xref: /openbmc/linux/net/ipv4/ipmr_base.c (revision b9b77222)
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 	int err;
39 
40 	mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
41 	if (!mrt)
42 		return ERR_PTR(-ENOMEM);
43 	mrt->id = id;
44 	write_pnet(&mrt->net, net);
45 
46 	mrt->ops = *ops;
47 	err = rhltable_init(&mrt->mfc_hash, mrt->ops.rht_params);
48 	if (err) {
49 		kfree(mrt);
50 		return ERR_PTR(err);
51 	}
52 	INIT_LIST_HEAD(&mrt->mfc_cache_list);
53 	INIT_LIST_HEAD(&mrt->mfc_unres_queue);
54 
55 	timer_setup(&mrt->ipmr_expire_timer, expire_func, 0);
56 
57 	mrt->mroute_reg_vif_num = -1;
58 	table_set(mrt, net);
59 	return mrt;
60 }
61 EXPORT_SYMBOL(mr_table_alloc);
62 
63 void *mr_mfc_find_parent(struct mr_table *mrt, void *hasharg, int parent)
64 {
65 	struct rhlist_head *tmp, *list;
66 	struct mr_mfc *c;
67 
68 	list = rhltable_lookup(&mrt->mfc_hash, hasharg, *mrt->ops.rht_params);
69 	rhl_for_each_entry_rcu(c, tmp, list, mnode)
70 		if (parent == -1 || parent == c->mfc_parent)
71 			return c;
72 
73 	return NULL;
74 }
75 EXPORT_SYMBOL(mr_mfc_find_parent);
76 
77 void *mr_mfc_find_any_parent(struct mr_table *mrt, int vifi)
78 {
79 	struct rhlist_head *tmp, *list;
80 	struct mr_mfc *c;
81 
82 	list = rhltable_lookup(&mrt->mfc_hash, mrt->ops.cmparg_any,
83 			       *mrt->ops.rht_params);
84 	rhl_for_each_entry_rcu(c, tmp, list, mnode)
85 		if (c->mfc_un.res.ttls[vifi] < 255)
86 			return c;
87 
88 	return NULL;
89 }
90 EXPORT_SYMBOL(mr_mfc_find_any_parent);
91 
92 void *mr_mfc_find_any(struct mr_table *mrt, int vifi, void *hasharg)
93 {
94 	struct rhlist_head *tmp, *list;
95 	struct mr_mfc *c, *proxy;
96 
97 	list = rhltable_lookup(&mrt->mfc_hash, hasharg, *mrt->ops.rht_params);
98 	rhl_for_each_entry_rcu(c, tmp, list, mnode) {
99 		if (c->mfc_un.res.ttls[vifi] < 255)
100 			return c;
101 
102 		/* It's ok if the vifi is part of the static tree */
103 		proxy = mr_mfc_find_any_parent(mrt, c->mfc_parent);
104 		if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
105 			return c;
106 	}
107 
108 	return mr_mfc_find_any_parent(mrt, vifi);
109 }
110 EXPORT_SYMBOL(mr_mfc_find_any);
111 
112 #ifdef CONFIG_PROC_FS
113 void *mr_vif_seq_idx(struct net *net, struct mr_vif_iter *iter, loff_t pos)
114 {
115 	struct mr_table *mrt = iter->mrt;
116 
117 	for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
118 		if (!VIF_EXISTS(mrt, iter->ct))
119 			continue;
120 		if (pos-- == 0)
121 			return &mrt->vif_table[iter->ct];
122 	}
123 	return NULL;
124 }
125 EXPORT_SYMBOL(mr_vif_seq_idx);
126 
127 void *mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
128 {
129 	struct mr_vif_iter *iter = seq->private;
130 	struct net *net = seq_file_net(seq);
131 	struct mr_table *mrt = iter->mrt;
132 
133 	++*pos;
134 	if (v == SEQ_START_TOKEN)
135 		return mr_vif_seq_idx(net, iter, 0);
136 
137 	while (++iter->ct < mrt->maxvif) {
138 		if (!VIF_EXISTS(mrt, iter->ct))
139 			continue;
140 		return &mrt->vif_table[iter->ct];
141 	}
142 	return NULL;
143 }
144 EXPORT_SYMBOL(mr_vif_seq_next);
145 
146 void *mr_mfc_seq_idx(struct net *net,
147 		     struct mr_mfc_iter *it, loff_t pos)
148 {
149 	struct mr_table *mrt = it->mrt;
150 	struct mr_mfc *mfc;
151 
152 	rcu_read_lock();
153 	it->cache = &mrt->mfc_cache_list;
154 	list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list)
155 		if (pos-- == 0)
156 			return mfc;
157 	rcu_read_unlock();
158 
159 	spin_lock_bh(it->lock);
160 	it->cache = &mrt->mfc_unres_queue;
161 	list_for_each_entry(mfc, it->cache, list)
162 		if (pos-- == 0)
163 			return mfc;
164 	spin_unlock_bh(it->lock);
165 
166 	it->cache = NULL;
167 	return NULL;
168 }
169 EXPORT_SYMBOL(mr_mfc_seq_idx);
170 
171 void *mr_mfc_seq_next(struct seq_file *seq, void *v,
172 		      loff_t *pos)
173 {
174 	struct mr_mfc_iter *it = seq->private;
175 	struct net *net = seq_file_net(seq);
176 	struct mr_table *mrt = it->mrt;
177 	struct mr_mfc *c = v;
178 
179 	++*pos;
180 
181 	if (v == SEQ_START_TOKEN)
182 		return mr_mfc_seq_idx(net, seq->private, 0);
183 
184 	if (c->list.next != it->cache)
185 		return list_entry(c->list.next, struct mr_mfc, list);
186 
187 	if (it->cache == &mrt->mfc_unres_queue)
188 		goto end_of_list;
189 
190 	/* exhausted cache_array, show unresolved */
191 	rcu_read_unlock();
192 	it->cache = &mrt->mfc_unres_queue;
193 
194 	spin_lock_bh(it->lock);
195 	if (!list_empty(it->cache))
196 		return list_first_entry(it->cache, struct mr_mfc, list);
197 
198 end_of_list:
199 	spin_unlock_bh(it->lock);
200 	it->cache = NULL;
201 
202 	return NULL;
203 }
204 EXPORT_SYMBOL(mr_mfc_seq_next);
205 #endif
206 
207 int mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
208 		   struct mr_mfc *c, struct rtmsg *rtm)
209 {
210 	struct rta_mfc_stats mfcs;
211 	struct nlattr *mp_attr;
212 	struct rtnexthop *nhp;
213 	unsigned long lastuse;
214 	int ct;
215 
216 	/* If cache is unresolved, don't try to parse IIF and OIF */
217 	if (c->mfc_parent >= MAXVIFS) {
218 		rtm->rtm_flags |= RTNH_F_UNRESOLVED;
219 		return -ENOENT;
220 	}
221 
222 	if (VIF_EXISTS(mrt, c->mfc_parent) &&
223 	    nla_put_u32(skb, RTA_IIF,
224 			mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
225 		return -EMSGSIZE;
226 
227 	if (c->mfc_flags & MFC_OFFLOAD)
228 		rtm->rtm_flags |= RTNH_F_OFFLOAD;
229 
230 	mp_attr = nla_nest_start(skb, RTA_MULTIPATH);
231 	if (!mp_attr)
232 		return -EMSGSIZE;
233 
234 	for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
235 		if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
236 			struct vif_device *vif;
237 
238 			nhp = nla_reserve_nohdr(skb, sizeof(*nhp));
239 			if (!nhp) {
240 				nla_nest_cancel(skb, mp_attr);
241 				return -EMSGSIZE;
242 			}
243 
244 			nhp->rtnh_flags = 0;
245 			nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
246 			vif = &mrt->vif_table[ct];
247 			nhp->rtnh_ifindex = vif->dev->ifindex;
248 			nhp->rtnh_len = sizeof(*nhp);
249 		}
250 	}
251 
252 	nla_nest_end(skb, mp_attr);
253 
254 	lastuse = READ_ONCE(c->mfc_un.res.lastuse);
255 	lastuse = time_after_eq(jiffies, lastuse) ? jiffies - lastuse : 0;
256 
257 	mfcs.mfcs_packets = c->mfc_un.res.pkt;
258 	mfcs.mfcs_bytes = c->mfc_un.res.bytes;
259 	mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
260 	if (nla_put_64bit(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs, RTA_PAD) ||
261 	    nla_put_u64_64bit(skb, RTA_EXPIRES, jiffies_to_clock_t(lastuse),
262 			      RTA_PAD))
263 		return -EMSGSIZE;
264 
265 	rtm->rtm_type = RTN_MULTICAST;
266 	return 1;
267 }
268 EXPORT_SYMBOL(mr_fill_mroute);
269 
270 int mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb,
271 		     struct mr_table *(*iter)(struct net *net,
272 					      struct mr_table *mrt),
273 		     int (*fill)(struct mr_table *mrt,
274 				 struct sk_buff *skb,
275 				 u32 portid, u32 seq, struct mr_mfc *c,
276 				 int cmd, int flags),
277 		     spinlock_t *lock)
278 {
279 	unsigned int t = 0, e = 0, s_t = cb->args[0], s_e = cb->args[1];
280 	struct net *net = sock_net(skb->sk);
281 	struct mr_table *mrt;
282 	struct mr_mfc *mfc;
283 
284 	rcu_read_lock();
285 	for (mrt = iter(net, NULL); mrt; mrt = iter(net, mrt)) {
286 		if (t < s_t)
287 			goto next_table;
288 		list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list) {
289 			if (e < s_e)
290 				goto next_entry;
291 			if (fill(mrt, skb, NETLINK_CB(cb->skb).portid,
292 				 cb->nlh->nlmsg_seq, mfc,
293 				 RTM_NEWROUTE, NLM_F_MULTI) < 0)
294 				goto done;
295 next_entry:
296 			e++;
297 		}
298 		e = 0;
299 		s_e = 0;
300 
301 		spin_lock_bh(lock);
302 		list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
303 			if (e < s_e)
304 				goto next_entry2;
305 			if (fill(mrt, skb, NETLINK_CB(cb->skb).portid,
306 				 cb->nlh->nlmsg_seq, mfc,
307 				 RTM_NEWROUTE, NLM_F_MULTI) < 0) {
308 				spin_unlock_bh(lock);
309 				goto done;
310 			}
311 next_entry2:
312 			e++;
313 		}
314 		spin_unlock_bh(lock);
315 		e = 0;
316 		s_e = 0;
317 next_table:
318 		t++;
319 	}
320 done:
321 	rcu_read_unlock();
322 
323 	cb->args[1] = e;
324 	cb->args[0] = t;
325 
326 	return skb->len;
327 }
328 EXPORT_SYMBOL(mr_rtm_dumproute);
329 
330 int mr_dump(struct net *net, struct notifier_block *nb, unsigned short family,
331 	    int (*rules_dump)(struct net *net,
332 			      struct notifier_block *nb),
333 	    struct mr_table *(*mr_iter)(struct net *net,
334 					struct mr_table *mrt),
335 	    rwlock_t *mrt_lock)
336 {
337 	struct mr_table *mrt;
338 	int err;
339 
340 	err = rules_dump(net, nb);
341 	if (err)
342 		return err;
343 
344 	for (mrt = mr_iter(net, NULL); mrt; mrt = mr_iter(net, mrt)) {
345 		struct vif_device *v = &mrt->vif_table[0];
346 		struct mr_mfc *mfc;
347 		int vifi;
348 
349 		/* Notifiy on table VIF entries */
350 		read_lock(mrt_lock);
351 		for (vifi = 0; vifi < mrt->maxvif; vifi++, v++) {
352 			if (!v->dev)
353 				continue;
354 
355 			mr_call_vif_notifier(nb, net, family,
356 					     FIB_EVENT_VIF_ADD,
357 					     v, vifi, mrt->id);
358 		}
359 		read_unlock(mrt_lock);
360 
361 		/* Notify on table MFC entries */
362 		list_for_each_entry_rcu(mfc, &mrt->mfc_cache_list, list)
363 			mr_call_mfc_notifier(nb, net, family,
364 					     FIB_EVENT_ENTRY_ADD,
365 					     mfc, mrt->id);
366 	}
367 
368 	return 0;
369 }
370 EXPORT_SYMBOL(mr_dump);
371