xref: /freebsd/sys/net/route/route_var.h (revision 96474d2a3fa895fb9636183403fc8ca7ccf60216)
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 	struct nh_control	*nh_control;	/* nexthop subsystem data */
74 	CK_STAILQ_HEAD(, rib_subscription)	rnh_subscribers;/* notification subscribers */
75 };
76 
77 #define	RIB_RLOCK_TRACKER	struct rm_priotracker _rib_tracker
78 #define	RIB_LOCK_INIT(rh)	rm_init(&(rh)->rib_lock, "rib head lock")
79 #define	RIB_LOCK_DESTROY(rh)	rm_destroy(&(rh)->rib_lock)
80 #define	RIB_RLOCK(rh)		rm_rlock(&(rh)->rib_lock, &_rib_tracker)
81 #define	RIB_RUNLOCK(rh)		rm_runlock(&(rh)->rib_lock, &_rib_tracker)
82 #define	RIB_WLOCK(rh)		rm_wlock(&(rh)->rib_lock)
83 #define	RIB_WUNLOCK(rh)		rm_wunlock(&(rh)->rib_lock)
84 #define	RIB_LOCK_ASSERT(rh)	rm_assert(&(rh)->rib_lock, RA_LOCKED)
85 #define	RIB_WLOCK_ASSERT(rh)	rm_assert(&(rh)->rib_lock, RA_WLOCKED)
86 
87 /* Constants */
88 #define	RIB_MAX_RETRIES	3
89 #define	RT_MAXFIBS	UINT16_MAX
90 
91 /* Macro for verifying fields in af-specific 'struct route' structures */
92 #define CHK_STRUCT_FIELD_GENERIC(_s1, _f1, _s2, _f2)			\
93 _Static_assert(sizeof(((_s1 *)0)->_f1) == sizeof(((_s2 *)0)->_f2),	\
94 		"Fields " #_f1 " and " #_f2 " size differs");		\
95 _Static_assert(__offsetof(_s1, _f1) == __offsetof(_s2, _f2),		\
96 		"Fields " #_f1 " and " #_f2 " offset differs");
97 
98 #define _CHK_ROUTE_FIELD(_route_new, _field) \
99 	CHK_STRUCT_FIELD_GENERIC(struct route, _field, _route_new, _field)
100 
101 #define CHK_STRUCT_ROUTE_FIELDS(_route_new)	\
102 	_CHK_ROUTE_FIELD(_route_new, ro_nh)	\
103 	_CHK_ROUTE_FIELD(_route_new, ro_lle)	\
104 	_CHK_ROUTE_FIELD(_route_new, ro_prepend)\
105 	_CHK_ROUTE_FIELD(_route_new, ro_plen)	\
106 	_CHK_ROUTE_FIELD(_route_new, ro_flags)	\
107 	_CHK_ROUTE_FIELD(_route_new, ro_mtu)	\
108 	_CHK_ROUTE_FIELD(_route_new, spare)
109 
110 #define CHK_STRUCT_ROUTE_COMPAT(_ro_new, _dst_new)				\
111 CHK_STRUCT_ROUTE_FIELDS(_ro_new);						\
112 _Static_assert(__offsetof(struct route, ro_dst) == __offsetof(_ro_new, _dst_new),\
113 		"ro_dst and " #_dst_new " are at different offset")
114 
115 struct rib_head *rt_tables_get_rnh(uint32_t table, sa_family_t family);
116 void rt_mpath_init_rnh(struct rib_head *rnh);
117 int rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum);
118 void rt_setmetrics(const struct rt_addrinfo *info, struct rtentry *rt);
119 #ifdef RADIX_MPATH
120 struct radix_node *rt_mpath_unlink(struct rib_head *rnh,
121     struct rt_addrinfo *info, struct rtentry *rto, int *perror);
122 #endif
123 struct rib_cmd_info;
124 
125 VNET_PCPUSTAT_DECLARE(struct rtstat, rtstat);
126 #define	RTSTAT_ADD(name, val)	\
127 	VNET_PCPUSTAT_ADD(struct rtstat, rtstat, name, (val))
128 #define	RTSTAT_INC(name)	RTSTAT_ADD(name, 1)
129 
130 /*
131  * Convert a 'struct radix_node *' to a 'struct rtentry *'.
132  * The operation can be done safely (in this code) because a
133  * 'struct rtentry' starts with two 'struct radix_node''s, the first
134  * one representing leaf nodes in the routing tree, which is
135  * what the code in radix.c passes us as a 'struct radix_node'.
136  *
137  * But because there are a lot of assumptions in this conversion,
138  * do not cast explicitly, but always use the macro below.
139  */
140 #define RNTORT(p)	((struct rtentry *)(p))
141 
142 struct rtentry {
143 	struct	radix_node rt_nodes[2];	/* tree glue, and other values */
144 	/*
145 	 * XXX struct rtentry must begin with a struct radix_node (or two!)
146 	 * because the code does some casts of a 'struct radix_node *'
147 	 * to a 'struct rtentry *'
148 	 */
149 #define	rt_key(r)	(*((struct sockaddr **)(&(r)->rt_nodes->rn_key)))
150 #define	rt_mask(r)	(*((struct sockaddr **)(&(r)->rt_nodes->rn_mask)))
151 #define	rt_key_const(r)		(*((const struct sockaddr * const *)(&(r)->rt_nodes->rn_key)))
152 #define	rt_mask_const(r)	(*((const struct sockaddr * const *)(&(r)->rt_nodes->rn_mask)))
153 
154 	/*
155 	 * 2 radix_node structurs above consists of 2x6 pointers, leaving
156 	 * 4 pointers (32 bytes) of the second cache line on amd64.
157 	 *
158 	 */
159 	struct nhop_object	*rt_nhop;	/* nexthop data */
160 	union {
161 		/*
162 		 * Destination address storage.
163 		 * sizeof(struct sockaddr_in6) == 28, however
164 		 * the dataplane-relevant part (e.g. address) lies
165 		 * at offset 8..24, making the address not crossing
166 		 * cacheline boundary.
167 		 */
168 		struct sockaddr_in	rt_dst4;
169 		struct sockaddr_in6	rt_dst6;
170 		struct sockaddr		rt_dst;
171 		char			rt_dstb[28];
172 	};
173 
174 	int		rte_flags;	/* up/down?, host/net */
175 	u_long		rt_weight;	/* absolute weight */
176 	u_long		rt_expire;	/* lifetime for route, e.g. redirect */
177 	struct rtentry	*rt_chain;	/* pointer to next rtentry to delete */
178 	struct epoch_context	rt_epoch_ctx;	/* net epoch tracker */
179 };
180 
181 /*
182  * With the split between the routing entry and the nexthop,
183  *  rt_flags has to be split between these 2 entries. As rtentry
184  *  mostly contains prefix data and is thought to be generic enough
185  *  so one can transparently change the nexthop pointer w/o requiring
186  *  any other rtentry changes, most of rt_flags shifts to the particular nexthop.
187  * /
188  *
189  * RTF_UP: rtentry, as an indication that it is linked.
190  * RTF_HOST: rtentry, nhop. The latter indication is needed for the datapath
191  * RTF_DYNAMIC: nhop, to make rtentry generic.
192  * RTF_MODIFIED: nhop, to make rtentry generic. (legacy)
193  * -- "native" path (nhop) properties:
194  * RTF_GATEWAY, RTF_STATIC, RTF_PROTO1, RTF_PROTO2, RTF_PROTO3, RTF_FIXEDMTU,
195  *  RTF_PINNED, RTF_REJECT, RTF_BLACKHOLE, RTF_BROADCAST
196  */
197 
198 /* Nexthop rt flags mask */
199 #define	NHOP_RT_FLAG_MASK	(RTF_GATEWAY | RTF_HOST | RTF_REJECT | RTF_DYNAMIC | \
200     RTF_MODIFIED | RTF_STATIC | RTF_BLACKHOLE | RTF_PROTO1 | RTF_PROTO2 | \
201     RTF_PROTO3 | RTF_FIXEDMTU | RTF_PINNED | RTF_BROADCAST)
202 
203 /* rtentry rt flag mask */
204 #define	RTE_RT_FLAG_MASK	(RTF_UP | RTF_HOST)
205 
206 /* Nexthop selection */
207 #define	_NH2MP(_nh)	((struct nhgrp_object *)(_nh))
208 #define	_SELECT_NHOP(_nh, _flowid)	\
209 	(_NH2MP(_nh))->nhops[(_flowid) % (_NH2MP(_nh))->mp_size]
210 #define	_RT_SELECT_NHOP(_nh, _flowid)	\
211 	((!NH_IS_MULTIPATH(_nh)) ? (_nh) : _SELECT_NHOP(_nh, _flowid))
212 #define	RT_SELECT_NHOP(_rt, _flowid)	_RT_SELECT_NHOP((_rt)->rt_nhop, _flowid)
213 
214 /* route_temporal.c */
215 void tmproutes_update(struct rib_head *rnh, struct rtentry *rt);
216 void tmproutes_init(struct rib_head *rh);
217 void tmproutes_destroy(struct rib_head *rh);
218 
219 /* route_ctl.c */
220 struct route_nhop_data {
221 	struct nhop_object	*rnd_nhop;
222 	uint32_t		rnd_weight;
223 };
224 int change_route_conditional(struct rib_head *rnh, struct rtentry *rt,
225     struct rt_addrinfo *info, struct route_nhop_data *nhd_orig,
226     struct route_nhop_data *nhd_new, struct rib_cmd_info *rc);
227 struct rtentry *lookup_prefix(struct rib_head *rnh,
228     const struct rt_addrinfo *info, struct route_nhop_data *rnd);
229 int check_info_match_nhop(const struct rt_addrinfo *info,
230     const struct rtentry *rt, const struct nhop_object *nh);
231 int can_override_nhop(const struct rt_addrinfo *info,
232     const struct nhop_object *nh);
233 
234 void vnet_rtzone_init(void);
235 void vnet_rtzone_destroy(void);
236 
237 /* subscriptions */
238 void rib_init_subscriptions(struct rib_head *rnh);
239 void rib_destroy_subscriptions(struct rib_head *rnh);
240 
241 /* Nexhops */
242 void nhops_init(void);
243 int nhops_init_rib(struct rib_head *rh);
244 void nhops_destroy_rib(struct rib_head *rh);
245 void nhop_ref_object(struct nhop_object *nh);
246 int nhop_try_ref_object(struct nhop_object *nh);
247 int nhop_ref_any(struct nhop_object *nh);
248 void nhop_free_any(struct nhop_object *nh);
249 
250 void nhop_set_type(struct nhop_object *nh, enum nhop_type nh_type);
251 void nhop_set_rtflags(struct nhop_object *nh, int rt_flags);
252 
253 int nhop_create_from_info(struct rib_head *rnh, struct rt_addrinfo *info,
254     struct nhop_object **nh_ret);
255 int nhop_create_from_nhop(struct rib_head *rnh, const struct nhop_object *nh_orig,
256     struct rt_addrinfo *info, struct nhop_object **pnh_priv);
257 
258 void nhops_update_ifmtu(struct rib_head *rh, struct ifnet *ifp, uint32_t mtu);
259 int nhops_dump_sysctl(struct rib_head *rh, struct sysctl_req *w);
260 
261 
262 #endif
263