xref: /openbmc/linux/include/linux/mroute_base.h (revision 068ac0db)
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 				       unsigned short family,
51 				       enum fib_event_type event_type,
52 				       struct vif_device *vif,
53 				       unsigned short vif_index, u32 tb_id,
54 				       struct netlink_ext_ack *extack)
55 {
56 	struct vif_entry_notifier_info info = {
57 		.info = {
58 			.family = family,
59 			.extack = extack,
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, 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 		},
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 				       unsigned short family,
176 				       enum fib_event_type event_type,
177 				       struct mr_mfc *mfc, u32 tb_id,
178 				       struct netlink_ext_ack *extack)
179 {
180 	struct mfc_entry_notifier_info info = {
181 		.info = {
182 			.family = family,
183 			.extack = extack,
184 		},
185 		.mfc = mfc,
186 		.tb_id = tb_id
187 	};
188 
189 	return call_fib_notifier(nb, 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 		},
202 		.mfc = mfc,
203 		.tb_id = tb_id
204 	};
205 
206 	ASSERT_RTNL();
207 	(*ipmr_seq)++;
208 	return call_fib_notifiers(net, event_type, &info.info);
209 }
210 
211 struct mr_table;
212 
213 /**
214  * struct mr_table_ops - callbacks and info for protocol-specific ops
215  * @rht_params: parameters for accessing the MFC hash
216  * @cmparg_any: a hash key to be used for matching on (*,*) routes
217  */
218 struct mr_table_ops {
219 	const struct rhashtable_params *rht_params;
220 	void *cmparg_any;
221 };
222 
223 /**
224  * struct mr_table - a multicast routing table
225  * @list: entry within a list of multicast routing tables
226  * @net: net where this table belongs
227  * @ops: protocol specific operations
228  * @id: identifier of the table
229  * @mroute_sk: socket associated with the table
230  * @ipmr_expire_timer: timer for handling unresolved routes
231  * @mfc_unres_queue: list of unresolved MFC entries
232  * @vif_table: array containing all possible vifs
233  * @mfc_hash: Hash table of all resolved routes for easy lookup
234  * @mfc_cache_list: list of resovled routes for possible traversal
235  * @maxvif: Identifier of highest value vif currently in use
236  * @cache_resolve_queue_len: current size of unresolved queue
237  * @mroute_do_assert: Whether to inform userspace on wrong ingress
238  * @mroute_do_pim: Whether to receive IGMP PIMv1
239  * @mroute_reg_vif_num: PIM-device vif index
240  */
241 struct mr_table {
242 	struct list_head	list;
243 	possible_net_t		net;
244 	struct mr_table_ops	ops;
245 	u32			id;
246 	struct sock __rcu	*mroute_sk;
247 	struct timer_list	ipmr_expire_timer;
248 	struct list_head	mfc_unres_queue;
249 	struct vif_device	vif_table[MAXVIFS];
250 	struct rhltable		mfc_hash;
251 	struct list_head	mfc_cache_list;
252 	int			maxvif;
253 	atomic_t		cache_resolve_queue_len;
254 	bool			mroute_do_assert;
255 	bool			mroute_do_pim;
256 	bool			mroute_do_wrvifwhole;
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_table_dump(struct mr_table *mrt, struct sk_buff *skb,
286 		  struct netlink_callback *cb,
287 		  int (*fill)(struct mr_table *mrt, struct sk_buff *skb,
288 			      u32 portid, u32 seq, struct mr_mfc *c,
289 			      int cmd, int flags),
290 		  spinlock_t *lock, struct fib_dump_filter *filter);
291 int mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb,
292 		     struct mr_table *(*iter)(struct net *net,
293 					      struct mr_table *mrt),
294 		     int (*fill)(struct mr_table *mrt,
295 				 struct sk_buff *skb,
296 				 u32 portid, u32 seq, struct mr_mfc *c,
297 				 int cmd, int flags),
298 		     spinlock_t *lock, struct fib_dump_filter *filter);
299 
300 int mr_dump(struct net *net, struct notifier_block *nb, unsigned short family,
301 	    int (*rules_dump)(struct net *net,
302 			      struct notifier_block *nb,
303 			      struct netlink_ext_ack *extack),
304 	    struct mr_table *(*mr_iter)(struct net *net,
305 					struct mr_table *mrt),
306 	    rwlock_t *mrt_lock, struct netlink_ext_ack *extack);
307 #else
308 static inline void vif_device_init(struct vif_device *v,
309 				   struct net_device *dev,
310 				   unsigned long rate_limit,
311 				   unsigned char threshold,
312 				   unsigned short flags,
313 				   unsigned short get_iflink_mask)
314 {
315 }
316 
317 static inline void *mr_mfc_find_parent(struct mr_table *mrt,
318 				       void *hasharg, int parent)
319 {
320 	return NULL;
321 }
322 
323 static inline void *mr_mfc_find_any_parent(struct mr_table *mrt,
324 					   int vifi)
325 {
326 	return NULL;
327 }
328 
329 static inline struct mr_mfc *mr_mfc_find_any(struct mr_table *mrt,
330 					     int vifi, void *hasharg)
331 {
332 	return NULL;
333 }
334 
335 static inline int mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
336 				 struct mr_mfc *c, struct rtmsg *rtm)
337 {
338 	return -EINVAL;
339 }
340 
341 static inline int
342 mr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb,
343 		 struct mr_table *(*iter)(struct net *net,
344 					  struct mr_table *mrt),
345 		 int (*fill)(struct mr_table *mrt,
346 			     struct sk_buff *skb,
347 			     u32 portid, u32 seq, struct mr_mfc *c,
348 			     int cmd, int flags),
349 		 spinlock_t *lock, struct fib_dump_filter *filter)
350 {
351 	return -EINVAL;
352 }
353 
354 static inline int mr_dump(struct net *net, struct notifier_block *nb,
355 			  unsigned short family,
356 			  int (*rules_dump)(struct net *net,
357 					    struct notifier_block *nb,
358 					    struct netlink_ext_ack *extack),
359 			  struct mr_table *(*mr_iter)(struct net *net,
360 						      struct mr_table *mrt),
361 			  rwlock_t *mrt_lock, struct netlink_ext_ack *extack)
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