1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Routing netlink socket interface: protocol independent part. 7 * 8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> 9 * 10 * This program is free software; you can redistribute it and/or 11 * modify it under the terms of the GNU General Public License 12 * as published by the Free Software Foundation; either version 13 * 2 of the License, or (at your option) any later version. 14 * 15 * Fixes: 16 * Vitaly E. Lavrov RTA_OK arithmetics was wrong. 17 */ 18 19 #include <linux/errno.h> 20 #include <linux/module.h> 21 #include <linux/types.h> 22 #include <linux/socket.h> 23 #include <linux/kernel.h> 24 #include <linux/timer.h> 25 #include <linux/string.h> 26 #include <linux/sockios.h> 27 #include <linux/net.h> 28 #include <linux/fcntl.h> 29 #include <linux/mm.h> 30 #include <linux/slab.h> 31 #include <linux/interrupt.h> 32 #include <linux/capability.h> 33 #include <linux/skbuff.h> 34 #include <linux/init.h> 35 #include <linux/security.h> 36 #include <linux/mutex.h> 37 #include <linux/if_addr.h> 38 #include <linux/pci.h> 39 40 #include <asm/uaccess.h> 41 #include <asm/system.h> 42 43 #include <linux/inet.h> 44 #include <linux/netdevice.h> 45 #include <net/ip.h> 46 #include <net/protocol.h> 47 #include <net/arp.h> 48 #include <net/route.h> 49 #include <net/udp.h> 50 #include <net/sock.h> 51 #include <net/pkt_sched.h> 52 #include <net/fib_rules.h> 53 #include <net/rtnetlink.h> 54 #include <net/net_namespace.h> 55 56 struct rtnl_link { 57 rtnl_doit_func doit; 58 rtnl_dumpit_func dumpit; 59 }; 60 61 static DEFINE_MUTEX(rtnl_mutex); 62 63 void rtnl_lock(void) 64 { 65 mutex_lock(&rtnl_mutex); 66 } 67 EXPORT_SYMBOL(rtnl_lock); 68 69 void __rtnl_unlock(void) 70 { 71 mutex_unlock(&rtnl_mutex); 72 } 73 74 void rtnl_unlock(void) 75 { 76 /* This fellow will unlock it for us. */ 77 netdev_run_todo(); 78 } 79 EXPORT_SYMBOL(rtnl_unlock); 80 81 int rtnl_trylock(void) 82 { 83 return mutex_trylock(&rtnl_mutex); 84 } 85 EXPORT_SYMBOL(rtnl_trylock); 86 87 int rtnl_is_locked(void) 88 { 89 return mutex_is_locked(&rtnl_mutex); 90 } 91 EXPORT_SYMBOL(rtnl_is_locked); 92 93 #ifdef CONFIG_PROVE_LOCKING 94 int lockdep_rtnl_is_held(void) 95 { 96 return lockdep_is_held(&rtnl_mutex); 97 } 98 EXPORT_SYMBOL(lockdep_rtnl_is_held); 99 #endif /* #ifdef CONFIG_PROVE_LOCKING */ 100 101 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1]; 102 103 static inline int rtm_msgindex(int msgtype) 104 { 105 int msgindex = msgtype - RTM_BASE; 106 107 /* 108 * msgindex < 0 implies someone tried to register a netlink 109 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that 110 * the message type has not been added to linux/rtnetlink.h 111 */ 112 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES); 113 114 return msgindex; 115 } 116 117 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex) 118 { 119 struct rtnl_link *tab; 120 121 if (protocol <= RTNL_FAMILY_MAX) 122 tab = rtnl_msg_handlers[protocol]; 123 else 124 tab = NULL; 125 126 if (tab == NULL || tab[msgindex].doit == NULL) 127 tab = rtnl_msg_handlers[PF_UNSPEC]; 128 129 return tab ? tab[msgindex].doit : NULL; 130 } 131 132 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex) 133 { 134 struct rtnl_link *tab; 135 136 if (protocol <= RTNL_FAMILY_MAX) 137 tab = rtnl_msg_handlers[protocol]; 138 else 139 tab = NULL; 140 141 if (tab == NULL || tab[msgindex].dumpit == NULL) 142 tab = rtnl_msg_handlers[PF_UNSPEC]; 143 144 return tab ? tab[msgindex].dumpit : NULL; 145 } 146 147 /** 148 * __rtnl_register - Register a rtnetlink message type 149 * @protocol: Protocol family or PF_UNSPEC 150 * @msgtype: rtnetlink message type 151 * @doit: Function pointer called for each request message 152 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message 153 * 154 * Registers the specified function pointers (at least one of them has 155 * to be non-NULL) to be called whenever a request message for the 156 * specified protocol family and message type is received. 157 * 158 * The special protocol family PF_UNSPEC may be used to define fallback 159 * function pointers for the case when no entry for the specific protocol 160 * family exists. 161 * 162 * Returns 0 on success or a negative error code. 163 */ 164 int __rtnl_register(int protocol, int msgtype, 165 rtnl_doit_func doit, rtnl_dumpit_func dumpit) 166 { 167 struct rtnl_link *tab; 168 int msgindex; 169 170 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 171 msgindex = rtm_msgindex(msgtype); 172 173 tab = rtnl_msg_handlers[protocol]; 174 if (tab == NULL) { 175 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL); 176 if (tab == NULL) 177 return -ENOBUFS; 178 179 rtnl_msg_handlers[protocol] = tab; 180 } 181 182 if (doit) 183 tab[msgindex].doit = doit; 184 185 if (dumpit) 186 tab[msgindex].dumpit = dumpit; 187 188 return 0; 189 } 190 EXPORT_SYMBOL_GPL(__rtnl_register); 191 192 /** 193 * rtnl_register - Register a rtnetlink message type 194 * 195 * Identical to __rtnl_register() but panics on failure. This is useful 196 * as failure of this function is very unlikely, it can only happen due 197 * to lack of memory when allocating the chain to store all message 198 * handlers for a protocol. Meant for use in init functions where lack 199 * of memory implies no sense in continueing. 200 */ 201 void rtnl_register(int protocol, int msgtype, 202 rtnl_doit_func doit, rtnl_dumpit_func dumpit) 203 { 204 if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0) 205 panic("Unable to register rtnetlink message handler, " 206 "protocol = %d, message type = %d\n", 207 protocol, msgtype); 208 } 209 EXPORT_SYMBOL_GPL(rtnl_register); 210 211 /** 212 * rtnl_unregister - Unregister a rtnetlink message type 213 * @protocol: Protocol family or PF_UNSPEC 214 * @msgtype: rtnetlink message type 215 * 216 * Returns 0 on success or a negative error code. 217 */ 218 int rtnl_unregister(int protocol, int msgtype) 219 { 220 int msgindex; 221 222 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 223 msgindex = rtm_msgindex(msgtype); 224 225 if (rtnl_msg_handlers[protocol] == NULL) 226 return -ENOENT; 227 228 rtnl_msg_handlers[protocol][msgindex].doit = NULL; 229 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL; 230 231 return 0; 232 } 233 EXPORT_SYMBOL_GPL(rtnl_unregister); 234 235 /** 236 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol 237 * @protocol : Protocol family or PF_UNSPEC 238 * 239 * Identical to calling rtnl_unregster() for all registered message types 240 * of a certain protocol family. 241 */ 242 void rtnl_unregister_all(int protocol) 243 { 244 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 245 246 kfree(rtnl_msg_handlers[protocol]); 247 rtnl_msg_handlers[protocol] = NULL; 248 } 249 EXPORT_SYMBOL_GPL(rtnl_unregister_all); 250 251 static LIST_HEAD(link_ops); 252 253 /** 254 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink. 255 * @ops: struct rtnl_link_ops * to register 256 * 257 * The caller must hold the rtnl_mutex. This function should be used 258 * by drivers that create devices during module initialization. It 259 * must be called before registering the devices. 260 * 261 * Returns 0 on success or a negative error code. 262 */ 263 int __rtnl_link_register(struct rtnl_link_ops *ops) 264 { 265 if (!ops->dellink) 266 ops->dellink = unregister_netdevice_queue; 267 268 list_add_tail(&ops->list, &link_ops); 269 return 0; 270 } 271 EXPORT_SYMBOL_GPL(__rtnl_link_register); 272 273 /** 274 * rtnl_link_register - Register rtnl_link_ops with rtnetlink. 275 * @ops: struct rtnl_link_ops * to register 276 * 277 * Returns 0 on success or a negative error code. 278 */ 279 int rtnl_link_register(struct rtnl_link_ops *ops) 280 { 281 int err; 282 283 rtnl_lock(); 284 err = __rtnl_link_register(ops); 285 rtnl_unlock(); 286 return err; 287 } 288 EXPORT_SYMBOL_GPL(rtnl_link_register); 289 290 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops) 291 { 292 struct net_device *dev; 293 LIST_HEAD(list_kill); 294 295 for_each_netdev(net, dev) { 296 if (dev->rtnl_link_ops == ops) 297 ops->dellink(dev, &list_kill); 298 } 299 unregister_netdevice_many(&list_kill); 300 } 301 302 /** 303 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink. 304 * @ops: struct rtnl_link_ops * to unregister 305 * 306 * The caller must hold the rtnl_mutex. 307 */ 308 void __rtnl_link_unregister(struct rtnl_link_ops *ops) 309 { 310 struct net *net; 311 312 for_each_net(net) { 313 __rtnl_kill_links(net, ops); 314 } 315 list_del(&ops->list); 316 } 317 EXPORT_SYMBOL_GPL(__rtnl_link_unregister); 318 319 /** 320 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink. 321 * @ops: struct rtnl_link_ops * to unregister 322 */ 323 void rtnl_link_unregister(struct rtnl_link_ops *ops) 324 { 325 rtnl_lock(); 326 __rtnl_link_unregister(ops); 327 rtnl_unlock(); 328 } 329 EXPORT_SYMBOL_GPL(rtnl_link_unregister); 330 331 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind) 332 { 333 const struct rtnl_link_ops *ops; 334 335 list_for_each_entry(ops, &link_ops, list) { 336 if (!strcmp(ops->kind, kind)) 337 return ops; 338 } 339 return NULL; 340 } 341 342 static size_t rtnl_link_get_size(const struct net_device *dev) 343 { 344 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 345 size_t size; 346 347 if (!ops) 348 return 0; 349 350 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */ 351 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */ 352 353 if (ops->get_size) 354 /* IFLA_INFO_DATA + nested data */ 355 size += nla_total_size(sizeof(struct nlattr)) + 356 ops->get_size(dev); 357 358 if (ops->get_xstats_size) 359 /* IFLA_INFO_XSTATS */ 360 size += nla_total_size(ops->get_xstats_size(dev)); 361 362 return size; 363 } 364 365 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev) 366 { 367 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 368 struct nlattr *linkinfo, *data; 369 int err = -EMSGSIZE; 370 371 linkinfo = nla_nest_start(skb, IFLA_LINKINFO); 372 if (linkinfo == NULL) 373 goto out; 374 375 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0) 376 goto err_cancel_link; 377 if (ops->fill_xstats) { 378 err = ops->fill_xstats(skb, dev); 379 if (err < 0) 380 goto err_cancel_link; 381 } 382 if (ops->fill_info) { 383 data = nla_nest_start(skb, IFLA_INFO_DATA); 384 if (data == NULL) 385 goto err_cancel_link; 386 err = ops->fill_info(skb, dev); 387 if (err < 0) 388 goto err_cancel_data; 389 nla_nest_end(skb, data); 390 } 391 392 nla_nest_end(skb, linkinfo); 393 return 0; 394 395 err_cancel_data: 396 nla_nest_cancel(skb, data); 397 err_cancel_link: 398 nla_nest_cancel(skb, linkinfo); 399 out: 400 return err; 401 } 402 403 static const int rtm_min[RTM_NR_FAMILIES] = 404 { 405 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)), 406 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)), 407 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)), 408 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)), 409 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)), 410 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)), 411 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)), 412 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)), 413 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)), 414 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)), 415 }; 416 417 static const int rta_max[RTM_NR_FAMILIES] = 418 { 419 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX, 420 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX, 421 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX, 422 [RTM_FAM(RTM_NEWRULE)] = FRA_MAX, 423 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX, 424 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX, 425 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX, 426 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX, 427 }; 428 429 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data) 430 { 431 struct rtattr *rta; 432 int size = RTA_LENGTH(attrlen); 433 434 rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size)); 435 rta->rta_type = attrtype; 436 rta->rta_len = size; 437 memcpy(RTA_DATA(rta), data, attrlen); 438 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size); 439 } 440 EXPORT_SYMBOL(__rta_fill); 441 442 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo) 443 { 444 struct sock *rtnl = net->rtnl; 445 int err = 0; 446 447 NETLINK_CB(skb).dst_group = group; 448 if (echo) 449 atomic_inc(&skb->users); 450 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL); 451 if (echo) 452 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT); 453 return err; 454 } 455 456 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid) 457 { 458 struct sock *rtnl = net->rtnl; 459 460 return nlmsg_unicast(rtnl, skb, pid); 461 } 462 EXPORT_SYMBOL(rtnl_unicast); 463 464 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group, 465 struct nlmsghdr *nlh, gfp_t flags) 466 { 467 struct sock *rtnl = net->rtnl; 468 int report = 0; 469 470 if (nlh) 471 report = nlmsg_report(nlh); 472 473 nlmsg_notify(rtnl, skb, pid, group, report, flags); 474 } 475 EXPORT_SYMBOL(rtnl_notify); 476 477 void rtnl_set_sk_err(struct net *net, u32 group, int error) 478 { 479 struct sock *rtnl = net->rtnl; 480 481 netlink_set_err(rtnl, 0, group, error); 482 } 483 EXPORT_SYMBOL(rtnl_set_sk_err); 484 485 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics) 486 { 487 struct nlattr *mx; 488 int i, valid = 0; 489 490 mx = nla_nest_start(skb, RTA_METRICS); 491 if (mx == NULL) 492 return -ENOBUFS; 493 494 for (i = 0; i < RTAX_MAX; i++) { 495 if (metrics[i]) { 496 valid++; 497 NLA_PUT_U32(skb, i+1, metrics[i]); 498 } 499 } 500 501 if (!valid) { 502 nla_nest_cancel(skb, mx); 503 return 0; 504 } 505 506 return nla_nest_end(skb, mx); 507 508 nla_put_failure: 509 nla_nest_cancel(skb, mx); 510 return -EMSGSIZE; 511 } 512 EXPORT_SYMBOL(rtnetlink_put_metrics); 513 514 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id, 515 u32 ts, u32 tsage, long expires, u32 error) 516 { 517 struct rta_cacheinfo ci = { 518 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse), 519 .rta_used = dst->__use, 520 .rta_clntref = atomic_read(&(dst->__refcnt)), 521 .rta_error = error, 522 .rta_id = id, 523 .rta_ts = ts, 524 .rta_tsage = tsage, 525 }; 526 527 if (expires) 528 ci.rta_expires = jiffies_to_clock_t(expires); 529 530 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci); 531 } 532 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo); 533 534 static void set_operstate(struct net_device *dev, unsigned char transition) 535 { 536 unsigned char operstate = dev->operstate; 537 538 switch (transition) { 539 case IF_OPER_UP: 540 if ((operstate == IF_OPER_DORMANT || 541 operstate == IF_OPER_UNKNOWN) && 542 !netif_dormant(dev)) 543 operstate = IF_OPER_UP; 544 break; 545 546 case IF_OPER_DORMANT: 547 if (operstate == IF_OPER_UP || 548 operstate == IF_OPER_UNKNOWN) 549 operstate = IF_OPER_DORMANT; 550 break; 551 } 552 553 if (dev->operstate != operstate) { 554 write_lock_bh(&dev_base_lock); 555 dev->operstate = operstate; 556 write_unlock_bh(&dev_base_lock); 557 netdev_state_change(dev); 558 } 559 } 560 561 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev, 562 const struct ifinfomsg *ifm) 563 { 564 unsigned int flags = ifm->ifi_flags; 565 566 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */ 567 if (ifm->ifi_change) 568 flags = (flags & ifm->ifi_change) | 569 (dev->flags & ~ifm->ifi_change); 570 571 return flags; 572 } 573 574 static void copy_rtnl_link_stats(struct rtnl_link_stats *a, 575 const struct rtnl_link_stats64 *b) 576 { 577 a->rx_packets = b->rx_packets; 578 a->tx_packets = b->tx_packets; 579 a->rx_bytes = b->rx_bytes; 580 a->tx_bytes = b->tx_bytes; 581 a->rx_errors = b->rx_errors; 582 a->tx_errors = b->tx_errors; 583 a->rx_dropped = b->rx_dropped; 584 a->tx_dropped = b->tx_dropped; 585 586 a->multicast = b->multicast; 587 a->collisions = b->collisions; 588 589 a->rx_length_errors = b->rx_length_errors; 590 a->rx_over_errors = b->rx_over_errors; 591 a->rx_crc_errors = b->rx_crc_errors; 592 a->rx_frame_errors = b->rx_frame_errors; 593 a->rx_fifo_errors = b->rx_fifo_errors; 594 a->rx_missed_errors = b->rx_missed_errors; 595 596 a->tx_aborted_errors = b->tx_aborted_errors; 597 a->tx_carrier_errors = b->tx_carrier_errors; 598 a->tx_fifo_errors = b->tx_fifo_errors; 599 a->tx_heartbeat_errors = b->tx_heartbeat_errors; 600 a->tx_window_errors = b->tx_window_errors; 601 602 a->rx_compressed = b->rx_compressed; 603 a->tx_compressed = b->tx_compressed; 604 } 605 606 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b) 607 { 608 memcpy(v, b, sizeof(*b)); 609 } 610 611 /* All VF info */ 612 static inline int rtnl_vfinfo_size(const struct net_device *dev) 613 { 614 if (dev->dev.parent && dev_is_pci(dev->dev.parent)) { 615 616 int num_vfs = dev_num_vf(dev->dev.parent); 617 size_t size = nla_total_size(sizeof(struct nlattr)); 618 size += nla_total_size(num_vfs * sizeof(struct nlattr)); 619 size += num_vfs * 620 (nla_total_size(sizeof(struct ifla_vf_mac)) + 621 nla_total_size(sizeof(struct ifla_vf_vlan)) + 622 nla_total_size(sizeof(struct ifla_vf_tx_rate))); 623 return size; 624 } else 625 return 0; 626 } 627 628 static size_t rtnl_port_size(const struct net_device *dev) 629 { 630 size_t port_size = nla_total_size(4) /* PORT_VF */ 631 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */ 632 + nla_total_size(sizeof(struct ifla_port_vsi)) 633 /* PORT_VSI_TYPE */ 634 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */ 635 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */ 636 + nla_total_size(1) /* PROT_VDP_REQUEST */ 637 + nla_total_size(2); /* PORT_VDP_RESPONSE */ 638 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr)); 639 size_t vf_port_size = nla_total_size(sizeof(struct nlattr)) 640 + port_size; 641 size_t port_self_size = nla_total_size(sizeof(struct nlattr)) 642 + port_size; 643 644 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent) 645 return 0; 646 if (dev_num_vf(dev->dev.parent)) 647 return port_self_size + vf_ports_size + 648 vf_port_size * dev_num_vf(dev->dev.parent); 649 else 650 return port_self_size; 651 } 652 653 static noinline size_t if_nlmsg_size(const struct net_device *dev) 654 { 655 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 656 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 657 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */ 658 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */ 659 + nla_total_size(sizeof(struct rtnl_link_ifmap)) 660 + nla_total_size(sizeof(struct rtnl_link_stats)) 661 + nla_total_size(sizeof(struct rtnl_link_stats64)) 662 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 663 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */ 664 + nla_total_size(4) /* IFLA_TXQLEN */ 665 + nla_total_size(4) /* IFLA_WEIGHT */ 666 + nla_total_size(4) /* IFLA_MTU */ 667 + nla_total_size(4) /* IFLA_LINK */ 668 + nla_total_size(4) /* IFLA_MASTER */ 669 + nla_total_size(1) /* IFLA_OPERSTATE */ 670 + nla_total_size(1) /* IFLA_LINKMODE */ 671 + nla_total_size(4) /* IFLA_NUM_VF */ 672 + rtnl_vfinfo_size(dev) /* IFLA_VFINFO_LIST */ 673 + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */ 674 + rtnl_link_get_size(dev); /* IFLA_LINKINFO */ 675 } 676 677 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev) 678 { 679 struct nlattr *vf_ports; 680 struct nlattr *vf_port; 681 int vf; 682 int err; 683 684 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS); 685 if (!vf_ports) 686 return -EMSGSIZE; 687 688 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) { 689 vf_port = nla_nest_start(skb, IFLA_VF_PORT); 690 if (!vf_port) 691 goto nla_put_failure; 692 NLA_PUT_U32(skb, IFLA_PORT_VF, vf); 693 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb); 694 if (err == -EMSGSIZE) 695 goto nla_put_failure; 696 if (err) { 697 nla_nest_cancel(skb, vf_port); 698 continue; 699 } 700 nla_nest_end(skb, vf_port); 701 } 702 703 nla_nest_end(skb, vf_ports); 704 705 return 0; 706 707 nla_put_failure: 708 nla_nest_cancel(skb, vf_ports); 709 return -EMSGSIZE; 710 } 711 712 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev) 713 { 714 struct nlattr *port_self; 715 int err; 716 717 port_self = nla_nest_start(skb, IFLA_PORT_SELF); 718 if (!port_self) 719 return -EMSGSIZE; 720 721 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb); 722 if (err) { 723 nla_nest_cancel(skb, port_self); 724 return (err == -EMSGSIZE) ? err : 0; 725 } 726 727 nla_nest_end(skb, port_self); 728 729 return 0; 730 } 731 732 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev) 733 { 734 int err; 735 736 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent) 737 return 0; 738 739 err = rtnl_port_self_fill(skb, dev); 740 if (err) 741 return err; 742 743 if (dev_num_vf(dev->dev.parent)) { 744 err = rtnl_vf_ports_fill(skb, dev); 745 if (err) 746 return err; 747 } 748 749 return 0; 750 } 751 752 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev, 753 int type, u32 pid, u32 seq, u32 change, 754 unsigned int flags) 755 { 756 struct ifinfomsg *ifm; 757 struct nlmsghdr *nlh; 758 struct rtnl_link_stats64 temp; 759 const struct rtnl_link_stats64 *stats; 760 struct nlattr *attr; 761 762 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags); 763 if (nlh == NULL) 764 return -EMSGSIZE; 765 766 ifm = nlmsg_data(nlh); 767 ifm->ifi_family = AF_UNSPEC; 768 ifm->__ifi_pad = 0; 769 ifm->ifi_type = dev->type; 770 ifm->ifi_index = dev->ifindex; 771 ifm->ifi_flags = dev_get_flags(dev); 772 ifm->ifi_change = change; 773 774 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name); 775 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len); 776 NLA_PUT_U8(skb, IFLA_OPERSTATE, 777 netif_running(dev) ? dev->operstate : IF_OPER_DOWN); 778 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode); 779 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu); 780 781 if (dev->ifindex != dev->iflink) 782 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink); 783 784 if (dev->master) 785 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex); 786 787 if (dev->qdisc) 788 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id); 789 790 if (dev->ifalias) 791 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias); 792 793 if (1) { 794 struct rtnl_link_ifmap map = { 795 .mem_start = dev->mem_start, 796 .mem_end = dev->mem_end, 797 .base_addr = dev->base_addr, 798 .irq = dev->irq, 799 .dma = dev->dma, 800 .port = dev->if_port, 801 }; 802 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map); 803 } 804 805 if (dev->addr_len) { 806 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr); 807 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast); 808 } 809 810 attr = nla_reserve(skb, IFLA_STATS, 811 sizeof(struct rtnl_link_stats)); 812 if (attr == NULL) 813 goto nla_put_failure; 814 815 stats = dev_get_stats(dev, &temp); 816 copy_rtnl_link_stats(nla_data(attr), stats); 817 818 attr = nla_reserve(skb, IFLA_STATS64, 819 sizeof(struct rtnl_link_stats64)); 820 if (attr == NULL) 821 goto nla_put_failure; 822 copy_rtnl_link_stats64(nla_data(attr), stats); 823 824 if (dev->dev.parent) 825 NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)); 826 827 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent) { 828 int i; 829 830 struct nlattr *vfinfo, *vf; 831 int num_vfs = dev_num_vf(dev->dev.parent); 832 833 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST); 834 if (!vfinfo) 835 goto nla_put_failure; 836 for (i = 0; i < num_vfs; i++) { 837 struct ifla_vf_info ivi; 838 struct ifla_vf_mac vf_mac; 839 struct ifla_vf_vlan vf_vlan; 840 struct ifla_vf_tx_rate vf_tx_rate; 841 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi)) 842 break; 843 vf_mac.vf = vf_vlan.vf = vf_tx_rate.vf = ivi.vf; 844 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac)); 845 vf_vlan.vlan = ivi.vlan; 846 vf_vlan.qos = ivi.qos; 847 vf_tx_rate.rate = ivi.tx_rate; 848 vf = nla_nest_start(skb, IFLA_VF_INFO); 849 if (!vf) { 850 nla_nest_cancel(skb, vfinfo); 851 goto nla_put_failure; 852 } 853 NLA_PUT(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac); 854 NLA_PUT(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan); 855 NLA_PUT(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate), &vf_tx_rate); 856 nla_nest_end(skb, vf); 857 } 858 nla_nest_end(skb, vfinfo); 859 } 860 861 if (rtnl_port_fill(skb, dev)) 862 goto nla_put_failure; 863 864 if (dev->rtnl_link_ops) { 865 if (rtnl_link_fill(skb, dev) < 0) 866 goto nla_put_failure; 867 } 868 869 return nlmsg_end(skb, nlh); 870 871 nla_put_failure: 872 nlmsg_cancel(skb, nlh); 873 return -EMSGSIZE; 874 } 875 876 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb) 877 { 878 struct net *net = sock_net(skb->sk); 879 int h, s_h; 880 int idx = 0, s_idx; 881 struct net_device *dev; 882 struct hlist_head *head; 883 struct hlist_node *node; 884 885 s_h = cb->args[0]; 886 s_idx = cb->args[1]; 887 888 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 889 idx = 0; 890 head = &net->dev_index_head[h]; 891 hlist_for_each_entry(dev, node, head, index_hlist) { 892 if (idx < s_idx) 893 goto cont; 894 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK, 895 NETLINK_CB(cb->skb).pid, 896 cb->nlh->nlmsg_seq, 0, 897 NLM_F_MULTI) <= 0) 898 goto out; 899 cont: 900 idx++; 901 } 902 } 903 out: 904 cb->args[1] = idx; 905 cb->args[0] = h; 906 907 return skb->len; 908 } 909 910 const struct nla_policy ifla_policy[IFLA_MAX+1] = { 911 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 }, 912 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 913 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 914 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) }, 915 [IFLA_MTU] = { .type = NLA_U32 }, 916 [IFLA_LINK] = { .type = NLA_U32 }, 917 [IFLA_TXQLEN] = { .type = NLA_U32 }, 918 [IFLA_WEIGHT] = { .type = NLA_U32 }, 919 [IFLA_OPERSTATE] = { .type = NLA_U8 }, 920 [IFLA_LINKMODE] = { .type = NLA_U8 }, 921 [IFLA_LINKINFO] = { .type = NLA_NESTED }, 922 [IFLA_NET_NS_PID] = { .type = NLA_U32 }, 923 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 }, 924 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED }, 925 [IFLA_VF_PORTS] = { .type = NLA_NESTED }, 926 [IFLA_PORT_SELF] = { .type = NLA_NESTED }, 927 }; 928 EXPORT_SYMBOL(ifla_policy); 929 930 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = { 931 [IFLA_INFO_KIND] = { .type = NLA_STRING }, 932 [IFLA_INFO_DATA] = { .type = NLA_NESTED }, 933 }; 934 935 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = { 936 [IFLA_VF_INFO] = { .type = NLA_NESTED }, 937 }; 938 939 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = { 940 [IFLA_VF_MAC] = { .type = NLA_BINARY, 941 .len = sizeof(struct ifla_vf_mac) }, 942 [IFLA_VF_VLAN] = { .type = NLA_BINARY, 943 .len = sizeof(struct ifla_vf_vlan) }, 944 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY, 945 .len = sizeof(struct ifla_vf_tx_rate) }, 946 }; 947 948 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = { 949 [IFLA_PORT_VF] = { .type = NLA_U32 }, 950 [IFLA_PORT_PROFILE] = { .type = NLA_STRING, 951 .len = PORT_PROFILE_MAX }, 952 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY, 953 .len = sizeof(struct ifla_port_vsi)}, 954 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY, 955 .len = PORT_UUID_MAX }, 956 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING, 957 .len = PORT_UUID_MAX }, 958 [IFLA_PORT_REQUEST] = { .type = NLA_U8, }, 959 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, }, 960 }; 961 962 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[]) 963 { 964 struct net *net; 965 /* Examine the link attributes and figure out which 966 * network namespace we are talking about. 967 */ 968 if (tb[IFLA_NET_NS_PID]) 969 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID])); 970 else 971 net = get_net(src_net); 972 return net; 973 } 974 EXPORT_SYMBOL(rtnl_link_get_net); 975 976 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[]) 977 { 978 if (dev) { 979 if (tb[IFLA_ADDRESS] && 980 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len) 981 return -EINVAL; 982 983 if (tb[IFLA_BROADCAST] && 984 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len) 985 return -EINVAL; 986 } 987 988 return 0; 989 } 990 991 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr) 992 { 993 int rem, err = -EINVAL; 994 struct nlattr *vf; 995 const struct net_device_ops *ops = dev->netdev_ops; 996 997 nla_for_each_nested(vf, attr, rem) { 998 switch (nla_type(vf)) { 999 case IFLA_VF_MAC: { 1000 struct ifla_vf_mac *ivm; 1001 ivm = nla_data(vf); 1002 err = -EOPNOTSUPP; 1003 if (ops->ndo_set_vf_mac) 1004 err = ops->ndo_set_vf_mac(dev, ivm->vf, 1005 ivm->mac); 1006 break; 1007 } 1008 case IFLA_VF_VLAN: { 1009 struct ifla_vf_vlan *ivv; 1010 ivv = nla_data(vf); 1011 err = -EOPNOTSUPP; 1012 if (ops->ndo_set_vf_vlan) 1013 err = ops->ndo_set_vf_vlan(dev, ivv->vf, 1014 ivv->vlan, 1015 ivv->qos); 1016 break; 1017 } 1018 case IFLA_VF_TX_RATE: { 1019 struct ifla_vf_tx_rate *ivt; 1020 ivt = nla_data(vf); 1021 err = -EOPNOTSUPP; 1022 if (ops->ndo_set_vf_tx_rate) 1023 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf, 1024 ivt->rate); 1025 break; 1026 } 1027 default: 1028 err = -EINVAL; 1029 break; 1030 } 1031 if (err) 1032 break; 1033 } 1034 return err; 1035 } 1036 1037 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm, 1038 struct nlattr **tb, char *ifname, int modified) 1039 { 1040 const struct net_device_ops *ops = dev->netdev_ops; 1041 int send_addr_notify = 0; 1042 int err; 1043 1044 if (tb[IFLA_NET_NS_PID]) { 1045 struct net *net = rtnl_link_get_net(dev_net(dev), tb); 1046 if (IS_ERR(net)) { 1047 err = PTR_ERR(net); 1048 goto errout; 1049 } 1050 err = dev_change_net_namespace(dev, net, ifname); 1051 put_net(net); 1052 if (err) 1053 goto errout; 1054 modified = 1; 1055 } 1056 1057 if (tb[IFLA_MAP]) { 1058 struct rtnl_link_ifmap *u_map; 1059 struct ifmap k_map; 1060 1061 if (!ops->ndo_set_config) { 1062 err = -EOPNOTSUPP; 1063 goto errout; 1064 } 1065 1066 if (!netif_device_present(dev)) { 1067 err = -ENODEV; 1068 goto errout; 1069 } 1070 1071 u_map = nla_data(tb[IFLA_MAP]); 1072 k_map.mem_start = (unsigned long) u_map->mem_start; 1073 k_map.mem_end = (unsigned long) u_map->mem_end; 1074 k_map.base_addr = (unsigned short) u_map->base_addr; 1075 k_map.irq = (unsigned char) u_map->irq; 1076 k_map.dma = (unsigned char) u_map->dma; 1077 k_map.port = (unsigned char) u_map->port; 1078 1079 err = ops->ndo_set_config(dev, &k_map); 1080 if (err < 0) 1081 goto errout; 1082 1083 modified = 1; 1084 } 1085 1086 if (tb[IFLA_ADDRESS]) { 1087 struct sockaddr *sa; 1088 int len; 1089 1090 if (!ops->ndo_set_mac_address) { 1091 err = -EOPNOTSUPP; 1092 goto errout; 1093 } 1094 1095 if (!netif_device_present(dev)) { 1096 err = -ENODEV; 1097 goto errout; 1098 } 1099 1100 len = sizeof(sa_family_t) + dev->addr_len; 1101 sa = kmalloc(len, GFP_KERNEL); 1102 if (!sa) { 1103 err = -ENOMEM; 1104 goto errout; 1105 } 1106 sa->sa_family = dev->type; 1107 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]), 1108 dev->addr_len); 1109 err = ops->ndo_set_mac_address(dev, sa); 1110 kfree(sa); 1111 if (err) 1112 goto errout; 1113 send_addr_notify = 1; 1114 modified = 1; 1115 } 1116 1117 if (tb[IFLA_MTU]) { 1118 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 1119 if (err < 0) 1120 goto errout; 1121 modified = 1; 1122 } 1123 1124 /* 1125 * Interface selected by interface index but interface 1126 * name provided implies that a name change has been 1127 * requested. 1128 */ 1129 if (ifm->ifi_index > 0 && ifname[0]) { 1130 err = dev_change_name(dev, ifname); 1131 if (err < 0) 1132 goto errout; 1133 modified = 1; 1134 } 1135 1136 if (tb[IFLA_IFALIAS]) { 1137 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]), 1138 nla_len(tb[IFLA_IFALIAS])); 1139 if (err < 0) 1140 goto errout; 1141 modified = 1; 1142 } 1143 1144 if (tb[IFLA_BROADCAST]) { 1145 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len); 1146 send_addr_notify = 1; 1147 } 1148 1149 if (ifm->ifi_flags || ifm->ifi_change) { 1150 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm)); 1151 if (err < 0) 1152 goto errout; 1153 } 1154 1155 if (tb[IFLA_TXQLEN]) 1156 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]); 1157 1158 if (tb[IFLA_OPERSTATE]) 1159 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 1160 1161 if (tb[IFLA_LINKMODE]) { 1162 write_lock_bh(&dev_base_lock); 1163 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]); 1164 write_unlock_bh(&dev_base_lock); 1165 } 1166 1167 if (tb[IFLA_VFINFO_LIST]) { 1168 struct nlattr *attr; 1169 int rem; 1170 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) { 1171 if (nla_type(attr) != IFLA_VF_INFO) { 1172 err = -EINVAL; 1173 goto errout; 1174 } 1175 err = do_setvfinfo(dev, attr); 1176 if (err < 0) 1177 goto errout; 1178 modified = 1; 1179 } 1180 } 1181 err = 0; 1182 1183 if (tb[IFLA_VF_PORTS]) { 1184 struct nlattr *port[IFLA_PORT_MAX+1]; 1185 struct nlattr *attr; 1186 int vf; 1187 int rem; 1188 1189 err = -EOPNOTSUPP; 1190 if (!ops->ndo_set_vf_port) 1191 goto errout; 1192 1193 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) { 1194 if (nla_type(attr) != IFLA_VF_PORT) 1195 continue; 1196 err = nla_parse_nested(port, IFLA_PORT_MAX, 1197 attr, ifla_port_policy); 1198 if (err < 0) 1199 goto errout; 1200 if (!port[IFLA_PORT_VF]) { 1201 err = -EOPNOTSUPP; 1202 goto errout; 1203 } 1204 vf = nla_get_u32(port[IFLA_PORT_VF]); 1205 err = ops->ndo_set_vf_port(dev, vf, port); 1206 if (err < 0) 1207 goto errout; 1208 modified = 1; 1209 } 1210 } 1211 err = 0; 1212 1213 if (tb[IFLA_PORT_SELF]) { 1214 struct nlattr *port[IFLA_PORT_MAX+1]; 1215 1216 err = nla_parse_nested(port, IFLA_PORT_MAX, 1217 tb[IFLA_PORT_SELF], ifla_port_policy); 1218 if (err < 0) 1219 goto errout; 1220 1221 err = -EOPNOTSUPP; 1222 if (ops->ndo_set_vf_port) 1223 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port); 1224 if (err < 0) 1225 goto errout; 1226 modified = 1; 1227 } 1228 err = 0; 1229 1230 errout: 1231 if (err < 0 && modified && net_ratelimit()) 1232 printk(KERN_WARNING "A link change request failed with " 1233 "some changes comitted already. Interface %s may " 1234 "have been left with an inconsistent configuration, " 1235 "please check.\n", dev->name); 1236 1237 if (send_addr_notify) 1238 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); 1239 return err; 1240 } 1241 1242 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) 1243 { 1244 struct net *net = sock_net(skb->sk); 1245 struct ifinfomsg *ifm; 1246 struct net_device *dev; 1247 int err; 1248 struct nlattr *tb[IFLA_MAX+1]; 1249 char ifname[IFNAMSIZ]; 1250 1251 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy); 1252 if (err < 0) 1253 goto errout; 1254 1255 if (tb[IFLA_IFNAME]) 1256 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 1257 else 1258 ifname[0] = '\0'; 1259 1260 err = -EINVAL; 1261 ifm = nlmsg_data(nlh); 1262 if (ifm->ifi_index > 0) 1263 dev = __dev_get_by_index(net, ifm->ifi_index); 1264 else if (tb[IFLA_IFNAME]) 1265 dev = __dev_get_by_name(net, ifname); 1266 else 1267 goto errout; 1268 1269 if (dev == NULL) { 1270 err = -ENODEV; 1271 goto errout; 1272 } 1273 1274 err = validate_linkmsg(dev, tb); 1275 if (err < 0) 1276 goto errout; 1277 1278 err = do_setlink(dev, ifm, tb, ifname, 0); 1279 errout: 1280 return err; 1281 } 1282 1283 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) 1284 { 1285 struct net *net = sock_net(skb->sk); 1286 const struct rtnl_link_ops *ops; 1287 struct net_device *dev; 1288 struct ifinfomsg *ifm; 1289 char ifname[IFNAMSIZ]; 1290 struct nlattr *tb[IFLA_MAX+1]; 1291 int err; 1292 1293 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy); 1294 if (err < 0) 1295 return err; 1296 1297 if (tb[IFLA_IFNAME]) 1298 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 1299 1300 ifm = nlmsg_data(nlh); 1301 if (ifm->ifi_index > 0) 1302 dev = __dev_get_by_index(net, ifm->ifi_index); 1303 else if (tb[IFLA_IFNAME]) 1304 dev = __dev_get_by_name(net, ifname); 1305 else 1306 return -EINVAL; 1307 1308 if (!dev) 1309 return -ENODEV; 1310 1311 ops = dev->rtnl_link_ops; 1312 if (!ops) 1313 return -EOPNOTSUPP; 1314 1315 ops->dellink(dev, NULL); 1316 return 0; 1317 } 1318 1319 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm) 1320 { 1321 unsigned int old_flags; 1322 int err; 1323 1324 old_flags = dev->flags; 1325 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) { 1326 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm)); 1327 if (err < 0) 1328 return err; 1329 } 1330 1331 dev->rtnl_link_state = RTNL_LINK_INITIALIZED; 1332 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U); 1333 1334 __dev_notify_flags(dev, old_flags); 1335 return 0; 1336 } 1337 EXPORT_SYMBOL(rtnl_configure_link); 1338 1339 struct net_device *rtnl_create_link(struct net *src_net, struct net *net, 1340 char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[]) 1341 { 1342 int err; 1343 struct net_device *dev; 1344 unsigned int num_queues = 1; 1345 unsigned int real_num_queues = 1; 1346 1347 if (ops->get_tx_queues) { 1348 err = ops->get_tx_queues(src_net, tb, &num_queues, 1349 &real_num_queues); 1350 if (err) 1351 goto err; 1352 } 1353 err = -ENOMEM; 1354 dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues); 1355 if (!dev) 1356 goto err; 1357 1358 dev_net_set(dev, net); 1359 dev->rtnl_link_ops = ops; 1360 dev->rtnl_link_state = RTNL_LINK_INITIALIZING; 1361 dev->real_num_tx_queues = real_num_queues; 1362 1363 if (strchr(dev->name, '%')) { 1364 err = dev_alloc_name(dev, dev->name); 1365 if (err < 0) 1366 goto err_free; 1367 } 1368 1369 if (tb[IFLA_MTU]) 1370 dev->mtu = nla_get_u32(tb[IFLA_MTU]); 1371 if (tb[IFLA_ADDRESS]) 1372 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]), 1373 nla_len(tb[IFLA_ADDRESS])); 1374 if (tb[IFLA_BROADCAST]) 1375 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]), 1376 nla_len(tb[IFLA_BROADCAST])); 1377 if (tb[IFLA_TXQLEN]) 1378 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]); 1379 if (tb[IFLA_OPERSTATE]) 1380 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 1381 if (tb[IFLA_LINKMODE]) 1382 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]); 1383 1384 return dev; 1385 1386 err_free: 1387 free_netdev(dev); 1388 err: 1389 return ERR_PTR(err); 1390 } 1391 EXPORT_SYMBOL(rtnl_create_link); 1392 1393 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) 1394 { 1395 struct net *net = sock_net(skb->sk); 1396 const struct rtnl_link_ops *ops; 1397 struct net_device *dev; 1398 struct ifinfomsg *ifm; 1399 char kind[MODULE_NAME_LEN]; 1400 char ifname[IFNAMSIZ]; 1401 struct nlattr *tb[IFLA_MAX+1]; 1402 struct nlattr *linkinfo[IFLA_INFO_MAX+1]; 1403 int err; 1404 1405 #ifdef CONFIG_MODULES 1406 replay: 1407 #endif 1408 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy); 1409 if (err < 0) 1410 return err; 1411 1412 if (tb[IFLA_IFNAME]) 1413 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 1414 else 1415 ifname[0] = '\0'; 1416 1417 ifm = nlmsg_data(nlh); 1418 if (ifm->ifi_index > 0) 1419 dev = __dev_get_by_index(net, ifm->ifi_index); 1420 else if (ifname[0]) 1421 dev = __dev_get_by_name(net, ifname); 1422 else 1423 dev = NULL; 1424 1425 err = validate_linkmsg(dev, tb); 1426 if (err < 0) 1427 return err; 1428 1429 if (tb[IFLA_LINKINFO]) { 1430 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX, 1431 tb[IFLA_LINKINFO], ifla_info_policy); 1432 if (err < 0) 1433 return err; 1434 } else 1435 memset(linkinfo, 0, sizeof(linkinfo)); 1436 1437 if (linkinfo[IFLA_INFO_KIND]) { 1438 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind)); 1439 ops = rtnl_link_ops_get(kind); 1440 } else { 1441 kind[0] = '\0'; 1442 ops = NULL; 1443 } 1444 1445 if (1) { 1446 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL; 1447 struct net *dest_net; 1448 1449 if (ops) { 1450 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) { 1451 err = nla_parse_nested(attr, ops->maxtype, 1452 linkinfo[IFLA_INFO_DATA], 1453 ops->policy); 1454 if (err < 0) 1455 return err; 1456 data = attr; 1457 } 1458 if (ops->validate) { 1459 err = ops->validate(tb, data); 1460 if (err < 0) 1461 return err; 1462 } 1463 } 1464 1465 if (dev) { 1466 int modified = 0; 1467 1468 if (nlh->nlmsg_flags & NLM_F_EXCL) 1469 return -EEXIST; 1470 if (nlh->nlmsg_flags & NLM_F_REPLACE) 1471 return -EOPNOTSUPP; 1472 1473 if (linkinfo[IFLA_INFO_DATA]) { 1474 if (!ops || ops != dev->rtnl_link_ops || 1475 !ops->changelink) 1476 return -EOPNOTSUPP; 1477 1478 err = ops->changelink(dev, tb, data); 1479 if (err < 0) 1480 return err; 1481 modified = 1; 1482 } 1483 1484 return do_setlink(dev, ifm, tb, ifname, modified); 1485 } 1486 1487 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) 1488 return -ENODEV; 1489 1490 if (ifm->ifi_index) 1491 return -EOPNOTSUPP; 1492 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO]) 1493 return -EOPNOTSUPP; 1494 1495 if (!ops) { 1496 #ifdef CONFIG_MODULES 1497 if (kind[0]) { 1498 __rtnl_unlock(); 1499 request_module("rtnl-link-%s", kind); 1500 rtnl_lock(); 1501 ops = rtnl_link_ops_get(kind); 1502 if (ops) 1503 goto replay; 1504 } 1505 #endif 1506 return -EOPNOTSUPP; 1507 } 1508 1509 if (!ifname[0]) 1510 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind); 1511 1512 dest_net = rtnl_link_get_net(net, tb); 1513 dev = rtnl_create_link(net, dest_net, ifname, ops, tb); 1514 1515 if (IS_ERR(dev)) 1516 err = PTR_ERR(dev); 1517 else if (ops->newlink) 1518 err = ops->newlink(net, dev, tb, data); 1519 else 1520 err = register_netdevice(dev); 1521 1522 if (err < 0 && !IS_ERR(dev)) 1523 free_netdev(dev); 1524 if (err < 0) 1525 goto out; 1526 1527 err = rtnl_configure_link(dev, ifm); 1528 if (err < 0) 1529 unregister_netdevice(dev); 1530 out: 1531 put_net(dest_net); 1532 return err; 1533 } 1534 } 1535 1536 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg) 1537 { 1538 struct net *net = sock_net(skb->sk); 1539 struct ifinfomsg *ifm; 1540 char ifname[IFNAMSIZ]; 1541 struct nlattr *tb[IFLA_MAX+1]; 1542 struct net_device *dev = NULL; 1543 struct sk_buff *nskb; 1544 int err; 1545 1546 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy); 1547 if (err < 0) 1548 return err; 1549 1550 if (tb[IFLA_IFNAME]) 1551 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 1552 1553 ifm = nlmsg_data(nlh); 1554 if (ifm->ifi_index > 0) 1555 dev = __dev_get_by_index(net, ifm->ifi_index); 1556 else if (tb[IFLA_IFNAME]) 1557 dev = __dev_get_by_name(net, ifname); 1558 else 1559 return -EINVAL; 1560 1561 if (dev == NULL) 1562 return -ENODEV; 1563 1564 nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL); 1565 if (nskb == NULL) 1566 return -ENOBUFS; 1567 1568 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid, 1569 nlh->nlmsg_seq, 0, 0); 1570 if (err < 0) { 1571 /* -EMSGSIZE implies BUG in if_nlmsg_size */ 1572 WARN_ON(err == -EMSGSIZE); 1573 kfree_skb(nskb); 1574 } else 1575 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid); 1576 1577 return err; 1578 } 1579 1580 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb) 1581 { 1582 int idx; 1583 int s_idx = cb->family; 1584 1585 if (s_idx == 0) 1586 s_idx = 1; 1587 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) { 1588 int type = cb->nlh->nlmsg_type-RTM_BASE; 1589 if (idx < s_idx || idx == PF_PACKET) 1590 continue; 1591 if (rtnl_msg_handlers[idx] == NULL || 1592 rtnl_msg_handlers[idx][type].dumpit == NULL) 1593 continue; 1594 if (idx > s_idx) 1595 memset(&cb->args[0], 0, sizeof(cb->args)); 1596 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb)) 1597 break; 1598 } 1599 cb->family = idx; 1600 1601 return skb->len; 1602 } 1603 1604 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change) 1605 { 1606 struct net *net = dev_net(dev); 1607 struct sk_buff *skb; 1608 int err = -ENOBUFS; 1609 1610 skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL); 1611 if (skb == NULL) 1612 goto errout; 1613 1614 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0); 1615 if (err < 0) { 1616 /* -EMSGSIZE implies BUG in if_nlmsg_size() */ 1617 WARN_ON(err == -EMSGSIZE); 1618 kfree_skb(skb); 1619 goto errout; 1620 } 1621 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL); 1622 return; 1623 errout: 1624 if (err < 0) 1625 rtnl_set_sk_err(net, RTNLGRP_LINK, err); 1626 } 1627 1628 /* Protected by RTNL sempahore. */ 1629 static struct rtattr **rta_buf; 1630 static int rtattr_max; 1631 1632 /* Process one rtnetlink message. */ 1633 1634 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh) 1635 { 1636 struct net *net = sock_net(skb->sk); 1637 rtnl_doit_func doit; 1638 int sz_idx, kind; 1639 int min_len; 1640 int family; 1641 int type; 1642 int err; 1643 1644 type = nlh->nlmsg_type; 1645 if (type > RTM_MAX) 1646 return -EOPNOTSUPP; 1647 1648 type -= RTM_BASE; 1649 1650 /* All the messages must have at least 1 byte length */ 1651 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg))) 1652 return 0; 1653 1654 family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family; 1655 sz_idx = type>>2; 1656 kind = type&3; 1657 1658 if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN)) 1659 return -EPERM; 1660 1661 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) { 1662 struct sock *rtnl; 1663 rtnl_dumpit_func dumpit; 1664 1665 dumpit = rtnl_get_dumpit(family, type); 1666 if (dumpit == NULL) 1667 return -EOPNOTSUPP; 1668 1669 __rtnl_unlock(); 1670 rtnl = net->rtnl; 1671 err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL); 1672 rtnl_lock(); 1673 return err; 1674 } 1675 1676 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *))); 1677 1678 min_len = rtm_min[sz_idx]; 1679 if (nlh->nlmsg_len < min_len) 1680 return -EINVAL; 1681 1682 if (nlh->nlmsg_len > min_len) { 1683 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len); 1684 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len); 1685 1686 while (RTA_OK(attr, attrlen)) { 1687 unsigned flavor = attr->rta_type; 1688 if (flavor) { 1689 if (flavor > rta_max[sz_idx]) 1690 return -EINVAL; 1691 rta_buf[flavor-1] = attr; 1692 } 1693 attr = RTA_NEXT(attr, attrlen); 1694 } 1695 } 1696 1697 doit = rtnl_get_doit(family, type); 1698 if (doit == NULL) 1699 return -EOPNOTSUPP; 1700 1701 return doit(skb, nlh, (void *)&rta_buf[0]); 1702 } 1703 1704 static void rtnetlink_rcv(struct sk_buff *skb) 1705 { 1706 rtnl_lock(); 1707 netlink_rcv_skb(skb, &rtnetlink_rcv_msg); 1708 rtnl_unlock(); 1709 } 1710 1711 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr) 1712 { 1713 struct net_device *dev = ptr; 1714 1715 switch (event) { 1716 case NETDEV_UP: 1717 case NETDEV_DOWN: 1718 case NETDEV_PRE_UP: 1719 case NETDEV_POST_INIT: 1720 case NETDEV_REGISTER: 1721 case NETDEV_CHANGE: 1722 case NETDEV_PRE_TYPE_CHANGE: 1723 case NETDEV_GOING_DOWN: 1724 case NETDEV_UNREGISTER: 1725 case NETDEV_UNREGISTER_BATCH: 1726 break; 1727 default: 1728 rtmsg_ifinfo(RTM_NEWLINK, dev, 0); 1729 break; 1730 } 1731 return NOTIFY_DONE; 1732 } 1733 1734 static struct notifier_block rtnetlink_dev_notifier = { 1735 .notifier_call = rtnetlink_event, 1736 }; 1737 1738 1739 static int __net_init rtnetlink_net_init(struct net *net) 1740 { 1741 struct sock *sk; 1742 sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX, 1743 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE); 1744 if (!sk) 1745 return -ENOMEM; 1746 net->rtnl = sk; 1747 return 0; 1748 } 1749 1750 static void __net_exit rtnetlink_net_exit(struct net *net) 1751 { 1752 netlink_kernel_release(net->rtnl); 1753 net->rtnl = NULL; 1754 } 1755 1756 static struct pernet_operations rtnetlink_net_ops = { 1757 .init = rtnetlink_net_init, 1758 .exit = rtnetlink_net_exit, 1759 }; 1760 1761 void __init rtnetlink_init(void) 1762 { 1763 int i; 1764 1765 rtattr_max = 0; 1766 for (i = 0; i < ARRAY_SIZE(rta_max); i++) 1767 if (rta_max[i] > rtattr_max) 1768 rtattr_max = rta_max[i]; 1769 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL); 1770 if (!rta_buf) 1771 panic("rtnetlink_init: cannot allocate rta_buf\n"); 1772 1773 if (register_pernet_subsys(&rtnetlink_net_ops)) 1774 panic("rtnetlink_init: cannot initialize rtnetlink\n"); 1775 1776 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV); 1777 register_netdevice_notifier(&rtnetlink_dev_notifier); 1778 1779 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo); 1780 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL); 1781 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL); 1782 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL); 1783 1784 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all); 1785 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all); 1786 } 1787 1788