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 50struct nh_control; 51typedef int rnh_preadd_entry_f_t(u_int fibnum, const struct sockaddr *addr, 52 const struct sockaddr *mask, struct nhop_object *nh); 53 54struct 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; /* datapath 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 rt_gen_t rnh_gen_rib; /* fib algo: rib generation counter */ 75 uint32_t rib_dying:1; /* rib is detaching */ 76 uint32_t rib_algo_fixed:1;/* fixed algorithm */ 77 uint32_t rib_algo_init:1;/* algo init done */ 78 struct nh_control *nh_control; /* nexthop subsystem data */ 79 CK_STAILQ_HEAD(, rib_subscription) rnh_subscribers;/* notification subscribers */ 80}; 81 82#define RIB_RLOCK_TRACKER struct rm_priotracker _rib_tracker 83#define RIB_LOCK_INIT(rh) rm_init(&(rh)->rib_lock, "rib head lock") 84#define RIB_LOCK_DESTROY(rh) rm_destroy(&(rh)->rib_lock) 85#define RIB_RLOCK(rh) rm_rlock(&(rh)->rib_lock, &_rib_tracker) 86#define RIB_RUNLOCK(rh) rm_runlock(&(rh)->rib_lock, &_rib_tracker) 87#define RIB_WLOCK(rh) rm_wlock(&(rh)->rib_lock) 88#define RIB_WUNLOCK(rh) rm_wunlock(&(rh)->rib_lock) 89#define RIB_LOCK_ASSERT(rh) rm_assert(&(rh)->rib_lock, RA_LOCKED) 90#define RIB_WLOCK_ASSERT(rh) rm_assert(&(rh)->rib_lock, RA_WLOCKED) 91 92/* Constants */ 93#define RIB_MAX_RETRIES 3 94#define RT_MAXFIBS UINT16_MAX 95#define RIB_MAX_MPATH_WIDTH 64 96 97/* Macro for verifying fields in af-specific 'struct route' structures */ 98#define CHK_STRUCT_FIELD_GENERIC(_s1, _f1, _s2, _f2) \ 99_Static_assert(sizeof(((_s1 *)0)->_f1) == sizeof(((_s2 *)0)->_f2), \ 100 "Fields " #_f1 " and " #_f2 " size differs"); \ 101_Static_assert(__offsetof(_s1, _f1) == __offsetof(_s2, _f2), \ 102 "Fields " #_f1 " and " #_f2 " offset differs"); 103 104#define _CHK_ROUTE_FIELD(_route_new, _field) \ 105 CHK_STRUCT_FIELD_GENERIC(struct route, _field, _route_new, _field) 106 107#define CHK_STRUCT_ROUTE_FIELDS(_route_new) \ 108 _CHK_ROUTE_FIELD(_route_new, ro_nh) \ 109 _CHK_ROUTE_FIELD(_route_new, ro_lle) \ 110 _CHK_ROUTE_FIELD(_route_new, ro_prepend)\ 111 _CHK_ROUTE_FIELD(_route_new, ro_plen) \ 112 _CHK_ROUTE_FIELD(_route_new, ro_flags) \ 113 _CHK_ROUTE_FIELD(_route_new, ro_mtu) \ 114 _CHK_ROUTE_FIELD(_route_new, spare) 115 116#define CHK_STRUCT_ROUTE_COMPAT(_ro_new, _dst_new) \ 117CHK_STRUCT_ROUTE_FIELDS(_ro_new); \ 118_Static_assert(__offsetof(struct route, ro_dst) == __offsetof(_ro_new, _dst_new),\ 119 "ro_dst and " #_dst_new " are at different offset") 120 121static inline void 122rib_bump_gen(struct rib_head *rnh) 123{ 124#ifdef FIB_ALGO 125 rnh->rnh_gen_rib++; 126#else 127 rnh->rnh_gen++; 128#endif 129} 130 131struct rib_head *rt_tables_get_rnh(uint32_t table, sa_family_t family); 132int rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum); 133struct rib_cmd_info; 134 135VNET_PCPUSTAT_DECLARE(struct rtstat, rtstat); 136#define RTSTAT_ADD(name, val) \ 137 VNET_PCPUSTAT_ADD(struct rtstat, rtstat, name, (val)) 138#define RTSTAT_INC(name) RTSTAT_ADD(name, 1) 139 140/* 141 * Convert a 'struct radix_node *' to a 'struct rtentry *'. 142 * The operation can be done safely (in this code) because a 143 * 'struct rtentry' starts with two 'struct radix_node''s, the first 144 * one representing leaf nodes in the routing tree, which is 145 * what the code in radix.c passes us as a 'struct radix_node'. 146 * 147 * But because there are a lot of assumptions in this conversion, 148 * do not cast explicitly, but always use the macro below. 149 */ 150#define RNTORT(p) ((struct rtentry *)(p)) 151 152struct rtentry { 153 struct radix_node rt_nodes[2]; /* tree glue, and other values */ 154 /* 155 * XXX struct rtentry must begin with a struct radix_node (or two!) 156 * because the code does some casts of a 'struct radix_node *' 157 * to a 'struct rtentry *' 158 */ 159#define rt_key(r) (*((struct sockaddr **)(&(r)->rt_nodes->rn_key))) 160#define rt_mask(r) (*((struct sockaddr **)(&(r)->rt_nodes->rn_mask))) 161#define rt_key_const(r) (*((const struct sockaddr * const *)(&(r)->rt_nodes->rn_key))) 162#define rt_mask_const(r) (*((const struct sockaddr * const *)(&(r)->rt_nodes->rn_mask))) 163 164 /* 165 * 2 radix_node structurs above consists of 2x6 pointers, leaving 166 * 4 pointers (32 bytes) of the second cache line on amd64. 167 * 168 */ 169 struct nhop_object *rt_nhop; /* nexthop data */ 170 union { 171 /* 172 * Destination address storage. 173 * sizeof(struct sockaddr_in6) == 28, however 174 * the dataplane-relevant part (e.g. address) lies 175 * at offset 8..24, making the address not crossing 176 * cacheline boundary. 177 */ 178 struct sockaddr_in rt_dst4; 179 struct sockaddr_in6 rt_dst6; 180 struct sockaddr rt_dst; 181 char rt_dstb[28]; 182 }; 183 184 int rte_flags; /* up/down?, host/net */ 185 u_long rt_weight; /* absolute weight */ 186 u_long rt_expire; /* lifetime for route, e.g. redirect */ 187 struct rtentry *rt_chain; /* pointer to next rtentry to delete */ 188 struct epoch_context rt_epoch_ctx; /* net epoch tracker */ 189}; 190 191/* 192 * With the split between the routing entry and the nexthop, 193 * rt_flags has to be split between these 2 entries. As rtentry 194 * mostly contains prefix data and is thought to be generic enough 195 * so one can transparently change the nexthop pointer w/o requiring 196 * any other rtentry changes, most of rt_flags shifts to the particular nexthop. 197 * / 198 * 199 * RTF_UP: rtentry, as an indication that it is linked. 200 * RTF_HOST: rtentry, nhop. The latter indication is needed for the datapath 201 * RTF_DYNAMIC: nhop, to make rtentry generic. 202 * RTF_MODIFIED: nhop, to make rtentry generic. (legacy) 203 * -- "native" path (nhop) properties: 204 * RTF_GATEWAY, RTF_STATIC, RTF_PROTO1, RTF_PROTO2, RTF_PROTO3, RTF_FIXEDMTU, 205 * RTF_PINNED, RTF_REJECT, RTF_BLACKHOLE, RTF_BROADCAST 206 */ 207 208/* Nexthop rt flags mask */ 209#define NHOP_RT_FLAG_MASK (RTF_GATEWAY | RTF_HOST | RTF_REJECT | RTF_DYNAMIC | \ 210 RTF_MODIFIED | RTF_STATIC | RTF_BLACKHOLE | RTF_PROTO1 | RTF_PROTO2 | \ 211 RTF_PROTO3 | RTF_FIXEDMTU | RTF_PINNED | RTF_BROADCAST) 212 213/* rtentry rt flag mask */ 214#define RTE_RT_FLAG_MASK (RTF_UP | RTF_HOST) 215 216/* route_temporal.c */ 217void tmproutes_update(struct rib_head *rnh, struct rtentry *rt); 218void tmproutes_init(struct rib_head *rh); 219void tmproutes_destroy(struct rib_head *rh); 220 221/* route_ctl.c */ 222struct route_nhop_data; 223int change_route_nhop(struct rib_head *rnh, struct rtentry *rt, 224 struct rt_addrinfo *info, struct route_nhop_data *rnd, 225 struct rib_cmd_info *rc); 226int change_route_conditional(struct rib_head *rnh, struct rtentry *rt, 227 struct rt_addrinfo *info, struct route_nhop_data *nhd_orig, 228 struct route_nhop_data *nhd_new, struct rib_cmd_info *rc); 229struct rtentry *lookup_prefix(struct rib_head *rnh, 230 const struct rt_addrinfo *info, struct route_nhop_data *rnd); 231 232bool nhop_can_multipath(const struct nhop_object *nh); 233bool match_nhop_gw(const struct nhop_object *nh, const struct sockaddr *gw); 234int check_info_match_nhop(const struct rt_addrinfo *info, 235 const struct rtentry *rt, const struct nhop_object *nh); 236int can_override_nhop(const struct rt_addrinfo *info, 237 const struct nhop_object *nh); 238 239void vnet_rtzone_init(void); 240void vnet_rtzone_destroy(void); 241 242/* subscriptions */ 243void rib_init_subscriptions(struct rib_head *rnh); 244void rib_destroy_subscriptions(struct rib_head *rnh); 245 246/* Nexhops */ 247void nhops_init(void); 248int nhops_init_rib(struct rib_head *rh); 249void nhops_destroy_rib(struct rib_head *rh); 250void nhop_ref_object(struct nhop_object *nh); 251int nhop_try_ref_object(struct nhop_object *nh); 252void nhop_ref_any(struct nhop_object *nh); 253void nhop_free_any(struct nhop_object *nh); 254 255void nhop_set_type(struct nhop_object *nh, enum nhop_type nh_type); 256void nhop_set_rtflags(struct nhop_object *nh, int rt_flags); 257 258int nhop_create_from_info(struct rib_head *rnh, struct rt_addrinfo *info, 259 struct nhop_object **nh_ret); 260int nhop_create_from_nhop(struct rib_head *rnh, const struct nhop_object *nh_orig, 261 struct rt_addrinfo *info, struct nhop_object **pnh_priv); 262 263void nhops_update_ifmtu(struct rib_head *rh, struct ifnet *ifp, uint32_t mtu); 264int nhops_dump_sysctl(struct rib_head *rh, struct sysctl_req *w); 265 266/* MULTIPATH */ 267#define MPF_MULTIPATH 0x08 /* need to be consistent with NHF_MULTIPATH */ 268 269struct nhgrp_object { 270 uint16_t nhg_flags; /* nexthop group flags */ 271 uint8_t nhg_size; /* dataplain group size */ 272 uint8_t spare; 273 struct nhop_object *nhops[0]; /* nhops */ 274}; 275 276static inline struct nhop_object * 277nhop_select(struct nhop_object *nh, uint32_t flowid) 278{ 279 280#ifdef ROUTE_MPATH 281 if (NH_IS_NHGRP(nh)) { 282 struct nhgrp_object *nhg = (struct nhgrp_object *)nh; 283 nh = nhg->nhops[flowid % nhg->nhg_size]; 284 } 285#endif 286 return (nh); 287} 288 289 290struct weightened_nhop; 291 292/* mpath_ctl.c */ 293int add_route_mpath(struct rib_head *rnh, struct rt_addrinfo *info, 294 struct rtentry *rt, struct route_nhop_data *rnd_add, 295 struct route_nhop_data *rnd_orig, struct rib_cmd_info *rc); 296int del_route_mpath(struct rib_head *rh, struct rt_addrinfo *info, 297 struct rtentry *rt, struct nhgrp_object *nhg, struct rib_cmd_info *rc); 298 299/* nhgrp.c */ 300int nhgrp_ctl_init(struct nh_control *ctl); 301void nhgrp_ctl_free(struct nh_control *ctl); 302void nhgrp_ctl_unlink_all(struct nh_control *ctl); 303 304 305/* nhgrp_ctl.c */ 306int nhgrp_dump_sysctl(struct rib_head *rh, struct sysctl_req *w); 307 308int nhgrp_get_group(struct rib_head *rh, struct weightened_nhop *wn, 309 int num_nhops, struct route_nhop_data *rnd); 310typedef bool nhgrp_filter_cb_t(const struct nhop_object *nh, void *data); 311int nhgrp_get_filtered_group(struct rib_head *rh, const struct nhgrp_object *src, 312 nhgrp_filter_cb_t flt_func, void *flt_data, struct route_nhop_data *rnd); 313int nhgrp_get_addition_group(struct rib_head *rnh, 314 struct route_nhop_data *rnd_orig, struct route_nhop_data *rnd_add, 315 struct route_nhop_data *rnd_new); 316 317void nhgrp_ref_object(struct nhgrp_object *nhg); 318uint32_t nhgrp_get_idx(const struct nhgrp_object *nhg); 319void nhgrp_free(struct nhgrp_object *nhg); 320 321/* rtsock */ 322int rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh, 323 int fibnum); 324int rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum); 325int rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum); 326 327 328/* lookup_framework.c */ 329void fib_grow_rtables(uint32_t new_num_tables); 330void fib_setup_family(int family, uint32_t num_tables); 331void fib_destroy_rib(struct rib_head *rh); 332void vnet_fib_init(void); 333void vnet_fib_destroy(void); 334 335/* Entropy data used for outbound hashing */ 336#define MPATH_ENTROPY_KEY_LEN 40 337extern uint8_t mpath_entropy_key[MPATH_ENTROPY_KEY_LEN]; 338 339#endif 340