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 void rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops) 303 { 304 rtnl_lock(); 305 __rtnl_kill_links(net, ops); 306 rtnl_unlock(); 307 } 308 EXPORT_SYMBOL_GPL(rtnl_kill_links); 309 310 /** 311 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink. 312 * @ops: struct rtnl_link_ops * to unregister 313 * 314 * The caller must hold the rtnl_mutex. 315 */ 316 void __rtnl_link_unregister(struct rtnl_link_ops *ops) 317 { 318 struct net *net; 319 320 for_each_net(net) { 321 __rtnl_kill_links(net, ops); 322 } 323 list_del(&ops->list); 324 } 325 EXPORT_SYMBOL_GPL(__rtnl_link_unregister); 326 327 /** 328 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink. 329 * @ops: struct rtnl_link_ops * to unregister 330 */ 331 void rtnl_link_unregister(struct rtnl_link_ops *ops) 332 { 333 rtnl_lock(); 334 __rtnl_link_unregister(ops); 335 rtnl_unlock(); 336 } 337 EXPORT_SYMBOL_GPL(rtnl_link_unregister); 338 339 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind) 340 { 341 const struct rtnl_link_ops *ops; 342 343 list_for_each_entry(ops, &link_ops, list) { 344 if (!strcmp(ops->kind, kind)) 345 return ops; 346 } 347 return NULL; 348 } 349 350 static size_t rtnl_link_get_size(const struct net_device *dev) 351 { 352 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 353 size_t size; 354 355 if (!ops) 356 return 0; 357 358 size = nlmsg_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */ 359 nlmsg_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */ 360 361 if (ops->get_size) 362 /* IFLA_INFO_DATA + nested data */ 363 size += nlmsg_total_size(sizeof(struct nlattr)) + 364 ops->get_size(dev); 365 366 if (ops->get_xstats_size) 367 size += ops->get_xstats_size(dev); /* IFLA_INFO_XSTATS */ 368 369 return size; 370 } 371 372 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev) 373 { 374 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 375 struct nlattr *linkinfo, *data; 376 int err = -EMSGSIZE; 377 378 linkinfo = nla_nest_start(skb, IFLA_LINKINFO); 379 if (linkinfo == NULL) 380 goto out; 381 382 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0) 383 goto err_cancel_link; 384 if (ops->fill_xstats) { 385 err = ops->fill_xstats(skb, dev); 386 if (err < 0) 387 goto err_cancel_link; 388 } 389 if (ops->fill_info) { 390 data = nla_nest_start(skb, IFLA_INFO_DATA); 391 if (data == NULL) 392 goto err_cancel_link; 393 err = ops->fill_info(skb, dev); 394 if (err < 0) 395 goto err_cancel_data; 396 nla_nest_end(skb, data); 397 } 398 399 nla_nest_end(skb, linkinfo); 400 return 0; 401 402 err_cancel_data: 403 nla_nest_cancel(skb, data); 404 err_cancel_link: 405 nla_nest_cancel(skb, linkinfo); 406 out: 407 return err; 408 } 409 410 static const int rtm_min[RTM_NR_FAMILIES] = 411 { 412 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)), 413 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)), 414 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)), 415 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)), 416 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)), 417 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)), 418 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)), 419 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)), 420 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)), 421 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)), 422 }; 423 424 static const int rta_max[RTM_NR_FAMILIES] = 425 { 426 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX, 427 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX, 428 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX, 429 [RTM_FAM(RTM_NEWRULE)] = FRA_MAX, 430 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX, 431 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX, 432 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX, 433 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX, 434 }; 435 436 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data) 437 { 438 struct rtattr *rta; 439 int size = RTA_LENGTH(attrlen); 440 441 rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size)); 442 rta->rta_type = attrtype; 443 rta->rta_len = size; 444 memcpy(RTA_DATA(rta), data, attrlen); 445 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size); 446 } 447 EXPORT_SYMBOL(__rta_fill); 448 449 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo) 450 { 451 struct sock *rtnl = net->rtnl; 452 int err = 0; 453 454 NETLINK_CB(skb).dst_group = group; 455 if (echo) 456 atomic_inc(&skb->users); 457 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL); 458 if (echo) 459 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT); 460 return err; 461 } 462 463 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid) 464 { 465 struct sock *rtnl = net->rtnl; 466 467 return nlmsg_unicast(rtnl, skb, pid); 468 } 469 EXPORT_SYMBOL(rtnl_unicast); 470 471 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group, 472 struct nlmsghdr *nlh, gfp_t flags) 473 { 474 struct sock *rtnl = net->rtnl; 475 int report = 0; 476 477 if (nlh) 478 report = nlmsg_report(nlh); 479 480 nlmsg_notify(rtnl, skb, pid, group, report, flags); 481 } 482 EXPORT_SYMBOL(rtnl_notify); 483 484 void rtnl_set_sk_err(struct net *net, u32 group, int error) 485 { 486 struct sock *rtnl = net->rtnl; 487 488 netlink_set_err(rtnl, 0, group, error); 489 } 490 EXPORT_SYMBOL(rtnl_set_sk_err); 491 492 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics) 493 { 494 struct nlattr *mx; 495 int i, valid = 0; 496 497 mx = nla_nest_start(skb, RTA_METRICS); 498 if (mx == NULL) 499 return -ENOBUFS; 500 501 for (i = 0; i < RTAX_MAX; i++) { 502 if (metrics[i]) { 503 valid++; 504 NLA_PUT_U32(skb, i+1, metrics[i]); 505 } 506 } 507 508 if (!valid) { 509 nla_nest_cancel(skb, mx); 510 return 0; 511 } 512 513 return nla_nest_end(skb, mx); 514 515 nla_put_failure: 516 nla_nest_cancel(skb, mx); 517 return -EMSGSIZE; 518 } 519 EXPORT_SYMBOL(rtnetlink_put_metrics); 520 521 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id, 522 u32 ts, u32 tsage, long expires, u32 error) 523 { 524 struct rta_cacheinfo ci = { 525 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse), 526 .rta_used = dst->__use, 527 .rta_clntref = atomic_read(&(dst->__refcnt)), 528 .rta_error = error, 529 .rta_id = id, 530 .rta_ts = ts, 531 .rta_tsage = tsage, 532 }; 533 534 if (expires) 535 ci.rta_expires = jiffies_to_clock_t(expires); 536 537 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci); 538 } 539 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo); 540 541 static void set_operstate(struct net_device *dev, unsigned char transition) 542 { 543 unsigned char operstate = dev->operstate; 544 545 switch (transition) { 546 case IF_OPER_UP: 547 if ((operstate == IF_OPER_DORMANT || 548 operstate == IF_OPER_UNKNOWN) && 549 !netif_dormant(dev)) 550 operstate = IF_OPER_UP; 551 break; 552 553 case IF_OPER_DORMANT: 554 if (operstate == IF_OPER_UP || 555 operstate == IF_OPER_UNKNOWN) 556 operstate = IF_OPER_DORMANT; 557 break; 558 } 559 560 if (dev->operstate != operstate) { 561 write_lock_bh(&dev_base_lock); 562 dev->operstate = operstate; 563 write_unlock_bh(&dev_base_lock); 564 netdev_state_change(dev); 565 } 566 } 567 568 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev, 569 const struct ifinfomsg *ifm) 570 { 571 unsigned int flags = ifm->ifi_flags; 572 573 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */ 574 if (ifm->ifi_change) 575 flags = (flags & ifm->ifi_change) | 576 (dev->flags & ~ifm->ifi_change); 577 578 return flags; 579 } 580 581 static void copy_rtnl_link_stats(struct rtnl_link_stats *a, 582 const struct net_device_stats *b) 583 { 584 a->rx_packets = b->rx_packets; 585 a->tx_packets = b->tx_packets; 586 a->rx_bytes = b->rx_bytes; 587 a->tx_bytes = b->tx_bytes; 588 a->rx_errors = b->rx_errors; 589 a->tx_errors = b->tx_errors; 590 a->rx_dropped = b->rx_dropped; 591 a->tx_dropped = b->tx_dropped; 592 593 a->multicast = b->multicast; 594 a->collisions = b->collisions; 595 596 a->rx_length_errors = b->rx_length_errors; 597 a->rx_over_errors = b->rx_over_errors; 598 a->rx_crc_errors = b->rx_crc_errors; 599 a->rx_frame_errors = b->rx_frame_errors; 600 a->rx_fifo_errors = b->rx_fifo_errors; 601 a->rx_missed_errors = b->rx_missed_errors; 602 603 a->tx_aborted_errors = b->tx_aborted_errors; 604 a->tx_carrier_errors = b->tx_carrier_errors; 605 a->tx_fifo_errors = b->tx_fifo_errors; 606 a->tx_heartbeat_errors = b->tx_heartbeat_errors; 607 a->tx_window_errors = b->tx_window_errors; 608 609 a->rx_compressed = b->rx_compressed; 610 a->tx_compressed = b->tx_compressed; 611 } 612 613 static void copy_rtnl_link_stats64(void *v, const struct net_device_stats *b) 614 { 615 struct rtnl_link_stats64 a; 616 617 a.rx_packets = b->rx_packets; 618 a.tx_packets = b->tx_packets; 619 a.rx_bytes = b->rx_bytes; 620 a.tx_bytes = b->tx_bytes; 621 a.rx_errors = b->rx_errors; 622 a.tx_errors = b->tx_errors; 623 a.rx_dropped = b->rx_dropped; 624 a.tx_dropped = b->tx_dropped; 625 626 a.multicast = b->multicast; 627 a.collisions = b->collisions; 628 629 a.rx_length_errors = b->rx_length_errors; 630 a.rx_over_errors = b->rx_over_errors; 631 a.rx_crc_errors = b->rx_crc_errors; 632 a.rx_frame_errors = b->rx_frame_errors; 633 a.rx_fifo_errors = b->rx_fifo_errors; 634 a.rx_missed_errors = b->rx_missed_errors; 635 636 a.tx_aborted_errors = b->tx_aborted_errors; 637 a.tx_carrier_errors = b->tx_carrier_errors; 638 a.tx_fifo_errors = b->tx_fifo_errors; 639 a.tx_heartbeat_errors = b->tx_heartbeat_errors; 640 a.tx_window_errors = b->tx_window_errors; 641 642 a.rx_compressed = b->rx_compressed; 643 a.tx_compressed = b->tx_compressed; 644 memcpy(v, &a, sizeof(a)); 645 } 646 647 /* All VF info */ 648 static inline int rtnl_vfinfo_size(const struct net_device *dev) 649 { 650 if (dev->dev.parent && dev_is_pci(dev->dev.parent)) { 651 652 int num_vfs = dev_num_vf(dev->dev.parent); 653 size_t size = nlmsg_total_size(sizeof(struct nlattr)); 654 size += nlmsg_total_size(num_vfs * sizeof(struct nlattr)); 655 size += num_vfs * (sizeof(struct ifla_vf_mac) + 656 sizeof(struct ifla_vf_vlan) + 657 sizeof(struct ifla_vf_tx_rate)); 658 return size; 659 } else 660 return 0; 661 } 662 663 static size_t rtnl_port_size(const struct net_device *dev) 664 { 665 size_t port_size = nla_total_size(4) /* PORT_VF */ 666 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */ 667 + nla_total_size(sizeof(struct ifla_port_vsi)) 668 /* PORT_VSI_TYPE */ 669 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */ 670 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */ 671 + nla_total_size(1) /* PROT_VDP_REQUEST */ 672 + nla_total_size(2); /* PORT_VDP_RESPONSE */ 673 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr)); 674 size_t vf_port_size = nla_total_size(sizeof(struct nlattr)) 675 + port_size; 676 size_t port_self_size = nla_total_size(sizeof(struct nlattr)) 677 + port_size; 678 679 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent) 680 return 0; 681 if (dev_num_vf(dev->dev.parent)) 682 return port_self_size + vf_ports_size + 683 vf_port_size * dev_num_vf(dev->dev.parent); 684 else 685 return port_self_size; 686 } 687 688 static inline size_t if_nlmsg_size(const struct net_device *dev) 689 { 690 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 691 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 692 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */ 693 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */ 694 + nla_total_size(sizeof(struct rtnl_link_ifmap)) 695 + nla_total_size(sizeof(struct rtnl_link_stats)) 696 + nla_total_size(sizeof(struct rtnl_link_stats64)) 697 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 698 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */ 699 + nla_total_size(4) /* IFLA_TXQLEN */ 700 + nla_total_size(4) /* IFLA_WEIGHT */ 701 + nla_total_size(4) /* IFLA_MTU */ 702 + nla_total_size(4) /* IFLA_LINK */ 703 + nla_total_size(4) /* IFLA_MASTER */ 704 + nla_total_size(1) /* IFLA_OPERSTATE */ 705 + nla_total_size(1) /* IFLA_LINKMODE */ 706 + nla_total_size(4) /* IFLA_NUM_VF */ 707 + rtnl_vfinfo_size(dev) /* IFLA_VFINFO_LIST */ 708 + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */ 709 + rtnl_link_get_size(dev); /* IFLA_LINKINFO */ 710 } 711 712 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev) 713 { 714 struct nlattr *vf_ports; 715 struct nlattr *vf_port; 716 int vf; 717 int err; 718 719 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS); 720 if (!vf_ports) 721 return -EMSGSIZE; 722 723 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) { 724 vf_port = nla_nest_start(skb, IFLA_VF_PORT); 725 if (!vf_port) { 726 nla_nest_cancel(skb, vf_ports); 727 return -EMSGSIZE; 728 } 729 NLA_PUT_U32(skb, IFLA_PORT_VF, vf); 730 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb); 731 if (err) { 732 nla_put_failure: 733 nla_nest_cancel(skb, vf_port); 734 continue; 735 } 736 nla_nest_end(skb, vf_port); 737 } 738 739 nla_nest_end(skb, vf_ports); 740 741 return 0; 742 } 743 744 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev) 745 { 746 struct nlattr *port_self; 747 int err; 748 749 port_self = nla_nest_start(skb, IFLA_PORT_SELF); 750 if (!port_self) 751 return -EMSGSIZE; 752 753 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb); 754 if (err) { 755 nla_nest_cancel(skb, port_self); 756 return err; 757 } 758 759 nla_nest_end(skb, port_self); 760 761 return 0; 762 } 763 764 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev) 765 { 766 int err; 767 768 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent) 769 return 0; 770 771 err = rtnl_port_self_fill(skb, dev); 772 if (err) 773 return err; 774 775 if (dev_num_vf(dev->dev.parent)) { 776 err = rtnl_vf_ports_fill(skb, dev); 777 if (err) 778 return err; 779 } 780 781 return 0; 782 } 783 784 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev, 785 int type, u32 pid, u32 seq, u32 change, 786 unsigned int flags) 787 { 788 struct ifinfomsg *ifm; 789 struct nlmsghdr *nlh; 790 const struct net_device_stats *stats; 791 struct nlattr *attr; 792 793 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags); 794 if (nlh == NULL) 795 return -EMSGSIZE; 796 797 ifm = nlmsg_data(nlh); 798 ifm->ifi_family = AF_UNSPEC; 799 ifm->__ifi_pad = 0; 800 ifm->ifi_type = dev->type; 801 ifm->ifi_index = dev->ifindex; 802 ifm->ifi_flags = dev_get_flags(dev); 803 ifm->ifi_change = change; 804 805 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name); 806 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len); 807 NLA_PUT_U8(skb, IFLA_OPERSTATE, 808 netif_running(dev) ? dev->operstate : IF_OPER_DOWN); 809 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode); 810 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu); 811 812 if (dev->ifindex != dev->iflink) 813 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink); 814 815 if (dev->master) 816 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex); 817 818 if (dev->qdisc) 819 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id); 820 821 if (dev->ifalias) 822 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias); 823 824 if (1) { 825 struct rtnl_link_ifmap map = { 826 .mem_start = dev->mem_start, 827 .mem_end = dev->mem_end, 828 .base_addr = dev->base_addr, 829 .irq = dev->irq, 830 .dma = dev->dma, 831 .port = dev->if_port, 832 }; 833 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map); 834 } 835 836 if (dev->addr_len) { 837 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr); 838 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast); 839 } 840 841 attr = nla_reserve(skb, IFLA_STATS, 842 sizeof(struct rtnl_link_stats)); 843 if (attr == NULL) 844 goto nla_put_failure; 845 846 stats = dev_get_stats(dev); 847 copy_rtnl_link_stats(nla_data(attr), stats); 848 849 attr = nla_reserve(skb, IFLA_STATS64, 850 sizeof(struct rtnl_link_stats64)); 851 if (attr == NULL) 852 goto nla_put_failure; 853 copy_rtnl_link_stats64(nla_data(attr), stats); 854 855 if (dev->dev.parent) 856 NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)); 857 858 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent) { 859 int i; 860 861 struct nlattr *vfinfo, *vf; 862 int num_vfs = dev_num_vf(dev->dev.parent); 863 864 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST); 865 if (!vfinfo) 866 goto nla_put_failure; 867 for (i = 0; i < num_vfs; i++) { 868 struct ifla_vf_info ivi; 869 struct ifla_vf_mac vf_mac; 870 struct ifla_vf_vlan vf_vlan; 871 struct ifla_vf_tx_rate vf_tx_rate; 872 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi)) 873 break; 874 vf_mac.vf = vf_vlan.vf = vf_tx_rate.vf = ivi.vf; 875 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac)); 876 vf_vlan.vlan = ivi.vlan; 877 vf_vlan.qos = ivi.qos; 878 vf_tx_rate.rate = ivi.tx_rate; 879 vf = nla_nest_start(skb, IFLA_VF_INFO); 880 if (!vf) { 881 nla_nest_cancel(skb, vfinfo); 882 goto nla_put_failure; 883 } 884 NLA_PUT(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac); 885 NLA_PUT(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan); 886 NLA_PUT(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate), &vf_tx_rate); 887 nla_nest_end(skb, vf); 888 } 889 nla_nest_end(skb, vfinfo); 890 } 891 892 if (rtnl_port_fill(skb, dev)) 893 goto nla_put_failure; 894 895 if (dev->rtnl_link_ops) { 896 if (rtnl_link_fill(skb, dev) < 0) 897 goto nla_put_failure; 898 } 899 900 return nlmsg_end(skb, nlh); 901 902 nla_put_failure: 903 nlmsg_cancel(skb, nlh); 904 return -EMSGSIZE; 905 } 906 907 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb) 908 { 909 struct net *net = sock_net(skb->sk); 910 int h, s_h; 911 int idx = 0, s_idx; 912 struct net_device *dev; 913 struct hlist_head *head; 914 struct hlist_node *node; 915 916 s_h = cb->args[0]; 917 s_idx = cb->args[1]; 918 919 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 920 idx = 0; 921 head = &net->dev_index_head[h]; 922 hlist_for_each_entry(dev, node, head, index_hlist) { 923 if (idx < s_idx) 924 goto cont; 925 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK, 926 NETLINK_CB(cb->skb).pid, 927 cb->nlh->nlmsg_seq, 0, 928 NLM_F_MULTI) <= 0) 929 goto out; 930 cont: 931 idx++; 932 } 933 } 934 out: 935 cb->args[1] = idx; 936 cb->args[0] = h; 937 938 return skb->len; 939 } 940 941 const struct nla_policy ifla_policy[IFLA_MAX+1] = { 942 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 }, 943 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 944 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 945 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) }, 946 [IFLA_MTU] = { .type = NLA_U32 }, 947 [IFLA_LINK] = { .type = NLA_U32 }, 948 [IFLA_TXQLEN] = { .type = NLA_U32 }, 949 [IFLA_WEIGHT] = { .type = NLA_U32 }, 950 [IFLA_OPERSTATE] = { .type = NLA_U8 }, 951 [IFLA_LINKMODE] = { .type = NLA_U8 }, 952 [IFLA_LINKINFO] = { .type = NLA_NESTED }, 953 [IFLA_NET_NS_PID] = { .type = NLA_U32 }, 954 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 }, 955 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED }, 956 [IFLA_VF_PORTS] = { .type = NLA_NESTED }, 957 [IFLA_PORT_SELF] = { .type = NLA_NESTED }, 958 }; 959 EXPORT_SYMBOL(ifla_policy); 960 961 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = { 962 [IFLA_INFO_KIND] = { .type = NLA_STRING }, 963 [IFLA_INFO_DATA] = { .type = NLA_NESTED }, 964 }; 965 966 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = { 967 [IFLA_VF_INFO] = { .type = NLA_NESTED }, 968 }; 969 970 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = { 971 [IFLA_VF_MAC] = { .type = NLA_BINARY, 972 .len = sizeof(struct ifla_vf_mac) }, 973 [IFLA_VF_VLAN] = { .type = NLA_BINARY, 974 .len = sizeof(struct ifla_vf_vlan) }, 975 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY, 976 .len = sizeof(struct ifla_vf_tx_rate) }, 977 }; 978 979 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = { 980 [IFLA_PORT_VF] = { .type = NLA_U32 }, 981 [IFLA_PORT_PROFILE] = { .type = NLA_STRING, 982 .len = PORT_PROFILE_MAX }, 983 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY, 984 .len = sizeof(struct ifla_port_vsi)}, 985 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY, 986 .len = PORT_UUID_MAX }, 987 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING, 988 .len = PORT_UUID_MAX }, 989 [IFLA_PORT_REQUEST] = { .type = NLA_U8, }, 990 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, }, 991 }; 992 993 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[]) 994 { 995 struct net *net; 996 /* Examine the link attributes and figure out which 997 * network namespace we are talking about. 998 */ 999 if (tb[IFLA_NET_NS_PID]) 1000 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID])); 1001 else 1002 net = get_net(src_net); 1003 return net; 1004 } 1005 EXPORT_SYMBOL(rtnl_link_get_net); 1006 1007 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[]) 1008 { 1009 if (dev) { 1010 if (tb[IFLA_ADDRESS] && 1011 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len) 1012 return -EINVAL; 1013 1014 if (tb[IFLA_BROADCAST] && 1015 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len) 1016 return -EINVAL; 1017 } 1018 1019 return 0; 1020 } 1021 1022 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr) 1023 { 1024 int rem, err = -EINVAL; 1025 struct nlattr *vf; 1026 const struct net_device_ops *ops = dev->netdev_ops; 1027 1028 nla_for_each_nested(vf, attr, rem) { 1029 switch (nla_type(vf)) { 1030 case IFLA_VF_MAC: { 1031 struct ifla_vf_mac *ivm; 1032 ivm = nla_data(vf); 1033 err = -EOPNOTSUPP; 1034 if (ops->ndo_set_vf_mac) 1035 err = ops->ndo_set_vf_mac(dev, ivm->vf, 1036 ivm->mac); 1037 break; 1038 } 1039 case IFLA_VF_VLAN: { 1040 struct ifla_vf_vlan *ivv; 1041 ivv = nla_data(vf); 1042 err = -EOPNOTSUPP; 1043 if (ops->ndo_set_vf_vlan) 1044 err = ops->ndo_set_vf_vlan(dev, ivv->vf, 1045 ivv->vlan, 1046 ivv->qos); 1047 break; 1048 } 1049 case IFLA_VF_TX_RATE: { 1050 struct ifla_vf_tx_rate *ivt; 1051 ivt = nla_data(vf); 1052 err = -EOPNOTSUPP; 1053 if (ops->ndo_set_vf_tx_rate) 1054 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf, 1055 ivt->rate); 1056 break; 1057 } 1058 default: 1059 err = -EINVAL; 1060 break; 1061 } 1062 if (err) 1063 break; 1064 } 1065 return err; 1066 } 1067 1068 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm, 1069 struct nlattr **tb, char *ifname, int modified) 1070 { 1071 const struct net_device_ops *ops = dev->netdev_ops; 1072 int send_addr_notify = 0; 1073 int err; 1074 1075 if (tb[IFLA_NET_NS_PID]) { 1076 struct net *net = rtnl_link_get_net(dev_net(dev), tb); 1077 if (IS_ERR(net)) { 1078 err = PTR_ERR(net); 1079 goto errout; 1080 } 1081 err = dev_change_net_namespace(dev, net, ifname); 1082 put_net(net); 1083 if (err) 1084 goto errout; 1085 modified = 1; 1086 } 1087 1088 if (tb[IFLA_MAP]) { 1089 struct rtnl_link_ifmap *u_map; 1090 struct ifmap k_map; 1091 1092 if (!ops->ndo_set_config) { 1093 err = -EOPNOTSUPP; 1094 goto errout; 1095 } 1096 1097 if (!netif_device_present(dev)) { 1098 err = -ENODEV; 1099 goto errout; 1100 } 1101 1102 u_map = nla_data(tb[IFLA_MAP]); 1103 k_map.mem_start = (unsigned long) u_map->mem_start; 1104 k_map.mem_end = (unsigned long) u_map->mem_end; 1105 k_map.base_addr = (unsigned short) u_map->base_addr; 1106 k_map.irq = (unsigned char) u_map->irq; 1107 k_map.dma = (unsigned char) u_map->dma; 1108 k_map.port = (unsigned char) u_map->port; 1109 1110 err = ops->ndo_set_config(dev, &k_map); 1111 if (err < 0) 1112 goto errout; 1113 1114 modified = 1; 1115 } 1116 1117 if (tb[IFLA_ADDRESS]) { 1118 struct sockaddr *sa; 1119 int len; 1120 1121 if (!ops->ndo_set_mac_address) { 1122 err = -EOPNOTSUPP; 1123 goto errout; 1124 } 1125 1126 if (!netif_device_present(dev)) { 1127 err = -ENODEV; 1128 goto errout; 1129 } 1130 1131 len = sizeof(sa_family_t) + dev->addr_len; 1132 sa = kmalloc(len, GFP_KERNEL); 1133 if (!sa) { 1134 err = -ENOMEM; 1135 goto errout; 1136 } 1137 sa->sa_family = dev->type; 1138 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]), 1139 dev->addr_len); 1140 err = ops->ndo_set_mac_address(dev, sa); 1141 kfree(sa); 1142 if (err) 1143 goto errout; 1144 send_addr_notify = 1; 1145 modified = 1; 1146 } 1147 1148 if (tb[IFLA_MTU]) { 1149 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 1150 if (err < 0) 1151 goto errout; 1152 modified = 1; 1153 } 1154 1155 /* 1156 * Interface selected by interface index but interface 1157 * name provided implies that a name change has been 1158 * requested. 1159 */ 1160 if (ifm->ifi_index > 0 && ifname[0]) { 1161 err = dev_change_name(dev, ifname); 1162 if (err < 0) 1163 goto errout; 1164 modified = 1; 1165 } 1166 1167 if (tb[IFLA_IFALIAS]) { 1168 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]), 1169 nla_len(tb[IFLA_IFALIAS])); 1170 if (err < 0) 1171 goto errout; 1172 modified = 1; 1173 } 1174 1175 if (tb[IFLA_BROADCAST]) { 1176 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len); 1177 send_addr_notify = 1; 1178 } 1179 1180 if (ifm->ifi_flags || ifm->ifi_change) { 1181 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm)); 1182 if (err < 0) 1183 goto errout; 1184 } 1185 1186 if (tb[IFLA_TXQLEN]) 1187 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]); 1188 1189 if (tb[IFLA_OPERSTATE]) 1190 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 1191 1192 if (tb[IFLA_LINKMODE]) { 1193 write_lock_bh(&dev_base_lock); 1194 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]); 1195 write_unlock_bh(&dev_base_lock); 1196 } 1197 1198 if (tb[IFLA_VFINFO_LIST]) { 1199 struct nlattr *attr; 1200 int rem; 1201 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) { 1202 if (nla_type(attr) != IFLA_VF_INFO) { 1203 err = -EINVAL; 1204 goto errout; 1205 } 1206 err = do_setvfinfo(dev, attr); 1207 if (err < 0) 1208 goto errout; 1209 modified = 1; 1210 } 1211 } 1212 err = 0; 1213 1214 if (tb[IFLA_VF_PORTS]) { 1215 struct nlattr *port[IFLA_PORT_MAX+1]; 1216 struct nlattr *attr; 1217 int vf; 1218 int rem; 1219 1220 err = -EOPNOTSUPP; 1221 if (!ops->ndo_set_vf_port) 1222 goto errout; 1223 1224 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) { 1225 if (nla_type(attr) != IFLA_VF_PORT) 1226 continue; 1227 err = nla_parse_nested(port, IFLA_PORT_MAX, 1228 attr, ifla_port_policy); 1229 if (err < 0) 1230 goto errout; 1231 if (!port[IFLA_PORT_VF]) { 1232 err = -EOPNOTSUPP; 1233 goto errout; 1234 } 1235 vf = nla_get_u32(port[IFLA_PORT_VF]); 1236 err = ops->ndo_set_vf_port(dev, vf, port); 1237 if (err < 0) 1238 goto errout; 1239 modified = 1; 1240 } 1241 } 1242 err = 0; 1243 1244 if (tb[IFLA_PORT_SELF]) { 1245 struct nlattr *port[IFLA_PORT_MAX+1]; 1246 1247 err = nla_parse_nested(port, IFLA_PORT_MAX, 1248 tb[IFLA_PORT_SELF], ifla_port_policy); 1249 if (err < 0) 1250 goto errout; 1251 1252 err = -EOPNOTSUPP; 1253 if (ops->ndo_set_vf_port) 1254 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port); 1255 if (err < 0) 1256 goto errout; 1257 modified = 1; 1258 } 1259 err = 0; 1260 1261 errout: 1262 if (err < 0 && modified && net_ratelimit()) 1263 printk(KERN_WARNING "A link change request failed with " 1264 "some changes comitted already. Interface %s may " 1265 "have been left with an inconsistent configuration, " 1266 "please check.\n", dev->name); 1267 1268 if (send_addr_notify) 1269 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); 1270 return err; 1271 } 1272 1273 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) 1274 { 1275 struct net *net = sock_net(skb->sk); 1276 struct ifinfomsg *ifm; 1277 struct net_device *dev; 1278 int err; 1279 struct nlattr *tb[IFLA_MAX+1]; 1280 char ifname[IFNAMSIZ]; 1281 1282 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy); 1283 if (err < 0) 1284 goto errout; 1285 1286 if (tb[IFLA_IFNAME]) 1287 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 1288 else 1289 ifname[0] = '\0'; 1290 1291 err = -EINVAL; 1292 ifm = nlmsg_data(nlh); 1293 if (ifm->ifi_index > 0) 1294 dev = __dev_get_by_index(net, ifm->ifi_index); 1295 else if (tb[IFLA_IFNAME]) 1296 dev = __dev_get_by_name(net, ifname); 1297 else 1298 goto errout; 1299 1300 if (dev == NULL) { 1301 err = -ENODEV; 1302 goto errout; 1303 } 1304 1305 err = validate_linkmsg(dev, tb); 1306 if (err < 0) 1307 goto errout; 1308 1309 err = do_setlink(dev, ifm, tb, ifname, 0); 1310 errout: 1311 return err; 1312 } 1313 1314 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) 1315 { 1316 struct net *net = sock_net(skb->sk); 1317 const struct rtnl_link_ops *ops; 1318 struct net_device *dev; 1319 struct ifinfomsg *ifm; 1320 char ifname[IFNAMSIZ]; 1321 struct nlattr *tb[IFLA_MAX+1]; 1322 int err; 1323 1324 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy); 1325 if (err < 0) 1326 return err; 1327 1328 if (tb[IFLA_IFNAME]) 1329 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 1330 1331 ifm = nlmsg_data(nlh); 1332 if (ifm->ifi_index > 0) 1333 dev = __dev_get_by_index(net, ifm->ifi_index); 1334 else if (tb[IFLA_IFNAME]) 1335 dev = __dev_get_by_name(net, ifname); 1336 else 1337 return -EINVAL; 1338 1339 if (!dev) 1340 return -ENODEV; 1341 1342 ops = dev->rtnl_link_ops; 1343 if (!ops) 1344 return -EOPNOTSUPP; 1345 1346 ops->dellink(dev, NULL); 1347 return 0; 1348 } 1349 1350 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm) 1351 { 1352 unsigned int old_flags; 1353 int err; 1354 1355 old_flags = dev->flags; 1356 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) { 1357 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm)); 1358 if (err < 0) 1359 return err; 1360 } 1361 1362 dev->rtnl_link_state = RTNL_LINK_INITIALIZED; 1363 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U); 1364 1365 __dev_notify_flags(dev, old_flags); 1366 return 0; 1367 } 1368 EXPORT_SYMBOL(rtnl_configure_link); 1369 1370 struct net_device *rtnl_create_link(struct net *src_net, struct net *net, 1371 char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[]) 1372 { 1373 int err; 1374 struct net_device *dev; 1375 unsigned int num_queues = 1; 1376 unsigned int real_num_queues = 1; 1377 1378 if (ops->get_tx_queues) { 1379 err = ops->get_tx_queues(src_net, tb, &num_queues, 1380 &real_num_queues); 1381 if (err) 1382 goto err; 1383 } 1384 err = -ENOMEM; 1385 dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues); 1386 if (!dev) 1387 goto err; 1388 1389 dev_net_set(dev, net); 1390 dev->rtnl_link_ops = ops; 1391 dev->rtnl_link_state = RTNL_LINK_INITIALIZING; 1392 dev->real_num_tx_queues = real_num_queues; 1393 1394 if (strchr(dev->name, '%')) { 1395 err = dev_alloc_name(dev, dev->name); 1396 if (err < 0) 1397 goto err_free; 1398 } 1399 1400 if (tb[IFLA_MTU]) 1401 dev->mtu = nla_get_u32(tb[IFLA_MTU]); 1402 if (tb[IFLA_ADDRESS]) 1403 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]), 1404 nla_len(tb[IFLA_ADDRESS])); 1405 if (tb[IFLA_BROADCAST]) 1406 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]), 1407 nla_len(tb[IFLA_BROADCAST])); 1408 if (tb[IFLA_TXQLEN]) 1409 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]); 1410 if (tb[IFLA_OPERSTATE]) 1411 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 1412 if (tb[IFLA_LINKMODE]) 1413 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]); 1414 1415 return dev; 1416 1417 err_free: 1418 free_netdev(dev); 1419 err: 1420 return ERR_PTR(err); 1421 } 1422 EXPORT_SYMBOL(rtnl_create_link); 1423 1424 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) 1425 { 1426 struct net *net = sock_net(skb->sk); 1427 const struct rtnl_link_ops *ops; 1428 struct net_device *dev; 1429 struct ifinfomsg *ifm; 1430 char kind[MODULE_NAME_LEN]; 1431 char ifname[IFNAMSIZ]; 1432 struct nlattr *tb[IFLA_MAX+1]; 1433 struct nlattr *linkinfo[IFLA_INFO_MAX+1]; 1434 int err; 1435 1436 #ifdef CONFIG_MODULES 1437 replay: 1438 #endif 1439 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy); 1440 if (err < 0) 1441 return err; 1442 1443 if (tb[IFLA_IFNAME]) 1444 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 1445 else 1446 ifname[0] = '\0'; 1447 1448 ifm = nlmsg_data(nlh); 1449 if (ifm->ifi_index > 0) 1450 dev = __dev_get_by_index(net, ifm->ifi_index); 1451 else if (ifname[0]) 1452 dev = __dev_get_by_name(net, ifname); 1453 else 1454 dev = NULL; 1455 1456 err = validate_linkmsg(dev, tb); 1457 if (err < 0) 1458 return err; 1459 1460 if (tb[IFLA_LINKINFO]) { 1461 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX, 1462 tb[IFLA_LINKINFO], ifla_info_policy); 1463 if (err < 0) 1464 return err; 1465 } else 1466 memset(linkinfo, 0, sizeof(linkinfo)); 1467 1468 if (linkinfo[IFLA_INFO_KIND]) { 1469 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind)); 1470 ops = rtnl_link_ops_get(kind); 1471 } else { 1472 kind[0] = '\0'; 1473 ops = NULL; 1474 } 1475 1476 if (1) { 1477 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL; 1478 struct net *dest_net; 1479 1480 if (ops) { 1481 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) { 1482 err = nla_parse_nested(attr, ops->maxtype, 1483 linkinfo[IFLA_INFO_DATA], 1484 ops->policy); 1485 if (err < 0) 1486 return err; 1487 data = attr; 1488 } 1489 if (ops->validate) { 1490 err = ops->validate(tb, data); 1491 if (err < 0) 1492 return err; 1493 } 1494 } 1495 1496 if (dev) { 1497 int modified = 0; 1498 1499 if (nlh->nlmsg_flags & NLM_F_EXCL) 1500 return -EEXIST; 1501 if (nlh->nlmsg_flags & NLM_F_REPLACE) 1502 return -EOPNOTSUPP; 1503 1504 if (linkinfo[IFLA_INFO_DATA]) { 1505 if (!ops || ops != dev->rtnl_link_ops || 1506 !ops->changelink) 1507 return -EOPNOTSUPP; 1508 1509 err = ops->changelink(dev, tb, data); 1510 if (err < 0) 1511 return err; 1512 modified = 1; 1513 } 1514 1515 return do_setlink(dev, ifm, tb, ifname, modified); 1516 } 1517 1518 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) 1519 return -ENODEV; 1520 1521 if (ifm->ifi_index) 1522 return -EOPNOTSUPP; 1523 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO]) 1524 return -EOPNOTSUPP; 1525 1526 if (!ops) { 1527 #ifdef CONFIG_MODULES 1528 if (kind[0]) { 1529 __rtnl_unlock(); 1530 request_module("rtnl-link-%s", kind); 1531 rtnl_lock(); 1532 ops = rtnl_link_ops_get(kind); 1533 if (ops) 1534 goto replay; 1535 } 1536 #endif 1537 return -EOPNOTSUPP; 1538 } 1539 1540 if (!ifname[0]) 1541 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind); 1542 1543 dest_net = rtnl_link_get_net(net, tb); 1544 dev = rtnl_create_link(net, dest_net, ifname, ops, tb); 1545 1546 if (IS_ERR(dev)) 1547 err = PTR_ERR(dev); 1548 else if (ops->newlink) 1549 err = ops->newlink(net, dev, tb, data); 1550 else 1551 err = register_netdevice(dev); 1552 1553 if (err < 0 && !IS_ERR(dev)) 1554 free_netdev(dev); 1555 if (err < 0) 1556 goto out; 1557 1558 err = rtnl_configure_link(dev, ifm); 1559 if (err < 0) 1560 unregister_netdevice(dev); 1561 out: 1562 put_net(dest_net); 1563 return err; 1564 } 1565 } 1566 1567 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg) 1568 { 1569 struct net *net = sock_net(skb->sk); 1570 struct ifinfomsg *ifm; 1571 char ifname[IFNAMSIZ]; 1572 struct nlattr *tb[IFLA_MAX+1]; 1573 struct net_device *dev = NULL; 1574 struct sk_buff *nskb; 1575 int err; 1576 1577 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy); 1578 if (err < 0) 1579 return err; 1580 1581 if (tb[IFLA_IFNAME]) 1582 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 1583 1584 ifm = nlmsg_data(nlh); 1585 if (ifm->ifi_index > 0) 1586 dev = __dev_get_by_index(net, ifm->ifi_index); 1587 else if (tb[IFLA_IFNAME]) 1588 dev = __dev_get_by_name(net, ifname); 1589 else 1590 return -EINVAL; 1591 1592 if (dev == NULL) 1593 return -ENODEV; 1594 1595 nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL); 1596 if (nskb == NULL) 1597 return -ENOBUFS; 1598 1599 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid, 1600 nlh->nlmsg_seq, 0, 0); 1601 if (err < 0) { 1602 /* -EMSGSIZE implies BUG in if_nlmsg_size */ 1603 WARN_ON(err == -EMSGSIZE); 1604 kfree_skb(nskb); 1605 } else 1606 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid); 1607 1608 return err; 1609 } 1610 1611 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb) 1612 { 1613 int idx; 1614 int s_idx = cb->family; 1615 1616 if (s_idx == 0) 1617 s_idx = 1; 1618 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) { 1619 int type = cb->nlh->nlmsg_type-RTM_BASE; 1620 if (idx < s_idx || idx == PF_PACKET) 1621 continue; 1622 if (rtnl_msg_handlers[idx] == NULL || 1623 rtnl_msg_handlers[idx][type].dumpit == NULL) 1624 continue; 1625 if (idx > s_idx) 1626 memset(&cb->args[0], 0, sizeof(cb->args)); 1627 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb)) 1628 break; 1629 } 1630 cb->family = idx; 1631 1632 return skb->len; 1633 } 1634 1635 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change) 1636 { 1637 struct net *net = dev_net(dev); 1638 struct sk_buff *skb; 1639 int err = -ENOBUFS; 1640 1641 skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL); 1642 if (skb == NULL) 1643 goto errout; 1644 1645 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0); 1646 if (err < 0) { 1647 /* -EMSGSIZE implies BUG in if_nlmsg_size() */ 1648 WARN_ON(err == -EMSGSIZE); 1649 kfree_skb(skb); 1650 goto errout; 1651 } 1652 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL); 1653 return; 1654 errout: 1655 if (err < 0) 1656 rtnl_set_sk_err(net, RTNLGRP_LINK, err); 1657 } 1658 1659 /* Protected by RTNL sempahore. */ 1660 static struct rtattr **rta_buf; 1661 static int rtattr_max; 1662 1663 /* Process one rtnetlink message. */ 1664 1665 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh) 1666 { 1667 struct net *net = sock_net(skb->sk); 1668 rtnl_doit_func doit; 1669 int sz_idx, kind; 1670 int min_len; 1671 int family; 1672 int type; 1673 int err; 1674 1675 type = nlh->nlmsg_type; 1676 if (type > RTM_MAX) 1677 return -EOPNOTSUPP; 1678 1679 type -= RTM_BASE; 1680 1681 /* All the messages must have at least 1 byte length */ 1682 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg))) 1683 return 0; 1684 1685 family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family; 1686 sz_idx = type>>2; 1687 kind = type&3; 1688 1689 if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN)) 1690 return -EPERM; 1691 1692 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) { 1693 struct sock *rtnl; 1694 rtnl_dumpit_func dumpit; 1695 1696 dumpit = rtnl_get_dumpit(family, type); 1697 if (dumpit == NULL) 1698 return -EOPNOTSUPP; 1699 1700 __rtnl_unlock(); 1701 rtnl = net->rtnl; 1702 err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL); 1703 rtnl_lock(); 1704 return err; 1705 } 1706 1707 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *))); 1708 1709 min_len = rtm_min[sz_idx]; 1710 if (nlh->nlmsg_len < min_len) 1711 return -EINVAL; 1712 1713 if (nlh->nlmsg_len > min_len) { 1714 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len); 1715 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len); 1716 1717 while (RTA_OK(attr, attrlen)) { 1718 unsigned flavor = attr->rta_type; 1719 if (flavor) { 1720 if (flavor > rta_max[sz_idx]) 1721 return -EINVAL; 1722 rta_buf[flavor-1] = attr; 1723 } 1724 attr = RTA_NEXT(attr, attrlen); 1725 } 1726 } 1727 1728 doit = rtnl_get_doit(family, type); 1729 if (doit == NULL) 1730 return -EOPNOTSUPP; 1731 1732 return doit(skb, nlh, (void *)&rta_buf[0]); 1733 } 1734 1735 static void rtnetlink_rcv(struct sk_buff *skb) 1736 { 1737 rtnl_lock(); 1738 netlink_rcv_skb(skb, &rtnetlink_rcv_msg); 1739 rtnl_unlock(); 1740 } 1741 1742 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr) 1743 { 1744 struct net_device *dev = ptr; 1745 1746 switch (event) { 1747 case NETDEV_UP: 1748 case NETDEV_DOWN: 1749 case NETDEV_PRE_UP: 1750 case NETDEV_POST_INIT: 1751 case NETDEV_REGISTER: 1752 case NETDEV_CHANGE: 1753 case NETDEV_PRE_TYPE_CHANGE: 1754 case NETDEV_GOING_DOWN: 1755 case NETDEV_UNREGISTER: 1756 case NETDEV_UNREGISTER_BATCH: 1757 break; 1758 default: 1759 rtmsg_ifinfo(RTM_NEWLINK, dev, 0); 1760 break; 1761 } 1762 return NOTIFY_DONE; 1763 } 1764 1765 static struct notifier_block rtnetlink_dev_notifier = { 1766 .notifier_call = rtnetlink_event, 1767 }; 1768 1769 1770 static int __net_init rtnetlink_net_init(struct net *net) 1771 { 1772 struct sock *sk; 1773 sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX, 1774 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE); 1775 if (!sk) 1776 return -ENOMEM; 1777 net->rtnl = sk; 1778 return 0; 1779 } 1780 1781 static void __net_exit rtnetlink_net_exit(struct net *net) 1782 { 1783 netlink_kernel_release(net->rtnl); 1784 net->rtnl = NULL; 1785 } 1786 1787 static struct pernet_operations rtnetlink_net_ops = { 1788 .init = rtnetlink_net_init, 1789 .exit = rtnetlink_net_exit, 1790 }; 1791 1792 void __init rtnetlink_init(void) 1793 { 1794 int i; 1795 1796 rtattr_max = 0; 1797 for (i = 0; i < ARRAY_SIZE(rta_max); i++) 1798 if (rta_max[i] > rtattr_max) 1799 rtattr_max = rta_max[i]; 1800 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL); 1801 if (!rta_buf) 1802 panic("rtnetlink_init: cannot allocate rta_buf\n"); 1803 1804 if (register_pernet_subsys(&rtnetlink_net_ops)) 1805 panic("rtnetlink_init: cannot initialize rtnetlink\n"); 1806 1807 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV); 1808 register_netdevice_notifier(&rtnetlink_dev_notifier); 1809 1810 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo); 1811 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL); 1812 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL); 1813 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL); 1814 1815 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all); 1816 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all); 1817 } 1818 1819