1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause 3 * 4 * Copyright (c) 2021 Ng Peng Nam Sean 5 * Copyright (c) 2022 Alexander V. Chernikov <melifaro@FreeBSD.org> 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 */ 28 29 /* 30 * This file contains socket and protocol bindings for netlink. 31 */ 32 33 #include <sys/param.h> 34 #include <sys/kernel.h> 35 #include <sys/malloc.h> 36 #include <sys/lock.h> 37 #include <sys/rmlock.h> 38 #include <sys/domain.h> 39 #include <sys/jail.h> 40 #include <sys/mbuf.h> 41 #include <sys/osd.h> 42 #include <sys/protosw.h> 43 #include <sys/proc.h> 44 #include <sys/ck.h> 45 #include <sys/socket.h> 46 #include <sys/socketvar.h> 47 #include <sys/sysent.h> 48 #include <sys/syslog.h> 49 #include <sys/priv.h> /* priv_check */ 50 51 #include <netlink/netlink.h> 52 #include <netlink/netlink_ctl.h> 53 #include <netlink/netlink_var.h> 54 55 #define DEBUG_MOD_NAME nl_domain 56 #define DEBUG_MAX_LEVEL LOG_DEBUG3 57 #include <netlink/netlink_debug.h> 58 _DECLARE_DEBUG(LOG_INFO); 59 60 _Static_assert((NLP_MAX_GROUPS % 64) == 0, 61 "NLP_MAX_GROUPS has to be multiple of 64"); 62 _Static_assert(NLP_MAX_GROUPS >= 64, 63 "NLP_MAX_GROUPS has to be at least 64"); 64 65 #define NLCTL_TRACKER struct rm_priotracker nl_tracker 66 #define NLCTL_RLOCK(_ctl) rm_rlock(&((_ctl)->ctl_lock), &nl_tracker) 67 #define NLCTL_RUNLOCK(_ctl) rm_runlock(&((_ctl)->ctl_lock), &nl_tracker) 68 69 #define NLCTL_WLOCK(_ctl) rm_wlock(&((_ctl)->ctl_lock)) 70 #define NLCTL_WUNLOCK(_ctl) rm_wunlock(&((_ctl)->ctl_lock)) 71 72 static u_long nl_sendspace = NLSNDQ; 73 SYSCTL_ULONG(_net_netlink, OID_AUTO, sendspace, CTLFLAG_RW, &nl_sendspace, 0, 74 "Default netlink socket send space"); 75 76 static u_long nl_recvspace = NLSNDQ; 77 SYSCTL_ULONG(_net_netlink, OID_AUTO, recvspace, CTLFLAG_RW, &nl_recvspace, 0, 78 "Default netlink socket receive space"); 79 80 extern u_long sb_max_adj; 81 static u_long nl_maxsockbuf = 512 * 1024 * 1024; /* 512M, XXX: init based on physmem */ 82 static int sysctl_handle_nl_maxsockbuf(SYSCTL_HANDLER_ARGS); 83 SYSCTL_OID(_net_netlink, OID_AUTO, nl_maxsockbuf, 84 CTLTYPE_ULONG | CTLFLAG_RW | CTLFLAG_MPSAFE, &nl_maxsockbuf, 0, 85 sysctl_handle_nl_maxsockbuf, "LU", 86 "Maximum Netlink socket buffer size"); 87 88 89 static unsigned int osd_slot_id = 0; 90 91 void 92 nl_osd_register(void) 93 { 94 osd_slot_id = osd_register(OSD_THREAD, NULL, NULL); 95 } 96 97 void 98 nl_osd_unregister(void) 99 { 100 osd_deregister(OSD_THREAD, osd_slot_id); 101 } 102 103 struct nlpcb * 104 _nl_get_thread_nlp(struct thread *td) 105 { 106 return (osd_get(OSD_THREAD, &td->td_osd, osd_slot_id)); 107 } 108 109 void 110 nl_set_thread_nlp(struct thread *td, struct nlpcb *nlp) 111 { 112 NLP_LOG(LOG_DEBUG2, nlp, "Set thread %p nlp to %p (slot %u)", td, nlp, osd_slot_id); 113 if (osd_set(OSD_THREAD, &td->td_osd, osd_slot_id, nlp) == 0) 114 return; 115 /* Failed, need to realloc */ 116 void **rsv = osd_reserve(osd_slot_id); 117 osd_set_reserved(OSD_THREAD, &td->td_osd, osd_slot_id, rsv, nlp); 118 } 119 120 /* 121 * Looks up a nlpcb struct based on the @portid. Need to claim nlsock_mtx. 122 * Returns nlpcb pointer if present else NULL 123 */ 124 static struct nlpcb * 125 nl_port_lookup(uint32_t port_id) 126 { 127 struct nlpcb *nlp; 128 129 CK_LIST_FOREACH(nlp, &V_nl_ctl->ctl_port_head, nl_port_next) { 130 if (nlp->nl_port == port_id) 131 return (nlp); 132 } 133 return (NULL); 134 } 135 136 static void 137 nl_add_group_locked(struct nlpcb *nlp, unsigned int group_id) 138 { 139 MPASS(group_id <= NLP_MAX_GROUPS); 140 --group_id; 141 142 /* TODO: add family handler callback */ 143 if (!nlp_unconstrained_vnet(nlp)) 144 return; 145 146 nlp->nl_groups[group_id / 64] |= (uint64_t)1 << (group_id % 64); 147 } 148 149 static void 150 nl_del_group_locked(struct nlpcb *nlp, unsigned int group_id) 151 { 152 MPASS(group_id <= NLP_MAX_GROUPS); 153 --group_id; 154 155 nlp->nl_groups[group_id / 64] &= ~((uint64_t)1 << (group_id % 64)); 156 } 157 158 static bool 159 nl_isset_group_locked(struct nlpcb *nlp, unsigned int group_id) 160 { 161 MPASS(group_id <= NLP_MAX_GROUPS); 162 --group_id; 163 164 return (nlp->nl_groups[group_id / 64] & ((uint64_t)1 << (group_id % 64))); 165 } 166 167 static uint32_t 168 nl_get_groups_compat(struct nlpcb *nlp) 169 { 170 uint32_t groups_mask = 0; 171 172 for (int i = 0; i < 32; i++) { 173 if (nl_isset_group_locked(nlp, i + 1)) 174 groups_mask |= (1 << i); 175 } 176 177 return (groups_mask); 178 } 179 180 static void 181 nl_send_one_group(struct mbuf *m, struct nlpcb *nlp, int num_messages, 182 int io_flags) 183 { 184 if (__predict_false(nlp->nl_flags & NLF_MSG_INFO)) 185 nl_add_msg_info(m); 186 nl_send_one(m, nlp, num_messages, io_flags); 187 } 188 189 /* 190 * Broadcasts message @m to the protocol @proto group specified by @group_id 191 */ 192 void 193 nl_send_group(struct mbuf *m, int num_messages, int proto, int group_id) 194 { 195 struct nlpcb *nlp_last = NULL; 196 struct nlpcb *nlp; 197 NLCTL_TRACKER; 198 199 IF_DEBUG_LEVEL(LOG_DEBUG2) { 200 struct nlmsghdr *hdr = mtod(m, struct nlmsghdr *); 201 NL_LOG(LOG_DEBUG2, "MCAST mbuf len %u msg type %d len %u to group %d/%d", 202 m->m_len, hdr->nlmsg_type, hdr->nlmsg_len, proto, group_id); 203 } 204 205 struct nl_control *ctl = atomic_load_ptr(&V_nl_ctl); 206 if (__predict_false(ctl == NULL)) { 207 /* 208 * Can be the case when notification is sent within VNET 209 * which doesn't have any netlink sockets. 210 */ 211 m_freem(m); 212 return; 213 } 214 215 NLCTL_RLOCK(ctl); 216 217 int io_flags = NL_IOF_UNTRANSLATED; 218 219 CK_LIST_FOREACH(nlp, &ctl->ctl_pcb_head, nl_next) { 220 if (nl_isset_group_locked(nlp, group_id) && nlp->nl_proto == proto) { 221 if (nlp_last != NULL) { 222 struct mbuf *m_copy; 223 m_copy = m_copym(m, 0, M_COPYALL, M_NOWAIT); 224 if (m_copy != NULL) 225 nl_send_one_group(m_copy, nlp_last, 226 num_messages, io_flags); 227 else { 228 NLP_LOCK(nlp_last); 229 if (nlp_last->nl_socket != NULL) 230 sorwakeup(nlp_last->nl_socket); 231 NLP_UNLOCK(nlp_last); 232 } 233 } 234 nlp_last = nlp; 235 } 236 } 237 if (nlp_last != NULL) 238 nl_send_one_group(m, nlp_last, num_messages, io_flags); 239 else 240 m_freem(m); 241 242 NLCTL_RUNLOCK(ctl); 243 } 244 245 bool 246 nl_has_listeners(int netlink_family, uint32_t groups_mask) 247 { 248 return (V_nl_ctl != NULL); 249 } 250 251 static uint32_t 252 nl_find_port(void) 253 { 254 /* 255 * app can open multiple netlink sockets. 256 * Start with current pid, if already taken, 257 * try random numbers in 65k..256k+65k space, 258 * avoiding clash with pids. 259 */ 260 if (nl_port_lookup(curproc->p_pid) == NULL) 261 return (curproc->p_pid); 262 for (int i = 0; i < 16; i++) { 263 uint32_t nl_port = (arc4random() % 65536) + 65536 * 4; 264 if (nl_port_lookup(nl_port) == 0) 265 return (nl_port); 266 NL_LOG(LOG_DEBUG3, "tried %u\n", nl_port); 267 } 268 return (curproc->p_pid); 269 } 270 271 static int 272 nl_bind_locked(struct nlpcb *nlp, struct sockaddr_nl *snl) 273 { 274 if (nlp->nl_bound) { 275 if (nlp->nl_port != snl->nl_pid) { 276 NL_LOG(LOG_DEBUG, 277 "bind() failed: program pid %d " 278 "is different from provided pid %d", 279 nlp->nl_port, snl->nl_pid); 280 return (EINVAL); // XXX: better error 281 } 282 } else { 283 if (snl->nl_pid == 0) 284 snl->nl_pid = nl_find_port(); 285 if (nl_port_lookup(snl->nl_pid) != NULL) 286 return (EADDRINUSE); 287 nlp->nl_port = snl->nl_pid; 288 nlp->nl_bound = true; 289 CK_LIST_INSERT_HEAD(&V_nl_ctl->ctl_port_head, nlp, nl_port_next); 290 } 291 for (int i = 0; i < 32; i++) { 292 if (snl->nl_groups & ((uint32_t)1 << i)) 293 nl_add_group_locked(nlp, i + 1); 294 else 295 nl_del_group_locked(nlp, i + 1); 296 } 297 298 return (0); 299 } 300 301 static int 302 nl_pru_attach(struct socket *so, int proto, struct thread *td) 303 { 304 struct nlpcb *nlp; 305 int error; 306 307 if (__predict_false(netlink_unloading != 0)) 308 return (EAFNOSUPPORT); 309 310 error = nl_verify_proto(proto); 311 if (error != 0) 312 return (error); 313 314 bool is_linux = SV_PROC_ABI(td->td_proc) == SV_ABI_LINUX; 315 NL_LOG(LOG_DEBUG2, "socket %p, %sPID %d: attaching socket to %s", 316 so, is_linux ? "(linux) " : "", curproc->p_pid, 317 nl_get_proto_name(proto)); 318 319 /* Create per-VNET state on first socket init */ 320 struct nl_control *ctl = atomic_load_ptr(&V_nl_ctl); 321 if (ctl == NULL) 322 ctl = vnet_nl_ctl_init(); 323 KASSERT(V_nl_ctl != NULL, ("nl_attach: vnet_sock_init() failed")); 324 325 MPASS(sotonlpcb(so) == NULL); 326 327 nlp = malloc(sizeof(struct nlpcb), M_PCB, M_WAITOK | M_ZERO); 328 error = soreserve(so, nl_sendspace, nl_recvspace); 329 if (error != 0) { 330 free(nlp, M_PCB); 331 return (error); 332 } 333 so->so_pcb = nlp; 334 nlp->nl_socket = so; 335 /* Copy so_cred to avoid having socket_var.h in every header */ 336 nlp->nl_cred = so->so_cred; 337 nlp->nl_proto = proto; 338 nlp->nl_process_id = curproc->p_pid; 339 nlp->nl_linux = is_linux; 340 nlp->nl_active = true; 341 nlp->nl_unconstrained_vnet = !jailed_without_vnet(so->so_cred); 342 nlp->nl_need_thread_setup = true; 343 NLP_LOCK_INIT(nlp); 344 refcount_init(&nlp->nl_refcount, 1); 345 nl_init_io(nlp); 346 347 nlp->nl_taskqueue = taskqueue_create("netlink_socket", M_WAITOK, 348 taskqueue_thread_enqueue, &nlp->nl_taskqueue); 349 TASK_INIT(&nlp->nl_task, 0, nl_taskqueue_handler, nlp); 350 taskqueue_start_threads(&nlp->nl_taskqueue, 1, PWAIT, 351 "netlink_socket (PID %u)", nlp->nl_process_id); 352 353 NLCTL_WLOCK(ctl); 354 /* XXX: check ctl is still alive */ 355 CK_LIST_INSERT_HEAD(&ctl->ctl_pcb_head, nlp, nl_next); 356 NLCTL_WUNLOCK(ctl); 357 358 soisconnected(so); 359 360 return (0); 361 } 362 363 static void 364 nl_pru_abort(struct socket *so) 365 { 366 NL_LOG(LOG_DEBUG3, "socket %p, PID %d", so, curproc->p_pid); 367 MPASS(sotonlpcb(so) != NULL); 368 soisdisconnected(so); 369 } 370 371 static int 372 nl_pru_bind(struct socket *so, struct sockaddr *sa, struct thread *td) 373 { 374 struct nl_control *ctl = atomic_load_ptr(&V_nl_ctl); 375 struct nlpcb *nlp = sotonlpcb(so); 376 struct sockaddr_nl *snl = (struct sockaddr_nl *)sa; 377 int error; 378 379 NL_LOG(LOG_DEBUG3, "socket %p, PID %d", so, curproc->p_pid); 380 if (snl->nl_len != sizeof(*snl)) { 381 NL_LOG(LOG_DEBUG, "socket %p, wrong sizeof(), ignoring bind()", so); 382 return (EINVAL); 383 } 384 385 386 NLCTL_WLOCK(ctl); 387 NLP_LOCK(nlp); 388 error = nl_bind_locked(nlp, snl); 389 NLP_UNLOCK(nlp); 390 NLCTL_WUNLOCK(ctl); 391 NL_LOG(LOG_DEBUG2, "socket %p, bind() to %u, groups %u, error %d", so, 392 snl->nl_pid, snl->nl_groups, error); 393 394 return (error); 395 } 396 397 398 static int 399 nl_assign_port(struct nlpcb *nlp, uint32_t port_id) 400 { 401 struct nl_control *ctl = atomic_load_ptr(&V_nl_ctl); 402 struct sockaddr_nl snl = { 403 .nl_pid = port_id, 404 }; 405 int error; 406 407 NLCTL_WLOCK(ctl); 408 NLP_LOCK(nlp); 409 snl.nl_groups = nl_get_groups_compat(nlp); 410 error = nl_bind_locked(nlp, &snl); 411 NLP_UNLOCK(nlp); 412 NLCTL_WUNLOCK(ctl); 413 414 NL_LOG(LOG_DEBUG3, "socket %p, port assign: %d, error: %d", nlp->nl_socket, port_id, error); 415 return (error); 416 } 417 418 /* 419 * nl_autobind_port binds a unused portid to @nlp 420 * @nlp: pcb data for the netlink socket 421 * @candidate_id: first id to consider 422 */ 423 static int 424 nl_autobind_port(struct nlpcb *nlp, uint32_t candidate_id) 425 { 426 struct nl_control *ctl = atomic_load_ptr(&V_nl_ctl); 427 uint32_t port_id = candidate_id; 428 NLCTL_TRACKER; 429 bool exist; 430 int error = EADDRINUSE; 431 432 for (int i = 0; i < 10; i++) { 433 NL_LOG(LOG_DEBUG3, "socket %p, trying to assign port %d", nlp->nl_socket, port_id); 434 NLCTL_RLOCK(ctl); 435 exist = nl_port_lookup(port_id) != 0; 436 NLCTL_RUNLOCK(ctl); 437 if (!exist) { 438 error = nl_assign_port(nlp, port_id); 439 if (error != EADDRINUSE) 440 break; 441 } 442 port_id++; 443 } 444 NL_LOG(LOG_DEBUG3, "socket %p, autobind to %d, error: %d", nlp->nl_socket, port_id, error); 445 return (error); 446 } 447 448 static int 449 nl_pru_connect(struct socket *so, struct sockaddr *sa, struct thread *td) 450 { 451 struct sockaddr_nl *snl = (struct sockaddr_nl *)sa; 452 struct nlpcb *nlp; 453 454 NL_LOG(LOG_DEBUG3, "socket %p, PID %d", so, curproc->p_pid); 455 if (snl->nl_len != sizeof(*snl)) { 456 NL_LOG(LOG_DEBUG, "socket %p, wrong sizeof(), ignoring bind()", so); 457 return (EINVAL); 458 } 459 460 nlp = sotonlpcb(so); 461 if (!nlp->nl_bound) { 462 int error = nl_autobind_port(nlp, td->td_proc->p_pid); 463 if (error != 0) { 464 NL_LOG(LOG_DEBUG, "socket %p, nl_autobind() failed: %d", so, error); 465 return (error); 466 } 467 } 468 /* XXX: Handle socket flags & multicast */ 469 soisconnected(so); 470 471 NL_LOG(LOG_DEBUG2, "socket %p, connect to %u", so, snl->nl_pid); 472 473 return (0); 474 } 475 476 static void 477 destroy_nlpcb(struct nlpcb *nlp) 478 { 479 NLP_LOCK(nlp); 480 nl_free_io(nlp); 481 NLP_LOCK_DESTROY(nlp); 482 free(nlp, M_PCB); 483 } 484 485 static void 486 destroy_nlpcb_epoch(epoch_context_t ctx) 487 { 488 struct nlpcb *nlp; 489 490 nlp = __containerof(ctx, struct nlpcb, nl_epoch_ctx); 491 492 destroy_nlpcb(nlp); 493 } 494 495 496 static void 497 nl_pru_detach(struct socket *so) 498 { 499 struct nl_control *ctl = atomic_load_ptr(&V_nl_ctl); 500 MPASS(sotonlpcb(so) != NULL); 501 struct nlpcb *nlp; 502 503 NL_LOG(LOG_DEBUG2, "detaching socket %p, PID %d", so, curproc->p_pid); 504 nlp = sotonlpcb(so); 505 506 /* Mark as inactive so no new work can be enqueued */ 507 NLP_LOCK(nlp); 508 bool was_bound = nlp->nl_bound; 509 nlp->nl_active = false; 510 NLP_UNLOCK(nlp); 511 512 /* Wait till all scheduled work has been completed */ 513 taskqueue_drain_all(nlp->nl_taskqueue); 514 taskqueue_free(nlp->nl_taskqueue); 515 516 NLCTL_WLOCK(ctl); 517 NLP_LOCK(nlp); 518 if (was_bound) { 519 CK_LIST_REMOVE(nlp, nl_port_next); 520 NL_LOG(LOG_DEBUG3, "socket %p, unlinking bound pid %u", so, nlp->nl_port); 521 } 522 CK_LIST_REMOVE(nlp, nl_next); 523 nlp->nl_socket = NULL; 524 NLP_UNLOCK(nlp); 525 NLCTL_WUNLOCK(ctl); 526 527 so->so_pcb = NULL; 528 529 NL_LOG(LOG_DEBUG3, "socket %p, detached", so); 530 531 /* XXX: is delayed free needed? */ 532 NET_EPOCH_CALL(destroy_nlpcb_epoch, &nlp->nl_epoch_ctx); 533 } 534 535 static int 536 nl_pru_disconnect(struct socket *so) 537 { 538 NL_LOG(LOG_DEBUG3, "socket %p, PID %d", so, curproc->p_pid); 539 MPASS(sotonlpcb(so) != NULL); 540 return (ENOTCONN); 541 } 542 543 static int 544 nl_pru_peeraddr(struct socket *so, struct sockaddr **sa) 545 { 546 NL_LOG(LOG_DEBUG3, "socket %p, PID %d", so, curproc->p_pid); 547 MPASS(sotonlpcb(so) != NULL); 548 return (ENOTCONN); 549 } 550 551 static int 552 nl_pru_shutdown(struct socket *so) 553 { 554 NL_LOG(LOG_DEBUG3, "socket %p, PID %d", so, curproc->p_pid); 555 MPASS(sotonlpcb(so) != NULL); 556 socantsendmore(so); 557 return (0); 558 } 559 560 static int 561 nl_pru_sockaddr(struct socket *so, struct sockaddr **sa) 562 { 563 struct sockaddr_nl *snl; 564 565 snl = malloc(sizeof(struct sockaddr_nl), M_SONAME, M_WAITOK | M_ZERO); 566 /* TODO: set other fields */ 567 snl->nl_len = sizeof(struct sockaddr_nl); 568 snl->nl_family = AF_NETLINK; 569 snl->nl_pid = sotonlpcb(so)->nl_port; 570 *sa = (struct sockaddr *)snl; 571 return (0); 572 } 573 574 static void 575 nl_pru_close(struct socket *so) 576 { 577 NL_LOG(LOG_DEBUG3, "socket %p, PID %d", so, curproc->p_pid); 578 MPASS(sotonlpcb(so) != NULL); 579 soisdisconnected(so); 580 } 581 582 static int 583 nl_pru_output(struct mbuf *m, struct socket *so, ...) 584 { 585 586 if (__predict_false(m == NULL || 587 ((m->m_len < sizeof(struct nlmsghdr)) && 588 (m = m_pullup(m, sizeof(struct nlmsghdr))) == NULL))) 589 return (ENOBUFS); 590 MPASS((m->m_flags & M_PKTHDR) != 0); 591 592 NL_LOG(LOG_DEBUG3, "sending message to kernel async processing"); 593 nl_receive_async(m, so); 594 return (0); 595 } 596 597 598 static int 599 nl_pru_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *sa, 600 struct mbuf *control, struct thread *td) 601 { 602 NL_LOG(LOG_DEBUG2, "sending message to kernel"); 603 604 if (__predict_false(control != NULL)) { 605 if (control->m_len) { 606 m_freem(control); 607 return (EINVAL); 608 } 609 m_freem(control); 610 } 611 612 return (nl_pru_output(m, so)); 613 } 614 615 static int 616 nl_pru_rcvd(struct socket *so, int flags) 617 { 618 NL_LOG(LOG_DEBUG3, "socket %p, PID %d", so, curproc->p_pid); 619 MPASS(sotonlpcb(so) != NULL); 620 621 nl_on_transmit(sotonlpcb(so)); 622 623 return (0); 624 } 625 626 static int 627 nl_getoptflag(int sopt_name) 628 { 629 switch (sopt_name) { 630 case NETLINK_CAP_ACK: 631 return (NLF_CAP_ACK); 632 case NETLINK_EXT_ACK: 633 return (NLF_EXT_ACK); 634 case NETLINK_GET_STRICT_CHK: 635 return (NLF_STRICT); 636 case NETLINK_MSG_INFO: 637 return (NLF_MSG_INFO); 638 } 639 640 return (0); 641 } 642 643 static int 644 nl_ctloutput(struct socket *so, struct sockopt *sopt) 645 { 646 struct nl_control *ctl = atomic_load_ptr(&V_nl_ctl); 647 struct nlpcb *nlp = sotonlpcb(so); 648 uint32_t flag; 649 int optval, error = 0; 650 NLCTL_TRACKER; 651 652 NL_LOG(LOG_DEBUG2, "%ssockopt(%p, %d)", (sopt->sopt_dir) ? "set" : "get", 653 so, sopt->sopt_name); 654 655 switch (sopt->sopt_dir) { 656 case SOPT_SET: 657 switch (sopt->sopt_name) { 658 case NETLINK_ADD_MEMBERSHIP: 659 case NETLINK_DROP_MEMBERSHIP: 660 error = sooptcopyin(sopt, &optval, sizeof(optval), sizeof(optval)); 661 if (error != 0) 662 break; 663 if (optval <= 0 || optval >= NLP_MAX_GROUPS) { 664 error = ERANGE; 665 break; 666 } 667 NL_LOG(LOG_DEBUG2, "ADD/DEL group %d", (uint32_t)optval); 668 669 NLCTL_WLOCK(ctl); 670 if (sopt->sopt_name == NETLINK_ADD_MEMBERSHIP) 671 nl_add_group_locked(nlp, optval); 672 else 673 nl_del_group_locked(nlp, optval); 674 NLCTL_WUNLOCK(ctl); 675 break; 676 case NETLINK_CAP_ACK: 677 case NETLINK_EXT_ACK: 678 case NETLINK_GET_STRICT_CHK: 679 case NETLINK_MSG_INFO: 680 error = sooptcopyin(sopt, &optval, sizeof(optval), sizeof(optval)); 681 if (error != 0) 682 break; 683 684 flag = nl_getoptflag(sopt->sopt_name); 685 686 if ((flag == NLF_MSG_INFO) && nlp->nl_linux) { 687 error = EINVAL; 688 break; 689 } 690 691 NLCTL_WLOCK(ctl); 692 if (optval != 0) 693 nlp->nl_flags |= flag; 694 else 695 nlp->nl_flags &= ~flag; 696 NLCTL_WUNLOCK(ctl); 697 break; 698 default: 699 error = ENOPROTOOPT; 700 } 701 break; 702 case SOPT_GET: 703 switch (sopt->sopt_name) { 704 case NETLINK_LIST_MEMBERSHIPS: 705 NLCTL_RLOCK(ctl); 706 optval = nl_get_groups_compat(nlp); 707 NLCTL_RUNLOCK(ctl); 708 error = sooptcopyout(sopt, &optval, sizeof(optval)); 709 break; 710 case NETLINK_CAP_ACK: 711 case NETLINK_EXT_ACK: 712 case NETLINK_GET_STRICT_CHK: 713 case NETLINK_MSG_INFO: 714 NLCTL_RLOCK(ctl); 715 optval = (nlp->nl_flags & nl_getoptflag(sopt->sopt_name)) != 0; 716 NLCTL_RUNLOCK(ctl); 717 error = sooptcopyout(sopt, &optval, sizeof(optval)); 718 break; 719 default: 720 error = ENOPROTOOPT; 721 } 722 break; 723 default: 724 error = ENOPROTOOPT; 725 } 726 727 return (error); 728 } 729 730 static int 731 sysctl_handle_nl_maxsockbuf(SYSCTL_HANDLER_ARGS) 732 { 733 int error = 0; 734 u_long tmp_maxsockbuf = nl_maxsockbuf; 735 736 error = sysctl_handle_long(oidp, &tmp_maxsockbuf, arg2, req); 737 if (error || !req->newptr) 738 return (error); 739 if (tmp_maxsockbuf < MSIZE + MCLBYTES) 740 return (EINVAL); 741 nl_maxsockbuf = tmp_maxsockbuf; 742 743 return (0); 744 } 745 746 static int 747 nl_setsbopt(struct socket *so, struct sockopt *sopt) 748 { 749 int error, optval; 750 bool result; 751 752 if (sopt->sopt_name != SO_RCVBUF) 753 return (sbsetopt(so, sopt)); 754 755 /* Allow to override max buffer size in certain conditions */ 756 757 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 758 if (error != 0) 759 return (error); 760 NL_LOG(LOG_DEBUG2, "socket %p, PID %d, SO_RCVBUF=%d", so, curproc->p_pid, optval); 761 if (optval > sb_max_adj) { 762 if (priv_check(curthread, PRIV_NET_ROUTE) != 0) 763 return (EPERM); 764 } 765 766 SOCK_RECVBUF_LOCK(so); 767 result = sbreserve_locked_limit(so, SO_RCV, optval, nl_maxsockbuf, curthread); 768 SOCK_RECVBUF_UNLOCK(so); 769 770 return (result ? 0 : ENOBUFS); 771 } 772 773 #define NETLINK_PROTOSW \ 774 .pr_flags = PR_ATOMIC | PR_ADDR | PR_WANTRCVD, \ 775 .pr_ctloutput = nl_ctloutput, \ 776 .pr_setsbopt = nl_setsbopt, \ 777 .pr_abort = nl_pru_abort, \ 778 .pr_attach = nl_pru_attach, \ 779 .pr_bind = nl_pru_bind, \ 780 .pr_connect = nl_pru_connect, \ 781 .pr_detach = nl_pru_detach, \ 782 .pr_disconnect = nl_pru_disconnect, \ 783 .pr_peeraddr = nl_pru_peeraddr, \ 784 .pr_send = nl_pru_send, \ 785 .pr_rcvd = nl_pru_rcvd, \ 786 .pr_shutdown = nl_pru_shutdown, \ 787 .pr_sockaddr = nl_pru_sockaddr, \ 788 .pr_close = nl_pru_close 789 790 static struct protosw netlink_raw_sw = { 791 .pr_type = SOCK_RAW, 792 NETLINK_PROTOSW 793 }; 794 795 static struct protosw netlink_dgram_sw = { 796 .pr_type = SOCK_DGRAM, 797 NETLINK_PROTOSW 798 }; 799 800 static struct domain netlinkdomain = { 801 .dom_family = PF_NETLINK, 802 .dom_name = "netlink", 803 .dom_flags = DOMF_UNLOADABLE, 804 .dom_nprotosw = 2, 805 .dom_protosw = { &netlink_raw_sw, &netlink_dgram_sw }, 806 }; 807 808 DOMAIN_SET(netlink); 809