xref: /linux/include/linux/mroute_base.h (revision f79e4d5f92a129a1159c973735007d4ddc8541f3)
1 #ifndef __LINUX_MROUTE_BASE_H
2 #define __LINUX_MROUTE_BASE_H
3 
4 #include <linux/netdevice.h>
5 #include <linux/rhashtable.h>
6 #include <linux/spinlock.h>
7 #include <net/net_namespace.h>
8 #include <net/sock.h>
9 #include <net/fib_notifier.h>
10 
11 /**
12  * struct vif_device - interface representor for multicast routing
13  * @dev: network device being used
14  * @bytes_in: statistic; bytes ingressing
15  * @bytes_out: statistic; bytes egresing
16  * @pkt_in: statistic; packets ingressing
17  * @pkt_out: statistic; packets egressing
18  * @rate_limit: Traffic shaping (NI)
19  * @threshold: TTL threshold
20  * @flags: Control flags
21  * @link: Physical interface index
22  * @dev_parent_id: device parent id
23  * @local: Local address
24  * @remote: Remote address for tunnels
25  */
26 struct vif_device {
27 	struct net_device *dev;
28 	unsigned long bytes_in, bytes_out;
29 	unsigned long pkt_in, pkt_out;
30 	unsigned long rate_limit;
31 	unsigned char threshold;
32 	unsigned short flags;
33 	int link;
34 
35 	/* Currently only used by ipmr */
36 	struct netdev_phys_item_id dev_parent_id;
37 	__be32 local, remote;
38 };
39 
40 struct vif_entry_notifier_info {
41 	struct fib_notifier_info info;
42 	struct net_device *dev;
43 	unsigned short vif_index;
44 	unsigned short vif_flags;
45 	u32 tb_id;
46 };
47 
48 static inline int mr_call_vif_notifier(struct notifier_block *nb,
49 				       struct net *net,
50 				       unsigned short family,
51 				       enum fib_event_type event_type,
52 				       struct vif_device *vif,
53 				       unsigned short vif_index, u32 tb_id)
54 {
55 	struct vif_entry_notifier_info info = {
56 		.info = {
57 			.family = family,
58 			.net = net,
59 		},
60 		.dev = vif->dev,
61 		.vif_index = vif_index,
62 		.vif_flags = vif->flags,
63 		.tb_id = tb_id,
64 	};
65 
66 	return call_fib_notifier(nb, net, event_type, &info.info);
67 }
68 
69 static inline int mr_call_vif_notifiers(struct net *net,
70 					unsigned short family,
71 					enum fib_event_type event_type,
72 					struct vif_device *vif,
73 					unsigned short vif_index, u32 tb_id,
74 					unsigned int *ipmr_seq)
75 {
76 	struct vif_entry_notifier_info info = {
77 		.info = {
78 			.family = family,
79 			.net = net,
80 		},
81 		.dev = vif->dev,
82 		.vif_index = vif_index,
83 		.vif_flags = vif->flags,
84 		.tb_id = tb_id,
85 	};
86 
87 	ASSERT_RTNL();
88 	(*ipmr_seq)++;
89 	return call_fib_notifiers(net, event_type, &info.info);
90 }
91 
92 #ifndef MAXVIFS
93 /* This one is nasty; value is defined in uapi using different symbols for
94  * mroute and morute6 but both map into same 32.
95  */
96 #define MAXVIFS	32
97 #endif
98 
99 #define VIF_EXISTS(_mrt, _idx) (!!((_mrt)->vif_table[_idx].dev))
100 
101 /* mfc_flags:
102  * MFC_STATIC - the entry was added statically (not by a routing daemon)
103  * MFC_OFFLOAD - the entry was offloaded to the hardware
104  */
105 enum {
106 	MFC_STATIC = BIT(0),
107 	MFC_OFFLOAD = BIT(1),
108 };
109 
110 /**
111  * struct mr_mfc - common multicast routing entries
112  * @mnode: rhashtable list
113  * @mfc_parent: source interface (iif)
114  * @mfc_flags: entry flags
115  * @expires: unresolved entry expire time
116  * @unresolved: unresolved cached skbs
117  * @last_assert: time of last assert
118  * @minvif: minimum VIF id
119  * @maxvif: maximum VIF id
120  * @bytes: bytes that have passed for this entry
121  * @pkt: packets that have passed for this entry
122  * @wrong_if: number of wrong source interface hits
123  * @lastuse: time of last use of the group (traffic or update)
124  * @ttls: OIF TTL threshold array
125  * @refcount: reference count for this entry
126  * @list: global entry list
127  * @rcu: used for entry destruction
128  * @free: Operation used for freeing an entry under RCU
129  */
130 struct mr_mfc {
131 	struct rhlist_head mnode;
132 	unsigned short mfc_parent;
133 	int mfc_flags;
134 
135 	union {
136 		struct {
137 			unsigned long expires;
138 			struct sk_buff_head unresolved;
139 		} unres;
140 		struct {
141 			unsigned long last_assert;
142 			int minvif;
143 			int maxvif;
144 			unsigned long bytes;
145 			unsigned long pkt;
146 			unsigned long wrong_if;
147 			unsigned long lastuse;
148 			unsigned char ttls[MAXVIFS];
149 			refcount_t refcount;
150 		} res;
151 	} mfc_un;
152 	struct list_head list;
153 	struct rcu_head	rcu;
154 	void (*free)(struct rcu_head *head);
155 };
156 
157 static inline void mr_cache_put(struct mr_mfc *c)
158 {
159 	if (refcount_dec_and_test(&c->mfc_un.res.refcount))
160 		call_rcu(&c->rcu, c->free);
161 }
162 
163 static inline void mr_cache_hold(struct mr_mfc *c)
164 {
165 	refcount_inc(&c->mfc_un.res.refcount);
166 }
167 
168 struct mfc_entry_notifier_info {
169 	struct fib_notifier_info info;
170 	struct mr_mfc *mfc;
171 	u32 tb_id;
172 };
173 
174 static inline int mr_call_mfc_notifier(struct notifier_block *nb,
175 				       struct net *net,
176 				       unsigned short family,
177 				       enum fib_event_type event_type,
178 				       struct mr_mfc *mfc, u32 tb_id)
179 {
180 	struct mfc_entry_notifier_info info = {
181 		.info = {
182 			.family = family,
183 			.net = net,
184 		},
185 		.mfc = mfc,
186 		.tb_id = tb_id
187 	};
188 
189 	return call_fib_notifier(nb, net, event_type, &info.info);
190 }
191 
192 static inline int mr_call_mfc_notifiers(struct net *net,
193 					unsigned short family,
194 					enum fib_event_type event_type,
195 					struct mr_mfc *mfc, u32 tb_id,
196 					unsigned int *ipmr_seq)
197 {
198 	struct mfc_entry_notifier_info info = {
199 		.info = {
200 			.family = family,
201 			.net = net,
202 		},
203 		.mfc = mfc,
204 		.tb_id = tb_id
205 	};
206 
207 	ASSERT_RTNL();
208 	(*ipmr_seq)++;
209 	return call_fib_notifiers(net, event_type, &info.info);
210 }
211 
212 struct mr_table;
213 
214 /**
215  * struct mr_table_ops - callbacks and info for protocol-specific ops
216  * @rht_params: parameters for accessing the MFC hash
217  * @cmparg_any: a hash key to be used for matching on (*,*) routes
218  */
219 struct mr_table_ops {
220 	const struct rhashtable_params *rht_params;
221 	void *cmparg_any;
222 };
223 
224 /**
225  * struct mr_table - a multicast routing table
226  * @list: entry within a list of multicast routing tables
227  * @net: net where this table belongs
228  * @ops: protocol specific operations
229  * @id: identifier of the table
230  * @mroute_sk: socket associated with the table
231  * @ipmr_expire_timer: timer for handling unresolved routes
232  * @mfc_unres_queue: list of unresolved MFC entries
233  * @vif_table: array containing all possible vifs
234  * @mfc_hash: Hash table of all resolved routes for easy lookup
235  * @mfc_cache_list: list of resovled routes for possible traversal
236  * @maxvif: Identifier of highest value vif currently in use
237  * @cache_resolve_queue_len: current size of unresolved queue
238  * @mroute_do_assert: Whether to inform userspace on wrong ingress
239  * @mroute_do_pim: Whether to receive IGMP PIMv1
240  * @mroute_reg_vif_num: PIM-device vif index
241  */
242 struct mr_table {
243 	struct list_head	list;
244 	possible_net_t		net;
245 	struct mr_table_ops	ops;
246 	u32			id;
247 	struct sock __rcu	*mroute_sk;
248 	struct timer_list	ipmr_expire_timer;
249 	struct list_head	mfc_unres_queue;
250 	struct vif_device	vif_table[MAXVIFS];
251 	struct rhltable		mfc_hash;
252 	struct list_head	mfc_cache_list;
253 	int			maxvif;
254 	atomic_t		cache_resolve_queue_len;
255 	bool			mroute_do_assert;
256 	bool			mroute_do_pim;
257 	int			mroute_reg_vif_num;
258 };
259 
260 #ifdef CONFIG_IP_MROUTE_COMMON
261 void vif_device_init(struct vif_device *v,
262 		     struct net_device *dev,
263 		     unsigned long rate_limit,
264 		     unsigned char threshold,
265 		     unsigned short flags,
266 		     unsigned short get_iflink_mask);
267 
268 struct mr_table *
269 mr_table_alloc(struct net *net, u32 id,
270 	       struct mr_table_ops *ops,
271 	       void (*expire_func)(struct timer_list *t),
272 	       void (*table_set)(struct mr_table *mrt,
273 				 struct net *net));
274 
275 /* These actually return 'struct mr_mfc *', but to avoid need for explicit
276  * castings they simply return void.
277  */
278 void *mr_mfc_find_parent(struct mr_table *mrt,
279 			 void *hasharg, int parent);
280 void *mr_mfc_find_any_parent(struct mr_table *mrt, int vifi);
281 void *mr_mfc_find_any(struct mr_table *mrt, int vifi, void *hasharg);
282 
283 int mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
284 		   struct mr_mfc *c, struct rtmsg *rtm);
285 int mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb,
286 		     struct mr_table *(*iter)(struct net *net,
287 					      struct mr_table *mrt),
288 		     int (*fill)(struct mr_table *mrt,
289 				 struct sk_buff *skb,
290 				 u32 portid, u32 seq, struct mr_mfc *c,
291 				 int cmd, int flags),
292 		     spinlock_t *lock);
293 
294 int mr_dump(struct net *net, struct notifier_block *nb, unsigned short family,
295 	    int (*rules_dump)(struct net *net,
296 			      struct notifier_block *nb),
297 	    struct mr_table *(*mr_iter)(struct net *net,
298 					struct mr_table *mrt),
299 	    rwlock_t *mrt_lock);
300 #else
301 static inline void vif_device_init(struct vif_device *v,
302 				   struct net_device *dev,
303 				   unsigned long rate_limit,
304 				   unsigned char threshold,
305 				   unsigned short flags,
306 				   unsigned short get_iflink_mask)
307 {
308 }
309 
310 static inline void *mr_mfc_find_parent(struct mr_table *mrt,
311 				       void *hasharg, int parent)
312 {
313 	return NULL;
314 }
315 
316 static inline void *mr_mfc_find_any_parent(struct mr_table *mrt,
317 					   int vifi)
318 {
319 	return NULL;
320 }
321 
322 static inline struct mr_mfc *mr_mfc_find_any(struct mr_table *mrt,
323 					     int vifi, void *hasharg)
324 {
325 	return NULL;
326 }
327 
328 static inline int mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
329 				 struct mr_mfc *c, struct rtmsg *rtm)
330 {
331 	return -EINVAL;
332 }
333 
334 static inline int
335 mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb,
336 		 struct mr_table *(*iter)(struct net *net,
337 					  struct mr_table *mrt),
338 		 int (*fill)(struct mr_table *mrt,
339 			     struct sk_buff *skb,
340 			     u32 portid, u32 seq, struct mr_mfc *c,
341 			     int cmd, int flags),
342 		 spinlock_t *lock)
343 {
344 	return -EINVAL;
345 }
346 
347 static inline int mr_dump(struct net *net, struct notifier_block *nb,
348 			  unsigned short family,
349 			  int (*rules_dump)(struct net *net,
350 					    struct notifier_block *nb),
351 			  struct mr_table *(*mr_iter)(struct net *net,
352 						      struct mr_table *mrt),
353 			  rwlock_t *mrt_lock)
354 {
355 	return -EINVAL;
356 }
357 #endif
358 
359 static inline void *mr_mfc_find(struct mr_table *mrt, void *hasharg)
360 {
361 	return mr_mfc_find_parent(mrt, hasharg, -1);
362 }
363 
364 #ifdef CONFIG_PROC_FS
365 struct mr_vif_iter {
366 	struct seq_net_private p;
367 	struct mr_table *mrt;
368 	int ct;
369 };
370 
371 struct mr_mfc_iter {
372 	struct seq_net_private p;
373 	struct mr_table *mrt;
374 	struct list_head *cache;
375 
376 	/* Lock protecting the mr_table's unresolved queue */
377 	spinlock_t *lock;
378 };
379 
380 #ifdef CONFIG_IP_MROUTE_COMMON
381 void *mr_vif_seq_idx(struct net *net, struct mr_vif_iter *iter, loff_t pos);
382 void *mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos);
383 
384 static inline void *mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
385 {
386 	return *pos ? mr_vif_seq_idx(seq_file_net(seq),
387 				     seq->private, *pos - 1)
388 		    : SEQ_START_TOKEN;
389 }
390 
391 /* These actually return 'struct mr_mfc *', but to avoid need for explicit
392  * castings they simply return void.
393  */
394 void *mr_mfc_seq_idx(struct net *net,
395 		     struct mr_mfc_iter *it, loff_t pos);
396 void *mr_mfc_seq_next(struct seq_file *seq, void *v,
397 		      loff_t *pos);
398 
399 static inline void *mr_mfc_seq_start(struct seq_file *seq, loff_t *pos,
400 				     struct mr_table *mrt, spinlock_t *lock)
401 {
402 	struct mr_mfc_iter *it = seq->private;
403 
404 	it->mrt = mrt;
405 	it->cache = NULL;
406 	it->lock = lock;
407 
408 	return *pos ? mr_mfc_seq_idx(seq_file_net(seq),
409 				     seq->private, *pos - 1)
410 		    : SEQ_START_TOKEN;
411 }
412 
413 static inline void mr_mfc_seq_stop(struct seq_file *seq, void *v)
414 {
415 	struct mr_mfc_iter *it = seq->private;
416 	struct mr_table *mrt = it->mrt;
417 
418 	if (it->cache == &mrt->mfc_unres_queue)
419 		spin_unlock_bh(it->lock);
420 	else if (it->cache == &mrt->mfc_cache_list)
421 		rcu_read_unlock();
422 }
423 #else
424 static inline void *mr_vif_seq_idx(struct net *net, struct mr_vif_iter *iter,
425 				   loff_t pos)
426 {
427 	return NULL;
428 }
429 
430 static inline void *mr_vif_seq_next(struct seq_file *seq,
431 				    void *v, loff_t *pos)
432 {
433 	return NULL;
434 }
435 
436 static inline void *mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
437 {
438 	return NULL;
439 }
440 
441 static inline void *mr_mfc_seq_idx(struct net *net,
442 				   struct mr_mfc_iter *it, loff_t pos)
443 {
444 	return NULL;
445 }
446 
447 static inline void *mr_mfc_seq_next(struct seq_file *seq, void *v,
448 				    loff_t *pos)
449 {
450 	return NULL;
451 }
452 
453 static inline void *mr_mfc_seq_start(struct seq_file *seq, loff_t *pos,
454 				     struct mr_table *mrt, spinlock_t *lock)
455 {
456 	return NULL;
457 }
458 
459 static inline void mr_mfc_seq_stop(struct seq_file *seq, void *v)
460 {
461 }
462 #endif
463 #endif
464 #endif
465