1/* Linux multicast routing support
2 * Common logic shared by IPv4 [ipmr] and IPv6 [ip6mr] implementation
3 */
4
5#include <linux/rhashtable.h>
6#include <linux/mroute_base.h>
7
8/* Sets everything common except 'dev', since that is done under locking */
9void vif_device_init(struct vif_device *v,
10 struct net_device *dev,
11 unsigned long rate_limit,
12 unsigned char threshold,
13 unsigned short flags,
14 unsigned short get_iflink_mask)
15{
16 RCU_INIT_POINTER(v->dev, NULL);
17 v->bytes_in = 0;
18 v->bytes_out = 0;
19 v->pkt_in = 0;
20 v->pkt_out = 0;
21 v->rate_limit = rate_limit;
22 v->flags = flags;
23 v->threshold = threshold;
24 if (v->flags & get_iflink_mask)
25 v->link = dev_get_iflink(dev);
26 else
27 v->link = dev->ifindex;
28}
29EXPORT_SYMBOL(vif_device_init);
30
31struct mr_table *
32mr_table_alloc(struct net *net, u32 id,
33 struct mr_table_ops *ops,
34 void (*expire_func)(struct timer_list *t),
35 void (*table_set)(struct mr_table *mrt,
36 struct net *net))
37{
38 struct mr_table *mrt;
39 int err;
40
41 mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
42 if (!mrt)
43 return ERR_PTR(error: -ENOMEM);
44 mrt->id = id;
45 write_pnet(pnet: &mrt->net, net);
46
47 mrt->ops = *ops;
48 err = rhltable_init(&mrt->mfc_hash, mrt->ops.rht_params);
49 if (err) {
50 kfree(objp: mrt);
51 return ERR_PTR(error: err);
52 }
53 INIT_LIST_HEAD(list: &mrt->mfc_cache_list);
54 INIT_LIST_HEAD(list: &mrt->mfc_unres_queue);
55
56 timer_setup(&mrt->ipmr_expire_timer, expire_func, 0);
57
58 mrt->mroute_reg_vif_num = -1;
59 table_set(mrt, net);
60 return mrt;
61}
62EXPORT_SYMBOL(mr_table_alloc);
63
64void *mr_mfc_find_parent(struct mr_table *mrt, void *hasharg, int parent)
65{
66 struct rhlist_head *tmp, *list;
67 struct mr_mfc *c;
68
69 list = rhltable_lookup(hlt: &mrt->mfc_hash, key: hasharg, params: *mrt->ops.rht_params);
70 rhl_for_each_entry_rcu(c, tmp, list, mnode)
71 if (parent == -1 || parent == c->mfc_parent)
72 return c;
73
74 return NULL;
75}
76EXPORT_SYMBOL(mr_mfc_find_parent);
77
78void *mr_mfc_find_any_parent(struct mr_table *mrt, int vifi)
79{
80 struct rhlist_head *tmp, *list;
81 struct mr_mfc *c;
82
83 list = rhltable_lookup(hlt: &mrt->mfc_hash, key: mrt->ops.cmparg_any,
84 params: *mrt->ops.rht_params);
85 rhl_for_each_entry_rcu(c, tmp, list, mnode)
86 if (c->mfc_un.res.ttls[vifi] < 255)
87 return c;
88
89 return NULL;
90}
91EXPORT_SYMBOL(mr_mfc_find_any_parent);
92
93void *mr_mfc_find_any(struct mr_table *mrt, int vifi, void *hasharg)
94{
95 struct rhlist_head *tmp, *list;
96 struct mr_mfc *c, *proxy;
97
98 list = rhltable_lookup(hlt: &mrt->mfc_hash, key: hasharg, params: *mrt->ops.rht_params);
99 rhl_for_each_entry_rcu(c, tmp, list, mnode) {
100 if (c->mfc_un.res.ttls[vifi] < 255)
101 return c;
102
103 /* It's ok if the vifi is part of the static tree */
104 proxy = mr_mfc_find_any_parent(mrt, c->mfc_parent);
105 if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
106 return c;
107 }
108
109 return mr_mfc_find_any_parent(mrt, vifi);
110}
111EXPORT_SYMBOL(mr_mfc_find_any);
112
113#ifdef CONFIG_PROC_FS
114void *mr_vif_seq_idx(struct net *net, struct mr_vif_iter *iter, loff_t pos)
115{
116 struct mr_table *mrt = iter->mrt;
117
118 for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
119 if (!VIF_EXISTS(mrt, iter->ct))
120 continue;
121 if (pos-- == 0)
122 return &mrt->vif_table[iter->ct];
123 }
124 return NULL;
125}
126EXPORT_SYMBOL(mr_vif_seq_idx);
127
128void *mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
129{
130 struct mr_vif_iter *iter = seq->private;
131 struct net *net = seq_file_net(seq);
132 struct mr_table *mrt = iter->mrt;
133
134 ++*pos;
135 if (v == SEQ_START_TOKEN)
136 return mr_vif_seq_idx(net, iter, 0);
137
138 while (++iter->ct < mrt->maxvif) {
139 if (!VIF_EXISTS(mrt, iter->ct))
140 continue;
141 return &mrt->vif_table[iter->ct];
142 }
143 return NULL;
144}
145EXPORT_SYMBOL(mr_vif_seq_next);
146
147void *mr_mfc_seq_idx(struct net *net,
148 struct mr_mfc_iter *it, loff_t pos)
149{
150 struct mr_table *mrt = it->mrt;
151 struct mr_mfc *mfc;
152
153 rcu_read_lock();
154 it->cache = &mrt->mfc_cache_list;
155 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list)
156 if (pos-- == 0)
157 return mfc;
158 rcu_read_unlock();
159
160 spin_lock_bh(lock: it->lock);
161 it->cache = &mrt->mfc_unres_queue;
162 list_for_each_entry(mfc, it->cache, list)
163 if (pos-- == 0)
164 return mfc;
165 spin_unlock_bh(lock: it->lock);
166
167 it->cache = NULL;
168 return NULL;
169}
170EXPORT_SYMBOL(mr_mfc_seq_idx);
171
172void *mr_mfc_seq_next(struct seq_file *seq, void *v,
173 loff_t *pos)
174{
175 struct mr_mfc_iter *it = seq->private;
176 struct net *net = seq_file_net(seq);
177 struct mr_table *mrt = it->mrt;
178 struct mr_mfc *c = v;
179
180 ++*pos;
181
182 if (v == SEQ_START_TOKEN)
183 return mr_mfc_seq_idx(net, seq->private, 0);
184
185 if (c->list.next != it->cache)
186 return list_entry(c->list.next, struct mr_mfc, list);
187
188 if (it->cache == &mrt->mfc_unres_queue)
189 goto end_of_list;
190
191 /* exhausted cache_array, show unresolved */
192 rcu_read_unlock();
193 it->cache = &mrt->mfc_unres_queue;
194
195 spin_lock_bh(lock: it->lock);
196 if (!list_empty(head: it->cache))
197 return list_first_entry(it->cache, struct mr_mfc, list);
198
199end_of_list:
200 spin_unlock_bh(lock: it->lock);
201 it->cache = NULL;
202
203 return NULL;
204}
205EXPORT_SYMBOL(mr_mfc_seq_next);
206#endif
207
208int mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
209 struct mr_mfc *c, struct rtmsg *rtm)
210{
211 struct net_device *vif_dev;
212 struct rta_mfc_stats mfcs;
213 struct nlattr *mp_attr;
214 struct rtnexthop *nhp;
215 unsigned long lastuse;
216 int ct;
217
218 /* If cache is unresolved, don't try to parse IIF and OIF */
219 if (c->mfc_parent >= MAXVIFS) {
220 rtm->rtm_flags |= RTNH_F_UNRESOLVED;
221 return -ENOENT;
222 }
223
224 rcu_read_lock();
225 vif_dev = rcu_dereference(mrt->vif_table[c->mfc_parent].dev);
226 if (vif_dev && nla_put_u32(skb, attrtype: RTA_IIF, value: vif_dev->ifindex) < 0) {
227 rcu_read_unlock();
228 return -EMSGSIZE;
229 }
230 rcu_read_unlock();
231
232 if (c->mfc_flags & MFC_OFFLOAD)
233 rtm->rtm_flags |= RTNH_F_OFFLOAD;
234
235 mp_attr = nla_nest_start_noflag(skb, attrtype: RTA_MULTIPATH);
236 if (!mp_attr)
237 return -EMSGSIZE;
238
239 rcu_read_lock();
240 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
241 struct vif_device *vif = &mrt->vif_table[ct];
242
243 vif_dev = rcu_dereference(vif->dev);
244 if (vif_dev && c->mfc_un.res.ttls[ct] < 255) {
245
246 nhp = nla_reserve_nohdr(skb, attrlen: sizeof(*nhp));
247 if (!nhp) {
248 rcu_read_unlock();
249 nla_nest_cancel(skb, start: mp_attr);
250 return -EMSGSIZE;
251 }
252
253 nhp->rtnh_flags = 0;
254 nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
255 nhp->rtnh_ifindex = vif_dev->ifindex;
256 nhp->rtnh_len = sizeof(*nhp);
257 }
258 }
259 rcu_read_unlock();
260
261 nla_nest_end(skb, start: mp_attr);
262
263 lastuse = READ_ONCE(c->mfc_un.res.lastuse);
264 lastuse = time_after_eq(jiffies, lastuse) ? jiffies - lastuse : 0;
265
266 mfcs.mfcs_packets = atomic_long_read(v: &c->mfc_un.res.pkt);
267 mfcs.mfcs_bytes = atomic_long_read(v: &c->mfc_un.res.bytes);
268 mfcs.mfcs_wrong_if = atomic_long_read(v: &c->mfc_un.res.wrong_if);
269 if (nla_put_64bit(skb, attrtype: RTA_MFC_STATS, attrlen: sizeof(mfcs), data: &mfcs, padattr: RTA_PAD) ||
270 nla_put_u64_64bit(skb, attrtype: RTA_EXPIRES, value: jiffies_to_clock_t(x: lastuse),
271 padattr: RTA_PAD))
272 return -EMSGSIZE;
273
274 rtm->rtm_type = RTN_MULTICAST;
275 return 1;
276}
277EXPORT_SYMBOL(mr_fill_mroute);
278
279static bool mr_mfc_uses_dev(const struct mr_table *mrt,
280 const struct mr_mfc *c,
281 const struct net_device *dev)
282{
283 int ct;
284
285 for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
286 const struct net_device *vif_dev;
287 const struct vif_device *vif;
288
289 vif = &mrt->vif_table[ct];
290 vif_dev = rcu_access_pointer(vif->dev);
291 if (vif_dev && c->mfc_un.res.ttls[ct] < 255 &&
292 vif_dev == dev)
293 return true;
294 }
295 return false;
296}
297
298int mr_table_dump(struct mr_table *mrt, struct sk_buff *skb,
299 struct netlink_callback *cb,
300 int (*fill)(struct mr_table *mrt, struct sk_buff *skb,
301 u32 portid, u32 seq, struct mr_mfc *c,
302 int cmd, int flags),
303 spinlock_t *lock, struct fib_dump_filter *filter)
304{
305 unsigned int e = 0, s_e = cb->args[1];
306 unsigned int flags = NLM_F_MULTI;
307 struct mr_mfc *mfc;
308 int err;
309
310 if (filter->filter_set)
311 flags |= NLM_F_DUMP_FILTERED;
312
313 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list,
314 lockdep_rtnl_is_held()) {
315 if (e < s_e)
316 goto next_entry;
317 if (filter->dev &&
318 !mr_mfc_uses_dev(mrt, c: mfc, dev: filter->dev))
319 goto next_entry;
320
321 err = fill(mrt, skb, NETLINK_CB(cb->skb).portid,
322 cb->nlh->nlmsg_seq, mfc, RTM_NEWROUTE, flags);
323 if (err < 0)
324 goto out;
325next_entry:
326 e++;
327 }
328
329 spin_lock_bh(lock);
330 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
331 if (e < s_e)
332 goto next_entry2;
333
334 err = fill(mrt, skb, NETLINK_CB(cb->skb).portid,
335 cb->nlh->nlmsg_seq, mfc, RTM_NEWROUTE, flags);
336 if (err < 0) {
337 spin_unlock_bh(lock);
338 goto out;
339 }
340next_entry2:
341 e++;
342 }
343 spin_unlock_bh(lock);
344 err = 0;
345out:
346 cb->args[1] = e;
347 return err;
348}
349EXPORT_SYMBOL(mr_table_dump);
350
351int mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb,
352 struct mr_table *(*iter)(struct net *net,
353 struct mr_table *mrt),
354 int (*fill)(struct mr_table *mrt,
355 struct sk_buff *skb,
356 u32 portid, u32 seq, struct mr_mfc *c,
357 int cmd, int flags),
358 spinlock_t *lock, struct fib_dump_filter *filter)
359{
360 unsigned int t = 0, s_t = cb->args[0];
361 struct net *net = sock_net(sk: skb->sk);
362 struct mr_table *mrt;
363 int err;
364
365 /* multicast does not track protocol or have route type other
366 * than RTN_MULTICAST
367 */
368 if (filter->filter_set) {
369 if (filter->protocol || filter->flags ||
370 (filter->rt_type && filter->rt_type != RTN_MULTICAST))
371 return skb->len;
372 }
373
374 rcu_read_lock();
375 for (mrt = iter(net, NULL); mrt; mrt = iter(net, mrt)) {
376 if (t < s_t)
377 goto next_table;
378
379 err = mr_table_dump(mrt, skb, cb, fill, lock, filter);
380 if (err < 0)
381 break;
382 cb->args[1] = 0;
383next_table:
384 t++;
385 }
386 rcu_read_unlock();
387
388 cb->args[0] = t;
389
390 return skb->len;
391}
392EXPORT_SYMBOL(mr_rtm_dumproute);
393
394int mr_dump(struct net *net, struct notifier_block *nb, unsigned short family,
395 int (*rules_dump)(struct net *net,
396 struct notifier_block *nb,
397 struct netlink_ext_ack *extack),
398 struct mr_table *(*mr_iter)(struct net *net,
399 struct mr_table *mrt),
400 struct netlink_ext_ack *extack)
401{
402 struct mr_table *mrt;
403 int err;
404
405 err = rules_dump(net, nb, extack);
406 if (err)
407 return err;
408
409 for (mrt = mr_iter(net, NULL); mrt; mrt = mr_iter(net, mrt)) {
410 struct vif_device *v = &mrt->vif_table[0];
411 struct net_device *vif_dev;
412 struct mr_mfc *mfc;
413 int vifi;
414
415 /* Notifiy on table VIF entries */
416 rcu_read_lock();
417 for (vifi = 0; vifi < mrt->maxvif; vifi++, v++) {
418 vif_dev = rcu_dereference(v->dev);
419 if (!vif_dev)
420 continue;
421
422 err = mr_call_vif_notifier(nb, family,
423 event_type: FIB_EVENT_VIF_ADD, vif: v,
424 vif_dev, vif_index: vifi,
425 tb_id: mrt->id, extack);
426 if (err)
427 break;
428 }
429 rcu_read_unlock();
430
431 if (err)
432 return err;
433
434 /* Notify on table MFC entries */
435 list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list) {
436 err = mr_call_mfc_notifier(nb, family,
437 event_type: FIB_EVENT_ENTRY_ADD,
438 mfc, tb_id: mrt->id, extack);
439 if (err)
440 return err;
441 }
442 }
443
444 return 0;
445}
446EXPORT_SYMBOL(mr_dump);
447