presence of MLDv1 Multicast Address Listeners
2124 * When Multicast Address Compatibility Mode is MLDv2, a router acts
2125 * using the MLDv2 protocol for that multicast address. When Multicast
2126 * Address Compatibility Mode is MLDv1, a router internally translates
2127 * the following MLDv1 messages for that multicast address to their
2128 * MLDv2 equivalents:
2130 * MLDv1 Message MLDv2 Equivalent
2131 * -------------- -----------------
2132 * Report IS_EX( {} )
2135 static void amt_mldv1_leave_handler(struct amt_dev *amt, struct sk_buff *skb,
2136 struct amt_tunnel_list *tunnel)
2138 struct mld_msg *mld = (struct mld_msg *)icmp6_hdr(skb);
2139 struct iphdr *iph = ip_hdr(skb);
2140 struct amt_group_node *gnode;
2141 union amt_addr group, host;
2143 memcpy(&group.ip6, &mld->mld_mca, sizeof(struct in6_addr));
2144 memset(&host, 0, sizeof(union amt_addr));
2145 host.ip4 = iph->saddr;
2147 gnode = amt_lookup_group(tunnel, &group, &host, true);
2149 amt_del_group(amt, gnode);
2154 static void amt_mldv2_report_handler(struct amt_dev *amt, struct sk_buff *skb,
2155 struct amt_tunnel_list *tunnel)
2157 struct mld2_report *mld2r = (struct mld2_report *)icmp6_hdr(skb);
2158 int len = skb_transport_offset(skb) + sizeof(*mld2r);
2159 void *zero_grec = (void *)&mldv2_zero_grec;
2160 struct ipv6hdr *ip6h = ipv6_hdr(skb);
2161 struct amt_group_node *gnode;
2162 union amt_addr group, host;
2163 struct mld2_grec *grec;
2167 for (i = 0; i < ntohs(mld2r->mld2r_ngrec); i++) {
2168 len += sizeof(*grec);
2169 if (!ipv6_mc_may_pull(skb, len))
2172 grec = (void *)(skb->data + len - sizeof(*grec));
2173 nsrcs = ntohs(grec->grec_nsrcs);
2175 len += nsrcs * sizeof(struct in6_addr);
2176 if (!ipv6_mc_may_pull(skb, len))
2179 memset(&group, 0, sizeof(union amt_addr));
2180 group.ip6 = grec->grec_mca;
2181 memset(&host, 0, sizeof(union amt_addr));
2182 host.ip6 = ip6h->saddr;
2183 gnode = amt_lookup_group(tunnel, &group, &host, true);
2185 gnode = amt_add_group(amt, tunnel, &group, &host,
2191 amt_add_srcs(amt, tunnel, gnode, grec, true);
2192 switch (grec->grec_type) {
2193 case MLD2_MODE_IS_INCLUDE:
2194 amt_mcast_is_in_handler(amt, tunnel, gnode, grec,
2197 case MLD2_MODE_IS_EXCLUDE:
2198 amt_mcast_is_ex_handler(amt, tunnel, gnode, grec,
2201 case MLD2_CHANGE_TO_INCLUDE:
2202 amt_mcast_to_in_handler(amt, tunnel, gnode, grec,
2205 case MLD2_CHANGE_TO_EXCLUDE:
2206 amt_mcast_to_ex_handler(amt, tunnel, gnode, grec,
2209 case MLD2_ALLOW_NEW_SOURCES:
2210 amt_mcast_allow_handler(amt, tunnel, gnode, grec,
2213 case MLD2_BLOCK_OLD_SOURCES:
2214 amt_mcast_block_handler(amt, tunnel, gnode, grec,
2220 amt_cleanup_srcs(amt, tunnel, gnode);
2224 /* caller held tunnel->lock */
2225 static void amt_mld_report_handler(struct amt_dev *amt, struct sk_buff *skb,
2226 struct amt_tunnel_list *tunnel)
2228 struct mld_msg *mld = (struct mld_msg *)icmp6_hdr(skb);
2230 switch (mld->mld_type) {
2231 case ICMPV6_MGM_REPORT:
2232 amt_mldv1_report_handler(amt, skb, tunnel);
2234 case ICMPV6_MLD2_REPORT:
2235 amt_mldv2_report_handler(amt, skb, tunnel);
2237 case ICMPV6_MGM_REDUCTION:
2238 amt_mldv1_leave_handler(amt, skb, tunnel);
2246 static bool amt_advertisement_handler(struct amt_dev *amt, struct sk_buff *skb)
2248 struct amt_header_advertisement *amta;
2251 hdr_size = sizeof(*amta) + sizeof(struct udphdr);
2252 if (!pskb_may_pull(skb, hdr_size))
2255 amta = (struct amt_header_advertisement *)(udp_hdr(skb) + 1);
2259 if (amta->reserved || amta->version)
2262 if (ipv4_is_loopback(amta->ip4) || ipv4_is_multicast(amta->ip4) ||
2263 ipv4_is_zeronet(amta->ip4))
2266 if (amt->status != AMT_STATUS_SENT_DISCOVERY ||
2267 amt->nonce != amta->nonce)
2270 amt->remote_ip = amta->ip4;
2271 netdev_dbg(amt->dev, "advertised remote ip = %pI4\n", &amt->remote_ip);
2272 mod_delayed_work(amt_wq, &amt->req_wq, 0);
2274 amt_update_gw_status(amt, AMT_STATUS_RECEIVED_ADVERTISEMENT, true);
2278 static bool amt_multicast_data_handler(struct amt_dev *amt, struct sk_buff *skb)
2280 struct amt_header_mcast_data *amtmd;
2281 int hdr_size, len, err;
2285 if (READ_ONCE(amt->status) != AMT_STATUS_SENT_UPDATE)
2288 hdr_size = sizeof(*amtmd) + sizeof(struct udphdr);
2289 if (!pskb_may_pull(skb, hdr_size))
2292 amtmd = (struct amt_header_mcast_data *)(udp_hdr(skb) + 1);
2293 if (amtmd->reserved || amtmd->version)
2296 if (iptunnel_pull_header(skb, hdr_size, htons(ETH_P_IP), false))
2299 skb_reset_network_header(skb);
2300 skb_push(skb, sizeof(*eth));
2301 skb_reset_mac_header(skb);
2302 skb_pull(skb, sizeof(*eth));
2305 if (!pskb_may_pull(skb, sizeof(*iph)))
2309 if (iph->version == 4) {
2310 if (!ipv4_is_multicast(iph->daddr))
2312 skb->protocol = htons(ETH_P_IP);
2313 eth->h_proto = htons(ETH_P_IP);
2314 ip_eth_mc_map(iph->daddr, eth->h_dest);
2315 #if IS_ENABLED(CONFIG_IPV6)
2316 } else if (iph->version == 6) {
2317 struct ipv6hdr *ip6h;
2319 if (!pskb_may_pull(skb, sizeof(*ip6h)))
2322 ip6h = ipv6_hdr(skb);
2323 if (!ipv6_addr_is_multicast(&ip6h->daddr))
2325 skb->protocol = htons(ETH_P_IPV6);
2326 eth->h_proto = htons(ETH_P_IPV6);
2327 ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
2333 skb->pkt_type = PACKET_MULTICAST;
2334 skb->ip_summed = CHECKSUM_NONE;
2336 err = gro_cells_receive(&amt->gro_cells, skb);
2337 if (likely(err == NET_RX_SUCCESS))
2338 dev_sw_netstats_rx_add(amt->dev, len);
2340 amt->dev->stats.rx_dropped++;
2345 static bool amt_membership_query_handler(struct amt_dev *amt,
2346 struct sk_buff *skb)
2348 struct amt_header_membership_query *amtmq;
2349 struct igmpv3_query *ihv3;
2350 struct ethhdr *eth, *oeth;
2354 hdr_size = sizeof(*amtmq) + sizeof(struct udphdr);
2355 if (!pskb_may_pull(skb, hdr_size))
2358 amtmq = (struct amt_header_membership_query *)(udp_hdr(skb) + 1);
2359 if (amtmq->reserved || amtmq->version)
2362 if (amtmq->nonce != amt->nonce)
2365 hdr_size -= sizeof(*eth);
2366 if (iptunnel_pull_header(skb, hdr_size, htons(ETH_P_TEB), false))
2369 oeth = eth_hdr(skb);
2370 skb_reset_mac_header(skb);
2371 skb_pull(skb, sizeof(*eth));
2372 skb_reset_network_header(skb);
2374 if (!pskb_may_pull(skb, sizeof(*iph)))
2378 if (iph->version == 4) {
2379 if (READ_ONCE(amt->ready4))
2382 if (!pskb_may_pull(skb, sizeof(*iph) + AMT_IPHDR_OPTS +
2386 if (!ipv4_is_multicast(iph->daddr))
2389 ihv3 = skb_pull(skb, sizeof(*iph) + AMT_IPHDR_OPTS);
2390 skb_reset_transport_header(skb);
2391 skb_push(skb, sizeof(*iph) + AMT_IPHDR_OPTS);
2392 WRITE_ONCE(amt->ready4, true);
2393 amt->mac = amtmq->response_mac;
2395 amt->qi = ihv3->qqic;
2396 skb->protocol = htons(ETH_P_IP);
2397 eth->h_proto = htons(ETH_P_IP);
2398 ip_eth_mc_map(iph->daddr, eth->h_dest);
2399 #if IS_ENABLED(CONFIG_IPV6)
2400 } else if (iph->version == 6) {
2401 struct mld2_query *mld2q;
2402 struct ipv6hdr *ip6h;
2404 if (READ_ONCE(amt->ready6))
2407 if (!pskb_may_pull(skb, sizeof(*ip6h) + AMT_IP6HDR_OPTS +
2411 ip6h = ipv6_hdr(skb);
2412 if (!ipv6_addr_is_multicast(&ip6h->daddr))
2415 mld2q = skb_pull(skb, sizeof(*ip6h) + AMT_IP6HDR_OPTS);
2416 skb_reset_transport_header(skb);
2417 skb_push(skb, sizeof(*ip6h) + AMT_IP6HDR_OPTS);
2418 WRITE_ONCE(amt->ready6, true);
2419 amt->mac = amtmq->response_mac;
2421 amt->qi = mld2q->mld2q_qqic;
2422 skb->protocol = htons(ETH_P_IPV6);
2423 eth->h_proto = htons(ETH_P_IPV6);
2424 ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
2430 ether_addr_copy(eth->h_source, oeth->h_source);
2431 skb->pkt_type = PACKET_MULTICAST;
2432 skb->ip_summed = CHECKSUM_NONE;
2435 if (__netif_rx(skb) == NET_RX_SUCCESS) {
2436 amt_update_gw_status(amt, AMT_STATUS_RECEIVED_QUERY, true);
2437 dev_sw_netstats_rx_add(amt->dev, len);
2439 amt->dev->stats.rx_dropped++;
2446 static bool amt_update_handler(struct amt_dev *amt, struct sk_buff *skb)
2448 struct amt_header_membership_update *amtmu;
2449 struct amt_tunnel_list *tunnel;
2456 hdr_size = sizeof(*amtmu) + sizeof(struct udphdr);
2457 if (!pskb_may_pull(skb, hdr_size))
2460 amtmu = (struct amt_header_membership_update *)(udp_hdr(skb) + 1);
2461 if (amtmu->reserved || amtmu->version)
2464 if (iptunnel_pull_header(skb, hdr_size, skb->protocol, false))
2467 skb_reset_network_header(skb);
2469 list_for_each_entry_rcu(tunnel, &amt->tunnel_list, list) {
2470 if (tunnel->ip4 == iph->saddr) {
2471 if ((amtmu->nonce == tunnel->nonce &&
2472 amtmu->response_mac == tunnel->mac)) {
2473 mod_delayed_work(amt_wq, &tunnel->gc_wq,
2474 msecs_to_jiffies(amt_gmi(amt))
2478 netdev_dbg(amt->dev, "Invalid MAC\n");
2487 if (!pskb_may_pull(skb, sizeof(*iph)))
2491 if (iph->version == 4) {
2492 if (ip_mc_check_igmp(skb)) {
2493 netdev_dbg(amt->dev, "Invalid IGMP\n");
2497 spin_lock_bh(&tunnel->lock);
2498 amt_igmp_report_handler(amt, skb, tunnel);
2499 spin_unlock_bh(&tunnel->lock);
2501 skb_push(skb, sizeof(struct ethhdr));
2502 skb_reset_mac_header(skb);
2504 skb->protocol = htons(ETH_P_IP);
2505 eth->h_proto = htons(ETH_P_IP);
2506 ip_eth_mc_map(iph->daddr, eth->h_dest);
2507 #if IS_ENABLED(CONFIG_IPV6)
2508 } else if (iph->version == 6) {
2509 struct ipv6hdr *ip6h = ipv6_hdr(skb);
2511 if (ipv6_mc_check_mld(skb)) {
2512 netdev_dbg(amt->dev, "Invalid MLD\n");
2516 spin_lock_bh(&tunnel->lock);
2517 amt_mld_report_handler(amt, skb, tunnel);
2518 spin_unlock_bh(&tunnel->lock);
2520 skb_push(skb, sizeof(struct ethhdr));
2521 skb_reset_mac_header(skb);
2523 skb->protocol = htons(ETH_P_IPV6);
2524 eth->h_proto = htons(ETH_P_IPV6);
2525 ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
2528 netdev_dbg(amt->dev, "Unsupported Protocol\n");
2532 skb_pull(skb, sizeof(struct ethhdr));
2533 skb->pkt_type = PACKET_MULTICAST;
2534 skb->ip_summed = CHECKSUM_NONE;
2536 if (__netif_rx(skb) == NET_RX_SUCCESS) {
2537 amt_update_relay_status(tunnel, AMT_STATUS_RECEIVED_UPDATE,
2539 dev_sw_netstats_rx_add(amt->dev, len);
2541 amt->dev->stats.rx_dropped++;
2547 static void amt_send_advertisement(struct amt_dev *amt, __be32 nonce,
2548 __be32 daddr, __be16 dport)
2550 struct amt_header_advertisement *amta;
2551 int hlen, tlen, offset;
2552 struct socket *sock;
2553 struct udphdr *udph;
2554 struct sk_buff *skb;
2562 sock = rcu_dereference(amt->sock);
2566 if (!netif_running(amt->stream_dev) || !netif_running(amt->dev))
2569 rt = ip_route_output_ports(amt->net, &fl4, sock->sk,
2570 daddr, amt->local_ip,
2571 dport, amt->relay_port,
2573 amt->stream_dev->ifindex);
2575 amt->dev->stats.tx_errors++;
2579 hlen = LL_RESERVED_SPACE(amt->dev);
2580 tlen = amt->dev->needed_tailroom;
2581 len = hlen + tlen + sizeof(*iph) + sizeof(*udph) + sizeof(*amta);
2582 skb = netdev_alloc_skb_ip_align(amt->dev, len);
2585 amt->dev->stats.tx_errors++;
2589 skb->priority = TC_PRIO_CONTROL;
2590 skb_dst_set(skb, &rt->dst);
2592 len = sizeof(*iph) + sizeof(*udph) + sizeof(*amta);
2593 skb_reset_network_header(skb);
2595 amta = skb_pull(skb, sizeof(*iph) + sizeof(*udph));
2597 amta->type = AMT_MSG_ADVERTISEMENT;
2599 amta->nonce = nonce;
2600 amta->ip4 = amt->local_ip;
2601 skb_push(skb, sizeof(*udph));
2602 skb_reset_transport_header(skb);
2603 udph = udp_hdr(skb);
2604 udph->source = amt->relay_port;
2606 udph->len = htons(sizeof(*amta) + sizeof(*udph));
2608 offset = skb_transport_offset(skb);
2609 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
2610 udph->check = csum_tcpudp_magic(amt->local_ip, daddr,
2611 sizeof(*udph) + sizeof(*amta),
2612 IPPROTO_UDP, skb->csum);
2614 skb_push(skb, sizeof(*iph));
2617 iph->ihl = (sizeof(struct iphdr)) >> 2;
2620 iph->ttl = ip4_dst_hoplimit(&rt->dst);
2622 iph->saddr = amt->local_ip;
2623 iph->protocol = IPPROTO_UDP;
2624 iph->tot_len = htons(len);
2626 skb->ip_summed = CHECKSUM_NONE;
2627 ip_select_ident(amt->net, skb, NULL);
2629 err = ip_local_out(amt->net, sock->sk, skb);
2630 if (unlikely(net_xmit_eval(err)))
2631 amt->dev->stats.tx_errors++;
2637 static bool amt_discovery_handler(struct amt_dev *amt, struct sk_buff *skb)
2639 struct amt_header_discovery *amtd;
2640 struct udphdr *udph;
2643 if (!pskb_may_pull(skb, sizeof(*udph) + sizeof(*amtd)))
2647 udph = udp_hdr(skb);
2648 amtd = (struct amt_header_discovery *)(udp_hdr(skb) + 1);
2650 if (amtd->reserved || amtd->version)
2653 amt_send_advertisement(amt, amtd->nonce, iph->saddr, udph->source);
2658 static bool amt_request_handler(struct amt_dev *amt, struct sk_buff *skb)
2660 struct amt_header_request *amtrh;
2661 struct amt_tunnel_list *tunnel;
2662 unsigned long long key;
2663 struct udphdr *udph;
2668 if (!pskb_may_pull(skb, sizeof(*udph) + sizeof(*amtrh)))
2672 udph = udp_hdr(skb);
2673 amtrh = (struct amt_header_request *)(udp_hdr(skb) + 1);
2675 if (amtrh->reserved1 || amtrh->reserved2 || amtrh->version)
2678 list_for_each_entry_rcu(tunnel, &amt->tunnel_list, list)
2679 if (tunnel->ip4 == iph->saddr)
2682 spin_lock_bh(&amt->lock);
2683 if (amt->nr_tunnels >= amt->max_tunnels) {
2684 spin_unlock_bh(&amt->lock);
2685 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0);
2689 tunnel = kzalloc(sizeof(*tunnel) +
2690 (sizeof(struct hlist_head) * amt->hash_buckets),
2693 spin_unlock_bh(&amt->lock);
2697 tunnel->source_port = udph->source;
2698 tunnel->ip4 = iph->saddr;
2700 memcpy(&key, &tunnel->key, sizeof(unsigned long long));
2702 spin_lock_init(&tunnel->lock);
2703 for (i = 0; i < amt->hash_buckets; i++)
2704 INIT_HLIST_HEAD(&tunnel->groups[i]);
2706 INIT_DELAYED_WORK(&tunnel->gc_wq, amt_tunnel_expire);
2708 list_add_tail_rcu(&tunnel->list, &amt->tunnel_list);
2709 tunnel->key = amt->key;
2710 __amt_update_relay_status(tunnel, AMT_STATUS_RECEIVED_REQUEST, true);
2712 mod_delayed_work(amt_wq, &tunnel->gc_wq,
2713 msecs_to_jiffies(amt_gmi(amt)));
2714 spin_unlock_bh(&amt->lock);
2717 tunnel->nonce = amtrh->nonce;
2718 mac = siphash_3u32((__force u32)tunnel->ip4,
2719 (__force u32)tunnel->source_port,
2720 (__force u32)tunnel->nonce,
2722 tunnel->mac = mac >> 16;
2724 if (!netif_running(amt->dev) || !netif_running(amt->stream_dev))
2728 amt_send_igmp_gq(amt, tunnel);
2730 amt_send_mld_gq(amt, tunnel);
2735 static void amt_gw_rcv(struct amt_dev *amt, struct sk_buff *skb)
2737 int type = amt_parse_type(skb);
2743 if (amt->mode == AMT_MODE_GATEWAY) {
2745 case AMT_MSG_ADVERTISEMENT:
2746 err = amt_advertisement_handler(amt, skb);
2748 case AMT_MSG_MEMBERSHIP_QUERY:
2749 err = amt_membership_query_handler(amt, skb);
2754 netdev_dbg(amt->dev, "Invalid type of Gateway\n");
2760 amt->dev->stats.rx_dropped++;
2767 static int amt_rcv(struct sock *sk, struct sk_buff *skb)
2769 struct amt_dev *amt;
2775 amt = rcu_dereference_sk_user_data(sk);
2782 skb->dev = amt->dev;
2784 type = amt_parse_type(skb);
2790 if (amt->mode == AMT_MODE_GATEWAY) {
2792 case AMT_MSG_ADVERTISEMENT:
2793 if (iph->saddr != amt->discovery_ip) {
2794 netdev_dbg(amt->dev, "Invalid Relay IP\n");
2798 if (amt_queue_event(amt, AMT_EVENT_RECEIVE, skb)) {
2799 netdev_dbg(amt->dev, "AMT Event queue full\n");
2804 case AMT_MSG_MULTICAST_DATA:
2805 if (iph->saddr != amt->remote_ip) {
2806 netdev_dbg(amt->dev, "Invalid Relay IP\n");
2810 err = amt_multicast_data_handler(amt, skb);
2815 case AMT_MSG_MEMBERSHIP_QUERY:
2816 if (iph->saddr != amt->remote_ip) {
2817 netdev_dbg(amt->dev, "Invalid Relay IP\n");
2821 if (amt_queue_event(amt, AMT_EVENT_RECEIVE, skb)) {
2822 netdev_dbg(amt->dev, "AMT Event queue full\n");
2829 netdev_dbg(amt->dev, "Invalid type of Gateway\n");
2834 case AMT_MSG_DISCOVERY:
2835 err = amt_discovery_handler(amt, skb);
2837 case AMT_MSG_REQUEST:
2838 err = amt_request_handler(amt, skb);
2840 case AMT_MSG_MEMBERSHIP_UPDATE:
2841 err = amt_update_handler(amt, skb);
2848 netdev_dbg(amt->dev, "Invalid type of relay\n");
2854 amt->dev->stats.rx_dropped++;
2860 rcu_read_unlock_bh();
2864 static void amt_event_work(struct work_struct *work)
2866 struct amt_dev *amt = container_of(work, struct amt_dev, event_wq);
2867 struct sk_buff *skb;
2871 for (i = 0; i < AMT_MAX_EVENTS; i++) {
2872 spin_lock_bh(&amt->lock);
2873 if (amt->nr_events == 0) {
2874 spin_unlock_bh(&amt->lock);
2877 event = amt->events[amt->event_idx].event;
2878 skb = amt->events[amt->event_idx].skb;
2879 amt->events[amt->event_idx].event = AMT_EVENT_NONE;
2880 amt->events[amt->event_idx].skb = NULL;
2883 amt->event_idx %= AMT_MAX_EVENTS;
2884 spin_unlock_bh(&amt->lock);
2887 case AMT_EVENT_RECEIVE:
2888 amt_gw_rcv(amt, skb);
2890 case AMT_EVENT_SEND_DISCOVERY:
2891 amt_event_send_discovery(amt);
2893 case AMT_EVENT_SEND_REQUEST:
2894 amt_event_send_request(amt);
2903 static int amt_err_lookup(struct sock *sk, struct sk_buff *skb)
2905 struct amt_dev *amt;
2909 amt = rcu_dereference_sk_user_data(sk);
2913 if (amt->mode != AMT_MODE_GATEWAY)
2916 type = amt_parse_type(skb);
2920 netdev_dbg(amt->dev, "Received IGMP Unreachable of %s\n",
2923 case AMT_MSG_DISCOVERY:
2925 case AMT_MSG_REQUEST:
2926 case AMT_MSG_MEMBERSHIP_UPDATE:
2927 if (READ_ONCE(amt->status) >= AMT_STATUS_RECEIVED_ADVERTISEMENT)
2928 mod_delayed_work(amt_wq, &amt->req_wq, 0);
2934 rcu_read_unlock_bh();
2937 rcu_read_unlock_bh();
2938 amt->dev->stats.rx_dropped++;
2942 static struct socket *amt_create_sock(struct net *net, __be16 port)
2944 struct udp_port_cfg udp_conf;
2945 struct socket *sock;
2948 memset(&udp_conf, 0, sizeof(udp_conf));
2949 udp_conf.family = AF_INET;
2950 udp_conf.local_ip.s_addr = htonl(INADDR_ANY);
2952 udp_conf.local_udp_port = port;
2954 err = udp_sock_create(net, &udp_conf, &sock);
2956 return ERR_PTR(err);
2961 static int amt_socket_create(struct amt_dev *amt)
2963 struct udp_tunnel_sock_cfg tunnel_cfg;
2964 struct socket *sock;
2966 sock = amt_create_sock(amt->net, amt->relay_port);
2968 return PTR_ERR(sock);
2970 /* Mark socket as an encapsulation socket */
2971 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
2972 tunnel_cfg.sk_user_data = amt;
2973 tunnel_cfg.encap_type = 1;
2974 tunnel_cfg.encap_rcv = amt_rcv;
2975 tunnel_cfg.encap_err_lookup = amt_err_lookup;
2976 tunnel_cfg.encap_destroy = NULL;
2977 setup_udp_tunnel_sock(amt->net, sock, &tunnel_cfg);
2979 rcu_assign_pointer(amt->sock, sock);
2983 static int amt_dev_open(struct net_device *dev)
2985 struct amt_dev *amt = netdev_priv(dev);
2988 amt->ready4 = false;
2989 amt->ready6 = false;
2993 err = amt_socket_create(amt);
3000 get_random_bytes(&amt->key, sizeof(siphash_key_t));
3002 amt->status = AMT_STATUS_INIT;
3003 if (amt->mode == AMT_MODE_GATEWAY) {
3004 mod_delayed_work(amt_wq, &amt->discovery_wq, 0);
3005 mod_delayed_work(amt_wq, &amt->req_wq, 0);
3006 } else if (amt->mode == AMT_MODE_RELAY) {
3007 mod_delayed_work(amt_wq, &amt->secret_wq,
3008 msecs_to_jiffies(AMT_SECRET_TIMEOUT));
3013 static int amt_dev_stop(struct net_device *dev)
3015 struct amt_dev *amt = netdev_priv(dev);
3016 struct amt_tunnel_list *tunnel, *tmp;
3017 struct socket *sock;
3018 struct sk_buff *skb;
3021 cancel_delayed_work_sync(&amt->req_wq);
3022 cancel_delayed_work_sync(&amt->discovery_wq);
3023 cancel_delayed_work_sync(&amt->secret_wq);
3026 sock = rtnl_dereference(amt->sock);
3027 RCU_INIT_POINTER(amt->sock, NULL);
3030 udp_tunnel_sock_release(sock);
3032 cancel_work_sync(&amt->event_wq);
3033 for (i = 0; i < AMT_MAX_EVENTS; i++) {
3034 skb = amt->events[i].skb;
3036 amt->events[i].event = AMT_EVENT_NONE;
3037 amt->events[i].skb = NULL;
3040 amt->ready4 = false;
3041 amt->ready6 = false;
3045 list_for_each_entry_safe(tunnel, tmp, &amt->tunnel_list, list) {
3046 list_del_rcu(&tunnel->list);
3048 cancel_delayed_work_sync(&tunnel->gc_wq);
3049 amt_clear_groups(tunnel);
3050 kfree_rcu(tunnel, rcu);
3056 static const struct device_type amt_type = {
3060 static int amt_dev_init(struct net_device *dev)
3062 struct amt_dev *amt = netdev_priv(dev);
3066 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
3070 err = gro_cells_init(&amt->gro_cells, dev);
3072 free_percpu(dev->tstats);
3079 static void amt_dev_uninit(struct net_device *dev)
3081 struct amt_dev *amt = netdev_priv(dev);
3083 gro_cells_destroy(&amt->gro_cells);
3084 free_percpu(dev->tstats);
3087 static const struct net_device_ops amt_netdev_ops = {
3088 .ndo_init = amt_dev_init,
3089 .ndo_uninit = amt_dev_uninit,
3090 .ndo_open = amt_dev_open,
3091 .ndo_stop = amt_dev_stop,
3092 .ndo_start_xmit = amt_dev_xmit,
3093 .ndo_get_stats64 = dev_get_tstats64,
3096 static void amt_link_setup(struct net_device *dev)
3098 dev->netdev_ops = &amt_netdev_ops;
3099 dev->needs_free_netdev = true;
3100 SET_NETDEV_DEVTYPE(dev, &amt_type);
3101 dev->min_mtu = ETH_MIN_MTU;
3102 dev->max_mtu = ETH_MAX_MTU;
3103 dev->type = ARPHRD_NONE;
3104 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3105 dev->hard_header_len = 0;
3107 dev->priv_flags |= IFF_NO_QUEUE;
3108 dev->features |= NETIF_F_LLTX;
3109 dev->features |= NETIF_F_GSO_SOFTWARE;
3110 dev->features |= NETIF_F_NETNS_LOCAL;
3111 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM;
3112 dev->hw_features |= NETIF_F_FRAGLIST | NETIF_F_RXCSUM;
3113 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
3114 eth_hw_addr_random(dev);
3115 eth_zero_addr(dev->broadcast);
3119 static const struct nla_policy amt_policy[IFLA_AMT_MAX + 1] = {
3120 [IFLA_AMT_MODE] = { .type = NLA_U32 },
3121 [IFLA_AMT_RELAY_PORT] = { .type = NLA_U16 },
3122 [IFLA_AMT_GATEWAY_PORT] = { .type = NLA_U16 },
3123 [IFLA_AMT_LINK] = { .type = NLA_U32 },
3124 [IFLA_AMT_LOCAL_IP] = { .len = sizeof_field(struct iphdr, daddr) },
3125 [IFLA_AMT_REMOTE_IP] = { .len = sizeof_field(struct iphdr, daddr) },
3126 [IFLA_AMT_DISCOVERY_IP] = { .len = sizeof_field(struct iphdr, daddr) },
3127 [IFLA_AMT_MAX_TUNNELS] = { .type = NLA_U32 },
3130 static int amt_validate(struct nlattr *tb[], struct nlattr *data[],
3131 struct netlink_ext_ack *extack)
3136 if (!data[IFLA_AMT_LINK]) {
3137 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_AMT_LINK],
3138 "Link attribute is required");
3142 if (!data[IFLA_AMT_MODE]) {
3143 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_AMT_MODE],
3144 "Mode attribute is required");
3148 if (nla_get_u32(data[IFLA_AMT_MODE]) > AMT_MODE_MAX) {
3149 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_AMT_MODE],
3150 "Mode attribute is not valid");
3154 if (!data[IFLA_AMT_LOCAL_IP]) {
3155 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_AMT_DISCOVERY_IP],
3156 "Local attribute is required");
3160 if (!data[IFLA_AMT_DISCOVERY_IP] &&
3161 nla_get_u32(data[IFLA_AMT_MODE]) == AMT_MODE_GATEWAY) {
3162 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_AMT_LOCAL_IP],
3163 "Discovery attribute is required");
3170 static int amt_newlink(struct net *net, struct net_device *dev,
3171 struct nlattr *tb[], struct nlattr *data[],
3172 struct netlink_ext_ack *extack)
3174 struct amt_dev *amt = netdev_priv(dev);
3178 amt->mode = nla_get_u32(data[IFLA_AMT_MODE]);
3180 if (data[IFLA_AMT_MAX_TUNNELS] &&
3181 nla_get_u32(data[IFLA_AMT_MAX_TUNNELS]))
3182 amt->max_tunnels = nla_get_u32(data[IFLA_AMT_MAX_TUNNELS]);
3184 amt->max_tunnels = AMT_MAX_TUNNELS;
3186 spin_lock_init(&amt->lock);
3187 amt->max_groups = AMT_MAX_GROUP;
3188 amt->max_sources = AMT_MAX_SOURCE;
3189 amt->hash_buckets = AMT_HSIZE;
3190 amt->nr_tunnels = 0;
3191 get_random_bytes(&amt->hash_seed, sizeof(amt->hash_seed));
3192 amt->stream_dev = dev_get_by_index(net,
3193 nla_get_u32(data[IFLA_AMT_LINK]));
3194 if (!amt->stream_dev) {
3195 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_AMT_LINK],
3196 "Can't find stream device");
3200 if (amt->stream_dev->type != ARPHRD_ETHER) {
3201 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_AMT_LINK],
3202 "Invalid stream device type");
3206 amt->local_ip = nla_get_in_addr(data[IFLA_AMT_LOCAL_IP]);
3207 if (ipv4_is_loopback(amt->local_ip) ||
3208 ipv4_is_zeronet(amt->local_ip) ||
3209 ipv4_is_multicast(amt->local_ip)) {
3210 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_AMT_LOCAL_IP],
3211 "Invalid Local address");
3215 if (data[IFLA_AMT_RELAY_PORT])
3216 amt->relay_port = nla_get_be16(data[IFLA_AMT_RELAY_PORT]);
3218 amt->relay_port = htons(IANA_AMT_UDP_PORT);
3220 if (data[IFLA_AMT_GATEWAY_PORT])
3221 amt->gw_port = nla_get_be16(data[IFLA_AMT_GATEWAY_PORT]);
3223 amt->gw_port = htons(IANA_AMT_UDP_PORT);
3225 if (!amt->relay_port) {
3226 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_AMT_DISCOVERY_IP],
3227 "relay port must not be 0");
3230 if (amt->mode == AMT_MODE_RELAY) {
3231 amt->qrv = READ_ONCE(amt->net->ipv4.sysctl_igmp_qrv);
3233 dev->needed_headroom = amt->stream_dev->needed_headroom +
3235 dev->mtu = amt->stream_dev->mtu - AMT_RELAY_HLEN;
3236 dev->max_mtu = dev->mtu;
3237 dev->min_mtu = ETH_MIN_MTU + AMT_RELAY_HLEN;
3239 if (!data[IFLA_AMT_DISCOVERY_IP]) {
3240 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_AMT_DISCOVERY_IP],
3241 "discovery must be set in gateway mode");
3244 if (!amt->gw_port) {
3245 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_AMT_DISCOVERY_IP],
3246 "gateway port must not be 0");
3250 amt->discovery_ip = nla_get_in_addr(data[IFLA_AMT_DISCOVERY_IP]);
3251 if (ipv4_is_loopback(amt->discovery_ip) ||
3252 ipv4_is_zeronet(amt->discovery_ip) ||
3253 ipv4_is_multicast(amt->discovery_ip)) {
3254 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_AMT_DISCOVERY_IP],
3255 "discovery must be unicast");
3259 dev->needed_headroom = amt->stream_dev->needed_headroom +
3261 dev->mtu = amt->stream_dev->mtu - AMT_GW_HLEN;
3262 dev->max_mtu = dev->mtu;
3263 dev->min_mtu = ETH_MIN_MTU + AMT_GW_HLEN;
3265 amt->qi = AMT_INIT_QUERY_INTERVAL;
3267 err = register_netdevice(dev);
3269 netdev_dbg(dev, "failed to register new netdev %d\n", err);
3273 err = netdev_upper_dev_link(amt->stream_dev, dev, extack);
3275 unregister_netdevice(dev);
3279 INIT_DELAYED_WORK(&amt->discovery_wq, amt_discovery_work);
3280 INIT_DELAYED_WORK(&amt->req_wq, amt_req_work);
3281 INIT_DELAYED_WORK(&amt->secret_wq, amt_secret_work);
3282 INIT_WORK(&amt->event_wq, amt_event_work);
3283 INIT_LIST_HEAD(&amt->tunnel_list);
3286 dev_put(amt->stream_dev);
3290 static void amt_dellink(struct net_device *dev, struct list_head *head)
3292 struct amt_dev *amt = netdev_priv(dev);
3294 unregister_netdevice_queue(dev, head);
3295 netdev_upper_dev_unlink(amt->stream_dev, dev);
3296 dev_put(amt->stream_dev);
3299 static size_t amt_get_size(const struct net_device *dev)
3301 return nla_total_size(sizeof(__u32)) + /* IFLA_AMT_MODE */
3302 nla_total_size(sizeof(__u16)) + /* IFLA_AMT_RELAY_PORT */
3303 nla_total_size(sizeof(__u16)) + /* IFLA_AMT_GATEWAY_PORT */
3304 nla_total_size(sizeof(__u32)) + /* IFLA_AMT_LINK */
3305 nla_total_size(sizeof(__u32)) + /* IFLA_MAX_TUNNELS */
3306 nla_total_size(sizeof(struct iphdr)) + /* IFLA_AMT_DISCOVERY_IP */
3307 nla_total_size(sizeof(struct iphdr)) + /* IFLA_AMT_REMOTE_IP */
3308 nla_total_size(sizeof(struct iphdr)); /* IFLA_AMT_LOCAL_IP */
3311 static int amt_fill_info(struct sk_buff *skb, const struct net_device *dev)
3313 struct amt_dev *amt = netdev_priv(dev);
3315 if (nla_put_u32(skb, IFLA_AMT_MODE, amt->mode))
3316 goto nla_put_failure;
3317 if (nla_put_be16(skb, IFLA_AMT_RELAY_PORT, amt->relay_port))
3318 goto nla_put_failure;
3319 if (nla_put_be16(skb, IFLA_AMT_GATEWAY_PORT, amt->gw_port))
3320 goto nla_put_failure;
3321 if (nla_put_u32(skb, IFLA_AMT_LINK, amt->stream_dev->ifindex))
3322 goto nla_put_failure;
3323 if (nla_put_in_addr(skb, IFLA_AMT_LOCAL_IP, amt->local_ip))
3324 goto nla_put_failure;
3325 if (nla_put_in_addr(skb, IFLA_AMT_DISCOVERY_IP, amt->discovery_ip))
3326 goto nla_put_failure;
3328 if (nla_put_in_addr(skb, IFLA_AMT_REMOTE_IP, amt->remote_ip))
3329 goto nla_put_failure;
3330 if (nla_put_u32(skb, IFLA_AMT_MAX_TUNNELS, amt->max_tunnels))
3331 goto nla_put_failure;
3339 static struct rtnl_link_ops amt_link_ops __read_mostly = {
3341 .maxtype = IFLA_AMT_MAX,
3342 .policy = amt_policy,
3343 .priv_size = sizeof(struct amt_dev),
3344 .setup = amt_link_setup,
3345 .validate = amt_validate,
3346 .newlink = amt_newlink,
3347 .dellink = amt_dellink,
3348 .get_size = amt_get_size,
3349 .fill_info = amt_fill_info,
3352 static struct net_device *amt_lookup_upper_dev(struct net_device *dev)
3354 struct net_device *upper_dev;
3355 struct amt_dev *amt;
3357 for_each_netdev(dev_net(dev), upper_dev) {
3358 if (netif_is_amt(upper_dev)) {
3359 amt = netdev_priv(upper_dev);
3360 if (amt->stream_dev == dev)
3368 static int amt_device_event(struct notifier_block *unused,
3369 unsigned long event, void *ptr)
3371 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3372 struct net_device *upper_dev;
3373 struct amt_dev *amt;
3377 upper_dev = amt_lookup_upper_dev(dev);
3380 amt = netdev_priv(upper_dev);
3383 case NETDEV_UNREGISTER:
3384 amt_dellink(amt->dev, &list);
3385 unregister_netdevice_many(&list);
3387 case NETDEV_CHANGEMTU:
3388 if (amt->mode == AMT_MODE_RELAY)
3389 new_mtu = dev->mtu - AMT_RELAY_HLEN;
3391 new_mtu = dev->mtu - AMT_GW_HLEN;
3393 dev_set_mtu(amt->dev, new_mtu);
3400 static struct notifier_block amt_notifier_block __read_mostly = {
3401 .notifier_call = amt_device_event,
3404 static int __init amt_init(void)
3408 err = register_netdevice_notifier(&amt_notifier_block);
3412 err = rtnl_link_register(&amt_link_ops);
3414 goto unregister_notifier;
3416 amt_wq = alloc_workqueue("amt", WQ_UNBOUND, 0);
3419 goto rtnl_unregister;
3422 spin_lock_init(&source_gc_lock);
3423 spin_lock_bh(&source_gc_lock);
3424 INIT_DELAYED_WORK(&source_gc_wq, amt_source_gc_work);
3425 mod_delayed_work(amt_wq, &source_gc_wq,
3426 msecs_to_jiffies(AMT_GC_INTERVAL));
3427 spin_unlock_bh(&source_gc_lock);
3432 rtnl_link_unregister(&amt_link_ops);
3433 unregister_notifier:
3434 unregister_netdevice_notifier(&amt_notifier_block);
3436 pr_err("error loading AMT module loaded\n");
3439 late_initcall(amt_init);
3441 static void __exit amt_fini(void)
3443 rtnl_link_unregister(&amt_link_ops);
3444 unregister_netdevice_notifier(&amt_notifier_block);
3445 cancel_delayed_work_sync(&source_gc_wq);
3446 __amt_source_gc_work();
3447 destroy_workqueue(amt_wq);
3449 module_exit(amt_fini);
3451 MODULE_LICENSE("GPL");
3452 MODULE_AUTHOR("Taehee Yoo <ap420073@gmail.com>");
3453 MODULE_ALIAS_RTNL_LINK("amt");