xref: /openbmc/linux/drivers/net/amt.c (revision a1117495)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* Copyright (c) 2021 Taehee Yoo <ap420073@gmail.com> */
3 
4 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
5 
6 #include <linux/module.h>
7 #include <linux/skbuff.h>
8 #include <linux/udp.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/net_namespace.h>
15 #include <net/ip.h>
16 #include <net/udp.h>
17 #include <net/udp_tunnel.h>
18 #include <net/icmp.h>
19 #include <net/mld.h>
20 #include <net/amt.h>
21 #include <uapi/linux/amt.h>
22 #include <linux/security.h>
23 #include <net/gro_cells.h>
24 #include <net/ipv6.h>
25 #include <net/if_inet6.h>
26 #include <net/ndisc.h>
27 #include <net/addrconf.h>
28 #include <net/ip6_route.h>
29 #include <net/inet_common.h>
30 #include <net/ip6_checksum.h>
31 
32 static struct workqueue_struct *amt_wq;
33 
34 static HLIST_HEAD(source_gc_list);
35 /* Lock for source_gc_list */
36 static spinlock_t source_gc_lock;
37 static struct delayed_work source_gc_wq;
38 static char *status_str[] = {
39 	"AMT_STATUS_INIT",
40 	"AMT_STATUS_SENT_DISCOVERY",
41 	"AMT_STATUS_RECEIVED_DISCOVERY",
42 	"AMT_STATUS_SENT_ADVERTISEMENT",
43 	"AMT_STATUS_RECEIVED_ADVERTISEMENT",
44 	"AMT_STATUS_SENT_REQUEST",
45 	"AMT_STATUS_RECEIVED_REQUEST",
46 	"AMT_STATUS_SENT_QUERY",
47 	"AMT_STATUS_RECEIVED_QUERY",
48 	"AMT_STATUS_SENT_UPDATE",
49 	"AMT_STATUS_RECEIVED_UPDATE",
50 };
51 
52 static char *type_str[] = {
53 	"AMT_MSG_DISCOVERY",
54 	"AMT_MSG_ADVERTISEMENT",
55 	"AMT_MSG_REQUEST",
56 	"AMT_MSG_MEMBERSHIP_QUERY",
57 	"AMT_MSG_MEMBERSHIP_UPDATE",
58 	"AMT_MSG_MULTICAST_DATA",
59 	"AMT_MSG_TEARDOWM",
60 };
61 
62 static char *action_str[] = {
63 	"AMT_ACT_GMI",
64 	"AMT_ACT_GMI_ZERO",
65 	"AMT_ACT_GT",
66 	"AMT_ACT_STATUS_FWD_NEW",
67 	"AMT_ACT_STATUS_D_FWD_NEW",
68 	"AMT_ACT_STATUS_NONE_NEW",
69 };
70 
71 static struct igmpv3_grec igmpv3_zero_grec;
72 
73 #if IS_ENABLED(CONFIG_IPV6)
74 #define MLD2_ALL_NODE_INIT { { { 0xff, 0x02, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0x01 } } }
75 static struct in6_addr mld2_all_node = MLD2_ALL_NODE_INIT;
76 static struct mld2_grec mldv2_zero_grec;
77 #endif
78 
79 static struct amt_skb_cb *amt_skb_cb(struct sk_buff *skb)
80 {
81 	BUILD_BUG_ON(sizeof(struct amt_skb_cb) + sizeof(struct qdisc_skb_cb) >
82 		     sizeof_field(struct sk_buff, cb));
83 
84 	return (struct amt_skb_cb *)((void *)skb->cb +
85 		sizeof(struct qdisc_skb_cb));
86 }
87 
88 static void __amt_source_gc_work(void)
89 {
90 	struct amt_source_node *snode;
91 	struct hlist_head gc_list;
92 	struct hlist_node *t;
93 
94 	spin_lock_bh(&source_gc_lock);
95 	hlist_move_list(&source_gc_list, &gc_list);
96 	spin_unlock_bh(&source_gc_lock);
97 
98 	hlist_for_each_entry_safe(snode, t, &gc_list, node) {
99 		hlist_del_rcu(&snode->node);
100 		kfree_rcu(snode, rcu);
101 	}
102 }
103 
104 static void amt_source_gc_work(struct work_struct *work)
105 {
106 	__amt_source_gc_work();
107 
108 	spin_lock_bh(&source_gc_lock);
109 	mod_delayed_work(amt_wq, &source_gc_wq,
110 			 msecs_to_jiffies(AMT_GC_INTERVAL));
111 	spin_unlock_bh(&source_gc_lock);
112 }
113 
114 static bool amt_addr_equal(union amt_addr *a, union amt_addr *b)
115 {
116 	return !memcmp(a, b, sizeof(union amt_addr));
117 }
118 
119 static u32 amt_source_hash(struct amt_tunnel_list *tunnel, union amt_addr *src)
120 {
121 	u32 hash = jhash(src, sizeof(*src), tunnel->amt->hash_seed);
122 
123 	return reciprocal_scale(hash, tunnel->amt->hash_buckets);
124 }
125 
126 static bool amt_status_filter(struct amt_source_node *snode,
127 			      enum amt_filter filter)
128 {
129 	bool rc = false;
130 
131 	switch (filter) {
132 	case AMT_FILTER_FWD:
133 		if (snode->status == AMT_SOURCE_STATUS_FWD &&
134 		    snode->flags == AMT_SOURCE_OLD)
135 			rc = true;
136 		break;
137 	case AMT_FILTER_D_FWD:
138 		if (snode->status == AMT_SOURCE_STATUS_D_FWD &&
139 		    snode->flags == AMT_SOURCE_OLD)
140 			rc = true;
141 		break;
142 	case AMT_FILTER_FWD_NEW:
143 		if (snode->status == AMT_SOURCE_STATUS_FWD &&
144 		    snode->flags == AMT_SOURCE_NEW)
145 			rc = true;
146 		break;
147 	case AMT_FILTER_D_FWD_NEW:
148 		if (snode->status == AMT_SOURCE_STATUS_D_FWD &&
149 		    snode->flags == AMT_SOURCE_NEW)
150 			rc = true;
151 		break;
152 	case AMT_FILTER_ALL:
153 		rc = true;
154 		break;
155 	case AMT_FILTER_NONE_NEW:
156 		if (snode->status == AMT_SOURCE_STATUS_NONE &&
157 		    snode->flags == AMT_SOURCE_NEW)
158 			rc = true;
159 		break;
160 	case AMT_FILTER_BOTH:
161 		if ((snode->status == AMT_SOURCE_STATUS_D_FWD ||
162 		     snode->status == AMT_SOURCE_STATUS_FWD) &&
163 		    snode->flags == AMT_SOURCE_OLD)
164 			rc = true;
165 		break;
166 	case AMT_FILTER_BOTH_NEW:
167 		if ((snode->status == AMT_SOURCE_STATUS_D_FWD ||
168 		     snode->status == AMT_SOURCE_STATUS_FWD) &&
169 		    snode->flags == AMT_SOURCE_NEW)
170 			rc = true;
171 		break;
172 	default:
173 		WARN_ON_ONCE(1);
174 		break;
175 	}
176 
177 	return rc;
178 }
179 
180 static struct amt_source_node *amt_lookup_src(struct amt_tunnel_list *tunnel,
181 					      struct amt_group_node *gnode,
182 					      enum amt_filter filter,
183 					      union amt_addr *src)
184 {
185 	u32 hash = amt_source_hash(tunnel, src);
186 	struct amt_source_node *snode;
187 
188 	hlist_for_each_entry_rcu(snode, &gnode->sources[hash], node)
189 		if (amt_status_filter(snode, filter) &&
190 		    amt_addr_equal(&snode->source_addr, src))
191 			return snode;
192 
193 	return NULL;
194 }
195 
196 static u32 amt_group_hash(struct amt_tunnel_list *tunnel, union amt_addr *group)
197 {
198 	u32 hash = jhash(group, sizeof(*group), tunnel->amt->hash_seed);
199 
200 	return reciprocal_scale(hash, tunnel->amt->hash_buckets);
201 }
202 
203 static struct amt_group_node *amt_lookup_group(struct amt_tunnel_list *tunnel,
204 					       union amt_addr *group,
205 					       union amt_addr *host,
206 					       bool v6)
207 {
208 	u32 hash = amt_group_hash(tunnel, group);
209 	struct amt_group_node *gnode;
210 
211 	hlist_for_each_entry_rcu(gnode, &tunnel->groups[hash], node) {
212 		if (amt_addr_equal(&gnode->group_addr, group) &&
213 		    amt_addr_equal(&gnode->host_addr, host) &&
214 		    gnode->v6 == v6)
215 			return gnode;
216 	}
217 
218 	return NULL;
219 }
220 
221 static void amt_destroy_source(struct amt_source_node *snode)
222 {
223 	struct amt_group_node *gnode = snode->gnode;
224 	struct amt_tunnel_list *tunnel;
225 
226 	tunnel = gnode->tunnel_list;
227 
228 	if (!gnode->v6) {
229 		netdev_dbg(snode->gnode->amt->dev,
230 			   "Delete source %pI4 from %pI4\n",
231 			   &snode->source_addr.ip4,
232 			   &gnode->group_addr.ip4);
233 #if IS_ENABLED(CONFIG_IPV6)
234 	} else {
235 		netdev_dbg(snode->gnode->amt->dev,
236 			   "Delete source %pI6 from %pI6\n",
237 			   &snode->source_addr.ip6,
238 			   &gnode->group_addr.ip6);
239 #endif
240 	}
241 
242 	cancel_delayed_work(&snode->source_timer);
243 	hlist_del_init_rcu(&snode->node);
244 	tunnel->nr_sources--;
245 	gnode->nr_sources--;
246 	spin_lock_bh(&source_gc_lock);
247 	hlist_add_head_rcu(&snode->node, &source_gc_list);
248 	spin_unlock_bh(&source_gc_lock);
249 }
250 
251 static void amt_del_group(struct amt_dev *amt, struct amt_group_node *gnode)
252 {
253 	struct amt_source_node *snode;
254 	struct hlist_node *t;
255 	int i;
256 
257 	if (cancel_delayed_work(&gnode->group_timer))
258 		dev_put(amt->dev);
259 	hlist_del_rcu(&gnode->node);
260 	gnode->tunnel_list->nr_groups--;
261 
262 	if (!gnode->v6)
263 		netdev_dbg(amt->dev, "Leave group %pI4\n",
264 			   &gnode->group_addr.ip4);
265 #if IS_ENABLED(CONFIG_IPV6)
266 	else
267 		netdev_dbg(amt->dev, "Leave group %pI6\n",
268 			   &gnode->group_addr.ip6);
269 #endif
270 	for (i = 0; i < amt->hash_buckets; i++)
271 		hlist_for_each_entry_safe(snode, t, &gnode->sources[i], node)
272 			amt_destroy_source(snode);
273 
274 	/* tunnel->lock was acquired outside of amt_del_group()
275 	 * But rcu_read_lock() was acquired too so It's safe.
276 	 */
277 	kfree_rcu(gnode, rcu);
278 }
279 
280 /* If a source timer expires with a router filter-mode for the group of
281  * INCLUDE, the router concludes that traffic from this particular
282  * source is no longer desired on the attached network, and deletes the
283  * associated source record.
284  */
285 static void amt_source_work(struct work_struct *work)
286 {
287 	struct amt_source_node *snode = container_of(to_delayed_work(work),
288 						     struct amt_source_node,
289 						     source_timer);
290 	struct amt_group_node *gnode = snode->gnode;
291 	struct amt_dev *amt = gnode->amt;
292 	struct amt_tunnel_list *tunnel;
293 
294 	tunnel = gnode->tunnel_list;
295 	spin_lock_bh(&tunnel->lock);
296 	rcu_read_lock();
297 	if (gnode->filter_mode == MCAST_INCLUDE) {
298 		amt_destroy_source(snode);
299 		if (!gnode->nr_sources)
300 			amt_del_group(amt, gnode);
301 	} else {
302 		/* When a router filter-mode for a group is EXCLUDE,
303 		 * source records are only deleted when the group timer expires
304 		 */
305 		snode->status = AMT_SOURCE_STATUS_D_FWD;
306 	}
307 	rcu_read_unlock();
308 	spin_unlock_bh(&tunnel->lock);
309 }
310 
311 static void amt_act_src(struct amt_tunnel_list *tunnel,
312 			struct amt_group_node *gnode,
313 			struct amt_source_node *snode,
314 			enum amt_act act)
315 {
316 	struct amt_dev *amt = tunnel->amt;
317 
318 	switch (act) {
319 	case AMT_ACT_GMI:
320 		mod_delayed_work(amt_wq, &snode->source_timer,
321 				 msecs_to_jiffies(amt_gmi(amt)));
322 		break;
323 	case AMT_ACT_GMI_ZERO:
324 		cancel_delayed_work(&snode->source_timer);
325 		break;
326 	case AMT_ACT_GT:
327 		mod_delayed_work(amt_wq, &snode->source_timer,
328 				 gnode->group_timer.timer.expires);
329 		break;
330 	case AMT_ACT_STATUS_FWD_NEW:
331 		snode->status = AMT_SOURCE_STATUS_FWD;
332 		snode->flags = AMT_SOURCE_NEW;
333 		break;
334 	case AMT_ACT_STATUS_D_FWD_NEW:
335 		snode->status = AMT_SOURCE_STATUS_D_FWD;
336 		snode->flags = AMT_SOURCE_NEW;
337 		break;
338 	case AMT_ACT_STATUS_NONE_NEW:
339 		cancel_delayed_work(&snode->source_timer);
340 		snode->status = AMT_SOURCE_STATUS_NONE;
341 		snode->flags = AMT_SOURCE_NEW;
342 		break;
343 	default:
344 		WARN_ON_ONCE(1);
345 		return;
346 	}
347 
348 	if (!gnode->v6)
349 		netdev_dbg(amt->dev, "Source %pI4 from %pI4 Acted %s\n",
350 			   &snode->source_addr.ip4,
351 			   &gnode->group_addr.ip4,
352 			   action_str[act]);
353 #if IS_ENABLED(CONFIG_IPV6)
354 	else
355 		netdev_dbg(amt->dev, "Source %pI6 from %pI6 Acted %s\n",
356 			   &snode->source_addr.ip6,
357 			   &gnode->group_addr.ip6,
358 			   action_str[act]);
359 #endif
360 }
361 
362 static struct amt_source_node *amt_alloc_snode(struct amt_group_node *gnode,
363 					       union amt_addr *src)
364 {
365 	struct amt_source_node *snode;
366 
367 	snode = kzalloc(sizeof(*snode), GFP_ATOMIC);
368 	if (!snode)
369 		return NULL;
370 
371 	memcpy(&snode->source_addr, src, sizeof(union amt_addr));
372 	snode->gnode = gnode;
373 	snode->status = AMT_SOURCE_STATUS_NONE;
374 	snode->flags = AMT_SOURCE_NEW;
375 	INIT_HLIST_NODE(&snode->node);
376 	INIT_DELAYED_WORK(&snode->source_timer, amt_source_work);
377 
378 	return snode;
379 }
380 
381 /* RFC 3810 - 7.2.2.  Definition of Filter Timers
382  *
383  *  Router Mode          Filter Timer         Actions/Comments
384  *  -----------       -----------------       ----------------
385  *
386  *    INCLUDE             Not Used            All listeners in
387  *                                            INCLUDE mode.
388  *
389  *    EXCLUDE             Timer > 0           At least one listener
390  *                                            in EXCLUDE mode.
391  *
392  *    EXCLUDE             Timer == 0          No more listeners in
393  *                                            EXCLUDE mode for the
394  *                                            multicast address.
395  *                                            If the Requested List
396  *                                            is empty, delete
397  *                                            Multicast Address
398  *                                            Record.  If not, switch
399  *                                            to INCLUDE filter mode;
400  *                                            the sources in the
401  *                                            Requested List are
402  *                                            moved to the Include
403  *                                            List, and the Exclude
404  *                                            List is deleted.
405  */
406 static void amt_group_work(struct work_struct *work)
407 {
408 	struct amt_group_node *gnode = container_of(to_delayed_work(work),
409 						    struct amt_group_node,
410 						    group_timer);
411 	struct amt_tunnel_list *tunnel = gnode->tunnel_list;
412 	struct amt_dev *amt = gnode->amt;
413 	struct amt_source_node *snode;
414 	bool delete_group = true;
415 	struct hlist_node *t;
416 	int i, buckets;
417 
418 	buckets = amt->hash_buckets;
419 
420 	spin_lock_bh(&tunnel->lock);
421 	if (gnode->filter_mode == MCAST_INCLUDE) {
422 		/* Not Used */
423 		spin_unlock_bh(&tunnel->lock);
424 		goto out;
425 	}
426 
427 	rcu_read_lock();
428 	for (i = 0; i < buckets; i++) {
429 		hlist_for_each_entry_safe(snode, t,
430 					  &gnode->sources[i], node) {
431 			if (!delayed_work_pending(&snode->source_timer) ||
432 			    snode->status == AMT_SOURCE_STATUS_D_FWD) {
433 				amt_destroy_source(snode);
434 			} else {
435 				delete_group = false;
436 				snode->status = AMT_SOURCE_STATUS_FWD;
437 			}
438 		}
439 	}
440 	if (delete_group)
441 		amt_del_group(amt, gnode);
442 	else
443 		gnode->filter_mode = MCAST_INCLUDE;
444 	rcu_read_unlock();
445 	spin_unlock_bh(&tunnel->lock);
446 out:
447 	dev_put(amt->dev);
448 }
449 
450 /* Non-existant group is created as INCLUDE {empty}:
451  *
452  * RFC 3376 - 5.1. Action on Change of Interface State
453  *
454  * If no interface state existed for that multicast address before
455  * the change (i.e., the change consisted of creating a new
456  * per-interface record), or if no state exists after the change
457  * (i.e., the change consisted of deleting a per-interface record),
458  * then the "non-existent" state is considered to have a filter mode
459  * of INCLUDE and an empty source list.
460  */
461 static struct amt_group_node *amt_add_group(struct amt_dev *amt,
462 					    struct amt_tunnel_list *tunnel,
463 					    union amt_addr *group,
464 					    union amt_addr *host,
465 					    bool v6)
466 {
467 	struct amt_group_node *gnode;
468 	u32 hash;
469 	int i;
470 
471 	if (tunnel->nr_groups >= amt->max_groups)
472 		return ERR_PTR(-ENOSPC);
473 
474 	gnode = kzalloc(sizeof(*gnode) +
475 			(sizeof(struct hlist_head) * amt->hash_buckets),
476 			GFP_ATOMIC);
477 	if (unlikely(!gnode))
478 		return ERR_PTR(-ENOMEM);
479 
480 	gnode->amt = amt;
481 	gnode->group_addr = *group;
482 	gnode->host_addr = *host;
483 	gnode->v6 = v6;
484 	gnode->tunnel_list = tunnel;
485 	gnode->filter_mode = MCAST_INCLUDE;
486 	INIT_HLIST_NODE(&gnode->node);
487 	INIT_DELAYED_WORK(&gnode->group_timer, amt_group_work);
488 	for (i = 0; i < amt->hash_buckets; i++)
489 		INIT_HLIST_HEAD(&gnode->sources[i]);
490 
491 	hash = amt_group_hash(tunnel, group);
492 	hlist_add_head_rcu(&gnode->node, &tunnel->groups[hash]);
493 	tunnel->nr_groups++;
494 
495 	if (!gnode->v6)
496 		netdev_dbg(amt->dev, "Join group %pI4\n",
497 			   &gnode->group_addr.ip4);
498 #if IS_ENABLED(CONFIG_IPV6)
499 	else
500 		netdev_dbg(amt->dev, "Join group %pI6\n",
501 			   &gnode->group_addr.ip6);
502 #endif
503 
504 	return gnode;
505 }
506 
507 static struct sk_buff *amt_build_igmp_gq(struct amt_dev *amt)
508 {
509 	u8 ra[AMT_IPHDR_OPTS] = { IPOPT_RA, 4, 0, 0 };
510 	int hlen = LL_RESERVED_SPACE(amt->dev);
511 	int tlen = amt->dev->needed_tailroom;
512 	struct igmpv3_query *ihv3;
513 	void *csum_start = NULL;
514 	__sum16 *csum = NULL;
515 	struct sk_buff *skb;
516 	struct ethhdr *eth;
517 	struct iphdr *iph;
518 	unsigned int len;
519 	int offset;
520 
521 	len = hlen + tlen + sizeof(*iph) + AMT_IPHDR_OPTS + sizeof(*ihv3);
522 	skb = netdev_alloc_skb_ip_align(amt->dev, len);
523 	if (!skb)
524 		return NULL;
525 
526 	skb_reserve(skb, hlen);
527 	skb_push(skb, sizeof(*eth));
528 	skb->protocol = htons(ETH_P_IP);
529 	skb_reset_mac_header(skb);
530 	skb->priority = TC_PRIO_CONTROL;
531 	skb_put(skb, sizeof(*iph));
532 	skb_put_data(skb, ra, sizeof(ra));
533 	skb_put(skb, sizeof(*ihv3));
534 	skb_pull(skb, sizeof(*eth));
535 	skb_reset_network_header(skb);
536 
537 	iph		= ip_hdr(skb);
538 	iph->version	= 4;
539 	iph->ihl	= (sizeof(struct iphdr) + AMT_IPHDR_OPTS) >> 2;
540 	iph->tos	= AMT_TOS;
541 	iph->tot_len	= htons(sizeof(*iph) + AMT_IPHDR_OPTS + sizeof(*ihv3));
542 	iph->frag_off	= htons(IP_DF);
543 	iph->ttl	= 1;
544 	iph->id		= 0;
545 	iph->protocol	= IPPROTO_IGMP;
546 	iph->daddr	= htonl(INADDR_ALLHOSTS_GROUP);
547 	iph->saddr	= htonl(INADDR_ANY);
548 	ip_send_check(iph);
549 
550 	eth = eth_hdr(skb);
551 	ether_addr_copy(eth->h_source, amt->dev->dev_addr);
552 	ip_eth_mc_map(htonl(INADDR_ALLHOSTS_GROUP), eth->h_dest);
553 	eth->h_proto = htons(ETH_P_IP);
554 
555 	ihv3		= skb_pull(skb, sizeof(*iph) + AMT_IPHDR_OPTS);
556 	skb_reset_transport_header(skb);
557 	ihv3->type	= IGMP_HOST_MEMBERSHIP_QUERY;
558 	ihv3->code	= 1;
559 	ihv3->group	= 0;
560 	ihv3->qqic	= amt->qi;
561 	ihv3->nsrcs	= 0;
562 	ihv3->resv	= 0;
563 	ihv3->suppress	= false;
564 	ihv3->qrv	= amt->net->ipv4.sysctl_igmp_qrv;
565 	ihv3->csum	= 0;
566 	csum		= &ihv3->csum;
567 	csum_start	= (void *)ihv3;
568 	*csum		= ip_compute_csum(csum_start, sizeof(*ihv3));
569 	offset		= skb_transport_offset(skb);
570 	skb->csum	= skb_checksum(skb, offset, skb->len - offset, 0);
571 	skb->ip_summed	= CHECKSUM_NONE;
572 
573 	skb_push(skb, sizeof(*eth) + sizeof(*iph) + AMT_IPHDR_OPTS);
574 
575 	return skb;
576 }
577 
578 static void __amt_update_gw_status(struct amt_dev *amt, enum amt_status status,
579 				   bool validate)
580 {
581 	if (validate && amt->status >= status)
582 		return;
583 	netdev_dbg(amt->dev, "Update GW status %s -> %s",
584 		   status_str[amt->status], status_str[status]);
585 	amt->status = status;
586 }
587 
588 static void __amt_update_relay_status(struct amt_tunnel_list *tunnel,
589 				      enum amt_status status,
590 				      bool validate)
591 {
592 	if (validate && tunnel->status >= status)
593 		return;
594 	netdev_dbg(tunnel->amt->dev,
595 		   "Update Tunnel(IP = %pI4, PORT = %u) status %s -> %s",
596 		   &tunnel->ip4, ntohs(tunnel->source_port),
597 		   status_str[tunnel->status], status_str[status]);
598 	tunnel->status = status;
599 }
600 
601 static void amt_update_gw_status(struct amt_dev *amt, enum amt_status status,
602 				 bool validate)
603 {
604 	spin_lock_bh(&amt->lock);
605 	__amt_update_gw_status(amt, status, validate);
606 	spin_unlock_bh(&amt->lock);
607 }
608 
609 static void amt_update_relay_status(struct amt_tunnel_list *tunnel,
610 				    enum amt_status status, bool validate)
611 {
612 	spin_lock_bh(&tunnel->lock);
613 	__amt_update_relay_status(tunnel, status, validate);
614 	spin_unlock_bh(&tunnel->lock);
615 }
616 
617 static void amt_send_discovery(struct amt_dev *amt)
618 {
619 	struct amt_header_discovery *amtd;
620 	int hlen, tlen, offset;
621 	struct socket *sock;
622 	struct udphdr *udph;
623 	struct sk_buff *skb;
624 	struct iphdr *iph;
625 	struct rtable *rt;
626 	struct flowi4 fl4;
627 	u32 len;
628 	int err;
629 
630 	rcu_read_lock();
631 	sock = rcu_dereference(amt->sock);
632 	if (!sock)
633 		goto out;
634 
635 	if (!netif_running(amt->stream_dev) || !netif_running(amt->dev))
636 		goto out;
637 
638 	rt = ip_route_output_ports(amt->net, &fl4, sock->sk,
639 				   amt->discovery_ip, amt->local_ip,
640 				   amt->gw_port, amt->relay_port,
641 				   IPPROTO_UDP, 0,
642 				   amt->stream_dev->ifindex);
643 	if (IS_ERR(rt)) {
644 		amt->dev->stats.tx_errors++;
645 		goto out;
646 	}
647 
648 	hlen = LL_RESERVED_SPACE(amt->dev);
649 	tlen = amt->dev->needed_tailroom;
650 	len = hlen + tlen + sizeof(*iph) + sizeof(*udph) + sizeof(*amtd);
651 	skb = netdev_alloc_skb_ip_align(amt->dev, len);
652 	if (!skb) {
653 		ip_rt_put(rt);
654 		amt->dev->stats.tx_errors++;
655 		goto out;
656 	}
657 
658 	skb->priority = TC_PRIO_CONTROL;
659 	skb_dst_set(skb, &rt->dst);
660 
661 	len = sizeof(*iph) + sizeof(*udph) + sizeof(*amtd);
662 	skb_reset_network_header(skb);
663 	skb_put(skb, len);
664 	amtd = skb_pull(skb, sizeof(*iph) + sizeof(*udph));
665 	amtd->version	= 0;
666 	amtd->type	= AMT_MSG_DISCOVERY;
667 	amtd->reserved	= 0;
668 	amtd->nonce	= amt->nonce;
669 	skb_push(skb, sizeof(*udph));
670 	skb_reset_transport_header(skb);
671 	udph		= udp_hdr(skb);
672 	udph->source	= amt->gw_port;
673 	udph->dest	= amt->relay_port;
674 	udph->len	= htons(sizeof(*udph) + sizeof(*amtd));
675 	udph->check	= 0;
676 	offset = skb_transport_offset(skb);
677 	skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
678 	udph->check = csum_tcpudp_magic(amt->local_ip, amt->discovery_ip,
679 					sizeof(*udph) + sizeof(*amtd),
680 					IPPROTO_UDP, skb->csum);
681 
682 	skb_push(skb, sizeof(*iph));
683 	iph		= ip_hdr(skb);
684 	iph->version	= 4;
685 	iph->ihl	= (sizeof(struct iphdr)) >> 2;
686 	iph->tos	= AMT_TOS;
687 	iph->frag_off	= 0;
688 	iph->ttl	= ip4_dst_hoplimit(&rt->dst);
689 	iph->daddr	= amt->discovery_ip;
690 	iph->saddr	= amt->local_ip;
691 	iph->protocol	= IPPROTO_UDP;
692 	iph->tot_len	= htons(len);
693 
694 	skb->ip_summed = CHECKSUM_NONE;
695 	ip_select_ident(amt->net, skb, NULL);
696 	ip_send_check(iph);
697 	err = ip_local_out(amt->net, sock->sk, skb);
698 	if (unlikely(net_xmit_eval(err)))
699 		amt->dev->stats.tx_errors++;
700 
701 	spin_lock_bh(&amt->lock);
702 	__amt_update_gw_status(amt, AMT_STATUS_SENT_DISCOVERY, true);
703 	spin_unlock_bh(&amt->lock);
704 out:
705 	rcu_read_unlock();
706 }
707 
708 static void amt_send_request(struct amt_dev *amt, bool v6)
709 {
710 	struct amt_header_request *amtrh;
711 	int hlen, tlen, offset;
712 	struct socket *sock;
713 	struct udphdr *udph;
714 	struct sk_buff *skb;
715 	struct iphdr *iph;
716 	struct rtable *rt;
717 	struct flowi4 fl4;
718 	u32 len;
719 	int err;
720 
721 	rcu_read_lock();
722 	sock = rcu_dereference(amt->sock);
723 	if (!sock)
724 		goto out;
725 
726 	if (!netif_running(amt->stream_dev) || !netif_running(amt->dev))
727 		goto out;
728 
729 	rt = ip_route_output_ports(amt->net, &fl4, sock->sk,
730 				   amt->remote_ip, amt->local_ip,
731 				   amt->gw_port, amt->relay_port,
732 				   IPPROTO_UDP, 0,
733 				   amt->stream_dev->ifindex);
734 	if (IS_ERR(rt)) {
735 		amt->dev->stats.tx_errors++;
736 		goto out;
737 	}
738 
739 	hlen = LL_RESERVED_SPACE(amt->dev);
740 	tlen = amt->dev->needed_tailroom;
741 	len = hlen + tlen + sizeof(*iph) + sizeof(*udph) + sizeof(*amtrh);
742 	skb = netdev_alloc_skb_ip_align(amt->dev, len);
743 	if (!skb) {
744 		ip_rt_put(rt);
745 		amt->dev->stats.tx_errors++;
746 		goto out;
747 	}
748 
749 	skb->priority = TC_PRIO_CONTROL;
750 	skb_dst_set(skb, &rt->dst);
751 
752 	len = sizeof(*iph) + sizeof(*udph) + sizeof(*amtrh);
753 	skb_reset_network_header(skb);
754 	skb_put(skb, len);
755 	amtrh = skb_pull(skb, sizeof(*iph) + sizeof(*udph));
756 	amtrh->version	 = 0;
757 	amtrh->type	 = AMT_MSG_REQUEST;
758 	amtrh->reserved1 = 0;
759 	amtrh->p	 = v6;
760 	amtrh->reserved2 = 0;
761 	amtrh->nonce	 = amt->nonce;
762 	skb_push(skb, sizeof(*udph));
763 	skb_reset_transport_header(skb);
764 	udph		= udp_hdr(skb);
765 	udph->source	= amt->gw_port;
766 	udph->dest	= amt->relay_port;
767 	udph->len	= htons(sizeof(*amtrh) + sizeof(*udph));
768 	udph->check	= 0;
769 	offset = skb_transport_offset(skb);
770 	skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
771 	udph->check = csum_tcpudp_magic(amt->local_ip, amt->remote_ip,
772 					sizeof(*udph) + sizeof(*amtrh),
773 					IPPROTO_UDP, skb->csum);
774 
775 	skb_push(skb, sizeof(*iph));
776 	iph		= ip_hdr(skb);
777 	iph->version	= 4;
778 	iph->ihl	= (sizeof(struct iphdr)) >> 2;
779 	iph->tos	= AMT_TOS;
780 	iph->frag_off	= 0;
781 	iph->ttl	= ip4_dst_hoplimit(&rt->dst);
782 	iph->daddr	= amt->remote_ip;
783 	iph->saddr	= amt->local_ip;
784 	iph->protocol	= IPPROTO_UDP;
785 	iph->tot_len	= htons(len);
786 
787 	skb->ip_summed = CHECKSUM_NONE;
788 	ip_select_ident(amt->net, skb, NULL);
789 	ip_send_check(iph);
790 	err = ip_local_out(amt->net, sock->sk, skb);
791 	if (unlikely(net_xmit_eval(err)))
792 		amt->dev->stats.tx_errors++;
793 
794 out:
795 	rcu_read_unlock();
796 }
797 
798 static void amt_send_igmp_gq(struct amt_dev *amt,
799 			     struct amt_tunnel_list *tunnel)
800 {
801 	struct sk_buff *skb;
802 
803 	skb = amt_build_igmp_gq(amt);
804 	if (!skb)
805 		return;
806 
807 	amt_skb_cb(skb)->tunnel = tunnel;
808 	dev_queue_xmit(skb);
809 }
810 
811 #if IS_ENABLED(CONFIG_IPV6)
812 static struct sk_buff *amt_build_mld_gq(struct amt_dev *amt)
813 {
814 	u8 ra[AMT_IP6HDR_OPTS] = { IPPROTO_ICMPV6, 0, IPV6_TLV_ROUTERALERT,
815 				   2, 0, 0, IPV6_TLV_PAD1, IPV6_TLV_PAD1 };
816 	int hlen = LL_RESERVED_SPACE(amt->dev);
817 	int tlen = amt->dev->needed_tailroom;
818 	struct mld2_query *mld2q;
819 	void *csum_start = NULL;
820 	struct ipv6hdr *ip6h;
821 	struct sk_buff *skb;
822 	struct ethhdr *eth;
823 	u32 len;
824 
825 	len = hlen + tlen + sizeof(*ip6h) + sizeof(ra) + sizeof(*mld2q);
826 	skb = netdev_alloc_skb_ip_align(amt->dev, len);
827 	if (!skb)
828 		return NULL;
829 
830 	skb_reserve(skb, hlen);
831 	skb_push(skb, sizeof(*eth));
832 	skb_reset_mac_header(skb);
833 	eth = eth_hdr(skb);
834 	skb->priority = TC_PRIO_CONTROL;
835 	skb->protocol = htons(ETH_P_IPV6);
836 	skb_put_zero(skb, sizeof(*ip6h));
837 	skb_put_data(skb, ra, sizeof(ra));
838 	skb_put_zero(skb, sizeof(*mld2q));
839 	skb_pull(skb, sizeof(*eth));
840 	skb_reset_network_header(skb);
841 	ip6h			= ipv6_hdr(skb);
842 	ip6h->payload_len	= htons(sizeof(ra) + sizeof(*mld2q));
843 	ip6h->nexthdr		= NEXTHDR_HOP;
844 	ip6h->hop_limit		= 1;
845 	ip6h->daddr		= mld2_all_node;
846 	ip6_flow_hdr(ip6h, 0, 0);
847 
848 	if (ipv6_dev_get_saddr(amt->net, amt->dev, &ip6h->daddr, 0,
849 			       &ip6h->saddr)) {
850 		amt->dev->stats.tx_errors++;
851 		kfree_skb(skb);
852 		return NULL;
853 	}
854 
855 	eth->h_proto = htons(ETH_P_IPV6);
856 	ether_addr_copy(eth->h_source, amt->dev->dev_addr);
857 	ipv6_eth_mc_map(&mld2_all_node, eth->h_dest);
858 
859 	skb_pull(skb, sizeof(*ip6h) + sizeof(ra));
860 	skb_reset_transport_header(skb);
861 	mld2q			= (struct mld2_query *)icmp6_hdr(skb);
862 	mld2q->mld2q_mrc	= htons(1);
863 	mld2q->mld2q_type	= ICMPV6_MGM_QUERY;
864 	mld2q->mld2q_code	= 0;
865 	mld2q->mld2q_cksum	= 0;
866 	mld2q->mld2q_resv1	= 0;
867 	mld2q->mld2q_resv2	= 0;
868 	mld2q->mld2q_suppress	= 0;
869 	mld2q->mld2q_qrv	= amt->qrv;
870 	mld2q->mld2q_nsrcs	= 0;
871 	mld2q->mld2q_qqic	= amt->qi;
872 	csum_start		= (void *)mld2q;
873 	mld2q->mld2q_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
874 					     sizeof(*mld2q),
875 					     IPPROTO_ICMPV6,
876 					     csum_partial(csum_start,
877 							  sizeof(*mld2q), 0));
878 
879 	skb->ip_summed = CHECKSUM_NONE;
880 	skb_push(skb, sizeof(*eth) + sizeof(*ip6h) + sizeof(ra));
881 	return skb;
882 }
883 
884 static void amt_send_mld_gq(struct amt_dev *amt, struct amt_tunnel_list *tunnel)
885 {
886 	struct sk_buff *skb;
887 
888 	skb = amt_build_mld_gq(amt);
889 	if (!skb)
890 		return;
891 
892 	amt_skb_cb(skb)->tunnel = tunnel;
893 	dev_queue_xmit(skb);
894 }
895 #else
896 static void amt_send_mld_gq(struct amt_dev *amt, struct amt_tunnel_list *tunnel)
897 {
898 }
899 #endif
900 
901 static void amt_secret_work(struct work_struct *work)
902 {
903 	struct amt_dev *amt = container_of(to_delayed_work(work),
904 					   struct amt_dev,
905 					   secret_wq);
906 
907 	spin_lock_bh(&amt->lock);
908 	get_random_bytes(&amt->key, sizeof(siphash_key_t));
909 	spin_unlock_bh(&amt->lock);
910 	mod_delayed_work(amt_wq, &amt->secret_wq,
911 			 msecs_to_jiffies(AMT_SECRET_TIMEOUT));
912 }
913 
914 static void amt_discovery_work(struct work_struct *work)
915 {
916 	struct amt_dev *amt = container_of(to_delayed_work(work),
917 					   struct amt_dev,
918 					   discovery_wq);
919 
920 	spin_lock_bh(&amt->lock);
921 	if (amt->status > AMT_STATUS_SENT_DISCOVERY)
922 		goto out;
923 	get_random_bytes(&amt->nonce, sizeof(__be32));
924 	spin_unlock_bh(&amt->lock);
925 
926 	amt_send_discovery(amt);
927 	spin_lock_bh(&amt->lock);
928 out:
929 	mod_delayed_work(amt_wq, &amt->discovery_wq,
930 			 msecs_to_jiffies(AMT_DISCOVERY_TIMEOUT));
931 	spin_unlock_bh(&amt->lock);
932 }
933 
934 static void amt_req_work(struct work_struct *work)
935 {
936 	struct amt_dev *amt = container_of(to_delayed_work(work),
937 					   struct amt_dev,
938 					   req_wq);
939 	u32 exp;
940 
941 	spin_lock_bh(&amt->lock);
942 	if (amt->status < AMT_STATUS_RECEIVED_ADVERTISEMENT)
943 		goto out;
944 
945 	if (amt->req_cnt++ > AMT_MAX_REQ_COUNT) {
946 		netdev_dbg(amt->dev, "Gateway is not ready");
947 		amt->qi = AMT_INIT_REQ_TIMEOUT;
948 		amt->ready4 = false;
949 		amt->ready6 = false;
950 		amt->remote_ip = 0;
951 		__amt_update_gw_status(amt, AMT_STATUS_INIT, false);
952 		amt->req_cnt = 0;
953 	}
954 	spin_unlock_bh(&amt->lock);
955 
956 	amt_send_request(amt, false);
957 	amt_send_request(amt, true);
958 	amt_update_gw_status(amt, AMT_STATUS_SENT_REQUEST, true);
959 	spin_lock_bh(&amt->lock);
960 out:
961 	exp = min_t(u32, (1 * (1 << amt->req_cnt)), AMT_MAX_REQ_TIMEOUT);
962 	mod_delayed_work(amt_wq, &amt->req_wq, msecs_to_jiffies(exp * 1000));
963 	spin_unlock_bh(&amt->lock);
964 }
965 
966 static bool amt_send_membership_update(struct amt_dev *amt,
967 				       struct sk_buff *skb,
968 				       bool v6)
969 {
970 	struct amt_header_membership_update *amtmu;
971 	struct socket *sock;
972 	struct iphdr *iph;
973 	struct flowi4 fl4;
974 	struct rtable *rt;
975 	int err;
976 
977 	sock = rcu_dereference_bh(amt->sock);
978 	if (!sock)
979 		return true;
980 
981 	err = skb_cow_head(skb, LL_RESERVED_SPACE(amt->dev) + sizeof(*amtmu) +
982 			   sizeof(*iph) + sizeof(struct udphdr));
983 	if (err)
984 		return true;
985 
986 	skb_reset_inner_headers(skb);
987 	memset(&fl4, 0, sizeof(struct flowi4));
988 	fl4.flowi4_oif         = amt->stream_dev->ifindex;
989 	fl4.daddr              = amt->remote_ip;
990 	fl4.saddr              = amt->local_ip;
991 	fl4.flowi4_tos         = AMT_TOS;
992 	fl4.flowi4_proto       = IPPROTO_UDP;
993 	rt = ip_route_output_key(amt->net, &fl4);
994 	if (IS_ERR(rt)) {
995 		netdev_dbg(amt->dev, "no route to %pI4\n", &amt->remote_ip);
996 		return true;
997 	}
998 
999 	amtmu			= skb_push(skb, sizeof(*amtmu));
1000 	amtmu->version		= 0;
1001 	amtmu->type		= AMT_MSG_MEMBERSHIP_UPDATE;
1002 	amtmu->reserved		= 0;
1003 	amtmu->nonce		= amt->nonce;
1004 	amtmu->response_mac	= amt->mac;
1005 
1006 	if (!v6)
1007 		skb_set_inner_protocol(skb, htons(ETH_P_IP));
1008 	else
1009 		skb_set_inner_protocol(skb, htons(ETH_P_IPV6));
1010 	udp_tunnel_xmit_skb(rt, sock->sk, skb,
1011 			    fl4.saddr,
1012 			    fl4.daddr,
1013 			    AMT_TOS,
1014 			    ip4_dst_hoplimit(&rt->dst),
1015 			    0,
1016 			    amt->gw_port,
1017 			    amt->relay_port,
1018 			    false,
1019 			    false);
1020 	amt_update_gw_status(amt, AMT_STATUS_SENT_UPDATE, true);
1021 	return false;
1022 }
1023 
1024 static void amt_send_multicast_data(struct amt_dev *amt,
1025 				    const struct sk_buff *oskb,
1026 				    struct amt_tunnel_list *tunnel,
1027 				    bool v6)
1028 {
1029 	struct amt_header_mcast_data *amtmd;
1030 	struct socket *sock;
1031 	struct sk_buff *skb;
1032 	struct iphdr *iph;
1033 	struct flowi4 fl4;
1034 	struct rtable *rt;
1035 
1036 	sock = rcu_dereference_bh(amt->sock);
1037 	if (!sock)
1038 		return;
1039 
1040 	skb = skb_copy_expand(oskb, sizeof(*amtmd) + sizeof(*iph) +
1041 			      sizeof(struct udphdr), 0, GFP_ATOMIC);
1042 	if (!skb)
1043 		return;
1044 
1045 	skb_reset_inner_headers(skb);
1046 	memset(&fl4, 0, sizeof(struct flowi4));
1047 	fl4.flowi4_oif         = amt->stream_dev->ifindex;
1048 	fl4.daddr              = tunnel->ip4;
1049 	fl4.saddr              = amt->local_ip;
1050 	fl4.flowi4_proto       = IPPROTO_UDP;
1051 	rt = ip_route_output_key(amt->net, &fl4);
1052 	if (IS_ERR(rt)) {
1053 		netdev_dbg(amt->dev, "no route to %pI4\n", &tunnel->ip4);
1054 		kfree_skb(skb);
1055 		return;
1056 	}
1057 
1058 	amtmd = skb_push(skb, sizeof(*amtmd));
1059 	amtmd->version = 0;
1060 	amtmd->reserved = 0;
1061 	amtmd->type = AMT_MSG_MULTICAST_DATA;
1062 
1063 	if (!v6)
1064 		skb_set_inner_protocol(skb, htons(ETH_P_IP));
1065 	else
1066 		skb_set_inner_protocol(skb, htons(ETH_P_IPV6));
1067 	udp_tunnel_xmit_skb(rt, sock->sk, skb,
1068 			    fl4.saddr,
1069 			    fl4.daddr,
1070 			    AMT_TOS,
1071 			    ip4_dst_hoplimit(&rt->dst),
1072 			    0,
1073 			    amt->relay_port,
1074 			    tunnel->source_port,
1075 			    false,
1076 			    false);
1077 }
1078 
1079 static bool amt_send_membership_query(struct amt_dev *amt,
1080 				      struct sk_buff *skb,
1081 				      struct amt_tunnel_list *tunnel,
1082 				      bool v6)
1083 {
1084 	struct amt_header_membership_query *amtmq;
1085 	struct socket *sock;
1086 	struct rtable *rt;
1087 	struct flowi4 fl4;
1088 	int err;
1089 
1090 	sock = rcu_dereference_bh(amt->sock);
1091 	if (!sock)
1092 		return true;
1093 
1094 	err = skb_cow_head(skb, LL_RESERVED_SPACE(amt->dev) + sizeof(*amtmq) +
1095 			   sizeof(struct iphdr) + sizeof(struct udphdr));
1096 	if (err)
1097 		return true;
1098 
1099 	skb_reset_inner_headers(skb);
1100 	memset(&fl4, 0, sizeof(struct flowi4));
1101 	fl4.flowi4_oif         = amt->stream_dev->ifindex;
1102 	fl4.daddr              = tunnel->ip4;
1103 	fl4.saddr              = amt->local_ip;
1104 	fl4.flowi4_tos         = AMT_TOS;
1105 	fl4.flowi4_proto       = IPPROTO_UDP;
1106 	rt = ip_route_output_key(amt->net, &fl4);
1107 	if (IS_ERR(rt)) {
1108 		netdev_dbg(amt->dev, "no route to %pI4\n", &tunnel->ip4);
1109 		return -1;
1110 	}
1111 
1112 	amtmq		= skb_push(skb, sizeof(*amtmq));
1113 	amtmq->version	= 0;
1114 	amtmq->type	= AMT_MSG_MEMBERSHIP_QUERY;
1115 	amtmq->reserved = 0;
1116 	amtmq->l	= 0;
1117 	amtmq->g	= 0;
1118 	amtmq->nonce	= tunnel->nonce;
1119 	amtmq->response_mac = tunnel->mac;
1120 
1121 	if (!v6)
1122 		skb_set_inner_protocol(skb, htons(ETH_P_IP));
1123 	else
1124 		skb_set_inner_protocol(skb, htons(ETH_P_IPV6));
1125 	udp_tunnel_xmit_skb(rt, sock->sk, skb,
1126 			    fl4.saddr,
1127 			    fl4.daddr,
1128 			    AMT_TOS,
1129 			    ip4_dst_hoplimit(&rt->dst),
1130 			    0,
1131 			    amt->relay_port,
1132 			    tunnel->source_port,
1133 			    false,
1134 			    false);
1135 	amt_update_relay_status(tunnel, AMT_STATUS_SENT_QUERY, true);
1136 	return false;
1137 }
1138 
1139 static netdev_tx_t amt_dev_xmit(struct sk_buff *skb, struct net_device *dev)
1140 {
1141 	struct amt_dev *amt = netdev_priv(dev);
1142 	struct amt_tunnel_list *tunnel;
1143 	struct amt_group_node *gnode;
1144 	union amt_addr group = {0,};
1145 #if IS_ENABLED(CONFIG_IPV6)
1146 	struct ipv6hdr *ip6h;
1147 	struct mld_msg *mld;
1148 #endif
1149 	bool report = false;
1150 	struct igmphdr *ih;
1151 	bool query = false;
1152 	struct iphdr *iph;
1153 	bool data = false;
1154 	bool v6 = false;
1155 	u32 hash;
1156 
1157 	iph = ip_hdr(skb);
1158 	if (iph->version == 4) {
1159 		if (!ipv4_is_multicast(iph->daddr))
1160 			goto free;
1161 
1162 		if (!ip_mc_check_igmp(skb)) {
1163 			ih = igmp_hdr(skb);
1164 			switch (ih->type) {
1165 			case IGMPV3_HOST_MEMBERSHIP_REPORT:
1166 			case IGMP_HOST_MEMBERSHIP_REPORT:
1167 				report = true;
1168 				break;
1169 			case IGMP_HOST_MEMBERSHIP_QUERY:
1170 				query = true;
1171 				break;
1172 			default:
1173 				goto free;
1174 			}
1175 		} else {
1176 			data = true;
1177 		}
1178 		v6 = false;
1179 		group.ip4 = iph->daddr;
1180 #if IS_ENABLED(CONFIG_IPV6)
1181 	} else if (iph->version == 6) {
1182 		ip6h = ipv6_hdr(skb);
1183 		if (!ipv6_addr_is_multicast(&ip6h->daddr))
1184 			goto free;
1185 
1186 		if (!ipv6_mc_check_mld(skb)) {
1187 			mld = (struct mld_msg *)skb_transport_header(skb);
1188 			switch (mld->mld_type) {
1189 			case ICMPV6_MGM_REPORT:
1190 			case ICMPV6_MLD2_REPORT:
1191 				report = true;
1192 				break;
1193 			case ICMPV6_MGM_QUERY:
1194 				query = true;
1195 				break;
1196 			default:
1197 				goto free;
1198 			}
1199 		} else {
1200 			data = true;
1201 		}
1202 		v6 = true;
1203 		group.ip6 = ip6h->daddr;
1204 #endif
1205 	} else {
1206 		dev->stats.tx_errors++;
1207 		goto free;
1208 	}
1209 
1210 	if (!pskb_may_pull(skb, sizeof(struct ethhdr)))
1211 		goto free;
1212 
1213 	skb_pull(skb, sizeof(struct ethhdr));
1214 
1215 	if (amt->mode == AMT_MODE_GATEWAY) {
1216 		/* Gateway only passes IGMP/MLD packets */
1217 		if (!report)
1218 			goto free;
1219 		if ((!v6 && !amt->ready4) || (v6 && !amt->ready6))
1220 			goto free;
1221 		if (amt_send_membership_update(amt, skb,  v6))
1222 			goto free;
1223 		goto unlock;
1224 	} else if (amt->mode == AMT_MODE_RELAY) {
1225 		if (query) {
1226 			tunnel = amt_skb_cb(skb)->tunnel;
1227 			if (!tunnel) {
1228 				WARN_ON(1);
1229 				goto free;
1230 			}
1231 
1232 			/* Do not forward unexpected query */
1233 			if (amt_send_membership_query(amt, skb, tunnel, v6))
1234 				goto free;
1235 			goto unlock;
1236 		}
1237 
1238 		if (!data)
1239 			goto free;
1240 		list_for_each_entry_rcu(tunnel, &amt->tunnel_list, list) {
1241 			hash = amt_group_hash(tunnel, &group);
1242 			hlist_for_each_entry_rcu(gnode, &tunnel->groups[hash],
1243 						 node) {
1244 				if (!v6) {
1245 					if (gnode->group_addr.ip4 == iph->daddr)
1246 						goto found;
1247 #if IS_ENABLED(CONFIG_IPV6)
1248 				} else {
1249 					if (ipv6_addr_equal(&gnode->group_addr.ip6,
1250 							    &ip6h->daddr))
1251 						goto found;
1252 #endif
1253 				}
1254 			}
1255 			continue;
1256 found:
1257 			amt_send_multicast_data(amt, skb, tunnel, v6);
1258 		}
1259 	}
1260 
1261 	dev_kfree_skb(skb);
1262 	return NETDEV_TX_OK;
1263 free:
1264 	dev_kfree_skb(skb);
1265 unlock:
1266 	dev->stats.tx_dropped++;
1267 	return NETDEV_TX_OK;
1268 }
1269 
1270 static int amt_parse_type(struct sk_buff *skb)
1271 {
1272 	struct amt_header *amth;
1273 
1274 	if (!pskb_may_pull(skb, sizeof(struct udphdr) +
1275 			   sizeof(struct amt_header)))
1276 		return -1;
1277 
1278 	amth = (struct amt_header *)(udp_hdr(skb) + 1);
1279 
1280 	if (amth->version != 0)
1281 		return -1;
1282 
1283 	if (amth->type >= __AMT_MSG_MAX || !amth->type)
1284 		return -1;
1285 	return amth->type;
1286 }
1287 
1288 static void amt_clear_groups(struct amt_tunnel_list *tunnel)
1289 {
1290 	struct amt_dev *amt = tunnel->amt;
1291 	struct amt_group_node *gnode;
1292 	struct hlist_node *t;
1293 	int i;
1294 
1295 	spin_lock_bh(&tunnel->lock);
1296 	rcu_read_lock();
1297 	for (i = 0; i < amt->hash_buckets; i++)
1298 		hlist_for_each_entry_safe(gnode, t, &tunnel->groups[i], node)
1299 			amt_del_group(amt, gnode);
1300 	rcu_read_unlock();
1301 	spin_unlock_bh(&tunnel->lock);
1302 }
1303 
1304 static void amt_tunnel_expire(struct work_struct *work)
1305 {
1306 	struct amt_tunnel_list *tunnel = container_of(to_delayed_work(work),
1307 						      struct amt_tunnel_list,
1308 						      gc_wq);
1309 	struct amt_dev *amt = tunnel->amt;
1310 
1311 	spin_lock_bh(&amt->lock);
1312 	rcu_read_lock();
1313 	list_del_rcu(&tunnel->list);
1314 	amt->nr_tunnels--;
1315 	amt_clear_groups(tunnel);
1316 	rcu_read_unlock();
1317 	spin_unlock_bh(&amt->lock);
1318 	kfree_rcu(tunnel, rcu);
1319 }
1320 
1321 static void amt_cleanup_srcs(struct amt_dev *amt,
1322 			     struct amt_tunnel_list *tunnel,
1323 			     struct amt_group_node *gnode)
1324 {
1325 	struct amt_source_node *snode;
1326 	struct hlist_node *t;
1327 	int i;
1328 
1329 	/* Delete old sources */
1330 	for (i = 0; i < amt->hash_buckets; i++) {
1331 		hlist_for_each_entry_safe(snode, t, &gnode->sources[i], node) {
1332 			if (snode->flags == AMT_SOURCE_OLD)
1333 				amt_destroy_source(snode);
1334 		}
1335 	}
1336 
1337 	/* switch from new to old */
1338 	for (i = 0; i < amt->hash_buckets; i++)  {
1339 		hlist_for_each_entry_rcu(snode, &gnode->sources[i], node) {
1340 			snode->flags = AMT_SOURCE_OLD;
1341 			if (!gnode->v6)
1342 				netdev_dbg(snode->gnode->amt->dev,
1343 					   "Add source as OLD %pI4 from %pI4\n",
1344 					   &snode->source_addr.ip4,
1345 					   &gnode->group_addr.ip4);
1346 #if IS_ENABLED(CONFIG_IPV6)
1347 			else
1348 				netdev_dbg(snode->gnode->amt->dev,
1349 					   "Add source as OLD %pI6 from %pI6\n",
1350 					   &snode->source_addr.ip6,
1351 					   &gnode->group_addr.ip6);
1352 #endif
1353 		}
1354 	}
1355 }
1356 
1357 static void amt_add_srcs(struct amt_dev *amt, struct amt_tunnel_list *tunnel,
1358 			 struct amt_group_node *gnode, void *grec,
1359 			 bool v6)
1360 {
1361 	struct igmpv3_grec *igmp_grec;
1362 	struct amt_source_node *snode;
1363 #if IS_ENABLED(CONFIG_IPV6)
1364 	struct mld2_grec *mld_grec;
1365 #endif
1366 	union amt_addr src = {0,};
1367 	u16 nsrcs;
1368 	u32 hash;
1369 	int i;
1370 
1371 	if (!v6) {
1372 		igmp_grec = (struct igmpv3_grec *)grec;
1373 		nsrcs = ntohs(igmp_grec->grec_nsrcs);
1374 	} else {
1375 #if IS_ENABLED(CONFIG_IPV6)
1376 		mld_grec = (struct mld2_grec *)grec;
1377 		nsrcs = ntohs(mld_grec->grec_nsrcs);
1378 #else
1379 	return;
1380 #endif
1381 	}
1382 	for (i = 0; i < nsrcs; i++) {
1383 		if (tunnel->nr_sources >= amt->max_sources)
1384 			return;
1385 		if (!v6)
1386 			src.ip4 = igmp_grec->grec_src[i];
1387 #if IS_ENABLED(CONFIG_IPV6)
1388 		else
1389 			memcpy(&src.ip6, &mld_grec->grec_src[i],
1390 			       sizeof(struct in6_addr));
1391 #endif
1392 		if (amt_lookup_src(tunnel, gnode, AMT_FILTER_ALL, &src))
1393 			continue;
1394 
1395 		snode = amt_alloc_snode(gnode, &src);
1396 		if (snode) {
1397 			hash = amt_source_hash(tunnel, &snode->source_addr);
1398 			hlist_add_head_rcu(&snode->node, &gnode->sources[hash]);
1399 			tunnel->nr_sources++;
1400 			gnode->nr_sources++;
1401 
1402 			if (!gnode->v6)
1403 				netdev_dbg(snode->gnode->amt->dev,
1404 					   "Add source as NEW %pI4 from %pI4\n",
1405 					   &snode->source_addr.ip4,
1406 					   &gnode->group_addr.ip4);
1407 #if IS_ENABLED(CONFIG_IPV6)
1408 			else
1409 				netdev_dbg(snode->gnode->amt->dev,
1410 					   "Add source as NEW %pI6 from %pI6\n",
1411 					   &snode->source_addr.ip6,
1412 					   &gnode->group_addr.ip6);
1413 #endif
1414 		}
1415 	}
1416 }
1417 
1418 /* Router State   Report Rec'd New Router State
1419  * ------------   ------------ ----------------
1420  * EXCLUDE (X,Y)  IS_IN (A)    EXCLUDE (X+A,Y-A)
1421  *
1422  * -----------+-----------+-----------+
1423  *            |    OLD    |    NEW    |
1424  * -----------+-----------+-----------+
1425  *    FWD     |     X     |    X+A    |
1426  * -----------+-----------+-----------+
1427  *    D_FWD   |     Y     |    Y-A    |
1428  * -----------+-----------+-----------+
1429  *    NONE    |           |     A     |
1430  * -----------+-----------+-----------+
1431  *
1432  * a) Received sources are NONE/NEW
1433  * b) All NONE will be deleted by amt_cleanup_srcs().
1434  * c) All OLD will be deleted by amt_cleanup_srcs().
1435  * d) After delete, NEW source will be switched to OLD.
1436  */
1437 static void amt_lookup_act_srcs(struct amt_tunnel_list *tunnel,
1438 				struct amt_group_node *gnode,
1439 				void *grec,
1440 				enum amt_ops ops,
1441 				enum amt_filter filter,
1442 				enum amt_act act,
1443 				bool v6)
1444 {
1445 	struct amt_dev *amt = tunnel->amt;
1446 	struct amt_source_node *snode;
1447 	struct igmpv3_grec *igmp_grec;
1448 #if IS_ENABLED(CONFIG_IPV6)
1449 	struct mld2_grec *mld_grec;
1450 #endif
1451 	union amt_addr src = {0,};
1452 	struct hlist_node *t;
1453 	u16 nsrcs;
1454 	int i, j;
1455 
1456 	if (!v6) {
1457 		igmp_grec = (struct igmpv3_grec *)grec;
1458 		nsrcs = ntohs(igmp_grec->grec_nsrcs);
1459 	} else {
1460 #if IS_ENABLED(CONFIG_IPV6)
1461 		mld_grec = (struct mld2_grec *)grec;
1462 		nsrcs = ntohs(mld_grec->grec_nsrcs);
1463 #else
1464 	return;
1465 #endif
1466 	}
1467 
1468 	memset(&src, 0, sizeof(union amt_addr));
1469 	switch (ops) {
1470 	case AMT_OPS_INT:
1471 		/* A*B */
1472 		for (i = 0; i < nsrcs; i++) {
1473 			if (!v6)
1474 				src.ip4 = igmp_grec->grec_src[i];
1475 #if IS_ENABLED(CONFIG_IPV6)
1476 			else
1477 				memcpy(&src.ip6, &mld_grec->grec_src[i],
1478 				       sizeof(struct in6_addr));
1479 #endif
1480 			snode = amt_lookup_src(tunnel, gnode, filter, &src);
1481 			if (!snode)
1482 				continue;
1483 			amt_act_src(tunnel, gnode, snode, act);
1484 		}
1485 		break;
1486 	case AMT_OPS_UNI:
1487 		/* A+B */
1488 		for (i = 0; i < amt->hash_buckets; i++) {
1489 			hlist_for_each_entry_safe(snode, t, &gnode->sources[i],
1490 						  node) {
1491 				if (amt_status_filter(snode, filter))
1492 					amt_act_src(tunnel, gnode, snode, act);
1493 			}
1494 		}
1495 		for (i = 0; i < nsrcs; i++) {
1496 			if (!v6)
1497 				src.ip4 = igmp_grec->grec_src[i];
1498 #if IS_ENABLED(CONFIG_IPV6)
1499 			else
1500 				memcpy(&src.ip6, &mld_grec->grec_src[i],
1501 				       sizeof(struct in6_addr));
1502 #endif
1503 			snode = amt_lookup_src(tunnel, gnode, filter, &src);
1504 			if (!snode)
1505 				continue;
1506 			amt_act_src(tunnel, gnode, snode, act);
1507 		}
1508 		break;
1509 	case AMT_OPS_SUB:
1510 		/* A-B */
1511 		for (i = 0; i < amt->hash_buckets; i++) {
1512 			hlist_for_each_entry_safe(snode, t, &gnode->sources[i],
1513 						  node) {
1514 				if (!amt_status_filter(snode, filter))
1515 					continue;
1516 				for (j = 0; j < nsrcs; j++) {
1517 					if (!v6)
1518 						src.ip4 = igmp_grec->grec_src[j];
1519 #if IS_ENABLED(CONFIG_IPV6)
1520 					else
1521 						memcpy(&src.ip6,
1522 						       &mld_grec->grec_src[j],
1523 						       sizeof(struct in6_addr));
1524 #endif
1525 					if (amt_addr_equal(&snode->source_addr,
1526 							   &src))
1527 						goto out_sub;
1528 				}
1529 				amt_act_src(tunnel, gnode, snode, act);
1530 				continue;
1531 out_sub:;
1532 			}
1533 		}
1534 		break;
1535 	case AMT_OPS_SUB_REV:
1536 		/* B-A */
1537 		for (i = 0; i < nsrcs; i++) {
1538 			if (!v6)
1539 				src.ip4 = igmp_grec->grec_src[i];
1540 #if IS_ENABLED(CONFIG_IPV6)
1541 			else
1542 				memcpy(&src.ip6, &mld_grec->grec_src[i],
1543 				       sizeof(struct in6_addr));
1544 #endif
1545 			snode = amt_lookup_src(tunnel, gnode, AMT_FILTER_ALL,
1546 					       &src);
1547 			if (!snode) {
1548 				snode = amt_lookup_src(tunnel, gnode,
1549 						       filter, &src);
1550 				if (snode)
1551 					amt_act_src(tunnel, gnode, snode, act);
1552 			}
1553 		}
1554 		break;
1555 	default:
1556 		netdev_dbg(amt->dev, "Invalid type\n");
1557 		return;
1558 	}
1559 }
1560 
1561 static void amt_mcast_is_in_handler(struct amt_dev *amt,
1562 				    struct amt_tunnel_list *tunnel,
1563 				    struct amt_group_node *gnode,
1564 				    void *grec, void *zero_grec, bool v6)
1565 {
1566 	if (gnode->filter_mode == MCAST_INCLUDE) {
1567 /* Router State   Report Rec'd New Router State        Actions
1568  * ------------   ------------ ----------------        -------
1569  * INCLUDE (A)    IS_IN (B)    INCLUDE (A+B)           (B)=GMI
1570  */
1571 		/* Update IS_IN (B) as FWD/NEW */
1572 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1573 				    AMT_FILTER_NONE_NEW,
1574 				    AMT_ACT_STATUS_FWD_NEW,
1575 				    v6);
1576 		/* Update INCLUDE (A) as NEW */
1577 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1578 				    AMT_FILTER_FWD,
1579 				    AMT_ACT_STATUS_FWD_NEW,
1580 				    v6);
1581 		/* (B)=GMI */
1582 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1583 				    AMT_FILTER_FWD_NEW,
1584 				    AMT_ACT_GMI,
1585 				    v6);
1586 	} else {
1587 /* State        Actions
1588  * ------------   ------------ ----------------        -------
1589  * EXCLUDE (X,Y)  IS_IN (A)    EXCLUDE (X+A,Y-A)       (A)=GMI
1590  */
1591 		/* Update (A) in (X, Y) as NONE/NEW */
1592 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1593 				    AMT_FILTER_BOTH,
1594 				    AMT_ACT_STATUS_NONE_NEW,
1595 				    v6);
1596 		/* Update FWD/OLD as FWD/NEW */
1597 		amt_lookup_act_srcs(tunnel, gnode, zero_grec, AMT_OPS_UNI,
1598 				    AMT_FILTER_FWD,
1599 				    AMT_ACT_STATUS_FWD_NEW,
1600 				    v6);
1601 		/* Update IS_IN (A) as FWD/NEW */
1602 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1603 				    AMT_FILTER_NONE_NEW,
1604 				    AMT_ACT_STATUS_FWD_NEW,
1605 				    v6);
1606 		/* Update EXCLUDE (, Y-A) as D_FWD_NEW */
1607 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB,
1608 				    AMT_FILTER_D_FWD,
1609 				    AMT_ACT_STATUS_D_FWD_NEW,
1610 				    v6);
1611 	}
1612 }
1613 
1614 static void amt_mcast_is_ex_handler(struct amt_dev *amt,
1615 				    struct amt_tunnel_list *tunnel,
1616 				    struct amt_group_node *gnode,
1617 				    void *grec, void *zero_grec, bool v6)
1618 {
1619 	if (gnode->filter_mode == MCAST_INCLUDE) {
1620 /* Router State   Report Rec'd  New Router State         Actions
1621  * ------------   ------------  ----------------         -------
1622  * INCLUDE (A)    IS_EX (B)     EXCLUDE (A*B,B-A)        (B-A)=0
1623  *                                                       Delete (A-B)
1624  *                                                       Group Timer=GMI
1625  */
1626 		/* EXCLUDE(A*B, ) */
1627 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1628 				    AMT_FILTER_FWD,
1629 				    AMT_ACT_STATUS_FWD_NEW,
1630 				    v6);
1631 		/* EXCLUDE(, B-A) */
1632 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB_REV,
1633 				    AMT_FILTER_FWD,
1634 				    AMT_ACT_STATUS_D_FWD_NEW,
1635 				    v6);
1636 		/* (B-A)=0 */
1637 		amt_lookup_act_srcs(tunnel, gnode, zero_grec, AMT_OPS_UNI,
1638 				    AMT_FILTER_D_FWD_NEW,
1639 				    AMT_ACT_GMI_ZERO,
1640 				    v6);
1641 		/* Group Timer=GMI */
1642 		if (!mod_delayed_work(amt_wq, &gnode->group_timer,
1643 				      msecs_to_jiffies(amt_gmi(amt))))
1644 			dev_hold(amt->dev);
1645 		gnode->filter_mode = MCAST_EXCLUDE;
1646 		/* Delete (A-B) will be worked by amt_cleanup_srcs(). */
1647 	} else {
1648 /* Router State   Report Rec'd  New Router State	Actions
1649  * ------------   ------------  ----------------	-------
1650  * EXCLUDE (X,Y)  IS_EX (A)     EXCLUDE (A-Y,Y*A)	(A-X-Y)=GMI
1651  *							Delete (X-A)
1652  *							Delete (Y-A)
1653  *							Group Timer=GMI
1654  */
1655 		/* EXCLUDE (A-Y, ) */
1656 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB_REV,
1657 				    AMT_FILTER_D_FWD,
1658 				    AMT_ACT_STATUS_FWD_NEW,
1659 				    v6);
1660 		/* EXCLUDE (, Y*A ) */
1661 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1662 				    AMT_FILTER_D_FWD,
1663 				    AMT_ACT_STATUS_D_FWD_NEW,
1664 				    v6);
1665 		/* (A-X-Y)=GMI */
1666 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB_REV,
1667 				    AMT_FILTER_BOTH_NEW,
1668 				    AMT_ACT_GMI,
1669 				    v6);
1670 		/* Group Timer=GMI */
1671 		if (!mod_delayed_work(amt_wq, &gnode->group_timer,
1672 				      msecs_to_jiffies(amt_gmi(amt))))
1673 			dev_hold(amt->dev);
1674 		/* Delete (X-A), (Y-A) will be worked by amt_cleanup_srcs(). */
1675 	}
1676 }
1677 
1678 static void amt_mcast_to_in_handler(struct amt_dev *amt,
1679 				    struct amt_tunnel_list *tunnel,
1680 				    struct amt_group_node *gnode,
1681 				    void *grec, void *zero_grec, bool v6)
1682 {
1683 	if (gnode->filter_mode == MCAST_INCLUDE) {
1684 /* Router State   Report Rec'd New Router State        Actions
1685  * ------------   ------------ ----------------        -------
1686  * INCLUDE (A)    TO_IN (B)    INCLUDE (A+B)           (B)=GMI
1687  *						       Send Q(G,A-B)
1688  */
1689 		/* Update TO_IN (B) sources as FWD/NEW */
1690 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1691 				    AMT_FILTER_NONE_NEW,
1692 				    AMT_ACT_STATUS_FWD_NEW,
1693 				    v6);
1694 		/* Update INCLUDE (A) sources as NEW */
1695 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1696 				    AMT_FILTER_FWD,
1697 				    AMT_ACT_STATUS_FWD_NEW,
1698 				    v6);
1699 		/* (B)=GMI */
1700 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1701 				    AMT_FILTER_FWD_NEW,
1702 				    AMT_ACT_GMI,
1703 				    v6);
1704 	} else {
1705 /* Router State   Report Rec'd New Router State        Actions
1706  * ------------   ------------ ----------------        -------
1707  * EXCLUDE (X,Y)  TO_IN (A)    EXCLUDE (X+A,Y-A)       (A)=GMI
1708  *						       Send Q(G,X-A)
1709  *						       Send Q(G)
1710  */
1711 		/* Update TO_IN (A) sources as FWD/NEW */
1712 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1713 				    AMT_FILTER_NONE_NEW,
1714 				    AMT_ACT_STATUS_FWD_NEW,
1715 				    v6);
1716 		/* Update EXCLUDE(X,) sources as FWD/NEW */
1717 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1718 				    AMT_FILTER_FWD,
1719 				    AMT_ACT_STATUS_FWD_NEW,
1720 				    v6);
1721 		/* EXCLUDE (, Y-A)
1722 		 * (A) are already switched to FWD_NEW.
1723 		 * So, D_FWD/OLD -> D_FWD/NEW is okay.
1724 		 */
1725 		amt_lookup_act_srcs(tunnel, gnode, zero_grec, AMT_OPS_UNI,
1726 				    AMT_FILTER_D_FWD,
1727 				    AMT_ACT_STATUS_D_FWD_NEW,
1728 				    v6);
1729 		/* (A)=GMI
1730 		 * Only FWD_NEW will have (A) sources.
1731 		 */
1732 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1733 				    AMT_FILTER_FWD_NEW,
1734 				    AMT_ACT_GMI,
1735 				    v6);
1736 	}
1737 }
1738 
1739 static void amt_mcast_to_ex_handler(struct amt_dev *amt,
1740 				    struct amt_tunnel_list *tunnel,
1741 				    struct amt_group_node *gnode,
1742 				    void *grec, void *zero_grec, bool v6)
1743 {
1744 	if (gnode->filter_mode == MCAST_INCLUDE) {
1745 /* Router State   Report Rec'd New Router State        Actions
1746  * ------------   ------------ ----------------        -------
1747  * INCLUDE (A)    TO_EX (B)    EXCLUDE (A*B,B-A)       (B-A)=0
1748  *						       Delete (A-B)
1749  *						       Send Q(G,A*B)
1750  *						       Group Timer=GMI
1751  */
1752 		/* EXCLUDE (A*B, ) */
1753 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1754 				    AMT_FILTER_FWD,
1755 				    AMT_ACT_STATUS_FWD_NEW,
1756 				    v6);
1757 		/* EXCLUDE (, B-A) */
1758 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB_REV,
1759 				    AMT_FILTER_FWD,
1760 				    AMT_ACT_STATUS_D_FWD_NEW,
1761 				    v6);
1762 		/* (B-A)=0 */
1763 		amt_lookup_act_srcs(tunnel, gnode, zero_grec, AMT_OPS_UNI,
1764 				    AMT_FILTER_D_FWD_NEW,
1765 				    AMT_ACT_GMI_ZERO,
1766 				    v6);
1767 		/* Group Timer=GMI */
1768 		if (!mod_delayed_work(amt_wq, &gnode->group_timer,
1769 				      msecs_to_jiffies(amt_gmi(amt))))
1770 			dev_hold(amt->dev);
1771 		gnode->filter_mode = MCAST_EXCLUDE;
1772 		/* Delete (A-B) will be worked by amt_cleanup_srcs(). */
1773 	} else {
1774 /* Router State   Report Rec'd New Router State        Actions
1775  * ------------   ------------ ----------------        -------
1776  * EXCLUDE (X,Y)  TO_EX (A)    EXCLUDE (A-Y,Y*A)       (A-X-Y)=Group Timer
1777  *						       Delete (X-A)
1778  *						       Delete (Y-A)
1779  *						       Send Q(G,A-Y)
1780  *						       Group Timer=GMI
1781  */
1782 		/* Update (A-X-Y) as NONE/OLD */
1783 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB_REV,
1784 				    AMT_FILTER_BOTH,
1785 				    AMT_ACT_GT,
1786 				    v6);
1787 		/* EXCLUDE (A-Y, ) */
1788 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB_REV,
1789 				    AMT_FILTER_D_FWD,
1790 				    AMT_ACT_STATUS_FWD_NEW,
1791 				    v6);
1792 		/* EXCLUDE (, Y*A) */
1793 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1794 				    AMT_FILTER_D_FWD,
1795 				    AMT_ACT_STATUS_D_FWD_NEW,
1796 				    v6);
1797 		/* Group Timer=GMI */
1798 		if (!mod_delayed_work(amt_wq, &gnode->group_timer,
1799 				      msecs_to_jiffies(amt_gmi(amt))))
1800 			dev_hold(amt->dev);
1801 		/* Delete (X-A), (Y-A) will be worked by amt_cleanup_srcs(). */
1802 	}
1803 }
1804 
1805 static void amt_mcast_allow_handler(struct amt_dev *amt,
1806 				    struct amt_tunnel_list *tunnel,
1807 				    struct amt_group_node *gnode,
1808 				    void *grec, void *zero_grec, bool v6)
1809 {
1810 	if (gnode->filter_mode == MCAST_INCLUDE) {
1811 /* Router State   Report Rec'd New Router State        Actions
1812  * ------------   ------------ ----------------        -------
1813  * INCLUDE (A)    ALLOW (B)    INCLUDE (A+B)	       (B)=GMI
1814  */
1815 		/* INCLUDE (A+B) */
1816 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1817 				    AMT_FILTER_FWD,
1818 				    AMT_ACT_STATUS_FWD_NEW,
1819 				    v6);
1820 		/* (B)=GMI */
1821 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1822 				    AMT_FILTER_FWD_NEW,
1823 				    AMT_ACT_GMI,
1824 				    v6);
1825 	} else {
1826 /* Router State   Report Rec'd New Router State        Actions
1827  * ------------   ------------ ----------------        -------
1828  * EXCLUDE (X,Y)  ALLOW (A)    EXCLUDE (X+A,Y-A)       (A)=GMI
1829  */
1830 		/* EXCLUDE (X+A, ) */
1831 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1832 				    AMT_FILTER_FWD,
1833 				    AMT_ACT_STATUS_FWD_NEW,
1834 				    v6);
1835 		/* EXCLUDE (, Y-A) */
1836 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB,
1837 				    AMT_FILTER_D_FWD,
1838 				    AMT_ACT_STATUS_D_FWD_NEW,
1839 				    v6);
1840 		/* (A)=GMI
1841 		 * All (A) source are now FWD/NEW status.
1842 		 */
1843 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_INT,
1844 				    AMT_FILTER_FWD_NEW,
1845 				    AMT_ACT_GMI,
1846 				    v6);
1847 	}
1848 }
1849 
1850 static void amt_mcast_block_handler(struct amt_dev *amt,
1851 				    struct amt_tunnel_list *tunnel,
1852 				    struct amt_group_node *gnode,
1853 				    void *grec, void *zero_grec, bool v6)
1854 {
1855 	if (gnode->filter_mode == MCAST_INCLUDE) {
1856 /* Router State   Report Rec'd New Router State        Actions
1857  * ------------   ------------ ----------------        -------
1858  * INCLUDE (A)    BLOCK (B)    INCLUDE (A)             Send Q(G,A*B)
1859  */
1860 		/* INCLUDE (A) */
1861 		amt_lookup_act_srcs(tunnel, gnode, zero_grec, AMT_OPS_UNI,
1862 				    AMT_FILTER_FWD,
1863 				    AMT_ACT_STATUS_FWD_NEW,
1864 				    v6);
1865 	} else {
1866 /* Router State   Report Rec'd New Router State        Actions
1867  * ------------   ------------ ----------------        -------
1868  * EXCLUDE (X,Y)  BLOCK (A)    EXCLUDE (X+(A-Y),Y)     (A-X-Y)=Group Timer
1869  *						       Send Q(G,A-Y)
1870  */
1871 		/* (A-X-Y)=Group Timer */
1872 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB_REV,
1873 				    AMT_FILTER_BOTH,
1874 				    AMT_ACT_GT,
1875 				    v6);
1876 		/* EXCLUDE (X, ) */
1877 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1878 				    AMT_FILTER_FWD,
1879 				    AMT_ACT_STATUS_FWD_NEW,
1880 				    v6);
1881 		/* EXCLUDE (X+(A-Y) */
1882 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_SUB_REV,
1883 				    AMT_FILTER_D_FWD,
1884 				    AMT_ACT_STATUS_FWD_NEW,
1885 				    v6);
1886 		/* EXCLUDE (, Y) */
1887 		amt_lookup_act_srcs(tunnel, gnode, grec, AMT_OPS_UNI,
1888 				    AMT_FILTER_D_FWD,
1889 				    AMT_ACT_STATUS_D_FWD_NEW,
1890 				    v6);
1891 	}
1892 }
1893 
1894 /* RFC 3376
1895  * 7.3.2. In the Presence of Older Version Group Members
1896  *
1897  * When Group Compatibility Mode is IGMPv2, a router internally
1898  * translates the following IGMPv2 messages for that group to their
1899  * IGMPv3 equivalents:
1900  *
1901  * IGMPv2 Message                IGMPv3 Equivalent
1902  * --------------                -----------------
1903  * Report                        IS_EX( {} )
1904  * Leave                         TO_IN( {} )
1905  */
1906 static void amt_igmpv2_report_handler(struct amt_dev *amt, struct sk_buff *skb,
1907 				      struct amt_tunnel_list *tunnel)
1908 {
1909 	struct igmphdr *ih = igmp_hdr(skb);
1910 	struct iphdr *iph = ip_hdr(skb);
1911 	struct amt_group_node *gnode;
1912 	union amt_addr group, host;
1913 
1914 	memset(&group, 0, sizeof(union amt_addr));
1915 	group.ip4 = ih->group;
1916 	memset(&host, 0, sizeof(union amt_addr));
1917 	host.ip4 = iph->saddr;
1918 
1919 	gnode = amt_lookup_group(tunnel, &group, &host, false);
1920 	if (!gnode) {
1921 		gnode = amt_add_group(amt, tunnel, &group, &host, false);
1922 		if (!IS_ERR(gnode)) {
1923 			gnode->filter_mode = MCAST_EXCLUDE;
1924 			if (!mod_delayed_work(amt_wq, &gnode->group_timer,
1925 					      msecs_to_jiffies(amt_gmi(amt))))
1926 				dev_hold(amt->dev);
1927 		}
1928 	}
1929 }
1930 
1931 /* RFC 3376
1932  * 7.3.2. In the Presence of Older Version Group Members
1933  *
1934  * When Group Compatibility Mode is IGMPv2, a router internally
1935  * translates the following IGMPv2 messages for that group to their
1936  * IGMPv3 equivalents:
1937  *
1938  * IGMPv2 Message                IGMPv3 Equivalent
1939  * --------------                -----------------
1940  * Report                        IS_EX( {} )
1941  * Leave                         TO_IN( {} )
1942  */
1943 static void amt_igmpv2_leave_handler(struct amt_dev *amt, struct sk_buff *skb,
1944 				     struct amt_tunnel_list *tunnel)
1945 {
1946 	struct igmphdr *ih = igmp_hdr(skb);
1947 	struct iphdr *iph = ip_hdr(skb);
1948 	struct amt_group_node *gnode;
1949 	union amt_addr group, host;
1950 
1951 	memset(&group, 0, sizeof(union amt_addr));
1952 	group.ip4 = ih->group;
1953 	memset(&host, 0, sizeof(union amt_addr));
1954 	host.ip4 = iph->saddr;
1955 
1956 	gnode = amt_lookup_group(tunnel, &group, &host, false);
1957 	if (gnode)
1958 		amt_del_group(amt, gnode);
1959 }
1960 
1961 static void amt_igmpv3_report_handler(struct amt_dev *amt, struct sk_buff *skb,
1962 				      struct amt_tunnel_list *tunnel)
1963 {
1964 	struct igmpv3_report *ihrv3 = igmpv3_report_hdr(skb);
1965 	int len = skb_transport_offset(skb) + sizeof(*ihrv3);
1966 	void *zero_grec = (void *)&igmpv3_zero_grec;
1967 	struct iphdr *iph = ip_hdr(skb);
1968 	struct amt_group_node *gnode;
1969 	union amt_addr group, host;
1970 	struct igmpv3_grec *grec;
1971 	u16 nsrcs;
1972 	int i;
1973 
1974 	for (i = 0; i < ntohs(ihrv3->ngrec); i++) {
1975 		len += sizeof(*grec);
1976 		if (!ip_mc_may_pull(skb, len))
1977 			break;
1978 
1979 		grec = (void *)(skb->data + len - sizeof(*grec));
1980 		nsrcs = ntohs(grec->grec_nsrcs);
1981 
1982 		len += nsrcs * sizeof(__be32);
1983 		if (!ip_mc_may_pull(skb, len))
1984 			break;
1985 
1986 		memset(&group, 0, sizeof(union amt_addr));
1987 		group.ip4 = grec->grec_mca;
1988 		memset(&host, 0, sizeof(union amt_addr));
1989 		host.ip4 = iph->saddr;
1990 		gnode = amt_lookup_group(tunnel, &group, &host, false);
1991 		if (!gnode) {
1992 			gnode = amt_add_group(amt, tunnel, &group, &host,
1993 					      false);
1994 			if (IS_ERR(gnode))
1995 				continue;
1996 		}
1997 
1998 		amt_add_srcs(amt, tunnel, gnode, grec, false);
1999 		switch (grec->grec_type) {
2000 		case IGMPV3_MODE_IS_INCLUDE:
2001 			amt_mcast_is_in_handler(amt, tunnel, gnode, grec,
2002 						zero_grec, false);
2003 			break;
2004 		case IGMPV3_MODE_IS_EXCLUDE:
2005 			amt_mcast_is_ex_handler(amt, tunnel, gnode, grec,
2006 						zero_grec, false);
2007 			break;
2008 		case IGMPV3_CHANGE_TO_INCLUDE:
2009 			amt_mcast_to_in_handler(amt, tunnel, gnode, grec,
2010 						zero_grec, false);
2011 			break;
2012 		case IGMPV3_CHANGE_TO_EXCLUDE:
2013 			amt_mcast_to_ex_handler(amt, tunnel, gnode, grec,
2014 						zero_grec, false);
2015 			break;
2016 		case IGMPV3_ALLOW_NEW_SOURCES:
2017 			amt_mcast_allow_handler(amt, tunnel, gnode, grec,
2018 						zero_grec, false);
2019 			break;
2020 		case IGMPV3_BLOCK_OLD_SOURCES:
2021 			amt_mcast_block_handler(amt, tunnel, gnode, grec,
2022 						zero_grec, false);
2023 			break;
2024 		default:
2025 			break;
2026 		}
2027 		amt_cleanup_srcs(amt, tunnel, gnode);
2028 	}
2029 }
2030 
2031 /* caller held tunnel->lock */
2032 static void amt_igmp_report_handler(struct amt_dev *amt, struct sk_buff *skb,
2033 				    struct amt_tunnel_list *tunnel)
2034 {
2035 	struct igmphdr *ih = igmp_hdr(skb);
2036 
2037 	switch (ih->type) {
2038 	case IGMPV3_HOST_MEMBERSHIP_REPORT:
2039 		amt_igmpv3_report_handler(amt, skb, tunnel);
2040 		break;
2041 	case IGMPV2_HOST_MEMBERSHIP_REPORT:
2042 		amt_igmpv2_report_handler(amt, skb, tunnel);
2043 		break;
2044 	case IGMP_HOST_LEAVE_MESSAGE:
2045 		amt_igmpv2_leave_handler(amt, skb, tunnel);
2046 		break;
2047 	default:
2048 		break;
2049 	}
2050 }
2051 
2052 #if IS_ENABLED(CONFIG_IPV6)
2053 /* RFC 3810
2054  * 8.3.2. In the Presence of MLDv1 Multicast Address Listeners
2055  *
2056  * When Multicast Address Compatibility Mode is MLDv2, a router acts
2057  * using the MLDv2 protocol for that multicast address.  When Multicast
2058  * Address Compatibility Mode is MLDv1, a router internally translates
2059  * the following MLDv1 messages for that multicast address to their
2060  * MLDv2 equivalents:
2061  *
2062  * MLDv1 Message                 MLDv2 Equivalent
2063  * --------------                -----------------
2064  * Report                        IS_EX( {} )
2065  * Done                          TO_IN( {} )
2066  */
2067 static void amt_mldv1_report_handler(struct amt_dev *amt, struct sk_buff *skb,
2068 				     struct amt_tunnel_list *tunnel)
2069 {
2070 	struct mld_msg *mld = (struct mld_msg *)icmp6_hdr(skb);
2071 	struct ipv6hdr *ip6h = ipv6_hdr(skb);
2072 	struct amt_group_node *gnode;
2073 	union amt_addr group, host;
2074 
2075 	memcpy(&group.ip6, &mld->mld_mca, sizeof(struct in6_addr));
2076 	memcpy(&host.ip6, &ip6h->saddr, sizeof(struct in6_addr));
2077 
2078 	gnode = amt_lookup_group(tunnel, &group, &host, true);
2079 	if (!gnode) {
2080 		gnode = amt_add_group(amt, tunnel, &group, &host, true);
2081 		if (!IS_ERR(gnode)) {
2082 			gnode->filter_mode = MCAST_EXCLUDE;
2083 			if (!mod_delayed_work(amt_wq, &gnode->group_timer,
2084 					      msecs_to_jiffies(amt_gmi(amt))))
2085 				dev_hold(amt->dev);
2086 		}
2087 	}
2088 }
2089 
2090 /* RFC 3810
2091  * 8.3.2. In the Presence of MLDv1 Multicast Address Listeners
2092  *
2093  * When Multicast Address Compatibility Mode is MLDv2, a router acts
2094  * using the MLDv2 protocol for that multicast address.  When Multicast
2095  * Address Compatibility Mode is MLDv1, a router internally translates
2096  * the following MLDv1 messages for that multicast address to their
2097  * MLDv2 equivalents:
2098  *
2099  * MLDv1 Message                 MLDv2 Equivalent
2100  * --------------                -----------------
2101  * Report                        IS_EX( {} )
2102  * Done                          TO_IN( {} )
2103  */
2104 static void amt_mldv1_leave_handler(struct amt_dev *amt, struct sk_buff *skb,
2105 				    struct amt_tunnel_list *tunnel)
2106 {
2107 	struct mld_msg *mld = (struct mld_msg *)icmp6_hdr(skb);
2108 	struct iphdr *iph = ip_hdr(skb);
2109 	struct amt_group_node *gnode;
2110 	union amt_addr group, host;
2111 
2112 	memcpy(&group.ip6, &mld->mld_mca, sizeof(struct in6_addr));
2113 	memset(&host, 0, sizeof(union amt_addr));
2114 	host.ip4 = iph->saddr;
2115 
2116 	gnode = amt_lookup_group(tunnel, &group, &host, true);
2117 	if (gnode) {
2118 		amt_del_group(amt, gnode);
2119 		return;
2120 	}
2121 }
2122 
2123 static void amt_mldv2_report_handler(struct amt_dev *amt, struct sk_buff *skb,
2124 				     struct amt_tunnel_list *tunnel)
2125 {
2126 	struct mld2_report *mld2r = (struct mld2_report *)icmp6_hdr(skb);
2127 	int len = skb_transport_offset(skb) + sizeof(*mld2r);
2128 	void *zero_grec = (void *)&mldv2_zero_grec;
2129 	struct ipv6hdr *ip6h = ipv6_hdr(skb);
2130 	struct amt_group_node *gnode;
2131 	union amt_addr group, host;
2132 	struct mld2_grec *grec;
2133 	u16 nsrcs;
2134 	int i;
2135 
2136 	for (i = 0; i < ntohs(mld2r->mld2r_ngrec); i++) {
2137 		len += sizeof(*grec);
2138 		if (!ipv6_mc_may_pull(skb, len))
2139 			break;
2140 
2141 		grec = (void *)(skb->data + len - sizeof(*grec));
2142 		nsrcs = ntohs(grec->grec_nsrcs);
2143 
2144 		len += nsrcs * sizeof(struct in6_addr);
2145 		if (!ipv6_mc_may_pull(skb, len))
2146 			break;
2147 
2148 		memset(&group, 0, sizeof(union amt_addr));
2149 		group.ip6 = grec->grec_mca;
2150 		memset(&host, 0, sizeof(union amt_addr));
2151 		host.ip6 = ip6h->saddr;
2152 		gnode = amt_lookup_group(tunnel, &group, &host, true);
2153 		if (!gnode) {
2154 			gnode = amt_add_group(amt, tunnel, &group, &host,
2155 					      ETH_P_IPV6);
2156 			if (IS_ERR(gnode))
2157 				continue;
2158 		}
2159 
2160 		amt_add_srcs(amt, tunnel, gnode, grec, true);
2161 		switch (grec->grec_type) {
2162 		case MLD2_MODE_IS_INCLUDE:
2163 			amt_mcast_is_in_handler(amt, tunnel, gnode, grec,
2164 						zero_grec, true);
2165 			break;
2166 		case MLD2_MODE_IS_EXCLUDE:
2167 			amt_mcast_is_ex_handler(amt, tunnel, gnode, grec,
2168 						zero_grec, true);
2169 			break;
2170 		case MLD2_CHANGE_TO_INCLUDE:
2171 			amt_mcast_to_in_handler(amt, tunnel, gnode, grec,
2172 						zero_grec, true);
2173 			break;
2174 		case MLD2_CHANGE_TO_EXCLUDE:
2175 			amt_mcast_to_ex_handler(amt, tunnel, gnode, grec,
2176 						zero_grec, true);
2177 			break;
2178 		case MLD2_ALLOW_NEW_SOURCES:
2179 			amt_mcast_allow_handler(amt, tunnel, gnode, grec,
2180 						zero_grec, true);
2181 			break;
2182 		case MLD2_BLOCK_OLD_SOURCES:
2183 			amt_mcast_block_handler(amt, tunnel, gnode, grec,
2184 						zero_grec, true);
2185 			break;
2186 		default:
2187 			break;
2188 		}
2189 		amt_cleanup_srcs(amt, tunnel, gnode);
2190 	}
2191 }
2192 
2193 /* caller held tunnel->lock */
2194 static void amt_mld_report_handler(struct amt_dev *amt, struct sk_buff *skb,
2195 				   struct amt_tunnel_list *tunnel)
2196 {
2197 	struct mld_msg *mld = (struct mld_msg *)icmp6_hdr(skb);
2198 
2199 	switch (mld->mld_type) {
2200 	case ICMPV6_MGM_REPORT:
2201 		amt_mldv1_report_handler(amt, skb, tunnel);
2202 		break;
2203 	case ICMPV6_MLD2_REPORT:
2204 		amt_mldv2_report_handler(amt, skb, tunnel);
2205 		break;
2206 	case ICMPV6_MGM_REDUCTION:
2207 		amt_mldv1_leave_handler(amt, skb, tunnel);
2208 		break;
2209 	default:
2210 		break;
2211 	}
2212 }
2213 #endif
2214 
2215 static bool amt_advertisement_handler(struct amt_dev *amt, struct sk_buff *skb)
2216 {
2217 	struct amt_header_advertisement *amta;
2218 	int hdr_size;
2219 
2220 	hdr_size = sizeof(*amta) - sizeof(struct amt_header);
2221 
2222 	if (!pskb_may_pull(skb, hdr_size))
2223 		return true;
2224 
2225 	amta = (struct amt_header_advertisement *)(udp_hdr(skb) + 1);
2226 	if (!amta->ip4)
2227 		return true;
2228 
2229 	if (amta->reserved || amta->version)
2230 		return true;
2231 
2232 	if (ipv4_is_loopback(amta->ip4) || ipv4_is_multicast(amta->ip4) ||
2233 	    ipv4_is_zeronet(amta->ip4))
2234 		return true;
2235 
2236 	amt->remote_ip = amta->ip4;
2237 	netdev_dbg(amt->dev, "advertised remote ip = %pI4\n", &amt->remote_ip);
2238 	mod_delayed_work(amt_wq, &amt->req_wq, 0);
2239 
2240 	amt_update_gw_status(amt, AMT_STATUS_RECEIVED_ADVERTISEMENT, true);
2241 	return false;
2242 }
2243 
2244 static bool amt_multicast_data_handler(struct amt_dev *amt, struct sk_buff *skb)
2245 {
2246 	struct amt_header_mcast_data *amtmd;
2247 	int hdr_size, len, err;
2248 	struct ethhdr *eth;
2249 	struct iphdr *iph;
2250 
2251 	amtmd = (struct amt_header_mcast_data *)(udp_hdr(skb) + 1);
2252 	if (amtmd->reserved || amtmd->version)
2253 		return true;
2254 
2255 	hdr_size = sizeof(*amtmd) + sizeof(struct udphdr);
2256 	if (iptunnel_pull_header(skb, hdr_size, htons(ETH_P_IP), false))
2257 		return true;
2258 	skb_reset_network_header(skb);
2259 	skb_push(skb, sizeof(*eth));
2260 	skb_reset_mac_header(skb);
2261 	skb_pull(skb, sizeof(*eth));
2262 	eth = eth_hdr(skb);
2263 	iph = ip_hdr(skb);
2264 	if (iph->version == 4) {
2265 		if (!ipv4_is_multicast(iph->daddr))
2266 			return true;
2267 		skb->protocol = htons(ETH_P_IP);
2268 		eth->h_proto = htons(ETH_P_IP);
2269 		ip_eth_mc_map(iph->daddr, eth->h_dest);
2270 #if IS_ENABLED(CONFIG_IPV6)
2271 	} else if (iph->version == 6) {
2272 		struct ipv6hdr *ip6h;
2273 
2274 		ip6h = ipv6_hdr(skb);
2275 		if (!ipv6_addr_is_multicast(&ip6h->daddr))
2276 			return true;
2277 		skb->protocol = htons(ETH_P_IPV6);
2278 		eth->h_proto = htons(ETH_P_IPV6);
2279 		ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
2280 #endif
2281 	} else {
2282 		return true;
2283 	}
2284 
2285 	skb->pkt_type = PACKET_MULTICAST;
2286 	skb->ip_summed = CHECKSUM_NONE;
2287 	len = skb->len;
2288 	err = gro_cells_receive(&amt->gro_cells, skb);
2289 	if (likely(err == NET_RX_SUCCESS))
2290 		dev_sw_netstats_rx_add(amt->dev, len);
2291 	else
2292 		amt->dev->stats.rx_dropped++;
2293 
2294 	return false;
2295 }
2296 
2297 static bool amt_membership_query_handler(struct amt_dev *amt,
2298 					 struct sk_buff *skb)
2299 {
2300 	struct amt_header_membership_query *amtmq;
2301 	struct igmpv3_query *ihv3;
2302 	struct ethhdr *eth, *oeth;
2303 	struct iphdr *iph;
2304 	int hdr_size, len;
2305 
2306 	hdr_size = sizeof(*amtmq) - sizeof(struct amt_header);
2307 
2308 	if (!pskb_may_pull(skb, hdr_size))
2309 		return true;
2310 
2311 	amtmq = (struct amt_header_membership_query *)(udp_hdr(skb) + 1);
2312 	if (amtmq->reserved || amtmq->version)
2313 		return true;
2314 
2315 	hdr_size = sizeof(*amtmq) + sizeof(struct udphdr) - sizeof(*eth);
2316 	if (iptunnel_pull_header(skb, hdr_size, htons(ETH_P_TEB), false))
2317 		return true;
2318 	oeth = eth_hdr(skb);
2319 	skb_reset_mac_header(skb);
2320 	skb_pull(skb, sizeof(*eth));
2321 	skb_reset_network_header(skb);
2322 	eth = eth_hdr(skb);
2323 	iph = ip_hdr(skb);
2324 	if (iph->version == 4) {
2325 		if (!ipv4_is_multicast(iph->daddr))
2326 			return true;
2327 		if (!pskb_may_pull(skb, sizeof(*iph) + AMT_IPHDR_OPTS +
2328 				   sizeof(*ihv3)))
2329 			return true;
2330 
2331 		ihv3 = skb_pull(skb, sizeof(*iph) + AMT_IPHDR_OPTS);
2332 		skb_reset_transport_header(skb);
2333 		skb_push(skb, sizeof(*iph) + AMT_IPHDR_OPTS);
2334 		spin_lock_bh(&amt->lock);
2335 		amt->ready4 = true;
2336 		amt->mac = amtmq->response_mac;
2337 		amt->req_cnt = 0;
2338 		amt->qi = ihv3->qqic;
2339 		spin_unlock_bh(&amt->lock);
2340 		skb->protocol = htons(ETH_P_IP);
2341 		eth->h_proto = htons(ETH_P_IP);
2342 		ip_eth_mc_map(iph->daddr, eth->h_dest);
2343 #if IS_ENABLED(CONFIG_IPV6)
2344 	} else if (iph->version == 6) {
2345 		struct ipv6hdr *ip6h = ipv6_hdr(skb);
2346 		struct mld2_query *mld2q;
2347 
2348 		if (!ipv6_addr_is_multicast(&ip6h->daddr))
2349 			return true;
2350 		if (!pskb_may_pull(skb, sizeof(*ip6h) + AMT_IP6HDR_OPTS +
2351 				   sizeof(*mld2q)))
2352 			return true;
2353 
2354 		mld2q = skb_pull(skb, sizeof(*ip6h) + AMT_IP6HDR_OPTS);
2355 		skb_reset_transport_header(skb);
2356 		skb_push(skb, sizeof(*ip6h) + AMT_IP6HDR_OPTS);
2357 		spin_lock_bh(&amt->lock);
2358 		amt->ready6 = true;
2359 		amt->mac = amtmq->response_mac;
2360 		amt->req_cnt = 0;
2361 		amt->qi = mld2q->mld2q_qqic;
2362 		spin_unlock_bh(&amt->lock);
2363 		skb->protocol = htons(ETH_P_IPV6);
2364 		eth->h_proto = htons(ETH_P_IPV6);
2365 		ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
2366 #endif
2367 	} else {
2368 		return true;
2369 	}
2370 
2371 	ether_addr_copy(eth->h_source, oeth->h_source);
2372 	skb->pkt_type = PACKET_MULTICAST;
2373 	skb->ip_summed = CHECKSUM_NONE;
2374 	len = skb->len;
2375 	if (netif_rx(skb) == NET_RX_SUCCESS) {
2376 		amt_update_gw_status(amt, AMT_STATUS_RECEIVED_QUERY, true);
2377 		dev_sw_netstats_rx_add(amt->dev, len);
2378 	} else {
2379 		amt->dev->stats.rx_dropped++;
2380 	}
2381 
2382 	return false;
2383 }
2384 
2385 static bool amt_update_handler(struct amt_dev *amt, struct sk_buff *skb)
2386 {
2387 	struct amt_header_membership_update *amtmu;
2388 	struct amt_tunnel_list *tunnel;
2389 	struct udphdr *udph;
2390 	struct ethhdr *eth;
2391 	struct iphdr *iph;
2392 	int len;
2393 
2394 	iph = ip_hdr(skb);
2395 	udph = udp_hdr(skb);
2396 
2397 	if (__iptunnel_pull_header(skb, sizeof(*udph), skb->protocol,
2398 				   false, false))
2399 		return true;
2400 
2401 	amtmu = (struct amt_header_membership_update *)skb->data;
2402 	if (amtmu->reserved || amtmu->version)
2403 		return true;
2404 
2405 	skb_pull(skb, sizeof(*amtmu));
2406 	skb_reset_network_header(skb);
2407 
2408 	list_for_each_entry_rcu(tunnel, &amt->tunnel_list, list) {
2409 		if (tunnel->ip4 == iph->saddr) {
2410 			if ((amtmu->nonce == tunnel->nonce &&
2411 			     amtmu->response_mac == tunnel->mac)) {
2412 				mod_delayed_work(amt_wq, &tunnel->gc_wq,
2413 						 msecs_to_jiffies(amt_gmi(amt))
2414 								  * 3);
2415 				goto report;
2416 			} else {
2417 				netdev_dbg(amt->dev, "Invalid MAC\n");
2418 				return true;
2419 			}
2420 		}
2421 	}
2422 
2423 	return false;
2424 
2425 report:
2426 	iph = ip_hdr(skb);
2427 	if (iph->version == 4) {
2428 		if (ip_mc_check_igmp(skb)) {
2429 			netdev_dbg(amt->dev, "Invalid IGMP\n");
2430 			return true;
2431 		}
2432 
2433 		spin_lock_bh(&tunnel->lock);
2434 		amt_igmp_report_handler(amt, skb, tunnel);
2435 		spin_unlock_bh(&tunnel->lock);
2436 
2437 		skb_push(skb, sizeof(struct ethhdr));
2438 		skb_reset_mac_header(skb);
2439 		eth = eth_hdr(skb);
2440 		skb->protocol = htons(ETH_P_IP);
2441 		eth->h_proto = htons(ETH_P_IP);
2442 		ip_eth_mc_map(iph->daddr, eth->h_dest);
2443 #if IS_ENABLED(CONFIG_IPV6)
2444 	} else if (iph->version == 6) {
2445 		struct ipv6hdr *ip6h = ipv6_hdr(skb);
2446 
2447 		if (ipv6_mc_check_mld(skb)) {
2448 			netdev_dbg(amt->dev, "Invalid MLD\n");
2449 			return true;
2450 		}
2451 
2452 		spin_lock_bh(&tunnel->lock);
2453 		amt_mld_report_handler(amt, skb, tunnel);
2454 		spin_unlock_bh(&tunnel->lock);
2455 
2456 		skb_push(skb, sizeof(struct ethhdr));
2457 		skb_reset_mac_header(skb);
2458 		eth = eth_hdr(skb);
2459 		skb->protocol = htons(ETH_P_IPV6);
2460 		eth->h_proto = htons(ETH_P_IPV6);
2461 		ipv6_eth_mc_map(&ip6h->daddr, eth->h_dest);
2462 #endif
2463 	} else {
2464 		netdev_dbg(amt->dev, "Unsupported Protocol\n");
2465 		return true;
2466 	}
2467 
2468 	skb_pull(skb, sizeof(struct ethhdr));
2469 	skb->pkt_type = PACKET_MULTICAST;
2470 	skb->ip_summed = CHECKSUM_NONE;
2471 	len = skb->len;
2472 	if (netif_rx(skb) == NET_RX_SUCCESS) {
2473 		amt_update_relay_status(tunnel, AMT_STATUS_RECEIVED_UPDATE,
2474 					true);
2475 		dev_sw_netstats_rx_add(amt->dev, len);
2476 	} else {
2477 		amt->dev->stats.rx_dropped++;
2478 	}
2479 
2480 	return false;
2481 }
2482 
2483 static void amt_send_advertisement(struct amt_dev *amt, __be32 nonce,
2484 				   __be32 daddr, __be16 dport)
2485 {
2486 	struct amt_header_advertisement *amta;
2487 	int hlen, tlen, offset;
2488 	struct socket *sock;
2489 	struct udphdr *udph;
2490 	struct sk_buff *skb;
2491 	struct iphdr *iph;
2492 	struct rtable *rt;
2493 	struct flowi4 fl4;
2494 	u32 len;
2495 	int err;
2496 
2497 	rcu_read_lock();
2498 	sock = rcu_dereference(amt->sock);
2499 	if (!sock)
2500 		goto out;
2501 
2502 	if (!netif_running(amt->stream_dev) || !netif_running(amt->dev))
2503 		goto out;
2504 
2505 	rt = ip_route_output_ports(amt->net, &fl4, sock->sk,
2506 				   daddr, amt->local_ip,
2507 				   dport, amt->relay_port,
2508 				   IPPROTO_UDP, 0,
2509 				   amt->stream_dev->ifindex);
2510 	if (IS_ERR(rt)) {
2511 		amt->dev->stats.tx_errors++;
2512 		goto out;
2513 	}
2514 
2515 	hlen = LL_RESERVED_SPACE(amt->dev);
2516 	tlen = amt->dev->needed_tailroom;
2517 	len = hlen + tlen + sizeof(*iph) + sizeof(*udph) + sizeof(*amta);
2518 	skb = netdev_alloc_skb_ip_align(amt->dev, len);
2519 	if (!skb) {
2520 		ip_rt_put(rt);
2521 		amt->dev->stats.tx_errors++;
2522 		goto out;
2523 	}
2524 
2525 	skb->priority = TC_PRIO_CONTROL;
2526 	skb_dst_set(skb, &rt->dst);
2527 
2528 	len = sizeof(*iph) + sizeof(*udph) + sizeof(*amta);
2529 	skb_reset_network_header(skb);
2530 	skb_put(skb, len);
2531 	amta = skb_pull(skb, sizeof(*iph) + sizeof(*udph));
2532 	amta->version	= 0;
2533 	amta->type	= AMT_MSG_ADVERTISEMENT;
2534 	amta->reserved	= 0;
2535 	amta->nonce	= nonce;
2536 	amta->ip4	= amt->local_ip;
2537 	skb_push(skb, sizeof(*udph));
2538 	skb_reset_transport_header(skb);
2539 	udph		= udp_hdr(skb);
2540 	udph->source	= amt->relay_port;
2541 	udph->dest	= dport;
2542 	udph->len	= htons(sizeof(*amta) + sizeof(*udph));
2543 	udph->check	= 0;
2544 	offset = skb_transport_offset(skb);
2545 	skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
2546 	udph->check = csum_tcpudp_magic(amt->local_ip, daddr,
2547 					sizeof(*udph) + sizeof(*amta),
2548 					IPPROTO_UDP, skb->csum);
2549 
2550 	skb_push(skb, sizeof(*iph));
2551 	iph		= ip_hdr(skb);
2552 	iph->version	= 4;
2553 	iph->ihl	= (sizeof(struct iphdr)) >> 2;
2554 	iph->tos	= AMT_TOS;
2555 	iph->frag_off	= 0;
2556 	iph->ttl	= ip4_dst_hoplimit(&rt->dst);
2557 	iph->daddr	= daddr;
2558 	iph->saddr	= amt->local_ip;
2559 	iph->protocol	= IPPROTO_UDP;
2560 	iph->tot_len	= htons(len);
2561 
2562 	skb->ip_summed = CHECKSUM_NONE;
2563 	ip_select_ident(amt->net, skb, NULL);
2564 	ip_send_check(iph);
2565 	err = ip_local_out(amt->net, sock->sk, skb);
2566 	if (unlikely(net_xmit_eval(err)))
2567 		amt->dev->stats.tx_errors++;
2568 
2569 out:
2570 	rcu_read_unlock();
2571 }
2572 
2573 static bool amt_discovery_handler(struct amt_dev *amt, struct sk_buff *skb)
2574 {
2575 	struct amt_header_discovery *amtd;
2576 	struct udphdr *udph;
2577 	struct iphdr *iph;
2578 
2579 	if (!pskb_may_pull(skb, sizeof(*udph) + sizeof(*amtd)))
2580 		return true;
2581 
2582 	iph = ip_hdr(skb);
2583 	udph = udp_hdr(skb);
2584 	amtd = (struct amt_header_discovery *)(udp_hdr(skb) + 1);
2585 
2586 	if (amtd->reserved || amtd->version)
2587 		return true;
2588 
2589 	amt_send_advertisement(amt, amtd->nonce, iph->saddr, udph->source);
2590 
2591 	return false;
2592 }
2593 
2594 static bool amt_request_handler(struct amt_dev *amt, struct sk_buff *skb)
2595 {
2596 	struct amt_header_request *amtrh;
2597 	struct amt_tunnel_list *tunnel;
2598 	unsigned long long key;
2599 	struct udphdr *udph;
2600 	struct iphdr *iph;
2601 	u64 mac;
2602 	int i;
2603 
2604 	if (!pskb_may_pull(skb, sizeof(*udph) + sizeof(*amtrh)))
2605 		return true;
2606 
2607 	iph = ip_hdr(skb);
2608 	udph = udp_hdr(skb);
2609 	amtrh = (struct amt_header_request *)(udp_hdr(skb) + 1);
2610 
2611 	if (amtrh->reserved1 || amtrh->reserved2 || amtrh->version)
2612 		return true;
2613 
2614 	list_for_each_entry_rcu(tunnel, &amt->tunnel_list, list)
2615 		if (tunnel->ip4 == iph->saddr)
2616 			goto send;
2617 
2618 	if (amt->nr_tunnels >= amt->max_tunnels) {
2619 		icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0);
2620 		return true;
2621 	}
2622 
2623 	tunnel = kzalloc(sizeof(*tunnel) +
2624 			 (sizeof(struct hlist_head) * amt->hash_buckets),
2625 			 GFP_ATOMIC);
2626 	if (!tunnel)
2627 		return true;
2628 
2629 	tunnel->source_port = udph->source;
2630 	tunnel->ip4 = iph->saddr;
2631 
2632 	memcpy(&key, &tunnel->key, sizeof(unsigned long long));
2633 	tunnel->amt = amt;
2634 	spin_lock_init(&tunnel->lock);
2635 	for (i = 0; i < amt->hash_buckets; i++)
2636 		INIT_HLIST_HEAD(&tunnel->groups[i]);
2637 
2638 	INIT_DELAYED_WORK(&tunnel->gc_wq, amt_tunnel_expire);
2639 
2640 	spin_lock_bh(&amt->lock);
2641 	list_add_tail_rcu(&tunnel->list, &amt->tunnel_list);
2642 	tunnel->key = amt->key;
2643 	amt_update_relay_status(tunnel, AMT_STATUS_RECEIVED_REQUEST, true);
2644 	amt->nr_tunnels++;
2645 	mod_delayed_work(amt_wq, &tunnel->gc_wq,
2646 			 msecs_to_jiffies(amt_gmi(amt)));
2647 	spin_unlock_bh(&amt->lock);
2648 
2649 send:
2650 	tunnel->nonce = amtrh->nonce;
2651 	mac = siphash_3u32((__force u32)tunnel->ip4,
2652 			   (__force u32)tunnel->source_port,
2653 			   (__force u32)tunnel->nonce,
2654 			   &tunnel->key);
2655 	tunnel->mac = mac >> 16;
2656 
2657 	if (!netif_running(amt->dev) || !netif_running(amt->stream_dev))
2658 		return true;
2659 
2660 	if (!amtrh->p)
2661 		amt_send_igmp_gq(amt, tunnel);
2662 	else
2663 		amt_send_mld_gq(amt, tunnel);
2664 
2665 	return false;
2666 }
2667 
2668 static int amt_rcv(struct sock *sk, struct sk_buff *skb)
2669 {
2670 	struct amt_dev *amt;
2671 	struct iphdr *iph;
2672 	int type;
2673 	bool err;
2674 
2675 	rcu_read_lock_bh();
2676 	amt = rcu_dereference_sk_user_data(sk);
2677 	if (!amt) {
2678 		err = true;
2679 		goto out;
2680 	}
2681 
2682 	skb->dev = amt->dev;
2683 	iph = ip_hdr(skb);
2684 	type = amt_parse_type(skb);
2685 	if (type == -1) {
2686 		err = true;
2687 		goto drop;
2688 	}
2689 
2690 	if (amt->mode == AMT_MODE_GATEWAY) {
2691 		switch (type) {
2692 		case AMT_MSG_ADVERTISEMENT:
2693 			if (iph->saddr != amt->discovery_ip) {
2694 				netdev_dbg(amt->dev, "Invalid Relay IP\n");
2695 				err = true;
2696 				goto drop;
2697 			}
2698 			if (amt_advertisement_handler(amt, skb))
2699 				amt->dev->stats.rx_dropped++;
2700 			goto out;
2701 		case AMT_MSG_MULTICAST_DATA:
2702 			if (iph->saddr != amt->remote_ip) {
2703 				netdev_dbg(amt->dev, "Invalid Relay IP\n");
2704 				err = true;
2705 				goto drop;
2706 			}
2707 			err = amt_multicast_data_handler(amt, skb);
2708 			if (err)
2709 				goto drop;
2710 			else
2711 				goto out;
2712 		case AMT_MSG_MEMBERSHIP_QUERY:
2713 			if (iph->saddr != amt->remote_ip) {
2714 				netdev_dbg(amt->dev, "Invalid Relay IP\n");
2715 				err = true;
2716 				goto drop;
2717 			}
2718 			err = amt_membership_query_handler(amt, skb);
2719 			if (err)
2720 				goto drop;
2721 			else
2722 				goto out;
2723 		default:
2724 			err = true;
2725 			netdev_dbg(amt->dev, "Invalid type of Gateway\n");
2726 			break;
2727 		}
2728 	} else {
2729 		switch (type) {
2730 		case AMT_MSG_DISCOVERY:
2731 			err = amt_discovery_handler(amt, skb);
2732 			break;
2733 		case AMT_MSG_REQUEST:
2734 			err = amt_request_handler(amt, skb);
2735 			break;
2736 		case AMT_MSG_MEMBERSHIP_UPDATE:
2737 			err = amt_update_handler(amt, skb);
2738 			if (err)
2739 				goto drop;
2740 			else
2741 				goto out;
2742 		default:
2743 			err = true;
2744 			netdev_dbg(amt->dev, "Invalid type of relay\n");
2745 			break;
2746 		}
2747 	}
2748 drop:
2749 	if (err) {
2750 		amt->dev->stats.rx_dropped++;
2751 		kfree_skb(skb);
2752 	} else {
2753 		consume_skb(skb);
2754 	}
2755 out:
2756 	rcu_read_unlock_bh();
2757 	return 0;
2758 }
2759 
2760 static int amt_err_lookup(struct sock *sk, struct sk_buff *skb)
2761 {
2762 	struct amt_dev *amt;
2763 	int type;
2764 
2765 	rcu_read_lock_bh();
2766 	amt = rcu_dereference_sk_user_data(sk);
2767 	if (!amt)
2768 		goto out;
2769 
2770 	if (amt->mode != AMT_MODE_GATEWAY)
2771 		goto drop;
2772 
2773 	type = amt_parse_type(skb);
2774 	if (type == -1)
2775 		goto drop;
2776 
2777 	netdev_dbg(amt->dev, "Received IGMP Unreachable of %s\n",
2778 		   type_str[type]);
2779 	switch (type) {
2780 	case AMT_MSG_DISCOVERY:
2781 		break;
2782 	case AMT_MSG_REQUEST:
2783 	case AMT_MSG_MEMBERSHIP_UPDATE:
2784 		if (amt->status >= AMT_STATUS_RECEIVED_ADVERTISEMENT)
2785 			mod_delayed_work(amt_wq, &amt->req_wq, 0);
2786 		break;
2787 	default:
2788 		goto drop;
2789 	}
2790 out:
2791 	rcu_read_unlock_bh();
2792 	return 0;
2793 drop:
2794 	rcu_read_unlock_bh();
2795 	amt->dev->stats.rx_dropped++;
2796 	return 0;
2797 }
2798 
2799 static struct socket *amt_create_sock(struct net *net, __be16 port)
2800 {
2801 	struct udp_port_cfg udp_conf;
2802 	struct socket *sock;
2803 	int err;
2804 
2805 	memset(&udp_conf, 0, sizeof(udp_conf));
2806 	udp_conf.family = AF_INET;
2807 	udp_conf.local_ip.s_addr = htonl(INADDR_ANY);
2808 
2809 	udp_conf.local_udp_port = port;
2810 
2811 	err = udp_sock_create(net, &udp_conf, &sock);
2812 	if (err < 0)
2813 		return ERR_PTR(err);
2814 
2815 	return sock;
2816 }
2817 
2818 static int amt_socket_create(struct amt_dev *amt)
2819 {
2820 	struct udp_tunnel_sock_cfg tunnel_cfg;
2821 	struct socket *sock;
2822 
2823 	sock = amt_create_sock(amt->net, amt->relay_port);
2824 	if (IS_ERR(sock))
2825 		return PTR_ERR(sock);
2826 
2827 	/* Mark socket as an encapsulation socket */
2828 	memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
2829 	tunnel_cfg.sk_user_data = amt;
2830 	tunnel_cfg.encap_type = 1;
2831 	tunnel_cfg.encap_rcv = amt_rcv;
2832 	tunnel_cfg.encap_err_lookup = amt_err_lookup;
2833 	tunnel_cfg.encap_destroy = NULL;
2834 	setup_udp_tunnel_sock(amt->net, sock, &tunnel_cfg);
2835 
2836 	rcu_assign_pointer(amt->sock, sock);
2837 	return 0;
2838 }
2839 
2840 static int amt_dev_open(struct net_device *dev)
2841 {
2842 	struct amt_dev *amt = netdev_priv(dev);
2843 	int err;
2844 
2845 	amt->ready4 = false;
2846 	amt->ready6 = false;
2847 
2848 	err = amt_socket_create(amt);
2849 	if (err)
2850 		return err;
2851 
2852 	amt->req_cnt = 0;
2853 	amt->remote_ip = 0;
2854 	get_random_bytes(&amt->key, sizeof(siphash_key_t));
2855 
2856 	amt->status = AMT_STATUS_INIT;
2857 	if (amt->mode == AMT_MODE_GATEWAY) {
2858 		mod_delayed_work(amt_wq, &amt->discovery_wq, 0);
2859 		mod_delayed_work(amt_wq, &amt->req_wq, 0);
2860 	} else if (amt->mode == AMT_MODE_RELAY) {
2861 		mod_delayed_work(amt_wq, &amt->secret_wq,
2862 				 msecs_to_jiffies(AMT_SECRET_TIMEOUT));
2863 	}
2864 	return err;
2865 }
2866 
2867 static int amt_dev_stop(struct net_device *dev)
2868 {
2869 	struct amt_dev *amt = netdev_priv(dev);
2870 	struct amt_tunnel_list *tunnel, *tmp;
2871 	struct socket *sock;
2872 
2873 	cancel_delayed_work_sync(&amt->req_wq);
2874 	cancel_delayed_work_sync(&amt->discovery_wq);
2875 	cancel_delayed_work_sync(&amt->secret_wq);
2876 
2877 	/* shutdown */
2878 	sock = rtnl_dereference(amt->sock);
2879 	RCU_INIT_POINTER(amt->sock, NULL);
2880 	synchronize_net();
2881 	if (sock)
2882 		udp_tunnel_sock_release(sock);
2883 
2884 	amt->ready4 = false;
2885 	amt->ready6 = false;
2886 	amt->req_cnt = 0;
2887 	amt->remote_ip = 0;
2888 
2889 	list_for_each_entry_safe(tunnel, tmp, &amt->tunnel_list, list) {
2890 		list_del_rcu(&tunnel->list);
2891 		amt->nr_tunnels--;
2892 		cancel_delayed_work_sync(&tunnel->gc_wq);
2893 		amt_clear_groups(tunnel);
2894 		kfree_rcu(tunnel, rcu);
2895 	}
2896 
2897 	return 0;
2898 }
2899 
2900 static const struct device_type amt_type = {
2901 	.name = "amt",
2902 };
2903 
2904 static int amt_dev_init(struct net_device *dev)
2905 {
2906 	struct amt_dev *amt = netdev_priv(dev);
2907 	int err;
2908 
2909 	amt->dev = dev;
2910 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2911 	if (!dev->tstats)
2912 		return -ENOMEM;
2913 
2914 	err = gro_cells_init(&amt->gro_cells, dev);
2915 	if (err) {
2916 		free_percpu(dev->tstats);
2917 		return err;
2918 	}
2919 
2920 	return 0;
2921 }
2922 
2923 static void amt_dev_uninit(struct net_device *dev)
2924 {
2925 	struct amt_dev *amt = netdev_priv(dev);
2926 
2927 	gro_cells_destroy(&amt->gro_cells);
2928 	free_percpu(dev->tstats);
2929 }
2930 
2931 static const struct net_device_ops amt_netdev_ops = {
2932 	.ndo_init               = amt_dev_init,
2933 	.ndo_uninit             = amt_dev_uninit,
2934 	.ndo_open		= amt_dev_open,
2935 	.ndo_stop		= amt_dev_stop,
2936 	.ndo_start_xmit         = amt_dev_xmit,
2937 	.ndo_get_stats64        = dev_get_tstats64,
2938 };
2939 
2940 static void amt_link_setup(struct net_device *dev)
2941 {
2942 	dev->netdev_ops         = &amt_netdev_ops;
2943 	dev->needs_free_netdev  = true;
2944 	SET_NETDEV_DEVTYPE(dev, &amt_type);
2945 	dev->min_mtu		= ETH_MIN_MTU;
2946 	dev->max_mtu		= ETH_MAX_MTU;
2947 	dev->type		= ARPHRD_NONE;
2948 	dev->flags		= IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
2949 	dev->hard_header_len	= 0;
2950 	dev->addr_len		= 0;
2951 	dev->priv_flags		|= IFF_NO_QUEUE;
2952 	dev->features		|= NETIF_F_LLTX;
2953 	dev->features		|= NETIF_F_GSO_SOFTWARE;
2954 	dev->features		|= NETIF_F_NETNS_LOCAL;
2955 	dev->hw_features	|= NETIF_F_SG | NETIF_F_HW_CSUM;
2956 	dev->hw_features	|= NETIF_F_FRAGLIST | NETIF_F_RXCSUM;
2957 	dev->hw_features	|= NETIF_F_GSO_SOFTWARE;
2958 	eth_hw_addr_random(dev);
2959 	eth_zero_addr(dev->broadcast);
2960 	ether_setup(dev);
2961 }
2962 
2963 static const struct nla_policy amt_policy[IFLA_AMT_MAX + 1] = {
2964 	[IFLA_AMT_MODE]		= { .type = NLA_U32 },
2965 	[IFLA_AMT_RELAY_PORT]	= { .type = NLA_U16 },
2966 	[IFLA_AMT_GATEWAY_PORT]	= { .type = NLA_U16 },
2967 	[IFLA_AMT_LINK]		= { .type = NLA_U32 },
2968 	[IFLA_AMT_LOCAL_IP]	= { .len = sizeof_field(struct iphdr, daddr) },
2969 	[IFLA_AMT_REMOTE_IP]	= { .len = sizeof_field(struct iphdr, daddr) },
2970 	[IFLA_AMT_DISCOVERY_IP]	= { .len = sizeof_field(struct iphdr, daddr) },
2971 	[IFLA_AMT_MAX_TUNNELS]	= { .type = NLA_U32 },
2972 };
2973 
2974 static int amt_validate(struct nlattr *tb[], struct nlattr *data[],
2975 			struct netlink_ext_ack *extack)
2976 {
2977 	if (!data)
2978 		return -EINVAL;
2979 
2980 	if (!data[IFLA_AMT_LINK]) {
2981 		NL_SET_ERR_MSG_ATTR(extack, data[IFLA_AMT_LINK],
2982 				    "Link attribute is required");
2983 		return -EINVAL;
2984 	}
2985 
2986 	if (!data[IFLA_AMT_MODE]) {
2987 		NL_SET_ERR_MSG_ATTR(extack, data[IFLA_AMT_MODE],
2988 				    "Mode attribute is required");
2989 		return -EINVAL;
2990 	}
2991 
2992 	if (nla_get_u32(data[IFLA_AMT_MODE]) > AMT_MODE_MAX) {
2993 		NL_SET_ERR_MSG_ATTR(extack, data[IFLA_AMT_MODE],
2994 				    "Mode attribute is not valid");
2995 		return -EINVAL;
2996 	}
2997 
2998 	if (!data[IFLA_AMT_LOCAL_IP]) {
2999 		NL_SET_ERR_MSG_ATTR(extack, data[IFLA_AMT_DISCOVERY_IP],
3000 				    "Local attribute is required");
3001 		return -EINVAL;
3002 	}
3003 
3004 	if (!data[IFLA_AMT_DISCOVERY_IP] &&
3005 	    nla_get_u32(data[IFLA_AMT_MODE]) == AMT_MODE_GATEWAY) {
3006 		NL_SET_ERR_MSG_ATTR(extack, data[IFLA_AMT_LOCAL_IP],
3007 				    "Discovery attribute is required");
3008 		return -EINVAL;
3009 	}
3010 
3011 	return 0;
3012 }
3013 
3014 static int amt_newlink(struct net *net, struct net_device *dev,
3015 		       struct nlattr *tb[], struct nlattr *data[],
3016 		       struct netlink_ext_ack *extack)
3017 {
3018 	struct amt_dev *amt = netdev_priv(dev);
3019 	int err = -EINVAL;
3020 
3021 	amt->net = net;
3022 	amt->mode = nla_get_u32(data[IFLA_AMT_MODE]);
3023 
3024 	if (data[IFLA_AMT_MAX_TUNNELS] &&
3025 	    nla_get_u32(data[IFLA_AMT_MAX_TUNNELS]))
3026 		amt->max_tunnels = nla_get_u32(data[IFLA_AMT_MAX_TUNNELS]);
3027 	else
3028 		amt->max_tunnels = AMT_MAX_TUNNELS;
3029 
3030 	spin_lock_init(&amt->lock);
3031 	amt->max_groups = AMT_MAX_GROUP;
3032 	amt->max_sources = AMT_MAX_SOURCE;
3033 	amt->hash_buckets = AMT_HSIZE;
3034 	amt->nr_tunnels = 0;
3035 	get_random_bytes(&amt->hash_seed, sizeof(amt->hash_seed));
3036 	amt->stream_dev = dev_get_by_index(net,
3037 					   nla_get_u32(data[IFLA_AMT_LINK]));
3038 	if (!amt->stream_dev) {
3039 		NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_AMT_LINK],
3040 				    "Can't find stream device");
3041 		return -ENODEV;
3042 	}
3043 
3044 	if (amt->stream_dev->type != ARPHRD_ETHER) {
3045 		NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_AMT_LINK],
3046 				    "Invalid stream device type");
3047 		goto err;
3048 	}
3049 
3050 	amt->local_ip = nla_get_in_addr(data[IFLA_AMT_LOCAL_IP]);
3051 	if (ipv4_is_loopback(amt->local_ip) ||
3052 	    ipv4_is_zeronet(amt->local_ip) ||
3053 	    ipv4_is_multicast(amt->local_ip)) {
3054 		NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_AMT_LOCAL_IP],
3055 				    "Invalid Local address");
3056 		goto err;
3057 	}
3058 
3059 	if (data[IFLA_AMT_RELAY_PORT])
3060 		amt->relay_port = nla_get_be16(data[IFLA_AMT_RELAY_PORT]);
3061 	else
3062 		amt->relay_port = htons(IANA_AMT_UDP_PORT);
3063 
3064 	if (data[IFLA_AMT_GATEWAY_PORT])
3065 		amt->gw_port = nla_get_be16(data[IFLA_AMT_GATEWAY_PORT]);
3066 	else
3067 		amt->gw_port = htons(IANA_AMT_UDP_PORT);
3068 
3069 	if (!amt->relay_port) {
3070 		NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_AMT_DISCOVERY_IP],
3071 				    "relay port must not be 0");
3072 		goto err;
3073 	}
3074 	if (amt->mode == AMT_MODE_RELAY) {
3075 		amt->qrv = amt->net->ipv4.sysctl_igmp_qrv;
3076 		amt->qri = 10;
3077 		dev->needed_headroom = amt->stream_dev->needed_headroom +
3078 				       AMT_RELAY_HLEN;
3079 		dev->mtu = amt->stream_dev->mtu - AMT_RELAY_HLEN;
3080 		dev->max_mtu = dev->mtu;
3081 		dev->min_mtu = ETH_MIN_MTU + AMT_RELAY_HLEN;
3082 	} else {
3083 		if (!data[IFLA_AMT_DISCOVERY_IP]) {
3084 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_AMT_DISCOVERY_IP],
3085 					    "discovery must be set in gateway mode");
3086 			goto err;
3087 		}
3088 		if (!amt->gw_port) {
3089 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_AMT_DISCOVERY_IP],
3090 					    "gateway port must not be 0");
3091 			goto err;
3092 		}
3093 		amt->remote_ip = 0;
3094 		amt->discovery_ip = nla_get_in_addr(data[IFLA_AMT_DISCOVERY_IP]);
3095 		if (ipv4_is_loopback(amt->discovery_ip) ||
3096 		    ipv4_is_zeronet(amt->discovery_ip) ||
3097 		    ipv4_is_multicast(amt->discovery_ip)) {
3098 			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_AMT_DISCOVERY_IP],
3099 					    "discovery must be unicast");
3100 			goto err;
3101 		}
3102 
3103 		dev->needed_headroom = amt->stream_dev->needed_headroom +
3104 				       AMT_GW_HLEN;
3105 		dev->mtu = amt->stream_dev->mtu - AMT_GW_HLEN;
3106 		dev->max_mtu = dev->mtu;
3107 		dev->min_mtu = ETH_MIN_MTU + AMT_GW_HLEN;
3108 	}
3109 	amt->qi = AMT_INIT_QUERY_INTERVAL;
3110 
3111 	err = register_netdevice(dev);
3112 	if (err < 0) {
3113 		netdev_dbg(dev, "failed to register new netdev %d\n", err);
3114 		goto err;
3115 	}
3116 
3117 	err = netdev_upper_dev_link(amt->stream_dev, dev, extack);
3118 	if (err < 0) {
3119 		unregister_netdevice(dev);
3120 		goto err;
3121 	}
3122 
3123 	INIT_DELAYED_WORK(&amt->discovery_wq, amt_discovery_work);
3124 	INIT_DELAYED_WORK(&amt->req_wq, amt_req_work);
3125 	INIT_DELAYED_WORK(&amt->secret_wq, amt_secret_work);
3126 	INIT_LIST_HEAD(&amt->tunnel_list);
3127 
3128 	return 0;
3129 err:
3130 	dev_put(amt->stream_dev);
3131 	return err;
3132 }
3133 
3134 static void amt_dellink(struct net_device *dev, struct list_head *head)
3135 {
3136 	struct amt_dev *amt = netdev_priv(dev);
3137 
3138 	unregister_netdevice_queue(dev, head);
3139 	netdev_upper_dev_unlink(amt->stream_dev, dev);
3140 	dev_put(amt->stream_dev);
3141 }
3142 
3143 static size_t amt_get_size(const struct net_device *dev)
3144 {
3145 	return nla_total_size(sizeof(__u32)) + /* IFLA_AMT_MODE */
3146 	       nla_total_size(sizeof(__u16)) + /* IFLA_AMT_RELAY_PORT */
3147 	       nla_total_size(sizeof(__u16)) + /* IFLA_AMT_GATEWAY_PORT */
3148 	       nla_total_size(sizeof(__u32)) + /* IFLA_AMT_LINK */
3149 	       nla_total_size(sizeof(__u32)) + /* IFLA_MAX_TUNNELS */
3150 	       nla_total_size(sizeof(struct iphdr)) + /* IFLA_AMT_DISCOVERY_IP */
3151 	       nla_total_size(sizeof(struct iphdr)) + /* IFLA_AMT_REMOTE_IP */
3152 	       nla_total_size(sizeof(struct iphdr)); /* IFLA_AMT_LOCAL_IP */
3153 }
3154 
3155 static int amt_fill_info(struct sk_buff *skb, const struct net_device *dev)
3156 {
3157 	struct amt_dev *amt = netdev_priv(dev);
3158 
3159 	if (nla_put_u32(skb, IFLA_AMT_MODE, amt->mode))
3160 		goto nla_put_failure;
3161 	if (nla_put_be16(skb, IFLA_AMT_RELAY_PORT, amt->relay_port))
3162 		goto nla_put_failure;
3163 	if (nla_put_be16(skb, IFLA_AMT_GATEWAY_PORT, amt->gw_port))
3164 		goto nla_put_failure;
3165 	if (nla_put_u32(skb, IFLA_AMT_LINK, amt->stream_dev->ifindex))
3166 		goto nla_put_failure;
3167 	if (nla_put_in_addr(skb, IFLA_AMT_LOCAL_IP, amt->local_ip))
3168 		goto nla_put_failure;
3169 	if (nla_put_in_addr(skb, IFLA_AMT_DISCOVERY_IP, amt->discovery_ip))
3170 		goto nla_put_failure;
3171 	if (amt->remote_ip)
3172 		if (nla_put_in_addr(skb, IFLA_AMT_REMOTE_IP, amt->remote_ip))
3173 			goto nla_put_failure;
3174 	if (nla_put_u32(skb, IFLA_AMT_MAX_TUNNELS, amt->max_tunnels))
3175 		goto nla_put_failure;
3176 
3177 	return 0;
3178 
3179 nla_put_failure:
3180 	return -EMSGSIZE;
3181 }
3182 
3183 static struct rtnl_link_ops amt_link_ops __read_mostly = {
3184 	.kind		= "amt",
3185 	.maxtype	= IFLA_AMT_MAX,
3186 	.policy		= amt_policy,
3187 	.priv_size	= sizeof(struct amt_dev),
3188 	.setup		= amt_link_setup,
3189 	.validate	= amt_validate,
3190 	.newlink	= amt_newlink,
3191 	.dellink	= amt_dellink,
3192 	.get_size       = amt_get_size,
3193 	.fill_info      = amt_fill_info,
3194 };
3195 
3196 static struct net_device *amt_lookup_upper_dev(struct net_device *dev)
3197 {
3198 	struct net_device *upper_dev;
3199 	struct amt_dev *amt;
3200 
3201 	for_each_netdev(dev_net(dev), upper_dev) {
3202 		if (netif_is_amt(upper_dev)) {
3203 			amt = netdev_priv(upper_dev);
3204 			if (amt->stream_dev == dev)
3205 				return upper_dev;
3206 		}
3207 	}
3208 
3209 	return NULL;
3210 }
3211 
3212 static int amt_device_event(struct notifier_block *unused,
3213 			    unsigned long event, void *ptr)
3214 {
3215 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3216 	struct net_device *upper_dev;
3217 	struct amt_dev *amt;
3218 	LIST_HEAD(list);
3219 	int new_mtu;
3220 
3221 	upper_dev = amt_lookup_upper_dev(dev);
3222 	if (!upper_dev)
3223 		return NOTIFY_DONE;
3224 	amt = netdev_priv(upper_dev);
3225 
3226 	switch (event) {
3227 	case NETDEV_UNREGISTER:
3228 		amt_dellink(amt->dev, &list);
3229 		unregister_netdevice_many(&list);
3230 		break;
3231 	case NETDEV_CHANGEMTU:
3232 		if (amt->mode == AMT_MODE_RELAY)
3233 			new_mtu = dev->mtu - AMT_RELAY_HLEN;
3234 		else
3235 			new_mtu = dev->mtu - AMT_GW_HLEN;
3236 
3237 		dev_set_mtu(amt->dev, new_mtu);
3238 		break;
3239 	}
3240 
3241 	return NOTIFY_DONE;
3242 }
3243 
3244 static struct notifier_block amt_notifier_block __read_mostly = {
3245 	.notifier_call = amt_device_event,
3246 };
3247 
3248 static int __init amt_init(void)
3249 {
3250 	int err;
3251 
3252 	err = register_netdevice_notifier(&amt_notifier_block);
3253 	if (err < 0)
3254 		goto err;
3255 
3256 	err = rtnl_link_register(&amt_link_ops);
3257 	if (err < 0)
3258 		goto unregister_notifier;
3259 
3260 	amt_wq = alloc_workqueue("amt", WQ_UNBOUND, 1);
3261 	if (!amt_wq) {
3262 		err = -ENOMEM;
3263 		goto rtnl_unregister;
3264 	}
3265 
3266 	spin_lock_init(&source_gc_lock);
3267 	spin_lock_bh(&source_gc_lock);
3268 	INIT_DELAYED_WORK(&source_gc_wq, amt_source_gc_work);
3269 	mod_delayed_work(amt_wq, &source_gc_wq,
3270 			 msecs_to_jiffies(AMT_GC_INTERVAL));
3271 	spin_unlock_bh(&source_gc_lock);
3272 
3273 	return 0;
3274 
3275 rtnl_unregister:
3276 	rtnl_link_unregister(&amt_link_ops);
3277 unregister_notifier:
3278 	unregister_netdevice_notifier(&amt_notifier_block);
3279 err:
3280 	pr_err("error loading AMT module loaded\n");
3281 	return err;
3282 }
3283 late_initcall(amt_init);
3284 
3285 static void __exit amt_fini(void)
3286 {
3287 	rtnl_link_unregister(&amt_link_ops);
3288 	unregister_netdevice_notifier(&amt_notifier_block);
3289 	cancel_delayed_work_sync(&source_gc_wq);
3290 	__amt_source_gc_work();
3291 	destroy_workqueue(amt_wq);
3292 }
3293 module_exit(amt_fini);
3294 
3295 MODULE_LICENSE("GPL");
3296 MODULE_AUTHOR("Taehee Yoo <ap420073@gmail.com>");
3297 MODULE_ALIAS_RTNL_LINK("amt");
3298