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 struct nh_control; 45 /* Sets prefix-specific nexthop flags (NHF_DEFAULT, RTF/NHF_HOST, RTF_BROADCAST,..) */ 46 typedef int rnh_set_nh_pfxflags_f_t(u_int fibnum, const struct sockaddr *addr, 47 const struct sockaddr *mask, struct nhop_object *nh); 48 /* Fills in family-specific details that are not yet set up (mtu, nhop type, ..) */ 49 typedef int rnh_augment_nh_f_t(u_int fibnum, struct nhop_object *nh); 50 51 struct rib_head { 52 struct radix_head head; 53 rn_matchaddr_f_t *rnh_matchaddr; /* longest match for sockaddr */ 54 rn_addaddr_f_t *rnh_addaddr; /* add based on sockaddr*/ 55 rn_deladdr_f_t *rnh_deladdr; /* remove based on sockaddr */ 56 rn_lookup_f_t *rnh_lookup; /* exact match for sockaddr */ 57 rn_walktree_t *rnh_walktree; /* traverse tree */ 58 rn_walktree_from_t *rnh_walktree_from; /* traverse tree below a */ 59 rnh_set_nh_pfxflags_f_t *rnh_set_nh_pfxflags; /* hook to alter record prior to insertion */ 60 rt_gen_t rnh_gen; /* datapath generation counter */ 61 int rnh_multipath; /* multipath capable ? */ 62 struct radix_node rnh_nodes[3]; /* empty tree for common case */ 63 struct rmlock rib_lock; /* config/data path lock */ 64 struct radix_mask_head rmhead; /* masks radix head */ 65 struct vnet *rib_vnet; /* vnet pointer */ 66 int rib_family; /* AF of the rtable */ 67 u_int rib_fibnum; /* fib number */ 68 struct callout expire_callout; /* Callout for expiring dynamic routes */ 69 time_t next_expire; /* Next expire run ts */ 70 uint32_t rnh_prefixes; /* Number of prefixes */ 71 rt_gen_t rnh_gen_rib; /* fib algo: rib generation counter */ 72 uint32_t rib_dying:1; /* rib is detaching */ 73 uint32_t rib_algo_fixed:1;/* fixed algorithm */ 74 uint32_t rib_algo_init:1;/* algo init done */ 75 struct nh_control *nh_control; /* nexthop subsystem data */ 76 rnh_augment_nh_f_t *rnh_augment_nh;/* hook to alter nexthop prior to insertion */ 77 CK_STAILQ_HEAD(, rib_subscription) rnh_subscribers;/* notification subscribers */ 78 }; 79 80 #define RIB_RLOCK_TRACKER struct rm_priotracker _rib_tracker 81 #define RIB_LOCK_INIT(rh) rm_init_flags(&(rh)->rib_lock, "rib head lock", RM_DUPOK) 82 #define RIB_LOCK_DESTROY(rh) rm_destroy(&(rh)->rib_lock) 83 #define RIB_RLOCK(rh) rm_rlock(&(rh)->rib_lock, &_rib_tracker) 84 #define RIB_RUNLOCK(rh) rm_runlock(&(rh)->rib_lock, &_rib_tracker) 85 #define RIB_WLOCK(rh) rm_wlock(&(rh)->rib_lock) 86 #define RIB_WUNLOCK(rh) rm_wunlock(&(rh)->rib_lock) 87 #define RIB_LOCK_ASSERT(rh) rm_assert(&(rh)->rib_lock, RA_LOCKED) 88 #define RIB_WLOCK_ASSERT(rh) rm_assert(&(rh)->rib_lock, RA_WLOCKED) 89 90 /* Constants */ 91 #define RIB_MAX_RETRIES 3 92 #define RT_MAXFIBS UINT16_MAX 93 #define RIB_MAX_MPATH_WIDTH 64 94 95 /* Macro for verifying fields in af-specific 'struct route' structures */ 96 #define CHK_STRUCT_FIELD_GENERIC(_s1, _f1, _s2, _f2) \ 97 _Static_assert(sizeof(((_s1 *)0)->_f1) == sizeof(((_s2 *)0)->_f2), \ 98 "Fields " #_f1 " and " #_f2 " size differs"); \ 99 _Static_assert(__offsetof(_s1, _f1) == __offsetof(_s2, _f2), \ 100 "Fields " #_f1 " and " #_f2 " offset differs"); 101 102 #define _CHK_ROUTE_FIELD(_route_new, _field) \ 103 CHK_STRUCT_FIELD_GENERIC(struct route, _field, _route_new, _field) 104 105 #define CHK_STRUCT_ROUTE_FIELDS(_route_new) \ 106 _CHK_ROUTE_FIELD(_route_new, ro_nh) \ 107 _CHK_ROUTE_FIELD(_route_new, ro_lle) \ 108 _CHK_ROUTE_FIELD(_route_new, ro_prepend)\ 109 _CHK_ROUTE_FIELD(_route_new, ro_plen) \ 110 _CHK_ROUTE_FIELD(_route_new, ro_flags) \ 111 _CHK_ROUTE_FIELD(_route_new, ro_mtu) \ 112 _CHK_ROUTE_FIELD(_route_new, spare) 113 114 #define CHK_STRUCT_ROUTE_COMPAT(_ro_new, _dst_new) \ 115 CHK_STRUCT_ROUTE_FIELDS(_ro_new); \ 116 _Static_assert(__offsetof(struct route, ro_dst) == __offsetof(_ro_new, _dst_new),\ 117 "ro_dst and " #_dst_new " are at different offset") 118 119 static inline void 120 rib_bump_gen(struct rib_head *rnh) 121 { 122 #ifdef FIB_ALGO 123 rnh->rnh_gen_rib++; 124 #else 125 rnh->rnh_gen++; 126 #endif 127 } 128 129 struct rib_head *rt_tables_get_rnh(uint32_t table, sa_family_t family); 130 int rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum); 131 struct rib_cmd_info; 132 133 VNET_PCPUSTAT_DECLARE(struct rtstat, rtstat); 134 #define RTSTAT_ADD(name, val) \ 135 VNET_PCPUSTAT_ADD(struct rtstat, rtstat, name, (val)) 136 #define RTSTAT_INC(name) RTSTAT_ADD(name, 1) 137 138 /* 139 * Convert a 'struct radix_node *' to a 'struct rtentry *'. 140 * The operation can be done safely (in this code) because a 141 * 'struct rtentry' starts with two 'struct radix_node''s, the first 142 * one representing leaf nodes in the routing tree, which is 143 * what the code in radix.c passes us as a 'struct radix_node'. 144 * 145 * But because there are a lot of assumptions in this conversion, 146 * do not cast explicitly, but always use the macro below. 147 */ 148 #define RNTORT(p) ((struct rtentry *)(p)) 149 150 struct rtentry { 151 struct radix_node rt_nodes[2]; /* tree glue, and other values */ 152 /* 153 * XXX struct rtentry must begin with a struct radix_node (or two!) 154 * because the code does some casts of a 'struct radix_node *' 155 * to a 'struct rtentry *' 156 */ 157 #define rt_key(r) (*((struct sockaddr **)(&(r)->rt_nodes->rn_key))) 158 #define rt_mask(r) (*((struct sockaddr **)(&(r)->rt_nodes->rn_mask))) 159 #define rt_key_const(r) (*((const struct sockaddr * const *)(&(r)->rt_nodes->rn_key))) 160 #define rt_mask_const(r) (*((const struct sockaddr * const *)(&(r)->rt_nodes->rn_mask))) 161 162 /* 163 * 2 radix_node structurs above consists of 2x6 pointers, leaving 164 * 4 pointers (32 bytes) of the second cache line on amd64. 165 * 166 */ 167 struct nhop_object *rt_nhop; /* nexthop data */ 168 union { 169 /* 170 * Destination address storage. 171 * sizeof(struct sockaddr_in6) == 28, however 172 * the dataplane-relevant part (e.g. address) lies 173 * at offset 8..24, making the address not crossing 174 * cacheline boundary. 175 */ 176 struct sockaddr_in rt_dst4; 177 struct sockaddr_in6 rt_dst6; 178 struct sockaddr rt_dst; 179 char rt_dstb[28]; 180 }; 181 182 int rte_flags; /* up/down?, host/net */ 183 u_long rt_weight; /* absolute weight */ 184 struct rtentry *rt_chain; /* pointer to next rtentry to delete */ 185 struct epoch_context rt_epoch_ctx; /* net epoch tracker */ 186 }; 187 188 /* 189 * With the split between the routing entry and the nexthop, 190 * rt_flags has to be split between these 2 entries. As rtentry 191 * mostly contains prefix data and is thought to be generic enough 192 * so one can transparently change the nexthop pointer w/o requiring 193 * any other rtentry changes, most of rt_flags shifts to the particular nexthop. 194 * / 195 * 196 * RTF_UP: rtentry, as an indication that it is linked. 197 * RTF_HOST: rtentry, nhop. The latter indication is needed for the datapath 198 * RTF_DYNAMIC: nhop, to make rtentry generic. 199 * RTF_MODIFIED: nhop, to make rtentry generic. (legacy) 200 * -- "native" path (nhop) properties: 201 * RTF_GATEWAY, RTF_STATIC, RTF_PROTO1, RTF_PROTO2, RTF_PROTO3, RTF_FIXEDMTU, 202 * RTF_PINNED, RTF_REJECT, RTF_BLACKHOLE, RTF_BROADCAST 203 */ 204 205 /* rtentry rt flag mask */ 206 #define RTE_RT_FLAG_MASK (RTF_UP | RTF_HOST) 207 208 /* route_temporal.c */ 209 void tmproutes_update(struct rib_head *rnh, struct rtentry *rt, struct nhop_object *nh); 210 void tmproutes_init(struct rib_head *rh); 211 void tmproutes_destroy(struct rib_head *rh); 212 213 /* route_ctl.c */ 214 struct route_nhop_data; 215 int change_route(struct rib_head *rnh, struct rtentry *rt, 216 struct route_nhop_data *rnd, struct rib_cmd_info *rc); 217 int change_route_conditional(struct rib_head *rnh, struct rtentry *rt, 218 struct route_nhop_data *nhd_orig, struct route_nhop_data *nhd_new, 219 struct rib_cmd_info *rc); 220 struct rtentry *lookup_prefix(struct rib_head *rnh, 221 const struct rt_addrinfo *info, struct route_nhop_data *rnd); 222 struct rtentry *lookup_prefix_rt(struct rib_head *rnh, const struct rtentry *rt, 223 struct route_nhop_data *rnd); 224 int rib_copy_route(struct rtentry *rt, const struct route_nhop_data *rnd_src, 225 struct rib_head *rh_dst, struct rib_cmd_info *rc); 226 227 bool nhop_can_multipath(const struct nhop_object *nh); 228 bool match_nhop_gw(const struct nhop_object *nh, const struct sockaddr *gw); 229 int check_info_match_nhop(const struct rt_addrinfo *info, 230 const struct rtentry *rt, const struct nhop_object *nh); 231 bool rib_can_4o6_nhop(void); 232 233 /* route_rtentry.c */ 234 void vnet_rtzone_init(void); 235 void vnet_rtzone_destroy(void); 236 void rt_free(struct rtentry *rt); 237 void rt_free_immediate(struct rtentry *rt); 238 struct rtentry *rt_alloc(struct rib_head *rnh, const struct sockaddr *dst, 239 struct sockaddr *netmask); 240 241 /* subscriptions */ 242 void rib_init_subscriptions(struct rib_head *rnh); 243 void rib_destroy_subscriptions(struct rib_head *rnh); 244 245 /* route_ifaddrs.c */ 246 void rib_copy_kernel_routes(struct rib_head *rh_src, struct rib_head *rh_dst); 247 248 /* Nexhops */ 249 void nhops_init(void); 250 int nhops_init_rib(struct rib_head *rh); 251 void nhops_destroy_rib(struct rib_head *rh); 252 void nhop_ref_object(struct nhop_object *nh); 253 int nhop_try_ref_object(struct nhop_object *nh); 254 void nhop_ref_any(struct nhop_object *nh); 255 void nhop_free_any(struct nhop_object *nh); 256 struct nhop_object *nhop_get_nhop_internal(struct rib_head *rnh, 257 struct nhop_object *nh, int *perror); 258 259 bool nhop_check_gateway(int upper_family, int neigh_family); 260 261 int nhop_create_from_info(struct rib_head *rnh, struct rt_addrinfo *info, 262 struct nhop_object **nh_ret); 263 int nhop_create_from_nhop(struct rib_head *rnh, const struct nhop_object *nh_orig, 264 struct rt_addrinfo *info, struct nhop_object **pnh_priv); 265 266 void nhops_update_ifmtu(struct rib_head *rh, struct ifnet *ifp, uint32_t mtu); 267 int nhops_dump_sysctl(struct rib_head *rh, struct sysctl_req *w); 268 269 /* MULTIPATH */ 270 #define MPF_MULTIPATH 0x08 /* need to be consistent with NHF_MULTIPATH */ 271 272 struct nhgrp_object { 273 uint16_t nhg_flags; /* nexthop group flags */ 274 uint8_t nhg_size; /* dataplain group size */ 275 uint8_t spare; 276 struct nhop_object *nhops[0]; /* nhops */ 277 }; 278 279 static inline struct nhop_object * 280 nhop_select(struct nhop_object *nh, uint32_t flowid) 281 { 282 283 #ifdef ROUTE_MPATH 284 if (NH_IS_NHGRP(nh)) { 285 struct nhgrp_object *nhg = (struct nhgrp_object *)nh; 286 nh = nhg->nhops[flowid % nhg->nhg_size]; 287 } 288 #endif 289 return (nh); 290 } 291 292 293 struct weightened_nhop; 294 295 /* mpath_ctl.c */ 296 int add_route_mpath(struct rib_head *rnh, struct rt_addrinfo *info, 297 struct rtentry *rt, struct route_nhop_data *rnd_add, 298 struct route_nhop_data *rnd_orig, struct rib_cmd_info *rc); 299 300 /* nhgrp.c */ 301 int nhgrp_ctl_init(struct nh_control *ctl); 302 void nhgrp_ctl_free(struct nh_control *ctl); 303 void nhgrp_ctl_unlink_all(struct nh_control *ctl); 304 305 306 /* nhgrp_ctl.c */ 307 int nhgrp_dump_sysctl(struct rib_head *rh, struct sysctl_req *w); 308 309 int nhgrp_get_filtered_group(struct rib_head *rh, const struct rtentry *rt, 310 const struct nhgrp_object *src, rib_filter_f_t flt_func, void *flt_data, 311 struct route_nhop_data *rnd); 312 int nhgrp_get_addition_group(struct rib_head *rnh, 313 struct route_nhop_data *rnd_orig, struct route_nhop_data *rnd_add, 314 struct route_nhop_data *rnd_new); 315 316 void nhgrp_ref_object(struct nhgrp_object *nhg); 317 uint32_t nhgrp_get_idx(const struct nhgrp_object *nhg); 318 void nhgrp_free(struct nhgrp_object *nhg); 319 320 /* rtsock */ 321 int rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh, 322 int fibnum); 323 int rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum); 324 int rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum); 325 326 327 /* lookup_framework.c */ 328 void fib_grow_rtables(uint32_t new_num_tables); 329 void fib_setup_family(int family, uint32_t num_tables); 330 void fib_destroy_rib(struct rib_head *rh); 331 void vnet_fib_init(void); 332 void vnet_fib_destroy(void); 333 334 /* Entropy data used for outbound hashing */ 335 #define MPATH_ENTROPY_KEY_LEN 40 336 extern uint8_t mpath_entropy_key[MPATH_ENTROPY_KEY_LEN]; 337 338 #endif 339