xref: /linux/include/linux/mroute_base.h (revision 8a922b7728a93d837954315c98b84f6b78de0c4f)
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  * @dev_tracker: refcount tracker for @dev reference
16  * @bytes_in: statistic; bytes ingressing
17  * @bytes_out: statistic; bytes egresing
18  * @pkt_in: statistic; packets ingressing
19  * @pkt_out: statistic; packets egressing
20  * @rate_limit: Traffic shaping (NI)
21  * @threshold: TTL threshold
22  * @flags: Control flags
23  * @link: Physical interface index
24  * @dev_parent_id: device parent id
25  * @local: Local address
26  * @remote: Remote address for tunnels
27  */
28 struct vif_device {
29 	struct net_device __rcu *dev;
30 	netdevice_tracker dev_tracker;
31 	unsigned long bytes_in, bytes_out;
32 	unsigned long pkt_in, pkt_out;
33 	unsigned long rate_limit;
34 	unsigned char threshold;
35 	unsigned short flags;
36 	int link;
37 
38 	/* Currently only used by ipmr */
39 	struct netdev_phys_item_id dev_parent_id;
40 	__be32 local, remote;
41 };
42 
43 struct vif_entry_notifier_info {
44 	struct fib_notifier_info info;
45 	struct net_device *dev;
46 	unsigned short vif_index;
47 	unsigned short vif_flags;
48 	u32 tb_id;
49 };
50 
51 static inline int mr_call_vif_notifier(struct notifier_block *nb,
52 				       unsigned short family,
53 				       enum fib_event_type event_type,
54 				       struct vif_device *vif,
55 				       struct net_device *vif_dev,
56 				       unsigned short vif_index, u32 tb_id,
57 				       struct netlink_ext_ack *extack)
58 {
59 	struct vif_entry_notifier_info info = {
60 		.info = {
61 			.family = family,
62 			.extack = extack,
63 		},
64 		.dev = vif_dev,
65 		.vif_index = vif_index,
66 		.vif_flags = vif->flags,
67 		.tb_id = tb_id,
68 	};
69 
70 	return call_fib_notifier(nb, event_type, &info.info);
71 }
72 
73 static inline int mr_call_vif_notifiers(struct net *net,
74 					unsigned short family,
75 					enum fib_event_type event_type,
76 					struct vif_device *vif,
77 					struct net_device *vif_dev,
78 					unsigned short vif_index, u32 tb_id,
79 					unsigned int *ipmr_seq)
80 {
81 	struct vif_entry_notifier_info info = {
82 		.info = {
83 			.family = family,
84 		},
85 		.dev = vif_dev,
86 		.vif_index = vif_index,
87 		.vif_flags = vif->flags,
88 		.tb_id = tb_id,
89 	};
90 
91 	ASSERT_RTNL();
92 	(*ipmr_seq)++;
93 	return call_fib_notifiers(net, event_type, &info.info);
94 }
95 
96 #ifndef MAXVIFS
97 /* This one is nasty; value is defined in uapi using different symbols for
98  * mroute and morute6 but both map into same 32.
99  */
100 #define MAXVIFS	32
101 #endif
102 
103 /* Note: This helper is deprecated. */
104 #define VIF_EXISTS(_mrt, _idx) (!!rcu_access_pointer((_mrt)->vif_table[_idx].dev))
105 
106 /* mfc_flags:
107  * MFC_STATIC - the entry was added statically (not by a routing daemon)
108  * MFC_OFFLOAD - the entry was offloaded to the hardware
109  */
110 enum {
111 	MFC_STATIC = BIT(0),
112 	MFC_OFFLOAD = BIT(1),
113 };
114 
115 /**
116  * struct mr_mfc - common multicast routing entries
117  * @mnode: rhashtable list
118  * @mfc_parent: source interface (iif)
119  * @mfc_flags: entry flags
120  * @expires: unresolved entry expire time
121  * @unresolved: unresolved cached skbs
122  * @last_assert: time of last assert
123  * @minvif: minimum VIF id
124  * @maxvif: maximum VIF id
125  * @bytes: bytes that have passed for this entry
126  * @pkt: packets that have passed for this entry
127  * @wrong_if: number of wrong source interface hits
128  * @lastuse: time of last use of the group (traffic or update)
129  * @ttls: OIF TTL threshold array
130  * @refcount: reference count for this entry
131  * @list: global entry list
132  * @rcu: used for entry destruction
133  * @free: Operation used for freeing an entry under RCU
134  */
135 struct mr_mfc {
136 	struct rhlist_head mnode;
137 	unsigned short mfc_parent;
138 	int mfc_flags;
139 
140 	union {
141 		struct {
142 			unsigned long expires;
143 			struct sk_buff_head unresolved;
144 		} unres;
145 		struct {
146 			unsigned long last_assert;
147 			int minvif;
148 			int maxvif;
149 			unsigned long bytes;
150 			unsigned long pkt;
151 			unsigned long wrong_if;
152 			unsigned long lastuse;
153 			unsigned char ttls[MAXVIFS];
154 			refcount_t refcount;
155 		} res;
156 	} mfc_un;
157 	struct list_head list;
158 	struct rcu_head	rcu;
159 	void (*free)(struct rcu_head *head);
160 };
161 
162 static inline void mr_cache_put(struct mr_mfc *c)
163 {
164 	if (refcount_dec_and_test(&c->mfc_un.res.refcount))
165 		call_rcu(&c->rcu, c->free);
166 }
167 
168 static inline void mr_cache_hold(struct mr_mfc *c)
169 {
170 	refcount_inc(&c->mfc_un.res.refcount);
171 }
172 
173 struct mfc_entry_notifier_info {
174 	struct fib_notifier_info info;
175 	struct mr_mfc *mfc;
176 	u32 tb_id;
177 };
178 
179 static inline int mr_call_mfc_notifier(struct notifier_block *nb,
180 				       unsigned short family,
181 				       enum fib_event_type event_type,
182 				       struct mr_mfc *mfc, u32 tb_id,
183 				       struct netlink_ext_ack *extack)
184 {
185 	struct mfc_entry_notifier_info info = {
186 		.info = {
187 			.family = family,
188 			.extack = extack,
189 		},
190 		.mfc = mfc,
191 		.tb_id = tb_id
192 	};
193 
194 	return call_fib_notifier(nb, event_type, &info.info);
195 }
196 
197 static inline int mr_call_mfc_notifiers(struct net *net,
198 					unsigned short family,
199 					enum fib_event_type event_type,
200 					struct mr_mfc *mfc, u32 tb_id,
201 					unsigned int *ipmr_seq)
202 {
203 	struct mfc_entry_notifier_info info = {
204 		.info = {
205 			.family = family,
206 		},
207 		.mfc = mfc,
208 		.tb_id = tb_id
209 	};
210 
211 	ASSERT_RTNL();
212 	(*ipmr_seq)++;
213 	return call_fib_notifiers(net, event_type, &info.info);
214 }
215 
216 struct mr_table;
217 
218 /**
219  * struct mr_table_ops - callbacks and info for protocol-specific ops
220  * @rht_params: parameters for accessing the MFC hash
221  * @cmparg_any: a hash key to be used for matching on (*,*) routes
222  */
223 struct mr_table_ops {
224 	const struct rhashtable_params *rht_params;
225 	void *cmparg_any;
226 };
227 
228 /**
229  * struct mr_table - a multicast routing table
230  * @list: entry within a list of multicast routing tables
231  * @net: net where this table belongs
232  * @ops: protocol specific operations
233  * @id: identifier of the table
234  * @mroute_sk: socket associated with the table
235  * @ipmr_expire_timer: timer for handling unresolved routes
236  * @mfc_unres_queue: list of unresolved MFC entries
237  * @vif_table: array containing all possible vifs
238  * @mfc_hash: Hash table of all resolved routes for easy lookup
239  * @mfc_cache_list: list of resovled routes for possible traversal
240  * @maxvif: Identifier of highest value vif currently in use
241  * @cache_resolve_queue_len: current size of unresolved queue
242  * @mroute_do_assert: Whether to inform userspace on wrong ingress
243  * @mroute_do_pim: Whether to receive IGMP PIMv1
244  * @mroute_reg_vif_num: PIM-device vif index
245  */
246 struct mr_table {
247 	struct list_head	list;
248 	possible_net_t		net;
249 	struct mr_table_ops	ops;
250 	u32			id;
251 	struct sock __rcu	*mroute_sk;
252 	struct timer_list	ipmr_expire_timer;
253 	struct list_head	mfc_unres_queue;
254 	struct vif_device	vif_table[MAXVIFS];
255 	struct rhltable		mfc_hash;
256 	struct list_head	mfc_cache_list;
257 	int			maxvif;
258 	atomic_t		cache_resolve_queue_len;
259 	bool			mroute_do_assert;
260 	bool			mroute_do_pim;
261 	bool			mroute_do_wrvifwhole;
262 	int			mroute_reg_vif_num;
263 };
264 
265 #ifdef CONFIG_IP_MROUTE_COMMON
266 void vif_device_init(struct vif_device *v,
267 		     struct net_device *dev,
268 		     unsigned long rate_limit,
269 		     unsigned char threshold,
270 		     unsigned short flags,
271 		     unsigned short get_iflink_mask);
272 
273 struct mr_table *
274 mr_table_alloc(struct net *net, u32 id,
275 	       struct mr_table_ops *ops,
276 	       void (*expire_func)(struct timer_list *t),
277 	       void (*table_set)(struct mr_table *mrt,
278 				 struct net *net));
279 
280 /* These actually return 'struct mr_mfc *', but to avoid need for explicit
281  * castings they simply return void.
282  */
283 void *mr_mfc_find_parent(struct mr_table *mrt,
284 			 void *hasharg, int parent);
285 void *mr_mfc_find_any_parent(struct mr_table *mrt, int vifi);
286 void *mr_mfc_find_any(struct mr_table *mrt, int vifi, void *hasharg);
287 
288 int mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
289 		   struct mr_mfc *c, struct rtmsg *rtm);
290 int mr_table_dump(struct mr_table *mrt, struct sk_buff *skb,
291 		  struct netlink_callback *cb,
292 		  int (*fill)(struct mr_table *mrt, struct sk_buff *skb,
293 			      u32 portid, u32 seq, struct mr_mfc *c,
294 			      int cmd, int flags),
295 		  spinlock_t *lock, struct fib_dump_filter *filter);
296 int mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb,
297 		     struct mr_table *(*iter)(struct net *net,
298 					      struct mr_table *mrt),
299 		     int (*fill)(struct mr_table *mrt,
300 				 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 int mr_dump(struct net *net, struct notifier_block *nb, unsigned short family,
306 	    int (*rules_dump)(struct net *net,
307 			      struct notifier_block *nb,
308 			      struct netlink_ext_ack *extack),
309 	    struct mr_table *(*mr_iter)(struct net *net,
310 					struct mr_table *mrt),
311 	    struct netlink_ext_ack *extack);
312 #else
313 static inline void vif_device_init(struct vif_device *v,
314 				   struct net_device *dev,
315 				   unsigned long rate_limit,
316 				   unsigned char threshold,
317 				   unsigned short flags,
318 				   unsigned short get_iflink_mask)
319 {
320 }
321 
322 static inline void *mr_mfc_find_parent(struct mr_table *mrt,
323 				       void *hasharg, int parent)
324 {
325 	return NULL;
326 }
327 
328 static inline void *mr_mfc_find_any_parent(struct mr_table *mrt,
329 					   int vifi)
330 {
331 	return NULL;
332 }
333 
334 static inline struct mr_mfc *mr_mfc_find_any(struct mr_table *mrt,
335 					     int vifi, void *hasharg)
336 {
337 	return NULL;
338 }
339 
340 static inline int mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
341 				 struct mr_mfc *c, struct rtmsg *rtm)
342 {
343 	return -EINVAL;
344 }
345 
346 static inline int
347 mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb,
348 		 struct mr_table *(*iter)(struct net *net,
349 					  struct mr_table *mrt),
350 		 int (*fill)(struct mr_table *mrt,
351 			     struct sk_buff *skb,
352 			     u32 portid, u32 seq, struct mr_mfc *c,
353 			     int cmd, int flags),
354 		 spinlock_t *lock, struct fib_dump_filter *filter)
355 {
356 	return -EINVAL;
357 }
358 
359 static inline int mr_dump(struct net *net, struct notifier_block *nb,
360 			  unsigned short family,
361 			  int (*rules_dump)(struct net *net,
362 					    struct notifier_block *nb,
363 					    struct netlink_ext_ack *extack),
364 			  struct mr_table *(*mr_iter)(struct net *net,
365 						      struct mr_table *mrt),
366 			  struct netlink_ext_ack *extack)
367 {
368 	return -EINVAL;
369 }
370 #endif
371 
372 static inline void *mr_mfc_find(struct mr_table *mrt, void *hasharg)
373 {
374 	return mr_mfc_find_parent(mrt, hasharg, -1);
375 }
376 
377 #ifdef CONFIG_PROC_FS
378 struct mr_vif_iter {
379 	struct seq_net_private p;
380 	struct mr_table *mrt;
381 	int ct;
382 };
383 
384 struct mr_mfc_iter {
385 	struct seq_net_private p;
386 	struct mr_table *mrt;
387 	struct list_head *cache;
388 
389 	/* Lock protecting the mr_table's unresolved queue */
390 	spinlock_t *lock;
391 };
392 
393 #ifdef CONFIG_IP_MROUTE_COMMON
394 void *mr_vif_seq_idx(struct net *net, struct mr_vif_iter *iter, loff_t pos);
395 void *mr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos);
396 
397 static inline void *mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
398 {
399 	return *pos ? mr_vif_seq_idx(seq_file_net(seq),
400 				     seq->private, *pos - 1)
401 		    : SEQ_START_TOKEN;
402 }
403 
404 /* These actually return 'struct mr_mfc *', but to avoid need for explicit
405  * castings they simply return void.
406  */
407 void *mr_mfc_seq_idx(struct net *net,
408 		     struct mr_mfc_iter *it, loff_t pos);
409 void *mr_mfc_seq_next(struct seq_file *seq, void *v,
410 		      loff_t *pos);
411 
412 static inline void *mr_mfc_seq_start(struct seq_file *seq, loff_t *pos,
413 				     struct mr_table *mrt, spinlock_t *lock)
414 {
415 	struct mr_mfc_iter *it = seq->private;
416 
417 	it->mrt = mrt;
418 	it->cache = NULL;
419 	it->lock = lock;
420 
421 	return *pos ? mr_mfc_seq_idx(seq_file_net(seq),
422 				     seq->private, *pos - 1)
423 		    : SEQ_START_TOKEN;
424 }
425 
426 static inline void mr_mfc_seq_stop(struct seq_file *seq, void *v)
427 {
428 	struct mr_mfc_iter *it = seq->private;
429 	struct mr_table *mrt = it->mrt;
430 
431 	if (it->cache == &mrt->mfc_unres_queue)
432 		spin_unlock_bh(it->lock);
433 	else if (it->cache == &mrt->mfc_cache_list)
434 		rcu_read_unlock();
435 }
436 #else
437 static inline void *mr_vif_seq_idx(struct net *net, struct mr_vif_iter *iter,
438 				   loff_t pos)
439 {
440 	return NULL;
441 }
442 
443 static inline void *mr_vif_seq_next(struct seq_file *seq,
444 				    void *v, loff_t *pos)
445 {
446 	return NULL;
447 }
448 
449 static inline void *mr_vif_seq_start(struct seq_file *seq, loff_t *pos)
450 {
451 	return NULL;
452 }
453 
454 static inline void *mr_mfc_seq_idx(struct net *net,
455 				   struct mr_mfc_iter *it, loff_t pos)
456 {
457 	return NULL;
458 }
459 
460 static inline void *mr_mfc_seq_next(struct seq_file *seq, void *v,
461 				    loff_t *pos)
462 {
463 	return NULL;
464 }
465 
466 static inline void *mr_mfc_seq_start(struct seq_file *seq, loff_t *pos,
467 				     struct mr_table *mrt, spinlock_t *lock)
468 {
469 	return NULL;
470 }
471 
472 static inline void mr_mfc_seq_stop(struct seq_file *seq, void *v)
473 {
474 }
475 #endif
476 #endif
477 #endif
478