drivers/net/vxlan/vxlan_vnifilter.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  *      Vxlan vni filter for collect metadata mode
   4  *
   5  *      Authors: Roopa Prabhu <roopa@nvidia.com>
   6  *
   7  */
   8
   9 #include <linux/kernel.h>
  10 #include <linux/slab.h>
  11 #include <linux/etherdevice.h>
  12 #include <linux/rhashtable.h>
  13 #include <net/rtnetlink.h>
  14 #include <net/net_namespace.h>
  15 #include <net/sock.h>
  16 #include <net/vxlan.h>
  17
  18 #include "vxlan_private.h"
  19
  20 static inline int vxlan_vni_cmp(struct rhashtable_compare_arg *arg,
  21                                 const void *ptr)
  22 {
  23         const struct vxlan_vni_node *vnode = ptr;
  24         __be32 vni = *(__be32 *)arg->key;
  25
  26         return vnode->vni != vni;
  27 }
  28
  29 const struct rhashtable_params vxlan_vni_rht_params = {
  30         .head_offset = offsetof(struct vxlan_vni_node, vnode),
  31         .key_offset = offsetof(struct vxlan_vni_node, vni),
  32         .key_len = sizeof(__be32),
  33         .nelem_hint = 3,
  34         .max_size = VXLAN_N_VID,
  35         .obj_cmpfn = vxlan_vni_cmp,
  36         .automatic_shrinking = true,
  37 };
  38
  39 static void vxlan_vs_add_del_vninode(struct vxlan_dev *vxlan,
  40                                      struct vxlan_vni_node *v,
  41                                      bool del)
  42 {
  43         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
  44         struct vxlan_dev_node *node;
  45         struct vxlan_sock *vs;
  46
  47         spin_lock(&vn->sock_lock);
  48         if (del) {
  49                 if (!hlist_unhashed(&v->hlist4.hlist))
  50                         hlist_del_init_rcu(&v->hlist4.hlist);
  51 #if IS_ENABLED(CONFIG_IPV6)
  52                 if (!hlist_unhashed(&v->hlist6.hlist))
  53                         hlist_del_init_rcu(&v->hlist6.hlist);
  54 #endif
  55                 goto out;
  56         }
  57
  58 #if IS_ENABLED(CONFIG_IPV6)
  59         vs = rtnl_dereference(vxlan->vn6_sock);
  60         if (vs && v) {
  61                 node = &v->hlist6;
  62                 hlist_add_head_rcu(&node->hlist, vni_head(vs, v->vni));
  63         }
  64 #endif
  65         vs = rtnl_dereference(vxlan->vn4_sock);
  66         if (vs && v) {
  67                 node = &v->hlist4;
  68                 hlist_add_head_rcu(&node->hlist, vni_head(vs, v->vni));
  69         }
  70 out:
  71         spin_unlock(&vn->sock_lock);
  72 }
  73
  74 void vxlan_vs_add_vnigrp(struct vxlan_dev *vxlan,
  75                          struct vxlan_sock *vs,
  76                          bool ipv6)
  77 {
  78         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
  79         struct vxlan_vni_group *vg = rtnl_dereference(vxlan->vnigrp);
  80         struct vxlan_vni_node *v, *tmp;
  81         struct vxlan_dev_node *node;
  82
  83         if (!vg)
  84                 return;
  85
  86         spin_lock(&vn->sock_lock);
  87         list_for_each_entry_safe(v, tmp, &vg->vni_list, vlist) {
  88 #if IS_ENABLED(CONFIG_IPV6)
  89                 if (ipv6)
  90                         node = &v->hlist6;
  91                 else
  92 #endif
  93                         node = &v->hlist4;
  94                 node->vxlan = vxlan;
  95                 hlist_add_head_rcu(&node->hlist, vni_head(vs, v->vni));
  96         }
  97         spin_unlock(&vn->sock_lock);
  98 }
  99
 100 void vxlan_vs_del_vnigrp(struct vxlan_dev *vxlan)
 101 {
 102         struct vxlan_vni_group *vg = rtnl_dereference(vxlan->vnigrp);
 103         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
 104         struct vxlan_vni_node *v, *tmp;
 105
 106         if (!vg)
 107                 return;
 108
 109         spin_lock(&vn->sock_lock);
 110         list_for_each_entry_safe(v, tmp, &vg->vni_list, vlist) {
 111                 hlist_del_init_rcu(&v->hlist4.hlist);
 112 #if IS_ENABLED(CONFIG_IPV6)
 113                 hlist_del_init_rcu(&v->hlist6.hlist);
 114 #endif
 115         }
 116         spin_unlock(&vn->sock_lock);
 117 }
 118
 119 static void vxlan_vnifilter_stats_get(const struct vxlan_vni_node *vninode,
 120                                       struct vxlan_vni_stats *dest)
 121 {
 122         int i;
 123
 124         memset(dest, 0, sizeof(*dest));
 125         for_each_possible_cpu(i) {
 126                 struct vxlan_vni_stats_pcpu *pstats;
 127                 struct vxlan_vni_stats temp;
 128                 unsigned int start;
 129
 130                 pstats = per_cpu_ptr(vninode->stats, i);
 131                 do {
 132                         start = u64_stats_fetch_begin(&pstats->syncp);
 133                         memcpy(&temp, &pstats->stats, sizeof(temp));
 134                 } while (u64_stats_fetch_retry(&pstats->syncp, start));
 135
 136                 dest->rx_packets += temp.rx_packets;
 137                 dest->rx_bytes += temp.rx_bytes;
 138                 dest->rx_drops += temp.rx_drops;
 139                 dest->rx_errors += temp.rx_errors;
 140                 dest->tx_packets += temp.tx_packets;
 141                 dest->tx_bytes += temp.tx_bytes;
 142                 dest->tx_drops += temp.tx_drops;
 143                 dest->tx_errors += temp.tx_errors;
 144         }
 145 }
 146
 147 static void vxlan_vnifilter_stats_add(struct vxlan_vni_node *vninode,
 148                                       int type, unsigned int len)
 149 {
 150         struct vxlan_vni_stats_pcpu *pstats = this_cpu_ptr(vninode->stats);
 151
 152         u64_stats_update_begin(&pstats->syncp);
 153         switch (type) {
 154         case VXLAN_VNI_STATS_RX:
 155                 pstats->stats.rx_bytes += len;
 156                 pstats->stats.rx_packets++;
 157                 break;
 158         case VXLAN_VNI_STATS_RX_DROPS:
 159                 pstats->stats.rx_drops++;
 160                 break;
 161         case VXLAN_VNI_STATS_RX_ERRORS:
 162                 pstats->stats.rx_errors++;
 163                 break;
 164         case VXLAN_VNI_STATS_TX:
 165                 pstats->stats.tx_bytes += len;
 166                 pstats->stats.tx_packets++;
 167                 break;
 168         case VXLAN_VNI_STATS_TX_DROPS:
 169                 pstats->stats.tx_drops++;
 170                 break;
 171         case VXLAN_VNI_STATS_TX_ERRORS:
 172                 pstats->stats.tx_errors++;
 173                 break;
 174         }
 175         u64_stats_update_end(&pstats->syncp);
 176 }
 177
 178 void vxlan_vnifilter_count(struct vxlan_dev *vxlan, __be32 vni,
 179                            struct vxlan_vni_node *vninode,
 180                            int type, unsigned int len)
 181 {
 182         struct vxlan_vni_node *vnode;
 183
 184         if (!(vxlan->cfg.flags & VXLAN_F_VNIFILTER))
 185                 return;
 186
 187         if (vninode) {
 188                 vnode = vninode;
 189         } else {
 190                 vnode = vxlan_vnifilter_lookup(vxlan, vni);
 191                 if (!vnode)
 192                         return;
 193         }
 194
 195         vxlan_vnifilter_stats_add(vnode, type, len);
 196 }
 197
 198 static u32 vnirange(struct vxlan_vni_node *vbegin,
 199                     struct vxlan_vni_node *vend)
 200 {
 201         return (be32_to_cpu(vend->vni) - be32_to_cpu(vbegin->vni));
 202 }
 203
 204 static size_t vxlan_vnifilter_entry_nlmsg_size(void)
 205 {
 206         return NLMSG_ALIGN(sizeof(struct tunnel_msg))
 207                 + nla_total_size(0) /* VXLAN_VNIFILTER_ENTRY */
 208                 + nla_total_size(sizeof(u32)) /* VXLAN_VNIFILTER_ENTRY_START */
 209                 + nla_total_size(sizeof(u32)) /* VXLAN_VNIFILTER_ENTRY_END */
 210                 + nla_total_size(sizeof(struct in6_addr));/* VXLAN_VNIFILTER_ENTRY_GROUP{6} */
 211 }
 212
 213 static int __vnifilter_entry_fill_stats(struct sk_buff *skb,
 214                                         const struct vxlan_vni_node *vbegin)
 215 {
 216         struct vxlan_vni_stats vstats;
 217         struct nlattr *vstats_attr;
 218
 219         vstats_attr = nla_nest_start(skb, VXLAN_VNIFILTER_ENTRY_STATS);
 220         if (!vstats_attr)
 221                 goto out_stats_err;
 222
 223         vxlan_vnifilter_stats_get(vbegin, &vstats);
 224         if (nla_put_u64_64bit(skb, VNIFILTER_ENTRY_STATS_RX_BYTES,
 225                               vstats.rx_bytes, VNIFILTER_ENTRY_STATS_PAD) ||
 226             nla_put_u64_64bit(skb, VNIFILTER_ENTRY_STATS_RX_PKTS,
 227                               vstats.rx_packets, VNIFILTER_ENTRY_STATS_PAD) ||
 228             nla_put_u64_64bit(skb, VNIFILTER_ENTRY_STATS_RX_DROPS,
 229                               vstats.rx_drops, VNIFILTER_ENTRY_STATS_PAD) ||
 230             nla_put_u64_64bit(skb, VNIFILTER_ENTRY_STATS_RX_ERRORS,
 231                               vstats.rx_errors, VNIFILTER_ENTRY_STATS_PAD) ||
 232             nla_put_u64_64bit(skb, VNIFILTER_ENTRY_STATS_TX_BYTES,
 233                               vstats.tx_bytes, VNIFILTER_ENTRY_STATS_PAD) ||
 234             nla_put_u64_64bit(skb, VNIFILTER_ENTRY_STATS_TX_PKTS,
 235                               vstats.tx_packets, VNIFILTER_ENTRY_STATS_PAD) ||
 236             nla_put_u64_64bit(skb, VNIFILTER_ENTRY_STATS_TX_DROPS,
 237                               vstats.tx_drops, VNIFILTER_ENTRY_STATS_PAD) ||
 238             nla_put_u64_64bit(skb, VNIFILTER_ENTRY_STATS_TX_ERRORS,
 239                               vstats.tx_errors, VNIFILTER_ENTRY_STATS_PAD))
 240                 goto out_stats_err;
 241
 242         nla_nest_end(skb, vstats_attr);
 243
 244         return 0;
 245
 246 out_stats_err:
 247         nla_nest_cancel(skb, vstats_attr);
 248         return -EMSGSIZE;
 249 }
 250
 251 static bool vxlan_fill_vni_filter_entry(struct sk_buff *skb,
 252                                         struct vxlan_vni_node *vbegin,
 253                                         struct vxlan_vni_node *vend,
 254                                         bool fill_stats)
 255 {
 256         struct nlattr *ventry;
 257         u32 vs = be32_to_cpu(vbegin->vni);
 258         u32 ve = 0;
 259
 260         if (vbegin != vend)
 261                 ve = be32_to_cpu(vend->vni);
 262
 263         ventry = nla_nest_start(skb, VXLAN_VNIFILTER_ENTRY);
 264         if (!ventry)
 265                 return false;
 266
 267         if (nla_put_u32(skb, VXLAN_VNIFILTER_ENTRY_START, vs))
 268                 goto out_err;
 269
 270         if (ve && nla_put_u32(skb, VXLAN_VNIFILTER_ENTRY_END, ve))
 271                 goto out_err;
 272
 273         if (!vxlan_addr_any(&vbegin->remote_ip)) {
 274                 if (vbegin->remote_ip.sa.sa_family == AF_INET) {
 275                         if (nla_put_in_addr(skb, VXLAN_VNIFILTER_ENTRY_GROUP,
 276                                             vbegin->remote_ip.sin.sin_addr.s_addr))
 277                                 goto out_err;
 278 #if IS_ENABLED(CONFIG_IPV6)
 279                 } else {
 280                         if (nla_put_in6_addr(skb, VXLAN_VNIFILTER_ENTRY_GROUP6,
 281                                              &vbegin->remote_ip.sin6.sin6_addr))
 282                                 goto out_err;
 283 #endif
 284                 }
 285         }
 286
 287         if (fill_stats && __vnifilter_entry_fill_stats(skb, vbegin))
 288                 goto out_err;
 289
 290         nla_nest_end(skb, ventry);
 291
 292         return true;
 293
 294 out_err:
 295         nla_nest_cancel(skb, ventry);
 296
 297         return false;
 298 }
 299
 300 static void vxlan_vnifilter_notify(const struct vxlan_dev *vxlan,
 301                                    struct vxlan_vni_node *vninode, int cmd)
 302 {
 303         struct tunnel_msg *tmsg;
 304         struct sk_buff *skb;
 305         struct nlmsghdr *nlh;
 306         struct net *net = dev_net(vxlan->dev);
 307         int err = -ENOBUFS;
 308
 309         skb = nlmsg_new(vxlan_vnifilter_entry_nlmsg_size(), GFP_KERNEL);
 310         if (!skb)
 311                 goto out_err;
 312
 313         err = -EMSGSIZE;
 314         nlh = nlmsg_put(skb, 0, 0, cmd, sizeof(*tmsg), 0);
 315         if (!nlh)
 316                 goto out_err;
 317         tmsg = nlmsg_data(nlh);
 318         memset(tmsg, 0, sizeof(*tmsg));
 319         tmsg->family = AF_BRIDGE;
 320         tmsg->ifindex = vxlan->dev->ifindex;
 321
 322         if (!vxlan_fill_vni_filter_entry(skb, vninode, vninode, false))
 323                 goto out_err;
 324
 325         nlmsg_end(skb, nlh);
 326         rtnl_notify(skb, net, 0, RTNLGRP_TUNNEL, NULL, GFP_KERNEL);
 327
 328         return;
 329
 330 out_err:
 331         rtnl_set_sk_err(net, RTNLGRP_TUNNEL, err);
 332
 333         kfree_skb(skb);
 334 }
 335
 336 static int vxlan_vnifilter_dump_dev(const struct net_device *dev,
 337                                     struct sk_buff *skb,
 338                                     struct netlink_callback *cb)
 339 {
 340         struct vxlan_vni_node *tmp, *v, *vbegin = NULL, *vend = NULL;
 341         struct vxlan_dev *vxlan = netdev_priv(dev);
 342         struct tunnel_msg *new_tmsg, *tmsg;
 343         int idx = 0, s_idx = cb->args[1];
 344         struct vxlan_vni_group *vg;
 345         struct nlmsghdr *nlh;
 346         bool dump_stats;
 347         int err = 0;
 348
 349         if (!(vxlan->cfg.flags & VXLAN_F_VNIFILTER))
 350                 return -EINVAL;
 351
 352         /* RCU needed because of the vni locking rules (rcu || rtnl) */
 353         vg = rcu_dereference(vxlan->vnigrp);
 354         if (!vg || !vg->num_vnis)
 355                 return 0;
 356
 357         tmsg = nlmsg_data(cb->nlh);
 358         dump_stats = !!(tmsg->flags & TUNNEL_MSG_FLAG_STATS);
 359
 360         nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
 361                         RTM_NEWTUNNEL, sizeof(*new_tmsg), NLM_F_MULTI);
 362         if (!nlh)
 363                 return -EMSGSIZE;
 364         new_tmsg = nlmsg_data(nlh);
 365         memset(new_tmsg, 0, sizeof(*new_tmsg));
 366         new_tmsg->family = PF_BRIDGE;
 367         new_tmsg->ifindex = dev->ifindex;
 368
 369         list_for_each_entry_safe(v, tmp, &vg->vni_list, vlist) {
 370                 if (idx < s_idx) {
 371                         idx++;
 372                         continue;
 373                 }
 374                 if (!vbegin) {
 375                         vbegin = v;
 376                         vend = v;
 377                         continue;
 378                 }
 379                 if (!dump_stats && vnirange(vend, v) == 1 &&
 380                     vxlan_addr_equal(&v->remote_ip, &vend->remote_ip)) {
 381                         goto update_end;
 382                 } else {
 383                         if (!vxlan_fill_vni_filter_entry(skb, vbegin, vend,
 384                                                          dump_stats)) {
 385                                 err = -EMSGSIZE;
 386                                 break;
 387                         }
 388                         idx += vnirange(vbegin, vend) + 1;
 389                         vbegin = v;
 390                 }
 391 update_end:
 392                 vend = v;
 393         }
 394
 395         if (!err && vbegin) {
 396                 if (!vxlan_fill_vni_filter_entry(skb, vbegin, vend, dump_stats))
 397                         err = -EMSGSIZE;
 398         }
 399
 400         cb->args[1] = err ? idx : 0;
 401
 402         nlmsg_end(skb, nlh);
 403
 404         return err;
 405 }
 406
 407 static int vxlan_vnifilter_dump(struct sk_buff *skb, struct netlink_callback *cb)
 408 {
 409         int idx = 0, err = 0, s_idx = cb->args[0];
 410         struct net *net = sock_net(skb->sk);
 411         struct tunnel_msg *tmsg;
 412         struct net_device *dev;
 413
 414         tmsg = nlmsg_data(cb->nlh);
 415
 416         if (tmsg->flags & ~TUNNEL_MSG_VALID_USER_FLAGS) {
 417                 NL_SET_ERR_MSG(cb->extack, "Invalid tunnelmsg flags in ancillary header");
 418                 return -EINVAL;
 419         }
 420
 421         rcu_read_lock();
 422         if (tmsg->ifindex) {
 423                 dev = dev_get_by_index_rcu(net, tmsg->ifindex);
 424                 if (!dev) {
 425                         err = -ENODEV;
 426                         goto out_err;
 427                 }
 428                 if (!netif_is_vxlan(dev)) {
 429                         NL_SET_ERR_MSG(cb->extack,
 430                                        "The device is not a vxlan device");
 431                         err = -EINVAL;
 432                         goto out_err;
 433                 }
 434                 err = vxlan_vnifilter_dump_dev(dev, skb, cb);
 435                 /* if the dump completed without an error we return 0 here */
 436                 if (err != -EMSGSIZE)
 437                         goto out_err;
 438         } else {
 439                 for_each_netdev_rcu(net, dev) {
 440                         if (!netif_is_vxlan(dev))
 441                                 continue;
 442                         if (idx < s_idx)
 443                                 goto skip;
 444                         err = vxlan_vnifilter_dump_dev(dev, skb, cb);
 445                         if (err == -EMSGSIZE)
 446                                 break;
 447 skip:
 448                         idx++;
 449                 }
 450         }
 451         cb->args[0] = idx;
 452         rcu_read_unlock();
 453
 454         return skb->len;
 455
 456 out_err:
 457         rcu_read_unlock();
 458
 459         return err;
 460 }
 461
 462 static const struct nla_policy vni_filter_entry_policy[VXLAN_VNIFILTER_ENTRY_MAX + 1] = {
 463         [VXLAN_VNIFILTER_ENTRY_START] = { .type = NLA_U32 },
 464         [VXLAN_VNIFILTER_ENTRY_END] = { .type = NLA_U32 },
 465         [VXLAN_VNIFILTER_ENTRY_GROUP]   = { .type = NLA_BINARY,
 466                                             .len = sizeof_field(struct iphdr, daddr) },
 467         [VXLAN_VNIFILTER_ENTRY_GROUP6]  = { .type = NLA_BINARY,
 468                                             .len = sizeof(struct in6_addr) },
 469 };
 470
 471 static const struct nla_policy vni_filter_policy[VXLAN_VNIFILTER_MAX + 1] = {
 472         [VXLAN_VNIFILTER_ENTRY] = { .type = NLA_NESTED },
 473 };
 474
 475 static int vxlan_update_default_fdb_entry(struct vxlan_dev *vxlan, __be32 vni,
 476                                           union vxlan_addr *old_remote_ip,
 477                                           union vxlan_addr *remote_ip,
 478                                           struct netlink_ext_ack *extack)
 479 {
 480         struct vxlan_rdst *dst = &vxlan->default_dst;
 481         u32 hash_index;
 482         int err = 0;
 483
 484         hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
 485         spin_lock_bh(&vxlan->hash_lock[hash_index]);
 486         if (remote_ip && !vxlan_addr_any(remote_ip)) {
 487                 err = vxlan_fdb_update(vxlan, all_zeros_mac,
 488                                        remote_ip,
 489                                        NUD_REACHABLE | NUD_PERMANENT,
 490                                        NLM_F_APPEND | NLM_F_CREATE,
 491                                        vxlan->cfg.dst_port,
 492                                        vni,
 493                                        vni,
 494                                        dst->remote_ifindex,
 495                                        NTF_SELF, 0, true, extack);
 496                 if (err) {
 497                         spin_unlock_bh(&vxlan->hash_lock[hash_index]);
 498                         return err;
 499                 }
 500         }
 501
 502         if (old_remote_ip && !vxlan_addr_any(old_remote_ip)) {
 503                 __vxlan_fdb_delete(vxlan, all_zeros_mac,
 504                                    *old_remote_ip,
 505                                    vxlan->cfg.dst_port,
 506                                    vni, vni,
 507                                    dst->remote_ifindex,
 508                                    true);
 509         }
 510         spin_unlock_bh(&vxlan->hash_lock[hash_index]);
 511
 512         return err;
 513 }
 514
 515 static int vxlan_vni_update_group(struct vxlan_dev *vxlan,
 516                                   struct vxlan_vni_node *vninode,
 517                                   union vxlan_addr *group,
 518                                   bool create, bool *changed,
 519                                   struct netlink_ext_ack *extack)
 520 {
 521         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
 522         struct vxlan_rdst *dst = &vxlan->default_dst;
 523         union vxlan_addr *newrip = NULL, *oldrip = NULL;
 524         union vxlan_addr old_remote_ip;
 525         int ret = 0;
 526
 527         memcpy(&old_remote_ip, &vninode->remote_ip, sizeof(old_remote_ip));
 528
 529         /* if per vni remote ip is not present use vxlan dev
 530          * default dst remote ip for fdb entry
 531          */
 532         if (group && !vxlan_addr_any(group)) {
 533                 newrip = group;
 534         } else {
 535                 if (!vxlan_addr_any(&dst->remote_ip))
 536                         newrip = &dst->remote_ip;
 537         }
 538
 539         /* if old rip exists, and no newrip,
 540          * explicitly delete old rip
 541          */
 542         if (!newrip && !vxlan_addr_any(&old_remote_ip))
 543                 oldrip = &old_remote_ip;
 544
 545         if (!newrip && !oldrip)
 546                 return 0;
 547
 548         if (!create && oldrip && newrip && vxlan_addr_equal(oldrip, newrip))
 549                 return 0;
 550
 551         ret = vxlan_update_default_fdb_entry(vxlan, vninode->vni,
 552                                              oldrip, newrip,
 553                                              extack);
 554         if (ret)
 555                 goto out;
 556
 557         if (group)
 558                 memcpy(&vninode->remote_ip, group, sizeof(vninode->remote_ip));
 559
 560         if (vxlan->dev->flags & IFF_UP) {
 561                 if (vxlan_addr_multicast(&old_remote_ip) &&
 562                     !vxlan_group_used(vn, vxlan, vninode->vni,
 563                                       &old_remote_ip,
 564                                       vxlan->default_dst.remote_ifindex)) {
 565                         ret = vxlan_igmp_leave(vxlan, &old_remote_ip,
 566                                                0);
 567                         if (ret)
 568                                 goto out;
 569                 }
 570
 571                 if (vxlan_addr_multicast(&vninode->remote_ip)) {
 572                         ret = vxlan_igmp_join(vxlan, &vninode->remote_ip, 0);
 573                         if (ret == -EADDRINUSE)
 574                                 ret = 0;
 575                         if (ret)
 576                                 goto out;
 577                 }
 578         }
 579
 580         *changed = true;
 581
 582         return 0;
 583 out:
 584         return ret;
 585 }
 586
 587 int vxlan_vnilist_update_group(struct vxlan_dev *vxlan,
 588                                union vxlan_addr *old_remote_ip,
 589                                union vxlan_addr *new_remote_ip,
 590                                struct netlink_ext_ack *extack)
 591 {
 592         struct list_head *headp, *hpos;
 593         struct vxlan_vni_group *vg;
 594         struct vxlan_vni_node *vent;
 595         int ret;
 596
 597         vg = rtnl_dereference(vxlan->vnigrp);
 598
 599         headp = &vg->vni_list;
 600         list_for_each_prev(hpos, headp) {
 601                 vent = list_entry(hpos, struct vxlan_vni_node, vlist);
 602                 if (vxlan_addr_any(&vent->remote_ip)) {
 603                         ret = vxlan_update_default_fdb_entry(vxlan, vent->vni,
 604                                                              old_remote_ip,
 605                                                              new_remote_ip,
 606                                                              extack);
 607                         if (ret)
 608                                 return ret;
 609                 }
 610         }
 611
 612         return 0;
 613 }
 614
 615 static void vxlan_vni_delete_group(struct vxlan_dev *vxlan,
 616                                    struct vxlan_vni_node *vninode)
 617 {
 618         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
 619         struct vxlan_rdst *dst = &vxlan->default_dst;
 620
 621         /* if per vni remote_ip not present, delete the
 622          * default dst remote_ip previously added for this vni
 623          */
 624         if (!vxlan_addr_any(&vninode->remote_ip) ||
 625             !vxlan_addr_any(&dst->remote_ip))
 626                 __vxlan_fdb_delete(vxlan, all_zeros_mac,
 627                                    (vxlan_addr_any(&vninode->remote_ip) ?
 628                                    dst->remote_ip : vninode->remote_ip),
 629                                    vxlan->cfg.dst_port,
 630                                    vninode->vni, vninode->vni,
 631                                    dst->remote_ifindex,
 632                                    true);
 633
 634         if (vxlan->dev->flags & IFF_UP) {
 635                 if (vxlan_addr_multicast(&vninode->remote_ip) &&
 636                     !vxlan_group_used(vn, vxlan, vninode->vni,
 637                                       &vninode->remote_ip,
 638                                       dst->remote_ifindex)) {
 639                         vxlan_igmp_leave(vxlan, &vninode->remote_ip, 0);
 640                 }
 641         }
 642 }
 643
 644 static int vxlan_vni_update(struct vxlan_dev *vxlan,
 645                             struct vxlan_vni_group *vg,
 646                             __be32 vni, union vxlan_addr *group,
 647                             bool *changed,
 648                             struct netlink_ext_ack *extack)
 649 {
 650         struct vxlan_vni_node *vninode;
 651         int ret;
 652
 653         vninode = rhashtable_lookup_fast(&vg->vni_hash, &vni,
 654                                          vxlan_vni_rht_params);
 655         if (!vninode)
 656                 return 0;
 657
 658         ret = vxlan_vni_update_group(vxlan, vninode, group, false, changed,
 659                                      extack);
 660         if (ret)
 661                 return ret;
 662
 663         if (changed)
 664                 vxlan_vnifilter_notify(vxlan, vninode, RTM_NEWTUNNEL);
 665
 666         return 0;
 667 }
 668
 669 static void __vxlan_vni_add_list(struct vxlan_vni_group *vg,
 670                                  struct vxlan_vni_node *v)
 671 {
 672         struct list_head *headp, *hpos;
 673         struct vxlan_vni_node *vent;
 674
 675         headp = &vg->vni_list;
 676         list_for_each_prev(hpos, headp) {
 677                 vent = list_entry(hpos, struct vxlan_vni_node, vlist);
 678                 if (be32_to_cpu(v->vni) < be32_to_cpu(vent->vni))
 679                         continue;
 680                 else
 681                         break;
 682         }
 683         list_add_rcu(&v->vlist, hpos);
 684         vg->num_vnis++;
 685 }
 686
 687 static void __vxlan_vni_del_list(struct vxlan_vni_group *vg,
 688                                  struct vxlan_vni_node *v)
 689 {
 690         list_del_rcu(&v->vlist);
 691         vg->num_vnis--;
 692 }
 693
 694 static struct vxlan_vni_node *vxlan_vni_alloc(struct vxlan_dev *vxlan,
 695                                               __be32 vni)
 696 {
 697         struct vxlan_vni_node *vninode;
 698
 699         vninode = kzalloc(sizeof(*vninode), GFP_KERNEL);
 700         if (!vninode)
 701                 return NULL;
 702         vninode->stats = netdev_alloc_pcpu_stats(struct vxlan_vni_stats_pcpu);
 703         if (!vninode->stats) {
 704                 kfree(vninode);
 705                 return NULL;
 706         }
 707         vninode->vni = vni;
 708         vninode->hlist4.vxlan = vxlan;
 709 #if IS_ENABLED(CONFIG_IPV6)
 710         vninode->hlist6.vxlan = vxlan;
 711 #endif
 712
 713         return vninode;
 714 }
 715
 716 static void vxlan_vni_free(struct vxlan_vni_node *vninode)
 717 {
 718         free_percpu(vninode->stats);
 719         kfree(vninode);
 720 }
 721
 722 static int vxlan_vni_add(struct vxlan_dev *vxlan,
 723                          struct vxlan_vni_group *vg,
 724                          u32 vni, union vxlan_addr *group,
 725                          struct netlink_ext_ack *extack)
 726 {
 727         struct vxlan_vni_node *vninode;
 728         __be32 v = cpu_to_be32(vni);
 729         bool changed = false;
 730         int err = 0;
 731
 732         if (vxlan_vnifilter_lookup(vxlan, v))
 733                 return vxlan_vni_update(vxlan, vg, v, group, &changed, extack);
 734
 735         err = vxlan_vni_in_use(vxlan->net, vxlan, &vxlan->cfg, v);
 736         if (err) {
 737                 NL_SET_ERR_MSG(extack, "VNI in use");
 738                 return err;
 739         }
 740
 741         vninode = vxlan_vni_alloc(vxlan, v);
 742         if (!vninode)
 743                 return -ENOMEM;
 744
 745         err = rhashtable_lookup_insert_fast(&vg->vni_hash,
 746                                             &vninode->vnode,
 747                                             vxlan_vni_rht_params);
 748         if (err) {
 749                 vxlan_vni_free(vninode);
 750                 return err;
 751         }
 752
 753         __vxlan_vni_add_list(vg, vninode);
 754
 755         if (vxlan->dev->flags & IFF_UP)
 756                 vxlan_vs_add_del_vninode(vxlan, vninode, false);
 757
 758         err = vxlan_vni_update_group(vxlan, vninode, group, true, &changed,
 759                                      extack);
 760
 761         if (changed)
 762                 vxlan_vnifilter_notify(vxlan, vninode, RTM_NEWTUNNEL);
 763
 764         return err;
 765 }
 766
 767 static void vxlan_vni_node_rcu_free(struct rcu_head *rcu)
 768 {
 769         struct vxlan_vni_node *v;
 770
 771         v = container_of(rcu, struct vxlan_vni_node, rcu);
 772         vxlan_vni_free(v);
 773 }
 774
 775 static int vxlan_vni_del(struct vxlan_dev *vxlan,
 776                          struct vxlan_vni_group *vg,
 777                          u32 vni, struct netlink_ext_ack *extack)
 778 {
 779         struct vxlan_vni_node *vninode;
 780         __be32 v = cpu_to_be32(vni);
 781         int err = 0;
 782
 783         vg = rtnl_dereference(vxlan->vnigrp);
 784
 785         vninode = rhashtable_lookup_fast(&vg->vni_hash, &v,
 786                                          vxlan_vni_rht_params);
 787         if (!vninode) {
 788                 err = -ENOENT;
 789                 goto out;
 790         }
 791
 792         vxlan_vni_delete_group(vxlan, vninode);
 793
 794         err = rhashtable_remove_fast(&vg->vni_hash,
 795                                      &vninode->vnode,
 796                                      vxlan_vni_rht_params);
 797         if (err)
 798                 goto out;
 799
 800         __vxlan_vni_del_list(vg, vninode);
 801
 802         vxlan_vnifilter_notify(vxlan, vninode, RTM_DELTUNNEL);
 803
 804         if (vxlan->dev->flags & IFF_UP)
 805                 vxlan_vs_add_del_vninode(vxlan, vninode, true);
 806
 807         call_rcu(&vninode->rcu, vxlan_vni_node_rcu_free);
 808
 809         return 0;
 810 out:
 811         return err;
 812 }
 813
 814 static int vxlan_vni_add_del(struct vxlan_dev *vxlan, __u32 start_vni,
 815                              __u32 end_vni, union vxlan_addr *group,
 816                              int cmd, struct netlink_ext_ack *extack)
 817 {
 818         struct vxlan_vni_group *vg;
 819         int v, err = 0;
 820
 821         vg = rtnl_dereference(vxlan->vnigrp);
 822
 823         for (v = start_vni; v <= end_vni; v++) {
 824                 switch (cmd) {
 825                 case RTM_NEWTUNNEL:
 826                         err = vxlan_vni_add(vxlan, vg, v, group, extack);
 827                         break;
 828                 case RTM_DELTUNNEL:
 829                         err = vxlan_vni_del(vxlan, vg, v, extack);
 830                         break;
 831                 default:
 832                         err = -EOPNOTSUPP;
 833                         break;
 834                 }
 835                 if (err)
 836                         goto out;
 837         }
 838
 839         return 0;
 840 out:
 841         return err;
 842 }
 843
 844 static int vxlan_process_vni_filter(struct vxlan_dev *vxlan,
 845                                     struct nlattr *nlvnifilter,
 846                                     int cmd, struct netlink_ext_ack *extack)
 847 {
 848         struct nlattr *vattrs[VXLAN_VNIFILTER_ENTRY_MAX + 1];
 849         u32 vni_start = 0, vni_end = 0;
 850         union vxlan_addr group;
 851         int err;
 852
 853         err = nla_parse_nested(vattrs,
 854                                VXLAN_VNIFILTER_ENTRY_MAX,
 855                                nlvnifilter, vni_filter_entry_policy,
 856                                extack);
 857         if (err)
 858                 return err;
 859
 860         if (vattrs[VXLAN_VNIFILTER_ENTRY_START]) {
 861                 vni_start = nla_get_u32(vattrs[VXLAN_VNIFILTER_ENTRY_START]);
 862                 vni_end = vni_start;
 863         }
 864
 865         if (vattrs[VXLAN_VNIFILTER_ENTRY_END])
 866                 vni_end = nla_get_u32(vattrs[VXLAN_VNIFILTER_ENTRY_END]);
 867
 868         if (!vni_start && !vni_end) {
 869                 NL_SET_ERR_MSG_ATTR(extack, nlvnifilter,
 870                                     "vni start nor end found in vni entry");
 871                 return -EINVAL;
 872         }
 873
 874         if (vattrs[VXLAN_VNIFILTER_ENTRY_GROUP]) {
 875                 group.sin.sin_addr.s_addr =
 876                         nla_get_in_addr(vattrs[VXLAN_VNIFILTER_ENTRY_GROUP]);
 877                 group.sa.sa_family = AF_INET;
 878         } else if (vattrs[VXLAN_VNIFILTER_ENTRY_GROUP6]) {
 879                 group.sin6.sin6_addr =
 880                         nla_get_in6_addr(vattrs[VXLAN_VNIFILTER_ENTRY_GROUP6]);
 881                 group.sa.sa_family = AF_INET6;
 882         } else {
 883                 memset(&group, 0, sizeof(group));
 884         }
 885
 886         if (vxlan_addr_multicast(&group) && !vxlan->default_dst.remote_ifindex) {
 887                 NL_SET_ERR_MSG(extack,
 888                                "Local interface required for multicast remote group");
 889
 890                 return -EINVAL;
 891         }
 892
 893         err = vxlan_vni_add_del(vxlan, vni_start, vni_end, &group, cmd,
 894                                 extack);
 895         if (err)
 896                 return err;
 897
 898         return 0;
 899 }
 900
 901 void vxlan_vnigroup_uninit(struct vxlan_dev *vxlan)
 902 {
 903         struct vxlan_vni_node *v, *tmp;
 904         struct vxlan_vni_group *vg;
 905
 906         vg = rtnl_dereference(vxlan->vnigrp);
 907         list_for_each_entry_safe(v, tmp, &vg->vni_list, vlist) {
 908                 rhashtable_remove_fast(&vg->vni_hash, &v->vnode,
 909                                        vxlan_vni_rht_params);
 910                 hlist_del_init_rcu(&v->hlist4.hlist);
 911 #if IS_ENABLED(CONFIG_IPV6)
 912                 hlist_del_init_rcu(&v->hlist6.hlist);
 913 #endif
 914                 __vxlan_vni_del_list(vg, v);
 915                 vxlan_vnifilter_notify(vxlan, v, RTM_DELTUNNEL);
 916                 call_rcu(&v->rcu, vxlan_vni_node_rcu_free);
 917         }
 918         rhashtable_destroy(&vg->vni_hash);
 919         kfree(vg);
 920 }
 921
 922 int vxlan_vnigroup_init(struct vxlan_dev *vxlan)
 923 {
 924         struct vxlan_vni_group *vg;
 925         int ret;
 926
 927         vg = kzalloc(sizeof(*vg), GFP_KERNEL);
 928         if (!vg)
 929                 return -ENOMEM;
 930         ret = rhashtable_init(&vg->vni_hash, &vxlan_vni_rht_params);
 931         if (ret) {
 932                 kfree(vg);
 933                 return ret;
 934         }
 935         INIT_LIST_HEAD(&vg->vni_list);
 936         rcu_assign_pointer(vxlan->vnigrp, vg);
 937
 938         return 0;
 939 }
 940
 941 static int vxlan_vnifilter_process(struct sk_buff *skb, struct nlmsghdr *nlh,
 942                                    struct netlink_ext_ack *extack)
 943 {
 944         struct net *net = sock_net(skb->sk);
 945         struct tunnel_msg *tmsg;
 946         struct vxlan_dev *vxlan;
 947         struct net_device *dev;
 948         struct nlattr *attr;
 949         int err, vnis = 0;
 950         int rem;
 951
 952         /* this should validate the header and check for remaining bytes */
 953         err = nlmsg_parse(nlh, sizeof(*tmsg), NULL, VXLAN_VNIFILTER_MAX,
 954                           vni_filter_policy, extack);
 955         if (err < 0)
 956                 return err;
 957
 958         tmsg = nlmsg_data(nlh);
 959         dev = __dev_get_by_index(net, tmsg->ifindex);
 960         if (!dev)
 961                 return -ENODEV;
 962
 963         if (!netif_is_vxlan(dev)) {
 964                 NL_SET_ERR_MSG_MOD(extack, "The device is not a vxlan device");
 965                 return -EINVAL;
 966         }
 967
 968         vxlan = netdev_priv(dev);
 969
 970         if (!(vxlan->cfg.flags & VXLAN_F_VNIFILTER))
 971                 return -EOPNOTSUPP;
 972
 973         nlmsg_for_each_attr(attr, nlh, sizeof(*tmsg), rem) {
 974                 switch (nla_type(attr)) {
 975                 case VXLAN_VNIFILTER_ENTRY:
 976                         err = vxlan_process_vni_filter(vxlan, attr,
 977                                                        nlh->nlmsg_type, extack);
 978                         break;
 979                 default:
 980                         continue;
 981                 }
 982                 vnis++;
 983                 if (err)
 984                         break;
 985         }
 986
 987         if (!vnis) {
 988                 NL_SET_ERR_MSG_MOD(extack, "No vnis found to process");
 989                 err = -EINVAL;
 990         }
 991
 992         return err;
 993 }
 994
 995 void vxlan_vnifilter_init(void)
 996 {
 997         rtnl_register_module(THIS_MODULE, PF_BRIDGE, RTM_GETTUNNEL, NULL,
 998                              vxlan_vnifilter_dump, 0);
 999         rtnl_register_module(THIS_MODULE, PF_BRIDGE, RTM_NEWTUNNEL,
1000                              vxlan_vnifilter_process, NULL, 0);
1001         rtnl_register_module(THIS_MODULE, PF_BRIDGE, RTM_DELTUNNEL,
1002                              vxlan_vnifilter_process, NULL, 0);
1003 }
1004
1005 void vxlan_vnifilter_uninit(void)
1006 {
1007         rtnl_unregister(PF_BRIDGE, RTM_GETTUNNEL);
1008         rtnl_unregister(PF_BRIDGE, RTM_NEWTUNNEL);
1009         rtnl_unregister(PF_BRIDGE, RTM_DELTUNNEL);
1010 }