xref: /freebsd/sys/net/route/route_var.h (revision ca53e5aedfebcc1b4091b68e01b2d5cae923f85e)
1 /*-
2  * Copyright (c) 2015-2016
3  * 	Alexander V. Chernikov <melifaro@FreeBSD.org>
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  * $FreeBSD$
30  */
31 
32 #ifndef _NET_ROUTE_VAR_H_
33 #define _NET_ROUTE_VAR_H_
34 
35 #ifndef RNF_NORMAL
36 #include <net/radix.h>
37 #endif
38 #include <sys/ck.h>
39 #include <sys/epoch.h>
40 #include <netinet/in.h>		/* struct sockaddr_in */
41 #include <sys/counter.h>
42 #include <net/route/nhop.h>
43 
44 #ifdef	RTDEBUG
45 #define	DPRINTF(_fmt, ...)	printf("%s: " _fmt "\n", __func__ , ## __VA_ARGS__)
46 #else
47 #define	DPRINTF(_fmt, ...)
48 #endif
49 
50 struct nh_control;
51 typedef int rnh_preadd_entry_f_t(u_int fibnum, const struct sockaddr *addr,
52 	const struct sockaddr *mask, struct nhop_object *nh);
53 
54 struct rib_head {
55 	struct radix_head	head;
56 	rn_matchaddr_f_t	*rnh_matchaddr;	/* longest match for sockaddr */
57 	rn_addaddr_f_t		*rnh_addaddr;	/* add based on sockaddr*/
58 	rn_deladdr_f_t		*rnh_deladdr;	/* remove based on sockaddr */
59 	rn_lookup_f_t		*rnh_lookup;	/* exact match for sockaddr */
60 	rn_walktree_t		*rnh_walktree;	/* traverse tree */
61 	rn_walktree_from_t	*rnh_walktree_from; /* traverse tree below a */
62 	rnh_preadd_entry_f_t	*rnh_preadd;	/* hook to alter record prior to insertion */
63 	rt_gen_t		rnh_gen;	/* generation counter */
64 	int			rnh_multipath;	/* multipath capable ? */
65 	struct radix_node	rnh_nodes[3];	/* empty tree for common case */
66 	struct rmlock		rib_lock;	/* config/data path lock */
67 	struct radix_mask_head	rmhead;		/* masks radix head */
68 	struct vnet		*rib_vnet;	/* vnet pointer */
69 	int			rib_family;	/* AF of the rtable */
70 	u_int			rib_fibnum;	/* fib number */
71 	struct callout		expire_callout;	/* Callout for expiring dynamic routes */
72 	time_t			next_expire;	/* Next expire run ts */
73 	uint32_t		rnh_prefixes;	/* Number of prefixes */
74 	struct nh_control	*nh_control;	/* nexthop subsystem data */
75 	CK_STAILQ_HEAD(, rib_subscription)	rnh_subscribers;/* notification subscribers */
76 };
77 
78 #define	RIB_RLOCK_TRACKER	struct rm_priotracker _rib_tracker
79 #define	RIB_LOCK_INIT(rh)	rm_init(&(rh)->rib_lock, "rib head lock")
80 #define	RIB_LOCK_DESTROY(rh)	rm_destroy(&(rh)->rib_lock)
81 #define	RIB_RLOCK(rh)		rm_rlock(&(rh)->rib_lock, &_rib_tracker)
82 #define	RIB_RUNLOCK(rh)		rm_runlock(&(rh)->rib_lock, &_rib_tracker)
83 #define	RIB_WLOCK(rh)		rm_wlock(&(rh)->rib_lock)
84 #define	RIB_WUNLOCK(rh)		rm_wunlock(&(rh)->rib_lock)
85 #define	RIB_LOCK_ASSERT(rh)	rm_assert(&(rh)->rib_lock, RA_LOCKED)
86 #define	RIB_WLOCK_ASSERT(rh)	rm_assert(&(rh)->rib_lock, RA_WLOCKED)
87 
88 /* Constants */
89 #define	RIB_MAX_RETRIES	3
90 #define	RT_MAXFIBS	UINT16_MAX
91 #define	RIB_MAX_MPATH_WIDTH	64
92 
93 /* Macro for verifying fields in af-specific 'struct route' structures */
94 #define CHK_STRUCT_FIELD_GENERIC(_s1, _f1, _s2, _f2)			\
95 _Static_assert(sizeof(((_s1 *)0)->_f1) == sizeof(((_s2 *)0)->_f2),	\
96 		"Fields " #_f1 " and " #_f2 " size differs");		\
97 _Static_assert(__offsetof(_s1, _f1) == __offsetof(_s2, _f2),		\
98 		"Fields " #_f1 " and " #_f2 " offset differs");
99 
100 #define _CHK_ROUTE_FIELD(_route_new, _field) \
101 	CHK_STRUCT_FIELD_GENERIC(struct route, _field, _route_new, _field)
102 
103 #define CHK_STRUCT_ROUTE_FIELDS(_route_new)	\
104 	_CHK_ROUTE_FIELD(_route_new, ro_nh)	\
105 	_CHK_ROUTE_FIELD(_route_new, ro_lle)	\
106 	_CHK_ROUTE_FIELD(_route_new, ro_prepend)\
107 	_CHK_ROUTE_FIELD(_route_new, ro_plen)	\
108 	_CHK_ROUTE_FIELD(_route_new, ro_flags)	\
109 	_CHK_ROUTE_FIELD(_route_new, ro_mtu)	\
110 	_CHK_ROUTE_FIELD(_route_new, spare)
111 
112 #define CHK_STRUCT_ROUTE_COMPAT(_ro_new, _dst_new)				\
113 CHK_STRUCT_ROUTE_FIELDS(_ro_new);						\
114 _Static_assert(__offsetof(struct route, ro_dst) == __offsetof(_ro_new, _dst_new),\
115 		"ro_dst and " #_dst_new " are at different offset")
116 
117 struct rib_head *rt_tables_get_rnh(uint32_t table, sa_family_t family);
118 int rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum);
119 struct rib_cmd_info;
120 
121 VNET_PCPUSTAT_DECLARE(struct rtstat, rtstat);
122 #define	RTSTAT_ADD(name, val)	\
123 	VNET_PCPUSTAT_ADD(struct rtstat, rtstat, name, (val))
124 #define	RTSTAT_INC(name)	RTSTAT_ADD(name, 1)
125 
126 /*
127  * Convert a 'struct radix_node *' to a 'struct rtentry *'.
128  * The operation can be done safely (in this code) because a
129  * 'struct rtentry' starts with two 'struct radix_node''s, the first
130  * one representing leaf nodes in the routing tree, which is
131  * what the code in radix.c passes us as a 'struct radix_node'.
132  *
133  * But because there are a lot of assumptions in this conversion,
134  * do not cast explicitly, but always use the macro below.
135  */
136 #define RNTORT(p)	((struct rtentry *)(p))
137 
138 struct rtentry {
139 	struct	radix_node rt_nodes[2];	/* tree glue, and other values */
140 	/*
141 	 * XXX struct rtentry must begin with a struct radix_node (or two!)
142 	 * because the code does some casts of a 'struct radix_node *'
143 	 * to a 'struct rtentry *'
144 	 */
145 #define	rt_key(r)	(*((struct sockaddr **)(&(r)->rt_nodes->rn_key)))
146 #define	rt_mask(r)	(*((struct sockaddr **)(&(r)->rt_nodes->rn_mask)))
147 #define	rt_key_const(r)		(*((const struct sockaddr * const *)(&(r)->rt_nodes->rn_key)))
148 #define	rt_mask_const(r)	(*((const struct sockaddr * const *)(&(r)->rt_nodes->rn_mask)))
149 
150 	/*
151 	 * 2 radix_node structurs above consists of 2x6 pointers, leaving
152 	 * 4 pointers (32 bytes) of the second cache line on amd64.
153 	 *
154 	 */
155 	struct nhop_object	*rt_nhop;	/* nexthop data */
156 	union {
157 		/*
158 		 * Destination address storage.
159 		 * sizeof(struct sockaddr_in6) == 28, however
160 		 * the dataplane-relevant part (e.g. address) lies
161 		 * at offset 8..24, making the address not crossing
162 		 * cacheline boundary.
163 		 */
164 		struct sockaddr_in	rt_dst4;
165 		struct sockaddr_in6	rt_dst6;
166 		struct sockaddr		rt_dst;
167 		char			rt_dstb[28];
168 	};
169 
170 	int		rte_flags;	/* up/down?, host/net */
171 	u_long		rt_weight;	/* absolute weight */
172 	u_long		rt_expire;	/* lifetime for route, e.g. redirect */
173 	struct rtentry	*rt_chain;	/* pointer to next rtentry to delete */
174 	struct epoch_context	rt_epoch_ctx;	/* net epoch tracker */
175 };
176 
177 /*
178  * With the split between the routing entry and the nexthop,
179  *  rt_flags has to be split between these 2 entries. As rtentry
180  *  mostly contains prefix data and is thought to be generic enough
181  *  so one can transparently change the nexthop pointer w/o requiring
182  *  any other rtentry changes, most of rt_flags shifts to the particular nexthop.
183  * /
184  *
185  * RTF_UP: rtentry, as an indication that it is linked.
186  * RTF_HOST: rtentry, nhop. The latter indication is needed for the datapath
187  * RTF_DYNAMIC: nhop, to make rtentry generic.
188  * RTF_MODIFIED: nhop, to make rtentry generic. (legacy)
189  * -- "native" path (nhop) properties:
190  * RTF_GATEWAY, RTF_STATIC, RTF_PROTO1, RTF_PROTO2, RTF_PROTO3, RTF_FIXEDMTU,
191  *  RTF_PINNED, RTF_REJECT, RTF_BLACKHOLE, RTF_BROADCAST
192  */
193 
194 /* Nexthop rt flags mask */
195 #define	NHOP_RT_FLAG_MASK	(RTF_GATEWAY | RTF_HOST | RTF_REJECT | RTF_DYNAMIC | \
196     RTF_MODIFIED | RTF_STATIC | RTF_BLACKHOLE | RTF_PROTO1 | RTF_PROTO2 | \
197     RTF_PROTO3 | RTF_FIXEDMTU | RTF_PINNED | RTF_BROADCAST)
198 
199 /* rtentry rt flag mask */
200 #define	RTE_RT_FLAG_MASK	(RTF_UP | RTF_HOST)
201 
202 /* route_temporal.c */
203 void tmproutes_update(struct rib_head *rnh, struct rtentry *rt);
204 void tmproutes_init(struct rib_head *rh);
205 void tmproutes_destroy(struct rib_head *rh);
206 
207 /* route_ctl.c */
208 struct route_nhop_data;
209 int change_route_nhop(struct rib_head *rnh, struct rtentry *rt,
210     struct rt_addrinfo *info, struct route_nhop_data *rnd,
211     struct rib_cmd_info *rc);
212 int change_route_conditional(struct rib_head *rnh, struct rtentry *rt,
213     struct rt_addrinfo *info, struct route_nhop_data *nhd_orig,
214     struct route_nhop_data *nhd_new, struct rib_cmd_info *rc);
215 struct rtentry *lookup_prefix(struct rib_head *rnh,
216     const struct rt_addrinfo *info, struct route_nhop_data *rnd);
217 
218 bool nhop_can_multipath(const struct nhop_object *nh);
219 bool match_nhop_gw(const struct nhop_object *nh, const struct sockaddr *gw);
220 int check_info_match_nhop(const struct rt_addrinfo *info,
221     const struct rtentry *rt, const struct nhop_object *nh);
222 int can_override_nhop(const struct rt_addrinfo *info,
223     const struct nhop_object *nh);
224 
225 void vnet_rtzone_init(void);
226 void vnet_rtzone_destroy(void);
227 
228 /* subscriptions */
229 void rib_init_subscriptions(struct rib_head *rnh);
230 void rib_destroy_subscriptions(struct rib_head *rnh);
231 
232 /* Nexhops */
233 void nhops_init(void);
234 int nhops_init_rib(struct rib_head *rh);
235 void nhops_destroy_rib(struct rib_head *rh);
236 void nhop_ref_object(struct nhop_object *nh);
237 int nhop_try_ref_object(struct nhop_object *nh);
238 void nhop_ref_any(struct nhop_object *nh);
239 void nhop_free_any(struct nhop_object *nh);
240 
241 void nhop_set_type(struct nhop_object *nh, enum nhop_type nh_type);
242 void nhop_set_rtflags(struct nhop_object *nh, int rt_flags);
243 
244 int nhop_create_from_info(struct rib_head *rnh, struct rt_addrinfo *info,
245     struct nhop_object **nh_ret);
246 int nhop_create_from_nhop(struct rib_head *rnh, const struct nhop_object *nh_orig,
247     struct rt_addrinfo *info, struct nhop_object **pnh_priv);
248 
249 void nhops_update_ifmtu(struct rib_head *rh, struct ifnet *ifp, uint32_t mtu);
250 int nhops_dump_sysctl(struct rib_head *rh, struct sysctl_req *w);
251 
252 /* MULTIPATH */
253 #define	MPF_MULTIPATH	0x08	/* need to be consistent with NHF_MULTIPATH */
254 
255 struct nhgrp_object {
256 	uint16_t		nhg_flags;	/* nexthop group flags */
257 	uint8_t			nhg_size;	/* dataplain group size */
258 	uint8_t			spare;
259 	struct nhop_object	*nhops[0];	/* nhops */
260 };
261 
262 static inline struct nhop_object *
263 nhop_select(struct nhop_object *nh, uint32_t flowid)
264 {
265 
266 #ifdef ROUTE_MPATH
267 	if (NH_IS_NHGRP(nh)) {
268 		struct nhgrp_object *nhg = (struct nhgrp_object *)nh;
269 		nh = nhg->nhops[flowid % nhg->nhg_size];
270 	}
271 #endif
272 	return (nh);
273 }
274 
275 
276 struct weightened_nhop;
277 
278 /* mpath_ctl.c */
279 int add_route_mpath(struct rib_head *rnh, struct rt_addrinfo *info,
280     struct rtentry *rt, struct route_nhop_data *rnd_add,
281     struct route_nhop_data *rnd_orig, struct rib_cmd_info *rc);
282 int del_route_mpath(struct rib_head *rh, struct rt_addrinfo *info,
283     struct rtentry *rt, struct nhgrp_object *nhg, struct rib_cmd_info *rc);
284 
285 /* nhgrp.c */
286 int nhgrp_ctl_init(struct nh_control *ctl);
287 void nhgrp_ctl_free(struct nh_control *ctl);
288 void nhgrp_ctl_unlink_all(struct nh_control *ctl);
289 
290 
291 /* nhgrp_ctl.c */
292 int nhgrp_dump_sysctl(struct rib_head *rh, struct sysctl_req *w);
293 
294 int nhgrp_get_group(struct rib_head *rh, struct weightened_nhop *wn,
295     int num_nhops, struct route_nhop_data *rnd);
296 typedef bool nhgrp_filter_cb_t(const struct nhop_object *nh, void *data);
297 int nhgrp_get_filtered_group(struct rib_head *rh, const struct nhgrp_object *src,
298     nhgrp_filter_cb_t flt_func, void *flt_data, struct route_nhop_data *rnd);
299 int nhgrp_get_addition_group(struct rib_head *rnh,
300     struct route_nhop_data *rnd_orig, struct route_nhop_data *rnd_add,
301     struct route_nhop_data *rnd_new);
302 
303 void nhgrp_ref_object(struct nhgrp_object *nhg);
304 uint32_t nhgrp_get_idx(const struct nhgrp_object *nhg);
305 void nhgrp_free(struct nhgrp_object *nhg);
306 
307 /* Entropy data used for outbound hashing */
308 #define MPATH_ENTROPY_KEY_LEN	40
309 extern uint8_t mpath_entropy_key[MPATH_ENTROPY_KEY_LEN];
310 
311 #endif
312