xref: /openbmc/linux/net/ipv4/ipmr_base.c (revision addee42a)
1 /* Linux multicast routing support
2  * Common logic shared by IPv4 [ipmr] and IPv6 [ip6mr] implementation
3  */
4 
5 #include <linux/mroute_base.h>
6 
7 /* Sets everything common except 'dev', since that is done under locking */
8 void vif_device_init(struct vif_device *v,
9 		     struct net_device *dev,
10 		     unsigned long rate_limit,
11 		     unsigned char threshold,
12 		     unsigned short flags,
13 		     unsigned short get_iflink_mask)
14 {
15 	v->dev = NULL;
16 	v->bytes_in = 0;
17 	v->bytes_out = 0;
18 	v->pkt_in = 0;
19 	v->pkt_out = 0;
20 	v->rate_limit = rate_limit;
21 	v->flags = flags;
22 	v->threshold = threshold;
23 	if (v->flags & get_iflink_mask)
24 		v->link = dev_get_iflink(dev);
25 	else
26 		v->link = dev->ifindex;
27 }
28 EXPORT_SYMBOL(vif_device_init);
29 
30 struct mr_table *
31 mr_table_alloc(struct net *net, u32 id,
32 	       struct mr_table_ops *ops,
33 	       void (*expire_func)(struct timer_list *t),
34 	       void (*table_set)(struct mr_table *mrt,
35 				 struct net *net))
36 {
37 	struct mr_table *mrt;
38 
39 	mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
40 	if (!mrt)
41 		return NULL;
42 	mrt->id = id;
43 	write_pnet(&mrt->net, net);
44 
45 	mrt->ops = *ops;
46 	if (rhltable_init(&mrt->mfc_hash, mrt->ops.rht_params)) {
47 		kfree(mrt);
48 		return NULL;
49 	}
50 	INIT_LIST_HEAD(&mrt->mfc_cache_list);
51 	INIT_LIST_HEAD(&mrt->mfc_unres_queue);
52 
53 	timer_setup(&mrt->ipmr_expire_timer, expire_func, 0);
54 
55 	mrt->mroute_reg_vif_num = -1;
56 	table_set(mrt, net);
57 	return mrt;
58 }
59 EXPORT_SYMBOL(mr_table_alloc);
60 
61 void *mr_mfc_find_parent(struct mr_table *mrt, void *hasharg, int parent)
62 {
63 	struct rhlist_head *tmp, *list;
64 	struct mr_mfc *c;
65 
66 	list = rhltable_lookup(&mrt->mfc_hash, hasharg, *mrt->ops.rht_params);
67 	rhl_for_each_entry_rcu(c, tmp, list, mnode)
68 		if (parent == -1 || parent == c->mfc_parent)
69 			return c;
70 
71 	return NULL;
72 }
73 EXPORT_SYMBOL(mr_mfc_find_parent);
74 
75 void *mr_mfc_find_any_parent(struct mr_table *mrt, int vifi)
76 {
77 	struct rhlist_head *tmp, *list;
78 	struct mr_mfc *c;
79 
80 	list = rhltable_lookup(&mrt->mfc_hash, mrt->ops.cmparg_any,
81 			       *mrt->ops.rht_params);
82 	rhl_for_each_entry_rcu(c, tmp, list, mnode)
83 		if (c->mfc_un.res.ttls[vifi] < 255)
84 			return c;
85 
86 	return NULL;
87 }
88 EXPORT_SYMBOL(mr_mfc_find_any_parent);
89 
90 void *mr_mfc_find_any(struct mr_table *mrt, int vifi, void *hasharg)
91 {
92 	struct rhlist_head *tmp, *list;
93 	struct mr_mfc *c, *proxy;
94 
95 	list = rhltable_lookup(&mrt->mfc_hash, hasharg, *mrt->ops.rht_params);
96 	rhl_for_each_entry_rcu(c, tmp, list, mnode) {
97 		if (c->mfc_un.res.ttls[vifi] < 255)
98 			return c;
99 
100 		/* It's ok if the vifi is part of the static tree */
101 		proxy = mr_mfc_find_any_parent(mrt, c->mfc_parent);
102 		if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
103 			return c;
104 	}
105 
106 	return mr_mfc_find_any_parent(mrt, vifi);
107 }
108 EXPORT_SYMBOL(mr_mfc_find_any);
109 
110 #ifdef CONFIG_PROC_FS
111 void *mr_vif_seq_idx(struct net *net, struct mr_vif_iter *iter, loff_t pos)
112 {
113 	struct mr_table *mrt = iter->mrt;
114 
115 	for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
116 		if (!VIF_EXISTS(mrt, iter->ct))
117 			continue;
118 		if (pos-- == 0)
119 			return &mrt->vif_table[iter->ct];
120 	}
121 	return NULL;
122 }
123 EXPORT_SYMBOL(mr_vif_seq_idx);
124 
125 void *mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
126 {
127 	struct mr_vif_iter *iter = seq->private;
128 	struct net *net = seq_file_net(seq);
129 	struct mr_table *mrt = iter->mrt;
130 
131 	++*pos;
132 	if (v == SEQ_START_TOKEN)
133 		return mr_vif_seq_idx(net, iter, 0);
134 
135 	while (++iter->ct < mrt->maxvif) {
136 		if (!VIF_EXISTS(mrt, iter->ct))
137 			continue;
138 		return &mrt->vif_table[iter->ct];
139 	}
140 	return NULL;
141 }
142 EXPORT_SYMBOL(mr_vif_seq_next);
143 
144 void *mr_mfc_seq_idx(struct net *net,
145 		     struct mr_mfc_iter *it, loff_t pos)
146 {
147 	struct mr_table *mrt = it->mrt;
148 	struct mr_mfc *mfc;
149 
150 	rcu_read_lock();
151 	it->cache = &mrt->mfc_cache_list;
152 	list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list)
153 		if (pos-- == 0)
154 			return mfc;
155 	rcu_read_unlock();
156 
157 	spin_lock_bh(it->lock);
158 	it->cache = &mrt->mfc_unres_queue;
159 	list_for_each_entry(mfc, it->cache, list)
160 		if (pos-- == 0)
161 			return mfc;
162 	spin_unlock_bh(it->lock);
163 
164 	it->cache = NULL;
165 	return NULL;
166 }
167 EXPORT_SYMBOL(mr_mfc_seq_idx);
168 
169 void *mr_mfc_seq_next(struct seq_file *seq, void *v,
170 		      loff_t *pos)
171 {
172 	struct mr_mfc_iter *it = seq->private;
173 	struct net *net = seq_file_net(seq);
174 	struct mr_table *mrt = it->mrt;
175 	struct mr_mfc *c = v;
176 
177 	++*pos;
178 
179 	if (v == SEQ_START_TOKEN)
180 		return mr_mfc_seq_idx(net, seq->private, 0);
181 
182 	if (c->list.next != it->cache)
183 		return list_entry(c->list.next, struct mr_mfc, list);
184 
185 	if (it->cache == &mrt->mfc_unres_queue)
186 		goto end_of_list;
187 
188 	/* exhausted cache_array, show unresolved */
189 	rcu_read_unlock();
190 	it->cache = &mrt->mfc_unres_queue;
191 
192 	spin_lock_bh(it->lock);
193 	if (!list_empty(it->cache))
194 		return list_first_entry(it->cache, struct mr_mfc, list);
195 
196 end_of_list:
197 	spin_unlock_bh(it->lock);
198 	it->cache = NULL;
199 
200 	return NULL;
201 }
202 EXPORT_SYMBOL(mr_mfc_seq_next);
203 #endif
204 
205 int mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
206 		   struct mr_mfc *c, struct rtmsg *rtm)
207 {
208 	struct rta_mfc_stats mfcs;
209 	struct nlattr *mp_attr;
210 	struct rtnexthop *nhp;
211 	unsigned long lastuse;
212 	int ct;
213 
214 	/* If cache is unresolved, don't try to parse IIF and OIF */
215 	if (c->mfc_parent >= MAXVIFS) {
216 		rtm->rtm_flags |= RTNH_F_UNRESOLVED;
217 		return -ENOENT;
218 	}
219 
220 	if (VIF_EXISTS(mrt, c->mfc_parent) &&
221 	    nla_put_u32(skb, RTA_IIF,
222 			mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
223 		return -EMSGSIZE;
224 
225 	if (c->mfc_flags & MFC_OFFLOAD)
226 		rtm->rtm_flags |= RTNH_F_OFFLOAD;
227 
228 	mp_attr = nla_nest_start(skb, RTA_MULTIPATH);
229 	if (!mp_attr)
230 		return -EMSGSIZE;
231 
232 	for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
233 		if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
234 			struct vif_device *vif;
235 
236 			nhp = nla_reserve_nohdr(skb, sizeof(*nhp));
237 			if (!nhp) {
238 				nla_nest_cancel(skb, mp_attr);
239 				return -EMSGSIZE;
240 			}
241 
242 			nhp->rtnh_flags = 0;
243 			nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
244 			vif = &mrt->vif_table[ct];
245 			nhp->rtnh_ifindex = vif->dev->ifindex;
246 			nhp->rtnh_len = sizeof(*nhp);
247 		}
248 	}
249 
250 	nla_nest_end(skb, mp_attr);
251 
252 	lastuse = READ_ONCE(c->mfc_un.res.lastuse);
253 	lastuse = time_after_eq(jiffies, lastuse) ? jiffies - lastuse : 0;
254 
255 	mfcs.mfcs_packets = c->mfc_un.res.pkt;
256 	mfcs.mfcs_bytes = c->mfc_un.res.bytes;
257 	mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
258 	if (nla_put_64bit(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs, RTA_PAD) ||
259 	    nla_put_u64_64bit(skb, RTA_EXPIRES, jiffies_to_clock_t(lastuse),
260 			      RTA_PAD))
261 		return -EMSGSIZE;
262 
263 	rtm->rtm_type = RTN_MULTICAST;
264 	return 1;
265 }
266 EXPORT_SYMBOL(mr_fill_mroute);
267 
268 int mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb,
269 		     struct mr_table *(*iter)(struct net *net,
270 					      struct mr_table *mrt),
271 		     int (*fill)(struct mr_table *mrt,
272 				 struct sk_buff *skb,
273 				 u32 portid, u32 seq, struct mr_mfc *c,
274 				 int cmd, int flags),
275 		     spinlock_t *lock)
276 {
277 	unsigned int t = 0, e = 0, s_t = cb->args[0], s_e = cb->args[1];
278 	struct net *net = sock_net(skb->sk);
279 	struct mr_table *mrt;
280 	struct mr_mfc *mfc;
281 
282 	rcu_read_lock();
283 	for (mrt = iter(net, NULL); mrt; mrt = iter(net, mrt)) {
284 		if (t < s_t)
285 			goto next_table;
286 		list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list) {
287 			if (e < s_e)
288 				goto next_entry;
289 			if (fill(mrt, skb, NETLINK_CB(cb->skb).portid,
290 				 cb->nlh->nlmsg_seq, mfc,
291 				 RTM_NEWROUTE, NLM_F_MULTI) < 0)
292 				goto done;
293 next_entry:
294 			e++;
295 		}
296 		e = 0;
297 		s_e = 0;
298 
299 		spin_lock_bh(lock);
300 		list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
301 			if (e < s_e)
302 				goto next_entry2;
303 			if (fill(mrt, skb, NETLINK_CB(cb->skb).portid,
304 				 cb->nlh->nlmsg_seq, mfc,
305 				 RTM_NEWROUTE, NLM_F_MULTI) < 0) {
306 				spin_unlock_bh(lock);
307 				goto done;
308 			}
309 next_entry2:
310 			e++;
311 		}
312 		spin_unlock_bh(lock);
313 		e = 0;
314 		s_e = 0;
315 next_table:
316 		t++;
317 	}
318 done:
319 	rcu_read_unlock();
320 
321 	cb->args[1] = e;
322 	cb->args[0] = t;
323 
324 	return skb->len;
325 }
326 EXPORT_SYMBOL(mr_rtm_dumproute);
327 
328 int mr_dump(struct net *net, struct notifier_block *nb, unsigned short family,
329 	    int (*rules_dump)(struct net *net,
330 			      struct notifier_block *nb),
331 	    struct mr_table *(*mr_iter)(struct net *net,
332 					struct mr_table *mrt),
333 	    rwlock_t *mrt_lock)
334 {
335 	struct mr_table *mrt;
336 	int err;
337 
338 	err = rules_dump(net, nb);
339 	if (err)
340 		return err;
341 
342 	for (mrt = mr_iter(net, NULL); mrt; mrt = mr_iter(net, mrt)) {
343 		struct vif_device *v = &mrt->vif_table[0];
344 		struct mr_mfc *mfc;
345 		int vifi;
346 
347 		/* Notifiy on table VIF entries */
348 		read_lock(mrt_lock);
349 		for (vifi = 0; vifi < mrt->maxvif; vifi++, v++) {
350 			if (!v->dev)
351 				continue;
352 
353 			mr_call_vif_notifier(nb, net, family,
354 					     FIB_EVENT_VIF_ADD,
355 					     v, vifi, mrt->id);
356 		}
357 		read_unlock(mrt_lock);
358 
359 		/* Notify on table MFC entries */
360 		list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list)
361 			mr_call_mfc_notifier(nb, net, family,
362 					     FIB_EVENT_ENTRY_ADD,
363 					     mfc, mrt->id);
364 	}
365 
366 	return 0;
367 }
368 EXPORT_SYMBOL(mr_dump);
369