bonding: IGMP handling cleanup
[linux-2.6.git] / net / ipv4 / igmp.c
1 /*
2  *      Linux NET3:     Internet Group Management Protocol  [IGMP]
3  *
4  *      This code implements the IGMP protocol as defined in RFC1112. There has
5  *      been a further revision of this protocol since which is now supported.
6  *
7  *      If you have trouble with this module be careful what gcc you have used,
8  *      the older version didn't come out right using gcc 2.5.8, the newer one
9  *      seems to fall out with gcc 2.6.2.
10  *
11  *      Authors:
12  *              Alan Cox <alan@lxorguk.ukuu.org.uk>
13  *
14  *      This program is free software; you can redistribute it and/or
15  *      modify it under the terms of the GNU General Public License
16  *      as published by the Free Software Foundation; either version
17  *      2 of the License, or (at your option) any later version.
18  *
19  *      Fixes:
20  *
21  *              Alan Cox        :       Added lots of __inline__ to optimise
22  *                                      the memory usage of all the tiny little
23  *                                      functions.
24  *              Alan Cox        :       Dumped the header building experiment.
25  *              Alan Cox        :       Minor tweaks ready for multicast routing
26  *                                      and extended IGMP protocol.
27  *              Alan Cox        :       Removed a load of inline directives. Gcc 2.5.8
28  *                                      writes utterly bogus code otherwise (sigh)
29  *                                      fixed IGMP loopback to behave in the manner
30  *                                      desired by mrouted, fixed the fact it has been
31  *                                      broken since 1.3.6 and cleaned up a few minor
32  *                                      points.
33  *
34  *              Chih-Jen Chang  :       Tried to revise IGMP to Version 2
35  *              Tsu-Sheng Tsao          E-mail: chihjenc@scf.usc.edu and tsusheng@scf.usc.edu
36  *                                      The enhancements are mainly based on Steve Deering's
37  *                                      ipmulti-3.5 source code.
38  *              Chih-Jen Chang  :       Added the igmp_get_mrouter_info and
39  *              Tsu-Sheng Tsao          igmp_set_mrouter_info to keep track of
40  *                                      the mrouted version on that device.
41  *              Chih-Jen Chang  :       Added the max_resp_time parameter to
42  *              Tsu-Sheng Tsao          igmp_heard_query(). Using this parameter
43  *                                      to identify the multicast router version
44  *                                      and do what the IGMP version 2 specified.
45  *              Chih-Jen Chang  :       Added a timer to revert to IGMP V2 router
46  *              Tsu-Sheng Tsao          if the specified time expired.
47  *              Alan Cox        :       Stop IGMP from 0.0.0.0 being accepted.
48  *              Alan Cox        :       Use GFP_ATOMIC in the right places.
49  *              Christian Daudt :       igmp timer wasn't set for local group
50  *                                      memberships but was being deleted,
51  *                                      which caused a "del_timer() called
52  *                                      from %p with timer not initialized\n"
53  *                                      message (960131).
54  *              Christian Daudt :       removed del_timer from
55  *                                      igmp_timer_expire function (960205).
56  *             Christian Daudt :       igmp_heard_report now only calls
57  *                                     igmp_timer_expire if tm->running is
58  *                                     true (960216).
59  *              Malcolm Beattie :       ttl comparison wrong in igmp_rcv made
60  *                                      igmp_heard_query never trigger. Expiry
61  *                                      miscalculation fixed in igmp_heard_query
62  *                                      and random() made to return unsigned to
63  *                                      prevent negative expiry times.
64  *              Alexey Kuznetsov:       Wrong group leaving behaviour, backport
65  *                                      fix from pending 2.1.x patches.
66  *              Alan Cox:               Forget to enable FDDI support earlier.
67  *              Alexey Kuznetsov:       Fixed leaving groups on device down.
68  *              Alexey Kuznetsov:       Accordance to igmp-v2-06 draft.
69  *              David L Stevens:        IGMPv3 support, with help from
70  *                                      Vinay Kulkarni
71  */
72
73 #include <linux/module.h>
74 #include <linux/slab.h>
75 #include <asm/uaccess.h>
76 #include <asm/system.h>
77 #include <linux/types.h>
78 #include <linux/kernel.h>
79 #include <linux/jiffies.h>
80 #include <linux/string.h>
81 #include <linux/socket.h>
82 #include <linux/sockios.h>
83 #include <linux/in.h>
84 #include <linux/inet.h>
85 #include <linux/netdevice.h>
86 #include <linux/skbuff.h>
87 #include <linux/inetdevice.h>
88 #include <linux/igmp.h>
89 #include <linux/if_arp.h>
90 #include <linux/rtnetlink.h>
91 #include <linux/times.h>
92
93 #include <net/net_namespace.h>
94 #include <net/arp.h>
95 #include <net/ip.h>
96 #include <net/protocol.h>
97 #include <net/route.h>
98 #include <net/sock.h>
99 #include <net/checksum.h>
100 #include <linux/netfilter_ipv4.h>
101 #ifdef CONFIG_IP_MROUTE
102 #include <linux/mroute.h>
103 #endif
104 #ifdef CONFIG_PROC_FS
105 #include <linux/proc_fs.h>
106 #include <linux/seq_file.h>
107 #endif
108
109 #define IP_MAX_MEMBERSHIPS      20
110 #define IP_MAX_MSF              10
111
112 #ifdef CONFIG_IP_MULTICAST
113 /* Parameter names and values are taken from igmp-v2-06 draft */
114
115 #define IGMP_V1_Router_Present_Timeout          (400*HZ)
116 #define IGMP_V2_Router_Present_Timeout          (400*HZ)
117 #define IGMP_Unsolicited_Report_Interval        (10*HZ)
118 #define IGMP_Query_Response_Interval            (10*HZ)
119 #define IGMP_Unsolicited_Report_Count           2
120
121
122 #define IGMP_Initial_Report_Delay               (1)
123
124 /* IGMP_Initial_Report_Delay is not from IGMP specs!
125  * IGMP specs require to report membership immediately after
126  * joining a group, but we delay the first report by a
127  * small interval. It seems more natural and still does not
128  * contradict to specs provided this delay is small enough.
129  */
130
131 #define IGMP_V1_SEEN(in_dev) \
132         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
133          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
134          ((in_dev)->mr_v1_seen && \
135           time_before(jiffies, (in_dev)->mr_v1_seen)))
136 #define IGMP_V2_SEEN(in_dev) \
137         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
138          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
139          ((in_dev)->mr_v2_seen && \
140           time_before(jiffies, (in_dev)->mr_v2_seen)))
141
142 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
143 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr);
144 static void igmpv3_clear_delrec(struct in_device *in_dev);
145 static int sf_setstate(struct ip_mc_list *pmc);
146 static void sf_markstate(struct ip_mc_list *pmc);
147 #endif
148 static void ip_mc_clear_src(struct ip_mc_list *pmc);
149 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
150                          int sfcount, __be32 *psfsrc, int delta);
151
152
153 static void ip_mc_list_reclaim(struct rcu_head *head)
154 {
155         kfree(container_of(head, struct ip_mc_list, rcu));
156 }
157
158 static void ip_ma_put(struct ip_mc_list *im)
159 {
160         if (atomic_dec_and_test(&im->refcnt)) {
161                 in_dev_put(im->interface);
162                 call_rcu(&im->rcu, ip_mc_list_reclaim);
163         }
164 }
165
166 #define for_each_pmc_rcu(in_dev, pmc)                           \
167         for (pmc = rcu_dereference(in_dev->mc_list);            \
168              pmc != NULL;                                       \
169              pmc = rcu_dereference(pmc->next_rcu))
170
171 #define for_each_pmc_rtnl(in_dev, pmc)                          \
172         for (pmc = rtnl_dereference(in_dev->mc_list);           \
173              pmc != NULL;                                       \
174              pmc = rtnl_dereference(pmc->next_rcu))
175
176 #ifdef CONFIG_IP_MULTICAST
177
178 /*
179  *      Timer management
180  */
181
182 static void igmp_stop_timer(struct ip_mc_list *im)
183 {
184         spin_lock_bh(&im->lock);
185         if (del_timer(&im->timer))
186                 atomic_dec(&im->refcnt);
187         im->tm_running = 0;
188         im->reporter = 0;
189         im->unsolicit_count = 0;
190         spin_unlock_bh(&im->lock);
191 }
192
193 /* It must be called with locked im->lock */
194 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
195 {
196         int tv = net_random() % max_delay;
197
198         im->tm_running = 1;
199         if (!mod_timer(&im->timer, jiffies+tv+2))
200                 atomic_inc(&im->refcnt);
201 }
202
203 static void igmp_gq_start_timer(struct in_device *in_dev)
204 {
205         int tv = net_random() % in_dev->mr_maxdelay;
206
207         in_dev->mr_gq_running = 1;
208         if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
209                 in_dev_hold(in_dev);
210 }
211
212 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
213 {
214         int tv = net_random() % delay;
215
216         if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
217                 in_dev_hold(in_dev);
218 }
219
220 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
221 {
222         spin_lock_bh(&im->lock);
223         im->unsolicit_count = 0;
224         if (del_timer(&im->timer)) {
225                 if ((long)(im->timer.expires-jiffies) < max_delay) {
226                         add_timer(&im->timer);
227                         im->tm_running = 1;
228                         spin_unlock_bh(&im->lock);
229                         return;
230                 }
231                 atomic_dec(&im->refcnt);
232         }
233         igmp_start_timer(im, max_delay);
234         spin_unlock_bh(&im->lock);
235 }
236
237
238 /*
239  *      Send an IGMP report.
240  */
241
242 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
243
244
245 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
246         int gdeleted, int sdeleted)
247 {
248         switch (type) {
249         case IGMPV3_MODE_IS_INCLUDE:
250         case IGMPV3_MODE_IS_EXCLUDE:
251                 if (gdeleted || sdeleted)
252                         return 0;
253                 if (!(pmc->gsquery && !psf->sf_gsresp)) {
254                         if (pmc->sfmode == MCAST_INCLUDE)
255                                 return 1;
256                         /* don't include if this source is excluded
257                          * in all filters
258                          */
259                         if (psf->sf_count[MCAST_INCLUDE])
260                                 return type == IGMPV3_MODE_IS_INCLUDE;
261                         return pmc->sfcount[MCAST_EXCLUDE] ==
262                                 psf->sf_count[MCAST_EXCLUDE];
263                 }
264                 return 0;
265         case IGMPV3_CHANGE_TO_INCLUDE:
266                 if (gdeleted || sdeleted)
267                         return 0;
268                 return psf->sf_count[MCAST_INCLUDE] != 0;
269         case IGMPV3_CHANGE_TO_EXCLUDE:
270                 if (gdeleted || sdeleted)
271                         return 0;
272                 if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
273                     psf->sf_count[MCAST_INCLUDE])
274                         return 0;
275                 return pmc->sfcount[MCAST_EXCLUDE] ==
276                         psf->sf_count[MCAST_EXCLUDE];
277         case IGMPV3_ALLOW_NEW_SOURCES:
278                 if (gdeleted || !psf->sf_crcount)
279                         return 0;
280                 return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
281         case IGMPV3_BLOCK_OLD_SOURCES:
282                 if (pmc->sfmode == MCAST_INCLUDE)
283                         return gdeleted || (psf->sf_crcount && sdeleted);
284                 return psf->sf_crcount && !gdeleted && !sdeleted;
285         }
286         return 0;
287 }
288
289 static int
290 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
291 {
292         struct ip_sf_list *psf;
293         int scount = 0;
294
295         for (psf=pmc->sources; psf; psf=psf->sf_next) {
296                 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
297                         continue;
298                 scount++;
299         }
300         return scount;
301 }
302
303 static struct sk_buff *igmpv3_newpack(struct net_device *dev, int size)
304 {
305         struct sk_buff *skb;
306         struct rtable *rt;
307         struct iphdr *pip;
308         struct igmpv3_report *pig;
309         struct net *net = dev_net(dev);
310
311         skb = alloc_skb(size + LL_ALLOCATED_SPACE(dev), GFP_ATOMIC);
312         if (skb == NULL)
313                 return NULL;
314
315         {
316                 struct flowi fl = { .oif = dev->ifindex,
317                                     .fl4_dst = IGMPV3_ALL_MCR,
318                                     .proto = IPPROTO_IGMP };
319                 if (ip_route_output_key(net, &rt, &fl)) {
320                         kfree_skb(skb);
321                         return NULL;
322                 }
323         }
324         if (rt->rt_src == 0) {
325                 kfree_skb(skb);
326                 ip_rt_put(rt);
327                 return NULL;
328         }
329
330         skb_dst_set(skb, &rt->dst);
331         skb->dev = dev;
332
333         skb_reserve(skb, LL_RESERVED_SPACE(dev));
334
335         skb_reset_network_header(skb);
336         pip = ip_hdr(skb);
337         skb_put(skb, sizeof(struct iphdr) + 4);
338
339         pip->version  = 4;
340         pip->ihl      = (sizeof(struct iphdr)+4)>>2;
341         pip->tos      = 0xc0;
342         pip->frag_off = htons(IP_DF);
343         pip->ttl      = 1;
344         pip->daddr    = rt->rt_dst;
345         pip->saddr    = rt->rt_src;
346         pip->protocol = IPPROTO_IGMP;
347         pip->tot_len  = 0;      /* filled in later */
348         ip_select_ident(pip, &rt->dst, NULL);
349         ((u8*)&pip[1])[0] = IPOPT_RA;
350         ((u8*)&pip[1])[1] = 4;
351         ((u8*)&pip[1])[2] = 0;
352         ((u8*)&pip[1])[3] = 0;
353
354         skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
355         skb_put(skb, sizeof(*pig));
356         pig = igmpv3_report_hdr(skb);
357         pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
358         pig->resv1 = 0;
359         pig->csum = 0;
360         pig->resv2 = 0;
361         pig->ngrec = 0;
362         return skb;
363 }
364
365 static int igmpv3_sendpack(struct sk_buff *skb)
366 {
367         struct igmphdr *pig = igmp_hdr(skb);
368         const int igmplen = skb->tail - skb->transport_header;
369
370         pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
371
372         return ip_local_out(skb);
373 }
374
375 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
376 {
377         return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
378 }
379
380 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
381         int type, struct igmpv3_grec **ppgr)
382 {
383         struct net_device *dev = pmc->interface->dev;
384         struct igmpv3_report *pih;
385         struct igmpv3_grec *pgr;
386
387         if (!skb)
388                 skb = igmpv3_newpack(dev, dev->mtu);
389         if (!skb)
390                 return NULL;
391         pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
392         pgr->grec_type = type;
393         pgr->grec_auxwords = 0;
394         pgr->grec_nsrcs = 0;
395         pgr->grec_mca = pmc->multiaddr;
396         pih = igmpv3_report_hdr(skb);
397         pih->ngrec = htons(ntohs(pih->ngrec)+1);
398         *ppgr = pgr;
399         return skb;
400 }
401
402 #define AVAILABLE(skb) ((skb) ? ((skb)->dev ? (skb)->dev->mtu - (skb)->len : \
403         skb_tailroom(skb)) : 0)
404
405 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
406         int type, int gdeleted, int sdeleted)
407 {
408         struct net_device *dev = pmc->interface->dev;
409         struct igmpv3_report *pih;
410         struct igmpv3_grec *pgr = NULL;
411         struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
412         int scount, stotal, first, isquery, truncate;
413
414         if (pmc->multiaddr == IGMP_ALL_HOSTS)
415                 return skb;
416
417         isquery = type == IGMPV3_MODE_IS_INCLUDE ||
418                   type == IGMPV3_MODE_IS_EXCLUDE;
419         truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
420                     type == IGMPV3_CHANGE_TO_EXCLUDE;
421
422         stotal = scount = 0;
423
424         psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
425
426         if (!*psf_list)
427                 goto empty_source;
428
429         pih = skb ? igmpv3_report_hdr(skb) : NULL;
430
431         /* EX and TO_EX get a fresh packet, if needed */
432         if (truncate) {
433                 if (pih && pih->ngrec &&
434                     AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
435                         if (skb)
436                                 igmpv3_sendpack(skb);
437                         skb = igmpv3_newpack(dev, dev->mtu);
438                 }
439         }
440         first = 1;
441         psf_prev = NULL;
442         for (psf=*psf_list; psf; psf=psf_next) {
443                 __be32 *psrc;
444
445                 psf_next = psf->sf_next;
446
447                 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
448                         psf_prev = psf;
449                         continue;
450                 }
451
452                 /* clear marks on query responses */
453                 if (isquery)
454                         psf->sf_gsresp = 0;
455
456                 if (AVAILABLE(skb) < sizeof(__be32) +
457                     first*sizeof(struct igmpv3_grec)) {
458                         if (truncate && !first)
459                                 break;   /* truncate these */
460                         if (pgr)
461                                 pgr->grec_nsrcs = htons(scount);
462                         if (skb)
463                                 igmpv3_sendpack(skb);
464                         skb = igmpv3_newpack(dev, dev->mtu);
465                         first = 1;
466                         scount = 0;
467                 }
468                 if (first) {
469                         skb = add_grhead(skb, pmc, type, &pgr);
470                         first = 0;
471                 }
472                 if (!skb)
473                         return NULL;
474                 psrc = (__be32 *)skb_put(skb, sizeof(__be32));
475                 *psrc = psf->sf_inaddr;
476                 scount++; stotal++;
477                 if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
478                      type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
479                         psf->sf_crcount--;
480                         if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
481                                 if (psf_prev)
482                                         psf_prev->sf_next = psf->sf_next;
483                                 else
484                                         *psf_list = psf->sf_next;
485                                 kfree(psf);
486                                 continue;
487                         }
488                 }
489                 psf_prev = psf;
490         }
491
492 empty_source:
493         if (!stotal) {
494                 if (type == IGMPV3_ALLOW_NEW_SOURCES ||
495                     type == IGMPV3_BLOCK_OLD_SOURCES)
496                         return skb;
497                 if (pmc->crcount || isquery) {
498                         /* make sure we have room for group header */
499                         if (skb && AVAILABLE(skb)<sizeof(struct igmpv3_grec)) {
500                                 igmpv3_sendpack(skb);
501                                 skb = NULL; /* add_grhead will get a new one */
502                         }
503                         skb = add_grhead(skb, pmc, type, &pgr);
504                 }
505         }
506         if (pgr)
507                 pgr->grec_nsrcs = htons(scount);
508
509         if (isquery)
510                 pmc->gsquery = 0;       /* clear query state on report */
511         return skb;
512 }
513
514 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
515 {
516         struct sk_buff *skb = NULL;
517         int type;
518
519         if (!pmc) {
520                 rcu_read_lock();
521                 for_each_pmc_rcu(in_dev, pmc) {
522                         if (pmc->multiaddr == IGMP_ALL_HOSTS)
523                                 continue;
524                         spin_lock_bh(&pmc->lock);
525                         if (pmc->sfcount[MCAST_EXCLUDE])
526                                 type = IGMPV3_MODE_IS_EXCLUDE;
527                         else
528                                 type = IGMPV3_MODE_IS_INCLUDE;
529                         skb = add_grec(skb, pmc, type, 0, 0);
530                         spin_unlock_bh(&pmc->lock);
531                 }
532                 rcu_read_unlock();
533         } else {
534                 spin_lock_bh(&pmc->lock);
535                 if (pmc->sfcount[MCAST_EXCLUDE])
536                         type = IGMPV3_MODE_IS_EXCLUDE;
537                 else
538                         type = IGMPV3_MODE_IS_INCLUDE;
539                 skb = add_grec(skb, pmc, type, 0, 0);
540                 spin_unlock_bh(&pmc->lock);
541         }
542         if (!skb)
543                 return 0;
544         return igmpv3_sendpack(skb);
545 }
546
547 /*
548  * remove zero-count source records from a source filter list
549  */
550 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
551 {
552         struct ip_sf_list *psf_prev, *psf_next, *psf;
553
554         psf_prev = NULL;
555         for (psf=*ppsf; psf; psf = psf_next) {
556                 psf_next = psf->sf_next;
557                 if (psf->sf_crcount == 0) {
558                         if (psf_prev)
559                                 psf_prev->sf_next = psf->sf_next;
560                         else
561                                 *ppsf = psf->sf_next;
562                         kfree(psf);
563                 } else
564                         psf_prev = psf;
565         }
566 }
567
568 static void igmpv3_send_cr(struct in_device *in_dev)
569 {
570         struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
571         struct sk_buff *skb = NULL;
572         int type, dtype;
573
574         rcu_read_lock();
575         spin_lock_bh(&in_dev->mc_tomb_lock);
576
577         /* deleted MCA's */
578         pmc_prev = NULL;
579         for (pmc=in_dev->mc_tomb; pmc; pmc=pmc_next) {
580                 pmc_next = pmc->next;
581                 if (pmc->sfmode == MCAST_INCLUDE) {
582                         type = IGMPV3_BLOCK_OLD_SOURCES;
583                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
584                         skb = add_grec(skb, pmc, type, 1, 0);
585                         skb = add_grec(skb, pmc, dtype, 1, 1);
586                 }
587                 if (pmc->crcount) {
588                         if (pmc->sfmode == MCAST_EXCLUDE) {
589                                 type = IGMPV3_CHANGE_TO_INCLUDE;
590                                 skb = add_grec(skb, pmc, type, 1, 0);
591                         }
592                         pmc->crcount--;
593                         if (pmc->crcount == 0) {
594                                 igmpv3_clear_zeros(&pmc->tomb);
595                                 igmpv3_clear_zeros(&pmc->sources);
596                         }
597                 }
598                 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
599                         if (pmc_prev)
600                                 pmc_prev->next = pmc_next;
601                         else
602                                 in_dev->mc_tomb = pmc_next;
603                         in_dev_put(pmc->interface);
604                         kfree(pmc);
605                 } else
606                         pmc_prev = pmc;
607         }
608         spin_unlock_bh(&in_dev->mc_tomb_lock);
609
610         /* change recs */
611         for_each_pmc_rcu(in_dev, pmc) {
612                 spin_lock_bh(&pmc->lock);
613                 if (pmc->sfcount[MCAST_EXCLUDE]) {
614                         type = IGMPV3_BLOCK_OLD_SOURCES;
615                         dtype = IGMPV3_ALLOW_NEW_SOURCES;
616                 } else {
617                         type = IGMPV3_ALLOW_NEW_SOURCES;
618                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
619                 }
620                 skb = add_grec(skb, pmc, type, 0, 0);
621                 skb = add_grec(skb, pmc, dtype, 0, 1);  /* deleted sources */
622
623                 /* filter mode changes */
624                 if (pmc->crcount) {
625                         if (pmc->sfmode == MCAST_EXCLUDE)
626                                 type = IGMPV3_CHANGE_TO_EXCLUDE;
627                         else
628                                 type = IGMPV3_CHANGE_TO_INCLUDE;
629                         skb = add_grec(skb, pmc, type, 0, 0);
630                         pmc->crcount--;
631                 }
632                 spin_unlock_bh(&pmc->lock);
633         }
634         rcu_read_unlock();
635
636         if (!skb)
637                 return;
638         (void) igmpv3_sendpack(skb);
639 }
640
641 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
642         int type)
643 {
644         struct sk_buff *skb;
645         struct iphdr *iph;
646         struct igmphdr *ih;
647         struct rtable *rt;
648         struct net_device *dev = in_dev->dev;
649         struct net *net = dev_net(dev);
650         __be32  group = pmc ? pmc->multiaddr : 0;
651         __be32  dst;
652
653         if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
654                 return igmpv3_send_report(in_dev, pmc);
655         else if (type == IGMP_HOST_LEAVE_MESSAGE)
656                 dst = IGMP_ALL_ROUTER;
657         else
658                 dst = group;
659
660         {
661                 struct flowi fl = { .oif = dev->ifindex,
662                                     .fl4_dst = dst,
663                                     .proto = IPPROTO_IGMP };
664                 if (ip_route_output_key(net, &rt, &fl))
665                         return -1;
666         }
667         if (rt->rt_src == 0) {
668                 ip_rt_put(rt);
669                 return -1;
670         }
671
672         skb = alloc_skb(IGMP_SIZE+LL_ALLOCATED_SPACE(dev), GFP_ATOMIC);
673         if (skb == NULL) {
674                 ip_rt_put(rt);
675                 return -1;
676         }
677
678         skb_dst_set(skb, &rt->dst);
679
680         skb_reserve(skb, LL_RESERVED_SPACE(dev));
681
682         skb_reset_network_header(skb);
683         iph = ip_hdr(skb);
684         skb_put(skb, sizeof(struct iphdr) + 4);
685
686         iph->version  = 4;
687         iph->ihl      = (sizeof(struct iphdr)+4)>>2;
688         iph->tos      = 0xc0;
689         iph->frag_off = htons(IP_DF);
690         iph->ttl      = 1;
691         iph->daddr    = dst;
692         iph->saddr    = rt->rt_src;
693         iph->protocol = IPPROTO_IGMP;
694         ip_select_ident(iph, &rt->dst, NULL);
695         ((u8*)&iph[1])[0] = IPOPT_RA;
696         ((u8*)&iph[1])[1] = 4;
697         ((u8*)&iph[1])[2] = 0;
698         ((u8*)&iph[1])[3] = 0;
699
700         ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
701         ih->type = type;
702         ih->code = 0;
703         ih->csum = 0;
704         ih->group = group;
705         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
706
707         return ip_local_out(skb);
708 }
709
710 static void igmp_gq_timer_expire(unsigned long data)
711 {
712         struct in_device *in_dev = (struct in_device *)data;
713
714         in_dev->mr_gq_running = 0;
715         igmpv3_send_report(in_dev, NULL);
716         __in_dev_put(in_dev);
717 }
718
719 static void igmp_ifc_timer_expire(unsigned long data)
720 {
721         struct in_device *in_dev = (struct in_device *)data;
722
723         igmpv3_send_cr(in_dev);
724         if (in_dev->mr_ifc_count) {
725                 in_dev->mr_ifc_count--;
726                 igmp_ifc_start_timer(in_dev, IGMP_Unsolicited_Report_Interval);
727         }
728         __in_dev_put(in_dev);
729 }
730
731 static void igmp_ifc_event(struct in_device *in_dev)
732 {
733         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
734                 return;
735         in_dev->mr_ifc_count = in_dev->mr_qrv ? in_dev->mr_qrv :
736                 IGMP_Unsolicited_Report_Count;
737         igmp_ifc_start_timer(in_dev, 1);
738 }
739
740
741 static void igmp_timer_expire(unsigned long data)
742 {
743         struct ip_mc_list *im=(struct ip_mc_list *)data;
744         struct in_device *in_dev = im->interface;
745
746         spin_lock(&im->lock);
747         im->tm_running = 0;
748
749         if (im->unsolicit_count) {
750                 im->unsolicit_count--;
751                 igmp_start_timer(im, IGMP_Unsolicited_Report_Interval);
752         }
753         im->reporter = 1;
754         spin_unlock(&im->lock);
755
756         if (IGMP_V1_SEEN(in_dev))
757                 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
758         else if (IGMP_V2_SEEN(in_dev))
759                 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
760         else
761                 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
762
763         ip_ma_put(im);
764 }
765
766 /* mark EXCLUDE-mode sources */
767 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
768 {
769         struct ip_sf_list *psf;
770         int i, scount;
771
772         scount = 0;
773         for (psf=pmc->sources; psf; psf=psf->sf_next) {
774                 if (scount == nsrcs)
775                         break;
776                 for (i=0; i<nsrcs; i++) {
777                         /* skip inactive filters */
778                         if (pmc->sfcount[MCAST_INCLUDE] ||
779                             pmc->sfcount[MCAST_EXCLUDE] !=
780                             psf->sf_count[MCAST_EXCLUDE])
781                                 continue;
782                         if (srcs[i] == psf->sf_inaddr) {
783                                 scount++;
784                                 break;
785                         }
786                 }
787         }
788         pmc->gsquery = 0;
789         if (scount == nsrcs)    /* all sources excluded */
790                 return 0;
791         return 1;
792 }
793
794 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
795 {
796         struct ip_sf_list *psf;
797         int i, scount;
798
799         if (pmc->sfmode == MCAST_EXCLUDE)
800                 return igmp_xmarksources(pmc, nsrcs, srcs);
801
802         /* mark INCLUDE-mode sources */
803         scount = 0;
804         for (psf=pmc->sources; psf; psf=psf->sf_next) {
805                 if (scount == nsrcs)
806                         break;
807                 for (i=0; i<nsrcs; i++)
808                         if (srcs[i] == psf->sf_inaddr) {
809                                 psf->sf_gsresp = 1;
810                                 scount++;
811                                 break;
812                         }
813         }
814         if (!scount) {
815                 pmc->gsquery = 0;
816                 return 0;
817         }
818         pmc->gsquery = 1;
819         return 1;
820 }
821
822 static void igmp_heard_report(struct in_device *in_dev, __be32 group)
823 {
824         struct ip_mc_list *im;
825
826         /* Timers are only set for non-local groups */
827
828         if (group == IGMP_ALL_HOSTS)
829                 return;
830
831         rcu_read_lock();
832         for_each_pmc_rcu(in_dev, im) {
833                 if (im->multiaddr == group) {
834                         igmp_stop_timer(im);
835                         break;
836                 }
837         }
838         rcu_read_unlock();
839 }
840
841 static void igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
842         int len)
843 {
844         struct igmphdr          *ih = igmp_hdr(skb);
845         struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
846         struct ip_mc_list       *im;
847         __be32                  group = ih->group;
848         int                     max_delay;
849         int                     mark = 0;
850
851
852         if (len == 8) {
853                 if (ih->code == 0) {
854                         /* Alas, old v1 router presents here. */
855
856                         max_delay = IGMP_Query_Response_Interval;
857                         in_dev->mr_v1_seen = jiffies +
858                                 IGMP_V1_Router_Present_Timeout;
859                         group = 0;
860                 } else {
861                         /* v2 router present */
862                         max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
863                         in_dev->mr_v2_seen = jiffies +
864                                 IGMP_V2_Router_Present_Timeout;
865                 }
866                 /* cancel the interface change timer */
867                 in_dev->mr_ifc_count = 0;
868                 if (del_timer(&in_dev->mr_ifc_timer))
869                         __in_dev_put(in_dev);
870                 /* clear deleted report items */
871                 igmpv3_clear_delrec(in_dev);
872         } else if (len < 12) {
873                 return; /* ignore bogus packet; freed by caller */
874         } else if (IGMP_V1_SEEN(in_dev)) {
875                 /* This is a v3 query with v1 queriers present */
876                 max_delay = IGMP_Query_Response_Interval;
877                 group = 0;
878         } else if (IGMP_V2_SEEN(in_dev)) {
879                 /* this is a v3 query with v2 queriers present;
880                  * Interpretation of the max_delay code is problematic here.
881                  * A real v2 host would use ih_code directly, while v3 has a
882                  * different encoding. We use the v3 encoding as more likely
883                  * to be intended in a v3 query.
884                  */
885                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
886         } else { /* v3 */
887                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
888                         return;
889
890                 ih3 = igmpv3_query_hdr(skb);
891                 if (ih3->nsrcs) {
892                         if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
893                                            + ntohs(ih3->nsrcs)*sizeof(__be32)))
894                                 return;
895                         ih3 = igmpv3_query_hdr(skb);
896                 }
897
898                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
899                 if (!max_delay)
900                         max_delay = 1;  /* can't mod w/ 0 */
901                 in_dev->mr_maxdelay = max_delay;
902                 if (ih3->qrv)
903                         in_dev->mr_qrv = ih3->qrv;
904                 if (!group) { /* general query */
905                         if (ih3->nsrcs)
906                                 return; /* no sources allowed */
907                         igmp_gq_start_timer(in_dev);
908                         return;
909                 }
910                 /* mark sources to include, if group & source-specific */
911                 mark = ih3->nsrcs != 0;
912         }
913
914         /*
915          * - Start the timers in all of our membership records
916          *   that the query applies to for the interface on
917          *   which the query arrived excl. those that belong
918          *   to a "local" group (224.0.0.X)
919          * - For timers already running check if they need to
920          *   be reset.
921          * - Use the igmp->igmp_code field as the maximum
922          *   delay possible
923          */
924         rcu_read_lock();
925         for_each_pmc_rcu(in_dev, im) {
926                 int changed;
927
928                 if (group && group != im->multiaddr)
929                         continue;
930                 if (im->multiaddr == IGMP_ALL_HOSTS)
931                         continue;
932                 spin_lock_bh(&im->lock);
933                 if (im->tm_running)
934                         im->gsquery = im->gsquery && mark;
935                 else
936                         im->gsquery = mark;
937                 changed = !im->gsquery ||
938                         igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
939                 spin_unlock_bh(&im->lock);
940                 if (changed)
941                         igmp_mod_timer(im, max_delay);
942         }
943         rcu_read_unlock();
944 }
945
946 /* called in rcu_read_lock() section */
947 int igmp_rcv(struct sk_buff *skb)
948 {
949         /* This basically follows the spec line by line -- see RFC1112 */
950         struct igmphdr *ih;
951         struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
952         int len = skb->len;
953
954         if (in_dev == NULL)
955                 goto drop;
956
957         if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
958                 goto drop;
959
960         switch (skb->ip_summed) {
961         case CHECKSUM_COMPLETE:
962                 if (!csum_fold(skb->csum))
963                         break;
964                 /* fall through */
965         case CHECKSUM_NONE:
966                 skb->csum = 0;
967                 if (__skb_checksum_complete(skb))
968                         goto drop;
969         }
970
971         ih = igmp_hdr(skb);
972         switch (ih->type) {
973         case IGMP_HOST_MEMBERSHIP_QUERY:
974                 igmp_heard_query(in_dev, skb, len);
975                 break;
976         case IGMP_HOST_MEMBERSHIP_REPORT:
977         case IGMPV2_HOST_MEMBERSHIP_REPORT:
978                 /* Is it our report looped back? */
979                 if (rt_is_output_route(skb_rtable(skb)))
980                         break;
981                 /* don't rely on MC router hearing unicast reports */
982                 if (skb->pkt_type == PACKET_MULTICAST ||
983                     skb->pkt_type == PACKET_BROADCAST)
984                         igmp_heard_report(in_dev, ih->group);
985                 break;
986         case IGMP_PIM:
987 #ifdef CONFIG_IP_PIMSM_V1
988                 return pim_rcv_v1(skb);
989 #endif
990         case IGMPV3_HOST_MEMBERSHIP_REPORT:
991         case IGMP_DVMRP:
992         case IGMP_TRACE:
993         case IGMP_HOST_LEAVE_MESSAGE:
994         case IGMP_MTRACE:
995         case IGMP_MTRACE_RESP:
996                 break;
997         default:
998                 break;
999         }
1000
1001 drop:
1002         kfree_skb(skb);
1003         return 0;
1004 }
1005
1006 #endif
1007
1008
1009 /*
1010  *      Add a filter to a device
1011  */
1012
1013 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1014 {
1015         char buf[MAX_ADDR_LEN];
1016         struct net_device *dev = in_dev->dev;
1017
1018         /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1019            We will get multicast token leakage, when IFF_MULTICAST
1020            is changed. This check should be done in dev->set_multicast_list
1021            routine. Something sort of:
1022            if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1023            --ANK
1024            */
1025         if (arp_mc_map(addr, buf, dev, 0) == 0)
1026                 dev_mc_add(dev, buf);
1027 }
1028
1029 /*
1030  *      Remove a filter from a device
1031  */
1032
1033 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1034 {
1035         char buf[MAX_ADDR_LEN];
1036         struct net_device *dev = in_dev->dev;
1037
1038         if (arp_mc_map(addr, buf, dev, 0) == 0)
1039                 dev_mc_del(dev, buf);
1040 }
1041
1042 #ifdef CONFIG_IP_MULTICAST
1043 /*
1044  * deleted ip_mc_list manipulation
1045  */
1046 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1047 {
1048         struct ip_mc_list *pmc;
1049
1050         /* this is an "ip_mc_list" for convenience; only the fields below
1051          * are actually used. In particular, the refcnt and users are not
1052          * used for management of the delete list. Using the same structure
1053          * for deleted items allows change reports to use common code with
1054          * non-deleted or query-response MCA's.
1055          */
1056         pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1057         if (!pmc)
1058                 return;
1059         spin_lock_bh(&im->lock);
1060         pmc->interface = im->interface;
1061         in_dev_hold(in_dev);
1062         pmc->multiaddr = im->multiaddr;
1063         pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1064                 IGMP_Unsolicited_Report_Count;
1065         pmc->sfmode = im->sfmode;
1066         if (pmc->sfmode == MCAST_INCLUDE) {
1067                 struct ip_sf_list *psf;
1068
1069                 pmc->tomb = im->tomb;
1070                 pmc->sources = im->sources;
1071                 im->tomb = im->sources = NULL;
1072                 for (psf=pmc->sources; psf; psf=psf->sf_next)
1073                         psf->sf_crcount = pmc->crcount;
1074         }
1075         spin_unlock_bh(&im->lock);
1076
1077         spin_lock_bh(&in_dev->mc_tomb_lock);
1078         pmc->next = in_dev->mc_tomb;
1079         in_dev->mc_tomb = pmc;
1080         spin_unlock_bh(&in_dev->mc_tomb_lock);
1081 }
1082
1083 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1084 {
1085         struct ip_mc_list *pmc, *pmc_prev;
1086         struct ip_sf_list *psf, *psf_next;
1087
1088         spin_lock_bh(&in_dev->mc_tomb_lock);
1089         pmc_prev = NULL;
1090         for (pmc=in_dev->mc_tomb; pmc; pmc=pmc->next) {
1091                 if (pmc->multiaddr == multiaddr)
1092                         break;
1093                 pmc_prev = pmc;
1094         }
1095         if (pmc) {
1096                 if (pmc_prev)
1097                         pmc_prev->next = pmc->next;
1098                 else
1099                         in_dev->mc_tomb = pmc->next;
1100         }
1101         spin_unlock_bh(&in_dev->mc_tomb_lock);
1102         if (pmc) {
1103                 for (psf=pmc->tomb; psf; psf=psf_next) {
1104                         psf_next = psf->sf_next;
1105                         kfree(psf);
1106                 }
1107                 in_dev_put(pmc->interface);
1108                 kfree(pmc);
1109         }
1110 }
1111
1112 static void igmpv3_clear_delrec(struct in_device *in_dev)
1113 {
1114         struct ip_mc_list *pmc, *nextpmc;
1115
1116         spin_lock_bh(&in_dev->mc_tomb_lock);
1117         pmc = in_dev->mc_tomb;
1118         in_dev->mc_tomb = NULL;
1119         spin_unlock_bh(&in_dev->mc_tomb_lock);
1120
1121         for (; pmc; pmc = nextpmc) {
1122                 nextpmc = pmc->next;
1123                 ip_mc_clear_src(pmc);
1124                 in_dev_put(pmc->interface);
1125                 kfree(pmc);
1126         }
1127         /* clear dead sources, too */
1128         rcu_read_lock();
1129         for_each_pmc_rcu(in_dev, pmc) {
1130                 struct ip_sf_list *psf, *psf_next;
1131
1132                 spin_lock_bh(&pmc->lock);
1133                 psf = pmc->tomb;
1134                 pmc->tomb = NULL;
1135                 spin_unlock_bh(&pmc->lock);
1136                 for (; psf; psf=psf_next) {
1137                         psf_next = psf->sf_next;
1138                         kfree(psf);
1139                 }
1140         }
1141         rcu_read_unlock();
1142 }
1143 #endif
1144
1145 static void igmp_group_dropped(struct ip_mc_list *im)
1146 {
1147         struct in_device *in_dev = im->interface;
1148 #ifdef CONFIG_IP_MULTICAST
1149         int reporter;
1150 #endif
1151
1152         if (im->loaded) {
1153                 im->loaded = 0;
1154                 ip_mc_filter_del(in_dev, im->multiaddr);
1155         }
1156
1157 #ifdef CONFIG_IP_MULTICAST
1158         if (im->multiaddr == IGMP_ALL_HOSTS)
1159                 return;
1160
1161         reporter = im->reporter;
1162         igmp_stop_timer(im);
1163
1164         if (!in_dev->dead) {
1165                 if (IGMP_V1_SEEN(in_dev))
1166                         goto done;
1167                 if (IGMP_V2_SEEN(in_dev)) {
1168                         if (reporter)
1169                                 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1170                         goto done;
1171                 }
1172                 /* IGMPv3 */
1173                 igmpv3_add_delrec(in_dev, im);
1174
1175                 igmp_ifc_event(in_dev);
1176         }
1177 done:
1178 #endif
1179         ip_mc_clear_src(im);
1180 }
1181
1182 static void igmp_group_added(struct ip_mc_list *im)
1183 {
1184         struct in_device *in_dev = im->interface;
1185
1186         if (im->loaded == 0) {
1187                 im->loaded = 1;
1188                 ip_mc_filter_add(in_dev, im->multiaddr);
1189         }
1190
1191 #ifdef CONFIG_IP_MULTICAST
1192         if (im->multiaddr == IGMP_ALL_HOSTS)
1193                 return;
1194
1195         if (in_dev->dead)
1196                 return;
1197         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1198                 spin_lock_bh(&im->lock);
1199                 igmp_start_timer(im, IGMP_Initial_Report_Delay);
1200                 spin_unlock_bh(&im->lock);
1201                 return;
1202         }
1203         /* else, v3 */
1204
1205         im->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1206                 IGMP_Unsolicited_Report_Count;
1207         igmp_ifc_event(in_dev);
1208 #endif
1209 }
1210
1211
1212 /*
1213  *      Multicast list managers
1214  */
1215
1216
1217 /*
1218  *      A socket has joined a multicast group on device dev.
1219  */
1220
1221 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1222 {
1223         struct ip_mc_list *im;
1224
1225         ASSERT_RTNL();
1226
1227         for_each_pmc_rtnl(in_dev, im) {
1228                 if (im->multiaddr == addr) {
1229                         im->users++;
1230                         ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1231                         goto out;
1232                 }
1233         }
1234
1235         im = kzalloc(sizeof(*im), GFP_KERNEL);
1236         if (!im)
1237                 goto out;
1238
1239         im->users = 1;
1240         im->interface = in_dev;
1241         in_dev_hold(in_dev);
1242         im->multiaddr = addr;
1243         /* initial mode is (EX, empty) */
1244         im->sfmode = MCAST_EXCLUDE;
1245         im->sfcount[MCAST_EXCLUDE] = 1;
1246         atomic_set(&im->refcnt, 1);
1247         spin_lock_init(&im->lock);
1248 #ifdef CONFIG_IP_MULTICAST
1249         setup_timer(&im->timer, &igmp_timer_expire, (unsigned long)im);
1250         im->unsolicit_count = IGMP_Unsolicited_Report_Count;
1251 #endif
1252
1253         im->next_rcu = in_dev->mc_list;
1254         in_dev->mc_count++;
1255         rcu_assign_pointer(in_dev->mc_list, im);
1256
1257 #ifdef CONFIG_IP_MULTICAST
1258         igmpv3_del_delrec(in_dev, im->multiaddr);
1259 #endif
1260         igmp_group_added(im);
1261         if (!in_dev->dead)
1262                 ip_rt_multicast_event(in_dev);
1263 out:
1264         return;
1265 }
1266 EXPORT_SYMBOL(ip_mc_inc_group);
1267
1268 /*
1269  *      Resend IGMP JOIN report; used for bonding.
1270  *      Called with rcu_read_lock()
1271  */
1272 void ip_mc_rejoin_groups(struct in_device *in_dev)
1273 {
1274 #ifdef CONFIG_IP_MULTICAST
1275         struct ip_mc_list *im;
1276         int type;
1277
1278         for_each_pmc_rcu(in_dev, im) {
1279                 if (im->multiaddr == IGMP_ALL_HOSTS)
1280                         continue;
1281
1282                 /* a failover is happening and switches
1283                  * must be notified immediately
1284                  */
1285                 if (IGMP_V1_SEEN(in_dev))
1286                         type = IGMP_HOST_MEMBERSHIP_REPORT;
1287                 else if (IGMP_V2_SEEN(in_dev))
1288                         type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1289                 else
1290                         type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1291                 igmp_send_report(in_dev, im, type);
1292         }
1293 #endif
1294 }
1295 EXPORT_SYMBOL(ip_mc_rejoin_groups);
1296
1297 /*
1298  *      A socket has left a multicast group on device dev
1299  */
1300
1301 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1302 {
1303         struct ip_mc_list *i;
1304         struct ip_mc_list __rcu **ip;
1305
1306         ASSERT_RTNL();
1307
1308         for (ip = &in_dev->mc_list;
1309              (i = rtnl_dereference(*ip)) != NULL;
1310              ip = &i->next_rcu) {
1311                 if (i->multiaddr == addr) {
1312                         if (--i->users == 0) {
1313                                 *ip = i->next_rcu;
1314                                 in_dev->mc_count--;
1315                                 igmp_group_dropped(i);
1316
1317                                 if (!in_dev->dead)
1318                                         ip_rt_multicast_event(in_dev);
1319
1320                                 ip_ma_put(i);
1321                                 return;
1322                         }
1323                         break;
1324                 }
1325         }
1326 }
1327 EXPORT_SYMBOL(ip_mc_dec_group);
1328
1329 /* Device changing type */
1330
1331 void ip_mc_unmap(struct in_device *in_dev)
1332 {
1333         struct ip_mc_list *pmc;
1334
1335         ASSERT_RTNL();
1336
1337         for_each_pmc_rtnl(in_dev, pmc)
1338                 igmp_group_dropped(pmc);
1339 }
1340
1341 void ip_mc_remap(struct in_device *in_dev)
1342 {
1343         struct ip_mc_list *pmc;
1344
1345         ASSERT_RTNL();
1346
1347         for_each_pmc_rtnl(in_dev, pmc)
1348                 igmp_group_added(pmc);
1349 }
1350
1351 /* Device going down */
1352
1353 void ip_mc_down(struct in_device *in_dev)
1354 {
1355         struct ip_mc_list *pmc;
1356
1357         ASSERT_RTNL();
1358
1359         for_each_pmc_rtnl(in_dev, pmc)
1360                 igmp_group_dropped(pmc);
1361
1362 #ifdef CONFIG_IP_MULTICAST
1363         in_dev->mr_ifc_count = 0;
1364         if (del_timer(&in_dev->mr_ifc_timer))
1365                 __in_dev_put(in_dev);
1366         in_dev->mr_gq_running = 0;
1367         if (del_timer(&in_dev->mr_gq_timer))
1368                 __in_dev_put(in_dev);
1369         igmpv3_clear_delrec(in_dev);
1370 #endif
1371
1372         ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1373 }
1374
1375 void ip_mc_init_dev(struct in_device *in_dev)
1376 {
1377         ASSERT_RTNL();
1378
1379         in_dev->mc_tomb = NULL;
1380 #ifdef CONFIG_IP_MULTICAST
1381         in_dev->mr_gq_running = 0;
1382         setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1383                         (unsigned long)in_dev);
1384         in_dev->mr_ifc_count = 0;
1385         in_dev->mc_count     = 0;
1386         setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1387                         (unsigned long)in_dev);
1388         in_dev->mr_qrv = IGMP_Unsolicited_Report_Count;
1389 #endif
1390
1391         spin_lock_init(&in_dev->mc_tomb_lock);
1392 }
1393
1394 /* Device going up */
1395
1396 void ip_mc_up(struct in_device *in_dev)
1397 {
1398         struct ip_mc_list *pmc;
1399
1400         ASSERT_RTNL();
1401
1402         ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1403
1404         for_each_pmc_rtnl(in_dev, pmc)
1405                 igmp_group_added(pmc);
1406 }
1407
1408 /*
1409  *      Device is about to be destroyed: clean up.
1410  */
1411
1412 void ip_mc_destroy_dev(struct in_device *in_dev)
1413 {
1414         struct ip_mc_list *i;
1415
1416         ASSERT_RTNL();
1417
1418         /* Deactivate timers */
1419         ip_mc_down(in_dev);
1420
1421         while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1422                 in_dev->mc_list = i->next_rcu;
1423                 in_dev->mc_count--;
1424
1425                 igmp_group_dropped(i);
1426                 ip_ma_put(i);
1427         }
1428 }
1429
1430 /* RTNL is locked */
1431 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1432 {
1433         struct flowi fl = { .fl4_dst = imr->imr_multiaddr.s_addr };
1434         struct rtable *rt;
1435         struct net_device *dev = NULL;
1436         struct in_device *idev = NULL;
1437
1438         if (imr->imr_ifindex) {
1439                 idev = inetdev_by_index(net, imr->imr_ifindex);
1440                 return idev;
1441         }
1442         if (imr->imr_address.s_addr) {
1443                 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1444                 if (!dev)
1445                         return NULL;
1446         }
1447
1448         if (!dev && !ip_route_output_key(net, &rt, &fl)) {
1449                 dev = rt->dst.dev;
1450                 ip_rt_put(rt);
1451         }
1452         if (dev) {
1453                 imr->imr_ifindex = dev->ifindex;
1454                 idev = __in_dev_get_rtnl(dev);
1455         }
1456         return idev;
1457 }
1458
1459 /*
1460  *      Join a socket to a group
1461  */
1462 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1463 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1464
1465
1466 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1467         __be32 *psfsrc)
1468 {
1469         struct ip_sf_list *psf, *psf_prev;
1470         int rv = 0;
1471
1472         psf_prev = NULL;
1473         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1474                 if (psf->sf_inaddr == *psfsrc)
1475                         break;
1476                 psf_prev = psf;
1477         }
1478         if (!psf || psf->sf_count[sfmode] == 0) {
1479                 /* source filter not found, or count wrong =>  bug */
1480                 return -ESRCH;
1481         }
1482         psf->sf_count[sfmode]--;
1483         if (psf->sf_count[sfmode] == 0) {
1484                 ip_rt_multicast_event(pmc->interface);
1485         }
1486         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1487 #ifdef CONFIG_IP_MULTICAST
1488                 struct in_device *in_dev = pmc->interface;
1489 #endif
1490
1491                 /* no more filters for this source */
1492                 if (psf_prev)
1493                         psf_prev->sf_next = psf->sf_next;
1494                 else
1495                         pmc->sources = psf->sf_next;
1496 #ifdef CONFIG_IP_MULTICAST
1497                 if (psf->sf_oldin &&
1498                     !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1499                         psf->sf_crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1500                                 IGMP_Unsolicited_Report_Count;
1501                         psf->sf_next = pmc->tomb;
1502                         pmc->tomb = psf;
1503                         rv = 1;
1504                 } else
1505 #endif
1506                         kfree(psf);
1507         }
1508         return rv;
1509 }
1510
1511 #ifndef CONFIG_IP_MULTICAST
1512 #define igmp_ifc_event(x)       do { } while (0)
1513 #endif
1514
1515 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1516                          int sfcount, __be32 *psfsrc, int delta)
1517 {
1518         struct ip_mc_list *pmc;
1519         int     changerec = 0;
1520         int     i, err;
1521
1522         if (!in_dev)
1523                 return -ENODEV;
1524         rcu_read_lock();
1525         for_each_pmc_rcu(in_dev, pmc) {
1526                 if (*pmca == pmc->multiaddr)
1527                         break;
1528         }
1529         if (!pmc) {
1530                 /* MCA not found?? bug */
1531                 rcu_read_unlock();
1532                 return -ESRCH;
1533         }
1534         spin_lock_bh(&pmc->lock);
1535         rcu_read_unlock();
1536 #ifdef CONFIG_IP_MULTICAST
1537         sf_markstate(pmc);
1538 #endif
1539         if (!delta) {
1540                 err = -EINVAL;
1541                 if (!pmc->sfcount[sfmode])
1542                         goto out_unlock;
1543                 pmc->sfcount[sfmode]--;
1544         }
1545         err = 0;
1546         for (i=0; i<sfcount; i++) {
1547                 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1548
1549                 changerec |= rv > 0;
1550                 if (!err && rv < 0)
1551                         err = rv;
1552         }
1553         if (pmc->sfmode == MCAST_EXCLUDE &&
1554             pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1555             pmc->sfcount[MCAST_INCLUDE]) {
1556 #ifdef CONFIG_IP_MULTICAST
1557                 struct ip_sf_list *psf;
1558 #endif
1559
1560                 /* filter mode change */
1561                 pmc->sfmode = MCAST_INCLUDE;
1562 #ifdef CONFIG_IP_MULTICAST
1563                 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1564                         IGMP_Unsolicited_Report_Count;
1565                 in_dev->mr_ifc_count = pmc->crcount;
1566                 for (psf=pmc->sources; psf; psf = psf->sf_next)
1567                         psf->sf_crcount = 0;
1568                 igmp_ifc_event(pmc->interface);
1569         } else if (sf_setstate(pmc) || changerec) {
1570                 igmp_ifc_event(pmc->interface);
1571 #endif
1572         }
1573 out_unlock:
1574         spin_unlock_bh(&pmc->lock);
1575         return err;
1576 }
1577
1578 /*
1579  * Add multicast single-source filter to the interface list
1580  */
1581 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1582         __be32 *psfsrc, int delta)
1583 {
1584         struct ip_sf_list *psf, *psf_prev;
1585
1586         psf_prev = NULL;
1587         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1588                 if (psf->sf_inaddr == *psfsrc)
1589                         break;
1590                 psf_prev = psf;
1591         }
1592         if (!psf) {
1593                 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1594                 if (!psf)
1595                         return -ENOBUFS;
1596                 psf->sf_inaddr = *psfsrc;
1597                 if (psf_prev) {
1598                         psf_prev->sf_next = psf;
1599                 } else
1600                         pmc->sources = psf;
1601         }
1602         psf->sf_count[sfmode]++;
1603         if (psf->sf_count[sfmode] == 1) {
1604                 ip_rt_multicast_event(pmc->interface);
1605         }
1606         return 0;
1607 }
1608
1609 #ifdef CONFIG_IP_MULTICAST
1610 static void sf_markstate(struct ip_mc_list *pmc)
1611 {
1612         struct ip_sf_list *psf;
1613         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1614
1615         for (psf=pmc->sources; psf; psf=psf->sf_next)
1616                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1617                         psf->sf_oldin = mca_xcount ==
1618                                 psf->sf_count[MCAST_EXCLUDE] &&
1619                                 !psf->sf_count[MCAST_INCLUDE];
1620                 } else
1621                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1622 }
1623
1624 static int sf_setstate(struct ip_mc_list *pmc)
1625 {
1626         struct ip_sf_list *psf, *dpsf;
1627         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1628         int qrv = pmc->interface->mr_qrv;
1629         int new_in, rv;
1630
1631         rv = 0;
1632         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1633                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1634                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1635                                 !psf->sf_count[MCAST_INCLUDE];
1636                 } else
1637                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1638                 if (new_in) {
1639                         if (!psf->sf_oldin) {
1640                                 struct ip_sf_list *prev = NULL;
1641
1642                                 for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next) {
1643                                         if (dpsf->sf_inaddr == psf->sf_inaddr)
1644                                                 break;
1645                                         prev = dpsf;
1646                                 }
1647                                 if (dpsf) {
1648                                         if (prev)
1649                                                 prev->sf_next = dpsf->sf_next;
1650                                         else
1651                                                 pmc->tomb = dpsf->sf_next;
1652                                         kfree(dpsf);
1653                                 }
1654                                 psf->sf_crcount = qrv;
1655                                 rv++;
1656                         }
1657                 } else if (psf->sf_oldin) {
1658
1659                         psf->sf_crcount = 0;
1660                         /*
1661                          * add or update "delete" records if an active filter
1662                          * is now inactive
1663                          */
1664                         for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next)
1665                                 if (dpsf->sf_inaddr == psf->sf_inaddr)
1666                                         break;
1667                         if (!dpsf) {
1668                                 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1669                                 if (!dpsf)
1670                                         continue;
1671                                 *dpsf = *psf;
1672                                 /* pmc->lock held by callers */
1673                                 dpsf->sf_next = pmc->tomb;
1674                                 pmc->tomb = dpsf;
1675                         }
1676                         dpsf->sf_crcount = qrv;
1677                         rv++;
1678                 }
1679         }
1680         return rv;
1681 }
1682 #endif
1683
1684 /*
1685  * Add multicast source filter list to the interface list
1686  */
1687 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1688                          int sfcount, __be32 *psfsrc, int delta)
1689 {
1690         struct ip_mc_list *pmc;
1691         int     isexclude;
1692         int     i, err;
1693
1694         if (!in_dev)
1695                 return -ENODEV;
1696         rcu_read_lock();
1697         for_each_pmc_rcu(in_dev, pmc) {
1698                 if (*pmca == pmc->multiaddr)
1699                         break;
1700         }
1701         if (!pmc) {
1702                 /* MCA not found?? bug */
1703                 rcu_read_unlock();
1704                 return -ESRCH;
1705         }
1706         spin_lock_bh(&pmc->lock);
1707         rcu_read_unlock();
1708
1709 #ifdef CONFIG_IP_MULTICAST
1710         sf_markstate(pmc);
1711 #endif
1712         isexclude = pmc->sfmode == MCAST_EXCLUDE;
1713         if (!delta)
1714                 pmc->sfcount[sfmode]++;
1715         err = 0;
1716         for (i=0; i<sfcount; i++) {
1717                 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i], delta);
1718                 if (err)
1719                         break;
1720         }
1721         if (err) {
1722                 int j;
1723
1724                 pmc->sfcount[sfmode]--;
1725                 for (j=0; j<i; j++)
1726                         (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1727         } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1728 #ifdef CONFIG_IP_MULTICAST
1729                 struct ip_sf_list *psf;
1730                 in_dev = pmc->interface;
1731 #endif
1732
1733                 /* filter mode change */
1734                 if (pmc->sfcount[MCAST_EXCLUDE])
1735                         pmc->sfmode = MCAST_EXCLUDE;
1736                 else if (pmc->sfcount[MCAST_INCLUDE])
1737                         pmc->sfmode = MCAST_INCLUDE;
1738 #ifdef CONFIG_IP_MULTICAST
1739                 /* else no filters; keep old mode for reports */
1740
1741                 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1742                         IGMP_Unsolicited_Report_Count;
1743                 in_dev->mr_ifc_count = pmc->crcount;
1744                 for (psf=pmc->sources; psf; psf = psf->sf_next)
1745                         psf->sf_crcount = 0;
1746                 igmp_ifc_event(in_dev);
1747         } else if (sf_setstate(pmc)) {
1748                 igmp_ifc_event(in_dev);
1749 #endif
1750         }
1751         spin_unlock_bh(&pmc->lock);
1752         return err;
1753 }
1754
1755 static void ip_mc_clear_src(struct ip_mc_list *pmc)
1756 {
1757         struct ip_sf_list *psf, *nextpsf;
1758
1759         for (psf=pmc->tomb; psf; psf=nextpsf) {
1760                 nextpsf = psf->sf_next;
1761                 kfree(psf);
1762         }
1763         pmc->tomb = NULL;
1764         for (psf=pmc->sources; psf; psf=nextpsf) {
1765                 nextpsf = psf->sf_next;
1766                 kfree(psf);
1767         }
1768         pmc->sources = NULL;
1769         pmc->sfmode = MCAST_EXCLUDE;
1770         pmc->sfcount[MCAST_INCLUDE] = 0;
1771         pmc->sfcount[MCAST_EXCLUDE] = 1;
1772 }
1773
1774
1775 /*
1776  * Join a multicast group
1777  */
1778 int ip_mc_join_group(struct sock *sk , struct ip_mreqn *imr)
1779 {
1780         int err;
1781         __be32 addr = imr->imr_multiaddr.s_addr;
1782         struct ip_mc_socklist *iml = NULL, *i;
1783         struct in_device *in_dev;
1784         struct inet_sock *inet = inet_sk(sk);
1785         struct net *net = sock_net(sk);
1786         int ifindex;
1787         int count = 0;
1788
1789         if (!ipv4_is_multicast(addr))
1790                 return -EINVAL;
1791
1792         rtnl_lock();
1793
1794         in_dev = ip_mc_find_dev(net, imr);
1795
1796         if (!in_dev) {
1797                 iml = NULL;
1798                 err = -ENODEV;
1799                 goto done;
1800         }
1801
1802         err = -EADDRINUSE;
1803         ifindex = imr->imr_ifindex;
1804         for_each_pmc_rtnl(inet, i) {
1805                 if (i->multi.imr_multiaddr.s_addr == addr &&
1806                     i->multi.imr_ifindex == ifindex)
1807                         goto done;
1808                 count++;
1809         }
1810         err = -ENOBUFS;
1811         if (count >= sysctl_igmp_max_memberships)
1812                 goto done;
1813         iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
1814         if (iml == NULL)
1815                 goto done;
1816
1817         memcpy(&iml->multi, imr, sizeof(*imr));
1818         iml->next_rcu = inet->mc_list;
1819         iml->sflist = NULL;
1820         iml->sfmode = MCAST_EXCLUDE;
1821         rcu_assign_pointer(inet->mc_list, iml);
1822         ip_mc_inc_group(in_dev, addr);
1823         err = 0;
1824 done:
1825         rtnl_unlock();
1826         return err;
1827 }
1828 EXPORT_SYMBOL(ip_mc_join_group);
1829
1830 static void ip_sf_socklist_reclaim(struct rcu_head *rp)
1831 {
1832         kfree(container_of(rp, struct ip_sf_socklist, rcu));
1833         /* sk_omem_alloc should have been decreased by the caller*/
1834 }
1835
1836 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
1837                            struct in_device *in_dev)
1838 {
1839         struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
1840         int err;
1841
1842         if (psf == NULL) {
1843                 /* any-source empty exclude case */
1844                 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1845                         iml->sfmode, 0, NULL, 0);
1846         }
1847         err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1848                         iml->sfmode, psf->sl_count, psf->sl_addr, 0);
1849         rcu_assign_pointer(iml->sflist, NULL);
1850         /* decrease mem now to avoid the memleak warning */
1851         atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
1852         call_rcu(&psf->rcu, ip_sf_socklist_reclaim);
1853         return err;
1854 }
1855
1856
1857 static void ip_mc_socklist_reclaim(struct rcu_head *rp)
1858 {
1859         kfree(container_of(rp, struct ip_mc_socklist, rcu));
1860         /* sk_omem_alloc should have been decreased by the caller*/
1861 }
1862
1863
1864 /*
1865  *      Ask a socket to leave a group.
1866  */
1867
1868 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
1869 {
1870         struct inet_sock *inet = inet_sk(sk);
1871         struct ip_mc_socklist *iml;
1872         struct ip_mc_socklist __rcu **imlp;
1873         struct in_device *in_dev;
1874         struct net *net = sock_net(sk);
1875         __be32 group = imr->imr_multiaddr.s_addr;
1876         u32 ifindex;
1877         int ret = -EADDRNOTAVAIL;
1878
1879         rtnl_lock();
1880         in_dev = ip_mc_find_dev(net, imr);
1881         ifindex = imr->imr_ifindex;
1882         for (imlp = &inet->mc_list;
1883              (iml = rtnl_dereference(*imlp)) != NULL;
1884              imlp = &iml->next_rcu) {
1885                 if (iml->multi.imr_multiaddr.s_addr != group)
1886                         continue;
1887                 if (ifindex) {
1888                         if (iml->multi.imr_ifindex != ifindex)
1889                                 continue;
1890                 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
1891                                 iml->multi.imr_address.s_addr)
1892                         continue;
1893
1894                 (void) ip_mc_leave_src(sk, iml, in_dev);
1895
1896                 *imlp = iml->next_rcu;
1897
1898                 if (in_dev)
1899                         ip_mc_dec_group(in_dev, group);
1900                 rtnl_unlock();
1901                 /* decrease mem now to avoid the memleak warning */
1902                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
1903                 call_rcu(&iml->rcu, ip_mc_socklist_reclaim);
1904                 return 0;
1905         }
1906         if (!in_dev)
1907                 ret = -ENODEV;
1908         rtnl_unlock();
1909         return ret;
1910 }
1911
1912 int ip_mc_source(int add, int omode, struct sock *sk, struct
1913         ip_mreq_source *mreqs, int ifindex)
1914 {
1915         int err;
1916         struct ip_mreqn imr;
1917         __be32 addr = mreqs->imr_multiaddr;
1918         struct ip_mc_socklist *pmc;
1919         struct in_device *in_dev = NULL;
1920         struct inet_sock *inet = inet_sk(sk);
1921         struct ip_sf_socklist *psl;
1922         struct net *net = sock_net(sk);
1923         int leavegroup = 0;
1924         int i, j, rv;
1925
1926         if (!ipv4_is_multicast(addr))
1927                 return -EINVAL;
1928
1929         rtnl_lock();
1930
1931         imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
1932         imr.imr_address.s_addr = mreqs->imr_interface;
1933         imr.imr_ifindex = ifindex;
1934         in_dev = ip_mc_find_dev(net, &imr);
1935
1936         if (!in_dev) {
1937                 err = -ENODEV;
1938                 goto done;
1939         }
1940         err = -EADDRNOTAVAIL;
1941
1942         for_each_pmc_rtnl(inet, pmc) {
1943                 if ((pmc->multi.imr_multiaddr.s_addr ==
1944                      imr.imr_multiaddr.s_addr) &&
1945                     (pmc->multi.imr_ifindex == imr.imr_ifindex))
1946                         break;
1947         }
1948         if (!pmc) {             /* must have a prior join */
1949                 err = -EINVAL;
1950                 goto done;
1951         }
1952         /* if a source filter was set, must be the same mode as before */
1953         if (pmc->sflist) {
1954                 if (pmc->sfmode != omode) {
1955                         err = -EINVAL;
1956                         goto done;
1957                 }
1958         } else if (pmc->sfmode != omode) {
1959                 /* allow mode switches for empty-set filters */
1960                 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
1961                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
1962                         NULL, 0);
1963                 pmc->sfmode = omode;
1964         }
1965
1966         psl = rtnl_dereference(pmc->sflist);
1967         if (!add) {
1968                 if (!psl)
1969                         goto done;      /* err = -EADDRNOTAVAIL */
1970                 rv = !0;
1971                 for (i=0; i<psl->sl_count; i++) {
1972                         rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
1973                                 sizeof(__be32));
1974                         if (rv == 0)
1975                                 break;
1976                 }
1977                 if (rv)         /* source not found */
1978                         goto done;      /* err = -EADDRNOTAVAIL */
1979
1980                 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
1981                 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
1982                         leavegroup = 1;
1983                         goto done;
1984                 }
1985
1986                 /* update the interface filter */
1987                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
1988                         &mreqs->imr_sourceaddr, 1);
1989
1990                 for (j=i+1; j<psl->sl_count; j++)
1991                         psl->sl_addr[j-1] = psl->sl_addr[j];
1992                 psl->sl_count--;
1993                 err = 0;
1994                 goto done;
1995         }
1996         /* else, add a new source to the filter */
1997
1998         if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
1999                 err = -ENOBUFS;
2000                 goto done;
2001         }
2002         if (!psl || psl->sl_count == psl->sl_max) {
2003                 struct ip_sf_socklist *newpsl;
2004                 int count = IP_SFBLOCK;
2005
2006                 if (psl)
2007                         count += psl->sl_max;
2008                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2009                 if (!newpsl) {
2010                         err = -ENOBUFS;
2011                         goto done;
2012                 }
2013                 newpsl->sl_max = count;
2014                 newpsl->sl_count = count - IP_SFBLOCK;
2015                 if (psl) {
2016                         for (i=0; i<psl->sl_count; i++)
2017                                 newpsl->sl_addr[i] = psl->sl_addr[i];
2018                         /* decrease mem now to avoid the memleak warning */
2019                         atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2020                         call_rcu(&psl->rcu, ip_sf_socklist_reclaim);
2021                 }
2022                 rcu_assign_pointer(pmc->sflist, newpsl);
2023                 psl = newpsl;
2024         }
2025         rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2026         for (i=0; i<psl->sl_count; i++) {
2027                 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2028                         sizeof(__be32));
2029                 if (rv == 0)
2030                         break;
2031         }
2032         if (rv == 0)            /* address already there is an error */
2033                 goto done;
2034         for (j=psl->sl_count-1; j>=i; j--)
2035                 psl->sl_addr[j+1] = psl->sl_addr[j];
2036         psl->sl_addr[i] = mreqs->imr_sourceaddr;
2037         psl->sl_count++;
2038         err = 0;
2039         /* update the interface list */
2040         ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2041                 &mreqs->imr_sourceaddr, 1);
2042 done:
2043         rtnl_unlock();
2044         if (leavegroup)
2045                 return ip_mc_leave_group(sk, &imr);
2046         return err;
2047 }
2048
2049 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2050 {
2051         int err = 0;
2052         struct ip_mreqn imr;
2053         __be32 addr = msf->imsf_multiaddr;
2054         struct ip_mc_socklist *pmc;
2055         struct in_device *in_dev;
2056         struct inet_sock *inet = inet_sk(sk);
2057         struct ip_sf_socklist *newpsl, *psl;
2058         struct net *net = sock_net(sk);
2059         int leavegroup = 0;
2060
2061         if (!ipv4_is_multicast(addr))
2062                 return -EINVAL;
2063         if (msf->imsf_fmode != MCAST_INCLUDE &&
2064             msf->imsf_fmode != MCAST_EXCLUDE)
2065                 return -EINVAL;
2066
2067         rtnl_lock();
2068
2069         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2070         imr.imr_address.s_addr = msf->imsf_interface;
2071         imr.imr_ifindex = ifindex;
2072         in_dev = ip_mc_find_dev(net, &imr);
2073
2074         if (!in_dev) {
2075                 err = -ENODEV;
2076                 goto done;
2077         }
2078
2079         /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2080         if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2081                 leavegroup = 1;
2082                 goto done;
2083         }
2084
2085         for_each_pmc_rtnl(inet, pmc) {
2086                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2087                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2088                         break;
2089         }
2090         if (!pmc) {             /* must have a prior join */
2091                 err = -EINVAL;
2092                 goto done;
2093         }
2094         if (msf->imsf_numsrc) {
2095                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2096                                                            GFP_KERNEL);
2097                 if (!newpsl) {
2098                         err = -ENOBUFS;
2099                         goto done;
2100                 }
2101                 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2102                 memcpy(newpsl->sl_addr, msf->imsf_slist,
2103                         msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2104                 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2105                         msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2106                 if (err) {
2107                         sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2108                         goto done;
2109                 }
2110         } else {
2111                 newpsl = NULL;
2112                 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2113                                      msf->imsf_fmode, 0, NULL, 0);
2114         }
2115         psl = rtnl_dereference(pmc->sflist);
2116         if (psl) {
2117                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2118                         psl->sl_count, psl->sl_addr, 0);
2119                 /* decrease mem now to avoid the memleak warning */
2120                 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2121                 call_rcu(&psl->rcu, ip_sf_socklist_reclaim);
2122         } else
2123                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2124                         0, NULL, 0);
2125         rcu_assign_pointer(pmc->sflist, newpsl);
2126         pmc->sfmode = msf->imsf_fmode;
2127         err = 0;
2128 done:
2129         rtnl_unlock();
2130         if (leavegroup)
2131                 err = ip_mc_leave_group(sk, &imr);
2132         return err;
2133 }
2134
2135 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2136         struct ip_msfilter __user *optval, int __user *optlen)
2137 {
2138         int err, len, count, copycount;
2139         struct ip_mreqn imr;
2140         __be32 addr = msf->imsf_multiaddr;
2141         struct ip_mc_socklist *pmc;
2142         struct in_device *in_dev;
2143         struct inet_sock *inet = inet_sk(sk);
2144         struct ip_sf_socklist *psl;
2145         struct net *net = sock_net(sk);
2146
2147         if (!ipv4_is_multicast(addr))
2148                 return -EINVAL;
2149
2150         rtnl_lock();
2151
2152         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2153         imr.imr_address.s_addr = msf->imsf_interface;
2154         imr.imr_ifindex = 0;
2155         in_dev = ip_mc_find_dev(net, &imr);
2156
2157         if (!in_dev) {
2158                 err = -ENODEV;
2159                 goto done;
2160         }
2161         err = -EADDRNOTAVAIL;
2162
2163         for_each_pmc_rtnl(inet, pmc) {
2164                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2165                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2166                         break;
2167         }
2168         if (!pmc)               /* must have a prior join */
2169                 goto done;
2170         msf->imsf_fmode = pmc->sfmode;
2171         psl = rtnl_dereference(pmc->sflist);
2172         rtnl_unlock();
2173         if (!psl) {
2174                 len = 0;
2175                 count = 0;
2176         } else {
2177                 count = psl->sl_count;
2178         }
2179         copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2180         len = copycount * sizeof(psl->sl_addr[0]);
2181         msf->imsf_numsrc = count;
2182         if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2183             copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2184                 return -EFAULT;
2185         }
2186         if (len &&
2187             copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2188                 return -EFAULT;
2189         return 0;
2190 done:
2191         rtnl_unlock();
2192         return err;
2193 }
2194
2195 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2196         struct group_filter __user *optval, int __user *optlen)
2197 {
2198         int err, i, count, copycount;
2199         struct sockaddr_in *psin;
2200         __be32 addr;
2201         struct ip_mc_socklist *pmc;
2202         struct inet_sock *inet = inet_sk(sk);
2203         struct ip_sf_socklist *psl;
2204
2205         psin = (struct sockaddr_in *)&gsf->gf_group;
2206         if (psin->sin_family != AF_INET)
2207                 return -EINVAL;
2208         addr = psin->sin_addr.s_addr;
2209         if (!ipv4_is_multicast(addr))
2210                 return -EINVAL;
2211
2212         rtnl_lock();
2213
2214         err = -EADDRNOTAVAIL;
2215
2216         for_each_pmc_rtnl(inet, pmc) {
2217                 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2218                     pmc->multi.imr_ifindex == gsf->gf_interface)
2219                         break;
2220         }
2221         if (!pmc)               /* must have a prior join */
2222                 goto done;
2223         gsf->gf_fmode = pmc->sfmode;
2224         psl = rtnl_dereference(pmc->sflist);
2225         rtnl_unlock();
2226         count = psl ? psl->sl_count : 0;
2227         copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2228         gsf->gf_numsrc = count;
2229         if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2230             copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2231                 return -EFAULT;
2232         }
2233         for (i=0; i<copycount; i++) {
2234                 struct sockaddr_storage ss;
2235
2236                 psin = (struct sockaddr_in *)&ss;
2237                 memset(&ss, 0, sizeof(ss));
2238                 psin->sin_family = AF_INET;
2239                 psin->sin_addr.s_addr = psl->sl_addr[i];
2240                 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2241                         return -EFAULT;
2242         }
2243         return 0;
2244 done:
2245         rtnl_unlock();
2246         return err;
2247 }
2248
2249 /*
2250  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2251  */
2252 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2253 {
2254         struct inet_sock *inet = inet_sk(sk);
2255         struct ip_mc_socklist *pmc;
2256         struct ip_sf_socklist *psl;
2257         int i;
2258         int ret;
2259
2260         ret = 1;
2261         if (!ipv4_is_multicast(loc_addr))
2262                 goto out;
2263
2264         rcu_read_lock();
2265         for_each_pmc_rcu(inet, pmc) {
2266                 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2267                     pmc->multi.imr_ifindex == dif)
2268                         break;
2269         }
2270         ret = inet->mc_all;
2271         if (!pmc)
2272                 goto unlock;
2273         psl = rcu_dereference(pmc->sflist);
2274         ret = (pmc->sfmode == MCAST_EXCLUDE);
2275         if (!psl)
2276                 goto unlock;
2277
2278         for (i=0; i<psl->sl_count; i++) {
2279                 if (psl->sl_addr[i] == rmt_addr)
2280                         break;
2281         }
2282         ret = 0;
2283         if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2284                 goto unlock;
2285         if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2286                 goto unlock;
2287         ret = 1;
2288 unlock:
2289         rcu_read_unlock();
2290 out:
2291         return ret;
2292 }
2293
2294 /*
2295  *      A socket is closing.
2296  */
2297
2298 void ip_mc_drop_socket(struct sock *sk)
2299 {
2300         struct inet_sock *inet = inet_sk(sk);
2301         struct ip_mc_socklist *iml;
2302         struct net *net = sock_net(sk);
2303
2304         if (inet->mc_list == NULL)
2305                 return;
2306
2307         rtnl_lock();
2308         while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2309                 struct in_device *in_dev;
2310
2311                 inet->mc_list = iml->next_rcu;
2312                 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2313                 (void) ip_mc_leave_src(sk, iml, in_dev);
2314                 if (in_dev != NULL)
2315                         ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2316                 /* decrease mem now to avoid the memleak warning */
2317                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2318                 call_rcu(&iml->rcu, ip_mc_socklist_reclaim);
2319         }
2320         rtnl_unlock();
2321 }
2322
2323 int ip_check_mc(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2324 {
2325         struct ip_mc_list *im;
2326         struct ip_sf_list *psf;
2327         int rv = 0;
2328
2329         rcu_read_lock();
2330         for_each_pmc_rcu(in_dev, im) {
2331                 if (im->multiaddr == mc_addr)
2332                         break;
2333         }
2334         if (im && proto == IPPROTO_IGMP) {
2335                 rv = 1;
2336         } else if (im) {
2337                 if (src_addr) {
2338                         for (psf=im->sources; psf; psf=psf->sf_next) {
2339                                 if (psf->sf_inaddr == src_addr)
2340                                         break;
2341                         }
2342                         if (psf)
2343                                 rv = psf->sf_count[MCAST_INCLUDE] ||
2344                                         psf->sf_count[MCAST_EXCLUDE] !=
2345                                         im->sfcount[MCAST_EXCLUDE];
2346                         else
2347                                 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2348                 } else
2349                         rv = 1; /* unspecified source; tentatively allow */
2350         }
2351         rcu_read_unlock();
2352         return rv;
2353 }
2354
2355 #if defined(CONFIG_PROC_FS)
2356 struct igmp_mc_iter_state {
2357         struct seq_net_private p;
2358         struct net_device *dev;
2359         struct in_device *in_dev;
2360 };
2361
2362 #define igmp_mc_seq_private(seq)        ((struct igmp_mc_iter_state *)(seq)->private)
2363
2364 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2365 {
2366         struct net *net = seq_file_net(seq);
2367         struct ip_mc_list *im = NULL;
2368         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2369
2370         state->in_dev = NULL;
2371         for_each_netdev_rcu(net, state->dev) {
2372                 struct in_device *in_dev;
2373
2374                 in_dev = __in_dev_get_rcu(state->dev);
2375                 if (!in_dev)
2376                         continue;
2377                 im = rcu_dereference(in_dev->mc_list);
2378                 if (im) {
2379                         state->in_dev = in_dev;
2380                         break;
2381                 }
2382         }
2383         return im;
2384 }
2385
2386 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2387 {
2388         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2389
2390         im = rcu_dereference(im->next_rcu);
2391         while (!im) {
2392                 state->dev = next_net_device_rcu(state->dev);
2393                 if (!state->dev) {
2394                         state->in_dev = NULL;
2395                         break;
2396                 }
2397                 state->in_dev = __in_dev_get_rcu(state->dev);
2398                 if (!state->in_dev)
2399                         continue;
2400                 im = rcu_dereference(state->in_dev->mc_list);
2401         }
2402         return im;
2403 }
2404
2405 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2406 {
2407         struct ip_mc_list *im = igmp_mc_get_first(seq);
2408         if (im)
2409                 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2410                         --pos;
2411         return pos ? NULL : im;
2412 }
2413
2414 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2415         __acquires(rcu)
2416 {
2417         rcu_read_lock();
2418         return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2419 }
2420
2421 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2422 {
2423         struct ip_mc_list *im;
2424         if (v == SEQ_START_TOKEN)
2425                 im = igmp_mc_get_first(seq);
2426         else
2427                 im = igmp_mc_get_next(seq, v);
2428         ++*pos;
2429         return im;
2430 }
2431
2432 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2433         __releases(rcu)
2434 {
2435         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2436
2437         state->in_dev = NULL;
2438         state->dev = NULL;
2439         rcu_read_unlock();
2440 }
2441
2442 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2443 {
2444         if (v == SEQ_START_TOKEN)
2445                 seq_puts(seq,
2446                          "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2447         else {
2448                 struct ip_mc_list *im = (struct ip_mc_list *)v;
2449                 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2450                 char   *querier;
2451 #ifdef CONFIG_IP_MULTICAST
2452                 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2453                           IGMP_V2_SEEN(state->in_dev) ? "V2" :
2454                           "V3";
2455 #else
2456                 querier = "NONE";
2457 #endif
2458
2459                 if (rcu_dereference(state->in_dev->mc_list) == im) {
2460                         seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2461                                    state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2462                 }
2463
2464                 seq_printf(seq,
2465                            "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2466                            im->multiaddr, im->users,
2467                            im->tm_running, im->tm_running ?
2468                            jiffies_to_clock_t(im->timer.expires-jiffies) : 0,
2469                            im->reporter);
2470         }
2471         return 0;
2472 }
2473
2474 static const struct seq_operations igmp_mc_seq_ops = {
2475         .start  =       igmp_mc_seq_start,
2476         .next   =       igmp_mc_seq_next,
2477         .stop   =       igmp_mc_seq_stop,
2478         .show   =       igmp_mc_seq_show,
2479 };
2480
2481 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2482 {
2483         return seq_open_net(inode, file, &igmp_mc_seq_ops,
2484                         sizeof(struct igmp_mc_iter_state));
2485 }
2486
2487 static const struct file_operations igmp_mc_seq_fops = {
2488         .owner          =       THIS_MODULE,
2489         .open           =       igmp_mc_seq_open,
2490         .read           =       seq_read,
2491         .llseek         =       seq_lseek,
2492         .release        =       seq_release_net,
2493 };
2494
2495 struct igmp_mcf_iter_state {
2496         struct seq_net_private p;
2497         struct net_device *dev;
2498         struct in_device *idev;
2499         struct ip_mc_list *im;
2500 };
2501
2502 #define igmp_mcf_seq_private(seq)       ((struct igmp_mcf_iter_state *)(seq)->private)
2503
2504 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2505 {
2506         struct net *net = seq_file_net(seq);
2507         struct ip_sf_list *psf = NULL;
2508         struct ip_mc_list *im = NULL;
2509         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2510
2511         state->idev = NULL;
2512         state->im = NULL;
2513         for_each_netdev_rcu(net, state->dev) {
2514                 struct in_device *idev;
2515                 idev = __in_dev_get_rcu(state->dev);
2516                 if (unlikely(idev == NULL))
2517                         continue;
2518                 im = rcu_dereference(idev->mc_list);
2519                 if (likely(im != NULL)) {
2520                         spin_lock_bh(&im->lock);
2521                         psf = im->sources;
2522                         if (likely(psf != NULL)) {
2523                                 state->im = im;
2524                                 state->idev = idev;
2525                                 break;
2526                         }
2527                         spin_unlock_bh(&im->lock);
2528                 }
2529         }
2530         return psf;
2531 }
2532
2533 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2534 {
2535         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2536
2537         psf = psf->sf_next;
2538         while (!psf) {
2539                 spin_unlock_bh(&state->im->lock);
2540                 state->im = state->im->next;
2541                 while (!state->im) {
2542                         state->dev = next_net_device_rcu(state->dev);
2543                         if (!state->dev) {
2544                                 state->idev = NULL;
2545                                 goto out;
2546                         }
2547                         state->idev = __in_dev_get_rcu(state->dev);
2548                         if (!state->idev)
2549                                 continue;
2550                         state->im = rcu_dereference(state->idev->mc_list);
2551                 }
2552                 if (!state->im)
2553                         break;
2554                 spin_lock_bh(&state->im->lock);
2555                 psf = state->im->sources;
2556         }
2557 out:
2558         return psf;
2559 }
2560
2561 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2562 {
2563         struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2564         if (psf)
2565                 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2566                         --pos;
2567         return pos ? NULL : psf;
2568 }
2569
2570 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2571         __acquires(rcu)
2572 {
2573         rcu_read_lock();
2574         return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2575 }
2576
2577 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2578 {
2579         struct ip_sf_list *psf;
2580         if (v == SEQ_START_TOKEN)
2581                 psf = igmp_mcf_get_first(seq);
2582         else
2583                 psf = igmp_mcf_get_next(seq, v);
2584         ++*pos;
2585         return psf;
2586 }
2587
2588 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2589         __releases(rcu)
2590 {
2591         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2592         if (likely(state->im != NULL)) {
2593                 spin_unlock_bh(&state->im->lock);
2594                 state->im = NULL;
2595         }
2596         state->idev = NULL;
2597         state->dev = NULL;
2598         rcu_read_unlock();
2599 }
2600
2601 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2602 {
2603         struct ip_sf_list *psf = (struct ip_sf_list *)v;
2604         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2605
2606         if (v == SEQ_START_TOKEN) {
2607                 seq_printf(seq,
2608                            "%3s %6s "
2609                            "%10s %10s %6s %6s\n", "Idx",
2610                            "Device", "MCA",
2611                            "SRC", "INC", "EXC");
2612         } else {
2613                 seq_printf(seq,
2614                            "%3d %6.6s 0x%08x "
2615                            "0x%08x %6lu %6lu\n",
2616                            state->dev->ifindex, state->dev->name,
2617                            ntohl(state->im->multiaddr),
2618                            ntohl(psf->sf_inaddr),
2619                            psf->sf_count[MCAST_INCLUDE],
2620                            psf->sf_count[MCAST_EXCLUDE]);
2621         }
2622         return 0;
2623 }
2624
2625 static const struct seq_operations igmp_mcf_seq_ops = {
2626         .start  =       igmp_mcf_seq_start,
2627         .next   =       igmp_mcf_seq_next,
2628         .stop   =       igmp_mcf_seq_stop,
2629         .show   =       igmp_mcf_seq_show,
2630 };
2631
2632 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2633 {
2634         return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2635                         sizeof(struct igmp_mcf_iter_state));
2636 }
2637
2638 static const struct file_operations igmp_mcf_seq_fops = {
2639         .owner          =       THIS_MODULE,
2640         .open           =       igmp_mcf_seq_open,
2641         .read           =       seq_read,
2642         .llseek         =       seq_lseek,
2643         .release        =       seq_release_net,
2644 };
2645
2646 static int __net_init igmp_net_init(struct net *net)
2647 {
2648         struct proc_dir_entry *pde;
2649
2650         pde = proc_net_fops_create(net, "igmp", S_IRUGO, &igmp_mc_seq_fops);
2651         if (!pde)
2652                 goto out_igmp;
2653         pde = proc_net_fops_create(net, "mcfilter", S_IRUGO, &igmp_mcf_seq_fops);
2654         if (!pde)
2655                 goto out_mcfilter;
2656         return 0;
2657
2658 out_mcfilter:
2659         proc_net_remove(net, "igmp");
2660 out_igmp:
2661         return -ENOMEM;
2662 }
2663
2664 static void __net_exit igmp_net_exit(struct net *net)
2665 {
2666         proc_net_remove(net, "mcfilter");
2667         proc_net_remove(net, "igmp");
2668 }
2669
2670 static struct pernet_operations igmp_net_ops = {
2671         .init = igmp_net_init,
2672         .exit = igmp_net_exit,
2673 };
2674
2675 int __init igmp_mc_proc_init(void)
2676 {
2677         return register_pernet_subsys(&igmp_net_ops);
2678 }
2679 #endif