1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 2020 Alexander V. Chernikov 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 25 * SUCH DAMAGE. 26 */ 27 28 #include <sys/cdefs.h> 29 __FBSDID("$FreeBSD$"); 30 #include "opt_inet.h" 31 #include "opt_inet6.h" 32 #include "opt_route.h" 33 34 #include <sys/param.h> 35 #include <sys/jail.h> 36 #include <sys/systm.h> 37 #include <sys/malloc.h> 38 #include <sys/mbuf.h> 39 #include <sys/socket.h> 40 #include <sys/sysctl.h> 41 #include <sys/syslog.h> 42 #include <sys/sysproto.h> 43 #include <sys/proc.h> 44 #include <sys/domain.h> 45 #include <sys/kernel.h> 46 #include <sys/lock.h> 47 #include <sys/rmlock.h> 48 49 #include <net/if.h> 50 #include <net/if_var.h> 51 #include <net/if_dl.h> 52 #include <net/route.h> 53 #include <net/route/route_ctl.h> 54 #include <net/route/route_var.h> 55 #include <net/route/nhop_utils.h> 56 #include <net/route/nhop.h> 57 #include <net/route/nhop_var.h> 58 #ifdef INET 59 #include <netinet/in_fib.h> 60 #endif 61 #ifdef INET6 62 #include <netinet6/in6_fib.h> 63 #endif 64 #include <net/vnet.h> 65 66 /* 67 * RIB helper functions. 68 */ 69 70 /* 71 * Calls @wa_f with @arg for each entry in the table specified by 72 * @af and @fibnum. 73 * 74 * @ss_t callback is called before and after the tree traversal 75 * while holding table lock. 76 * 77 * Table is traversed under read lock unless @wlock is set. 78 */ 79 void 80 rib_walk_ext_internal(struct rib_head *rnh, bool wlock, rib_walktree_f_t *wa_f, 81 rib_walk_hook_f_t *hook_f, void *arg) 82 { 83 RIB_RLOCK_TRACKER; 84 85 if (wlock) 86 RIB_WLOCK(rnh); 87 else 88 RIB_RLOCK(rnh); 89 if (hook_f != NULL) 90 hook_f(rnh, RIB_WALK_HOOK_PRE, arg); 91 rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f, arg); 92 if (hook_f != NULL) 93 hook_f(rnh, RIB_WALK_HOOK_POST, arg); 94 if (wlock) 95 RIB_WUNLOCK(rnh); 96 else 97 RIB_RUNLOCK(rnh); 98 } 99 100 void 101 rib_walk_ext(uint32_t fibnum, int family, bool wlock, rib_walktree_f_t *wa_f, 102 rib_walk_hook_f_t *hook_f, void *arg) 103 { 104 struct rib_head *rnh; 105 106 if ((rnh = rt_tables_get_rnh(fibnum, family)) != NULL) 107 rib_walk_ext_internal(rnh, wlock, wa_f, hook_f, arg); 108 } 109 110 /* 111 * Calls @wa_f with @arg for each entry in the table specified by 112 * @af and @fibnum. 113 * 114 * Table is traversed under read lock unless @wlock is set. 115 */ 116 void 117 rib_walk(uint32_t fibnum, int family, bool wlock, rib_walktree_f_t *wa_f, 118 void *arg) 119 { 120 121 rib_walk_ext(fibnum, family, wlock, wa_f, NULL, arg); 122 } 123 124 /* 125 * Iterates over all existing fibs in system calling 126 * @hook_f function before/after traversing each fib. 127 * Calls @wa_f function for each element in current fib. 128 * If af is not AF_UNSPEC, iterates over fibs in particular 129 * address family. 130 */ 131 void 132 rib_foreach_table_walk(int family, bool wlock, rib_walktree_f_t *wa_f, 133 rib_walk_hook_f_t *hook_f, void *arg) 134 { 135 136 for (uint32_t fibnum = 0; fibnum < rt_numfibs; fibnum++) { 137 /* Do we want some specific family? */ 138 if (family != AF_UNSPEC) { 139 rib_walk_ext(fibnum, family, wlock, wa_f, hook_f, arg); 140 continue; 141 } 142 143 for (int i = 1; i <= AF_MAX; i++) 144 rib_walk_ext(fibnum, i, wlock, wa_f, hook_f, arg); 145 } 146 } 147 148 /* 149 * Iterates over all existing fibs in system and deletes each element 150 * for which @filter_f function returns non-zero value. 151 * If @family is not AF_UNSPEC, iterates over fibs in particular 152 * address family. 153 */ 154 void 155 rib_foreach_table_walk_del(int family, rib_filter_f_t *filter_f, void *arg) 156 { 157 158 for (uint32_t fibnum = 0; fibnum < rt_numfibs; fibnum++) { 159 /* Do we want some specific family? */ 160 if (family != AF_UNSPEC) { 161 rib_walk_del(fibnum, family, filter_f, arg, 0); 162 continue; 163 } 164 165 for (int i = 1; i <= AF_MAX; i++) 166 rib_walk_del(fibnum, i, filter_f, arg, 0); 167 } 168 } 169 170 171 /* 172 * Wrapper for the control plane functions for performing af-agnostic 173 * lookups. 174 * @fibnum: fib to perform the lookup. 175 * @dst: sockaddr with family and addr filled in. IPv6 addresses needs to be in 176 * deembedded from. 177 * @flags: fib(9) flags. 178 * @flowid: flow id for path selection in multipath use case. 179 * 180 * Returns nhop_object or NULL. 181 * 182 * Requires NET_EPOCH. 183 * 184 */ 185 struct nhop_object * 186 rib_lookup(uint32_t fibnum, const struct sockaddr *dst, uint32_t flags, 187 uint32_t flowid) 188 { 189 struct nhop_object *nh; 190 191 nh = NULL; 192 193 switch (dst->sa_family) { 194 #ifdef INET 195 case AF_INET: 196 { 197 const struct sockaddr_in *a = (const struct sockaddr_in *)dst; 198 nh = fib4_lookup(fibnum, a->sin_addr, 0, flags, flowid); 199 break; 200 } 201 #endif 202 #ifdef INET6 203 case AF_INET6: 204 { 205 const struct sockaddr_in6 *a = (const struct sockaddr_in6*)dst; 206 nh = fib6_lookup(fibnum, &a->sin6_addr, a->sin6_scope_id, 207 flags, flowid); 208 break; 209 } 210 #endif 211 } 212 213 return (nh); 214 } 215 216 #ifdef ROUTE_MPATH 217 static void 218 decompose_change_notification(struct rib_cmd_info *rc, route_notification_t *cb, 219 void *cbdata) 220 { 221 uint32_t num_old, num_new; 222 uint32_t nh_idx_old, nh_idx_new; 223 struct weightened_nhop *wn_old, *wn_new; 224 struct weightened_nhop tmp = { NULL, 0 }; 225 uint32_t idx_old = 0, idx_new = 0; 226 227 struct rib_cmd_info rc_del = { .rc_cmd = RTM_DELETE, .rc_rt = rc->rc_rt }; 228 struct rib_cmd_info rc_add = { .rc_cmd = RTM_ADD, .rc_rt = rc->rc_rt }; 229 230 if (NH_IS_NHGRP(rc->rc_nh_old)) { 231 wn_old = nhgrp_get_nhops((struct nhgrp_object *)rc->rc_nh_old, &num_old); 232 } else { 233 tmp.nh = rc->rc_nh_old; 234 tmp.weight = rc->rc_nh_weight; 235 wn_old = &tmp; 236 num_old = 1; 237 } 238 if (NH_IS_NHGRP(rc->rc_nh_new)) { 239 wn_new = nhgrp_get_nhops((struct nhgrp_object *)rc->rc_nh_new, &num_new); 240 } else { 241 tmp.nh = rc->rc_nh_new; 242 tmp.weight = rc->rc_nh_weight; 243 wn_new = &tmp; 244 num_new = 1; 245 } 246 247 /* Use the fact that each @wn array is sorted */ 248 /* 249 * Want to convert into set of add and delete operations 250 * [1] -> [1, 2] = A{2} 251 * [2] -> [1, 2] = A{1} 252 * [1, 2, 4]->[1, 3, 4] = A{2}, D{3} 253 * [1, 2, 4]->[1, 4] = D{2} 254 * [1, 2, 4] -> [3, 4] = D{1}, C{2,3} OR C{1,3}, D{2} OR D{1},D{2},A{3} 255 * [1, 2] -> [3, 4] = 256 * 257 */ 258 idx_old = 0; 259 while ((idx_old < num_old) && (idx_new < num_new)) { 260 nh_idx_old = wn_old[idx_old].nh->nh_priv->nh_idx; 261 nh_idx_new = wn_new[idx_new].nh->nh_priv->nh_idx; 262 263 if (nh_idx_old == nh_idx_new) { 264 if (wn_old[idx_old].weight != wn_new[idx_new].weight) { 265 /* Update weight by providing del/add notifications */ 266 rc_del.rc_nh_old = wn_old[idx_old].nh; 267 rc_del.rc_nh_weight = wn_old[idx_old].weight; 268 cb(&rc_del, cbdata); 269 270 rc_add.rc_nh_new = wn_new[idx_new].nh; 271 rc_add.rc_nh_weight = wn_new[idx_new].weight; 272 cb(&rc_add, cbdata); 273 } 274 idx_old++; 275 idx_new++; 276 } else if (nh_idx_old < nh_idx_new) { 277 /* 278 * [1, ~2~, 4], [1, ~3~, 4] 279 * [1, ~2~, 5], [1, ~3~, 4] 280 * [1, ~2~], [1, ~3~, 4] 281 */ 282 if ((idx_old + 1 >= num_old) || 283 (wn_old[idx_old + 1].nh->nh_priv->nh_idx > nh_idx_new)) { 284 /* Add new unless the next old item is still <= new */ 285 rc_add.rc_nh_new = wn_new[idx_new].nh; 286 rc_add.rc_nh_weight = wn_new[idx_new].weight; 287 cb(&rc_add, cbdata); 288 idx_new++; 289 } 290 /* In any case, delete current old */ 291 rc_del.rc_nh_old = wn_old[idx_old].nh; 292 rc_del.rc_nh_weight = wn_old[idx_old].weight; 293 cb(&rc_del, cbdata); 294 idx_old++; 295 } else { 296 /* 297 * nh_idx_old > nh_idx_new 298 * 299 * [1, ~3~, 4], [1, ~2~, 4] 300 * [1, ~3~, 5], [1, ~2~, 4] 301 * [1, ~3~, 4], [1, ~2~] 302 */ 303 if ((idx_new + 1 >= num_new) || 304 (wn_new[idx_new + 1].nh->nh_priv->nh_idx > nh_idx_old)) { 305 /* No next item or next item is > current one */ 306 rc_add.rc_nh_new = wn_new[idx_new].nh; 307 rc_add.rc_nh_weight = wn_new[idx_new].weight; 308 cb(&rc_add, cbdata); 309 idx_new++; 310 } 311 /* In any case, delete current old */ 312 rc_del.rc_nh_old = wn_old[idx_old].nh; 313 rc_del.rc_nh_weight = wn_old[idx_old].weight; 314 cb(&rc_del, cbdata); 315 idx_old++; 316 } 317 } 318 319 while (idx_old < num_old) { 320 rc_del.rc_nh_old = wn_old[idx_old].nh; 321 rc_del.rc_nh_weight = wn_old[idx_old].weight; 322 cb(&rc_del, cbdata); 323 idx_old++; 324 } 325 326 while (idx_new < num_new) { 327 rc_add.rc_nh_new = wn_new[idx_new].nh; 328 rc_add.rc_nh_weight = wn_new[idx_new].weight; 329 cb(&rc_add, cbdata); 330 idx_new++; 331 } 332 } 333 334 /* 335 * Decompose multipath cmd info @rc into a list of add/del/change 336 * single-path operations, calling @cb callback for each operation. 337 * Assumes at least one of the nexthops in @rc is multipath. 338 */ 339 void 340 rib_decompose_notification(struct rib_cmd_info *rc, route_notification_t *cb, 341 void *cbdata) 342 { 343 struct weightened_nhop *wn; 344 uint32_t num_nhops; 345 struct rib_cmd_info rc_new; 346 347 rc_new = *rc; 348 DPRINTF("cb=%p cmd=%d nh_old=%p nh_new=%p", 349 cb, rc->cmd, rc->nh_old, rc->nh_new); 350 switch (rc->rc_cmd) { 351 case RTM_ADD: 352 if (!NH_IS_NHGRP(rc->rc_nh_new)) 353 return; 354 wn = nhgrp_get_nhops((struct nhgrp_object *)rc->rc_nh_new, &num_nhops); 355 for (uint32_t i = 0; i < num_nhops; i++) { 356 rc_new.rc_nh_new = wn[i].nh; 357 rc_new.rc_nh_weight = wn[i].weight; 358 cb(&rc_new, cbdata); 359 } 360 break; 361 case RTM_DELETE: 362 if (!NH_IS_NHGRP(rc->rc_nh_old)) 363 return; 364 wn = nhgrp_get_nhops((struct nhgrp_object *)rc->rc_nh_old, &num_nhops); 365 for (uint32_t i = 0; i < num_nhops; i++) { 366 rc_new.rc_nh_old = wn[i].nh; 367 rc_new.rc_nh_weight = wn[i].weight; 368 cb(&rc_new, cbdata); 369 } 370 break; 371 case RTM_CHANGE: 372 if (!NH_IS_NHGRP(rc->rc_nh_old) && !NH_IS_NHGRP(rc->rc_nh_new)) 373 return; 374 decompose_change_notification(rc, cb, cbdata); 375 break; 376 } 377 } 378 #endif 379