net/ipv4/nexthop.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /* Generic nexthop implementation
   3  *
   4  * Copyright (c) 2017-19 Cumulus Networks
   5  * Copyright (c) 2017-19 David Ahern <dsa@cumulusnetworks.com>
   6  */
   7
   8 #include <linux/nexthop.h>
   9 #include <linux/rtnetlink.h>
  10 #include <linux/slab.h>
  11 #include <net/arp.h>
  12 #include <net/ipv6_stubs.h>
  13 #include <net/lwtunnel.h>
  14 #include <net/ndisc.h>
  15 #include <net/nexthop.h>
  16 #include <net/route.h>
  17 #include <net/sock.h>
  18
  19 static void remove_nexthop(struct net *net, struct nexthop *nh,
  20                            struct nl_info *nlinfo);
  21
  22 #define NH_DEV_HASHBITS  8
  23 #define NH_DEV_HASHSIZE (1U << NH_DEV_HASHBITS)
  24
  25 static const struct nla_policy rtm_nh_policy[NHA_MAX + 1] = {
  26         [NHA_UNSPEC]            = { .strict_start_type = NHA_UNSPEC + 1 },
  27         [NHA_ID]                = { .type = NLA_U32 },
  28         [NHA_GROUP]             = { .type = NLA_BINARY },
  29         [NHA_GROUP_TYPE]        = { .type = NLA_U16 },
  30         [NHA_BLACKHOLE]         = { .type = NLA_FLAG },
  31         [NHA_OIF]               = { .type = NLA_U32 },
  32         [NHA_GATEWAY]           = { .type = NLA_BINARY },
  33         [NHA_ENCAP_TYPE]        = { .type = NLA_U16 },
  34         [NHA_ENCAP]             = { .type = NLA_NESTED },
  35         [NHA_GROUPS]            = { .type = NLA_FLAG },
  36         [NHA_MASTER]            = { .type = NLA_U32 },
  37 };
  38
  39 static unsigned int nh_dev_hashfn(unsigned int val)
  40 {
  41         unsigned int mask = NH_DEV_HASHSIZE - 1;
  42
  43         return (val ^
  44                 (val >> NH_DEV_HASHBITS) ^
  45                 (val >> (NH_DEV_HASHBITS * 2))) & mask;
  46 }
  47
  48 static void nexthop_devhash_add(struct net *net, struct nh_info *nhi)
  49 {
  50         struct net_device *dev = nhi->fib_nhc.nhc_dev;
  51         struct hlist_head *head;
  52         unsigned int hash;
  53
  54         WARN_ON(!dev);
  55
  56         hash = nh_dev_hashfn(dev->ifindex);
  57         head = &net->nexthop.devhash[hash];
  58         hlist_add_head(&nhi->dev_hash, head);
  59 }
  60
  61 static void nexthop_free_mpath(struct nexthop *nh)
  62 {
  63         struct nh_group *nhg;
  64         int i;
  65
  66         nhg = rcu_dereference_raw(nh->nh_grp);
  67         for (i = 0; i < nhg->num_nh; ++i)
  68                 WARN_ON(nhg->nh_entries[i].nh);
  69
  70         kfree(nhg);
  71 }
  72
  73 static void nexthop_free_single(struct nexthop *nh)
  74 {
  75         struct nh_info *nhi;
  76
  77         nhi = rcu_dereference_raw(nh->nh_info);
  78         switch (nhi->family) {
  79         case AF_INET:
  80                 fib_nh_release(nh->net, &nhi->fib_nh);
  81                 break;
  82         case AF_INET6:
  83                 ipv6_stub->fib6_nh_release(&nhi->fib6_nh);
  84                 break;
  85         }
  86         kfree(nhi);
  87 }
  88
  89 void nexthop_free_rcu(struct rcu_head *head)
  90 {
  91         struct nexthop *nh = container_of(head, struct nexthop, rcu);
  92
  93         if (nh->is_group)
  94                 nexthop_free_mpath(nh);
  95         else
  96                 nexthop_free_single(nh);
  97
  98         kfree(nh);
  99 }
 100 EXPORT_SYMBOL_GPL(nexthop_free_rcu);
 101
 102 static struct nexthop *nexthop_alloc(void)
 103 {
 104         struct nexthop *nh;
 105
 106         nh = kzalloc(sizeof(struct nexthop), GFP_KERNEL);
 107         if (nh) {
 108                 INIT_LIST_HEAD(&nh->fi_list);
 109                 INIT_LIST_HEAD(&nh->grp_list);
 110         }
 111         return nh;
 112 }
 113
 114 static struct nh_group *nexthop_grp_alloc(u16 num_nh)
 115 {
 116         size_t sz = offsetof(struct nexthop, nh_grp)
 117                     + sizeof(struct nh_group)
 118                     + sizeof(struct nh_grp_entry) * num_nh;
 119         struct nh_group *nhg;
 120
 121         nhg = kzalloc(sz, GFP_KERNEL);
 122         if (nhg)
 123                 nhg->num_nh = num_nh;
 124
 125         return nhg;
 126 }
 127
 128 static void nh_base_seq_inc(struct net *net)
 129 {
 130         while (++net->nexthop.seq == 0)
 131                 ;
 132 }
 133
 134 /* no reference taken; rcu lock or rtnl must be held */
 135 struct nexthop *nexthop_find_by_id(struct net *net, u32 id)
 136 {
 137         struct rb_node **pp, *parent = NULL, *next;
 138
 139         pp = &net->nexthop.rb_root.rb_node;
 140         while (1) {
 141                 struct nexthop *nh;
 142
 143                 next = rcu_dereference_raw(*pp);
 144                 if (!next)
 145                         break;
 146                 parent = next;
 147
 148                 nh = rb_entry(parent, struct nexthop, rb_node);
 149                 if (id < nh->id)
 150                         pp = &next->rb_left;
 151                 else if (id > nh->id)
 152                         pp = &next->rb_right;
 153                 else
 154                         return nh;
 155         }
 156         return NULL;
 157 }
 158 EXPORT_SYMBOL_GPL(nexthop_find_by_id);
 159
 160 /* used for auto id allocation; called with rtnl held */
 161 static u32 nh_find_unused_id(struct net *net)
 162 {
 163         u32 id_start = net->nexthop.last_id_allocated;
 164
 165         while (1) {
 166                 net->nexthop.last_id_allocated++;
 167                 if (net->nexthop.last_id_allocated == id_start)
 168                         break;
 169
 170                 if (!nexthop_find_by_id(net, net->nexthop.last_id_allocated))
 171                         return net->nexthop.last_id_allocated;
 172         }
 173         return 0;
 174 }
 175
 176 static int nla_put_nh_group(struct sk_buff *skb, struct nh_group *nhg)
 177 {
 178         struct nexthop_grp *p;
 179         size_t len = nhg->num_nh * sizeof(*p);
 180         struct nlattr *nla;
 181         u16 group_type = 0;
 182         int i;
 183
 184         if (nhg->mpath)
 185                 group_type = NEXTHOP_GRP_TYPE_MPATH;
 186
 187         if (nla_put_u16(skb, NHA_GROUP_TYPE, group_type))
 188                 goto nla_put_failure;
 189
 190         nla = nla_reserve(skb, NHA_GROUP, len);
 191         if (!nla)
 192                 goto nla_put_failure;
 193
 194         p = nla_data(nla);
 195         for (i = 0; i < nhg->num_nh; ++i) {
 196                 p->id = nhg->nh_entries[i].nh->id;
 197                 p->weight = nhg->nh_entries[i].weight - 1;
 198                 p += 1;
 199         }
 200
 201         return 0;
 202
 203 nla_put_failure:
 204         return -EMSGSIZE;
 205 }
 206
 207 static int nh_fill_node(struct sk_buff *skb, struct nexthop *nh,
 208                         int event, u32 portid, u32 seq, unsigned int nlflags)
 209 {
 210         struct fib6_nh *fib6_nh;
 211         struct fib_nh *fib_nh;
 212         struct nlmsghdr *nlh;
 213         struct nh_info *nhi;
 214         struct nhmsg *nhm;
 215
 216         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*nhm), nlflags);
 217         if (!nlh)
 218                 return -EMSGSIZE;
 219
 220         nhm = nlmsg_data(nlh);
 221         nhm->nh_family = AF_UNSPEC;
 222         nhm->nh_flags = nh->nh_flags;
 223         nhm->nh_protocol = nh->protocol;
 224         nhm->nh_scope = 0;
 225         nhm->resvd = 0;
 226
 227         if (nla_put_u32(skb, NHA_ID, nh->id))
 228                 goto nla_put_failure;
 229
 230         if (nh->is_group) {
 231                 struct nh_group *nhg = rtnl_dereference(nh->nh_grp);
 232
 233                 if (nla_put_nh_group(skb, nhg))
 234                         goto nla_put_failure;
 235                 goto out;
 236         }
 237
 238         nhi = rtnl_dereference(nh->nh_info);
 239         nhm->nh_family = nhi->family;
 240         if (nhi->reject_nh) {
 241                 if (nla_put_flag(skb, NHA_BLACKHOLE))
 242                         goto nla_put_failure;
 243                 goto out;
 244         } else {
 245                 const struct net_device *dev;
 246
 247                 dev = nhi->fib_nhc.nhc_dev;
 248                 if (dev && nla_put_u32(skb, NHA_OIF, dev->ifindex))
 249                         goto nla_put_failure;
 250         }
 251
 252         nhm->nh_scope = nhi->fib_nhc.nhc_scope;
 253         switch (nhi->family) {
 254         case AF_INET:
 255                 fib_nh = &nhi->fib_nh;
 256                 if (fib_nh->fib_nh_gw_family &&
 257                     nla_put_u32(skb, NHA_GATEWAY, fib_nh->fib_nh_gw4))
 258                         goto nla_put_failure;
 259                 break;
 260
 261         case AF_INET6:
 262                 fib6_nh = &nhi->fib6_nh;
 263                 if (fib6_nh->fib_nh_gw_family &&
 264                     nla_put_in6_addr(skb, NHA_GATEWAY, &fib6_nh->fib_nh_gw6))
 265                         goto nla_put_failure;
 266                 break;
 267         }
 268
 269         if (nhi->fib_nhc.nhc_lwtstate &&
 270             lwtunnel_fill_encap(skb, nhi->fib_nhc.nhc_lwtstate,
 271                                 NHA_ENCAP, NHA_ENCAP_TYPE) < 0)
 272                 goto nla_put_failure;
 273
 274 out:
 275         nlmsg_end(skb, nlh);
 276         return 0;
 277
 278 nla_put_failure:
 279         return -EMSGSIZE;
 280 }
 281
 282 static size_t nh_nlmsg_size_grp(struct nexthop *nh)
 283 {
 284         struct nh_group *nhg = rtnl_dereference(nh->nh_grp);
 285         size_t sz = sizeof(struct nexthop_grp) * nhg->num_nh;
 286
 287         return nla_total_size(sz) +
 288                nla_total_size(2);  /* NHA_GROUP_TYPE */
 289 }
 290
 291 static size_t nh_nlmsg_size_single(struct nexthop *nh)
 292 {
 293         struct nh_info *nhi = rtnl_dereference(nh->nh_info);
 294         size_t sz;
 295
 296         /* covers NHA_BLACKHOLE since NHA_OIF and BLACKHOLE
 297          * are mutually exclusive
 298          */
 299         sz = nla_total_size(4);  /* NHA_OIF */
 300
 301         switch (nhi->family) {
 302         case AF_INET:
 303                 if (nhi->fib_nh.fib_nh_gw_family)
 304                         sz += nla_total_size(4);  /* NHA_GATEWAY */
 305                 break;
 306
 307         case AF_INET6:
 308                 /* NHA_GATEWAY */
 309                 if (nhi->fib6_nh.fib_nh_gw_family)
 310                         sz += nla_total_size(sizeof(const struct in6_addr));
 311                 break;
 312         }
 313
 314         if (nhi->fib_nhc.nhc_lwtstate) {
 315                 sz += lwtunnel_get_encap_size(nhi->fib_nhc.nhc_lwtstate);
 316                 sz += nla_total_size(2);  /* NHA_ENCAP_TYPE */
 317         }
 318
 319         return sz;
 320 }
 321
 322 static size_t nh_nlmsg_size(struct nexthop *nh)
 323 {
 324         size_t sz = nla_total_size(4);    /* NHA_ID */
 325
 326         if (nh->is_group)
 327                 sz += nh_nlmsg_size_grp(nh);
 328         else
 329                 sz += nh_nlmsg_size_single(nh);
 330
 331         return sz;
 332 }
 333
 334 static void nexthop_notify(int event, struct nexthop *nh, struct nl_info *info)
 335 {
 336         unsigned int nlflags = info->nlh ? info->nlh->nlmsg_flags : 0;
 337         u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
 338         struct sk_buff *skb;
 339         int err = -ENOBUFS;
 340
 341         skb = nlmsg_new(nh_nlmsg_size(nh), gfp_any());
 342         if (!skb)
 343                 goto errout;
 344
 345         err = nh_fill_node(skb, nh, event, info->portid, seq, nlflags);
 346         if (err < 0) {
 347                 /* -EMSGSIZE implies BUG in nh_nlmsg_size() */
 348                 WARN_ON(err == -EMSGSIZE);
 349                 kfree_skb(skb);
 350                 goto errout;
 351         }
 352
 353         rtnl_notify(skb, info->nl_net, info->portid, RTNLGRP_NEXTHOP,
 354                     info->nlh, gfp_any());
 355         return;
 356 errout:
 357         if (err < 0)
 358                 rtnl_set_sk_err(info->nl_net, RTNLGRP_NEXTHOP, err);
 359 }
 360
 361 static bool valid_group_nh(struct nexthop *nh, unsigned int npaths,
 362                            struct netlink_ext_ack *extack)
 363 {
 364         if (nh->is_group) {
 365                 struct nh_group *nhg = rtnl_dereference(nh->nh_grp);
 366
 367                 /* nested multipath (group within a group) is not
 368                  * supported
 369                  */
 370                 if (nhg->mpath) {
 371                         NL_SET_ERR_MSG(extack,
 372                                        "Multipath group can not be a nexthop within a group");
 373                         return false;
 374                 }
 375         } else {
 376                 struct nh_info *nhi = rtnl_dereference(nh->nh_info);
 377
 378                 if (nhi->reject_nh && npaths > 1) {
 379                         NL_SET_ERR_MSG(extack,
 380                                        "Blackhole nexthop can not be used in a group with more than 1 path");
 381                         return false;
 382                 }
 383         }
 384
 385         return true;
 386 }
 387
 388 static int nh_check_attr_group(struct net *net, struct nlattr *tb[],
 389                                struct netlink_ext_ack *extack)
 390 {
 391         unsigned int len = nla_len(tb[NHA_GROUP]);
 392         struct nexthop_grp *nhg;
 393         unsigned int i, j;
 394
 395         if (len & (sizeof(struct nexthop_grp) - 1)) {
 396                 NL_SET_ERR_MSG(extack,
 397                                "Invalid length for nexthop group attribute");
 398                 return -EINVAL;
 399         }
 400
 401         /* convert len to number of nexthop ids */
 402         len /= sizeof(*nhg);
 403
 404         nhg = nla_data(tb[NHA_GROUP]);
 405         for (i = 0; i < len; ++i) {
 406                 if (nhg[i].resvd1 || nhg[i].resvd2) {
 407                         NL_SET_ERR_MSG(extack, "Reserved fields in nexthop_grp must be 0");
 408                         return -EINVAL;
 409                 }
 410                 if (nhg[i].weight > 254) {
 411                         NL_SET_ERR_MSG(extack, "Invalid value for weight");
 412                         return -EINVAL;
 413                 }
 414                 for (j = i + 1; j < len; ++j) {
 415                         if (nhg[i].id == nhg[j].id) {
 416                                 NL_SET_ERR_MSG(extack, "Nexthop id can not be used twice in a group");
 417                                 return -EINVAL;
 418                         }
 419                 }
 420         }
 421
 422         nhg = nla_data(tb[NHA_GROUP]);
 423         for (i = 0; i < len; ++i) {
 424                 struct nexthop *nh;
 425
 426                 nh = nexthop_find_by_id(net, nhg[i].id);
 427                 if (!nh) {
 428                         NL_SET_ERR_MSG(extack, "Invalid nexthop id");
 429                         return -EINVAL;
 430                 }
 431                 if (!valid_group_nh(nh, len, extack))
 432                         return -EINVAL;
 433         }
 434         for (i = NHA_GROUP + 1; i < __NHA_MAX; ++i) {
 435                 if (!tb[i])
 436                         continue;
 437
 438                 NL_SET_ERR_MSG(extack,
 439                                "No other attributes can be set in nexthop groups");
 440                 return -EINVAL;
 441         }
 442
 443         return 0;
 444 }
 445
 446 static bool ipv6_good_nh(const struct fib6_nh *nh)
 447 {
 448         int state = NUD_REACHABLE;
 449         struct neighbour *n;
 450
 451         rcu_read_lock_bh();
 452
 453         n = __ipv6_neigh_lookup_noref_stub(nh->fib_nh_dev, &nh->fib_nh_gw6);
 454         if (n)
 455                 state = n->nud_state;
 456
 457         rcu_read_unlock_bh();
 458
 459         return !!(state & NUD_VALID);
 460 }
 461
 462 static bool ipv4_good_nh(const struct fib_nh *nh)
 463 {
 464         int state = NUD_REACHABLE;
 465         struct neighbour *n;
 466
 467         rcu_read_lock_bh();
 468
 469         n = __ipv4_neigh_lookup_noref(nh->fib_nh_dev,
 470                                       (__force u32)nh->fib_nh_gw4);
 471         if (n)
 472                 state = n->nud_state;
 473
 474         rcu_read_unlock_bh();
 475
 476         return !!(state & NUD_VALID);
 477 }
 478
 479 struct nexthop *nexthop_select_path(struct nexthop *nh, int hash)
 480 {
 481         struct nexthop *rc = NULL;
 482         struct nh_group *nhg;
 483         int i;
 484
 485         if (!nh->is_group)
 486                 return nh;
 487
 488         nhg = rcu_dereference(nh->nh_grp);
 489         for (i = 0; i < nhg->num_nh; ++i) {
 490                 struct nh_grp_entry *nhge = &nhg->nh_entries[i];
 491                 struct nh_info *nhi;
 492
 493                 if (hash > atomic_read(&nhge->upper_bound))
 494                         continue;
 495
 496                 /* nexthops always check if it is good and does
 497                  * not rely on a sysctl for this behavior
 498                  */
 499                 nhi = rcu_dereference(nhge->nh->nh_info);
 500                 switch (nhi->family) {
 501                 case AF_INET:
 502                         if (ipv4_good_nh(&nhi->fib_nh))
 503                                 return nhge->nh;
 504                         break;
 505                 case AF_INET6:
 506                         if (ipv6_good_nh(&nhi->fib6_nh))
 507                                 return nhge->nh;
 508                         break;
 509                 }
 510
 511                 if (!rc)
 512                         rc = nhge->nh;
 513         }
 514
 515         return rc;
 516 }
 517 EXPORT_SYMBOL_GPL(nexthop_select_path);
 518
 519 static int nexthop_check_scope(struct nexthop *nh, u8 scope,
 520                                struct netlink_ext_ack *extack)
 521 {
 522         struct nh_info *nhi;
 523
 524         nhi = rtnl_dereference(nh->nh_info);
 525         if (scope == RT_SCOPE_HOST && nhi->fib_nhc.nhc_gw_family) {
 526                 NL_SET_ERR_MSG(extack,
 527                                "Route with host scope can not have a gateway");
 528                 return -EINVAL;
 529         }
 530
 531         if (nhi->fib_nhc.nhc_flags & RTNH_F_ONLINK && scope >= RT_SCOPE_LINK) {
 532                 NL_SET_ERR_MSG(extack, "Scope mismatch with nexthop");
 533                 return -EINVAL;
 534         }
 535
 536         return 0;
 537 }
 538
 539 /* Invoked by fib add code to verify nexthop by id is ok with
 540  * config for prefix; parts of fib_check_nh not done when nexthop
 541  * object is used.
 542  */
 543 int fib_check_nexthop(struct nexthop *nh, u8 scope,
 544                       struct netlink_ext_ack *extack)
 545 {
 546         int err = 0;
 547
 548         if (nh->is_group) {
 549                 struct nh_group *nhg;
 550
 551                 if (scope == RT_SCOPE_HOST) {
 552                         NL_SET_ERR_MSG(extack, "Route with host scope can not have multiple nexthops");
 553                         err = -EINVAL;
 554                         goto out;
 555                 }
 556
 557                 nhg = rtnl_dereference(nh->nh_grp);
 558                 /* all nexthops in a group have the same scope */
 559                 err = nexthop_check_scope(nhg->nh_entries[0].nh, scope, extack);
 560         } else {
 561                 err = nexthop_check_scope(nh, scope, extack);
 562         }
 563 out:
 564         return err;
 565 }
 566
 567 static void nh_group_rebalance(struct nh_group *nhg)
 568 {
 569         int total = 0;
 570         int w = 0;
 571         int i;
 572
 573         for (i = 0; i < nhg->num_nh; ++i)
 574                 total += nhg->nh_entries[i].weight;
 575
 576         for (i = 0; i < nhg->num_nh; ++i) {
 577                 struct nh_grp_entry *nhge = &nhg->nh_entries[i];
 578                 int upper_bound;
 579
 580                 w += nhge->weight;
 581                 upper_bound = DIV_ROUND_CLOSEST_ULL((u64)w << 31, total) - 1;
 582                 atomic_set(&nhge->upper_bound, upper_bound);
 583         }
 584 }
 585
 586 static void remove_nh_grp_entry(struct nh_grp_entry *nhge,
 587                                 struct nh_group *nhg,
 588                                 struct nl_info *nlinfo)
 589 {
 590         struct nexthop *nh = nhge->nh;
 591         struct nh_grp_entry *nhges;
 592         bool found = false;
 593         int i;
 594
 595         WARN_ON(!nh);
 596
 597         nhges = nhg->nh_entries;
 598         for (i = 0; i < nhg->num_nh; ++i) {
 599                 if (found) {
 600                         nhges[i-1].nh = nhges[i].nh;
 601                         nhges[i-1].weight = nhges[i].weight;
 602                         list_del(&nhges[i].nh_list);
 603                         list_add(&nhges[i-1].nh_list, &nhges[i-1].nh->grp_list);
 604                 } else if (nhg->nh_entries[i].nh == nh) {
 605                         found = true;
 606                 }
 607         }
 608
 609         if (WARN_ON(!found))
 610                 return;
 611
 612         nhg->num_nh--;
 613         nhg->nh_entries[nhg->num_nh].nh = NULL;
 614
 615         nh_group_rebalance(nhg);
 616
 617         nexthop_put(nh);
 618
 619         if (nlinfo)
 620                 nexthop_notify(RTM_NEWNEXTHOP, nhge->nh_parent, nlinfo);
 621 }
 622
 623 static void remove_nexthop_from_groups(struct net *net, struct nexthop *nh,
 624                                        struct nl_info *nlinfo)
 625 {
 626         struct nh_grp_entry *nhge, *tmp;
 627
 628         list_for_each_entry_safe(nhge, tmp, &nh->grp_list, nh_list) {
 629                 struct nh_group *nhg;
 630
 631                 list_del(&nhge->nh_list);
 632                 nhg = rtnl_dereference(nhge->nh_parent->nh_grp);
 633                 remove_nh_grp_entry(nhge, nhg, nlinfo);
 634
 635                 /* if this group has no more entries then remove it */
 636                 if (!nhg->num_nh)
 637                         remove_nexthop(net, nhge->nh_parent, nlinfo);
 638         }
 639 }
 640
 641 static void remove_nexthop_group(struct nexthop *nh, struct nl_info *nlinfo)
 642 {
 643         struct nh_group *nhg = rcu_dereference_rtnl(nh->nh_grp);
 644         int i, num_nh = nhg->num_nh;
 645
 646         for (i = 0; i < num_nh; ++i) {
 647                 struct nh_grp_entry *nhge = &nhg->nh_entries[i];
 648
 649                 if (WARN_ON(!nhge->nh))
 650                         continue;
 651
 652                 list_del(&nhge->nh_list);
 653                 nexthop_put(nhge->nh);
 654                 nhge->nh = NULL;
 655                 nhg->num_nh--;
 656         }
 657 }
 658
 659 static void __remove_nexthop_fib(struct net *net, struct nexthop *nh)
 660 {
 661         bool do_flush = false;
 662         struct fib_info *fi;
 663
 664         list_for_each_entry(fi, &nh->fi_list, nh_list) {
 665                 fi->fib_flags |= RTNH_F_DEAD;
 666                 do_flush = true;
 667         }
 668         if (do_flush)
 669                 fib_flush(net);
 670 }
 671
 672 static void __remove_nexthop(struct net *net, struct nexthop *nh,
 673                              struct nl_info *nlinfo)
 674 {
 675         __remove_nexthop_fib(net, nh);
 676
 677         if (nh->is_group) {
 678                 remove_nexthop_group(nh, nlinfo);
 679         } else {
 680                 struct nh_info *nhi;
 681
 682                 nhi = rtnl_dereference(nh->nh_info);
 683                 if (nhi->fib_nhc.nhc_dev)
 684                         hlist_del(&nhi->dev_hash);
 685
 686                 remove_nexthop_from_groups(net, nh, nlinfo);
 687         }
 688 }
 689
 690 static void remove_nexthop(struct net *net, struct nexthop *nh,
 691                            struct nl_info *nlinfo)
 692 {
 693         /* remove from the tree */
 694         rb_erase(&nh->rb_node, &net->nexthop.rb_root);
 695
 696         if (nlinfo)
 697                 nexthop_notify(RTM_DELNEXTHOP, nh, nlinfo);
 698
 699         __remove_nexthop(net, nh, nlinfo);
 700         nh_base_seq_inc(net);
 701
 702         nexthop_put(nh);
 703 }
 704
 705 static int replace_nexthop(struct net *net, struct nexthop *old,
 706                            struct nexthop *new, struct netlink_ext_ack *extack)
 707 {
 708         return -EEXIST;
 709 }
 710
 711 /* called with rtnl_lock held */
 712 static int insert_nexthop(struct net *net, struct nexthop *new_nh,
 713                           struct nh_config *cfg, struct netlink_ext_ack *extack)
 714 {
 715         struct rb_node **pp, *parent = NULL, *next;
 716         struct rb_root *root = &net->nexthop.rb_root;
 717         bool replace = !!(cfg->nlflags & NLM_F_REPLACE);
 718         bool create = !!(cfg->nlflags & NLM_F_CREATE);
 719         u32 new_id = new_nh->id;
 720         int rc = -EEXIST;
 721
 722         pp = &root->rb_node;
 723         while (1) {
 724                 struct nexthop *nh;
 725
 726                 next = rtnl_dereference(*pp);
 727                 if (!next)
 728                         break;
 729
 730                 parent = next;
 731
 732                 nh = rb_entry(parent, struct nexthop, rb_node);
 733                 if (new_id < nh->id) {
 734                         pp = &next->rb_left;
 735                 } else if (new_id > nh->id) {
 736                         pp = &next->rb_right;
 737                 } else if (replace) {
 738                         rc = replace_nexthop(net, nh, new_nh, extack);
 739                         if (!rc)
 740                                 new_nh = nh; /* send notification with old nh */
 741                         goto out;
 742                 } else {
 743                         /* id already exists and not a replace */
 744                         goto out;
 745                 }
 746         }
 747
 748         if (replace && !create) {
 749                 NL_SET_ERR_MSG(extack, "Replace specified without create and no entry exists");
 750                 rc = -ENOENT;
 751                 goto out;
 752         }
 753
 754         rb_link_node_rcu(&new_nh->rb_node, parent, pp);
 755         rb_insert_color(&new_nh->rb_node, root);
 756         rc = 0;
 757 out:
 758         if (!rc) {
 759                 nh_base_seq_inc(net);
 760                 nexthop_notify(RTM_NEWNEXTHOP, new_nh, &cfg->nlinfo);
 761         }
 762
 763         return rc;
 764 }
 765
 766 /* rtnl */
 767 /* remove all nexthops tied to a device being deleted */
 768 static void nexthop_flush_dev(struct net_device *dev)
 769 {
 770         unsigned int hash = nh_dev_hashfn(dev->ifindex);
 771         struct net *net = dev_net(dev);
 772         struct hlist_head *head = &net->nexthop.devhash[hash];
 773         struct hlist_node *n;
 774         struct nh_info *nhi;
 775
 776         hlist_for_each_entry_safe(nhi, n, head, dev_hash) {
 777                 if (nhi->fib_nhc.nhc_dev != dev)
 778                         continue;
 779
 780                 remove_nexthop(net, nhi->nh_parent, NULL);
 781         }
 782 }
 783
 784 /* rtnl; called when net namespace is deleted */
 785 static void flush_all_nexthops(struct net *net)
 786 {
 787         struct rb_root *root = &net->nexthop.rb_root;
 788         struct rb_node *node;
 789         struct nexthop *nh;
 790
 791         while ((node = rb_first(root))) {
 792                 nh = rb_entry(node, struct nexthop, rb_node);
 793                 remove_nexthop(net, nh, NULL);
 794                 cond_resched();
 795         }
 796 }
 797
 798 static struct nexthop *nexthop_create_group(struct net *net,
 799                                             struct nh_config *cfg)
 800 {
 801         struct nlattr *grps_attr = cfg->nh_grp;
 802         struct nexthop_grp *entry = nla_data(grps_attr);
 803         struct nh_group *nhg;
 804         struct nexthop *nh;
 805         int i;
 806
 807         nh = nexthop_alloc();
 808         if (!nh)
 809                 return ERR_PTR(-ENOMEM);
 810
 811         nh->is_group = 1;
 812
 813         nhg = nexthop_grp_alloc(nla_len(grps_attr) / sizeof(*entry));
 814         if (!nhg) {
 815                 kfree(nh);
 816                 return ERR_PTR(-ENOMEM);
 817         }
 818
 819         for (i = 0; i < nhg->num_nh; ++i) {
 820                 struct nexthop *nhe;
 821                 struct nh_info *nhi;
 822
 823                 nhe = nexthop_find_by_id(net, entry[i].id);
 824                 if (!nexthop_get(nhe))
 825                         goto out_no_nh;
 826
 827                 nhi = rtnl_dereference(nhe->nh_info);
 828                 if (nhi->family == AF_INET)
 829                         nhg->has_v4 = true;
 830
 831                 nhg->nh_entries[i].nh = nhe;
 832                 nhg->nh_entries[i].weight = entry[i].weight + 1;
 833                 list_add(&nhg->nh_entries[i].nh_list, &nhe->grp_list);
 834                 nhg->nh_entries[i].nh_parent = nh;
 835         }
 836
 837         if (cfg->nh_grp_type == NEXTHOP_GRP_TYPE_MPATH) {
 838                 nhg->mpath = 1;
 839                 nh_group_rebalance(nhg);
 840         }
 841
 842         rcu_assign_pointer(nh->nh_grp, nhg);
 843
 844         return nh;
 845
 846 out_no_nh:
 847         for (; i >= 0; --i)
 848                 nexthop_put(nhg->nh_entries[i].nh);
 849
 850         kfree(nhg);
 851         kfree(nh);
 852
 853         return ERR_PTR(-ENOENT);
 854 }
 855
 856 static int nh_create_ipv4(struct net *net, struct nexthop *nh,
 857                           struct nh_info *nhi, struct nh_config *cfg,
 858                           struct netlink_ext_ack *extack)
 859 {
 860         struct fib_nh *fib_nh = &nhi->fib_nh;
 861         struct fib_config fib_cfg = {
 862                 .fc_oif   = cfg->nh_ifindex,
 863                 .fc_gw4   = cfg->gw.ipv4,
 864                 .fc_gw_family = cfg->gw.ipv4 ? AF_INET : 0,
 865                 .fc_flags = cfg->nh_flags,
 866                 .fc_encap = cfg->nh_encap,
 867                 .fc_encap_type = cfg->nh_encap_type,
 868         };
 869         u32 tb_id = l3mdev_fib_table(cfg->dev);
 870         int err = -EINVAL;
 871
 872         err = fib_nh_init(net, fib_nh, &fib_cfg, 1, extack);
 873         if (err) {
 874                 fib_nh_release(net, fib_nh);
 875                 goto out;
 876         }
 877
 878         /* sets nh_dev if successful */
 879         err = fib_check_nh(net, fib_nh, tb_id, 0, extack);
 880         if (!err) {
 881                 nh->nh_flags = fib_nh->fib_nh_flags;
 882                 fib_info_update_nhc_saddr(net, &fib_nh->nh_common,
 883                                           fib_nh->fib_nh_scope);
 884         } else {
 885                 fib_nh_release(net, fib_nh);
 886         }
 887 out:
 888         return err;
 889 }
 890
 891 static int nh_create_ipv6(struct net *net,  struct nexthop *nh,
 892                           struct nh_info *nhi, struct nh_config *cfg,
 893                           struct netlink_ext_ack *extack)
 894 {
 895         struct fib6_nh *fib6_nh = &nhi->fib6_nh;
 896         struct fib6_config fib6_cfg = {
 897                 .fc_table = l3mdev_fib_table(cfg->dev),
 898                 .fc_ifindex = cfg->nh_ifindex,
 899                 .fc_gateway = cfg->gw.ipv6,
 900                 .fc_flags = cfg->nh_flags,
 901                 .fc_encap = cfg->nh_encap,
 902                 .fc_encap_type = cfg->nh_encap_type,
 903         };
 904         int err;
 905
 906         if (!ipv6_addr_any(&cfg->gw.ipv6))
 907                 fib6_cfg.fc_flags |= RTF_GATEWAY;
 908
 909         /* sets nh_dev if successful */
 910         err = ipv6_stub->fib6_nh_init(net, fib6_nh, &fib6_cfg, GFP_KERNEL,
 911                                       extack);
 912         if (err)
 913                 ipv6_stub->fib6_nh_release(fib6_nh);
 914         else
 915                 nh->nh_flags = fib6_nh->fib_nh_flags;
 916
 917         return err;
 918 }
 919
 920 static struct nexthop *nexthop_create(struct net *net, struct nh_config *cfg,
 921                                       struct netlink_ext_ack *extack)
 922 {
 923         struct nh_info *nhi;
 924         struct nexthop *nh;
 925         int err = 0;
 926
 927         nh = nexthop_alloc();
 928         if (!nh)
 929                 return ERR_PTR(-ENOMEM);
 930
 931         nhi = kzalloc(sizeof(*nhi), GFP_KERNEL);
 932         if (!nhi) {
 933                 kfree(nh);
 934                 return ERR_PTR(-ENOMEM);
 935         }
 936
 937         nh->nh_flags = cfg->nh_flags;
 938         nh->net = net;
 939
 940         nhi->nh_parent = nh;
 941         nhi->family = cfg->nh_family;
 942         nhi->fib_nhc.nhc_scope = RT_SCOPE_LINK;
 943
 944         if (cfg->nh_blackhole) {
 945                 nhi->reject_nh = 1;
 946                 cfg->nh_ifindex = net->loopback_dev->ifindex;
 947         }
 948
 949         switch (cfg->nh_family) {
 950         case AF_INET:
 951                 err = nh_create_ipv4(net, nh, nhi, cfg, extack);
 952                 break;
 953         case AF_INET6:
 954                 err = nh_create_ipv6(net, nh, nhi, cfg, extack);
 955                 break;
 956         }
 957
 958         if (err) {
 959                 kfree(nhi);
 960                 kfree(nh);
 961                 return ERR_PTR(err);
 962         }
 963
 964         /* add the entry to the device based hash */
 965         nexthop_devhash_add(net, nhi);
 966
 967         rcu_assign_pointer(nh->nh_info, nhi);
 968
 969         return nh;
 970 }
 971
 972 /* called with rtnl lock held */
 973 static struct nexthop *nexthop_add(struct net *net, struct nh_config *cfg,
 974                                    struct netlink_ext_ack *extack)
 975 {
 976         struct nexthop *nh;
 977         int err;
 978
 979         if (cfg->nlflags & NLM_F_REPLACE && !cfg->nh_id) {
 980                 NL_SET_ERR_MSG(extack, "Replace requires nexthop id");
 981                 return ERR_PTR(-EINVAL);
 982         }
 983
 984         if (!cfg->nh_id) {
 985                 cfg->nh_id = nh_find_unused_id(net);
 986                 if (!cfg->nh_id) {
 987                         NL_SET_ERR_MSG(extack, "No unused id");
 988                         return ERR_PTR(-EINVAL);
 989                 }
 990         }
 991
 992         if (cfg->nh_grp)
 993                 nh = nexthop_create_group(net, cfg);
 994         else
 995                 nh = nexthop_create(net, cfg, extack);
 996
 997         if (IS_ERR(nh))
 998                 return nh;
 999
1000         refcount_set(&nh->refcnt, 1);
1001         nh->id = cfg->nh_id;
1002         nh->protocol = cfg->nh_protocol;
1003         nh->net = net;
1004
1005         err = insert_nexthop(net, nh, cfg, extack);
1006         if (err) {
1007                 __remove_nexthop(net, nh, NULL);
1008                 nexthop_put(nh);
1009                 nh = ERR_PTR(err);
1010         }
1011
1012         return nh;
1013 }
1014
1015 static int rtm_to_nh_config(struct net *net, struct sk_buff *skb,
1016                             struct nlmsghdr *nlh, struct nh_config *cfg,
1017                             struct netlink_ext_ack *extack)
1018 {
1019         struct nhmsg *nhm = nlmsg_data(nlh);
1020         struct nlattr *tb[NHA_MAX + 1];
1021         int err;
1022
1023         err = nlmsg_parse(nlh, sizeof(*nhm), tb, NHA_MAX, rtm_nh_policy,
1024                           extack);
1025         if (err < 0)
1026                 return err;
1027
1028         err = -EINVAL;
1029         if (nhm->resvd || nhm->nh_scope) {
1030                 NL_SET_ERR_MSG(extack, "Invalid values in ancillary header");
1031                 goto out;
1032         }
1033         if (nhm->nh_flags & ~NEXTHOP_VALID_USER_FLAGS) {
1034                 NL_SET_ERR_MSG(extack, "Invalid nexthop flags in ancillary header");
1035                 goto out;
1036         }
1037
1038         switch (nhm->nh_family) {
1039         case AF_INET:
1040         case AF_INET6:
1041                 break;
1042         case AF_UNSPEC:
1043                 if (tb[NHA_GROUP])
1044                         break;
1045                 /* fallthrough */
1046         default:
1047                 NL_SET_ERR_MSG(extack, "Invalid address family");
1048                 goto out;
1049         }
1050
1051         if (tb[NHA_GROUPS] || tb[NHA_MASTER]) {
1052                 NL_SET_ERR_MSG(extack, "Invalid attributes in request");
1053                 goto out;
1054         }
1055
1056         memset(cfg, 0, sizeof(*cfg));
1057         cfg->nlflags = nlh->nlmsg_flags;
1058         cfg->nlinfo.portid = NETLINK_CB(skb).portid;
1059         cfg->nlinfo.nlh = nlh;
1060         cfg->nlinfo.nl_net = net;
1061
1062         cfg->nh_family = nhm->nh_family;
1063         cfg->nh_protocol = nhm->nh_protocol;
1064         cfg->nh_flags = nhm->nh_flags;
1065
1066         if (tb[NHA_ID])
1067                 cfg->nh_id = nla_get_u32(tb[NHA_ID]);
1068
1069         if (tb[NHA_GROUP]) {
1070                 if (nhm->nh_family != AF_UNSPEC) {
1071                         NL_SET_ERR_MSG(extack, "Invalid family for group");
1072                         goto out;
1073                 }
1074                 cfg->nh_grp = tb[NHA_GROUP];
1075
1076                 cfg->nh_grp_type = NEXTHOP_GRP_TYPE_MPATH;
1077                 if (tb[NHA_GROUP_TYPE])
1078                         cfg->nh_grp_type = nla_get_u16(tb[NHA_GROUP_TYPE]);
1079
1080                 if (cfg->nh_grp_type > NEXTHOP_GRP_TYPE_MAX) {
1081                         NL_SET_ERR_MSG(extack, "Invalid group type");
1082                         goto out;
1083                 }
1084                 err = nh_check_attr_group(net, tb, extack);
1085
1086                 /* no other attributes should be set */
1087                 goto out;
1088         }
1089
1090         if (tb[NHA_BLACKHOLE]) {
1091                 if (tb[NHA_GATEWAY] || tb[NHA_OIF] ||
1092                     tb[NHA_ENCAP]   || tb[NHA_ENCAP_TYPE]) {
1093                         NL_SET_ERR_MSG(extack, "Blackhole attribute can not be used with gateway or oif");
1094                         goto out;
1095                 }
1096
1097                 cfg->nh_blackhole = 1;
1098                 err = 0;
1099                 goto out;
1100         }
1101
1102         if (!tb[NHA_OIF]) {
1103                 NL_SET_ERR_MSG(extack, "Device attribute required for non-blackhole nexthops");
1104                 goto out;
1105         }
1106
1107         cfg->nh_ifindex = nla_get_u32(tb[NHA_OIF]);
1108         if (cfg->nh_ifindex)
1109                 cfg->dev = __dev_get_by_index(net, cfg->nh_ifindex);
1110
1111         if (!cfg->dev) {
1112                 NL_SET_ERR_MSG(extack, "Invalid device index");
1113                 goto out;
1114         } else if (!(cfg->dev->flags & IFF_UP)) {
1115                 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
1116                 err = -ENETDOWN;
1117                 goto out;
1118         } else if (!netif_carrier_ok(cfg->dev)) {
1119                 NL_SET_ERR_MSG(extack, "Carrier for nexthop device is down");
1120                 err = -ENETDOWN;
1121                 goto out;
1122         }
1123
1124         err = -EINVAL;
1125         if (tb[NHA_GATEWAY]) {
1126                 struct nlattr *gwa = tb[NHA_GATEWAY];
1127
1128                 switch (cfg->nh_family) {
1129                 case AF_INET:
1130                         if (nla_len(gwa) != sizeof(u32)) {
1131                                 NL_SET_ERR_MSG(extack, "Invalid gateway");
1132                                 goto out;
1133                         }
1134                         cfg->gw.ipv4 = nla_get_be32(gwa);
1135                         break;
1136                 case AF_INET6:
1137                         if (nla_len(gwa) != sizeof(struct in6_addr)) {
1138                                 NL_SET_ERR_MSG(extack, "Invalid gateway");
1139                                 goto out;
1140                         }
1141                         cfg->gw.ipv6 = nla_get_in6_addr(gwa);
1142                         break;
1143                 default:
1144                         NL_SET_ERR_MSG(extack,
1145                                        "Unknown address family for gateway");
1146                         goto out;
1147                 }
1148         } else {
1149                 /* device only nexthop (no gateway) */
1150                 if (cfg->nh_flags & RTNH_F_ONLINK) {
1151                         NL_SET_ERR_MSG(extack,
1152                                        "ONLINK flag can not be set for nexthop without a gateway");
1153                         goto out;
1154                 }
1155         }
1156
1157         if (tb[NHA_ENCAP]) {
1158                 cfg->nh_encap = tb[NHA_ENCAP];
1159
1160                 if (!tb[NHA_ENCAP_TYPE]) {
1161                         NL_SET_ERR_MSG(extack, "LWT encapsulation type is missing");
1162                         goto out;
1163                 }
1164
1165                 cfg->nh_encap_type = nla_get_u16(tb[NHA_ENCAP_TYPE]);
1166                 err = lwtunnel_valid_encap_type(cfg->nh_encap_type, extack);
1167                 if (err < 0)
1168                         goto out;
1169
1170         } else if (tb[NHA_ENCAP_TYPE]) {
1171                 NL_SET_ERR_MSG(extack, "LWT encapsulation attribute is missing");
1172                 goto out;
1173         }
1174
1175
1176         err = 0;
1177 out:
1178         return err;
1179 }
1180
1181 /* rtnl */
1182 static int rtm_new_nexthop(struct sk_buff *skb, struct nlmsghdr *nlh,
1183                            struct netlink_ext_ack *extack)
1184 {
1185         struct net *net = sock_net(skb->sk);
1186         struct nh_config cfg;
1187         struct nexthop *nh;
1188         int err;
1189
1190         err = rtm_to_nh_config(net, skb, nlh, &cfg, extack);
1191         if (!err) {
1192                 nh = nexthop_add(net, &cfg, extack);
1193                 if (IS_ERR(nh))
1194                         err = PTR_ERR(nh);
1195         }
1196
1197         return err;
1198 }
1199
1200 static int nh_valid_get_del_req(struct nlmsghdr *nlh, u32 *id,
1201                                 struct netlink_ext_ack *extack)
1202 {
1203         struct nhmsg *nhm = nlmsg_data(nlh);
1204         struct nlattr *tb[NHA_MAX + 1];
1205         int err, i;
1206
1207         err = nlmsg_parse(nlh, sizeof(*nhm), tb, NHA_MAX, rtm_nh_policy,
1208                           extack);
1209         if (err < 0)
1210                 return err;
1211
1212         err = -EINVAL;
1213         for (i = 0; i < __NHA_MAX; ++i) {
1214                 if (!tb[i])
1215                         continue;
1216
1217                 switch (i) {
1218                 case NHA_ID:
1219                         break;
1220                 default:
1221                         NL_SET_ERR_MSG_ATTR(extack, tb[i],
1222                                             "Unexpected attribute in request");
1223                         goto out;
1224                 }
1225         }
1226         if (nhm->nh_protocol || nhm->resvd || nhm->nh_scope || nhm->nh_flags) {
1227                 NL_SET_ERR_MSG(extack, "Invalid values in header");
1228                 goto out;
1229         }
1230
1231         if (!tb[NHA_ID]) {
1232                 NL_SET_ERR_MSG(extack, "Nexthop id is missing");
1233                 goto out;
1234         }
1235
1236         *id = nla_get_u32(tb[NHA_ID]);
1237         if (!(*id))
1238                 NL_SET_ERR_MSG(extack, "Invalid nexthop id");
1239         else
1240                 err = 0;
1241 out:
1242         return err;
1243 }
1244
1245 /* rtnl */
1246 static int rtm_del_nexthop(struct sk_buff *skb, struct nlmsghdr *nlh,
1247                            struct netlink_ext_ack *extack)
1248 {
1249         struct net *net = sock_net(skb->sk);
1250         struct nl_info nlinfo = {
1251                 .nlh = nlh,
1252                 .nl_net = net,
1253                 .portid = NETLINK_CB(skb).portid,
1254         };
1255         struct nexthop *nh;
1256         int err;
1257         u32 id;
1258
1259         err = nh_valid_get_del_req(nlh, &id, extack);
1260         if (err)
1261                 return err;
1262
1263         nh = nexthop_find_by_id(net, id);
1264         if (!nh)
1265                 return -ENOENT;
1266
1267         remove_nexthop(net, nh, &nlinfo);
1268
1269         return 0;
1270 }
1271
1272 /* rtnl */
1273 static int rtm_get_nexthop(struct sk_buff *in_skb, struct nlmsghdr *nlh,
1274                            struct netlink_ext_ack *extack)
1275 {
1276         struct net *net = sock_net(in_skb->sk);
1277         struct sk_buff *skb = NULL;
1278         struct nexthop *nh;
1279         int err;
1280         u32 id;
1281
1282         err = nh_valid_get_del_req(nlh, &id, extack);
1283         if (err)
1284                 return err;
1285
1286         err = -ENOBUFS;
1287         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1288         if (!skb)
1289                 goto out;
1290
1291         err = -ENOENT;
1292         nh = nexthop_find_by_id(net, id);
1293         if (!nh)
1294                 goto errout_free;
1295
1296         err = nh_fill_node(skb, nh, RTM_NEWNEXTHOP, NETLINK_CB(in_skb).portid,
1297                            nlh->nlmsg_seq, 0);
1298         if (err < 0) {
1299                 WARN_ON(err == -EMSGSIZE);
1300                 goto errout_free;
1301         }
1302
1303         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
1304 out:
1305         return err;
1306 errout_free:
1307         kfree_skb(skb);
1308         goto out;
1309 }
1310
1311 static bool nh_dump_filtered(struct nexthop *nh, int dev_idx, int master_idx,
1312                              bool group_filter, u8 family)
1313 {
1314         const struct net_device *dev;
1315         const struct nh_info *nhi;
1316
1317         if (group_filter && !nh->is_group)
1318                 return true;
1319
1320         if (!dev_idx && !master_idx && !family)
1321                 return false;
1322
1323         if (nh->is_group)
1324                 return true;
1325
1326         nhi = rtnl_dereference(nh->nh_info);
1327         if (family && nhi->family != family)
1328                 return true;
1329
1330         dev = nhi->fib_nhc.nhc_dev;
1331         if (dev_idx && (!dev || dev->ifindex != dev_idx))
1332                 return true;
1333
1334         if (master_idx) {
1335                 struct net_device *master;
1336
1337                 if (!dev)
1338                         return true;
1339
1340                 master = netdev_master_upper_dev_get((struct net_device *)dev);
1341                 if (!master || master->ifindex != master_idx)
1342                         return true;
1343         }
1344
1345         return false;
1346 }
1347
1348 static int nh_valid_dump_req(const struct nlmsghdr *nlh, int *dev_idx,
1349                              int *master_idx, bool *group_filter,
1350                              struct netlink_callback *cb)
1351 {
1352         struct netlink_ext_ack *extack = cb->extack;
1353         struct nlattr *tb[NHA_MAX + 1];
1354         struct nhmsg *nhm;
1355         int err, i;
1356         u32 idx;
1357
1358         err = nlmsg_parse(nlh, sizeof(*nhm), tb, NHA_MAX, rtm_nh_policy,
1359                           NULL);
1360         if (err < 0)
1361                 return err;
1362
1363         for (i = 0; i <= NHA_MAX; ++i) {
1364                 if (!tb[i])
1365                         continue;
1366
1367                 switch (i) {
1368                 case NHA_OIF:
1369                         idx = nla_get_u32(tb[i]);
1370                         if (idx > INT_MAX) {
1371                                 NL_SET_ERR_MSG(extack, "Invalid device index");
1372                                 return -EINVAL;
1373                         }
1374                         *dev_idx = idx;
1375                         break;
1376                 case NHA_MASTER:
1377                         idx = nla_get_u32(tb[i]);
1378                         if (idx > INT_MAX) {
1379                                 NL_SET_ERR_MSG(extack, "Invalid master device index");
1380                                 return -EINVAL;
1381                         }
1382                         *master_idx = idx;
1383                         break;
1384                 case NHA_GROUPS:
1385                         *group_filter = true;
1386                         break;
1387                 default:
1388                         NL_SET_ERR_MSG(extack, "Unsupported attribute in dump request");
1389                         return -EINVAL;
1390                 }
1391         }
1392
1393         nhm = nlmsg_data(nlh);
1394         if (nhm->nh_protocol || nhm->resvd || nhm->nh_scope || nhm->nh_flags) {
1395                 NL_SET_ERR_MSG(extack, "Invalid values in header for nexthop dump request");
1396                 return -EINVAL;
1397         }
1398
1399         return 0;
1400 }
1401
1402 /* rtnl */
1403 static int rtm_dump_nexthop(struct sk_buff *skb, struct netlink_callback *cb)
1404 {
1405         struct nhmsg *nhm = nlmsg_data(cb->nlh);
1406         int dev_filter_idx = 0, master_idx = 0;
1407         struct net *net = sock_net(skb->sk);
1408         struct rb_root *root = &net->nexthop.rb_root;
1409         bool group_filter = false;
1410         struct rb_node *node;
1411         int idx = 0, s_idx;
1412         int err;
1413
1414         err = nh_valid_dump_req(cb->nlh, &dev_filter_idx, &master_idx,
1415                                 &group_filter, cb);
1416         if (err < 0)
1417                 return err;
1418
1419         s_idx = cb->args[0];
1420         for (node = rb_first(root); node; node = rb_next(node)) {
1421                 struct nexthop *nh;
1422
1423                 if (idx < s_idx)
1424                         goto cont;
1425
1426                 nh = rb_entry(node, struct nexthop, rb_node);
1427                 if (nh_dump_filtered(nh, dev_filter_idx, master_idx,
1428                                      group_filter, nhm->nh_family))
1429                         goto cont;
1430
1431                 err = nh_fill_node(skb, nh, RTM_NEWNEXTHOP,
1432                                    NETLINK_CB(cb->skb).portid,
1433                                    cb->nlh->nlmsg_seq, NLM_F_MULTI);
1434                 if (err < 0) {
1435                         if (likely(skb->len))
1436                                 goto out;
1437
1438                         goto out_err;
1439                 }
1440 cont:
1441                 idx++;
1442         }
1443
1444 out:
1445         err = skb->len;
1446 out_err:
1447         cb->args[0] = idx;
1448         cb->seq = net->nexthop.seq;
1449         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1450
1451         return err;
1452 }
1453
1454 static void nexthop_sync_mtu(struct net_device *dev, u32 orig_mtu)
1455 {
1456         unsigned int hash = nh_dev_hashfn(dev->ifindex);
1457         struct net *net = dev_net(dev);
1458         struct hlist_head *head = &net->nexthop.devhash[hash];
1459         struct hlist_node *n;
1460         struct nh_info *nhi;
1461
1462         hlist_for_each_entry_safe(nhi, n, head, dev_hash) {
1463                 if (nhi->fib_nhc.nhc_dev == dev) {
1464                         if (nhi->family == AF_INET)
1465                                 fib_nhc_update_mtu(&nhi->fib_nhc, dev->mtu,
1466                                                    orig_mtu);
1467                 }
1468         }
1469 }
1470
1471 /* rtnl */
1472 static int nh_netdev_event(struct notifier_block *this,
1473                            unsigned long event, void *ptr)
1474 {
1475         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1476         struct netdev_notifier_info_ext *info_ext;
1477
1478         switch (event) {
1479         case NETDEV_DOWN:
1480         case NETDEV_UNREGISTER:
1481                 nexthop_flush_dev(dev);
1482                 break;
1483         case NETDEV_CHANGE:
1484                 if (!(dev_get_flags(dev) & (IFF_RUNNING | IFF_LOWER_UP)))
1485                         nexthop_flush_dev(dev);
1486                 break;
1487         case NETDEV_CHANGEMTU:
1488                 info_ext = ptr;
1489                 nexthop_sync_mtu(dev, info_ext->ext.mtu);
1490                 rt_cache_flush(dev_net(dev));
1491                 break;
1492         }
1493         return NOTIFY_DONE;
1494 }
1495
1496 static struct notifier_block nh_netdev_notifier = {
1497         .notifier_call = nh_netdev_event,
1498 };
1499
1500 static void __net_exit nexthop_net_exit(struct net *net)
1501 {
1502         rtnl_lock();
1503         flush_all_nexthops(net);
1504         rtnl_unlock();
1505         kfree(net->nexthop.devhash);
1506 }
1507
1508 static int __net_init nexthop_net_init(struct net *net)
1509 {
1510         size_t sz = sizeof(struct hlist_head) * NH_DEV_HASHSIZE;
1511
1512         net->nexthop.rb_root = RB_ROOT;
1513         net->nexthop.devhash = kzalloc(sz, GFP_KERNEL);
1514         if (!net->nexthop.devhash)
1515                 return -ENOMEM;
1516
1517         return 0;
1518 }
1519
1520 static struct pernet_operations nexthop_net_ops = {
1521         .init = nexthop_net_init,
1522         .exit = nexthop_net_exit,
1523 };
1524
1525 static int __init nexthop_init(void)
1526 {
1527         register_pernet_subsys(&nexthop_net_ops);
1528
1529         register_netdevice_notifier(&nh_netdev_notifier);
1530
1531         rtnl_register(PF_UNSPEC, RTM_NEWNEXTHOP, rtm_new_nexthop, NULL, 0);
1532         rtnl_register(PF_UNSPEC, RTM_DELNEXTHOP, rtm_del_nexthop, NULL, 0);
1533         rtnl_register(PF_UNSPEC, RTM_GETNEXTHOP, rtm_get_nexthop,
1534                       rtm_dump_nexthop, 0);
1535
1536         rtnl_register(PF_INET, RTM_NEWNEXTHOP, rtm_new_nexthop, NULL, 0);
1537         rtnl_register(PF_INET, RTM_GETNEXTHOP, NULL, rtm_dump_nexthop, 0);
1538
1539         rtnl_register(PF_INET6, RTM_NEWNEXTHOP, rtm_new_nexthop, NULL, 0);
1540         rtnl_register(PF_INET6, RTM_GETNEXTHOP, NULL, rtm_dump_nexthop, 0);
1541
1542         return 0;
1543 }
1544 subsys_initcall(nexthop_init);