ipip: add module alias for tunl0 tunnel device
[linux-2.6.git] / net / ipv4 / igmp.c
1 /*
2  *      Linux NET3:     Internet Group Management Protocol  [IGMP]
3  *
4  *      This code implements the IGMP protocol as defined in RFC1112. There has
5  *      been a further revision of this protocol since which is now supported.
6  *
7  *      If you have trouble with this module be careful what gcc you have used,
8  *      the older version didn't come out right using gcc 2.5.8, the newer one
9  *      seems to fall out with gcc 2.6.2.
10  *
11  *      Authors:
12  *              Alan Cox <alan@lxorguk.ukuu.org.uk>
13  *
14  *      This program is free software; you can redistribute it and/or
15  *      modify it under the terms of the GNU General Public License
16  *      as published by the Free Software Foundation; either version
17  *      2 of the License, or (at your option) any later version.
18  *
19  *      Fixes:
20  *
21  *              Alan Cox        :       Added lots of __inline__ to optimise
22  *                                      the memory usage of all the tiny little
23  *                                      functions.
24  *              Alan Cox        :       Dumped the header building experiment.
25  *              Alan Cox        :       Minor tweaks ready for multicast routing
26  *                                      and extended IGMP protocol.
27  *              Alan Cox        :       Removed a load of inline directives. Gcc 2.5.8
28  *                                      writes utterly bogus code otherwise (sigh)
29  *                                      fixed IGMP loopback to behave in the manner
30  *                                      desired by mrouted, fixed the fact it has been
31  *                                      broken since 1.3.6 and cleaned up a few minor
32  *                                      points.
33  *
34  *              Chih-Jen Chang  :       Tried to revise IGMP to Version 2
35  *              Tsu-Sheng Tsao          E-mail: chihjenc@scf.usc.edu and tsusheng@scf.usc.edu
36  *                                      The enhancements are mainly based on Steve Deering's
37  *                                      ipmulti-3.5 source code.
38  *              Chih-Jen Chang  :       Added the igmp_get_mrouter_info and
39  *              Tsu-Sheng Tsao          igmp_set_mrouter_info to keep track of
40  *                                      the mrouted version on that device.
41  *              Chih-Jen Chang  :       Added the max_resp_time parameter to
42  *              Tsu-Sheng Tsao          igmp_heard_query(). Using this parameter
43  *                                      to identify the multicast router version
44  *                                      and do what the IGMP version 2 specified.
45  *              Chih-Jen Chang  :       Added a timer to revert to IGMP V2 router
46  *              Tsu-Sheng Tsao          if the specified time expired.
47  *              Alan Cox        :       Stop IGMP from 0.0.0.0 being accepted.
48  *              Alan Cox        :       Use GFP_ATOMIC in the right places.
49  *              Christian Daudt :       igmp timer wasn't set for local group
50  *                                      memberships but was being deleted,
51  *                                      which caused a "del_timer() called
52  *                                      from %p with timer not initialized\n"
53  *                                      message (960131).
54  *              Christian Daudt :       removed del_timer from
55  *                                      igmp_timer_expire function (960205).
56  *             Christian Daudt :       igmp_heard_report now only calls
57  *                                     igmp_timer_expire if tm->running is
58  *                                     true (960216).
59  *              Malcolm Beattie :       ttl comparison wrong in igmp_rcv made
60  *                                      igmp_heard_query never trigger. Expiry
61  *                                      miscalculation fixed in igmp_heard_query
62  *                                      and random() made to return unsigned to
63  *                                      prevent negative expiry times.
64  *              Alexey Kuznetsov:       Wrong group leaving behaviour, backport
65  *                                      fix from pending 2.1.x patches.
66  *              Alan Cox:               Forget to enable FDDI support earlier.
67  *              Alexey Kuznetsov:       Fixed leaving groups on device down.
68  *              Alexey Kuznetsov:       Accordance to igmp-v2-06 draft.
69  *              David L Stevens:        IGMPv3 support, with help from
70  *                                      Vinay Kulkarni
71  */
72
73 #include <linux/module.h>
74 #include <linux/slab.h>
75 #include <asm/uaccess.h>
76 #include <asm/system.h>
77 #include <linux/types.h>
78 #include <linux/kernel.h>
79 #include <linux/jiffies.h>
80 #include <linux/string.h>
81 #include <linux/socket.h>
82 #include <linux/sockios.h>
83 #include <linux/in.h>
84 #include <linux/inet.h>
85 #include <linux/netdevice.h>
86 #include <linux/skbuff.h>
87 #include <linux/inetdevice.h>
88 #include <linux/igmp.h>
89 #include <linux/if_arp.h>
90 #include <linux/rtnetlink.h>
91 #include <linux/times.h>
92
93 #include <net/net_namespace.h>
94 #include <net/arp.h>
95 #include <net/ip.h>
96 #include <net/protocol.h>
97 #include <net/route.h>
98 #include <net/sock.h>
99 #include <net/checksum.h>
100 #include <linux/netfilter_ipv4.h>
101 #ifdef CONFIG_IP_MROUTE
102 #include <linux/mroute.h>
103 #endif
104 #ifdef CONFIG_PROC_FS
105 #include <linux/proc_fs.h>
106 #include <linux/seq_file.h>
107 #endif
108
109 #define IP_MAX_MEMBERSHIPS      20
110 #define IP_MAX_MSF              10
111
112 #ifdef CONFIG_IP_MULTICAST
113 /* Parameter names and values are taken from igmp-v2-06 draft */
114
115 #define IGMP_V1_Router_Present_Timeout          (400*HZ)
116 #define IGMP_V2_Router_Present_Timeout          (400*HZ)
117 #define IGMP_Unsolicited_Report_Interval        (10*HZ)
118 #define IGMP_Query_Response_Interval            (10*HZ)
119 #define IGMP_Unsolicited_Report_Count           2
120
121
122 #define IGMP_Initial_Report_Delay               (1)
123
124 /* IGMP_Initial_Report_Delay is not from IGMP specs!
125  * IGMP specs require to report membership immediately after
126  * joining a group, but we delay the first report by a
127  * small interval. It seems more natural and still does not
128  * contradict to specs provided this delay is small enough.
129  */
130
131 #define IGMP_V1_SEEN(in_dev) \
132         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 1 || \
133          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 1 || \
134          ((in_dev)->mr_v1_seen && \
135           time_before(jiffies, (in_dev)->mr_v1_seen)))
136 #define IGMP_V2_SEEN(in_dev) \
137         (IPV4_DEVCONF_ALL(dev_net(in_dev->dev), FORCE_IGMP_VERSION) == 2 || \
138          IN_DEV_CONF_GET((in_dev), FORCE_IGMP_VERSION) == 2 || \
139          ((in_dev)->mr_v2_seen && \
140           time_before(jiffies, (in_dev)->mr_v2_seen)))
141
142 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im);
143 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr);
144 static void igmpv3_clear_delrec(struct in_device *in_dev);
145 static int sf_setstate(struct ip_mc_list *pmc);
146 static void sf_markstate(struct ip_mc_list *pmc);
147 #endif
148 static void ip_mc_clear_src(struct ip_mc_list *pmc);
149 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
150                          int sfcount, __be32 *psfsrc, int delta);
151
152
153 static void ip_mc_list_reclaim(struct rcu_head *head)
154 {
155         kfree(container_of(head, struct ip_mc_list, rcu));
156 }
157
158 static void ip_ma_put(struct ip_mc_list *im)
159 {
160         if (atomic_dec_and_test(&im->refcnt)) {
161                 in_dev_put(im->interface);
162                 call_rcu(&im->rcu, ip_mc_list_reclaim);
163         }
164 }
165
166 #define for_each_pmc_rcu(in_dev, pmc)                           \
167         for (pmc = rcu_dereference(in_dev->mc_list);            \
168              pmc != NULL;                                       \
169              pmc = rcu_dereference(pmc->next_rcu))
170
171 #define for_each_pmc_rtnl(in_dev, pmc)                          \
172         for (pmc = rtnl_dereference(in_dev->mc_list);           \
173              pmc != NULL;                                       \
174              pmc = rtnl_dereference(pmc->next_rcu))
175
176 #ifdef CONFIG_IP_MULTICAST
177
178 /*
179  *      Timer management
180  */
181
182 static void igmp_stop_timer(struct ip_mc_list *im)
183 {
184         spin_lock_bh(&im->lock);
185         if (del_timer(&im->timer))
186                 atomic_dec(&im->refcnt);
187         im->tm_running = 0;
188         im->reporter = 0;
189         im->unsolicit_count = 0;
190         spin_unlock_bh(&im->lock);
191 }
192
193 /* It must be called with locked im->lock */
194 static void igmp_start_timer(struct ip_mc_list *im, int max_delay)
195 {
196         int tv = net_random() % max_delay;
197
198         im->tm_running = 1;
199         if (!mod_timer(&im->timer, jiffies+tv+2))
200                 atomic_inc(&im->refcnt);
201 }
202
203 static void igmp_gq_start_timer(struct in_device *in_dev)
204 {
205         int tv = net_random() % in_dev->mr_maxdelay;
206
207         in_dev->mr_gq_running = 1;
208         if (!mod_timer(&in_dev->mr_gq_timer, jiffies+tv+2))
209                 in_dev_hold(in_dev);
210 }
211
212 static void igmp_ifc_start_timer(struct in_device *in_dev, int delay)
213 {
214         int tv = net_random() % delay;
215
216         if (!mod_timer(&in_dev->mr_ifc_timer, jiffies+tv+2))
217                 in_dev_hold(in_dev);
218 }
219
220 static void igmp_mod_timer(struct ip_mc_list *im, int max_delay)
221 {
222         spin_lock_bh(&im->lock);
223         im->unsolicit_count = 0;
224         if (del_timer(&im->timer)) {
225                 if ((long)(im->timer.expires-jiffies) < max_delay) {
226                         add_timer(&im->timer);
227                         im->tm_running = 1;
228                         spin_unlock_bh(&im->lock);
229                         return;
230                 }
231                 atomic_dec(&im->refcnt);
232         }
233         igmp_start_timer(im, max_delay);
234         spin_unlock_bh(&im->lock);
235 }
236
237
238 /*
239  *      Send an IGMP report.
240  */
241
242 #define IGMP_SIZE (sizeof(struct igmphdr)+sizeof(struct iphdr)+4)
243
244
245 static int is_in(struct ip_mc_list *pmc, struct ip_sf_list *psf, int type,
246         int gdeleted, int sdeleted)
247 {
248         switch (type) {
249         case IGMPV3_MODE_IS_INCLUDE:
250         case IGMPV3_MODE_IS_EXCLUDE:
251                 if (gdeleted || sdeleted)
252                         return 0;
253                 if (!(pmc->gsquery && !psf->sf_gsresp)) {
254                         if (pmc->sfmode == MCAST_INCLUDE)
255                                 return 1;
256                         /* don't include if this source is excluded
257                          * in all filters
258                          */
259                         if (psf->sf_count[MCAST_INCLUDE])
260                                 return type == IGMPV3_MODE_IS_INCLUDE;
261                         return pmc->sfcount[MCAST_EXCLUDE] ==
262                                 psf->sf_count[MCAST_EXCLUDE];
263                 }
264                 return 0;
265         case IGMPV3_CHANGE_TO_INCLUDE:
266                 if (gdeleted || sdeleted)
267                         return 0;
268                 return psf->sf_count[MCAST_INCLUDE] != 0;
269         case IGMPV3_CHANGE_TO_EXCLUDE:
270                 if (gdeleted || sdeleted)
271                         return 0;
272                 if (pmc->sfcount[MCAST_EXCLUDE] == 0 ||
273                     psf->sf_count[MCAST_INCLUDE])
274                         return 0;
275                 return pmc->sfcount[MCAST_EXCLUDE] ==
276                         psf->sf_count[MCAST_EXCLUDE];
277         case IGMPV3_ALLOW_NEW_SOURCES:
278                 if (gdeleted || !psf->sf_crcount)
279                         return 0;
280                 return (pmc->sfmode == MCAST_INCLUDE) ^ sdeleted;
281         case IGMPV3_BLOCK_OLD_SOURCES:
282                 if (pmc->sfmode == MCAST_INCLUDE)
283                         return gdeleted || (psf->sf_crcount && sdeleted);
284                 return psf->sf_crcount && !gdeleted && !sdeleted;
285         }
286         return 0;
287 }
288
289 static int
290 igmp_scount(struct ip_mc_list *pmc, int type, int gdeleted, int sdeleted)
291 {
292         struct ip_sf_list *psf;
293         int scount = 0;
294
295         for (psf=pmc->sources; psf; psf=psf->sf_next) {
296                 if (!is_in(pmc, psf, type, gdeleted, sdeleted))
297                         continue;
298                 scount++;
299         }
300         return scount;
301 }
302
303 #define igmp_skb_size(skb) (*(unsigned int *)((skb)->cb))
304
305 static struct sk_buff *igmpv3_newpack(struct net_device *dev, int size)
306 {
307         struct sk_buff *skb;
308         struct rtable *rt;
309         struct iphdr *pip;
310         struct igmpv3_report *pig;
311         struct net *net = dev_net(dev);
312
313         while (1) {
314                 skb = alloc_skb(size + LL_ALLOCATED_SPACE(dev),
315                                 GFP_ATOMIC | __GFP_NOWARN);
316                 if (skb)
317                         break;
318                 size >>= 1;
319                 if (size < 256)
320                         return NULL;
321         }
322         igmp_skb_size(skb) = size;
323
324         {
325                 struct flowi fl = { .oif = dev->ifindex,
326                                     .fl4_dst = IGMPV3_ALL_MCR,
327                                     .proto = IPPROTO_IGMP };
328                 if (ip_route_output_key(net, &rt, &fl)) {
329                         kfree_skb(skb);
330                         return NULL;
331                 }
332         }
333         if (rt->rt_src == 0) {
334                 kfree_skb(skb);
335                 ip_rt_put(rt);
336                 return NULL;
337         }
338
339         skb_dst_set(skb, &rt->dst);
340         skb->dev = dev;
341
342         skb_reserve(skb, LL_RESERVED_SPACE(dev));
343
344         skb_reset_network_header(skb);
345         pip = ip_hdr(skb);
346         skb_put(skb, sizeof(struct iphdr) + 4);
347
348         pip->version  = 4;
349         pip->ihl      = (sizeof(struct iphdr)+4)>>2;
350         pip->tos      = 0xc0;
351         pip->frag_off = htons(IP_DF);
352         pip->ttl      = 1;
353         pip->daddr    = rt->rt_dst;
354         pip->saddr    = rt->rt_src;
355         pip->protocol = IPPROTO_IGMP;
356         pip->tot_len  = 0;      /* filled in later */
357         ip_select_ident(pip, &rt->dst, NULL);
358         ((u8*)&pip[1])[0] = IPOPT_RA;
359         ((u8*)&pip[1])[1] = 4;
360         ((u8*)&pip[1])[2] = 0;
361         ((u8*)&pip[1])[3] = 0;
362
363         skb->transport_header = skb->network_header + sizeof(struct iphdr) + 4;
364         skb_put(skb, sizeof(*pig));
365         pig = igmpv3_report_hdr(skb);
366         pig->type = IGMPV3_HOST_MEMBERSHIP_REPORT;
367         pig->resv1 = 0;
368         pig->csum = 0;
369         pig->resv2 = 0;
370         pig->ngrec = 0;
371         return skb;
372 }
373
374 static int igmpv3_sendpack(struct sk_buff *skb)
375 {
376         struct igmphdr *pig = igmp_hdr(skb);
377         const int igmplen = skb->tail - skb->transport_header;
378
379         pig->csum = ip_compute_csum(igmp_hdr(skb), igmplen);
380
381         return ip_local_out(skb);
382 }
383
384 static int grec_size(struct ip_mc_list *pmc, int type, int gdel, int sdel)
385 {
386         return sizeof(struct igmpv3_grec) + 4*igmp_scount(pmc, type, gdel, sdel);
387 }
388
389 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ip_mc_list *pmc,
390         int type, struct igmpv3_grec **ppgr)
391 {
392         struct net_device *dev = pmc->interface->dev;
393         struct igmpv3_report *pih;
394         struct igmpv3_grec *pgr;
395
396         if (!skb)
397                 skb = igmpv3_newpack(dev, dev->mtu);
398         if (!skb)
399                 return NULL;
400         pgr = (struct igmpv3_grec *)skb_put(skb, sizeof(struct igmpv3_grec));
401         pgr->grec_type = type;
402         pgr->grec_auxwords = 0;
403         pgr->grec_nsrcs = 0;
404         pgr->grec_mca = pmc->multiaddr;
405         pih = igmpv3_report_hdr(skb);
406         pih->ngrec = htons(ntohs(pih->ngrec)+1);
407         *ppgr = pgr;
408         return skb;
409 }
410
411 #define AVAILABLE(skb) ((skb) ? ((skb)->dev ? igmp_skb_size(skb) - (skb)->len : \
412         skb_tailroom(skb)) : 0)
413
414 static struct sk_buff *add_grec(struct sk_buff *skb, struct ip_mc_list *pmc,
415         int type, int gdeleted, int sdeleted)
416 {
417         struct net_device *dev = pmc->interface->dev;
418         struct igmpv3_report *pih;
419         struct igmpv3_grec *pgr = NULL;
420         struct ip_sf_list *psf, *psf_next, *psf_prev, **psf_list;
421         int scount, stotal, first, isquery, truncate;
422
423         if (pmc->multiaddr == IGMP_ALL_HOSTS)
424                 return skb;
425
426         isquery = type == IGMPV3_MODE_IS_INCLUDE ||
427                   type == IGMPV3_MODE_IS_EXCLUDE;
428         truncate = type == IGMPV3_MODE_IS_EXCLUDE ||
429                     type == IGMPV3_CHANGE_TO_EXCLUDE;
430
431         stotal = scount = 0;
432
433         psf_list = sdeleted ? &pmc->tomb : &pmc->sources;
434
435         if (!*psf_list)
436                 goto empty_source;
437
438         pih = skb ? igmpv3_report_hdr(skb) : NULL;
439
440         /* EX and TO_EX get a fresh packet, if needed */
441         if (truncate) {
442                 if (pih && pih->ngrec &&
443                     AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
444                         if (skb)
445                                 igmpv3_sendpack(skb);
446                         skb = igmpv3_newpack(dev, dev->mtu);
447                 }
448         }
449         first = 1;
450         psf_prev = NULL;
451         for (psf=*psf_list; psf; psf=psf_next) {
452                 __be32 *psrc;
453
454                 psf_next = psf->sf_next;
455
456                 if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
457                         psf_prev = psf;
458                         continue;
459                 }
460
461                 /* clear marks on query responses */
462                 if (isquery)
463                         psf->sf_gsresp = 0;
464
465                 if (AVAILABLE(skb) < sizeof(__be32) +
466                     first*sizeof(struct igmpv3_grec)) {
467                         if (truncate && !first)
468                                 break;   /* truncate these */
469                         if (pgr)
470                                 pgr->grec_nsrcs = htons(scount);
471                         if (skb)
472                                 igmpv3_sendpack(skb);
473                         skb = igmpv3_newpack(dev, dev->mtu);
474                         first = 1;
475                         scount = 0;
476                 }
477                 if (first) {
478                         skb = add_grhead(skb, pmc, type, &pgr);
479                         first = 0;
480                 }
481                 if (!skb)
482                         return NULL;
483                 psrc = (__be32 *)skb_put(skb, sizeof(__be32));
484                 *psrc = psf->sf_inaddr;
485                 scount++; stotal++;
486                 if ((type == IGMPV3_ALLOW_NEW_SOURCES ||
487                      type == IGMPV3_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
488                         psf->sf_crcount--;
489                         if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
490                                 if (psf_prev)
491                                         psf_prev->sf_next = psf->sf_next;
492                                 else
493                                         *psf_list = psf->sf_next;
494                                 kfree(psf);
495                                 continue;
496                         }
497                 }
498                 psf_prev = psf;
499         }
500
501 empty_source:
502         if (!stotal) {
503                 if (type == IGMPV3_ALLOW_NEW_SOURCES ||
504                     type == IGMPV3_BLOCK_OLD_SOURCES)
505                         return skb;
506                 if (pmc->crcount || isquery) {
507                         /* make sure we have room for group header */
508                         if (skb && AVAILABLE(skb)<sizeof(struct igmpv3_grec)) {
509                                 igmpv3_sendpack(skb);
510                                 skb = NULL; /* add_grhead will get a new one */
511                         }
512                         skb = add_grhead(skb, pmc, type, &pgr);
513                 }
514         }
515         if (pgr)
516                 pgr->grec_nsrcs = htons(scount);
517
518         if (isquery)
519                 pmc->gsquery = 0;       /* clear query state on report */
520         return skb;
521 }
522
523 static int igmpv3_send_report(struct in_device *in_dev, struct ip_mc_list *pmc)
524 {
525         struct sk_buff *skb = NULL;
526         int type;
527
528         if (!pmc) {
529                 rcu_read_lock();
530                 for_each_pmc_rcu(in_dev, pmc) {
531                         if (pmc->multiaddr == IGMP_ALL_HOSTS)
532                                 continue;
533                         spin_lock_bh(&pmc->lock);
534                         if (pmc->sfcount[MCAST_EXCLUDE])
535                                 type = IGMPV3_MODE_IS_EXCLUDE;
536                         else
537                                 type = IGMPV3_MODE_IS_INCLUDE;
538                         skb = add_grec(skb, pmc, type, 0, 0);
539                         spin_unlock_bh(&pmc->lock);
540                 }
541                 rcu_read_unlock();
542         } else {
543                 spin_lock_bh(&pmc->lock);
544                 if (pmc->sfcount[MCAST_EXCLUDE])
545                         type = IGMPV3_MODE_IS_EXCLUDE;
546                 else
547                         type = IGMPV3_MODE_IS_INCLUDE;
548                 skb = add_grec(skb, pmc, type, 0, 0);
549                 spin_unlock_bh(&pmc->lock);
550         }
551         if (!skb)
552                 return 0;
553         return igmpv3_sendpack(skb);
554 }
555
556 /*
557  * remove zero-count source records from a source filter list
558  */
559 static void igmpv3_clear_zeros(struct ip_sf_list **ppsf)
560 {
561         struct ip_sf_list *psf_prev, *psf_next, *psf;
562
563         psf_prev = NULL;
564         for (psf=*ppsf; psf; psf = psf_next) {
565                 psf_next = psf->sf_next;
566                 if (psf->sf_crcount == 0) {
567                         if (psf_prev)
568                                 psf_prev->sf_next = psf->sf_next;
569                         else
570                                 *ppsf = psf->sf_next;
571                         kfree(psf);
572                 } else
573                         psf_prev = psf;
574         }
575 }
576
577 static void igmpv3_send_cr(struct in_device *in_dev)
578 {
579         struct ip_mc_list *pmc, *pmc_prev, *pmc_next;
580         struct sk_buff *skb = NULL;
581         int type, dtype;
582
583         rcu_read_lock();
584         spin_lock_bh(&in_dev->mc_tomb_lock);
585
586         /* deleted MCA's */
587         pmc_prev = NULL;
588         for (pmc=in_dev->mc_tomb; pmc; pmc=pmc_next) {
589                 pmc_next = pmc->next;
590                 if (pmc->sfmode == MCAST_INCLUDE) {
591                         type = IGMPV3_BLOCK_OLD_SOURCES;
592                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
593                         skb = add_grec(skb, pmc, type, 1, 0);
594                         skb = add_grec(skb, pmc, dtype, 1, 1);
595                 }
596                 if (pmc->crcount) {
597                         if (pmc->sfmode == MCAST_EXCLUDE) {
598                                 type = IGMPV3_CHANGE_TO_INCLUDE;
599                                 skb = add_grec(skb, pmc, type, 1, 0);
600                         }
601                         pmc->crcount--;
602                         if (pmc->crcount == 0) {
603                                 igmpv3_clear_zeros(&pmc->tomb);
604                                 igmpv3_clear_zeros(&pmc->sources);
605                         }
606                 }
607                 if (pmc->crcount == 0 && !pmc->tomb && !pmc->sources) {
608                         if (pmc_prev)
609                                 pmc_prev->next = pmc_next;
610                         else
611                                 in_dev->mc_tomb = pmc_next;
612                         in_dev_put(pmc->interface);
613                         kfree(pmc);
614                 } else
615                         pmc_prev = pmc;
616         }
617         spin_unlock_bh(&in_dev->mc_tomb_lock);
618
619         /* change recs */
620         for_each_pmc_rcu(in_dev, pmc) {
621                 spin_lock_bh(&pmc->lock);
622                 if (pmc->sfcount[MCAST_EXCLUDE]) {
623                         type = IGMPV3_BLOCK_OLD_SOURCES;
624                         dtype = IGMPV3_ALLOW_NEW_SOURCES;
625                 } else {
626                         type = IGMPV3_ALLOW_NEW_SOURCES;
627                         dtype = IGMPV3_BLOCK_OLD_SOURCES;
628                 }
629                 skb = add_grec(skb, pmc, type, 0, 0);
630                 skb = add_grec(skb, pmc, dtype, 0, 1);  /* deleted sources */
631
632                 /* filter mode changes */
633                 if (pmc->crcount) {
634                         if (pmc->sfmode == MCAST_EXCLUDE)
635                                 type = IGMPV3_CHANGE_TO_EXCLUDE;
636                         else
637                                 type = IGMPV3_CHANGE_TO_INCLUDE;
638                         skb = add_grec(skb, pmc, type, 0, 0);
639                         pmc->crcount--;
640                 }
641                 spin_unlock_bh(&pmc->lock);
642         }
643         rcu_read_unlock();
644
645         if (!skb)
646                 return;
647         (void) igmpv3_sendpack(skb);
648 }
649
650 static int igmp_send_report(struct in_device *in_dev, struct ip_mc_list *pmc,
651         int type)
652 {
653         struct sk_buff *skb;
654         struct iphdr *iph;
655         struct igmphdr *ih;
656         struct rtable *rt;
657         struct net_device *dev = in_dev->dev;
658         struct net *net = dev_net(dev);
659         __be32  group = pmc ? pmc->multiaddr : 0;
660         __be32  dst;
661
662         if (type == IGMPV3_HOST_MEMBERSHIP_REPORT)
663                 return igmpv3_send_report(in_dev, pmc);
664         else if (type == IGMP_HOST_LEAVE_MESSAGE)
665                 dst = IGMP_ALL_ROUTER;
666         else
667                 dst = group;
668
669         {
670                 struct flowi fl = { .oif = dev->ifindex,
671                                     .fl4_dst = dst,
672                                     .proto = IPPROTO_IGMP };
673                 if (ip_route_output_key(net, &rt, &fl))
674                         return -1;
675         }
676         if (rt->rt_src == 0) {
677                 ip_rt_put(rt);
678                 return -1;
679         }
680
681         skb = alloc_skb(IGMP_SIZE+LL_ALLOCATED_SPACE(dev), GFP_ATOMIC);
682         if (skb == NULL) {
683                 ip_rt_put(rt);
684                 return -1;
685         }
686
687         skb_dst_set(skb, &rt->dst);
688
689         skb_reserve(skb, LL_RESERVED_SPACE(dev));
690
691         skb_reset_network_header(skb);
692         iph = ip_hdr(skb);
693         skb_put(skb, sizeof(struct iphdr) + 4);
694
695         iph->version  = 4;
696         iph->ihl      = (sizeof(struct iphdr)+4)>>2;
697         iph->tos      = 0xc0;
698         iph->frag_off = htons(IP_DF);
699         iph->ttl      = 1;
700         iph->daddr    = dst;
701         iph->saddr    = rt->rt_src;
702         iph->protocol = IPPROTO_IGMP;
703         ip_select_ident(iph, &rt->dst, NULL);
704         ((u8*)&iph[1])[0] = IPOPT_RA;
705         ((u8*)&iph[1])[1] = 4;
706         ((u8*)&iph[1])[2] = 0;
707         ((u8*)&iph[1])[3] = 0;
708
709         ih = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
710         ih->type = type;
711         ih->code = 0;
712         ih->csum = 0;
713         ih->group = group;
714         ih->csum = ip_compute_csum((void *)ih, sizeof(struct igmphdr));
715
716         return ip_local_out(skb);
717 }
718
719 static void igmp_gq_timer_expire(unsigned long data)
720 {
721         struct in_device *in_dev = (struct in_device *)data;
722
723         in_dev->mr_gq_running = 0;
724         igmpv3_send_report(in_dev, NULL);
725         __in_dev_put(in_dev);
726 }
727
728 static void igmp_ifc_timer_expire(unsigned long data)
729 {
730         struct in_device *in_dev = (struct in_device *)data;
731
732         igmpv3_send_cr(in_dev);
733         if (in_dev->mr_ifc_count) {
734                 in_dev->mr_ifc_count--;
735                 igmp_ifc_start_timer(in_dev, IGMP_Unsolicited_Report_Interval);
736         }
737         __in_dev_put(in_dev);
738 }
739
740 static void igmp_ifc_event(struct in_device *in_dev)
741 {
742         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev))
743                 return;
744         in_dev->mr_ifc_count = in_dev->mr_qrv ? in_dev->mr_qrv :
745                 IGMP_Unsolicited_Report_Count;
746         igmp_ifc_start_timer(in_dev, 1);
747 }
748
749
750 static void igmp_timer_expire(unsigned long data)
751 {
752         struct ip_mc_list *im=(struct ip_mc_list *)data;
753         struct in_device *in_dev = im->interface;
754
755         spin_lock(&im->lock);
756         im->tm_running = 0;
757
758         if (im->unsolicit_count) {
759                 im->unsolicit_count--;
760                 igmp_start_timer(im, IGMP_Unsolicited_Report_Interval);
761         }
762         im->reporter = 1;
763         spin_unlock(&im->lock);
764
765         if (IGMP_V1_SEEN(in_dev))
766                 igmp_send_report(in_dev, im, IGMP_HOST_MEMBERSHIP_REPORT);
767         else if (IGMP_V2_SEEN(in_dev))
768                 igmp_send_report(in_dev, im, IGMPV2_HOST_MEMBERSHIP_REPORT);
769         else
770                 igmp_send_report(in_dev, im, IGMPV3_HOST_MEMBERSHIP_REPORT);
771
772         ip_ma_put(im);
773 }
774
775 /* mark EXCLUDE-mode sources */
776 static int igmp_xmarksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
777 {
778         struct ip_sf_list *psf;
779         int i, scount;
780
781         scount = 0;
782         for (psf=pmc->sources; psf; psf=psf->sf_next) {
783                 if (scount == nsrcs)
784                         break;
785                 for (i=0; i<nsrcs; i++) {
786                         /* skip inactive filters */
787                         if (pmc->sfcount[MCAST_INCLUDE] ||
788                             pmc->sfcount[MCAST_EXCLUDE] !=
789                             psf->sf_count[MCAST_EXCLUDE])
790                                 continue;
791                         if (srcs[i] == psf->sf_inaddr) {
792                                 scount++;
793                                 break;
794                         }
795                 }
796         }
797         pmc->gsquery = 0;
798         if (scount == nsrcs)    /* all sources excluded */
799                 return 0;
800         return 1;
801 }
802
803 static int igmp_marksources(struct ip_mc_list *pmc, int nsrcs, __be32 *srcs)
804 {
805         struct ip_sf_list *psf;
806         int i, scount;
807
808         if (pmc->sfmode == MCAST_EXCLUDE)
809                 return igmp_xmarksources(pmc, nsrcs, srcs);
810
811         /* mark INCLUDE-mode sources */
812         scount = 0;
813         for (psf=pmc->sources; psf; psf=psf->sf_next) {
814                 if (scount == nsrcs)
815                         break;
816                 for (i=0; i<nsrcs; i++)
817                         if (srcs[i] == psf->sf_inaddr) {
818                                 psf->sf_gsresp = 1;
819                                 scount++;
820                                 break;
821                         }
822         }
823         if (!scount) {
824                 pmc->gsquery = 0;
825                 return 0;
826         }
827         pmc->gsquery = 1;
828         return 1;
829 }
830
831 static void igmp_heard_report(struct in_device *in_dev, __be32 group)
832 {
833         struct ip_mc_list *im;
834
835         /* Timers are only set for non-local groups */
836
837         if (group == IGMP_ALL_HOSTS)
838                 return;
839
840         rcu_read_lock();
841         for_each_pmc_rcu(in_dev, im) {
842                 if (im->multiaddr == group) {
843                         igmp_stop_timer(im);
844                         break;
845                 }
846         }
847         rcu_read_unlock();
848 }
849
850 static void igmp_heard_query(struct in_device *in_dev, struct sk_buff *skb,
851         int len)
852 {
853         struct igmphdr          *ih = igmp_hdr(skb);
854         struct igmpv3_query *ih3 = igmpv3_query_hdr(skb);
855         struct ip_mc_list       *im;
856         __be32                  group = ih->group;
857         int                     max_delay;
858         int                     mark = 0;
859
860
861         if (len == 8) {
862                 if (ih->code == 0) {
863                         /* Alas, old v1 router presents here. */
864
865                         max_delay = IGMP_Query_Response_Interval;
866                         in_dev->mr_v1_seen = jiffies +
867                                 IGMP_V1_Router_Present_Timeout;
868                         group = 0;
869                 } else {
870                         /* v2 router present */
871                         max_delay = ih->code*(HZ/IGMP_TIMER_SCALE);
872                         in_dev->mr_v2_seen = jiffies +
873                                 IGMP_V2_Router_Present_Timeout;
874                 }
875                 /* cancel the interface change timer */
876                 in_dev->mr_ifc_count = 0;
877                 if (del_timer(&in_dev->mr_ifc_timer))
878                         __in_dev_put(in_dev);
879                 /* clear deleted report items */
880                 igmpv3_clear_delrec(in_dev);
881         } else if (len < 12) {
882                 return; /* ignore bogus packet; freed by caller */
883         } else if (IGMP_V1_SEEN(in_dev)) {
884                 /* This is a v3 query with v1 queriers present */
885                 max_delay = IGMP_Query_Response_Interval;
886                 group = 0;
887         } else if (IGMP_V2_SEEN(in_dev)) {
888                 /* this is a v3 query with v2 queriers present;
889                  * Interpretation of the max_delay code is problematic here.
890                  * A real v2 host would use ih_code directly, while v3 has a
891                  * different encoding. We use the v3 encoding as more likely
892                  * to be intended in a v3 query.
893                  */
894                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
895         } else { /* v3 */
896                 if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)))
897                         return;
898
899                 ih3 = igmpv3_query_hdr(skb);
900                 if (ih3->nsrcs) {
901                         if (!pskb_may_pull(skb, sizeof(struct igmpv3_query)
902                                            + ntohs(ih3->nsrcs)*sizeof(__be32)))
903                                 return;
904                         ih3 = igmpv3_query_hdr(skb);
905                 }
906
907                 max_delay = IGMPV3_MRC(ih3->code)*(HZ/IGMP_TIMER_SCALE);
908                 if (!max_delay)
909                         max_delay = 1;  /* can't mod w/ 0 */
910                 in_dev->mr_maxdelay = max_delay;
911                 if (ih3->qrv)
912                         in_dev->mr_qrv = ih3->qrv;
913                 if (!group) { /* general query */
914                         if (ih3->nsrcs)
915                                 return; /* no sources allowed */
916                         igmp_gq_start_timer(in_dev);
917                         return;
918                 }
919                 /* mark sources to include, if group & source-specific */
920                 mark = ih3->nsrcs != 0;
921         }
922
923         /*
924          * - Start the timers in all of our membership records
925          *   that the query applies to for the interface on
926          *   which the query arrived excl. those that belong
927          *   to a "local" group (224.0.0.X)
928          * - For timers already running check if they need to
929          *   be reset.
930          * - Use the igmp->igmp_code field as the maximum
931          *   delay possible
932          */
933         rcu_read_lock();
934         for_each_pmc_rcu(in_dev, im) {
935                 int changed;
936
937                 if (group && group != im->multiaddr)
938                         continue;
939                 if (im->multiaddr == IGMP_ALL_HOSTS)
940                         continue;
941                 spin_lock_bh(&im->lock);
942                 if (im->tm_running)
943                         im->gsquery = im->gsquery && mark;
944                 else
945                         im->gsquery = mark;
946                 changed = !im->gsquery ||
947                         igmp_marksources(im, ntohs(ih3->nsrcs), ih3->srcs);
948                 spin_unlock_bh(&im->lock);
949                 if (changed)
950                         igmp_mod_timer(im, max_delay);
951         }
952         rcu_read_unlock();
953 }
954
955 /* called in rcu_read_lock() section */
956 int igmp_rcv(struct sk_buff *skb)
957 {
958         /* This basically follows the spec line by line -- see RFC1112 */
959         struct igmphdr *ih;
960         struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
961         int len = skb->len;
962
963         if (in_dev == NULL)
964                 goto drop;
965
966         if (!pskb_may_pull(skb, sizeof(struct igmphdr)))
967                 goto drop;
968
969         switch (skb->ip_summed) {
970         case CHECKSUM_COMPLETE:
971                 if (!csum_fold(skb->csum))
972                         break;
973                 /* fall through */
974         case CHECKSUM_NONE:
975                 skb->csum = 0;
976                 if (__skb_checksum_complete(skb))
977                         goto drop;
978         }
979
980         ih = igmp_hdr(skb);
981         switch (ih->type) {
982         case IGMP_HOST_MEMBERSHIP_QUERY:
983                 igmp_heard_query(in_dev, skb, len);
984                 break;
985         case IGMP_HOST_MEMBERSHIP_REPORT:
986         case IGMPV2_HOST_MEMBERSHIP_REPORT:
987                 /* Is it our report looped back? */
988                 if (rt_is_output_route(skb_rtable(skb)))
989                         break;
990                 /* don't rely on MC router hearing unicast reports */
991                 if (skb->pkt_type == PACKET_MULTICAST ||
992                     skb->pkt_type == PACKET_BROADCAST)
993                         igmp_heard_report(in_dev, ih->group);
994                 break;
995         case IGMP_PIM:
996 #ifdef CONFIG_IP_PIMSM_V1
997                 return pim_rcv_v1(skb);
998 #endif
999         case IGMPV3_HOST_MEMBERSHIP_REPORT:
1000         case IGMP_DVMRP:
1001         case IGMP_TRACE:
1002         case IGMP_HOST_LEAVE_MESSAGE:
1003         case IGMP_MTRACE:
1004         case IGMP_MTRACE_RESP:
1005                 break;
1006         default:
1007                 break;
1008         }
1009
1010 drop:
1011         kfree_skb(skb);
1012         return 0;
1013 }
1014
1015 #endif
1016
1017
1018 /*
1019  *      Add a filter to a device
1020  */
1021
1022 static void ip_mc_filter_add(struct in_device *in_dev, __be32 addr)
1023 {
1024         char buf[MAX_ADDR_LEN];
1025         struct net_device *dev = in_dev->dev;
1026
1027         /* Checking for IFF_MULTICAST here is WRONG-WRONG-WRONG.
1028            We will get multicast token leakage, when IFF_MULTICAST
1029            is changed. This check should be done in dev->set_multicast_list
1030            routine. Something sort of:
1031            if (dev->mc_list && dev->flags&IFF_MULTICAST) { do it; }
1032            --ANK
1033            */
1034         if (arp_mc_map(addr, buf, dev, 0) == 0)
1035                 dev_mc_add(dev, buf);
1036 }
1037
1038 /*
1039  *      Remove a filter from a device
1040  */
1041
1042 static void ip_mc_filter_del(struct in_device *in_dev, __be32 addr)
1043 {
1044         char buf[MAX_ADDR_LEN];
1045         struct net_device *dev = in_dev->dev;
1046
1047         if (arp_mc_map(addr, buf, dev, 0) == 0)
1048                 dev_mc_del(dev, buf);
1049 }
1050
1051 #ifdef CONFIG_IP_MULTICAST
1052 /*
1053  * deleted ip_mc_list manipulation
1054  */
1055 static void igmpv3_add_delrec(struct in_device *in_dev, struct ip_mc_list *im)
1056 {
1057         struct ip_mc_list *pmc;
1058
1059         /* this is an "ip_mc_list" for convenience; only the fields below
1060          * are actually used. In particular, the refcnt and users are not
1061          * used for management of the delete list. Using the same structure
1062          * for deleted items allows change reports to use common code with
1063          * non-deleted or query-response MCA's.
1064          */
1065         pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
1066         if (!pmc)
1067                 return;
1068         spin_lock_bh(&im->lock);
1069         pmc->interface = im->interface;
1070         in_dev_hold(in_dev);
1071         pmc->multiaddr = im->multiaddr;
1072         pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1073                 IGMP_Unsolicited_Report_Count;
1074         pmc->sfmode = im->sfmode;
1075         if (pmc->sfmode == MCAST_INCLUDE) {
1076                 struct ip_sf_list *psf;
1077
1078                 pmc->tomb = im->tomb;
1079                 pmc->sources = im->sources;
1080                 im->tomb = im->sources = NULL;
1081                 for (psf=pmc->sources; psf; psf=psf->sf_next)
1082                         psf->sf_crcount = pmc->crcount;
1083         }
1084         spin_unlock_bh(&im->lock);
1085
1086         spin_lock_bh(&in_dev->mc_tomb_lock);
1087         pmc->next = in_dev->mc_tomb;
1088         in_dev->mc_tomb = pmc;
1089         spin_unlock_bh(&in_dev->mc_tomb_lock);
1090 }
1091
1092 static void igmpv3_del_delrec(struct in_device *in_dev, __be32 multiaddr)
1093 {
1094         struct ip_mc_list *pmc, *pmc_prev;
1095         struct ip_sf_list *psf, *psf_next;
1096
1097         spin_lock_bh(&in_dev->mc_tomb_lock);
1098         pmc_prev = NULL;
1099         for (pmc=in_dev->mc_tomb; pmc; pmc=pmc->next) {
1100                 if (pmc->multiaddr == multiaddr)
1101                         break;
1102                 pmc_prev = pmc;
1103         }
1104         if (pmc) {
1105                 if (pmc_prev)
1106                         pmc_prev->next = pmc->next;
1107                 else
1108                         in_dev->mc_tomb = pmc->next;
1109         }
1110         spin_unlock_bh(&in_dev->mc_tomb_lock);
1111         if (pmc) {
1112                 for (psf=pmc->tomb; psf; psf=psf_next) {
1113                         psf_next = psf->sf_next;
1114                         kfree(psf);
1115                 }
1116                 in_dev_put(pmc->interface);
1117                 kfree(pmc);
1118         }
1119 }
1120
1121 static void igmpv3_clear_delrec(struct in_device *in_dev)
1122 {
1123         struct ip_mc_list *pmc, *nextpmc;
1124
1125         spin_lock_bh(&in_dev->mc_tomb_lock);
1126         pmc = in_dev->mc_tomb;
1127         in_dev->mc_tomb = NULL;
1128         spin_unlock_bh(&in_dev->mc_tomb_lock);
1129
1130         for (; pmc; pmc = nextpmc) {
1131                 nextpmc = pmc->next;
1132                 ip_mc_clear_src(pmc);
1133                 in_dev_put(pmc->interface);
1134                 kfree(pmc);
1135         }
1136         /* clear dead sources, too */
1137         rcu_read_lock();
1138         for_each_pmc_rcu(in_dev, pmc) {
1139                 struct ip_sf_list *psf, *psf_next;
1140
1141                 spin_lock_bh(&pmc->lock);
1142                 psf = pmc->tomb;
1143                 pmc->tomb = NULL;
1144                 spin_unlock_bh(&pmc->lock);
1145                 for (; psf; psf=psf_next) {
1146                         psf_next = psf->sf_next;
1147                         kfree(psf);
1148                 }
1149         }
1150         rcu_read_unlock();
1151 }
1152 #endif
1153
1154 static void igmp_group_dropped(struct ip_mc_list *im)
1155 {
1156         struct in_device *in_dev = im->interface;
1157 #ifdef CONFIG_IP_MULTICAST
1158         int reporter;
1159 #endif
1160
1161         if (im->loaded) {
1162                 im->loaded = 0;
1163                 ip_mc_filter_del(in_dev, im->multiaddr);
1164         }
1165
1166 #ifdef CONFIG_IP_MULTICAST
1167         if (im->multiaddr == IGMP_ALL_HOSTS)
1168                 return;
1169
1170         reporter = im->reporter;
1171         igmp_stop_timer(im);
1172
1173         if (!in_dev->dead) {
1174                 if (IGMP_V1_SEEN(in_dev))
1175                         goto done;
1176                 if (IGMP_V2_SEEN(in_dev)) {
1177                         if (reporter)
1178                                 igmp_send_report(in_dev, im, IGMP_HOST_LEAVE_MESSAGE);
1179                         goto done;
1180                 }
1181                 /* IGMPv3 */
1182                 igmpv3_add_delrec(in_dev, im);
1183
1184                 igmp_ifc_event(in_dev);
1185         }
1186 done:
1187 #endif
1188         ip_mc_clear_src(im);
1189 }
1190
1191 static void igmp_group_added(struct ip_mc_list *im)
1192 {
1193         struct in_device *in_dev = im->interface;
1194
1195         if (im->loaded == 0) {
1196                 im->loaded = 1;
1197                 ip_mc_filter_add(in_dev, im->multiaddr);
1198         }
1199
1200 #ifdef CONFIG_IP_MULTICAST
1201         if (im->multiaddr == IGMP_ALL_HOSTS)
1202                 return;
1203
1204         if (in_dev->dead)
1205                 return;
1206         if (IGMP_V1_SEEN(in_dev) || IGMP_V2_SEEN(in_dev)) {
1207                 spin_lock_bh(&im->lock);
1208                 igmp_start_timer(im, IGMP_Initial_Report_Delay);
1209                 spin_unlock_bh(&im->lock);
1210                 return;
1211         }
1212         /* else, v3 */
1213
1214         im->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1215                 IGMP_Unsolicited_Report_Count;
1216         igmp_ifc_event(in_dev);
1217 #endif
1218 }
1219
1220
1221 /*
1222  *      Multicast list managers
1223  */
1224
1225
1226 /*
1227  *      A socket has joined a multicast group on device dev.
1228  */
1229
1230 void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
1231 {
1232         struct ip_mc_list *im;
1233
1234         ASSERT_RTNL();
1235
1236         for_each_pmc_rtnl(in_dev, im) {
1237                 if (im->multiaddr == addr) {
1238                         im->users++;
1239                         ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
1240                         goto out;
1241                 }
1242         }
1243
1244         im = kzalloc(sizeof(*im), GFP_KERNEL);
1245         if (!im)
1246                 goto out;
1247
1248         im->users = 1;
1249         im->interface = in_dev;
1250         in_dev_hold(in_dev);
1251         im->multiaddr = addr;
1252         /* initial mode is (EX, empty) */
1253         im->sfmode = MCAST_EXCLUDE;
1254         im->sfcount[MCAST_EXCLUDE] = 1;
1255         atomic_set(&im->refcnt, 1);
1256         spin_lock_init(&im->lock);
1257 #ifdef CONFIG_IP_MULTICAST
1258         setup_timer(&im->timer, &igmp_timer_expire, (unsigned long)im);
1259         im->unsolicit_count = IGMP_Unsolicited_Report_Count;
1260 #endif
1261
1262         im->next_rcu = in_dev->mc_list;
1263         in_dev->mc_count++;
1264         rcu_assign_pointer(in_dev->mc_list, im);
1265
1266 #ifdef CONFIG_IP_MULTICAST
1267         igmpv3_del_delrec(in_dev, im->multiaddr);
1268 #endif
1269         igmp_group_added(im);
1270         if (!in_dev->dead)
1271                 ip_rt_multicast_event(in_dev);
1272 out:
1273         return;
1274 }
1275 EXPORT_SYMBOL(ip_mc_inc_group);
1276
1277 /*
1278  *      Resend IGMP JOIN report; used for bonding.
1279  *      Called with rcu_read_lock()
1280  */
1281 void ip_mc_rejoin_groups(struct in_device *in_dev)
1282 {
1283 #ifdef CONFIG_IP_MULTICAST
1284         struct ip_mc_list *im;
1285         int type;
1286
1287         for_each_pmc_rcu(in_dev, im) {
1288                 if (im->multiaddr == IGMP_ALL_HOSTS)
1289                         continue;
1290
1291                 /* a failover is happening and switches
1292                  * must be notified immediately
1293                  */
1294                 if (IGMP_V1_SEEN(in_dev))
1295                         type = IGMP_HOST_MEMBERSHIP_REPORT;
1296                 else if (IGMP_V2_SEEN(in_dev))
1297                         type = IGMPV2_HOST_MEMBERSHIP_REPORT;
1298                 else
1299                         type = IGMPV3_HOST_MEMBERSHIP_REPORT;
1300                 igmp_send_report(in_dev, im, type);
1301         }
1302 #endif
1303 }
1304 EXPORT_SYMBOL(ip_mc_rejoin_groups);
1305
1306 /*
1307  *      A socket has left a multicast group on device dev
1308  */
1309
1310 void ip_mc_dec_group(struct in_device *in_dev, __be32 addr)
1311 {
1312         struct ip_mc_list *i;
1313         struct ip_mc_list __rcu **ip;
1314
1315         ASSERT_RTNL();
1316
1317         for (ip = &in_dev->mc_list;
1318              (i = rtnl_dereference(*ip)) != NULL;
1319              ip = &i->next_rcu) {
1320                 if (i->multiaddr == addr) {
1321                         if (--i->users == 0) {
1322                                 *ip = i->next_rcu;
1323                                 in_dev->mc_count--;
1324                                 igmp_group_dropped(i);
1325
1326                                 if (!in_dev->dead)
1327                                         ip_rt_multicast_event(in_dev);
1328
1329                                 ip_ma_put(i);
1330                                 return;
1331                         }
1332                         break;
1333                 }
1334         }
1335 }
1336 EXPORT_SYMBOL(ip_mc_dec_group);
1337
1338 /* Device changing type */
1339
1340 void ip_mc_unmap(struct in_device *in_dev)
1341 {
1342         struct ip_mc_list *pmc;
1343
1344         ASSERT_RTNL();
1345
1346         for_each_pmc_rtnl(in_dev, pmc)
1347                 igmp_group_dropped(pmc);
1348 }
1349
1350 void ip_mc_remap(struct in_device *in_dev)
1351 {
1352         struct ip_mc_list *pmc;
1353
1354         ASSERT_RTNL();
1355
1356         for_each_pmc_rtnl(in_dev, pmc)
1357                 igmp_group_added(pmc);
1358 }
1359
1360 /* Device going down */
1361
1362 void ip_mc_down(struct in_device *in_dev)
1363 {
1364         struct ip_mc_list *pmc;
1365
1366         ASSERT_RTNL();
1367
1368         for_each_pmc_rtnl(in_dev, pmc)
1369                 igmp_group_dropped(pmc);
1370
1371 #ifdef CONFIG_IP_MULTICAST
1372         in_dev->mr_ifc_count = 0;
1373         if (del_timer(&in_dev->mr_ifc_timer))
1374                 __in_dev_put(in_dev);
1375         in_dev->mr_gq_running = 0;
1376         if (del_timer(&in_dev->mr_gq_timer))
1377                 __in_dev_put(in_dev);
1378         igmpv3_clear_delrec(in_dev);
1379 #endif
1380
1381         ip_mc_dec_group(in_dev, IGMP_ALL_HOSTS);
1382 }
1383
1384 void ip_mc_init_dev(struct in_device *in_dev)
1385 {
1386         ASSERT_RTNL();
1387
1388         in_dev->mc_tomb = NULL;
1389 #ifdef CONFIG_IP_MULTICAST
1390         in_dev->mr_gq_running = 0;
1391         setup_timer(&in_dev->mr_gq_timer, igmp_gq_timer_expire,
1392                         (unsigned long)in_dev);
1393         in_dev->mr_ifc_count = 0;
1394         in_dev->mc_count     = 0;
1395         setup_timer(&in_dev->mr_ifc_timer, igmp_ifc_timer_expire,
1396                         (unsigned long)in_dev);
1397         in_dev->mr_qrv = IGMP_Unsolicited_Report_Count;
1398 #endif
1399
1400         spin_lock_init(&in_dev->mc_tomb_lock);
1401 }
1402
1403 /* Device going up */
1404
1405 void ip_mc_up(struct in_device *in_dev)
1406 {
1407         struct ip_mc_list *pmc;
1408
1409         ASSERT_RTNL();
1410
1411         ip_mc_inc_group(in_dev, IGMP_ALL_HOSTS);
1412
1413         for_each_pmc_rtnl(in_dev, pmc)
1414                 igmp_group_added(pmc);
1415 }
1416
1417 /*
1418  *      Device is about to be destroyed: clean up.
1419  */
1420
1421 void ip_mc_destroy_dev(struct in_device *in_dev)
1422 {
1423         struct ip_mc_list *i;
1424
1425         ASSERT_RTNL();
1426
1427         /* Deactivate timers */
1428         ip_mc_down(in_dev);
1429
1430         while ((i = rtnl_dereference(in_dev->mc_list)) != NULL) {
1431                 in_dev->mc_list = i->next_rcu;
1432                 in_dev->mc_count--;
1433
1434                 igmp_group_dropped(i);
1435                 ip_ma_put(i);
1436         }
1437 }
1438
1439 /* RTNL is locked */
1440 static struct in_device *ip_mc_find_dev(struct net *net, struct ip_mreqn *imr)
1441 {
1442         struct flowi fl = { .fl4_dst = imr->imr_multiaddr.s_addr };
1443         struct rtable *rt;
1444         struct net_device *dev = NULL;
1445         struct in_device *idev = NULL;
1446
1447         if (imr->imr_ifindex) {
1448                 idev = inetdev_by_index(net, imr->imr_ifindex);
1449                 return idev;
1450         }
1451         if (imr->imr_address.s_addr) {
1452                 dev = __ip_dev_find(net, imr->imr_address.s_addr, false);
1453                 if (!dev)
1454                         return NULL;
1455         }
1456
1457         if (!dev && !ip_route_output_key(net, &rt, &fl)) {
1458                 dev = rt->dst.dev;
1459                 ip_rt_put(rt);
1460         }
1461         if (dev) {
1462                 imr->imr_ifindex = dev->ifindex;
1463                 idev = __in_dev_get_rtnl(dev);
1464         }
1465         return idev;
1466 }
1467
1468 /*
1469  *      Join a socket to a group
1470  */
1471 int sysctl_igmp_max_memberships __read_mostly = IP_MAX_MEMBERSHIPS;
1472 int sysctl_igmp_max_msf __read_mostly = IP_MAX_MSF;
1473
1474
1475 static int ip_mc_del1_src(struct ip_mc_list *pmc, int sfmode,
1476         __be32 *psfsrc)
1477 {
1478         struct ip_sf_list *psf, *psf_prev;
1479         int rv = 0;
1480
1481         psf_prev = NULL;
1482         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1483                 if (psf->sf_inaddr == *psfsrc)
1484                         break;
1485                 psf_prev = psf;
1486         }
1487         if (!psf || psf->sf_count[sfmode] == 0) {
1488                 /* source filter not found, or count wrong =>  bug */
1489                 return -ESRCH;
1490         }
1491         psf->sf_count[sfmode]--;
1492         if (psf->sf_count[sfmode] == 0) {
1493                 ip_rt_multicast_event(pmc->interface);
1494         }
1495         if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
1496 #ifdef CONFIG_IP_MULTICAST
1497                 struct in_device *in_dev = pmc->interface;
1498 #endif
1499
1500                 /* no more filters for this source */
1501                 if (psf_prev)
1502                         psf_prev->sf_next = psf->sf_next;
1503                 else
1504                         pmc->sources = psf->sf_next;
1505 #ifdef CONFIG_IP_MULTICAST
1506                 if (psf->sf_oldin &&
1507                     !IGMP_V1_SEEN(in_dev) && !IGMP_V2_SEEN(in_dev)) {
1508                         psf->sf_crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1509                                 IGMP_Unsolicited_Report_Count;
1510                         psf->sf_next = pmc->tomb;
1511                         pmc->tomb = psf;
1512                         rv = 1;
1513                 } else
1514 #endif
1515                         kfree(psf);
1516         }
1517         return rv;
1518 }
1519
1520 #ifndef CONFIG_IP_MULTICAST
1521 #define igmp_ifc_event(x)       do { } while (0)
1522 #endif
1523
1524 static int ip_mc_del_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1525                          int sfcount, __be32 *psfsrc, int delta)
1526 {
1527         struct ip_mc_list *pmc;
1528         int     changerec = 0;
1529         int     i, err;
1530
1531         if (!in_dev)
1532                 return -ENODEV;
1533         rcu_read_lock();
1534         for_each_pmc_rcu(in_dev, pmc) {
1535                 if (*pmca == pmc->multiaddr)
1536                         break;
1537         }
1538         if (!pmc) {
1539                 /* MCA not found?? bug */
1540                 rcu_read_unlock();
1541                 return -ESRCH;
1542         }
1543         spin_lock_bh(&pmc->lock);
1544         rcu_read_unlock();
1545 #ifdef CONFIG_IP_MULTICAST
1546         sf_markstate(pmc);
1547 #endif
1548         if (!delta) {
1549                 err = -EINVAL;
1550                 if (!pmc->sfcount[sfmode])
1551                         goto out_unlock;
1552                 pmc->sfcount[sfmode]--;
1553         }
1554         err = 0;
1555         for (i=0; i<sfcount; i++) {
1556                 int rv = ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1557
1558                 changerec |= rv > 0;
1559                 if (!err && rv < 0)
1560                         err = rv;
1561         }
1562         if (pmc->sfmode == MCAST_EXCLUDE &&
1563             pmc->sfcount[MCAST_EXCLUDE] == 0 &&
1564             pmc->sfcount[MCAST_INCLUDE]) {
1565 #ifdef CONFIG_IP_MULTICAST
1566                 struct ip_sf_list *psf;
1567 #endif
1568
1569                 /* filter mode change */
1570                 pmc->sfmode = MCAST_INCLUDE;
1571 #ifdef CONFIG_IP_MULTICAST
1572                 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1573                         IGMP_Unsolicited_Report_Count;
1574                 in_dev->mr_ifc_count = pmc->crcount;
1575                 for (psf=pmc->sources; psf; psf = psf->sf_next)
1576                         psf->sf_crcount = 0;
1577                 igmp_ifc_event(pmc->interface);
1578         } else if (sf_setstate(pmc) || changerec) {
1579                 igmp_ifc_event(pmc->interface);
1580 #endif
1581         }
1582 out_unlock:
1583         spin_unlock_bh(&pmc->lock);
1584         return err;
1585 }
1586
1587 /*
1588  * Add multicast single-source filter to the interface list
1589  */
1590 static int ip_mc_add1_src(struct ip_mc_list *pmc, int sfmode,
1591         __be32 *psfsrc, int delta)
1592 {
1593         struct ip_sf_list *psf, *psf_prev;
1594
1595         psf_prev = NULL;
1596         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1597                 if (psf->sf_inaddr == *psfsrc)
1598                         break;
1599                 psf_prev = psf;
1600         }
1601         if (!psf) {
1602                 psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
1603                 if (!psf)
1604                         return -ENOBUFS;
1605                 psf->sf_inaddr = *psfsrc;
1606                 if (psf_prev) {
1607                         psf_prev->sf_next = psf;
1608                 } else
1609                         pmc->sources = psf;
1610         }
1611         psf->sf_count[sfmode]++;
1612         if (psf->sf_count[sfmode] == 1) {
1613                 ip_rt_multicast_event(pmc->interface);
1614         }
1615         return 0;
1616 }
1617
1618 #ifdef CONFIG_IP_MULTICAST
1619 static void sf_markstate(struct ip_mc_list *pmc)
1620 {
1621         struct ip_sf_list *psf;
1622         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1623
1624         for (psf=pmc->sources; psf; psf=psf->sf_next)
1625                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1626                         psf->sf_oldin = mca_xcount ==
1627                                 psf->sf_count[MCAST_EXCLUDE] &&
1628                                 !psf->sf_count[MCAST_INCLUDE];
1629                 } else
1630                         psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
1631 }
1632
1633 static int sf_setstate(struct ip_mc_list *pmc)
1634 {
1635         struct ip_sf_list *psf, *dpsf;
1636         int mca_xcount = pmc->sfcount[MCAST_EXCLUDE];
1637         int qrv = pmc->interface->mr_qrv;
1638         int new_in, rv;
1639
1640         rv = 0;
1641         for (psf=pmc->sources; psf; psf=psf->sf_next) {
1642                 if (pmc->sfcount[MCAST_EXCLUDE]) {
1643                         new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
1644                                 !psf->sf_count[MCAST_INCLUDE];
1645                 } else
1646                         new_in = psf->sf_count[MCAST_INCLUDE] != 0;
1647                 if (new_in) {
1648                         if (!psf->sf_oldin) {
1649                                 struct ip_sf_list *prev = NULL;
1650
1651                                 for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next) {
1652                                         if (dpsf->sf_inaddr == psf->sf_inaddr)
1653                                                 break;
1654                                         prev = dpsf;
1655                                 }
1656                                 if (dpsf) {
1657                                         if (prev)
1658                                                 prev->sf_next = dpsf->sf_next;
1659                                         else
1660                                                 pmc->tomb = dpsf->sf_next;
1661                                         kfree(dpsf);
1662                                 }
1663                                 psf->sf_crcount = qrv;
1664                                 rv++;
1665                         }
1666                 } else if (psf->sf_oldin) {
1667
1668                         psf->sf_crcount = 0;
1669                         /*
1670                          * add or update "delete" records if an active filter
1671                          * is now inactive
1672                          */
1673                         for (dpsf=pmc->tomb; dpsf; dpsf=dpsf->sf_next)
1674                                 if (dpsf->sf_inaddr == psf->sf_inaddr)
1675                                         break;
1676                         if (!dpsf) {
1677                                 dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
1678                                 if (!dpsf)
1679                                         continue;
1680                                 *dpsf = *psf;
1681                                 /* pmc->lock held by callers */
1682                                 dpsf->sf_next = pmc->tomb;
1683                                 pmc->tomb = dpsf;
1684                         }
1685                         dpsf->sf_crcount = qrv;
1686                         rv++;
1687                 }
1688         }
1689         return rv;
1690 }
1691 #endif
1692
1693 /*
1694  * Add multicast source filter list to the interface list
1695  */
1696 static int ip_mc_add_src(struct in_device *in_dev, __be32 *pmca, int sfmode,
1697                          int sfcount, __be32 *psfsrc, int delta)
1698 {
1699         struct ip_mc_list *pmc;
1700         int     isexclude;
1701         int     i, err;
1702
1703         if (!in_dev)
1704                 return -ENODEV;
1705         rcu_read_lock();
1706         for_each_pmc_rcu(in_dev, pmc) {
1707                 if (*pmca == pmc->multiaddr)
1708                         break;
1709         }
1710         if (!pmc) {
1711                 /* MCA not found?? bug */
1712                 rcu_read_unlock();
1713                 return -ESRCH;
1714         }
1715         spin_lock_bh(&pmc->lock);
1716         rcu_read_unlock();
1717
1718 #ifdef CONFIG_IP_MULTICAST
1719         sf_markstate(pmc);
1720 #endif
1721         isexclude = pmc->sfmode == MCAST_EXCLUDE;
1722         if (!delta)
1723                 pmc->sfcount[sfmode]++;
1724         err = 0;
1725         for (i=0; i<sfcount; i++) {
1726                 err = ip_mc_add1_src(pmc, sfmode, &psfsrc[i], delta);
1727                 if (err)
1728                         break;
1729         }
1730         if (err) {
1731                 int j;
1732
1733                 pmc->sfcount[sfmode]--;
1734                 for (j=0; j<i; j++)
1735                         (void) ip_mc_del1_src(pmc, sfmode, &psfsrc[i]);
1736         } else if (isexclude != (pmc->sfcount[MCAST_EXCLUDE] != 0)) {
1737 #ifdef CONFIG_IP_MULTICAST
1738                 struct ip_sf_list *psf;
1739                 in_dev = pmc->interface;
1740 #endif
1741
1742                 /* filter mode change */
1743                 if (pmc->sfcount[MCAST_EXCLUDE])
1744                         pmc->sfmode = MCAST_EXCLUDE;
1745                 else if (pmc->sfcount[MCAST_INCLUDE])
1746                         pmc->sfmode = MCAST_INCLUDE;
1747 #ifdef CONFIG_IP_MULTICAST
1748                 /* else no filters; keep old mode for reports */
1749
1750                 pmc->crcount = in_dev->mr_qrv ? in_dev->mr_qrv :
1751                         IGMP_Unsolicited_Report_Count;
1752                 in_dev->mr_ifc_count = pmc->crcount;
1753                 for (psf=pmc->sources; psf; psf = psf->sf_next)
1754                         psf->sf_crcount = 0;
1755                 igmp_ifc_event(in_dev);
1756         } else if (sf_setstate(pmc)) {
1757                 igmp_ifc_event(in_dev);
1758 #endif
1759         }
1760         spin_unlock_bh(&pmc->lock);
1761         return err;
1762 }
1763
1764 static void ip_mc_clear_src(struct ip_mc_list *pmc)
1765 {
1766         struct ip_sf_list *psf, *nextpsf;
1767
1768         for (psf=pmc->tomb; psf; psf=nextpsf) {
1769                 nextpsf = psf->sf_next;
1770                 kfree(psf);
1771         }
1772         pmc->tomb = NULL;
1773         for (psf=pmc->sources; psf; psf=nextpsf) {
1774                 nextpsf = psf->sf_next;
1775                 kfree(psf);
1776         }
1777         pmc->sources = NULL;
1778         pmc->sfmode = MCAST_EXCLUDE;
1779         pmc->sfcount[MCAST_INCLUDE] = 0;
1780         pmc->sfcount[MCAST_EXCLUDE] = 1;
1781 }
1782
1783
1784 /*
1785  * Join a multicast group
1786  */
1787 int ip_mc_join_group(struct sock *sk , struct ip_mreqn *imr)
1788 {
1789         int err;
1790         __be32 addr = imr->imr_multiaddr.s_addr;
1791         struct ip_mc_socklist *iml = NULL, *i;
1792         struct in_device *in_dev;
1793         struct inet_sock *inet = inet_sk(sk);
1794         struct net *net = sock_net(sk);
1795         int ifindex;
1796         int count = 0;
1797
1798         if (!ipv4_is_multicast(addr))
1799                 return -EINVAL;
1800
1801         rtnl_lock();
1802
1803         in_dev = ip_mc_find_dev(net, imr);
1804
1805         if (!in_dev) {
1806                 iml = NULL;
1807                 err = -ENODEV;
1808                 goto done;
1809         }
1810
1811         err = -EADDRINUSE;
1812         ifindex = imr->imr_ifindex;
1813         for_each_pmc_rtnl(inet, i) {
1814                 if (i->multi.imr_multiaddr.s_addr == addr &&
1815                     i->multi.imr_ifindex == ifindex)
1816                         goto done;
1817                 count++;
1818         }
1819         err = -ENOBUFS;
1820         if (count >= sysctl_igmp_max_memberships)
1821                 goto done;
1822         iml = sock_kmalloc(sk, sizeof(*iml), GFP_KERNEL);
1823         if (iml == NULL)
1824                 goto done;
1825
1826         memcpy(&iml->multi, imr, sizeof(*imr));
1827         iml->next_rcu = inet->mc_list;
1828         iml->sflist = NULL;
1829         iml->sfmode = MCAST_EXCLUDE;
1830         rcu_assign_pointer(inet->mc_list, iml);
1831         ip_mc_inc_group(in_dev, addr);
1832         err = 0;
1833 done:
1834         rtnl_unlock();
1835         return err;
1836 }
1837 EXPORT_SYMBOL(ip_mc_join_group);
1838
1839 static void ip_sf_socklist_reclaim(struct rcu_head *rp)
1840 {
1841         kfree(container_of(rp, struct ip_sf_socklist, rcu));
1842         /* sk_omem_alloc should have been decreased by the caller*/
1843 }
1844
1845 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
1846                            struct in_device *in_dev)
1847 {
1848         struct ip_sf_socklist *psf = rtnl_dereference(iml->sflist);
1849         int err;
1850
1851         if (psf == NULL) {
1852                 /* any-source empty exclude case */
1853                 return ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1854                         iml->sfmode, 0, NULL, 0);
1855         }
1856         err = ip_mc_del_src(in_dev, &iml->multi.imr_multiaddr.s_addr,
1857                         iml->sfmode, psf->sl_count, psf->sl_addr, 0);
1858         rcu_assign_pointer(iml->sflist, NULL);
1859         /* decrease mem now to avoid the memleak warning */
1860         atomic_sub(IP_SFLSIZE(psf->sl_max), &sk->sk_omem_alloc);
1861         call_rcu(&psf->rcu, ip_sf_socklist_reclaim);
1862         return err;
1863 }
1864
1865
1866 static void ip_mc_socklist_reclaim(struct rcu_head *rp)
1867 {
1868         kfree(container_of(rp, struct ip_mc_socklist, rcu));
1869         /* sk_omem_alloc should have been decreased by the caller*/
1870 }
1871
1872
1873 /*
1874  *      Ask a socket to leave a group.
1875  */
1876
1877 int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr)
1878 {
1879         struct inet_sock *inet = inet_sk(sk);
1880         struct ip_mc_socklist *iml;
1881         struct ip_mc_socklist __rcu **imlp;
1882         struct in_device *in_dev;
1883         struct net *net = sock_net(sk);
1884         __be32 group = imr->imr_multiaddr.s_addr;
1885         u32 ifindex;
1886         int ret = -EADDRNOTAVAIL;
1887
1888         rtnl_lock();
1889         in_dev = ip_mc_find_dev(net, imr);
1890         ifindex = imr->imr_ifindex;
1891         for (imlp = &inet->mc_list;
1892              (iml = rtnl_dereference(*imlp)) != NULL;
1893              imlp = &iml->next_rcu) {
1894                 if (iml->multi.imr_multiaddr.s_addr != group)
1895                         continue;
1896                 if (ifindex) {
1897                         if (iml->multi.imr_ifindex != ifindex)
1898                                 continue;
1899                 } else if (imr->imr_address.s_addr && imr->imr_address.s_addr !=
1900                                 iml->multi.imr_address.s_addr)
1901                         continue;
1902
1903                 (void) ip_mc_leave_src(sk, iml, in_dev);
1904
1905                 *imlp = iml->next_rcu;
1906
1907                 if (in_dev)
1908                         ip_mc_dec_group(in_dev, group);
1909                 rtnl_unlock();
1910                 /* decrease mem now to avoid the memleak warning */
1911                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
1912                 call_rcu(&iml->rcu, ip_mc_socklist_reclaim);
1913                 return 0;
1914         }
1915         if (!in_dev)
1916                 ret = -ENODEV;
1917         rtnl_unlock();
1918         return ret;
1919 }
1920
1921 int ip_mc_source(int add, int omode, struct sock *sk, struct
1922         ip_mreq_source *mreqs, int ifindex)
1923 {
1924         int err;
1925         struct ip_mreqn imr;
1926         __be32 addr = mreqs->imr_multiaddr;
1927         struct ip_mc_socklist *pmc;
1928         struct in_device *in_dev = NULL;
1929         struct inet_sock *inet = inet_sk(sk);
1930         struct ip_sf_socklist *psl;
1931         struct net *net = sock_net(sk);
1932         int leavegroup = 0;
1933         int i, j, rv;
1934
1935         if (!ipv4_is_multicast(addr))
1936                 return -EINVAL;
1937
1938         rtnl_lock();
1939
1940         imr.imr_multiaddr.s_addr = mreqs->imr_multiaddr;
1941         imr.imr_address.s_addr = mreqs->imr_interface;
1942         imr.imr_ifindex = ifindex;
1943         in_dev = ip_mc_find_dev(net, &imr);
1944
1945         if (!in_dev) {
1946                 err = -ENODEV;
1947                 goto done;
1948         }
1949         err = -EADDRNOTAVAIL;
1950
1951         for_each_pmc_rtnl(inet, pmc) {
1952                 if ((pmc->multi.imr_multiaddr.s_addr ==
1953                      imr.imr_multiaddr.s_addr) &&
1954                     (pmc->multi.imr_ifindex == imr.imr_ifindex))
1955                         break;
1956         }
1957         if (!pmc) {             /* must have a prior join */
1958                 err = -EINVAL;
1959                 goto done;
1960         }
1961         /* if a source filter was set, must be the same mode as before */
1962         if (pmc->sflist) {
1963                 if (pmc->sfmode != omode) {
1964                         err = -EINVAL;
1965                         goto done;
1966                 }
1967         } else if (pmc->sfmode != omode) {
1968                 /* allow mode switches for empty-set filters */
1969                 ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 0, NULL, 0);
1970                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, pmc->sfmode, 0,
1971                         NULL, 0);
1972                 pmc->sfmode = omode;
1973         }
1974
1975         psl = rtnl_dereference(pmc->sflist);
1976         if (!add) {
1977                 if (!psl)
1978                         goto done;      /* err = -EADDRNOTAVAIL */
1979                 rv = !0;
1980                 for (i=0; i<psl->sl_count; i++) {
1981                         rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
1982                                 sizeof(__be32));
1983                         if (rv == 0)
1984                                 break;
1985                 }
1986                 if (rv)         /* source not found */
1987                         goto done;      /* err = -EADDRNOTAVAIL */
1988
1989                 /* special case - (INCLUDE, empty) == LEAVE_GROUP */
1990                 if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
1991                         leavegroup = 1;
1992                         goto done;
1993                 }
1994
1995                 /* update the interface filter */
1996                 ip_mc_del_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
1997                         &mreqs->imr_sourceaddr, 1);
1998
1999                 for (j=i+1; j<psl->sl_count; j++)
2000                         psl->sl_addr[j-1] = psl->sl_addr[j];
2001                 psl->sl_count--;
2002                 err = 0;
2003                 goto done;
2004         }
2005         /* else, add a new source to the filter */
2006
2007         if (psl && psl->sl_count >= sysctl_igmp_max_msf) {
2008                 err = -ENOBUFS;
2009                 goto done;
2010         }
2011         if (!psl || psl->sl_count == psl->sl_max) {
2012                 struct ip_sf_socklist *newpsl;
2013                 int count = IP_SFBLOCK;
2014
2015                 if (psl)
2016                         count += psl->sl_max;
2017                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(count), GFP_KERNEL);
2018                 if (!newpsl) {
2019                         err = -ENOBUFS;
2020                         goto done;
2021                 }
2022                 newpsl->sl_max = count;
2023                 newpsl->sl_count = count - IP_SFBLOCK;
2024                 if (psl) {
2025                         for (i=0; i<psl->sl_count; i++)
2026                                 newpsl->sl_addr[i] = psl->sl_addr[i];
2027                         /* decrease mem now to avoid the memleak warning */
2028                         atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2029                         call_rcu(&psl->rcu, ip_sf_socklist_reclaim);
2030                 }
2031                 rcu_assign_pointer(pmc->sflist, newpsl);
2032                 psl = newpsl;
2033         }
2034         rv = 1; /* > 0 for insert logic below if sl_count is 0 */
2035         for (i=0; i<psl->sl_count; i++) {
2036                 rv = memcmp(&psl->sl_addr[i], &mreqs->imr_sourceaddr,
2037                         sizeof(__be32));
2038                 if (rv == 0)
2039                         break;
2040         }
2041         if (rv == 0)            /* address already there is an error */
2042                 goto done;
2043         for (j=psl->sl_count-1; j>=i; j--)
2044                 psl->sl_addr[j+1] = psl->sl_addr[j];
2045         psl->sl_addr[i] = mreqs->imr_sourceaddr;
2046         psl->sl_count++;
2047         err = 0;
2048         /* update the interface list */
2049         ip_mc_add_src(in_dev, &mreqs->imr_multiaddr, omode, 1,
2050                 &mreqs->imr_sourceaddr, 1);
2051 done:
2052         rtnl_unlock();
2053         if (leavegroup)
2054                 return ip_mc_leave_group(sk, &imr);
2055         return err;
2056 }
2057
2058 int ip_mc_msfilter(struct sock *sk, struct ip_msfilter *msf, int ifindex)
2059 {
2060         int err = 0;
2061         struct ip_mreqn imr;
2062         __be32 addr = msf->imsf_multiaddr;
2063         struct ip_mc_socklist *pmc;
2064         struct in_device *in_dev;
2065         struct inet_sock *inet = inet_sk(sk);
2066         struct ip_sf_socklist *newpsl, *psl;
2067         struct net *net = sock_net(sk);
2068         int leavegroup = 0;
2069
2070         if (!ipv4_is_multicast(addr))
2071                 return -EINVAL;
2072         if (msf->imsf_fmode != MCAST_INCLUDE &&
2073             msf->imsf_fmode != MCAST_EXCLUDE)
2074                 return -EINVAL;
2075
2076         rtnl_lock();
2077
2078         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2079         imr.imr_address.s_addr = msf->imsf_interface;
2080         imr.imr_ifindex = ifindex;
2081         in_dev = ip_mc_find_dev(net, &imr);
2082
2083         if (!in_dev) {
2084                 err = -ENODEV;
2085                 goto done;
2086         }
2087
2088         /* special case - (INCLUDE, empty) == LEAVE_GROUP */
2089         if (msf->imsf_fmode == MCAST_INCLUDE && msf->imsf_numsrc == 0) {
2090                 leavegroup = 1;
2091                 goto done;
2092         }
2093
2094         for_each_pmc_rtnl(inet, pmc) {
2095                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2096                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2097                         break;
2098         }
2099         if (!pmc) {             /* must have a prior join */
2100                 err = -EINVAL;
2101                 goto done;
2102         }
2103         if (msf->imsf_numsrc) {
2104                 newpsl = sock_kmalloc(sk, IP_SFLSIZE(msf->imsf_numsrc),
2105                                                            GFP_KERNEL);
2106                 if (!newpsl) {
2107                         err = -ENOBUFS;
2108                         goto done;
2109                 }
2110                 newpsl->sl_max = newpsl->sl_count = msf->imsf_numsrc;
2111                 memcpy(newpsl->sl_addr, msf->imsf_slist,
2112                         msf->imsf_numsrc * sizeof(msf->imsf_slist[0]));
2113                 err = ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2114                         msf->imsf_fmode, newpsl->sl_count, newpsl->sl_addr, 0);
2115                 if (err) {
2116                         sock_kfree_s(sk, newpsl, IP_SFLSIZE(newpsl->sl_max));
2117                         goto done;
2118                 }
2119         } else {
2120                 newpsl = NULL;
2121                 (void) ip_mc_add_src(in_dev, &msf->imsf_multiaddr,
2122                                      msf->imsf_fmode, 0, NULL, 0);
2123         }
2124         psl = rtnl_dereference(pmc->sflist);
2125         if (psl) {
2126                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2127                         psl->sl_count, psl->sl_addr, 0);
2128                 /* decrease mem now to avoid the memleak warning */
2129                 atomic_sub(IP_SFLSIZE(psl->sl_max), &sk->sk_omem_alloc);
2130                 call_rcu(&psl->rcu, ip_sf_socklist_reclaim);
2131         } else
2132                 (void) ip_mc_del_src(in_dev, &msf->imsf_multiaddr, pmc->sfmode,
2133                         0, NULL, 0);
2134         rcu_assign_pointer(pmc->sflist, newpsl);
2135         pmc->sfmode = msf->imsf_fmode;
2136         err = 0;
2137 done:
2138         rtnl_unlock();
2139         if (leavegroup)
2140                 err = ip_mc_leave_group(sk, &imr);
2141         return err;
2142 }
2143
2144 int ip_mc_msfget(struct sock *sk, struct ip_msfilter *msf,
2145         struct ip_msfilter __user *optval, int __user *optlen)
2146 {
2147         int err, len, count, copycount;
2148         struct ip_mreqn imr;
2149         __be32 addr = msf->imsf_multiaddr;
2150         struct ip_mc_socklist *pmc;
2151         struct in_device *in_dev;
2152         struct inet_sock *inet = inet_sk(sk);
2153         struct ip_sf_socklist *psl;
2154         struct net *net = sock_net(sk);
2155
2156         if (!ipv4_is_multicast(addr))
2157                 return -EINVAL;
2158
2159         rtnl_lock();
2160
2161         imr.imr_multiaddr.s_addr = msf->imsf_multiaddr;
2162         imr.imr_address.s_addr = msf->imsf_interface;
2163         imr.imr_ifindex = 0;
2164         in_dev = ip_mc_find_dev(net, &imr);
2165
2166         if (!in_dev) {
2167                 err = -ENODEV;
2168                 goto done;
2169         }
2170         err = -EADDRNOTAVAIL;
2171
2172         for_each_pmc_rtnl(inet, pmc) {
2173                 if (pmc->multi.imr_multiaddr.s_addr == msf->imsf_multiaddr &&
2174                     pmc->multi.imr_ifindex == imr.imr_ifindex)
2175                         break;
2176         }
2177         if (!pmc)               /* must have a prior join */
2178                 goto done;
2179         msf->imsf_fmode = pmc->sfmode;
2180         psl = rtnl_dereference(pmc->sflist);
2181         rtnl_unlock();
2182         if (!psl) {
2183                 len = 0;
2184                 count = 0;
2185         } else {
2186                 count = psl->sl_count;
2187         }
2188         copycount = count < msf->imsf_numsrc ? count : msf->imsf_numsrc;
2189         len = copycount * sizeof(psl->sl_addr[0]);
2190         msf->imsf_numsrc = count;
2191         if (put_user(IP_MSFILTER_SIZE(copycount), optlen) ||
2192             copy_to_user(optval, msf, IP_MSFILTER_SIZE(0))) {
2193                 return -EFAULT;
2194         }
2195         if (len &&
2196             copy_to_user(&optval->imsf_slist[0], psl->sl_addr, len))
2197                 return -EFAULT;
2198         return 0;
2199 done:
2200         rtnl_unlock();
2201         return err;
2202 }
2203
2204 int ip_mc_gsfget(struct sock *sk, struct group_filter *gsf,
2205         struct group_filter __user *optval, int __user *optlen)
2206 {
2207         int err, i, count, copycount;
2208         struct sockaddr_in *psin;
2209         __be32 addr;
2210         struct ip_mc_socklist *pmc;
2211         struct inet_sock *inet = inet_sk(sk);
2212         struct ip_sf_socklist *psl;
2213
2214         psin = (struct sockaddr_in *)&gsf->gf_group;
2215         if (psin->sin_family != AF_INET)
2216                 return -EINVAL;
2217         addr = psin->sin_addr.s_addr;
2218         if (!ipv4_is_multicast(addr))
2219                 return -EINVAL;
2220
2221         rtnl_lock();
2222
2223         err = -EADDRNOTAVAIL;
2224
2225         for_each_pmc_rtnl(inet, pmc) {
2226                 if (pmc->multi.imr_multiaddr.s_addr == addr &&
2227                     pmc->multi.imr_ifindex == gsf->gf_interface)
2228                         break;
2229         }
2230         if (!pmc)               /* must have a prior join */
2231                 goto done;
2232         gsf->gf_fmode = pmc->sfmode;
2233         psl = rtnl_dereference(pmc->sflist);
2234         rtnl_unlock();
2235         count = psl ? psl->sl_count : 0;
2236         copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
2237         gsf->gf_numsrc = count;
2238         if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
2239             copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
2240                 return -EFAULT;
2241         }
2242         for (i=0; i<copycount; i++) {
2243                 struct sockaddr_storage ss;
2244
2245                 psin = (struct sockaddr_in *)&ss;
2246                 memset(&ss, 0, sizeof(ss));
2247                 psin->sin_family = AF_INET;
2248                 psin->sin_addr.s_addr = psl->sl_addr[i];
2249                 if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
2250                         return -EFAULT;
2251         }
2252         return 0;
2253 done:
2254         rtnl_unlock();
2255         return err;
2256 }
2257
2258 /*
2259  * check if a multicast source filter allows delivery for a given <src,dst,intf>
2260  */
2261 int ip_mc_sf_allow(struct sock *sk, __be32 loc_addr, __be32 rmt_addr, int dif)
2262 {
2263         struct inet_sock *inet = inet_sk(sk);
2264         struct ip_mc_socklist *pmc;
2265         struct ip_sf_socklist *psl;
2266         int i;
2267         int ret;
2268
2269         ret = 1;
2270         if (!ipv4_is_multicast(loc_addr))
2271                 goto out;
2272
2273         rcu_read_lock();
2274         for_each_pmc_rcu(inet, pmc) {
2275                 if (pmc->multi.imr_multiaddr.s_addr == loc_addr &&
2276                     pmc->multi.imr_ifindex == dif)
2277                         break;
2278         }
2279         ret = inet->mc_all;
2280         if (!pmc)
2281                 goto unlock;
2282         psl = rcu_dereference(pmc->sflist);
2283         ret = (pmc->sfmode == MCAST_EXCLUDE);
2284         if (!psl)
2285                 goto unlock;
2286
2287         for (i=0; i<psl->sl_count; i++) {
2288                 if (psl->sl_addr[i] == rmt_addr)
2289                         break;
2290         }
2291         ret = 0;
2292         if (pmc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
2293                 goto unlock;
2294         if (pmc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
2295                 goto unlock;
2296         ret = 1;
2297 unlock:
2298         rcu_read_unlock();
2299 out:
2300         return ret;
2301 }
2302
2303 /*
2304  *      A socket is closing.
2305  */
2306
2307 void ip_mc_drop_socket(struct sock *sk)
2308 {
2309         struct inet_sock *inet = inet_sk(sk);
2310         struct ip_mc_socklist *iml;
2311         struct net *net = sock_net(sk);
2312
2313         if (inet->mc_list == NULL)
2314                 return;
2315
2316         rtnl_lock();
2317         while ((iml = rtnl_dereference(inet->mc_list)) != NULL) {
2318                 struct in_device *in_dev;
2319
2320                 inet->mc_list = iml->next_rcu;
2321                 in_dev = inetdev_by_index(net, iml->multi.imr_ifindex);
2322                 (void) ip_mc_leave_src(sk, iml, in_dev);
2323                 if (in_dev != NULL)
2324                         ip_mc_dec_group(in_dev, iml->multi.imr_multiaddr.s_addr);
2325                 /* decrease mem now to avoid the memleak warning */
2326                 atomic_sub(sizeof(*iml), &sk->sk_omem_alloc);
2327                 call_rcu(&iml->rcu, ip_mc_socklist_reclaim);
2328         }
2329         rtnl_unlock();
2330 }
2331
2332 int ip_check_mc(struct in_device *in_dev, __be32 mc_addr, __be32 src_addr, u16 proto)
2333 {
2334         struct ip_mc_list *im;
2335         struct ip_sf_list *psf;
2336         int rv = 0;
2337
2338         rcu_read_lock();
2339         for_each_pmc_rcu(in_dev, im) {
2340                 if (im->multiaddr == mc_addr)
2341                         break;
2342         }
2343         if (im && proto == IPPROTO_IGMP) {
2344                 rv = 1;
2345         } else if (im) {
2346                 if (src_addr) {
2347                         for (psf=im->sources; psf; psf=psf->sf_next) {
2348                                 if (psf->sf_inaddr == src_addr)
2349                                         break;
2350                         }
2351                         if (psf)
2352                                 rv = psf->sf_count[MCAST_INCLUDE] ||
2353                                         psf->sf_count[MCAST_EXCLUDE] !=
2354                                         im->sfcount[MCAST_EXCLUDE];
2355                         else
2356                                 rv = im->sfcount[MCAST_EXCLUDE] != 0;
2357                 } else
2358                         rv = 1; /* unspecified source; tentatively allow */
2359         }
2360         rcu_read_unlock();
2361         return rv;
2362 }
2363
2364 #if defined(CONFIG_PROC_FS)
2365 struct igmp_mc_iter_state {
2366         struct seq_net_private p;
2367         struct net_device *dev;
2368         struct in_device *in_dev;
2369 };
2370
2371 #define igmp_mc_seq_private(seq)        ((struct igmp_mc_iter_state *)(seq)->private)
2372
2373 static inline struct ip_mc_list *igmp_mc_get_first(struct seq_file *seq)
2374 {
2375         struct net *net = seq_file_net(seq);
2376         struct ip_mc_list *im = NULL;
2377         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2378
2379         state->in_dev = NULL;
2380         for_each_netdev_rcu(net, state->dev) {
2381                 struct in_device *in_dev;
2382
2383                 in_dev = __in_dev_get_rcu(state->dev);
2384                 if (!in_dev)
2385                         continue;
2386                 im = rcu_dereference(in_dev->mc_list);
2387                 if (im) {
2388                         state->in_dev = in_dev;
2389                         break;
2390                 }
2391         }
2392         return im;
2393 }
2394
2395 static struct ip_mc_list *igmp_mc_get_next(struct seq_file *seq, struct ip_mc_list *im)
2396 {
2397         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2398
2399         im = rcu_dereference(im->next_rcu);
2400         while (!im) {
2401                 state->dev = next_net_device_rcu(state->dev);
2402                 if (!state->dev) {
2403                         state->in_dev = NULL;
2404                         break;
2405                 }
2406                 state->in_dev = __in_dev_get_rcu(state->dev);
2407                 if (!state->in_dev)
2408                         continue;
2409                 im = rcu_dereference(state->in_dev->mc_list);
2410         }
2411         return im;
2412 }
2413
2414 static struct ip_mc_list *igmp_mc_get_idx(struct seq_file *seq, loff_t pos)
2415 {
2416         struct ip_mc_list *im = igmp_mc_get_first(seq);
2417         if (im)
2418                 while (pos && (im = igmp_mc_get_next(seq, im)) != NULL)
2419                         --pos;
2420         return pos ? NULL : im;
2421 }
2422
2423 static void *igmp_mc_seq_start(struct seq_file *seq, loff_t *pos)
2424         __acquires(rcu)
2425 {
2426         rcu_read_lock();
2427         return *pos ? igmp_mc_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2428 }
2429
2430 static void *igmp_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2431 {
2432         struct ip_mc_list *im;
2433         if (v == SEQ_START_TOKEN)
2434                 im = igmp_mc_get_first(seq);
2435         else
2436                 im = igmp_mc_get_next(seq, v);
2437         ++*pos;
2438         return im;
2439 }
2440
2441 static void igmp_mc_seq_stop(struct seq_file *seq, void *v)
2442         __releases(rcu)
2443 {
2444         struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2445
2446         state->in_dev = NULL;
2447         state->dev = NULL;
2448         rcu_read_unlock();
2449 }
2450
2451 static int igmp_mc_seq_show(struct seq_file *seq, void *v)
2452 {
2453         if (v == SEQ_START_TOKEN)
2454                 seq_puts(seq,
2455                          "Idx\tDevice    : Count Querier\tGroup    Users Timer\tReporter\n");
2456         else {
2457                 struct ip_mc_list *im = (struct ip_mc_list *)v;
2458                 struct igmp_mc_iter_state *state = igmp_mc_seq_private(seq);
2459                 char   *querier;
2460 #ifdef CONFIG_IP_MULTICAST
2461                 querier = IGMP_V1_SEEN(state->in_dev) ? "V1" :
2462                           IGMP_V2_SEEN(state->in_dev) ? "V2" :
2463                           "V3";
2464 #else
2465                 querier = "NONE";
2466 #endif
2467
2468                 if (rcu_dereference(state->in_dev->mc_list) == im) {
2469                         seq_printf(seq, "%d\t%-10s: %5d %7s\n",
2470                                    state->dev->ifindex, state->dev->name, state->in_dev->mc_count, querier);
2471                 }
2472
2473                 seq_printf(seq,
2474                            "\t\t\t\t%08X %5d %d:%08lX\t\t%d\n",
2475                            im->multiaddr, im->users,
2476                            im->tm_running, im->tm_running ?
2477                            jiffies_to_clock_t(im->timer.expires-jiffies) : 0,
2478                            im->reporter);
2479         }
2480         return 0;
2481 }
2482
2483 static const struct seq_operations igmp_mc_seq_ops = {
2484         .start  =       igmp_mc_seq_start,
2485         .next   =       igmp_mc_seq_next,
2486         .stop   =       igmp_mc_seq_stop,
2487         .show   =       igmp_mc_seq_show,
2488 };
2489
2490 static int igmp_mc_seq_open(struct inode *inode, struct file *file)
2491 {
2492         return seq_open_net(inode, file, &igmp_mc_seq_ops,
2493                         sizeof(struct igmp_mc_iter_state));
2494 }
2495
2496 static const struct file_operations igmp_mc_seq_fops = {
2497         .owner          =       THIS_MODULE,
2498         .open           =       igmp_mc_seq_open,
2499         .read           =       seq_read,
2500         .llseek         =       seq_lseek,
2501         .release        =       seq_release_net,
2502 };
2503
2504 struct igmp_mcf_iter_state {
2505         struct seq_net_private p;
2506         struct net_device *dev;
2507         struct in_device *idev;
2508         struct ip_mc_list *im;
2509 };
2510
2511 #define igmp_mcf_seq_private(seq)       ((struct igmp_mcf_iter_state *)(seq)->private)
2512
2513 static inline struct ip_sf_list *igmp_mcf_get_first(struct seq_file *seq)
2514 {
2515         struct net *net = seq_file_net(seq);
2516         struct ip_sf_list *psf = NULL;
2517         struct ip_mc_list *im = NULL;
2518         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2519
2520         state->idev = NULL;
2521         state->im = NULL;
2522         for_each_netdev_rcu(net, state->dev) {
2523                 struct in_device *idev;
2524                 idev = __in_dev_get_rcu(state->dev);
2525                 if (unlikely(idev == NULL))
2526                         continue;
2527                 im = rcu_dereference(idev->mc_list);
2528                 if (likely(im != NULL)) {
2529                         spin_lock_bh(&im->lock);
2530                         psf = im->sources;
2531                         if (likely(psf != NULL)) {
2532                                 state->im = im;
2533                                 state->idev = idev;
2534                                 break;
2535                         }
2536                         spin_unlock_bh(&im->lock);
2537                 }
2538         }
2539         return psf;
2540 }
2541
2542 static struct ip_sf_list *igmp_mcf_get_next(struct seq_file *seq, struct ip_sf_list *psf)
2543 {
2544         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2545
2546         psf = psf->sf_next;
2547         while (!psf) {
2548                 spin_unlock_bh(&state->im->lock);
2549                 state->im = state->im->next;
2550                 while (!state->im) {
2551                         state->dev = next_net_device_rcu(state->dev);
2552                         if (!state->dev) {
2553                                 state->idev = NULL;
2554                                 goto out;
2555                         }
2556                         state->idev = __in_dev_get_rcu(state->dev);
2557                         if (!state->idev)
2558                                 continue;
2559                         state->im = rcu_dereference(state->idev->mc_list);
2560                 }
2561                 if (!state->im)
2562                         break;
2563                 spin_lock_bh(&state->im->lock);
2564                 psf = state->im->sources;
2565         }
2566 out:
2567         return psf;
2568 }
2569
2570 static struct ip_sf_list *igmp_mcf_get_idx(struct seq_file *seq, loff_t pos)
2571 {
2572         struct ip_sf_list *psf = igmp_mcf_get_first(seq);
2573         if (psf)
2574                 while (pos && (psf = igmp_mcf_get_next(seq, psf)) != NULL)
2575                         --pos;
2576         return pos ? NULL : psf;
2577 }
2578
2579 static void *igmp_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2580         __acquires(rcu)
2581 {
2582         rcu_read_lock();
2583         return *pos ? igmp_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2584 }
2585
2586 static void *igmp_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2587 {
2588         struct ip_sf_list *psf;
2589         if (v == SEQ_START_TOKEN)
2590                 psf = igmp_mcf_get_first(seq);
2591         else
2592                 psf = igmp_mcf_get_next(seq, v);
2593         ++*pos;
2594         return psf;
2595 }
2596
2597 static void igmp_mcf_seq_stop(struct seq_file *seq, void *v)
2598         __releases(rcu)
2599 {
2600         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2601         if (likely(state->im != NULL)) {
2602                 spin_unlock_bh(&state->im->lock);
2603                 state->im = NULL;
2604         }
2605         state->idev = NULL;
2606         state->dev = NULL;
2607         rcu_read_unlock();
2608 }
2609
2610 static int igmp_mcf_seq_show(struct seq_file *seq, void *v)
2611 {
2612         struct ip_sf_list *psf = (struct ip_sf_list *)v;
2613         struct igmp_mcf_iter_state *state = igmp_mcf_seq_private(seq);
2614
2615         if (v == SEQ_START_TOKEN) {
2616                 seq_printf(seq,
2617                            "%3s %6s "
2618                            "%10s %10s %6s %6s\n", "Idx",
2619                            "Device", "MCA",
2620                            "SRC", "INC", "EXC");
2621         } else {
2622                 seq_printf(seq,
2623                            "%3d %6.6s 0x%08x "
2624                            "0x%08x %6lu %6lu\n",
2625                            state->dev->ifindex, state->dev->name,
2626                            ntohl(state->im->multiaddr),
2627                            ntohl(psf->sf_inaddr),
2628                            psf->sf_count[MCAST_INCLUDE],
2629                            psf->sf_count[MCAST_EXCLUDE]);
2630         }
2631         return 0;
2632 }
2633
2634 static const struct seq_operations igmp_mcf_seq_ops = {
2635         .start  =       igmp_mcf_seq_start,
2636         .next   =       igmp_mcf_seq_next,
2637         .stop   =       igmp_mcf_seq_stop,
2638         .show   =       igmp_mcf_seq_show,
2639 };
2640
2641 static int igmp_mcf_seq_open(struct inode *inode, struct file *file)
2642 {
2643         return seq_open_net(inode, file, &igmp_mcf_seq_ops,
2644                         sizeof(struct igmp_mcf_iter_state));
2645 }
2646
2647 static const struct file_operations igmp_mcf_seq_fops = {
2648         .owner          =       THIS_MODULE,
2649         .open           =       igmp_mcf_seq_open,
2650         .read           =       seq_read,
2651         .llseek         =       seq_lseek,
2652         .release        =       seq_release_net,
2653 };
2654
2655 static int __net_init igmp_net_init(struct net *net)
2656 {
2657         struct proc_dir_entry *pde;
2658
2659         pde = proc_net_fops_create(net, "igmp", S_IRUGO, &igmp_mc_seq_fops);
2660         if (!pde)
2661                 goto out_igmp;
2662         pde = proc_net_fops_create(net, "mcfilter", S_IRUGO, &igmp_mcf_seq_fops);
2663         if (!pde)
2664                 goto out_mcfilter;
2665         return 0;
2666
2667 out_mcfilter:
2668         proc_net_remove(net, "igmp");
2669 out_igmp:
2670         return -ENOMEM;
2671 }
2672
2673 static void __net_exit igmp_net_exit(struct net *net)
2674 {
2675         proc_net_remove(net, "mcfilter");
2676         proc_net_remove(net, "igmp");
2677 }
2678
2679 static struct pernet_operations igmp_net_ops = {
2680         .init = igmp_net_init,
2681         .exit = igmp_net_exit,
2682 };
2683
2684 int __init igmp_mc_proc_init(void)
2685 {
2686         return register_pernet_subsys(&igmp_net_ops);
2687 }
2688 #endif