igmp: Avoid zero delay when receiving odd mixture of IGMP queries
[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 static void ip_ma_put(struct ip_mc_list *im)
153 {
154         if (atomic_dec_and_test(&im->refcnt)) {
155                 in_dev_put(im->interface);
156                 kfree_rcu(im, rcu);
157         }
158 }
159
160 #define for_each_pmc_rcu(in_dev, pmc)                           \
161         for (pmc = rcu_dereference(in_dev->mc_list);            \
162              pmc != NULL;                                       \
163              pmc = rcu_dereference(pmc->next_rcu))
164
165 #define for_each_pmc_rtnl(in_dev, pmc)                          \
166         for (pmc = rtnl_dereference(in_dev->mc_list);           \
167              pmc != NULL;                                       \
168              pmc = rtnl_dereference(pmc->next_rcu))
169
170 #ifdef CONFIG_IP_MULTICAST
171
172 /*
173  *      Timer management
174  */
175
176 static void igmp_stop_timer(struct ip_mc_list *im)
177 {
178         spin_lock_bh(&im->lock);
179         if (del_timer(&im->timer))
180                 atomic_dec(&im->refcnt);
181         im->tm_running = 0;
182         im->reporter = 0;
183         im->unsolicit_count = 0;
184         spin_unlock_bh(&im->lock);
185 }
186
187 /* It must be called with locked im->lock */
188 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
189 {
190         int tv = net_random() % max_delay;
191
192         im->tm_running = 1;
193         if (!mod_timer(&im->timer, jiffies+tv+2))
194                 atomic_inc(&im->refcnt);
195 }
196
197 static void igmp_gq_start_timer(struct in_device *in_dev)
198 {
199         int tv = net_random() % in_dev->mr_maxdelay;
200
201         in_dev->mr_gq_running = 1;
202         if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
203                 in_dev_hold(in_dev);
204 }
205
206 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
207 {
208         int tv = net_random() % delay;
209
210         if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
211                 in_dev_hold(in_dev);
212 }
213
214 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
215 {
216         spin_lock_bh(&im->lock);
217         im->unsolicit_count = 0;
218         if (del_timer(&im->timer)) {
219                 if ((long)(im->timer.expires-jiffies) < max_delay) {
220                         add_timer(&im->timer);
221                         im->tm_running = 1;
222                         spin_unlock_bh(&im->lock);
223                         return;
224                 }
225                 atomic_dec(&im->refcnt);
226         }
227         igmp_start_timer(im, max_delay);
228         spin_unlock_bh(&im->lock);
229 }
230
231
232 /*
233  *      Send an IGMP report.
234  */
235
236 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
237
238
239 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
240         int gdeleted, int sdeleted)
241 {
242         switch (type) {
243         case IGMPV3_MODE_IS_INCLUDE:
244         case IGMPV3_MODE_IS_EXCLUDE:
245                 if (gdeleted || sdeleted)
246                         return 0;
247                 if (!(pmc->gsquery && !psf->sf_gsresp)) {
248                         if (pmc->sfmode == MCAST_INCLUDE)
249                                 return 1;
250                         /* don't include if this source is excluded
251                          * in all filters
252                          */
253                         if (psf->sf_count[MCAST_INCLUDE])
254                                 return type == IGMPV3_MODE_IS_INCLUDE;
255                         return pmc->sfcount[MCAST_EXCLUDE] ==
256                                 psf->sf_count[MCAST_EXCLUDE];
257                 }
258                 return 0;
259         case IGMPV3_CHANGE_TO_INCLUDE:
260                 if (gdeleted || sdeleted)
261                         return 0;
262                 return psf->sf_count[MCAST_INCLUDE] != 0;
263         case IGMPV3_CHANGE_TO_EXCLUDE:
264                 if (gdeleted || sdeleted)
265                         return 0;
266                 if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
267                     psf->sf_count[MCAST_INCLUDE])
268                         return 0;
269                 return pmc->sfcount[MCAST_EXCLUDE] ==
270                         psf->sf_count[MCAST_EXCLUDE];
271         case IGMPV3_ALLOW_NEW_SOURCES:
272                 if (gdeleted || !psf->sf_crcount)
273                         return 0;
274                 return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
275         case IGMPV3_BLOCK_OLD_SOURCES:
276                 if (pmc->sfmode == MCAST_INCLUDE)
277                         return gdeleted || (psf->sf_crcount && sdeleted);
278                 return psf->sf_crcount && !gdeleted && !sdeleted;
279         }
280         return 0;
281 }
282
283 static int
284 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
285 {
286         struct ip_sf_list *psf;
287         int scount = 0;
288
289         for (psf=pmc->sources; psf; psf=psf->sf_next) {
290                 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
291                         continue;
292                 scount++;
293         }
294         return scount;
295 }
296
297 #define igmp_skb_size(skb) (*(unsigned int *)((skb)->cb))
298
299 static struct sk_buff *igmpv3_newpack(struct net_device *dev, int size)
300 {
301         struct sk_buff *skb;
302         struct rtable *rt;
303         struct iphdr *pip;
304         struct igmpv3_report *pig;
305         struct net *net = dev_net(dev);
306         struct flowi4 fl4;
307
308         while (1) {
309                 skb = alloc_skb(size + LL_ALLOCATED_SPACE(dev),
310                                 GFP_ATOMIC | __GFP_NOWARN);
311                 if (skb)
312                         break;
313                 size >>= 1;
314                 if (size < 256)
315                         return NULL;
316         }
317         igmp_skb_size(skb) = size;
318
319         rt = ip_route_output_ports(net, &fl4, NULL, IGMPV3_ALL_MCR, 0,
320                                    0, 0,
321                                    IPPROTO_IGMP, 0, dev->ifindex);
322         if (IS_ERR(rt)) {
323                 kfree_skb(skb);
324                 return NULL;
325         }
326
327         skb_dst_set(skb, &rt->dst);
328         skb->dev = dev;
329
330         skb_reserve(skb, LL_RESERVED_SPACE(dev));
331
332         skb_reset_network_header(skb);
333         pip = ip_hdr(skb);
334         skb_put(skb, sizeof(struct iphdr) + 4);
335
336         pip->version  = 4;
337         pip->ihl      = (sizeof(struct iphdr)+4)>>2;
338         pip->tos      = 0xc0;
339         pip->frag_off = htons(IP_DF);
340         pip->ttl      = 1;
341         pip->daddr    = fl4.daddr;
342         pip->saddr    = fl4.saddr;
343         pip->protocol = IPPROTO_IGMP;
344         pip->tot_len  = 0;      /* filled in later */
345         ip_select_ident(pip, &rt->dst, NULL);
346         ((u8*)&pip[1])[0] = IPOPT_RA;
347         ((u8*)&pip[1])[1] = 4;
348         ((u8*)&pip[1])[2] = 0;
349         ((u8*)&pip[1])[3] = 0;
350
351         skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
352         skb_put(skb, sizeof(*pig));
353         pig = igmpv3_report_hdr(skb);
354         pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
355         pig->resv1 = 0;
356         pig->csum = 0;
357         pig->resv2 = 0;
358         pig->ngrec = 0;
359         return skb;
360 }
361
362 static int igmpv3_sendpack(struct sk_buff *skb)
363 {
364         struct igmphdr *pig = igmp_hdr(skb);
365         const int igmplen = skb->tail - skb->transport_header;
366
367         pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
368
369         return ip_local_out(skb);
370 }
371
372 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
373 {
374         return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
375 }
376
377 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
378         int type, struct igmpv3_grec **ppgr)
379 {
380         struct net_device *dev = pmc->interface->dev;
381         struct igmpv3_report *pih;
382         struct igmpv3_grec *pgr;
383
384         if (!skb)
385                 skb = igmpv3_newpack(dev, dev->mtu);
386         if (!skb)
387                 return NULL;
388         pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
389         pgr->grec_type = type;
390         pgr->grec_auxwords = 0;
391         pgr->grec_nsrcs = 0;
392         pgr->grec_mca = pmc->multiaddr;
393         pih = igmpv3_report_hdr(skb);
394         pih->ngrec = htons(ntohs(pih->ngrec)+1);
395         *ppgr = pgr;
396         return skb;
397 }
398
399 #define AVAILABLE(skb) ((skb) ? ((skb)->dev ? igmp_skb_size(skb) - (skb)->len : \
400         skb_tailroom(skb)) : 0)
401
402 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
403         int type, int gdeleted, int sdeleted)
404 {
405         struct net_device *dev = pmc->interface->dev;
406         struct igmpv3_report *pih;
407         struct igmpv3_grec *pgr = NULL;
408         struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
409         int scount, stotal, first, isquery, truncate;
410
411         if (pmc->multiaddr == IGMP_ALL_HOSTS)
412                 return skb;
413
414         isquery = type == IGMPV3_MODE_IS_INCLUDE ||
415                   type == IGMPV3_MODE_IS_EXCLUDE;
416         truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
417                     type == IGMPV3_CHANGE_TO_EXCLUDE;
418
419         stotal = scount = 0;
420
421         psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
422
423         if (!*psf_list)
424                 goto empty_source;
425
426         pih = skb ? igmpv3_report_hdr(skb) : NULL;
427
428         /* EX and TO_EX get a fresh packet, if needed */
429         if (truncate) {
430                 if (pih && pih->ngrec &&
431                     AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
432                         if (skb)
433                                 igmpv3_sendpack(skb);
434                         skb = igmpv3_newpack(dev, dev->mtu);
435                 }
436         }
437         first = 1;
438         psf_prev = NULL;
439         for (psf=*psf_list; psf; psf=psf_next) {
440                 __be32 *psrc;
441
442                 psf_next = psf->sf_next;
443
444                 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
445                         psf_prev = psf;
446                         continue;
447                 }
448
449                 /* clear marks on query responses */
450                 if (isquery)
451                         psf->sf_gsresp = 0;
452
453                 if (AVAILABLE(skb) < sizeof(__be32) +
454                     first*sizeof(struct igmpv3_grec)) {
455                         if (truncate && !first)
456                                 break;   /* truncate these */
457                         if (pgr)
458                                 pgr->grec_nsrcs = htons(scount);
459                         if (skb)
460                                 igmpv3_sendpack(skb);
461                         skb = igmpv3_newpack(dev, dev->mtu);
462                         first = 1;
463                         scount = 0;
464                 }
465                 if (first) {
466                         skb = add_grhead(skb, pmc, type, &pgr);
467                         first = 0;
468                 }
469                 if (!skb)
470                         return NULL;
471                 psrc = (__be32 *)skb_put(skb, sizeof(__be32));
472                 *psrc = psf->sf_inaddr;
473                 scount++; stotal++;
474                 if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
475                      type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
476                         psf->sf_crcount--;
477                         if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
478                                 if (psf_prev)
479                                         psf_prev->sf_next = psf->sf_next;
480                                 else
481                                         *psf_list = psf->sf_next;
482                                 kfree(psf);
483                                 continue;
484                         }
485                 }
486                 psf_prev = psf;
487         }
488
489 empty_source:
490         if (!stotal) {
491                 if (type == IGMPV3_ALLOW_NEW_SOURCES ||
492                     type == IGMPV3_BLOCK_OLD_SOURCES)
493                         return skb;
494                 if (pmc->crcount || isquery) {
495                         /* make sure we have room for group header */
496                         if (skb && AVAILABLE(skb)<sizeof(struct igmpv3_grec)) {
497                                 igmpv3_sendpack(skb);
498                                 skb = NULL; /* add_grhead will get a new one */
499                         }
500                         skb = add_grhead(skb, pmc, type, &pgr);
501                 }
502         }
503         if (pgr)
504                 pgr->grec_nsrcs = htons(scount);
505
506         if (isquery)
507                 pmc->gsquery = 0;       /* clear query state on report */
508         return skb;
509 }
510
511 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
512 {
513         struct sk_buff *skb = NULL;
514         int type;
515
516         if (!pmc) {
517                 rcu_read_lock();
518                 for_each_pmc_rcu(in_dev, pmc) {
519                         if (pmc->multiaddr == IGMP_ALL_HOSTS)
520                                 continue;
521                         spin_lock_bh(&pmc->lock);
522                         if (pmc->sfcount[MCAST_EXCLUDE])
523                                 type = IGMPV3_MODE_IS_EXCLUDE;
524                         else
525                                 type = IGMPV3_MODE_IS_INCLUDE;
526                         skb = add_grec(skb, pmc, type, 0, 0);
527                         spin_unlock_bh(&pmc->lock);
528                 }
529                 rcu_read_unlock();
530         } else {
531                 spin_lock_bh(&pmc->lock);
532                 if (pmc->sfcount[MCAST_EXCLUDE])
533                         type = IGMPV3_MODE_IS_EXCLUDE;
534                 else
535                         type = IGMPV3_MODE_IS_INCLUDE;
536                 skb = add_grec(skb, pmc, type, 0, 0);
537                 spin_unlock_bh(&pmc->lock);
538         }
539         if (!skb)
540                 return 0;
541         return igmpv3_sendpack(skb);
542 }
543
544 /*
545  * remove zero-count source records from a source filter list
546  */
547 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
548 {
549         struct ip_sf_list *psf_prev, *psf_next, *psf;
550
551         psf_prev = NULL;
552         for (psf=*ppsf; psf; psf = psf_next) {
553                 psf_next = psf->sf_next;
554                 if (psf->sf_crcount == 0) {
555                         if (psf_prev)
556                                 psf_prev->sf_next = psf->sf_next;
557                         else
558                                 *ppsf = psf->sf_next;
559                         kfree(psf);
560                 } else
561                         psf_prev = psf;
562         }
563 }
564
565 static void igmpv3_send_cr(struct in_device *in_dev)
566 {
567         struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
568         struct sk_buff *skb = NULL;
569         int type, dtype;
570
571         rcu_read_lock();
572         spin_lock_bh(&in_dev->mc_tomb_lock);
573
574         /* deleted MCA's */
575         pmc_prev = NULL;
576         for (pmc=in_dev->mc_tomb; pmc; pmc=pmc_next) {
577                 pmc_next = pmc->next;
578                 if (pmc->sfmode == MCAST_INCLUDE) {
579                         type = IGMPV3_BLOCK_OLD_SOURCES;
580                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
581                         skb = add_grec(skb, pmc, type, 1, 0);
582                         skb = add_grec(skb, pmc, dtype, 1, 1);
583                 }
584                 if (pmc->crcount) {
585                         if (pmc->sfmode == MCAST_EXCLUDE) {
586                                 type = IGMPV3_CHANGE_TO_INCLUDE;
587                                 skb = add_grec(skb, pmc, type, 1, 0);
588                         }
589                         pmc->crcount--;
590                         if (pmc->crcount == 0) {
591                                 igmpv3_clear_zeros(&pmc->tomb);
592                                 igmpv3_clear_zeros(&pmc->sources);
593                         }
594                 }
595                 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
596                         if (pmc_prev)
597                                 pmc_prev->next = pmc_next;
598                         else
599                                 in_dev->mc_tomb = pmc_next;
600                         in_dev_put(pmc->interface);
601                         kfree(pmc);
602                 } else
603                         pmc_prev = pmc;
604         }
605         spin_unlock_bh(&in_dev->mc_tomb_lock);
606
607         /* change recs */
608         for_each_pmc_rcu(in_dev, pmc) {
609                 spin_lock_bh(&pmc->lock);
610                 if (pmc->sfcount[MCAST_EXCLUDE]) {
611                         type = IGMPV3_BLOCK_OLD_SOURCES;
612                         dtype = IGMPV3_ALLOW_NEW_SOURCES;
613                 } else {
614                         type = IGMPV3_ALLOW_NEW_SOURCES;
615                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
616                 }
617                 skb = add_grec(skb, pmc, type, 0, 0);
618                 skb = add_grec(skb, pmc, dtype, 0, 1);  /* deleted sources */
619
620                 /* filter mode changes */
621                 if (pmc->crcount) {
622                         if (pmc->sfmode == MCAST_EXCLUDE)
623                                 type = IGMPV3_CHANGE_TO_EXCLUDE;
624                         else
625                                 type = IGMPV3_CHANGE_TO_INCLUDE;
626                         skb = add_grec(skb, pmc, type, 0, 0);
627                         pmc->crcount--;
628                 }
629                 spin_unlock_bh(&pmc->lock);
630         }
631         rcu_read_unlock();
632
633         if (!skb)
634                 return;
635         (void) igmpv3_sendpack(skb);
636 }
637
638 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
639         int type)
640 {
641         struct sk_buff *skb;
642         struct iphdr *iph;
643         struct igmphdr *ih;
644         struct rtable *rt;
645         struct net_device *dev = in_dev->dev;
646         struct net *net = dev_net(dev);
647         __be32  group = pmc ? pmc->multiaddr : 0;
648         struct flowi4 fl4;
649         __be32  dst;
650
651         if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
652                 return igmpv3_send_report(in_dev, pmc);
653         else if (type == IGMP_HOST_LEAVE_MESSAGE)
654                 dst = IGMP_ALL_ROUTER;
655         else
656                 dst = group;
657
658         rt = ip_route_output_ports(net, &fl4, NULL, dst, 0,
659                                    0, 0,
660                                    IPPROTO_IGMP, 0, dev->ifindex);
661         if (IS_ERR(rt))
662                 return -1;
663
664         skb = alloc_skb(IGMP_SIZE+LL_ALLOCATED_SPACE(dev), GFP_ATOMIC);
665         if (skb == NULL) {
666                 ip_rt_put(rt);
667                 return -1;
668         }
669
670         skb_dst_set(skb, &rt->dst);
671
672         skb_reserve(skb, LL_RESERVED_SPACE(dev));
673
674         skb_reset_network_header(skb);
675         iph = ip_hdr(skb);
676         skb_put(skb, sizeof(struct iphdr) + 4);
677
678         iph->version  = 4;
679         iph->ihl      = (sizeof(struct iphdr)+4)>>2;
680         iph->tos      = 0xc0;
681         iph->frag_off = htons(IP_DF);
682         iph->ttl      = 1;
683         iph->daddr    = dst;
684         iph->saddr    = fl4.saddr;
685         iph->protocol = IPPROTO_IGMP;
686         ip_select_ident(iph, &rt->dst, NULL);
687         ((u8*)&iph[1])[0] = IPOPT_RA;
688         ((u8*)&iph[1])[1] = 4;
689         ((u8*)&iph[1])[2] = 0;
690         ((u8*)&iph[1])[3] = 0;
691
692         ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
693         ih->type = type;
694         ih->code = 0;
695         ih->csum = 0;
696         ih->group = group;
697         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
698
699         return ip_local_out(skb);
700 }
701
702 static void igmp_gq_timer_expire(unsigned long data)
703 {
704         struct in_device *in_dev = (struct in_device *)data;
705
706         in_dev->mr_gq_running = 0;
707         igmpv3_send_report(in_dev, NULL);
708         __in_dev_put(in_dev);
709 }
710
711 static void igmp_ifc_timer_expire(unsigned long data)
712 {
713         struct in_device *in_dev = (struct in_device *)data;
714
715         igmpv3_send_cr(in_dev);
716         if (in_dev->mr_ifc_count) {
717                 in_dev->mr_ifc_count--;
718                 igmp_ifc_start_timer(in_dev, IGMP_Unsolicited_Report_Interval);
719         }
720         __in_dev_put(in_dev);
721 }
722
723 static void igmp_ifc_event(struct in_device *in_dev)
724 {
725         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
726                 return;
727         in_dev->mr_ifc_count = in_dev->mr_qrv ? in_dev->mr_qrv :
728                 IGMP_Unsolicited_Report_Count;
729         igmp_ifc_start_timer(in_dev, 1);
730 }
731
732
733 static void igmp_timer_expire(unsigned long data)
734 {
735         struct ip_mc_list *im=(struct ip_mc_list *)data;
736         struct in_device *in_dev = im->interface;
737
738         spin_lock(&im->lock);
739         im->tm_running = 0;
740
741         if (im->unsolicit_count) {
742                 im->unsolicit_count--;
743                 igmp_start_timer(im, IGMP_Unsolicited_Report_Interval);
744         }
745         im->reporter = 1;
746         spin_unlock(&im->lock);
747
748         if (IGMP_V1_SEEN(in_dev))
749                 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
750         else if (IGMP_V2_SEEN(in_dev))
751                 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
752         else
753                 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
754
755         ip_ma_put(im);
756 }
757
758 /* mark EXCLUDE-mode sources */
759 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
760 {
761         struct ip_sf_list *psf;
762         int i, scount;
763
764         scount = 0;
765         for (psf=pmc->sources; psf; psf=psf->sf_next) {
766                 if (scount == nsrcs)
767                         break;
768                 for (i=0; i<nsrcs; i++) {
769                         /* skip inactive filters */
770                         if (psf->sf_count[MCAST_INCLUDE] ||
771                             pmc->sfcount[MCAST_EXCLUDE] !=
772                             psf->sf_count[MCAST_EXCLUDE])
773                                 continue;
774                         if (srcs[i] == psf->sf_inaddr) {
775                                 scount++;
776                                 break;
777                         }
778                 }
779         }
780         pmc->gsquery = 0;
781         if (scount == nsrcs)    /* all sources excluded */
782                 return 0;
783         return 1;
784 }
785
786 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
787 {
788         struct ip_sf_list *psf;
789         int i, scount;
790
791         if (pmc->sfmode == MCAST_EXCLUDE)
792                 return igmp_xmarksources(pmc, nsrcs, srcs);
793
794         /* mark INCLUDE-mode sources */
795         scount = 0;
796         for (psf=pmc->sources; psf; psf=psf->sf_next) {
797                 if (scount == nsrcs)
798                         break;
799                 for (i=0; i<nsrcs; i++)
800                         if (srcs[i] == psf->sf_inaddr) {
801                                 psf->sf_gsresp = 1;
802                                 scount++;
803                                 break;
804                         }
805         }
806         if (!scount) {
807                 pmc->gsquery = 0;
808                 return 0;
809         }
810         pmc->gsquery = 1;
811         return 1;
812 }
813
814 static void igmp_heard_report(struct in_device *in_dev, __be32 group)
815 {
816         struct ip_mc_list *im;
817
818         /* Timers are only set for non-local groups */
819
820         if (group == IGMP_ALL_HOSTS)
821                 return;
822
823         rcu_read_lock();
824         for_each_pmc_rcu(in_dev, im) {
825                 if (im->multiaddr == group) {
826                         igmp_stop_timer(im);
827                         break;
828                 }
829         }
830         rcu_read_unlock();
831 }
832
833 static void igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
834         int len)
835 {
836         struct igmphdr          *ih = igmp_hdr(skb);
837         struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
838         struct ip_mc_list       *im;
839         __be32                  group = ih->group;
840         int                     max_delay;
841         int                     mark = 0;
842
843
844         if (len == 8) {
845                 if (ih->code == 0) {
846                         /* Alas, old v1 router presents here. */
847
848                         max_delay = IGMP_Query_Response_Interval;
849                         in_dev->mr_v1_seen = jiffies +
850                                 IGMP_V1_Router_Present_Timeout;
851                         group = 0;
852                 } else {
853                         /* v2 router present */
854                         max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
855                         in_dev->mr_v2_seen = jiffies +
856                                 IGMP_V2_Router_Present_Timeout;
857                 }
858                 /* cancel the interface change timer */
859                 in_dev->mr_ifc_count = 0;
860                 if (del_timer(&in_dev->mr_ifc_timer))
861                         __in_dev_put(in_dev);
862                 /* clear deleted report items */
863                 igmpv3_clear_delrec(in_dev);
864         } else if (len < 12) {
865                 return; /* ignore bogus packet; freed by caller */
866         } else if (IGMP_V1_SEEN(in_dev)) {
867                 /* This is a v3 query with v1 queriers present */
868                 max_delay = IGMP_Query_Response_Interval;
869                 group = 0;
870         } else if (IGMP_V2_SEEN(in_dev)) {
871                 /* this is a v3 query with v2 queriers present;
872                  * Interpretation of the max_delay code is problematic here.
873                  * A real v2 host would use ih_code directly, while v3 has a
874                  * different encoding. We use the v3 encoding as more likely
875                  * to be intended in a v3 query.
876                  */
877                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
878                 if (!max_delay)
879                         max_delay = 1;  /* can't mod w/ 0 */
880         } else { /* v3 */
881                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
882                         return;
883
884                 ih3 = igmpv3_query_hdr(skb);
885                 if (ih3->nsrcs) {
886                         if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
887                                            + ntohs(ih3->nsrcs)*sizeof(__be32)))
888                                 return;
889                         ih3 = igmpv3_query_hdr(skb);
890                 }
891
892                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
893                 if (!max_delay)
894                         max_delay = 1;  /* can't mod w/ 0 */
895                 in_dev->mr_maxdelay = max_delay;
896                 if (ih3->qrv)
897                         in_dev->mr_qrv = ih3->qrv;
898                 if (!group) { /* general query */
899                         if (ih3->nsrcs)
900                                 return; /* no sources allowed */
901                         igmp_gq_start_timer(in_dev);
902                         return;
903                 }
904                 /* mark sources to include, if group & source-specific */
905                 mark = ih3->nsrcs != 0;
906         }
907
908         /*
909          * - Start the timers in all of our membership records
910          *   that the query applies to for the interface on
911          *   which the query arrived excl. those that belong
912          *   to a "local" group (224.0.0.X)
913          * - For timers already running check if they need to
914          *   be reset.
915          * - Use the igmp->igmp_code field as the maximum
916          *   delay possible
917          */
918         rcu_read_lock();
919         for_each_pmc_rcu(in_dev, im) {
920                 int changed;
921
922                 if (group && group != im->multiaddr)
923                         continue;
924                 if (im->multiaddr == IGMP_ALL_HOSTS)
925                         continue;
926                 spin_lock_bh(&im->lock);
927                 if (im->tm_running)
928                         im->gsquery = im->gsquery && mark;
929                 else
930                         im->gsquery = mark;
931                 changed = !im->gsquery ||
932                         igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
933                 spin_unlock_bh(&im->lock);
934                 if (changed)
935                         igmp_mod_timer(im, max_delay);
936         }
937         rcu_read_unlock();
938 }
939
940 /* called in rcu_read_lock() section */
941 int igmp_rcv(struct sk_buff *skb)
942 {
943         /* This basically follows the spec line by line -- see RFC1112 */
944         struct igmphdr *ih;
945         struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
946         int len = skb->len;
947
948         if (in_dev == NULL)
949                 goto drop;
950
951         if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
952                 goto drop;
953
954         switch (skb->ip_summed) {
955         case CHECKSUM_COMPLETE:
956                 if (!csum_fold(skb->csum))
957                         break;
958                 /* fall through */
959         case CHECKSUM_NONE:
960                 skb->csum = 0;
961                 if (__skb_checksum_complete(skb))
962                         goto drop;
963         }
964
965         ih = igmp_hdr(skb);
966         switch (ih->type) {
967         case IGMP_HOST_MEMBERSHIP_QUERY:
968                 igmp_heard_query(in_dev, skb, len);
969                 break;
970         case IGMP_HOST_MEMBERSHIP_REPORT:
971         case IGMPV2_HOST_MEMBERSHIP_REPORT:
972                 /* Is it our report looped back? */
973                 if (rt_is_output_route(skb_rtable(skb)))
974                         break;
975                 /* don't rely on MC router hearing unicast reports */
976                 if (skb->pkt_type == PACKET_MULTICAST ||
977                     skb->pkt_type == PACKET_BROADCAST)
978                         igmp_heard_report(in_dev, ih->group);
979                 break;
980         case IGMP_PIM:
981 #ifdef CONFIG_IP_PIMSM_V1
982                 return pim_rcv_v1(skb);
983 #endif
984         case IGMPV3_HOST_MEMBERSHIP_REPORT:
985         case IGMP_DVMRP:
986         case IGMP_TRACE:
987         case IGMP_HOST_LEAVE_MESSAGE:
988         case IGMP_MTRACE:
989         case IGMP_MTRACE_RESP:
990                 break;
991         default:
992                 break;
993         }
994
995 drop:
996         kfree_skb(skb);
997         return 0;
998 }
999
1000 #endif
1001
1002
1003 /*
1004  *      Add a filter to a device
1005  */
1006
1007 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1008 {
1009         char buf[MAX_ADDR_LEN];
1010         struct net_device *dev = in_dev->dev;
1011
1012         /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1013            We will get multicast token leakage, when IFF_MULTICAST
1014            is changed. This check should be done in dev->set_multicast_list
1015            routine. Something sort of:
1016            if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1017            --ANK
1018            */
1019         if (arp_mc_map(addr, buf, dev, 0) == 0)
1020                 dev_mc_add(dev, buf);
1021 }
1022
1023 /*
1024  *      Remove a filter from a device
1025  */
1026
1027 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1028 {
1029         char buf[MAX_ADDR_LEN];
1030         struct net_device *dev = in_dev->dev;
1031
1032         if (arp_mc_map(addr, buf, dev, 0) == 0)
1033                 dev_mc_del(dev, buf);
1034 }
1035
1036 #ifdef CONFIG_IP_MULTICAST
1037 /*
1038  * deleted ip_mc_list manipulation
1039  */
1040 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1041 {
1042         struct ip_mc_list *pmc;
1043
1044         /* this is an "ip_mc_list" for convenience; only the fields below
1045          * are actually used. In particular, the refcnt and users are not
1046          * used for management of the delete list. Using the same structure
1047          * for deleted items allows change reports to use common code with
1048          * non-deleted or query-response MCA's.
1049          */
1050         pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1051         if (!pmc)
1052                 return;
1053         spin_lock_bh(&im->lock);
1054         pmc->interface = im->interface;
1055         in_dev_hold(in_dev);
1056         pmc->multiaddr = im->multiaddr;
1057         pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1058                 IGMP_Unsolicited_Report_Count;
1059         pmc->sfmode = im->sfmode;
1060         if (pmc->sfmode == MCAST_INCLUDE) {
1061                 struct ip_sf_list *psf;
1062
1063                 pmc->tomb = im->tomb;
1064                 pmc->sources = im->sources;
1065                 im->tomb = im->sources = NULL;
1066                 for (psf=pmc->sources; psf; psf=psf->sf_next)
1067                         psf->sf_crcount = pmc->crcount;
1068         }
1069         spin_unlock_bh(&im->lock);
1070
1071         spin_lock_bh(&in_dev->mc_tomb_lock);
1072         pmc->next = in_dev->mc_tomb;
1073         in_dev->mc_tomb = pmc;
1074         spin_unlock_bh(&in_dev->mc_tomb_lock);
1075 }
1076
1077 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1078 {
1079         struct ip_mc_list *pmc, *pmc_prev;
1080         struct ip_sf_list *psf, *psf_next;
1081
1082         spin_lock_bh(&in_dev->mc_tomb_lock);
1083         pmc_prev = NULL;
1084         for (pmc=in_dev->mc_tomb; pmc; pmc=pmc->next) {
1085                 if (pmc->multiaddr == multiaddr)
1086                         break;
1087                 pmc_prev = pmc;
1088         }
1089         if (pmc) {
1090                 if (pmc_prev)
1091                         pmc_prev->next = pmc->next;
1092                 else
1093                         in_dev->mc_tomb = pmc->next;
1094         }
1095         spin_unlock_bh(&in_dev->mc_tomb_lock);
1096         if (pmc) {
1097                 for (psf=pmc->tomb; psf; psf=psf_next) {
1098                         psf_next = psf->sf_next;
1099                         kfree(psf);
1100                 }
1101                 in_dev_put(pmc->interface);
1102                 kfree(pmc);
1103         }
1104 }
1105
1106 static void igmpv3_clear_delrec(struct in_device *in_dev)
1107 {
1108         struct ip_mc_list *pmc, *nextpmc;
1109
1110         spin_lock_bh(&in_dev->mc_tomb_lock);
1111         pmc = in_dev->mc_tomb;
1112         in_dev->mc_tomb = NULL;
1113         spin_unlock_bh(&in_dev->mc_tomb_lock);
1114
1115         for (; pmc; pmc = nextpmc) {
1116                 nextpmc = pmc->next;
1117                 ip_mc_clear_src(pmc);
1118                 in_dev_put(pmc->interface);
1119                 kfree(pmc);
1120         }
1121         /* clear dead sources, too */
1122         rcu_read_lock();
1123         for_each_pmc_rcu(in_dev, pmc) {
1124                 struct ip_sf_list *psf, *psf_next;
1125
1126                 spin_lock_bh(&pmc->lock);
1127                 psf = pmc->tomb;
1128                 pmc->tomb = NULL;
1129                 spin_unlock_bh(&pmc->lock);
1130                 for (; psf; psf=psf_next) {
1131                         psf_next = psf->sf_next;
1132                         kfree(psf);
1133                 }
1134         }
1135         rcu_read_unlock();
1136 }
1137 #endif
1138
1139 static void igmp_group_dropped(struct ip_mc_list *im)
1140 {
1141         struct in_device *in_dev = im->interface;
1142 #ifdef CONFIG_IP_MULTICAST
1143         int reporter;
1144 #endif
1145
1146         if (im->loaded) {
1147                 im->loaded = 0;
1148                 ip_mc_filter_del(in_dev, im->multiaddr);
1149         }
1150
1151 #ifdef CONFIG_IP_MULTICAST
1152         if (im->multiaddr == IGMP_ALL_HOSTS)
1153                 return;
1154
1155         reporter = im->reporter;
1156         igmp_stop_timer(im);
1157
1158         if (!in_dev->dead) {
1159                 if (IGMP_V1_SEEN(in_dev))
1160                         return;
1161                 if (IGMP_V2_SEEN(in_dev)) {
1162                         if (reporter)
1163                                 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1164                         return;
1165                 }
1166                 /* IGMPv3 */
1167                 igmpv3_add_delrec(in_dev, im);
1168
1169                 igmp_ifc_event(in_dev);
1170         }
1171 #endif
1172 }
1173
1174 static void igmp_group_added(struct ip_mc_list *im)
1175 {
1176         struct in_device *in_dev = im->interface;
1177
1178         if (im->loaded == 0) {
1179                 im->loaded = 1;
1180                 ip_mc_filter_add(in_dev, im->multiaddr);
1181         }
1182
1183 #ifdef CONFIG_IP_MULTICAST
1184         if (im->multiaddr == IGMP_ALL_HOSTS)
1185                 return;
1186
1187         if (in_dev->dead)
1188                 return;
1189         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1190                 spin_lock_bh(&im->lock);
1191                 igmp_start_timer(im, IGMP_Initial_Report_Delay);
1192                 spin_unlock_bh(&im->lock);
1193                 return;
1194         }
1195         /* else, v3 */
1196
1197         im->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1198                 IGMP_Unsolicited_Report_Count;
1199         igmp_ifc_event(in_dev);
1200 #endif
1201 }
1202
1203
1204 /*
1205  *      Multicast list managers
1206  */
1207
1208
1209 /*
1210  *      A socket has joined a multicast group on device dev.
1211  */
1212
1213 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1214 {
1215         struct ip_mc_list *im;
1216
1217         ASSERT_RTNL();
1218
1219         for_each_pmc_rtnl(in_dev, im) {
1220                 if (im->multiaddr == addr) {
1221                         im->users++;
1222                         ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1223                         goto out;
1224                 }
1225         }
1226
1227         im = kzalloc(sizeof(*im), GFP_KERNEL);
1228         if (!im)
1229                 goto out;
1230
1231         im->users = 1;
1232         im->interface = in_dev;
1233         in_dev_hold(in_dev);
1234         im->multiaddr = addr;
1235         /* initial mode is (EX, empty) */
1236         im->sfmode = MCAST_EXCLUDE;
1237         im->sfcount[MCAST_EXCLUDE] = 1;
1238         atomic_set(&im->refcnt, 1);
1239         spin_lock_init(&im->lock);
1240 #ifdef CONFIG_IP_MULTICAST
1241         setup_timer(&im->timer, &igmp_timer_expire, (unsigned long)im);
1242         im->unsolicit_count = IGMP_Unsolicited_Report_Count;
1243 #endif
1244
1245         im->next_rcu = in_dev->mc_list;
1246         in_dev->mc_count++;
1247         rcu_assign_pointer(in_dev->mc_list, im);
1248
1249 #ifdef CONFIG_IP_MULTICAST
1250         igmpv3_del_delrec(in_dev, im->multiaddr);
1251 #endif
1252         igmp_group_added(im);
1253         if (!in_dev->dead)
1254                 ip_rt_multicast_event(in_dev);
1255 out:
1256         return;
1257 }
1258 EXPORT_SYMBOL(ip_mc_inc_group);
1259
1260 /*
1261  *      Resend IGMP JOIN report; used for bonding.
1262  *      Called with rcu_read_lock()
1263  */
1264 void ip_mc_rejoin_groups(struct in_device *in_dev)
1265 {
1266 #ifdef CONFIG_IP_MULTICAST
1267         struct ip_mc_list *im;
1268         int type;
1269
1270         for_each_pmc_rcu(in_dev, im) {
1271                 if (im->multiaddr == IGMP_ALL_HOSTS)
1272                         continue;
1273
1274                 /* a failover is happening and switches
1275                  * must be notified immediately
1276                  */
1277                 if (IGMP_V1_SEEN(in_dev))
1278                         type = IGMP_HOST_MEMBERSHIP_REPORT;
1279                 else if (IGMP_V2_SEEN(in_dev))
1280                         type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1281                 else
1282                         type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1283                 igmp_send_report(in_dev, im, type);
1284         }
1285 #endif
1286 }
1287 EXPORT_SYMBOL(ip_mc_rejoin_groups);
1288
1289 /*
1290  *      A socket has left a multicast group on device dev
1291  */
1292
1293 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1294 {
1295         struct ip_mc_list *i;
1296         struct ip_mc_list __rcu **ip;
1297
1298         ASSERT_RTNL();
1299
1300         for (ip = &in_dev->mc_list;
1301              (i = rtnl_dereference(*ip)) != NULL;
1302              ip = &i->next_rcu) {
1303                 if (i->multiaddr == addr) {
1304                         if (--i->users == 0) {
1305                                 *ip = i->next_rcu;
1306                                 in_dev->mc_count--;
1307                                 igmp_group_dropped(i);
1308                                 ip_mc_clear_src(i);
1309
1310                                 if (!in_dev->dead)
1311                                         ip_rt_multicast_event(in_dev);
1312
1313                                 ip_ma_put(i);
1314                                 return;
1315                         }
1316                         break;
1317                 }
1318         }
1319 }
1320 EXPORT_SYMBOL(ip_mc_dec_group);
1321
1322 /* Device changing type */
1323
1324 void ip_mc_unmap(struct in_device *in_dev)
1325 {
1326         struct ip_mc_list *pmc;
1327
1328         ASSERT_RTNL();
1329
1330         for_each_pmc_rtnl(in_dev, pmc)
1331                 igmp_group_dropped(pmc);
1332 }
1333
1334 void ip_mc_remap(struct in_device *in_dev)
1335 {
1336         struct ip_mc_list *pmc;
1337
1338         ASSERT_RTNL();
1339
1340         for_each_pmc_rtnl(in_dev, pmc)
1341                 igmp_group_added(pmc);
1342 }
1343
1344 /* Device going down */
1345
1346 void ip_mc_down(struct in_device *in_dev)
1347 {
1348         struct ip_mc_list *pmc;
1349
1350         ASSERT_RTNL();
1351
1352         for_each_pmc_rtnl(in_dev, pmc)
1353                 igmp_group_dropped(pmc);
1354
1355 #ifdef CONFIG_IP_MULTICAST
1356         in_dev->mr_ifc_count = 0;
1357         if (del_timer(&in_dev->mr_ifc_timer))
1358                 __in_dev_put(in_dev);
1359         in_dev->mr_gq_running = 0;
1360         if (del_timer(&in_dev->mr_gq_timer))
1361                 __in_dev_put(in_dev);
1362         igmpv3_clear_delrec(in_dev);
1363 #endif
1364
1365         ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1366 }
1367
1368 void ip_mc_init_dev(struct in_device *in_dev)
1369 {
1370         ASSERT_RTNL();
1371
1372         in_dev->mc_tomb = NULL;
1373 #ifdef CONFIG_IP_MULTICAST
1374         in_dev->mr_gq_running = 0;
1375         setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1376                         (unsigned long)in_dev);
1377         in_dev->mr_ifc_count = 0;
1378         in_dev->mc_count     = 0;
1379         setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1380                         (unsigned long)in_dev);
1381         in_dev->mr_qrv = IGMP_Unsolicited_Report_Count;
1382 #endif
1383
1384         spin_lock_init(&in_dev->mc_tomb_lock);
1385 }
1386
1387 /* Device going up */
1388
1389 void ip_mc_up(struct in_device *in_dev)
1390 {
1391         struct ip_mc_list *pmc;
1392
1393         ASSERT_RTNL();
1394
1395         ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1396
1397         for_each_pmc_rtnl(in_dev, pmc)
1398                 igmp_group_added(pmc);
1399 }
1400
1401 /*
1402  *      Device is about to be destroyed: clean up.
1403  */
1404
1405 void ip_mc_destroy_dev(struct in_device *in_dev)
1406 {
1407         struct ip_mc_list *i;
1408
1409         ASSERT_RTNL();
1410
1411         /* Deactivate timers */
1412         ip_mc_down(in_dev);
1413
1414         while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1415                 in_dev->mc_list = i->next_rcu;
1416                 in_dev->mc_count--;
1417
1418                 /* We've dropped the groups in ip_mc_down already */
1419                 ip_mc_clear_src(i);
1420                 ip_ma_put(i);
1421         }
1422 }
1423
1424 /* RTNL is locked */
1425 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1426 {
1427         struct net_device *dev = NULL;
1428         struct in_device *idev = NULL;
1429
1430         if (imr->imr_ifindex) {
1431                 idev = inetdev_by_index(net, imr->imr_ifindex);
1432                 return idev;
1433         }
1434         if (imr->imr_address.s_addr) {
1435                 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1436                 if (!dev)
1437                         return NULL;
1438         }
1439
1440         if (!dev) {
1441                 struct rtable *rt = ip_route_output(net,
1442                                                     imr->imr_multiaddr.s_addr,
1443                                                     0, 0, 0);
1444                 if (!IS_ERR(rt)) {
1445                         dev = rt->dst.dev;
1446                         ip_rt_put(rt);
1447                 }
1448         }
1449         if (dev) {
1450                 imr->imr_ifindex = dev->ifindex;
1451                 idev = __in_dev_get_rtnl(dev);
1452         }
1453         return idev;
1454 }
1455
1456 /*
1457  *      Join a socket to a group
1458  */
1459 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1460 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1461
1462
1463 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1464         __be32 *psfsrc)
1465 {
1466         struct ip_sf_list *psf, *psf_prev;
1467         int rv = 0;
1468
1469         psf_prev = NULL;
1470         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1471                 if (psf->sf_inaddr == *psfsrc)
1472                         break;
1473                 psf_prev = psf;
1474         }
1475         if (!psf || psf->sf_count[sfmode] == 0) {
1476                 /* source filter not found, or count wrong =>  bug */
1477                 return -ESRCH;
1478         }
1479         psf->sf_count[sfmode]--;
1480         if (psf->sf_count[sfmode] == 0) {
1481                 ip_rt_multicast_event(pmc->interface);
1482         }
1483         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1484 #ifdef CONFIG_IP_MULTICAST
1485                 struct in_device *in_dev = pmc->interface;
1486 #endif
1487
1488                 /* no more filters for this source */
1489                 if (psf_prev)
1490                         psf_prev->sf_next = psf->sf_next;
1491                 else
1492                         pmc->sources = psf->sf_next;
1493 #ifdef CONFIG_IP_MULTICAST
1494                 if (psf->sf_oldin &&
1495                     !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1496                         psf->sf_crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1497                                 IGMP_Unsolicited_Report_Count;
1498                         psf->sf_next = pmc->tomb;
1499                         pmc->tomb = psf;
1500                         rv = 1;
1501                 } else
1502 #endif
1503                         kfree(psf);
1504         }
1505         return rv;
1506 }
1507
1508 #ifndef CONFIG_IP_MULTICAST
1509 #define igmp_ifc_event(x)       do { } while (0)
1510 #endif
1511
1512 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1513                          int sfcount, __be32 *psfsrc, int delta)
1514 {
1515         struct ip_mc_list *pmc;
1516         int     changerec = 0;
1517         int     i, err;
1518
1519         if (!in_dev)
1520                 return -ENODEV;
1521         rcu_read_lock();
1522         for_each_pmc_rcu(in_dev, pmc) {
1523                 if (*pmca == pmc->multiaddr)
1524                         break;
1525         }
1526         if (!pmc) {
1527                 /* MCA not found?? bug */
1528                 rcu_read_unlock();
1529                 return -ESRCH;
1530         }
1531         spin_lock_bh(&pmc->lock);
1532         rcu_read_unlock();
1533 #ifdef CONFIG_IP_MULTICAST
1534         sf_markstate(pmc);
1535 #endif
1536         if (!delta) {
1537                 err = -EINVAL;
1538                 if (!pmc->sfcount[sfmode])
1539                         goto out_unlock;
1540                 pmc->sfcount[sfmode]--;
1541         }
1542         err = 0;
1543         for (i=0; i<sfcount; i++) {
1544                 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1545
1546                 changerec |= rv > 0;
1547                 if (!err && rv < 0)
1548                         err = rv;
1549         }
1550         if (pmc->sfmode == MCAST_EXCLUDE &&
1551             pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1552             pmc->sfcount[MCAST_INCLUDE]) {
1553 #ifdef CONFIG_IP_MULTICAST
1554                 struct ip_sf_list *psf;
1555 #endif
1556
1557                 /* filter mode change */
1558                 pmc->sfmode = MCAST_INCLUDE;
1559 #ifdef CONFIG_IP_MULTICAST
1560                 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1561                         IGMP_Unsolicited_Report_Count;
1562                 in_dev->mr_ifc_count = pmc->crcount;
1563                 for (psf=pmc->sources; psf; psf = psf->sf_next)
1564                         psf->sf_crcount = 0;
1565                 igmp_ifc_event(pmc->interface);
1566         } else if (sf_setstate(pmc) || changerec) {
1567                 igmp_ifc_event(pmc->interface);
1568 #endif
1569         }
1570 out_unlock:
1571         spin_unlock_bh(&pmc->lock);
1572         return err;
1573 }
1574
1575 /*
1576  * Add multicast single-source filter to the interface list
1577  */
1578 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1579         __be32 *psfsrc, int delta)
1580 {
1581         struct ip_sf_list *psf, *psf_prev;
1582
1583         psf_prev = NULL;
1584         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1585                 if (psf->sf_inaddr == *psfsrc)
1586                         break;
1587                 psf_prev = psf;
1588         }
1589         if (!psf) {
1590                 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1591                 if (!psf)
1592                         return -ENOBUFS;
1593                 psf->sf_inaddr = *psfsrc;
1594                 if (psf_prev) {
1595                         psf_prev->sf_next = psf;
1596                 } else
1597                         pmc->sources = psf;
1598         }
1599         psf->sf_count[sfmode]++;
1600         if (psf->sf_count[sfmode] == 1) {
1601                 ip_rt_multicast_event(pmc->interface);
1602         }
1603         return 0;
1604 }
1605
1606 #ifdef CONFIG_IP_MULTICAST
1607 static void sf_markstate(struct ip_mc_list *pmc)
1608 {
1609         struct ip_sf_list *psf;
1610         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1611
1612         for (psf=pmc->sources; psf; psf=psf->sf_next)
1613                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1614                         psf->sf_oldin = mca_xcount ==
1615                                 psf->sf_count[MCAST_EXCLUDE] &&
1616                                 !psf->sf_count[MCAST_INCLUDE];
1617                 } else
1618                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1619 }
1620
1621 static int sf_setstate(struct ip_mc_list *pmc)
1622 {
1623         struct ip_sf_list *psf, *dpsf;
1624         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1625         int qrv = pmc->interface->mr_qrv;
1626         int new_in, rv;
1627
1628         rv = 0;
1629         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1630                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1631                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1632                                 !psf->sf_count[MCAST_INCLUDE];
1633                 } else
1634                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1635                 if (new_in) {
1636                         if (!psf->sf_oldin) {
1637                                 struct ip_sf_list *prev = NULL;
1638
1639                                 for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next) {
1640                                         if (dpsf->sf_inaddr == psf->sf_inaddr)
1641                                                 break;
1642                                         prev = dpsf;
1643                                 }
1644                                 if (dpsf) {
1645                                         if (prev)
1646                                                 prev->sf_next = dpsf->sf_next;
1647                                         else
1648                                                 pmc->tomb = dpsf->sf_next;
1649                                         kfree(dpsf);
1650                                 }
1651                                 psf->sf_crcount = qrv;
1652                                 rv++;
1653                         }
1654                 } else if (psf->sf_oldin) {
1655
1656                         psf->sf_crcount = 0;
1657                         /*
1658                          * add or update "delete" records if an active filter
1659                          * is now inactive
1660                          */
1661                         for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next)
1662                                 if (dpsf->sf_inaddr == psf->sf_inaddr)
1663                                         break;
1664                         if (!dpsf) {
1665                                 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1666                                 if (!dpsf)
1667                                         continue;
1668                                 *dpsf = *psf;
1669                                 /* pmc->lock held by callers */
1670                                 dpsf->sf_next = pmc->tomb;
1671                                 pmc->tomb = dpsf;
1672                         }
1673                         dpsf->sf_crcount = qrv;
1674                         rv++;
1675                 }
1676         }
1677         return rv;
1678 }
1679 #endif
1680
1681 /*
1682  * Add multicast source filter list to the interface list
1683  */
1684 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1685                          int sfcount, __be32 *psfsrc, int delta)
1686 {
1687         struct ip_mc_list *pmc;
1688         int     isexclude;
1689         int     i, err;
1690
1691         if (!in_dev)
1692                 return -ENODEV;
1693         rcu_read_lock();
1694         for_each_pmc_rcu(in_dev, pmc) {
1695                 if (*pmca == pmc->multiaddr)
1696                         break;
1697         }
1698         if (!pmc) {
1699                 /* MCA not found?? bug */
1700                 rcu_read_unlock();
1701                 return -ESRCH;
1702         }
1703         spin_lock_bh(&pmc->lock);
1704         rcu_read_unlock();
1705
1706 #ifdef CONFIG_IP_MULTICAST
1707         sf_markstate(pmc);
1708 #endif
1709         isexclude = pmc->sfmode == MCAST_EXCLUDE;
1710         if (!delta)
1711                 pmc->sfcount[sfmode]++;
1712         err = 0;
1713         for (i=0; i<sfcount; i++) {
1714                 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i], delta);
1715                 if (err)
1716                         break;
1717         }
1718         if (err) {
1719                 int j;
1720
1721                 pmc->sfcount[sfmode]--;
1722                 for (j=0; j<i; j++)
1723                         (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[j]);
1724         } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1725 #ifdef CONFIG_IP_MULTICAST
1726                 struct ip_sf_list *psf;
1727                 in_dev = pmc->interface;
1728 #endif
1729
1730                 /* filter mode change */
1731                 if (pmc->sfcount[MCAST_EXCLUDE])
1732                         pmc->sfmode = MCAST_EXCLUDE;
1733                 else if (pmc->sfcount[MCAST_INCLUDE])
1734                         pmc->sfmode = MCAST_INCLUDE;
1735 #ifdef CONFIG_IP_MULTICAST
1736                 /* else no filters; keep old mode for reports */
1737
1738                 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1739                         IGMP_Unsolicited_Report_Count;
1740                 in_dev->mr_ifc_count = pmc->crcount;
1741                 for (psf=pmc->sources; psf; psf = psf->sf_next)
1742                         psf->sf_crcount = 0;
1743                 igmp_ifc_event(in_dev);
1744         } else if (sf_setstate(pmc)) {
1745                 igmp_ifc_event(in_dev);
1746 #endif
1747         }
1748         spin_unlock_bh(&pmc->lock);
1749         return err;
1750 }
1751
1752 static void ip_mc_clear_src(struct ip_mc_list *pmc)
1753 {
1754         struct ip_sf_list *psf, *nextpsf;
1755
1756         for (psf=pmc->tomb; psf; psf=nextpsf) {
1757                 nextpsf = psf->sf_next;
1758                 kfree(psf);
1759         }
1760         pmc->tomb = NULL;
1761         for (psf=pmc->sources; psf; psf=nextpsf) {
1762                 nextpsf = psf->sf_next;
1763                 kfree(psf);
1764         }
1765         pmc->sources = NULL;
1766         pmc->sfmode = MCAST_EXCLUDE;
1767         pmc->sfcount[MCAST_INCLUDE] = 0;
1768         pmc->sfcount[MCAST_EXCLUDE] = 1;
1769 }
1770
1771
1772 /*
1773  * Join a multicast group
1774  */
1775 int ip_mc_join_group(struct sock *sk , struct ip_mreqn *imr)
1776 {
1777         int err;
1778         __be32 addr = imr->imr_multiaddr.s_addr;
1779         struct ip_mc_socklist *iml = NULL, *i;
1780         struct in_device *in_dev;
1781         struct inet_sock *inet = inet_sk(sk);
1782         struct net *net = sock_net(sk);
1783         int ifindex;
1784         int count = 0;
1785
1786         if (!ipv4_is_multicast(addr))
1787                 return -EINVAL;
1788
1789         rtnl_lock();
1790
1791         in_dev = ip_mc_find_dev(net, imr);
1792
1793         if (!in_dev) {
1794                 iml = NULL;
1795                 err = -ENODEV;
1796                 goto done;
1797         }
1798
1799         err = -EADDRINUSE;
1800         ifindex = imr->imr_ifindex;
1801         for_each_pmc_rtnl(inet, i) {
1802                 if (i->multi.imr_multiaddr.s_addr == addr &&
1803                     i->multi.imr_ifindex == ifindex)
1804                         goto done;
1805                 count++;
1806         }
1807         err = -ENOBUFS;
1808         if (count >= sysctl_igmp_max_memberships)
1809                 goto done;
1810         iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
1811         if (iml == NULL)
1812                 goto done;
1813
1814         memcpy(&iml->multi, imr, sizeof(*imr));
1815         iml->next_rcu = inet->mc_list;
1816         iml->sflist = NULL;
1817         iml->sfmode = MCAST_EXCLUDE;
1818         rcu_assign_pointer(inet->mc_list, iml);
1819         ip_mc_inc_group(in_dev, addr);
1820         err = 0;
1821 done:
1822         rtnl_unlock();
1823         return err;
1824 }
1825 EXPORT_SYMBOL(ip_mc_join_group);
1826
1827 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
1828                            struct in_device *in_dev)
1829 {
1830         struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
1831         int err;
1832
1833         if (psf == NULL) {
1834                 /* any-source empty exclude case */
1835                 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1836                         iml->sfmode, 0, NULL, 0);
1837         }
1838         err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1839                         iml->sfmode, psf->sl_count, psf->sl_addr, 0);
1840         rcu_assign_pointer(iml->sflist, NULL);
1841         /* decrease mem now to avoid the memleak warning */
1842         atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
1843         kfree_rcu(psf, rcu);
1844         return err;
1845 }
1846
1847 /*
1848  *      Ask a socket to leave a group.
1849  */
1850
1851 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
1852 {
1853         struct inet_sock *inet = inet_sk(sk);
1854         struct ip_mc_socklist *iml;
1855         struct ip_mc_socklist __rcu **imlp;
1856         struct in_device *in_dev;
1857         struct net *net = sock_net(sk);
1858         __be32 group = imr->imr_multiaddr.s_addr;
1859         u32 ifindex;
1860         int ret = -EADDRNOTAVAIL;
1861
1862         rtnl_lock();
1863         in_dev = ip_mc_find_dev(net, imr);
1864         ifindex = imr->imr_ifindex;
1865         for (imlp = &inet->mc_list;
1866              (iml = rtnl_dereference(*imlp)) != NULL;
1867              imlp = &iml->next_rcu) {
1868                 if (iml->multi.imr_multiaddr.s_addr != group)
1869                         continue;
1870                 if (ifindex) {
1871                         if (iml->multi.imr_ifindex != ifindex)
1872                                 continue;
1873                 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
1874                                 iml->multi.imr_address.s_addr)
1875                         continue;
1876
1877                 (void) ip_mc_leave_src(sk, iml, in_dev);
1878
1879                 *imlp = iml->next_rcu;
1880
1881                 if (in_dev)
1882                         ip_mc_dec_group(in_dev, group);
1883                 rtnl_unlock();
1884                 /* decrease mem now to avoid the memleak warning */
1885                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
1886                 kfree_rcu(iml, rcu);
1887                 return 0;
1888         }
1889         if (!in_dev)
1890                 ret = -ENODEV;
1891         rtnl_unlock();
1892         return ret;
1893 }
1894
1895 int ip_mc_source(int add, int omode, struct sock *sk, struct
1896         ip_mreq_source *mreqs, int ifindex)
1897 {
1898         int err;
1899         struct ip_mreqn imr;
1900         __be32 addr = mreqs->imr_multiaddr;
1901         struct ip_mc_socklist *pmc;
1902         struct in_device *in_dev = NULL;
1903         struct inet_sock *inet = inet_sk(sk);
1904         struct ip_sf_socklist *psl;
1905         struct net *net = sock_net(sk);
1906         int leavegroup = 0;
1907         int i, j, rv;
1908
1909         if (!ipv4_is_multicast(addr))
1910                 return -EINVAL;
1911
1912         rtnl_lock();
1913
1914         imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
1915         imr.imr_address.s_addr = mreqs->imr_interface;
1916         imr.imr_ifindex = ifindex;
1917         in_dev = ip_mc_find_dev(net, &imr);
1918
1919         if (!in_dev) {
1920                 err = -ENODEV;
1921                 goto done;
1922         }
1923         err = -EADDRNOTAVAIL;
1924
1925         for_each_pmc_rtnl(inet, pmc) {
1926                 if ((pmc->multi.imr_multiaddr.s_addr ==
1927                      imr.imr_multiaddr.s_addr) &&
1928                     (pmc->multi.imr_ifindex == imr.imr_ifindex))
1929                         break;
1930         }
1931         if (!pmc) {             /* must have a prior join */
1932                 err = -EINVAL;
1933                 goto done;
1934         }
1935         /* if a source filter was set, must be the same mode as before */
1936         if (pmc->sflist) {
1937                 if (pmc->sfmode != omode) {
1938                         err = -EINVAL;
1939                         goto done;
1940                 }
1941         } else if (pmc->sfmode != omode) {
1942                 /* allow mode switches for empty-set filters */
1943                 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
1944                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
1945                         NULL, 0);
1946                 pmc->sfmode = omode;
1947         }
1948
1949         psl = rtnl_dereference(pmc->sflist);
1950         if (!add) {
1951                 if (!psl)
1952                         goto done;      /* err = -EADDRNOTAVAIL */
1953                 rv = !0;
1954                 for (i=0; i<psl->sl_count; i++) {
1955                         rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
1956                                 sizeof(__be32));
1957                         if (rv == 0)
1958                                 break;
1959                 }
1960                 if (rv)         /* source not found */
1961                         goto done;      /* err = -EADDRNOTAVAIL */
1962
1963                 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
1964                 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
1965                         leavegroup = 1;
1966                         goto done;
1967                 }
1968
1969                 /* update the interface filter */
1970                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
1971                         &mreqs->imr_sourceaddr, 1);
1972
1973                 for (j=i+1; j<psl->sl_count; j++)
1974                         psl->sl_addr[j-1] = psl->sl_addr[j];
1975                 psl->sl_count--;
1976                 err = 0;
1977                 goto done;
1978         }
1979         /* else, add a new source to the filter */
1980
1981         if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
1982                 err = -ENOBUFS;
1983                 goto done;
1984         }
1985         if (!psl || psl->sl_count == psl->sl_max) {
1986                 struct ip_sf_socklist *newpsl;
1987                 int count = IP_SFBLOCK;
1988
1989                 if (psl)
1990                         count += psl->sl_max;
1991                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
1992                 if (!newpsl) {
1993                         err = -ENOBUFS;
1994                         goto done;
1995                 }
1996                 newpsl->sl_max = count;
1997                 newpsl->sl_count = count - IP_SFBLOCK;
1998                 if (psl) {
1999                         for (i=0; i<psl->sl_count; i++)
2000                                 newpsl->sl_addr[i] = psl->sl_addr[i];
2001                         /* decrease mem now to avoid the memleak warning */
2002                         atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2003                         kfree_rcu(psl, rcu);
2004                 }
2005                 rcu_assign_pointer(pmc->sflist, newpsl);
2006                 psl = newpsl;
2007         }
2008         rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2009         for (i=0; i<psl->sl_count; i++) {
2010                 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2011                         sizeof(__be32));
2012                 if (rv == 0)
2013                         break;
2014         }
2015         if (rv == 0)            /* address already there is an error */
2016                 goto done;
2017         for (j=psl->sl_count-1; j>=i; j--)
2018                 psl->sl_addr[j+1] = psl->sl_addr[j];
2019         psl->sl_addr[i] = mreqs->imr_sourceaddr;
2020         psl->sl_count++;
2021         err = 0;
2022         /* update the interface list */
2023         ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2024                 &mreqs->imr_sourceaddr, 1);
2025 done:
2026         rtnl_unlock();
2027         if (leavegroup)
2028                 return ip_mc_leave_group(sk, &imr);
2029         return err;
2030 }
2031
2032 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2033 {
2034         int err = 0;
2035         struct ip_mreqn imr;
2036         __be32 addr = msf->imsf_multiaddr;
2037         struct ip_mc_socklist *pmc;
2038         struct in_device *in_dev;
2039         struct inet_sock *inet = inet_sk(sk);
2040         struct ip_sf_socklist *newpsl, *psl;
2041         struct net *net = sock_net(sk);
2042         int leavegroup = 0;
2043
2044         if (!ipv4_is_multicast(addr))
2045                 return -EINVAL;
2046         if (msf->imsf_fmode != MCAST_INCLUDE &&
2047             msf->imsf_fmode != MCAST_EXCLUDE)
2048                 return -EINVAL;
2049
2050         rtnl_lock();
2051
2052         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2053         imr.imr_address.s_addr = msf->imsf_interface;
2054         imr.imr_ifindex = ifindex;
2055         in_dev = ip_mc_find_dev(net, &imr);
2056
2057         if (!in_dev) {
2058                 err = -ENODEV;
2059                 goto done;
2060         }
2061
2062         /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2063         if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2064                 leavegroup = 1;
2065                 goto done;
2066         }
2067
2068         for_each_pmc_rtnl(inet, pmc) {
2069                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2070                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2071                         break;
2072         }
2073         if (!pmc) {             /* must have a prior join */
2074                 err = -EINVAL;
2075                 goto done;
2076         }
2077         if (msf->imsf_numsrc) {
2078                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2079                                                            GFP_KERNEL);
2080                 if (!newpsl) {
2081                         err = -ENOBUFS;
2082                         goto done;
2083                 }
2084                 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2085                 memcpy(newpsl->sl_addr, msf->imsf_slist,
2086                         msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2087                 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2088                         msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2089                 if (err) {
2090                         sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2091                         goto done;
2092                 }
2093         } else {
2094                 newpsl = NULL;
2095                 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2096                                      msf->imsf_fmode, 0, NULL, 0);
2097         }
2098         psl = rtnl_dereference(pmc->sflist);
2099         if (psl) {
2100                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2101                         psl->sl_count, psl->sl_addr, 0);
2102                 /* decrease mem now to avoid the memleak warning */
2103                 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2104                 kfree_rcu(psl, rcu);
2105         } else
2106                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2107                         0, NULL, 0);
2108         rcu_assign_pointer(pmc->sflist, newpsl);
2109         pmc->sfmode = msf->imsf_fmode;
2110         err = 0;
2111 done:
2112         rtnl_unlock();
2113         if (leavegroup)
2114                 err = ip_mc_leave_group(sk, &imr);
2115         return err;
2116 }
2117
2118 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2119         struct ip_msfilter __user *optval, int __user *optlen)
2120 {
2121         int err, len, count, copycount;
2122         struct ip_mreqn imr;
2123         __be32 addr = msf->imsf_multiaddr;
2124         struct ip_mc_socklist *pmc;
2125         struct in_device *in_dev;
2126         struct inet_sock *inet = inet_sk(sk);
2127         struct ip_sf_socklist *psl;
2128         struct net *net = sock_net(sk);
2129
2130         if (!ipv4_is_multicast(addr))
2131                 return -EINVAL;
2132
2133         rtnl_lock();
2134
2135         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2136         imr.imr_address.s_addr = msf->imsf_interface;
2137         imr.imr_ifindex = 0;
2138         in_dev = ip_mc_find_dev(net, &imr);
2139
2140         if (!in_dev) {
2141                 err = -ENODEV;
2142                 goto done;
2143         }
2144         err = -EADDRNOTAVAIL;
2145
2146         for_each_pmc_rtnl(inet, pmc) {
2147                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2148                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2149                         break;
2150         }
2151         if (!pmc)               /* must have a prior join */
2152                 goto done;
2153         msf->imsf_fmode = pmc->sfmode;
2154         psl = rtnl_dereference(pmc->sflist);
2155         rtnl_unlock();
2156         if (!psl) {
2157                 len = 0;
2158                 count = 0;
2159         } else {
2160                 count = psl->sl_count;
2161         }
2162         copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2163         len = copycount * sizeof(psl->sl_addr[0]);
2164         msf->imsf_numsrc = count;
2165         if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2166             copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2167                 return -EFAULT;
2168         }
2169         if (len &&
2170             copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2171                 return -EFAULT;
2172         return 0;
2173 done:
2174         rtnl_unlock();
2175         return err;
2176 }
2177
2178 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2179         struct group_filter __user *optval, int __user *optlen)
2180 {
2181         int err, i, count, copycount;
2182         struct sockaddr_in *psin;
2183         __be32 addr;
2184         struct ip_mc_socklist *pmc;
2185         struct inet_sock *inet = inet_sk(sk);
2186         struct ip_sf_socklist *psl;
2187
2188         psin = (struct sockaddr_in *)&gsf->gf_group;
2189         if (psin->sin_family != AF_INET)
2190                 return -EINVAL;
2191         addr = psin->sin_addr.s_addr;
2192         if (!ipv4_is_multicast(addr))
2193                 return -EINVAL;
2194
2195         rtnl_lock();
2196
2197         err = -EADDRNOTAVAIL;
2198
2199         for_each_pmc_rtnl(inet, pmc) {
2200                 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2201                     pmc->multi.imr_ifindex == gsf->gf_interface)
2202                         break;
2203         }
2204         if (!pmc)               /* must have a prior join */
2205                 goto done;
2206         gsf->gf_fmode = pmc->sfmode;
2207         psl = rtnl_dereference(pmc->sflist);
2208         rtnl_unlock();
2209         count = psl ? psl->sl_count : 0;
2210         copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2211         gsf->gf_numsrc = count;
2212         if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2213             copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2214                 return -EFAULT;
2215         }
2216         for (i=0; i<copycount; i++) {
2217                 struct sockaddr_storage ss;
2218
2219                 psin = (struct sockaddr_in *)&ss;
2220                 memset(&ss, 0, sizeof(ss));
2221                 psin->sin_family = AF_INET;
2222                 psin->sin_addr.s_addr = psl->sl_addr[i];
2223                 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2224                         return -EFAULT;
2225         }
2226         return 0;
2227 done:
2228         rtnl_unlock();
2229         return err;
2230 }
2231
2232 /*
2233  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2234  */
2235 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2236 {
2237         struct inet_sock *inet = inet_sk(sk);
2238         struct ip_mc_socklist *pmc;
2239         struct ip_sf_socklist *psl;
2240         int i;
2241         int ret;
2242
2243         ret = 1;
2244         if (!ipv4_is_multicast(loc_addr))
2245                 goto out;
2246
2247         rcu_read_lock();
2248         for_each_pmc_rcu(inet, pmc) {
2249                 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2250                     pmc->multi.imr_ifindex == dif)
2251                         break;
2252         }
2253         ret = inet->mc_all;
2254         if (!pmc)
2255                 goto unlock;
2256         psl = rcu_dereference(pmc->sflist);
2257         ret = (pmc->sfmode == MCAST_EXCLUDE);
2258         if (!psl)
2259                 goto unlock;
2260
2261         for (i=0; i<psl->sl_count; i++) {
2262                 if (psl->sl_addr[i] == rmt_addr)
2263                         break;
2264         }
2265         ret = 0;
2266         if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2267                 goto unlock;
2268         if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2269                 goto unlock;
2270         ret = 1;
2271 unlock:
2272         rcu_read_unlock();
2273 out:
2274         return ret;
2275 }
2276
2277 /*
2278  *      A socket is closing.
2279  */
2280
2281 void ip_mc_drop_socket(struct sock *sk)
2282 {
2283         struct inet_sock *inet = inet_sk(sk);
2284         struct ip_mc_socklist *iml;
2285         struct net *net = sock_net(sk);
2286
2287         if (inet->mc_list == NULL)
2288                 return;
2289
2290         rtnl_lock();
2291         while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2292                 struct in_device *in_dev;
2293
2294                 inet->mc_list = iml->next_rcu;
2295                 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2296                 (void) ip_mc_leave_src(sk, iml, in_dev);
2297                 if (in_dev != NULL)
2298                         ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2299                 /* decrease mem now to avoid the memleak warning */
2300                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2301                 kfree_rcu(iml, rcu);
2302         }
2303         rtnl_unlock();
2304 }
2305
2306 /* called with rcu_read_lock() */
2307 int ip_check_mc_rcu(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2308 {
2309         struct ip_mc_list *im;
2310         struct ip_sf_list *psf;
2311         int rv = 0;
2312
2313         for_each_pmc_rcu(in_dev, im) {
2314                 if (im->multiaddr == mc_addr)
2315                         break;
2316         }
2317         if (im && proto == IPPROTO_IGMP) {
2318                 rv = 1;
2319         } else if (im) {
2320                 if (src_addr) {
2321                         for (psf=im->sources; psf; psf=psf->sf_next) {
2322                                 if (psf->sf_inaddr == src_addr)
2323                                         break;
2324                         }
2325                         if (psf)
2326                                 rv = psf->sf_count[MCAST_INCLUDE] ||
2327                                         psf->sf_count[MCAST_EXCLUDE] !=
2328                                         im->sfcount[MCAST_EXCLUDE];
2329                         else
2330                                 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2331                 } else
2332                         rv = 1; /* unspecified source; tentatively allow */
2333         }
2334         return rv;
2335 }
2336
2337 #if defined(CONFIG_PROC_FS)
2338 struct igmp_mc_iter_state {
2339         struct seq_net_private p;
2340         struct net_device *dev;
2341         struct in_device *in_dev;
2342 };
2343
2344 #define igmp_mc_seq_private(seq)        ((struct igmp_mc_iter_state *)(seq)->private)
2345
2346 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2347 {
2348         struct net *net = seq_file_net(seq);
2349         struct ip_mc_list *im = NULL;
2350         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2351
2352         state->in_dev = NULL;
2353         for_each_netdev_rcu(net, state->dev) {
2354                 struct in_device *in_dev;
2355
2356                 in_dev = __in_dev_get_rcu(state->dev);
2357                 if (!in_dev)
2358                         continue;
2359                 im = rcu_dereference(in_dev->mc_list);
2360                 if (im) {
2361                         state->in_dev = in_dev;
2362                         break;
2363                 }
2364         }
2365         return im;
2366 }
2367
2368 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2369 {
2370         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2371
2372         im = rcu_dereference(im->next_rcu);
2373         while (!im) {
2374                 state->dev = next_net_device_rcu(state->dev);
2375                 if (!state->dev) {
2376                         state->in_dev = NULL;
2377                         break;
2378                 }
2379                 state->in_dev = __in_dev_get_rcu(state->dev);
2380                 if (!state->in_dev)
2381                         continue;
2382                 im = rcu_dereference(state->in_dev->mc_list);
2383         }
2384         return im;
2385 }
2386
2387 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2388 {
2389         struct ip_mc_list *im = igmp_mc_get_first(seq);
2390         if (im)
2391                 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2392                         --pos;
2393         return pos ? NULL : im;
2394 }
2395
2396 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2397         __acquires(rcu)
2398 {
2399         rcu_read_lock();
2400         return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2401 }
2402
2403 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2404 {
2405         struct ip_mc_list *im;
2406         if (v == SEQ_START_TOKEN)
2407                 im = igmp_mc_get_first(seq);
2408         else
2409                 im = igmp_mc_get_next(seq, v);
2410         ++*pos;
2411         return im;
2412 }
2413
2414 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2415         __releases(rcu)
2416 {
2417         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2418
2419         state->in_dev = NULL;
2420         state->dev = NULL;
2421         rcu_read_unlock();
2422 }
2423
2424 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2425 {
2426         if (v == SEQ_START_TOKEN)
2427                 seq_puts(seq,
2428                          "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2429         else {
2430                 struct ip_mc_list *im = (struct ip_mc_list *)v;
2431                 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2432                 char   *querier;
2433 #ifdef CONFIG_IP_MULTICAST
2434                 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2435                           IGMP_V2_SEEN(state->in_dev) ? "V2" :
2436                           "V3";
2437 #else
2438                 querier = "NONE";
2439 #endif
2440
2441                 if (rcu_dereference(state->in_dev->mc_list) == im) {
2442                         seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2443                                    state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2444                 }
2445
2446                 seq_printf(seq,
2447                            "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2448                            im->multiaddr, im->users,
2449                            im->tm_running, im->tm_running ?
2450                            jiffies_to_clock_t(im->timer.expires-jiffies) : 0,
2451                            im->reporter);
2452         }
2453         return 0;
2454 }
2455
2456 static const struct seq_operations igmp_mc_seq_ops = {
2457         .start  =       igmp_mc_seq_start,
2458         .next   =       igmp_mc_seq_next,
2459         .stop   =       igmp_mc_seq_stop,
2460         .show   =       igmp_mc_seq_show,
2461 };
2462
2463 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2464 {
2465         return seq_open_net(inode, file, &igmp_mc_seq_ops,
2466                         sizeof(struct igmp_mc_iter_state));
2467 }
2468
2469 static const struct file_operations igmp_mc_seq_fops = {
2470         .owner          =       THIS_MODULE,
2471         .open           =       igmp_mc_seq_open,
2472         .read           =       seq_read,
2473         .llseek         =       seq_lseek,
2474         .release        =       seq_release_net,
2475 };
2476
2477 struct igmp_mcf_iter_state {
2478         struct seq_net_private p;
2479         struct net_device *dev;
2480         struct in_device *idev;
2481         struct ip_mc_list *im;
2482 };
2483
2484 #define igmp_mcf_seq_private(seq)       ((struct igmp_mcf_iter_state *)(seq)->private)
2485
2486 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2487 {
2488         struct net *net = seq_file_net(seq);
2489         struct ip_sf_list *psf = NULL;
2490         struct ip_mc_list *im = NULL;
2491         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2492
2493         state->idev = NULL;
2494         state->im = NULL;
2495         for_each_netdev_rcu(net, state->dev) {
2496                 struct in_device *idev;
2497                 idev = __in_dev_get_rcu(state->dev);
2498                 if (unlikely(idev == NULL))
2499                         continue;
2500                 im = rcu_dereference(idev->mc_list);
2501                 if (likely(im != NULL)) {
2502                         spin_lock_bh(&im->lock);
2503                         psf = im->sources;
2504                         if (likely(psf != NULL)) {
2505                                 state->im = im;
2506                                 state->idev = idev;
2507                                 break;
2508                         }
2509                         spin_unlock_bh(&im->lock);
2510                 }
2511         }
2512         return psf;
2513 }
2514
2515 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2516 {
2517         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2518
2519         psf = psf->sf_next;
2520         while (!psf) {
2521                 spin_unlock_bh(&state->im->lock);
2522                 state->im = state->im->next;
2523                 while (!state->im) {
2524                         state->dev = next_net_device_rcu(state->dev);
2525                         if (!state->dev) {
2526                                 state->idev = NULL;
2527                                 goto out;
2528                         }
2529                         state->idev = __in_dev_get_rcu(state->dev);
2530                         if (!state->idev)
2531                                 continue;
2532                         state->im = rcu_dereference(state->idev->mc_list);
2533                 }
2534                 if (!state->im)
2535                         break;
2536                 spin_lock_bh(&state->im->lock);
2537                 psf = state->im->sources;
2538         }
2539 out:
2540         return psf;
2541 }
2542
2543 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2544 {
2545         struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2546         if (psf)
2547                 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2548                         --pos;
2549         return pos ? NULL : psf;
2550 }
2551
2552 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2553         __acquires(rcu)
2554 {
2555         rcu_read_lock();
2556         return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2557 }
2558
2559 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2560 {
2561         struct ip_sf_list *psf;
2562         if (v == SEQ_START_TOKEN)
2563                 psf = igmp_mcf_get_first(seq);
2564         else
2565                 psf = igmp_mcf_get_next(seq, v);
2566         ++*pos;
2567         return psf;
2568 }
2569
2570 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2571         __releases(rcu)
2572 {
2573         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2574         if (likely(state->im != NULL)) {
2575                 spin_unlock_bh(&state->im->lock);
2576                 state->im = NULL;
2577         }
2578         state->idev = NULL;
2579         state->dev = NULL;
2580         rcu_read_unlock();
2581 }
2582
2583 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2584 {
2585         struct ip_sf_list *psf = (struct ip_sf_list *)v;
2586         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2587
2588         if (v == SEQ_START_TOKEN) {
2589                 seq_printf(seq,
2590                            "%3s %6s "
2591                            "%10s %10s %6s %6s\n", "Idx",
2592                            "Device", "MCA",
2593                            "SRC", "INC", "EXC");
2594         } else {
2595                 seq_printf(seq,
2596                            "%3d %6.6s 0x%08x "
2597                            "0x%08x %6lu %6lu\n",
2598                            state->dev->ifindex, state->dev->name,
2599                            ntohl(state->im->multiaddr),
2600                            ntohl(psf->sf_inaddr),
2601                            psf->sf_count[MCAST_INCLUDE],
2602                            psf->sf_count[MCAST_EXCLUDE]);
2603         }
2604         return 0;
2605 }
2606
2607 static const struct seq_operations igmp_mcf_seq_ops = {
2608         .start  =       igmp_mcf_seq_start,
2609         .next   =       igmp_mcf_seq_next,
2610         .stop   =       igmp_mcf_seq_stop,
2611         .show   =       igmp_mcf_seq_show,
2612 };
2613
2614 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2615 {
2616         return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2617                         sizeof(struct igmp_mcf_iter_state));
2618 }
2619
2620 static const struct file_operations igmp_mcf_seq_fops = {
2621         .owner          =       THIS_MODULE,
2622         .open           =       igmp_mcf_seq_open,
2623         .read           =       seq_read,
2624         .llseek         =       seq_lseek,
2625         .release        =       seq_release_net,
2626 };
2627
2628 static int __net_init igmp_net_init(struct net *net)
2629 {
2630         struct proc_dir_entry *pde;
2631
2632         pde = proc_net_fops_create(net, "igmp", S_IRUGO, &igmp_mc_seq_fops);
2633         if (!pde)
2634                 goto out_igmp;
2635         pde = proc_net_fops_create(net, "mcfilter", S_IRUGO, &igmp_mcf_seq_fops);
2636         if (!pde)
2637                 goto out_mcfilter;
2638         return 0;
2639
2640 out_mcfilter:
2641         proc_net_remove(net, "igmp");
2642 out_igmp:
2643         return -ENOMEM;
2644 }
2645
2646 static void __net_exit igmp_net_exit(struct net *net)
2647 {
2648         proc_net_remove(net, "mcfilter");
2649         proc_net_remove(net, "igmp");
2650 }
2651
2652 static struct pernet_operations igmp_net_ops = {
2653         .init = igmp_net_init,
2654         .exit = igmp_net_exit,
2655 };
2656
2657 int __init igmp_mc_proc_init(void)
2658 {
2659         return register_pernet_subsys(&igmp_net_ops);
2660 }
2661 #endif