xref: /linux/include/linux/mroute_base.h (revision 71e2f4dd5a65bd8dbca0b77661e75eea471168f8)
1 #ifndef __LINUX_MROUTE_BASE_H
2 #define __LINUX_MROUTE_BASE_H
3 
4 #include <linux/netdevice.h>
5 #include <linux/rhashtable-types.h>
6 #include <linux/spinlock.h>
7 #include <net/net_namespace.h>
8 #include <net/sock.h>
9 #include <net/fib_notifier.h>
10 #include <net/ip_fib.h>
11 
12 /**
13  * struct vif_device - interface representor for multicast routing
14  * @dev: network device being used
15  * @bytes_in: statistic; bytes ingressing
16  * @bytes_out: statistic; bytes egresing
17  * @pkt_in: statistic; packets ingressing
18  * @pkt_out: statistic; packets egressing
19  * @rate_limit: Traffic shaping (NI)
20  * @threshold: TTL threshold
21  * @flags: Control flags
22  * @link: Physical interface index
23  * @dev_parent_id: device parent id
24  * @local: Local address
25  * @remote: Remote address for tunnels
26  */
27 struct vif_device {
28 	struct net_device *dev;
29 	unsigned long bytes_in, bytes_out;
30 	unsigned long pkt_in, pkt_out;
31 	unsigned long rate_limit;
32 	unsigned char threshold;
33 	unsigned short flags;
34 	int link;
35 
36 	/* Currently only used by ipmr */
37 	struct netdev_phys_item_id dev_parent_id;
38 	__be32 local, remote;
39 };
40 
41 struct vif_entry_notifier_info {
42 	struct fib_notifier_info info;
43 	struct net_device *dev;
44 	unsigned short vif_index;
45 	unsigned short vif_flags;
46 	u32 tb_id;
47 };
48 
49 static inline int mr_call_vif_notifier(struct notifier_block *nb,
50 				       struct net *net,
51 				       unsigned short family,
52 				       enum fib_event_type event_type,
53 				       struct vif_device *vif,
54 				       unsigned short vif_index, u32 tb_id)
55 {
56 	struct vif_entry_notifier_info info = {
57 		.info = {
58 			.family = family,
59 			.net = net,
60 		},
61 		.dev = vif->dev,
62 		.vif_index = vif_index,
63 		.vif_flags = vif->flags,
64 		.tb_id = tb_id,
65 	};
66 
67 	return call_fib_notifier(nb, net, event_type, &info.info);
68 }
69 
70 static inline int mr_call_vif_notifiers(struct net *net,
71 					unsigned short family,
72 					enum fib_event_type event_type,
73 					struct vif_device *vif,
74 					unsigned short vif_index, u32 tb_id,
75 					unsigned int *ipmr_seq)
76 {
77 	struct vif_entry_notifier_info info = {
78 		.info = {
79 			.family = family,
80 			.net = net,
81 		},
82 		.dev = vif->dev,
83 		.vif_index = vif_index,
84 		.vif_flags = vif->flags,
85 		.tb_id = tb_id,
86 	};
87 
88 	ASSERT_RTNL();
89 	(*ipmr_seq)++;
90 	return call_fib_notifiers(net, event_type, &info.info);
91 }
92 
93 #ifndef MAXVIFS
94 /* This one is nasty; value is defined in uapi using different symbols for
95  * mroute and morute6 but both map into same 32.
96  */
97 #define MAXVIFS	32
98 #endif
99 
100 #define VIF_EXISTS(_mrt, _idx) (!!((_mrt)->vif_table[_idx].dev))
101 
102 /* mfc_flags:
103  * MFC_STATIC - the entry was added statically (not by a routing daemon)
104  * MFC_OFFLOAD - the entry was offloaded to the hardware
105  */
106 enum {
107 	MFC_STATIC = BIT(0),
108 	MFC_OFFLOAD = BIT(1),
109 };
110 
111 /**
112  * struct mr_mfc - common multicast routing entries
113  * @mnode: rhashtable list
114  * @mfc_parent: source interface (iif)
115  * @mfc_flags: entry flags
116  * @expires: unresolved entry expire time
117  * @unresolved: unresolved cached skbs
118  * @last_assert: time of last assert
119  * @minvif: minimum VIF id
120  * @maxvif: maximum VIF id
121  * @bytes: bytes that have passed for this entry
122  * @pkt: packets that have passed for this entry
123  * @wrong_if: number of wrong source interface hits
124  * @lastuse: time of last use of the group (traffic or update)
125  * @ttls: OIF TTL threshold array
126  * @refcount: reference count for this entry
127  * @list: global entry list
128  * @rcu: used for entry destruction
129  * @free: Operation used for freeing an entry under RCU
130  */
131 struct mr_mfc {
132 	struct rhlist_head mnode;
133 	unsigned short mfc_parent;
134 	int mfc_flags;
135 
136 	union {
137 		struct {
138 			unsigned long expires;
139 			struct sk_buff_head unresolved;
140 		} unres;
141 		struct {
142 			unsigned long last_assert;
143 			int minvif;
144 			int maxvif;
145 			unsigned long bytes;
146 			unsigned long pkt;
147 			unsigned long wrong_if;
148 			unsigned long lastuse;
149 			unsigned char ttls[MAXVIFS];
150 			refcount_t refcount;
151 		} res;
152 	} mfc_un;
153 	struct list_head list;
154 	struct rcu_head	rcu;
155 	void (*free)(struct rcu_head *head);
156 };
157 
158 static inline void mr_cache_put(struct mr_mfc *c)
159 {
160 	if (refcount_dec_and_test(&c->mfc_un.res.refcount))
161 		call_rcu(&c->rcu, c->free);
162 }
163 
164 static inline void mr_cache_hold(struct mr_mfc *c)
165 {
166 	refcount_inc(&c->mfc_un.res.refcount);
167 }
168 
169 struct mfc_entry_notifier_info {
170 	struct fib_notifier_info info;
171 	struct mr_mfc *mfc;
172 	u32 tb_id;
173 };
174 
175 static inline int mr_call_mfc_notifier(struct notifier_block *nb,
176 				       struct net *net,
177 				       unsigned short family,
178 				       enum fib_event_type event_type,
179 				       struct mr_mfc *mfc, u32 tb_id)
180 {
181 	struct mfc_entry_notifier_info info = {
182 		.info = {
183 			.family = family,
184 			.net = net,
185 		},
186 		.mfc = mfc,
187 		.tb_id = tb_id
188 	};
189 
190 	return call_fib_notifier(nb, net, event_type, &info.info);
191 }
192 
193 static inline int mr_call_mfc_notifiers(struct net *net,
194 					unsigned short family,
195 					enum fib_event_type event_type,
196 					struct mr_mfc *mfc, u32 tb_id,
197 					unsigned int *ipmr_seq)
198 {
199 	struct mfc_entry_notifier_info info = {
200 		.info = {
201 			.family = family,
202 			.net = net,
203 		},
204 		.mfc = mfc,
205 		.tb_id = tb_id
206 	};
207 
208 	ASSERT_RTNL();
209 	(*ipmr_seq)++;
210 	return call_fib_notifiers(net, event_type, &info.info);
211 }
212 
213 struct mr_table;
214 
215 /**
216  * struct mr_table_ops - callbacks and info for protocol-specific ops
217  * @rht_params: parameters for accessing the MFC hash
218  * @cmparg_any: a hash key to be used for matching on (*,*) routes
219  */
220 struct mr_table_ops {
221 	const struct rhashtable_params *rht_params;
222 	void *cmparg_any;
223 };
224 
225 /**
226  * struct mr_table - a multicast routing table
227  * @list: entry within a list of multicast routing tables
228  * @net: net where this table belongs
229  * @ops: protocol specific operations
230  * @id: identifier of the table
231  * @mroute_sk: socket associated with the table
232  * @ipmr_expire_timer: timer for handling unresolved routes
233  * @mfc_unres_queue: list of unresolved MFC entries
234  * @vif_table: array containing all possible vifs
235  * @mfc_hash: Hash table of all resolved routes for easy lookup
236  * @mfc_cache_list: list of resovled routes for possible traversal
237  * @maxvif: Identifier of highest value vif currently in use
238  * @cache_resolve_queue_len: current size of unresolved queue
239  * @mroute_do_assert: Whether to inform userspace on wrong ingress
240  * @mroute_do_pim: Whether to receive IGMP PIMv1
241  * @mroute_reg_vif_num: PIM-device vif index
242  */
243 struct mr_table {
244 	struct list_head	list;
245 	possible_net_t		net;
246 	struct mr_table_ops	ops;
247 	u32			id;
248 	struct sock __rcu	*mroute_sk;
249 	struct timer_list	ipmr_expire_timer;
250 	struct list_head	mfc_unres_queue;
251 	struct vif_device	vif_table[MAXVIFS];
252 	struct rhltable		mfc_hash;
253 	struct list_head	mfc_cache_list;
254 	int			maxvif;
255 	atomic_t		cache_resolve_queue_len;
256 	bool			mroute_do_assert;
257 	bool			mroute_do_pim;
258 	bool			mroute_do_wrvifwhole;
259 	int			mroute_reg_vif_num;
260 };
261 
262 #ifdef CONFIG_IP_MROUTE_COMMON
263 void vif_device_init(struct vif_device *v,
264 		     struct net_device *dev,
265 		     unsigned long rate_limit,
266 		     unsigned char threshold,
267 		     unsigned short flags,
268 		     unsigned short get_iflink_mask);
269 
270 struct mr_table *
271 mr_table_alloc(struct net *net, u32 id,
272 	       struct mr_table_ops *ops,
273 	       void (*expire_func)(struct timer_list *t),
274 	       void (*table_set)(struct mr_table *mrt,
275 				 struct net *net));
276 
277 /* These actually return 'struct mr_mfc *', but to avoid need for explicit
278  * castings they simply return void.
279  */
280 void *mr_mfc_find_parent(struct mr_table *mrt,
281 			 void *hasharg, int parent);
282 void *mr_mfc_find_any_parent(struct mr_table *mrt, int vifi);
283 void *mr_mfc_find_any(struct mr_table *mrt, int vifi, void *hasharg);
284 
285 int mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
286 		   struct mr_mfc *c, struct rtmsg *rtm);
287 int mr_table_dump(struct mr_table *mrt, struct sk_buff *skb,
288 		  struct netlink_callback *cb,
289 		  int (*fill)(struct mr_table *mrt, struct sk_buff *skb,
290 			      u32 portid, u32 seq, struct mr_mfc *c,
291 			      int cmd, int flags),
292 		  spinlock_t *lock, struct fib_dump_filter *filter);
293 int mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb,
294 		     struct mr_table *(*iter)(struct net *net,
295 					      struct mr_table *mrt),
296 		     int (*fill)(struct mr_table *mrt,
297 				 struct sk_buff *skb,
298 				 u32 portid, u32 seq, struct mr_mfc *c,
299 				 int cmd, int flags),
300 		     spinlock_t *lock, struct fib_dump_filter *filter);
301 
302 int mr_dump(struct net *net, struct notifier_block *nb, unsigned short family,
303 	    int (*rules_dump)(struct net *net,
304 			      struct notifier_block *nb),
305 	    struct mr_table *(*mr_iter)(struct net *net,
306 					struct mr_table *mrt),
307 	    rwlock_t *mrt_lock);
308 #else
309 static inline void vif_device_init(struct vif_device *v,
310 				   struct net_device *dev,
311 				   unsigned long rate_limit,
312 				   unsigned char threshold,
313 				   unsigned short flags,
314 				   unsigned short get_iflink_mask)
315 {
316 }
317 
318 static inline void *mr_mfc_find_parent(struct mr_table *mrt,
319 				       void *hasharg, int parent)
320 {
321 	return NULL;
322 }
323 
324 static inline void *mr_mfc_find_any_parent(struct mr_table *mrt,
325 					   int vifi)
326 {
327 	return NULL;
328 }
329 
330 static inline struct mr_mfc *mr_mfc_find_any(struct mr_table *mrt,
331 					     int vifi, void *hasharg)
332 {
333 	return NULL;
334 }
335 
336 static inline int mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
337 				 struct mr_mfc *c, struct rtmsg *rtm)
338 {
339 	return -EINVAL;
340 }
341 
342 static inline int
343 mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb,
344 		 struct mr_table *(*iter)(struct net *net,
345 					  struct mr_table *mrt),
346 		 int (*fill)(struct mr_table *mrt,
347 			     struct sk_buff *skb,
348 			     u32 portid, u32 seq, struct mr_mfc *c,
349 			     int cmd, int flags),
350 		 spinlock_t *lock, struct fib_dump_filter *filter)
351 {
352 	return -EINVAL;
353 }
354 
355 static inline int mr_dump(struct net *net, struct notifier_block *nb,
356 			  unsigned short family,
357 			  int (*rules_dump)(struct net *net,
358 					    struct notifier_block *nb),
359 			  struct mr_table *(*mr_iter)(struct net *net,
360 						      struct mr_table *mrt),
361 			  rwlock_t *mrt_lock)
362 {
363 	return -EINVAL;
364 }
365 #endif
366 
367 static inline void *mr_mfc_find(struct mr_table *mrt, void *hasharg)
368 {
369 	return mr_mfc_find_parent(mrt, hasharg, -1);
370 }
371 
372 #ifdef CONFIG_PROC_FS
373 struct mr_vif_iter {
374 	struct seq_net_private p;
375 	struct mr_table *mrt;
376 	int ct;
377 };
378 
379 struct mr_mfc_iter {
380 	struct seq_net_private p;
381 	struct mr_table *mrt;
382 	struct list_head *cache;
383 
384 	/* Lock protecting the mr_table's unresolved queue */
385 	spinlock_t *lock;
386 };
387 
388 #ifdef CONFIG_IP_MROUTE_COMMON
389 void *mr_vif_seq_idx(struct net *net, struct mr_vif_iter *iter, loff_t pos);
390 void *mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos);
391 
392 static inline void *mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
393 {
394 	return *pos ? mr_vif_seq_idx(seq_file_net(seq),
395 				     seq->private, *pos - 1)
396 		    : SEQ_START_TOKEN;
397 }
398 
399 /* These actually return 'struct mr_mfc *', but to avoid need for explicit
400  * castings they simply return void.
401  */
402 void *mr_mfc_seq_idx(struct net *net,
403 		     struct mr_mfc_iter *it, loff_t pos);
404 void *mr_mfc_seq_next(struct seq_file *seq, void *v,
405 		      loff_t *pos);
406 
407 static inline void *mr_mfc_seq_start(struct seq_file *seq, loff_t *pos,
408 				     struct mr_table *mrt, spinlock_t *lock)
409 {
410 	struct mr_mfc_iter *it = seq->private;
411 
412 	it->mrt = mrt;
413 	it->cache = NULL;
414 	it->lock = lock;
415 
416 	return *pos ? mr_mfc_seq_idx(seq_file_net(seq),
417 				     seq->private, *pos - 1)
418 		    : SEQ_START_TOKEN;
419 }
420 
421 static inline void mr_mfc_seq_stop(struct seq_file *seq, void *v)
422 {
423 	struct mr_mfc_iter *it = seq->private;
424 	struct mr_table *mrt = it->mrt;
425 
426 	if (it->cache == &mrt->mfc_unres_queue)
427 		spin_unlock_bh(it->lock);
428 	else if (it->cache == &mrt->mfc_cache_list)
429 		rcu_read_unlock();
430 }
431 #else
432 static inline void *mr_vif_seq_idx(struct net *net, struct mr_vif_iter *iter,
433 				   loff_t pos)
434 {
435 	return NULL;
436 }
437 
438 static inline void *mr_vif_seq_next(struct seq_file *seq,
439 				    void *v, loff_t *pos)
440 {
441 	return NULL;
442 }
443 
444 static inline void *mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
445 {
446 	return NULL;
447 }
448 
449 static inline void *mr_mfc_seq_idx(struct net *net,
450 				   struct mr_mfc_iter *it, loff_t pos)
451 {
452 	return NULL;
453 }
454 
455 static inline void *mr_mfc_seq_next(struct seq_file *seq, void *v,
456 				    loff_t *pos)
457 {
458 	return NULL;
459 }
460 
461 static inline void *mr_mfc_seq_start(struct seq_file *seq, loff_t *pos,
462 				     struct mr_table *mrt, spinlock_t *lock)
463 {
464 	return NULL;
465 }
466 
467 static inline void mr_mfc_seq_stop(struct seq_file *seq, void *v)
468 {
469 }
470 #endif
471 #endif
472 #endif
473