1 /* 2 * NET3 IP device support routines. 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of the GNU General Public License 6 * as published by the Free Software Foundation; either version 7 * 2 of the License, or (at your option) any later version. 8 * 9 * Derived from the IP parts of dev.c 1.0.19 10 * Authors: Ross Biro 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 12 * Mark Evans, <evansmp@uhura.aston.ac.uk> 13 * 14 * Additional Authors: 15 * Alan Cox, <gw4pts@gw4pts.ampr.org> 16 * Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> 17 * 18 * Changes: 19 * Alexey Kuznetsov: pa_* fields are replaced with ifaddr 20 * lists. 21 * Cyrus Durgin: updated for kmod 22 * Matthias Andree: in devinet_ioctl, compare label and 23 * address (4.4BSD alias style support), 24 * fall back to comparing just the label 25 * if no match found. 26 */ 27 28 29 #include <asm/uaccess.h> 30 #include <asm/system.h> 31 #include <linux/bitops.h> 32 #include <linux/capability.h> 33 #include <linux/module.h> 34 #include <linux/types.h> 35 #include <linux/kernel.h> 36 #include <linux/string.h> 37 #include <linux/mm.h> 38 #include <linux/socket.h> 39 #include <linux/sockios.h> 40 #include <linux/in.h> 41 #include <linux/errno.h> 42 #include <linux/interrupt.h> 43 #include <linux/if_addr.h> 44 #include <linux/if_ether.h> 45 #include <linux/inet.h> 46 #include <linux/netdevice.h> 47 #include <linux/etherdevice.h> 48 #include <linux/skbuff.h> 49 #include <linux/init.h> 50 #include <linux/notifier.h> 51 #include <linux/inetdevice.h> 52 #include <linux/igmp.h> 53 #include <linux/slab.h> 54 #ifdef CONFIG_SYSCTL 55 #include <linux/sysctl.h> 56 #endif 57 #include <linux/kmod.h> 58 59 #include <net/arp.h> 60 #include <net/ip.h> 61 #include <net/route.h> 62 #include <net/ip_fib.h> 63 #include <net/rtnetlink.h> 64 #include <net/net_namespace.h> 65 66 static struct ipv4_devconf ipv4_devconf = { 67 .data = { 68 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1, 69 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1, 70 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1, 71 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1, 72 }, 73 }; 74 75 static struct ipv4_devconf ipv4_devconf_dflt = { 76 .data = { 77 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1, 78 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1, 79 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1, 80 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1, 81 [IPV4_DEVCONF_ACCEPT_SOURCE_ROUTE - 1] = 1, 82 }, 83 }; 84 85 #define IPV4_DEVCONF_DFLT(net, attr) \ 86 IPV4_DEVCONF((*net->ipv4.devconf_dflt), attr) 87 88 static const struct nla_policy ifa_ipv4_policy[IFA_MAX+1] = { 89 [IFA_LOCAL] = { .type = NLA_U32 }, 90 [IFA_ADDRESS] = { .type = NLA_U32 }, 91 [IFA_BROADCAST] = { .type = NLA_U32 }, 92 [IFA_LABEL] = { .type = NLA_STRING, .len = IFNAMSIZ - 1 }, 93 }; 94 95 static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32); 96 97 static BLOCKING_NOTIFIER_HEAD(inetaddr_chain); 98 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap, 99 int destroy); 100 #ifdef CONFIG_SYSCTL 101 static void devinet_sysctl_register(struct in_device *idev); 102 static void devinet_sysctl_unregister(struct in_device *idev); 103 #else 104 static inline void devinet_sysctl_register(struct in_device *idev) 105 { 106 } 107 static inline void devinet_sysctl_unregister(struct in_device *idev) 108 { 109 } 110 #endif 111 112 /* Locks all the inet devices. */ 113 114 static struct in_ifaddr *inet_alloc_ifa(void) 115 { 116 return kzalloc(sizeof(struct in_ifaddr), GFP_KERNEL); 117 } 118 119 static void inet_rcu_free_ifa(struct rcu_head *head) 120 { 121 struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head); 122 if (ifa->ifa_dev) 123 in_dev_put(ifa->ifa_dev); 124 kfree(ifa); 125 } 126 127 static inline void inet_free_ifa(struct in_ifaddr *ifa) 128 { 129 call_rcu(&ifa->rcu_head, inet_rcu_free_ifa); 130 } 131 132 void in_dev_finish_destroy(struct in_device *idev) 133 { 134 struct net_device *dev = idev->dev; 135 136 WARN_ON(idev->ifa_list); 137 WARN_ON(idev->mc_list); 138 #ifdef NET_REFCNT_DEBUG 139 printk(KERN_DEBUG "in_dev_finish_destroy: %p=%s\n", 140 idev, dev ? dev->name : "NIL"); 141 #endif 142 dev_put(dev); 143 if (!idev->dead) 144 pr_err("Freeing alive in_device %p\n", idev); 145 else 146 kfree(idev); 147 } 148 EXPORT_SYMBOL(in_dev_finish_destroy); 149 150 static struct in_device *inetdev_init(struct net_device *dev) 151 { 152 struct in_device *in_dev; 153 154 ASSERT_RTNL(); 155 156 in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL); 157 if (!in_dev) 158 goto out; 159 memcpy(&in_dev->cnf, dev_net(dev)->ipv4.devconf_dflt, 160 sizeof(in_dev->cnf)); 161 in_dev->cnf.sysctl = NULL; 162 in_dev->dev = dev; 163 in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl); 164 if (!in_dev->arp_parms) 165 goto out_kfree; 166 if (IPV4_DEVCONF(in_dev->cnf, FORWARDING)) 167 dev_disable_lro(dev); 168 /* Reference in_dev->dev */ 169 dev_hold(dev); 170 /* Account for reference dev->ip_ptr (below) */ 171 in_dev_hold(in_dev); 172 173 devinet_sysctl_register(in_dev); 174 ip_mc_init_dev(in_dev); 175 if (dev->flags & IFF_UP) 176 ip_mc_up(in_dev); 177 178 /* we can receive as soon as ip_ptr is set -- do this last */ 179 rcu_assign_pointer(dev->ip_ptr, in_dev); 180 out: 181 return in_dev; 182 out_kfree: 183 kfree(in_dev); 184 in_dev = NULL; 185 goto out; 186 } 187 188 static void in_dev_rcu_put(struct rcu_head *head) 189 { 190 struct in_device *idev = container_of(head, struct in_device, rcu_head); 191 in_dev_put(idev); 192 } 193 194 static void inetdev_destroy(struct in_device *in_dev) 195 { 196 struct in_ifaddr *ifa; 197 struct net_device *dev; 198 199 ASSERT_RTNL(); 200 201 dev = in_dev->dev; 202 203 in_dev->dead = 1; 204 205 ip_mc_destroy_dev(in_dev); 206 207 while ((ifa = in_dev->ifa_list) != NULL) { 208 inet_del_ifa(in_dev, &in_dev->ifa_list, 0); 209 inet_free_ifa(ifa); 210 } 211 212 dev->ip_ptr = NULL; 213 214 devinet_sysctl_unregister(in_dev); 215 neigh_parms_release(&arp_tbl, in_dev->arp_parms); 216 arp_ifdown(dev); 217 218 call_rcu(&in_dev->rcu_head, in_dev_rcu_put); 219 } 220 221 int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b) 222 { 223 rcu_read_lock(); 224 for_primary_ifa(in_dev) { 225 if (inet_ifa_match(a, ifa)) { 226 if (!b || inet_ifa_match(b, ifa)) { 227 rcu_read_unlock(); 228 return 1; 229 } 230 } 231 } endfor_ifa(in_dev); 232 rcu_read_unlock(); 233 return 0; 234 } 235 236 static void __inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap, 237 int destroy, struct nlmsghdr *nlh, u32 pid) 238 { 239 struct in_ifaddr *promote = NULL; 240 struct in_ifaddr *ifa, *ifa1 = *ifap; 241 struct in_ifaddr *last_prim = in_dev->ifa_list; 242 struct in_ifaddr *prev_prom = NULL; 243 int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev); 244 245 ASSERT_RTNL(); 246 247 /* 1. Deleting primary ifaddr forces deletion all secondaries 248 * unless alias promotion is set 249 **/ 250 251 if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) { 252 struct in_ifaddr **ifap1 = &ifa1->ifa_next; 253 254 while ((ifa = *ifap1) != NULL) { 255 if (!(ifa->ifa_flags & IFA_F_SECONDARY) && 256 ifa1->ifa_scope <= ifa->ifa_scope) 257 last_prim = ifa; 258 259 if (!(ifa->ifa_flags & IFA_F_SECONDARY) || 260 ifa1->ifa_mask != ifa->ifa_mask || 261 !inet_ifa_match(ifa1->ifa_address, ifa)) { 262 ifap1 = &ifa->ifa_next; 263 prev_prom = ifa; 264 continue; 265 } 266 267 if (!do_promote) { 268 *ifap1 = ifa->ifa_next; 269 270 rtmsg_ifa(RTM_DELADDR, ifa, nlh, pid); 271 blocking_notifier_call_chain(&inetaddr_chain, 272 NETDEV_DOWN, ifa); 273 inet_free_ifa(ifa); 274 } else { 275 promote = ifa; 276 break; 277 } 278 } 279 } 280 281 /* 2. Unlink it */ 282 283 *ifap = ifa1->ifa_next; 284 285 /* 3. Announce address deletion */ 286 287 /* Send message first, then call notifier. 288 At first sight, FIB update triggered by notifier 289 will refer to already deleted ifaddr, that could confuse 290 netlink listeners. It is not true: look, gated sees 291 that route deleted and if it still thinks that ifaddr 292 is valid, it will try to restore deleted routes... Grr. 293 So that, this order is correct. 294 */ 295 rtmsg_ifa(RTM_DELADDR, ifa1, nlh, pid); 296 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1); 297 298 if (promote) { 299 300 if (prev_prom) { 301 prev_prom->ifa_next = promote->ifa_next; 302 promote->ifa_next = last_prim->ifa_next; 303 last_prim->ifa_next = promote; 304 } 305 306 promote->ifa_flags &= ~IFA_F_SECONDARY; 307 rtmsg_ifa(RTM_NEWADDR, promote, nlh, pid); 308 blocking_notifier_call_chain(&inetaddr_chain, 309 NETDEV_UP, promote); 310 for (ifa = promote->ifa_next; ifa; ifa = ifa->ifa_next) { 311 if (ifa1->ifa_mask != ifa->ifa_mask || 312 !inet_ifa_match(ifa1->ifa_address, ifa)) 313 continue; 314 fib_add_ifaddr(ifa); 315 } 316 317 } 318 if (destroy) 319 inet_free_ifa(ifa1); 320 } 321 322 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap, 323 int destroy) 324 { 325 __inet_del_ifa(in_dev, ifap, destroy, NULL, 0); 326 } 327 328 static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh, 329 u32 pid) 330 { 331 struct in_device *in_dev = ifa->ifa_dev; 332 struct in_ifaddr *ifa1, **ifap, **last_primary; 333 334 ASSERT_RTNL(); 335 336 if (!ifa->ifa_local) { 337 inet_free_ifa(ifa); 338 return 0; 339 } 340 341 ifa->ifa_flags &= ~IFA_F_SECONDARY; 342 last_primary = &in_dev->ifa_list; 343 344 for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL; 345 ifap = &ifa1->ifa_next) { 346 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) && 347 ifa->ifa_scope <= ifa1->ifa_scope) 348 last_primary = &ifa1->ifa_next; 349 if (ifa1->ifa_mask == ifa->ifa_mask && 350 inet_ifa_match(ifa1->ifa_address, ifa)) { 351 if (ifa1->ifa_local == ifa->ifa_local) { 352 inet_free_ifa(ifa); 353 return -EEXIST; 354 } 355 if (ifa1->ifa_scope != ifa->ifa_scope) { 356 inet_free_ifa(ifa); 357 return -EINVAL; 358 } 359 ifa->ifa_flags |= IFA_F_SECONDARY; 360 } 361 } 362 363 if (!(ifa->ifa_flags & IFA_F_SECONDARY)) { 364 net_srandom(ifa->ifa_local); 365 ifap = last_primary; 366 } 367 368 ifa->ifa_next = *ifap; 369 *ifap = ifa; 370 371 /* Send message first, then call notifier. 372 Notifier will trigger FIB update, so that 373 listeners of netlink will know about new ifaddr */ 374 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, pid); 375 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa); 376 377 return 0; 378 } 379 380 static int inet_insert_ifa(struct in_ifaddr *ifa) 381 { 382 return __inet_insert_ifa(ifa, NULL, 0); 383 } 384 385 static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa) 386 { 387 struct in_device *in_dev = __in_dev_get_rtnl(dev); 388 389 ASSERT_RTNL(); 390 391 if (!in_dev) { 392 inet_free_ifa(ifa); 393 return -ENOBUFS; 394 } 395 ipv4_devconf_setall(in_dev); 396 if (ifa->ifa_dev != in_dev) { 397 WARN_ON(ifa->ifa_dev); 398 in_dev_hold(in_dev); 399 ifa->ifa_dev = in_dev; 400 } 401 if (ipv4_is_loopback(ifa->ifa_local)) 402 ifa->ifa_scope = RT_SCOPE_HOST; 403 return inet_insert_ifa(ifa); 404 } 405 406 struct in_device *inetdev_by_index(struct net *net, int ifindex) 407 { 408 struct net_device *dev; 409 struct in_device *in_dev = NULL; 410 411 rcu_read_lock(); 412 dev = dev_get_by_index_rcu(net, ifindex); 413 if (dev) 414 in_dev = in_dev_get(dev); 415 rcu_read_unlock(); 416 return in_dev; 417 } 418 EXPORT_SYMBOL(inetdev_by_index); 419 420 /* Called only from RTNL semaphored context. No locks. */ 421 422 struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix, 423 __be32 mask) 424 { 425 ASSERT_RTNL(); 426 427 for_primary_ifa(in_dev) { 428 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa)) 429 return ifa; 430 } endfor_ifa(in_dev); 431 return NULL; 432 } 433 434 static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) 435 { 436 struct net *net = sock_net(skb->sk); 437 struct nlattr *tb[IFA_MAX+1]; 438 struct in_device *in_dev; 439 struct ifaddrmsg *ifm; 440 struct in_ifaddr *ifa, **ifap; 441 int err = -EINVAL; 442 443 ASSERT_RTNL(); 444 445 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy); 446 if (err < 0) 447 goto errout; 448 449 ifm = nlmsg_data(nlh); 450 in_dev = inetdev_by_index(net, ifm->ifa_index); 451 if (in_dev == NULL) { 452 err = -ENODEV; 453 goto errout; 454 } 455 456 __in_dev_put(in_dev); 457 458 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL; 459 ifap = &ifa->ifa_next) { 460 if (tb[IFA_LOCAL] && 461 ifa->ifa_local != nla_get_be32(tb[IFA_LOCAL])) 462 continue; 463 464 if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label)) 465 continue; 466 467 if (tb[IFA_ADDRESS] && 468 (ifm->ifa_prefixlen != ifa->ifa_prefixlen || 469 !inet_ifa_match(nla_get_be32(tb[IFA_ADDRESS]), ifa))) 470 continue; 471 472 __inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).pid); 473 return 0; 474 } 475 476 err = -EADDRNOTAVAIL; 477 errout: 478 return err; 479 } 480 481 static struct in_ifaddr *rtm_to_ifaddr(struct net *net, struct nlmsghdr *nlh) 482 { 483 struct nlattr *tb[IFA_MAX+1]; 484 struct in_ifaddr *ifa; 485 struct ifaddrmsg *ifm; 486 struct net_device *dev; 487 struct in_device *in_dev; 488 int err; 489 490 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy); 491 if (err < 0) 492 goto errout; 493 494 ifm = nlmsg_data(nlh); 495 err = -EINVAL; 496 if (ifm->ifa_prefixlen > 32 || tb[IFA_LOCAL] == NULL) 497 goto errout; 498 499 dev = __dev_get_by_index(net, ifm->ifa_index); 500 err = -ENODEV; 501 if (dev == NULL) 502 goto errout; 503 504 in_dev = __in_dev_get_rtnl(dev); 505 err = -ENOBUFS; 506 if (in_dev == NULL) 507 goto errout; 508 509 ifa = inet_alloc_ifa(); 510 if (ifa == NULL) 511 /* 512 * A potential indev allocation can be left alive, it stays 513 * assigned to its device and is destroy with it. 514 */ 515 goto errout; 516 517 ipv4_devconf_setall(in_dev); 518 in_dev_hold(in_dev); 519 520 if (tb[IFA_ADDRESS] == NULL) 521 tb[IFA_ADDRESS] = tb[IFA_LOCAL]; 522 523 ifa->ifa_prefixlen = ifm->ifa_prefixlen; 524 ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen); 525 ifa->ifa_flags = ifm->ifa_flags; 526 ifa->ifa_scope = ifm->ifa_scope; 527 ifa->ifa_dev = in_dev; 528 529 ifa->ifa_local = nla_get_be32(tb[IFA_LOCAL]); 530 ifa->ifa_address = nla_get_be32(tb[IFA_ADDRESS]); 531 532 if (tb[IFA_BROADCAST]) 533 ifa->ifa_broadcast = nla_get_be32(tb[IFA_BROADCAST]); 534 535 if (tb[IFA_LABEL]) 536 nla_strlcpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ); 537 else 538 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ); 539 540 return ifa; 541 542 errout: 543 return ERR_PTR(err); 544 } 545 546 static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg) 547 { 548 struct net *net = sock_net(skb->sk); 549 struct in_ifaddr *ifa; 550 551 ASSERT_RTNL(); 552 553 ifa = rtm_to_ifaddr(net, nlh); 554 if (IS_ERR(ifa)) 555 return PTR_ERR(ifa); 556 557 return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).pid); 558 } 559 560 /* 561 * Determine a default network mask, based on the IP address. 562 */ 563 564 static inline int inet_abc_len(__be32 addr) 565 { 566 int rc = -1; /* Something else, probably a multicast. */ 567 568 if (ipv4_is_zeronet(addr)) 569 rc = 0; 570 else { 571 __u32 haddr = ntohl(addr); 572 573 if (IN_CLASSA(haddr)) 574 rc = 8; 575 else if (IN_CLASSB(haddr)) 576 rc = 16; 577 else if (IN_CLASSC(haddr)) 578 rc = 24; 579 } 580 581 return rc; 582 } 583 584 585 int devinet_ioctl(struct net *net, unsigned int cmd, void __user *arg) 586 { 587 struct ifreq ifr; 588 struct sockaddr_in sin_orig; 589 struct sockaddr_in *sin = (struct sockaddr_in *)&ifr.ifr_addr; 590 struct in_device *in_dev; 591 struct in_ifaddr **ifap = NULL; 592 struct in_ifaddr *ifa = NULL; 593 struct net_device *dev; 594 char *colon; 595 int ret = -EFAULT; 596 int tryaddrmatch = 0; 597 598 /* 599 * Fetch the caller's info block into kernel space 600 */ 601 602 if (copy_from_user(&ifr, arg, sizeof(struct ifreq))) 603 goto out; 604 ifr.ifr_name[IFNAMSIZ - 1] = 0; 605 606 /* save original address for comparison */ 607 memcpy(&sin_orig, sin, sizeof(*sin)); 608 609 colon = strchr(ifr.ifr_name, ':'); 610 if (colon) 611 *colon = 0; 612 613 dev_load(net, ifr.ifr_name); 614 615 switch (cmd) { 616 case SIOCGIFADDR: /* Get interface address */ 617 case SIOCGIFBRDADDR: /* Get the broadcast address */ 618 case SIOCGIFDSTADDR: /* Get the destination address */ 619 case SIOCGIFNETMASK: /* Get the netmask for the interface */ 620 /* Note that these ioctls will not sleep, 621 so that we do not impose a lock. 622 One day we will be forced to put shlock here (I mean SMP) 623 */ 624 tryaddrmatch = (sin_orig.sin_family == AF_INET); 625 memset(sin, 0, sizeof(*sin)); 626 sin->sin_family = AF_INET; 627 break; 628 629 case SIOCSIFFLAGS: 630 ret = -EACCES; 631 if (!capable(CAP_NET_ADMIN)) 632 goto out; 633 break; 634 case SIOCSIFADDR: /* Set interface address (and family) */ 635 case SIOCSIFBRDADDR: /* Set the broadcast address */ 636 case SIOCSIFDSTADDR: /* Set the destination address */ 637 case SIOCSIFNETMASK: /* Set the netmask for the interface */ 638 ret = -EACCES; 639 if (!capable(CAP_NET_ADMIN)) 640 goto out; 641 ret = -EINVAL; 642 if (sin->sin_family != AF_INET) 643 goto out; 644 break; 645 default: 646 ret = -EINVAL; 647 goto out; 648 } 649 650 rtnl_lock(); 651 652 ret = -ENODEV; 653 dev = __dev_get_by_name(net, ifr.ifr_name); 654 if (!dev) 655 goto done; 656 657 if (colon) 658 *colon = ':'; 659 660 in_dev = __in_dev_get_rtnl(dev); 661 if (in_dev) { 662 if (tryaddrmatch) { 663 /* Matthias Andree */ 664 /* compare label and address (4.4BSD style) */ 665 /* note: we only do this for a limited set of ioctls 666 and only if the original address family was AF_INET. 667 This is checked above. */ 668 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL; 669 ifap = &ifa->ifa_next) { 670 if (!strcmp(ifr.ifr_name, ifa->ifa_label) && 671 sin_orig.sin_addr.s_addr == 672 ifa->ifa_address) { 673 break; /* found */ 674 } 675 } 676 } 677 /* we didn't get a match, maybe the application is 678 4.3BSD-style and passed in junk so we fall back to 679 comparing just the label */ 680 if (!ifa) { 681 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL; 682 ifap = &ifa->ifa_next) 683 if (!strcmp(ifr.ifr_name, ifa->ifa_label)) 684 break; 685 } 686 } 687 688 ret = -EADDRNOTAVAIL; 689 if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS) 690 goto done; 691 692 switch (cmd) { 693 case SIOCGIFADDR: /* Get interface address */ 694 sin->sin_addr.s_addr = ifa->ifa_local; 695 goto rarok; 696 697 case SIOCGIFBRDADDR: /* Get the broadcast address */ 698 sin->sin_addr.s_addr = ifa->ifa_broadcast; 699 goto rarok; 700 701 case SIOCGIFDSTADDR: /* Get the destination address */ 702 sin->sin_addr.s_addr = ifa->ifa_address; 703 goto rarok; 704 705 case SIOCGIFNETMASK: /* Get the netmask for the interface */ 706 sin->sin_addr.s_addr = ifa->ifa_mask; 707 goto rarok; 708 709 case SIOCSIFFLAGS: 710 if (colon) { 711 ret = -EADDRNOTAVAIL; 712 if (!ifa) 713 break; 714 ret = 0; 715 if (!(ifr.ifr_flags & IFF_UP)) 716 inet_del_ifa(in_dev, ifap, 1); 717 break; 718 } 719 ret = dev_change_flags(dev, ifr.ifr_flags); 720 break; 721 722 case SIOCSIFADDR: /* Set interface address (and family) */ 723 ret = -EINVAL; 724 if (inet_abc_len(sin->sin_addr.s_addr) < 0) 725 break; 726 727 if (!ifa) { 728 ret = -ENOBUFS; 729 ifa = inet_alloc_ifa(); 730 if (!ifa) 731 break; 732 if (colon) 733 memcpy(ifa->ifa_label, ifr.ifr_name, IFNAMSIZ); 734 else 735 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ); 736 } else { 737 ret = 0; 738 if (ifa->ifa_local == sin->sin_addr.s_addr) 739 break; 740 inet_del_ifa(in_dev, ifap, 0); 741 ifa->ifa_broadcast = 0; 742 ifa->ifa_scope = 0; 743 } 744 745 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr; 746 747 if (!(dev->flags & IFF_POINTOPOINT)) { 748 ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address); 749 ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen); 750 if ((dev->flags & IFF_BROADCAST) && 751 ifa->ifa_prefixlen < 31) 752 ifa->ifa_broadcast = ifa->ifa_address | 753 ~ifa->ifa_mask; 754 } else { 755 ifa->ifa_prefixlen = 32; 756 ifa->ifa_mask = inet_make_mask(32); 757 } 758 ret = inet_set_ifa(dev, ifa); 759 break; 760 761 case SIOCSIFBRDADDR: /* Set the broadcast address */ 762 ret = 0; 763 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) { 764 inet_del_ifa(in_dev, ifap, 0); 765 ifa->ifa_broadcast = sin->sin_addr.s_addr; 766 inet_insert_ifa(ifa); 767 } 768 break; 769 770 case SIOCSIFDSTADDR: /* Set the destination address */ 771 ret = 0; 772 if (ifa->ifa_address == sin->sin_addr.s_addr) 773 break; 774 ret = -EINVAL; 775 if (inet_abc_len(sin->sin_addr.s_addr) < 0) 776 break; 777 ret = 0; 778 inet_del_ifa(in_dev, ifap, 0); 779 ifa->ifa_address = sin->sin_addr.s_addr; 780 inet_insert_ifa(ifa); 781 break; 782 783 case SIOCSIFNETMASK: /* Set the netmask for the interface */ 784 785 /* 786 * The mask we set must be legal. 787 */ 788 ret = -EINVAL; 789 if (bad_mask(sin->sin_addr.s_addr, 0)) 790 break; 791 ret = 0; 792 if (ifa->ifa_mask != sin->sin_addr.s_addr) { 793 __be32 old_mask = ifa->ifa_mask; 794 inet_del_ifa(in_dev, ifap, 0); 795 ifa->ifa_mask = sin->sin_addr.s_addr; 796 ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask); 797 798 /* See if current broadcast address matches 799 * with current netmask, then recalculate 800 * the broadcast address. Otherwise it's a 801 * funny address, so don't touch it since 802 * the user seems to know what (s)he's doing... 803 */ 804 if ((dev->flags & IFF_BROADCAST) && 805 (ifa->ifa_prefixlen < 31) && 806 (ifa->ifa_broadcast == 807 (ifa->ifa_local|~old_mask))) { 808 ifa->ifa_broadcast = (ifa->ifa_local | 809 ~sin->sin_addr.s_addr); 810 } 811 inet_insert_ifa(ifa); 812 } 813 break; 814 } 815 done: 816 rtnl_unlock(); 817 out: 818 return ret; 819 rarok: 820 rtnl_unlock(); 821 ret = copy_to_user(arg, &ifr, sizeof(struct ifreq)) ? -EFAULT : 0; 822 goto out; 823 } 824 825 static int inet_gifconf(struct net_device *dev, char __user *buf, int len) 826 { 827 struct in_device *in_dev = __in_dev_get_rtnl(dev); 828 struct in_ifaddr *ifa; 829 struct ifreq ifr; 830 int done = 0; 831 832 if (!in_dev) 833 goto out; 834 835 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) { 836 if (!buf) { 837 done += sizeof(ifr); 838 continue; 839 } 840 if (len < (int) sizeof(ifr)) 841 break; 842 memset(&ifr, 0, sizeof(struct ifreq)); 843 if (ifa->ifa_label) 844 strcpy(ifr.ifr_name, ifa->ifa_label); 845 else 846 strcpy(ifr.ifr_name, dev->name); 847 848 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET; 849 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr = 850 ifa->ifa_local; 851 852 if (copy_to_user(buf, &ifr, sizeof(struct ifreq))) { 853 done = -EFAULT; 854 break; 855 } 856 buf += sizeof(struct ifreq); 857 len -= sizeof(struct ifreq); 858 done += sizeof(struct ifreq); 859 } 860 out: 861 return done; 862 } 863 864 __be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope) 865 { 866 __be32 addr = 0; 867 struct in_device *in_dev; 868 struct net *net = dev_net(dev); 869 870 rcu_read_lock(); 871 in_dev = __in_dev_get_rcu(dev); 872 if (!in_dev) 873 goto no_in_dev; 874 875 for_primary_ifa(in_dev) { 876 if (ifa->ifa_scope > scope) 877 continue; 878 if (!dst || inet_ifa_match(dst, ifa)) { 879 addr = ifa->ifa_local; 880 break; 881 } 882 if (!addr) 883 addr = ifa->ifa_local; 884 } endfor_ifa(in_dev); 885 886 if (addr) 887 goto out_unlock; 888 no_in_dev: 889 890 /* Not loopback addresses on loopback should be preferred 891 in this case. It is importnat that lo is the first interface 892 in dev_base list. 893 */ 894 for_each_netdev_rcu(net, dev) { 895 in_dev = __in_dev_get_rcu(dev); 896 if (!in_dev) 897 continue; 898 899 for_primary_ifa(in_dev) { 900 if (ifa->ifa_scope != RT_SCOPE_LINK && 901 ifa->ifa_scope <= scope) { 902 addr = ifa->ifa_local; 903 goto out_unlock; 904 } 905 } endfor_ifa(in_dev); 906 } 907 out_unlock: 908 rcu_read_unlock(); 909 return addr; 910 } 911 EXPORT_SYMBOL(inet_select_addr); 912 913 static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst, 914 __be32 local, int scope) 915 { 916 int same = 0; 917 __be32 addr = 0; 918 919 for_ifa(in_dev) { 920 if (!addr && 921 (local == ifa->ifa_local || !local) && 922 ifa->ifa_scope <= scope) { 923 addr = ifa->ifa_local; 924 if (same) 925 break; 926 } 927 if (!same) { 928 same = (!local || inet_ifa_match(local, ifa)) && 929 (!dst || inet_ifa_match(dst, ifa)); 930 if (same && addr) { 931 if (local || !dst) 932 break; 933 /* Is the selected addr into dst subnet? */ 934 if (inet_ifa_match(addr, ifa)) 935 break; 936 /* No, then can we use new local src? */ 937 if (ifa->ifa_scope <= scope) { 938 addr = ifa->ifa_local; 939 break; 940 } 941 /* search for large dst subnet for addr */ 942 same = 0; 943 } 944 } 945 } endfor_ifa(in_dev); 946 947 return same ? addr : 0; 948 } 949 950 /* 951 * Confirm that local IP address exists using wildcards: 952 * - in_dev: only on this interface, 0=any interface 953 * - dst: only in the same subnet as dst, 0=any dst 954 * - local: address, 0=autoselect the local address 955 * - scope: maximum allowed scope value for the local address 956 */ 957 __be32 inet_confirm_addr(struct in_device *in_dev, 958 __be32 dst, __be32 local, int scope) 959 { 960 __be32 addr = 0; 961 struct net_device *dev; 962 struct net *net; 963 964 if (scope != RT_SCOPE_LINK) 965 return confirm_addr_indev(in_dev, dst, local, scope); 966 967 net = dev_net(in_dev->dev); 968 rcu_read_lock(); 969 for_each_netdev_rcu(net, dev) { 970 in_dev = __in_dev_get_rcu(dev); 971 if (in_dev) { 972 addr = confirm_addr_indev(in_dev, dst, local, scope); 973 if (addr) 974 break; 975 } 976 } 977 rcu_read_unlock(); 978 979 return addr; 980 } 981 982 /* 983 * Device notifier 984 */ 985 986 int register_inetaddr_notifier(struct notifier_block *nb) 987 { 988 return blocking_notifier_chain_register(&inetaddr_chain, nb); 989 } 990 EXPORT_SYMBOL(register_inetaddr_notifier); 991 992 int unregister_inetaddr_notifier(struct notifier_block *nb) 993 { 994 return blocking_notifier_chain_unregister(&inetaddr_chain, nb); 995 } 996 EXPORT_SYMBOL(unregister_inetaddr_notifier); 997 998 /* Rename ifa_labels for a device name change. Make some effort to preserve 999 * existing alias numbering and to create unique labels if possible. 1000 */ 1001 static void inetdev_changename(struct net_device *dev, struct in_device *in_dev) 1002 { 1003 struct in_ifaddr *ifa; 1004 int named = 0; 1005 1006 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) { 1007 char old[IFNAMSIZ], *dot; 1008 1009 memcpy(old, ifa->ifa_label, IFNAMSIZ); 1010 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ); 1011 if (named++ == 0) 1012 goto skip; 1013 dot = strchr(old, ':'); 1014 if (dot == NULL) { 1015 sprintf(old, ":%d", named); 1016 dot = old; 1017 } 1018 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ) 1019 strcat(ifa->ifa_label, dot); 1020 else 1021 strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot); 1022 skip: 1023 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0); 1024 } 1025 } 1026 1027 static inline bool inetdev_valid_mtu(unsigned mtu) 1028 { 1029 return mtu >= 68; 1030 } 1031 1032 /* Called only under RTNL semaphore */ 1033 1034 static int inetdev_event(struct notifier_block *this, unsigned long event, 1035 void *ptr) 1036 { 1037 struct net_device *dev = ptr; 1038 struct in_device *in_dev = __in_dev_get_rtnl(dev); 1039 1040 ASSERT_RTNL(); 1041 1042 if (!in_dev) { 1043 if (event == NETDEV_REGISTER) { 1044 in_dev = inetdev_init(dev); 1045 if (!in_dev) 1046 return notifier_from_errno(-ENOMEM); 1047 if (dev->flags & IFF_LOOPBACK) { 1048 IN_DEV_CONF_SET(in_dev, NOXFRM, 1); 1049 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1); 1050 } 1051 } else if (event == NETDEV_CHANGEMTU) { 1052 /* Re-enabling IP */ 1053 if (inetdev_valid_mtu(dev->mtu)) 1054 in_dev = inetdev_init(dev); 1055 } 1056 goto out; 1057 } 1058 1059 switch (event) { 1060 case NETDEV_REGISTER: 1061 printk(KERN_DEBUG "inetdev_event: bug\n"); 1062 dev->ip_ptr = NULL; 1063 break; 1064 case NETDEV_UP: 1065 if (!inetdev_valid_mtu(dev->mtu)) 1066 break; 1067 if (dev->flags & IFF_LOOPBACK) { 1068 struct in_ifaddr *ifa = inet_alloc_ifa(); 1069 1070 if (ifa) { 1071 ifa->ifa_local = 1072 ifa->ifa_address = htonl(INADDR_LOOPBACK); 1073 ifa->ifa_prefixlen = 8; 1074 ifa->ifa_mask = inet_make_mask(8); 1075 in_dev_hold(in_dev); 1076 ifa->ifa_dev = in_dev; 1077 ifa->ifa_scope = RT_SCOPE_HOST; 1078 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ); 1079 inet_insert_ifa(ifa); 1080 } 1081 } 1082 ip_mc_up(in_dev); 1083 /* fall through */ 1084 case NETDEV_NOTIFY_PEERS: 1085 case NETDEV_CHANGEADDR: 1086 /* Send gratuitous ARP to notify of link change */ 1087 if (IN_DEV_ARP_NOTIFY(in_dev)) { 1088 struct in_ifaddr *ifa = in_dev->ifa_list; 1089 1090 if (ifa) 1091 arp_send(ARPOP_REQUEST, ETH_P_ARP, 1092 ifa->ifa_address, dev, 1093 ifa->ifa_address, NULL, 1094 dev->dev_addr, NULL); 1095 } 1096 break; 1097 case NETDEV_DOWN: 1098 ip_mc_down(in_dev); 1099 break; 1100 case NETDEV_PRE_TYPE_CHANGE: 1101 ip_mc_unmap(in_dev); 1102 break; 1103 case NETDEV_POST_TYPE_CHANGE: 1104 ip_mc_remap(in_dev); 1105 break; 1106 case NETDEV_CHANGEMTU: 1107 if (inetdev_valid_mtu(dev->mtu)) 1108 break; 1109 /* disable IP when MTU is not enough */ 1110 case NETDEV_UNREGISTER: 1111 inetdev_destroy(in_dev); 1112 break; 1113 case NETDEV_CHANGENAME: 1114 /* Do not notify about label change, this event is 1115 * not interesting to applications using netlink. 1116 */ 1117 inetdev_changename(dev, in_dev); 1118 1119 devinet_sysctl_unregister(in_dev); 1120 devinet_sysctl_register(in_dev); 1121 break; 1122 } 1123 out: 1124 return NOTIFY_DONE; 1125 } 1126 1127 static struct notifier_block ip_netdev_notifier = { 1128 .notifier_call = inetdev_event, 1129 }; 1130 1131 static inline size_t inet_nlmsg_size(void) 1132 { 1133 return NLMSG_ALIGN(sizeof(struct ifaddrmsg)) 1134 + nla_total_size(4) /* IFA_ADDRESS */ 1135 + nla_total_size(4) /* IFA_LOCAL */ 1136 + nla_total_size(4) /* IFA_BROADCAST */ 1137 + nla_total_size(IFNAMSIZ); /* IFA_LABEL */ 1138 } 1139 1140 static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa, 1141 u32 pid, u32 seq, int event, unsigned int flags) 1142 { 1143 struct ifaddrmsg *ifm; 1144 struct nlmsghdr *nlh; 1145 1146 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*ifm), flags); 1147 if (nlh == NULL) 1148 return -EMSGSIZE; 1149 1150 ifm = nlmsg_data(nlh); 1151 ifm->ifa_family = AF_INET; 1152 ifm->ifa_prefixlen = ifa->ifa_prefixlen; 1153 ifm->ifa_flags = ifa->ifa_flags|IFA_F_PERMANENT; 1154 ifm->ifa_scope = ifa->ifa_scope; 1155 ifm->ifa_index = ifa->ifa_dev->dev->ifindex; 1156 1157 if (ifa->ifa_address) 1158 NLA_PUT_BE32(skb, IFA_ADDRESS, ifa->ifa_address); 1159 1160 if (ifa->ifa_local) 1161 NLA_PUT_BE32(skb, IFA_LOCAL, ifa->ifa_local); 1162 1163 if (ifa->ifa_broadcast) 1164 NLA_PUT_BE32(skb, IFA_BROADCAST, ifa->ifa_broadcast); 1165 1166 if (ifa->ifa_label[0]) 1167 NLA_PUT_STRING(skb, IFA_LABEL, ifa->ifa_label); 1168 1169 return nlmsg_end(skb, nlh); 1170 1171 nla_put_failure: 1172 nlmsg_cancel(skb, nlh); 1173 return -EMSGSIZE; 1174 } 1175 1176 static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb) 1177 { 1178 struct net *net = sock_net(skb->sk); 1179 int h, s_h; 1180 int idx, s_idx; 1181 int ip_idx, s_ip_idx; 1182 struct net_device *dev; 1183 struct in_device *in_dev; 1184 struct in_ifaddr *ifa; 1185 struct hlist_head *head; 1186 struct hlist_node *node; 1187 1188 s_h = cb->args[0]; 1189 s_idx = idx = cb->args[1]; 1190 s_ip_idx = ip_idx = cb->args[2]; 1191 1192 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 1193 idx = 0; 1194 head = &net->dev_index_head[h]; 1195 rcu_read_lock(); 1196 hlist_for_each_entry_rcu(dev, node, head, index_hlist) { 1197 if (idx < s_idx) 1198 goto cont; 1199 if (h > s_h || idx > s_idx) 1200 s_ip_idx = 0; 1201 in_dev = __in_dev_get_rcu(dev); 1202 if (!in_dev) 1203 goto cont; 1204 1205 for (ifa = in_dev->ifa_list, ip_idx = 0; ifa; 1206 ifa = ifa->ifa_next, ip_idx++) { 1207 if (ip_idx < s_ip_idx) 1208 continue; 1209 if (inet_fill_ifaddr(skb, ifa, 1210 NETLINK_CB(cb->skb).pid, 1211 cb->nlh->nlmsg_seq, 1212 RTM_NEWADDR, NLM_F_MULTI) <= 0) { 1213 rcu_read_unlock(); 1214 goto done; 1215 } 1216 } 1217 cont: 1218 idx++; 1219 } 1220 rcu_read_unlock(); 1221 } 1222 1223 done: 1224 cb->args[0] = h; 1225 cb->args[1] = idx; 1226 cb->args[2] = ip_idx; 1227 1228 return skb->len; 1229 } 1230 1231 static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh, 1232 u32 pid) 1233 { 1234 struct sk_buff *skb; 1235 u32 seq = nlh ? nlh->nlmsg_seq : 0; 1236 int err = -ENOBUFS; 1237 struct net *net; 1238 1239 net = dev_net(ifa->ifa_dev->dev); 1240 skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL); 1241 if (skb == NULL) 1242 goto errout; 1243 1244 err = inet_fill_ifaddr(skb, ifa, pid, seq, event, 0); 1245 if (err < 0) { 1246 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */ 1247 WARN_ON(err == -EMSGSIZE); 1248 kfree_skb(skb); 1249 goto errout; 1250 } 1251 rtnl_notify(skb, net, pid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL); 1252 return; 1253 errout: 1254 if (err < 0) 1255 rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err); 1256 } 1257 1258 #ifdef CONFIG_SYSCTL 1259 1260 static void devinet_copy_dflt_conf(struct net *net, int i) 1261 { 1262 struct net_device *dev; 1263 1264 rcu_read_lock(); 1265 for_each_netdev_rcu(net, dev) { 1266 struct in_device *in_dev; 1267 1268 in_dev = __in_dev_get_rcu(dev); 1269 if (in_dev && !test_bit(i, in_dev->cnf.state)) 1270 in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i]; 1271 } 1272 rcu_read_unlock(); 1273 } 1274 1275 /* called with RTNL locked */ 1276 static void inet_forward_change(struct net *net) 1277 { 1278 struct net_device *dev; 1279 int on = IPV4_DEVCONF_ALL(net, FORWARDING); 1280 1281 IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on; 1282 IPV4_DEVCONF_DFLT(net, FORWARDING) = on; 1283 1284 for_each_netdev(net, dev) { 1285 struct in_device *in_dev; 1286 if (on) 1287 dev_disable_lro(dev); 1288 rcu_read_lock(); 1289 in_dev = __in_dev_get_rcu(dev); 1290 if (in_dev) 1291 IN_DEV_CONF_SET(in_dev, FORWARDING, on); 1292 rcu_read_unlock(); 1293 } 1294 } 1295 1296 static int devinet_conf_proc(ctl_table *ctl, int write, 1297 void __user *buffer, 1298 size_t *lenp, loff_t *ppos) 1299 { 1300 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos); 1301 1302 if (write) { 1303 struct ipv4_devconf *cnf = ctl->extra1; 1304 struct net *net = ctl->extra2; 1305 int i = (int *)ctl->data - cnf->data; 1306 1307 set_bit(i, cnf->state); 1308 1309 if (cnf == net->ipv4.devconf_dflt) 1310 devinet_copy_dflt_conf(net, i); 1311 } 1312 1313 return ret; 1314 } 1315 1316 static int devinet_sysctl_forward(ctl_table *ctl, int write, 1317 void __user *buffer, 1318 size_t *lenp, loff_t *ppos) 1319 { 1320 int *valp = ctl->data; 1321 int val = *valp; 1322 loff_t pos = *ppos; 1323 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos); 1324 1325 if (write && *valp != val) { 1326 struct net *net = ctl->extra2; 1327 1328 if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) { 1329 if (!rtnl_trylock()) { 1330 /* Restore the original values before restarting */ 1331 *valp = val; 1332 *ppos = pos; 1333 return restart_syscall(); 1334 } 1335 if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) { 1336 inet_forward_change(net); 1337 } else if (*valp) { 1338 struct ipv4_devconf *cnf = ctl->extra1; 1339 struct in_device *idev = 1340 container_of(cnf, struct in_device, cnf); 1341 dev_disable_lro(idev->dev); 1342 } 1343 rtnl_unlock(); 1344 rt_cache_flush(net, 0); 1345 } 1346 } 1347 1348 return ret; 1349 } 1350 1351 int ipv4_doint_and_flush(ctl_table *ctl, int write, 1352 void __user *buffer, 1353 size_t *lenp, loff_t *ppos) 1354 { 1355 int *valp = ctl->data; 1356 int val = *valp; 1357 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos); 1358 struct net *net = ctl->extra2; 1359 1360 if (write && *valp != val) 1361 rt_cache_flush(net, 0); 1362 1363 return ret; 1364 } 1365 1366 #define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \ 1367 { \ 1368 .procname = name, \ 1369 .data = ipv4_devconf.data + \ 1370 IPV4_DEVCONF_ ## attr - 1, \ 1371 .maxlen = sizeof(int), \ 1372 .mode = mval, \ 1373 .proc_handler = proc, \ 1374 .extra1 = &ipv4_devconf, \ 1375 } 1376 1377 #define DEVINET_SYSCTL_RW_ENTRY(attr, name) \ 1378 DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc) 1379 1380 #define DEVINET_SYSCTL_RO_ENTRY(attr, name) \ 1381 DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc) 1382 1383 #define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \ 1384 DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc) 1385 1386 #define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \ 1387 DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush) 1388 1389 static struct devinet_sysctl_table { 1390 struct ctl_table_header *sysctl_header; 1391 struct ctl_table devinet_vars[__IPV4_DEVCONF_MAX]; 1392 char *dev_name; 1393 } devinet_sysctl = { 1394 .devinet_vars = { 1395 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding", 1396 devinet_sysctl_forward), 1397 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"), 1398 1399 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"), 1400 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"), 1401 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"), 1402 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"), 1403 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"), 1404 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE, 1405 "accept_source_route"), 1406 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"), 1407 DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"), 1408 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"), 1409 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"), 1410 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"), 1411 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"), 1412 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"), 1413 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"), 1414 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"), 1415 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"), 1416 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"), 1417 DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"), 1418 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"), 1419 1420 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"), 1421 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"), 1422 DEVINET_SYSCTL_FLUSHING_ENTRY(FORCE_IGMP_VERSION, 1423 "force_igmp_version"), 1424 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES, 1425 "promote_secondaries"), 1426 }, 1427 }; 1428 1429 static int __devinet_sysctl_register(struct net *net, char *dev_name, 1430 struct ipv4_devconf *p) 1431 { 1432 int i; 1433 struct devinet_sysctl_table *t; 1434 1435 #define DEVINET_CTL_PATH_DEV 3 1436 1437 struct ctl_path devinet_ctl_path[] = { 1438 { .procname = "net", }, 1439 { .procname = "ipv4", }, 1440 { .procname = "conf", }, 1441 { /* to be set */ }, 1442 { }, 1443 }; 1444 1445 t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL); 1446 if (!t) 1447 goto out; 1448 1449 for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) { 1450 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf; 1451 t->devinet_vars[i].extra1 = p; 1452 t->devinet_vars[i].extra2 = net; 1453 } 1454 1455 /* 1456 * Make a copy of dev_name, because '.procname' is regarded as const 1457 * by sysctl and we wouldn't want anyone to change it under our feet 1458 * (see SIOCSIFNAME). 1459 */ 1460 t->dev_name = kstrdup(dev_name, GFP_KERNEL); 1461 if (!t->dev_name) 1462 goto free; 1463 1464 devinet_ctl_path[DEVINET_CTL_PATH_DEV].procname = t->dev_name; 1465 1466 t->sysctl_header = register_net_sysctl_table(net, devinet_ctl_path, 1467 t->devinet_vars); 1468 if (!t->sysctl_header) 1469 goto free_procname; 1470 1471 p->sysctl = t; 1472 return 0; 1473 1474 free_procname: 1475 kfree(t->dev_name); 1476 free: 1477 kfree(t); 1478 out: 1479 return -ENOBUFS; 1480 } 1481 1482 static void __devinet_sysctl_unregister(struct ipv4_devconf *cnf) 1483 { 1484 struct devinet_sysctl_table *t = cnf->sysctl; 1485 1486 if (t == NULL) 1487 return; 1488 1489 cnf->sysctl = NULL; 1490 unregister_sysctl_table(t->sysctl_header); 1491 kfree(t->dev_name); 1492 kfree(t); 1493 } 1494 1495 static void devinet_sysctl_register(struct in_device *idev) 1496 { 1497 neigh_sysctl_register(idev->dev, idev->arp_parms, "ipv4", NULL); 1498 __devinet_sysctl_register(dev_net(idev->dev), idev->dev->name, 1499 &idev->cnf); 1500 } 1501 1502 static void devinet_sysctl_unregister(struct in_device *idev) 1503 { 1504 __devinet_sysctl_unregister(&idev->cnf); 1505 neigh_sysctl_unregister(idev->arp_parms); 1506 } 1507 1508 static struct ctl_table ctl_forward_entry[] = { 1509 { 1510 .procname = "ip_forward", 1511 .data = &ipv4_devconf.data[ 1512 IPV4_DEVCONF_FORWARDING - 1], 1513 .maxlen = sizeof(int), 1514 .mode = 0644, 1515 .proc_handler = devinet_sysctl_forward, 1516 .extra1 = &ipv4_devconf, 1517 .extra2 = &init_net, 1518 }, 1519 { }, 1520 }; 1521 1522 static __net_initdata struct ctl_path net_ipv4_path[] = { 1523 { .procname = "net", }, 1524 { .procname = "ipv4", }, 1525 { }, 1526 }; 1527 #endif 1528 1529 static __net_init int devinet_init_net(struct net *net) 1530 { 1531 int err; 1532 struct ipv4_devconf *all, *dflt; 1533 #ifdef CONFIG_SYSCTL 1534 struct ctl_table *tbl = ctl_forward_entry; 1535 struct ctl_table_header *forw_hdr; 1536 #endif 1537 1538 err = -ENOMEM; 1539 all = &ipv4_devconf; 1540 dflt = &ipv4_devconf_dflt; 1541 1542 if (!net_eq(net, &init_net)) { 1543 all = kmemdup(all, sizeof(ipv4_devconf), GFP_KERNEL); 1544 if (all == NULL) 1545 goto err_alloc_all; 1546 1547 dflt = kmemdup(dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL); 1548 if (dflt == NULL) 1549 goto err_alloc_dflt; 1550 1551 #ifdef CONFIG_SYSCTL 1552 tbl = kmemdup(tbl, sizeof(ctl_forward_entry), GFP_KERNEL); 1553 if (tbl == NULL) 1554 goto err_alloc_ctl; 1555 1556 tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1]; 1557 tbl[0].extra1 = all; 1558 tbl[0].extra2 = net; 1559 #endif 1560 } 1561 1562 #ifdef CONFIG_SYSCTL 1563 err = __devinet_sysctl_register(net, "all", all); 1564 if (err < 0) 1565 goto err_reg_all; 1566 1567 err = __devinet_sysctl_register(net, "default", dflt); 1568 if (err < 0) 1569 goto err_reg_dflt; 1570 1571 err = -ENOMEM; 1572 forw_hdr = register_net_sysctl_table(net, net_ipv4_path, tbl); 1573 if (forw_hdr == NULL) 1574 goto err_reg_ctl; 1575 net->ipv4.forw_hdr = forw_hdr; 1576 #endif 1577 1578 net->ipv4.devconf_all = all; 1579 net->ipv4.devconf_dflt = dflt; 1580 return 0; 1581 1582 #ifdef CONFIG_SYSCTL 1583 err_reg_ctl: 1584 __devinet_sysctl_unregister(dflt); 1585 err_reg_dflt: 1586 __devinet_sysctl_unregister(all); 1587 err_reg_all: 1588 if (tbl != ctl_forward_entry) 1589 kfree(tbl); 1590 err_alloc_ctl: 1591 #endif 1592 if (dflt != &ipv4_devconf_dflt) 1593 kfree(dflt); 1594 err_alloc_dflt: 1595 if (all != &ipv4_devconf) 1596 kfree(all); 1597 err_alloc_all: 1598 return err; 1599 } 1600 1601 static __net_exit void devinet_exit_net(struct net *net) 1602 { 1603 #ifdef CONFIG_SYSCTL 1604 struct ctl_table *tbl; 1605 1606 tbl = net->ipv4.forw_hdr->ctl_table_arg; 1607 unregister_net_sysctl_table(net->ipv4.forw_hdr); 1608 __devinet_sysctl_unregister(net->ipv4.devconf_dflt); 1609 __devinet_sysctl_unregister(net->ipv4.devconf_all); 1610 kfree(tbl); 1611 #endif 1612 kfree(net->ipv4.devconf_dflt); 1613 kfree(net->ipv4.devconf_all); 1614 } 1615 1616 static __net_initdata struct pernet_operations devinet_ops = { 1617 .init = devinet_init_net, 1618 .exit = devinet_exit_net, 1619 }; 1620 1621 void __init devinet_init(void) 1622 { 1623 register_pernet_subsys(&devinet_ops); 1624 1625 register_gifconf(PF_INET, inet_gifconf); 1626 register_netdevice_notifier(&ip_netdev_notifier); 1627 1628 rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL); 1629 rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL); 1630 rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr); 1631 } 1632 1633