net/mptcp/pm_netlink.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /* Multipath TCP
   3  *
   4  * Copyright (c) 2020, Red Hat, Inc.
   5  */
   6
   7 #define pr_fmt(fmt) "MPTCP: " fmt
   8
   9 #include <linux/inet.h>
  10 #include <linux/kernel.h>
  11 #include <net/tcp.h>
  12 #include <net/netns/generic.h>
  13 #include <net/mptcp.h>
  14 #include <net/genetlink.h>
  15 #include <uapi/linux/mptcp.h>
  16
  17 #include "protocol.h"
  18 #include "mib.h"
  19
  20 /* forward declaration */
  21 static struct genl_family mptcp_genl_family;
  22
  23 static int pm_nl_pernet_id;
  24
  25 struct mptcp_pm_addr_entry {
  26         struct list_head        list;
  27         struct mptcp_addr_info  addr;
  28         u8                      flags;
  29         int                     ifindex;
  30         struct socket           *lsk;
  31 };
  32
  33 struct mptcp_pm_add_entry {
  34         struct list_head        list;
  35         struct mptcp_addr_info  addr;
  36         struct timer_list       add_timer;
  37         struct mptcp_sock       *sock;
  38         u8                      retrans_times;
  39 };
  40
  41 #define MAX_ADDR_ID             255
  42 #define BITMAP_SZ DIV_ROUND_UP(MAX_ADDR_ID + 1, BITS_PER_LONG)
  43
  44 struct pm_nl_pernet {
  45         /* protects pernet updates */
  46         spinlock_t              lock;
  47         struct list_head        local_addr_list;
  48         unsigned int            addrs;
  49         unsigned int            add_addr_signal_max;
  50         unsigned int            add_addr_accept_max;
  51         unsigned int            local_addr_max;
  52         unsigned int            subflows_max;
  53         unsigned int            next_id;
  54         unsigned long           id_bitmap[BITMAP_SZ];
  55 };
  56
  57 #define MPTCP_PM_ADDR_MAX       8
  58 #define ADD_ADDR_RETRANS_MAX    3
  59
  60 static bool addresses_equal(const struct mptcp_addr_info *a,
  61                             struct mptcp_addr_info *b, bool use_port)
  62 {
  63         bool addr_equals = false;
  64
  65         if (a->family == b->family) {
  66                 if (a->family == AF_INET)
  67                         addr_equals = a->addr.s_addr == b->addr.s_addr;
  68 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
  69                 else
  70                         addr_equals = !ipv6_addr_cmp(&a->addr6, &b->addr6);
  71         } else if (a->family == AF_INET) {
  72                 if (ipv6_addr_v4mapped(&b->addr6))
  73                         addr_equals = a->addr.s_addr == b->addr6.s6_addr32[3];
  74         } else if (b->family == AF_INET) {
  75                 if (ipv6_addr_v4mapped(&a->addr6))
  76                         addr_equals = a->addr6.s6_addr32[3] == b->addr.s_addr;
  77 #endif
  78         }
  79
  80         if (!addr_equals)
  81                 return false;
  82         if (!use_port)
  83                 return true;
  84
  85         return a->port == b->port;
  86 }
  87
  88 static bool address_zero(const struct mptcp_addr_info *addr)
  89 {
  90         struct mptcp_addr_info zero;
  91
  92         memset(&zero, 0, sizeof(zero));
  93         zero.family = addr->family;
  94
  95         return addresses_equal(addr, &zero, true);
  96 }
  97
  98 static void local_address(const struct sock_common *skc,
  99                           struct mptcp_addr_info *addr)
 100 {
 101         addr->family = skc->skc_family;
 102         addr->port = htons(skc->skc_num);
 103         if (addr->family == AF_INET)
 104                 addr->addr.s_addr = skc->skc_rcv_saddr;
 105 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 106         else if (addr->family == AF_INET6)
 107                 addr->addr6 = skc->skc_v6_rcv_saddr;
 108 #endif
 109 }
 110
 111 static void remote_address(const struct sock_common *skc,
 112                            struct mptcp_addr_info *addr)
 113 {
 114         addr->family = skc->skc_family;
 115         addr->port = skc->skc_dport;
 116         if (addr->family == AF_INET)
 117                 addr->addr.s_addr = skc->skc_daddr;
 118 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 119         else if (addr->family == AF_INET6)
 120                 addr->addr6 = skc->skc_v6_daddr;
 121 #endif
 122 }
 123
 124 static bool lookup_subflow_by_saddr(const struct list_head *list,
 125                                     struct mptcp_addr_info *saddr)
 126 {
 127         struct mptcp_subflow_context *subflow;
 128         struct mptcp_addr_info cur;
 129         struct sock_common *skc;
 130
 131         list_for_each_entry(subflow, list, node) {
 132                 skc = (struct sock_common *)mptcp_subflow_tcp_sock(subflow);
 133
 134                 local_address(skc, &cur);
 135                 if (addresses_equal(&cur, saddr, saddr->port))
 136                         return true;
 137         }
 138
 139         return false;
 140 }
 141
 142 static bool lookup_subflow_by_daddr(const struct list_head *list,
 143                                     struct mptcp_addr_info *daddr)
 144 {
 145         struct mptcp_subflow_context *subflow;
 146         struct mptcp_addr_info cur;
 147         struct sock_common *skc;
 148
 149         list_for_each_entry(subflow, list, node) {
 150                 skc = (struct sock_common *)mptcp_subflow_tcp_sock(subflow);
 151
 152                 remote_address(skc, &cur);
 153                 if (addresses_equal(&cur, daddr, daddr->port))
 154                         return true;
 155         }
 156
 157         return false;
 158 }
 159
 160 static struct mptcp_pm_addr_entry *
 161 select_local_address(const struct pm_nl_pernet *pernet,
 162                      struct mptcp_sock *msk)
 163 {
 164         struct mptcp_pm_addr_entry *entry, *ret = NULL;
 165         struct sock *sk = (struct sock *)msk;
 166
 167         msk_owned_by_me(msk);
 168
 169         rcu_read_lock();
 170         __mptcp_flush_join_list(msk);
 171         list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
 172                 if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW))
 173                         continue;
 174
 175                 if (entry->addr.family != sk->sk_family) {
 176 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 177                         if ((entry->addr.family == AF_INET &&
 178                              !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) ||
 179                             (sk->sk_family == AF_INET &&
 180                              !ipv6_addr_v4mapped(&entry->addr.addr6)))
 181 #endif
 182                                 continue;
 183                 }
 184
 185                 /* avoid any address already in use by subflows and
 186                  * pending join
 187                  */
 188                 if (!lookup_subflow_by_saddr(&msk->conn_list, &entry->addr)) {
 189                         ret = entry;
 190                         break;
 191                 }
 192         }
 193         rcu_read_unlock();
 194         return ret;
 195 }
 196
 197 static struct mptcp_pm_addr_entry *
 198 select_signal_address(struct pm_nl_pernet *pernet, unsigned int pos)
 199 {
 200         struct mptcp_pm_addr_entry *entry, *ret = NULL;
 201         int i = 0;
 202
 203         rcu_read_lock();
 204         /* do not keep any additional per socket state, just signal
 205          * the address list in order.
 206          * Note: removal from the local address list during the msk life-cycle
 207          * can lead to additional addresses not being announced.
 208          */
 209         list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
 210                 if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL))
 211                         continue;
 212                 if (i++ == pos) {
 213                         ret = entry;
 214                         break;
 215                 }
 216         }
 217         rcu_read_unlock();
 218         return ret;
 219 }
 220
 221 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk)
 222 {
 223         struct pm_nl_pernet *pernet;
 224
 225         pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
 226         return READ_ONCE(pernet->add_addr_signal_max);
 227 }
 228 EXPORT_SYMBOL_GPL(mptcp_pm_get_add_addr_signal_max);
 229
 230 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk)
 231 {
 232         struct pm_nl_pernet *pernet;
 233
 234         pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
 235         return READ_ONCE(pernet->add_addr_accept_max);
 236 }
 237 EXPORT_SYMBOL_GPL(mptcp_pm_get_add_addr_accept_max);
 238
 239 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk)
 240 {
 241         struct pm_nl_pernet *pernet;
 242
 243         pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
 244         return READ_ONCE(pernet->subflows_max);
 245 }
 246 EXPORT_SYMBOL_GPL(mptcp_pm_get_subflows_max);
 247
 248 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk)
 249 {
 250         struct pm_nl_pernet *pernet;
 251
 252         pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
 253         return READ_ONCE(pernet->local_addr_max);
 254 }
 255 EXPORT_SYMBOL_GPL(mptcp_pm_get_local_addr_max);
 256
 257 static void check_work_pending(struct mptcp_sock *msk)
 258 {
 259         if (msk->pm.add_addr_signaled == mptcp_pm_get_add_addr_signal_max(msk) &&
 260             (msk->pm.local_addr_used == mptcp_pm_get_local_addr_max(msk) ||
 261              msk->pm.subflows == mptcp_pm_get_subflows_max(msk)))
 262                 WRITE_ONCE(msk->pm.work_pending, false);
 263 }
 264
 265 struct mptcp_pm_add_entry *
 266 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk,
 267                                 struct mptcp_addr_info *addr)
 268 {
 269         struct mptcp_pm_add_entry *entry;
 270
 271         lockdep_assert_held(&msk->pm.lock);
 272
 273         list_for_each_entry(entry, &msk->pm.anno_list, list) {
 274                 if (addresses_equal(&entry->addr, addr, true))
 275                         return entry;
 276         }
 277
 278         return NULL;
 279 }
 280
 281 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk)
 282 {
 283         struct mptcp_pm_add_entry *entry;
 284         struct mptcp_addr_info saddr;
 285         bool ret = false;
 286
 287         local_address((struct sock_common *)sk, &saddr);
 288
 289         spin_lock_bh(&msk->pm.lock);
 290         list_for_each_entry(entry, &msk->pm.anno_list, list) {
 291                 if (addresses_equal(&entry->addr, &saddr, true)) {
 292                         ret = true;
 293                         goto out;
 294                 }
 295         }
 296
 297 out:
 298         spin_unlock_bh(&msk->pm.lock);
 299         return ret;
 300 }
 301
 302 static void mptcp_pm_add_timer(struct timer_list *timer)
 303 {
 304         struct mptcp_pm_add_entry *entry = from_timer(entry, timer, add_timer);
 305         struct mptcp_sock *msk = entry->sock;
 306         struct sock *sk = (struct sock *)msk;
 307
 308         pr_debug("msk=%p", msk);
 309
 310         if (!msk)
 311                 return;
 312
 313         if (inet_sk_state_load(sk) == TCP_CLOSE)
 314                 return;
 315
 316         if (!entry->addr.id)
 317                 return;
 318
 319         if (mptcp_pm_should_add_signal(msk)) {
 320                 sk_reset_timer(sk, timer, jiffies + TCP_RTO_MAX / 8);
 321                 goto out;
 322         }
 323
 324         spin_lock_bh(&msk->pm.lock);
 325
 326         if (!mptcp_pm_should_add_signal(msk)) {
 327                 pr_debug("retransmit ADD_ADDR id=%d", entry->addr.id);
 328                 mptcp_pm_announce_addr(msk, &entry->addr, false);
 329                 mptcp_pm_add_addr_send_ack(msk);
 330                 entry->retrans_times++;
 331         }
 332
 333         if (entry->retrans_times < ADD_ADDR_RETRANS_MAX)
 334                 sk_reset_timer(sk, timer,
 335                                jiffies + mptcp_get_add_addr_timeout(sock_net(sk)));
 336
 337         spin_unlock_bh(&msk->pm.lock);
 338
 339         if (entry->retrans_times == ADD_ADDR_RETRANS_MAX)
 340                 mptcp_pm_subflow_established(msk);
 341
 342 out:
 343         __sock_put(sk);
 344 }
 345
 346 struct mptcp_pm_add_entry *
 347 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
 348                        struct mptcp_addr_info *addr, bool check_id)
 349 {
 350         struct mptcp_pm_add_entry *entry;
 351         struct sock *sk = (struct sock *)msk;
 352
 353         spin_lock_bh(&msk->pm.lock);
 354         entry = mptcp_lookup_anno_list_by_saddr(msk, addr);
 355         if (entry && (!check_id || entry->addr.id == addr->id))
 356                 entry->retrans_times = ADD_ADDR_RETRANS_MAX;
 357         spin_unlock_bh(&msk->pm.lock);
 358
 359         if (entry && (!check_id || entry->addr.id == addr->id))
 360                 sk_stop_timer_sync(sk, &entry->add_timer);
 361
 362         return entry;
 363 }
 364
 365 static bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
 366                                      struct mptcp_pm_addr_entry *entry)
 367 {
 368         struct mptcp_pm_add_entry *add_entry = NULL;
 369         struct sock *sk = (struct sock *)msk;
 370         struct net *net = sock_net(sk);
 371
 372         lockdep_assert_held(&msk->pm.lock);
 373
 374         if (mptcp_lookup_anno_list_by_saddr(msk, &entry->addr))
 375                 return false;
 376
 377         add_entry = kmalloc(sizeof(*add_entry), GFP_ATOMIC);
 378         if (!add_entry)
 379                 return false;
 380
 381         list_add(&add_entry->list, &msk->pm.anno_list);
 382
 383         add_entry->addr = entry->addr;
 384         add_entry->sock = msk;
 385         add_entry->retrans_times = 0;
 386
 387         timer_setup(&add_entry->add_timer, mptcp_pm_add_timer, 0);
 388         sk_reset_timer(sk, &add_entry->add_timer,
 389                        jiffies + mptcp_get_add_addr_timeout(net));
 390
 391         return true;
 392 }
 393
 394 void mptcp_pm_free_anno_list(struct mptcp_sock *msk)
 395 {
 396         struct mptcp_pm_add_entry *entry, *tmp;
 397         struct sock *sk = (struct sock *)msk;
 398         LIST_HEAD(free_list);
 399
 400         pr_debug("msk=%p", msk);
 401
 402         spin_lock_bh(&msk->pm.lock);
 403         list_splice_init(&msk->pm.anno_list, &free_list);
 404         spin_unlock_bh(&msk->pm.lock);
 405
 406         list_for_each_entry_safe(entry, tmp, &free_list, list) {
 407                 sk_stop_timer_sync(sk, &entry->add_timer);
 408                 kfree(entry);
 409         }
 410 }
 411
 412 static void mptcp_pm_create_subflow_or_signal_addr(struct mptcp_sock *msk)
 413 {
 414         struct sock *sk = (struct sock *)msk;
 415         struct mptcp_pm_addr_entry *local;
 416         unsigned int add_addr_signal_max;
 417         unsigned int local_addr_max;
 418         struct pm_nl_pernet *pernet;
 419         unsigned int subflows_max;
 420
 421         pernet = net_generic(sock_net(sk), pm_nl_pernet_id);
 422
 423         add_addr_signal_max = mptcp_pm_get_add_addr_signal_max(msk);
 424         local_addr_max = mptcp_pm_get_local_addr_max(msk);
 425         subflows_max = mptcp_pm_get_subflows_max(msk);
 426
 427         pr_debug("local %d:%d signal %d:%d subflows %d:%d\n",
 428                  msk->pm.local_addr_used, local_addr_max,
 429                  msk->pm.add_addr_signaled, add_addr_signal_max,
 430                  msk->pm.subflows, subflows_max);
 431
 432         /* check first for announce */
 433         if (msk->pm.add_addr_signaled < add_addr_signal_max) {
 434                 local = select_signal_address(pernet,
 435                                               msk->pm.add_addr_signaled);
 436
 437                 if (local) {
 438                         if (mptcp_pm_alloc_anno_list(msk, local)) {
 439                                 msk->pm.add_addr_signaled++;
 440                                 mptcp_pm_announce_addr(msk, &local->addr, false);
 441                                 mptcp_pm_nl_addr_send_ack(msk);
 442                         }
 443                 } else {
 444                         /* pick failed, avoid fourther attempts later */
 445                         msk->pm.local_addr_used = add_addr_signal_max;
 446                 }
 447
 448                 check_work_pending(msk);
 449         }
 450
 451         /* check if should create a new subflow */
 452         if (msk->pm.local_addr_used < local_addr_max &&
 453             msk->pm.subflows < subflows_max &&
 454             !READ_ONCE(msk->pm.remote_deny_join_id0)) {
 455                 local = select_local_address(pernet, msk);
 456                 if (local) {
 457                         struct mptcp_addr_info remote = { 0 };
 458
 459                         msk->pm.local_addr_used++;
 460                         msk->pm.subflows++;
 461                         check_work_pending(msk);
 462                         remote_address((struct sock_common *)sk, &remote);
 463                         spin_unlock_bh(&msk->pm.lock);
 464                         __mptcp_subflow_connect(sk, &local->addr, &remote,
 465                                                 local->flags, local->ifindex);
 466                         spin_lock_bh(&msk->pm.lock);
 467                         return;
 468                 }
 469
 470                 /* lookup failed, avoid fourther attempts later */
 471                 msk->pm.local_addr_used = local_addr_max;
 472                 check_work_pending(msk);
 473         }
 474 }
 475
 476 static void mptcp_pm_nl_fully_established(struct mptcp_sock *msk)
 477 {
 478         mptcp_pm_create_subflow_or_signal_addr(msk);
 479 }
 480
 481 static void mptcp_pm_nl_subflow_established(struct mptcp_sock *msk)
 482 {
 483         mptcp_pm_create_subflow_or_signal_addr(msk);
 484 }
 485
 486 static void mptcp_pm_nl_add_addr_received(struct mptcp_sock *msk)
 487 {
 488         struct sock *sk = (struct sock *)msk;
 489         unsigned int add_addr_accept_max;
 490         struct mptcp_addr_info remote;
 491         struct mptcp_addr_info local;
 492         unsigned int subflows_max;
 493
 494         add_addr_accept_max = mptcp_pm_get_add_addr_accept_max(msk);
 495         subflows_max = mptcp_pm_get_subflows_max(msk);
 496
 497         pr_debug("accepted %d:%d remote family %d",
 498                  msk->pm.add_addr_accepted, add_addr_accept_max,
 499                  msk->pm.remote.family);
 500
 501         if (lookup_subflow_by_daddr(&msk->conn_list, &msk->pm.remote))
 502                 goto add_addr_echo;
 503
 504         msk->pm.add_addr_accepted++;
 505         msk->pm.subflows++;
 506         if (msk->pm.add_addr_accepted >= add_addr_accept_max ||
 507             msk->pm.subflows >= subflows_max)
 508                 WRITE_ONCE(msk->pm.accept_addr, false);
 509
 510         /* connect to the specified remote address, using whatever
 511          * local address the routing configuration will pick.
 512          */
 513         remote = msk->pm.remote;
 514         if (!remote.port)
 515                 remote.port = sk->sk_dport;
 516         memset(&local, 0, sizeof(local));
 517         local.family = remote.family;
 518
 519         spin_unlock_bh(&msk->pm.lock);
 520         __mptcp_subflow_connect(sk, &local, &remote, 0, 0);
 521         spin_lock_bh(&msk->pm.lock);
 522
 523 add_addr_echo:
 524         mptcp_pm_announce_addr(msk, &msk->pm.remote, true);
 525         mptcp_pm_nl_addr_send_ack(msk);
 526 }
 527
 528 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk)
 529 {
 530         struct mptcp_subflow_context *subflow;
 531
 532         msk_owned_by_me(msk);
 533         lockdep_assert_held(&msk->pm.lock);
 534
 535         if (!mptcp_pm_should_add_signal(msk) &&
 536             !mptcp_pm_should_rm_signal(msk))
 537                 return;
 538
 539         __mptcp_flush_join_list(msk);
 540         subflow = list_first_entry_or_null(&msk->conn_list, typeof(*subflow), node);
 541         if (subflow) {
 542                 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
 543                 bool slow;
 544
 545                 spin_unlock_bh(&msk->pm.lock);
 546                 pr_debug("send ack for %s%s%s",
 547                          mptcp_pm_should_add_signal(msk) ? "add_addr" : "rm_addr",
 548                          mptcp_pm_should_add_signal_ipv6(msk) ? " [ipv6]" : "",
 549                          mptcp_pm_should_add_signal_port(msk) ? " [port]" : "");
 550
 551                 slow = lock_sock_fast(ssk);
 552                 tcp_send_ack(ssk);
 553                 unlock_sock_fast(ssk, slow);
 554                 spin_lock_bh(&msk->pm.lock);
 555         }
 556 }
 557
 558 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
 559                                  struct mptcp_addr_info *addr,
 560                                  u8 bkup)
 561 {
 562         struct mptcp_subflow_context *subflow;
 563
 564         pr_debug("bkup=%d", bkup);
 565
 566         mptcp_for_each_subflow(msk, subflow) {
 567                 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
 568                 struct sock *sk = (struct sock *)msk;
 569                 struct mptcp_addr_info local;
 570                 bool slow;
 571
 572                 local_address((struct sock_common *)ssk, &local);
 573                 if (!addresses_equal(&local, addr, addr->port))
 574                         continue;
 575
 576                 subflow->backup = bkup;
 577                 subflow->send_mp_prio = 1;
 578                 subflow->request_bkup = bkup;
 579                 __MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPPRIOTX);
 580
 581                 spin_unlock_bh(&msk->pm.lock);
 582                 pr_debug("send ack for mp_prio");
 583                 slow = lock_sock_fast(ssk);
 584                 tcp_send_ack(ssk);
 585                 unlock_sock_fast(ssk, slow);
 586                 spin_lock_bh(&msk->pm.lock);
 587
 588                 return 0;
 589         }
 590
 591         return -EINVAL;
 592 }
 593
 594 static void mptcp_pm_nl_rm_addr_or_subflow(struct mptcp_sock *msk,
 595                                            const struct mptcp_rm_list *rm_list,
 596                                            enum linux_mptcp_mib_field rm_type)
 597 {
 598         struct mptcp_subflow_context *subflow, *tmp;
 599         struct sock *sk = (struct sock *)msk;
 600         u8 i;
 601
 602         pr_debug("%s rm_list_nr %d",
 603                  rm_type == MPTCP_MIB_RMADDR ? "address" : "subflow", rm_list->nr);
 604
 605         msk_owned_by_me(msk);
 606
 607         if (!rm_list->nr)
 608                 return;
 609
 610         if (list_empty(&msk->conn_list))
 611                 return;
 612
 613         for (i = 0; i < rm_list->nr; i++) {
 614                 list_for_each_entry_safe(subflow, tmp, &msk->conn_list, node) {
 615                         struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
 616                         int how = RCV_SHUTDOWN | SEND_SHUTDOWN;
 617                         u8 id = subflow->local_id;
 618
 619                         if (rm_type == MPTCP_MIB_RMADDR)
 620                                 id = subflow->remote_id;
 621
 622                         if (rm_list->ids[i] != id)
 623                                 continue;
 624
 625                         pr_debug(" -> %s rm_list_ids[%d]=%u local_id=%u remote_id=%u",
 626                                  rm_type == MPTCP_MIB_RMADDR ? "address" : "subflow",
 627                                  i, rm_list->ids[i], subflow->local_id, subflow->remote_id);
 628                         spin_unlock_bh(&msk->pm.lock);
 629                         mptcp_subflow_shutdown(sk, ssk, how);
 630                         mptcp_close_ssk(sk, ssk, subflow);
 631                         spin_lock_bh(&msk->pm.lock);
 632
 633                         if (rm_type == MPTCP_MIB_RMADDR) {
 634                                 msk->pm.add_addr_accepted--;
 635                                 WRITE_ONCE(msk->pm.accept_addr, true);
 636                         } else if (rm_type == MPTCP_MIB_RMSUBFLOW) {
 637                                 msk->pm.local_addr_used--;
 638                         }
 639                         msk->pm.subflows--;
 640                         __MPTCP_INC_STATS(sock_net(sk), rm_type);
 641                 }
 642         }
 643 }
 644
 645 static void mptcp_pm_nl_rm_addr_received(struct mptcp_sock *msk)
 646 {
 647         mptcp_pm_nl_rm_addr_or_subflow(msk, &msk->pm.rm_list_rx, MPTCP_MIB_RMADDR);
 648 }
 649
 650 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
 651                                      const struct mptcp_rm_list *rm_list)
 652 {
 653         mptcp_pm_nl_rm_addr_or_subflow(msk, rm_list, MPTCP_MIB_RMSUBFLOW);
 654 }
 655
 656 void mptcp_pm_nl_work(struct mptcp_sock *msk)
 657 {
 658         struct mptcp_pm_data *pm = &msk->pm;
 659
 660         msk_owned_by_me(msk);
 661
 662         spin_lock_bh(&msk->pm.lock);
 663
 664         pr_debug("msk=%p status=%x", msk, pm->status);
 665         if (pm->status & BIT(MPTCP_PM_ADD_ADDR_RECEIVED)) {
 666                 pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_RECEIVED);
 667                 mptcp_pm_nl_add_addr_received(msk);
 668         }
 669         if (pm->status & BIT(MPTCP_PM_ADD_ADDR_SEND_ACK)) {
 670                 pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_SEND_ACK);
 671                 mptcp_pm_nl_addr_send_ack(msk);
 672         }
 673         if (pm->status & BIT(MPTCP_PM_RM_ADDR_RECEIVED)) {
 674                 pm->status &= ~BIT(MPTCP_PM_RM_ADDR_RECEIVED);
 675                 mptcp_pm_nl_rm_addr_received(msk);
 676         }
 677         if (pm->status & BIT(MPTCP_PM_ESTABLISHED)) {
 678                 pm->status &= ~BIT(MPTCP_PM_ESTABLISHED);
 679                 mptcp_pm_nl_fully_established(msk);
 680         }
 681         if (pm->status & BIT(MPTCP_PM_SUBFLOW_ESTABLISHED)) {
 682                 pm->status &= ~BIT(MPTCP_PM_SUBFLOW_ESTABLISHED);
 683                 mptcp_pm_nl_subflow_established(msk);
 684         }
 685
 686         spin_unlock_bh(&msk->pm.lock);
 687 }
 688
 689 static bool address_use_port(struct mptcp_pm_addr_entry *entry)
 690 {
 691         return (entry->flags &
 692                 (MPTCP_PM_ADDR_FLAG_SIGNAL | MPTCP_PM_ADDR_FLAG_SUBFLOW)) ==
 693                 MPTCP_PM_ADDR_FLAG_SIGNAL;
 694 }
 695
 696 static int mptcp_pm_nl_append_new_local_addr(struct pm_nl_pernet *pernet,
 697                                              struct mptcp_pm_addr_entry *entry)
 698 {
 699         struct mptcp_pm_addr_entry *cur;
 700         unsigned int addr_max;
 701         int ret = -EINVAL;
 702
 703         spin_lock_bh(&pernet->lock);
 704         /* to keep the code simple, don't do IDR-like allocation for address ID,
 705          * just bail when we exceed limits
 706          */
 707         if (pernet->next_id == MAX_ADDR_ID)
 708                 pernet->next_id = 1;
 709         if (pernet->addrs >= MPTCP_PM_ADDR_MAX)
 710                 goto out;
 711         if (test_bit(entry->addr.id, pernet->id_bitmap))
 712                 goto out;
 713
 714         /* do not insert duplicate address, differentiate on port only
 715          * singled addresses
 716          */
 717         list_for_each_entry(cur, &pernet->local_addr_list, list) {
 718                 if (addresses_equal(&cur->addr, &entry->addr,
 719                                     address_use_port(entry) &&
 720                                     address_use_port(cur)))
 721                         goto out;
 722         }
 723
 724         if (!entry->addr.id) {
 725 find_next:
 726                 entry->addr.id = find_next_zero_bit(pernet->id_bitmap,
 727                                                     MAX_ADDR_ID + 1,
 728                                                     pernet->next_id);
 729                 if ((!entry->addr.id || entry->addr.id > MAX_ADDR_ID) &&
 730                     pernet->next_id != 1) {
 731                         pernet->next_id = 1;
 732                         goto find_next;
 733                 }
 734         }
 735
 736         if (!entry->addr.id || entry->addr.id > MAX_ADDR_ID)
 737                 goto out;
 738
 739         __set_bit(entry->addr.id, pernet->id_bitmap);
 740         if (entry->addr.id > pernet->next_id)
 741                 pernet->next_id = entry->addr.id;
 742
 743         if (entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL) {
 744                 addr_max = pernet->add_addr_signal_max;
 745                 WRITE_ONCE(pernet->add_addr_signal_max, addr_max + 1);
 746         }
 747         if (entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW) {
 748                 addr_max = pernet->local_addr_max;
 749                 WRITE_ONCE(pernet->local_addr_max, addr_max + 1);
 750         }
 751
 752         pernet->addrs++;
 753         list_add_tail_rcu(&entry->list, &pernet->local_addr_list);
 754         ret = entry->addr.id;
 755
 756 out:
 757         spin_unlock_bh(&pernet->lock);
 758         return ret;
 759 }
 760
 761 static int mptcp_pm_nl_create_listen_socket(struct sock *sk,
 762                                             struct mptcp_pm_addr_entry *entry)
 763 {
 764         struct sockaddr_storage addr;
 765         struct mptcp_sock *msk;
 766         struct socket *ssock;
 767         int backlog = 1024;
 768         int err;
 769
 770         err = sock_create_kern(sock_net(sk), entry->addr.family,
 771                                SOCK_STREAM, IPPROTO_MPTCP, &entry->lsk);
 772         if (err)
 773                 return err;
 774
 775         msk = mptcp_sk(entry->lsk->sk);
 776         if (!msk) {
 777                 err = -EINVAL;
 778                 goto out;
 779         }
 780
 781         ssock = __mptcp_nmpc_socket(msk);
 782         if (!ssock) {
 783                 err = -EINVAL;
 784                 goto out;
 785         }
 786
 787         mptcp_info2sockaddr(&entry->addr, &addr, entry->addr.family);
 788         err = kernel_bind(ssock, (struct sockaddr *)&addr,
 789                           sizeof(struct sockaddr_in));
 790         if (err) {
 791                 pr_warn("kernel_bind error, err=%d", err);
 792                 goto out;
 793         }
 794
 795         err = kernel_listen(ssock, backlog);
 796         if (err) {
 797                 pr_warn("kernel_listen error, err=%d", err);
 798                 goto out;
 799         }
 800
 801         return 0;
 802
 803 out:
 804         sock_release(entry->lsk);
 805         return err;
 806 }
 807
 808 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc)
 809 {
 810         struct mptcp_pm_addr_entry *entry;
 811         struct mptcp_addr_info skc_local;
 812         struct mptcp_addr_info msk_local;
 813         struct pm_nl_pernet *pernet;
 814         int ret = -1;
 815
 816         if (WARN_ON_ONCE(!msk))
 817                 return -1;
 818
 819         /* The 0 ID mapping is defined by the first subflow, copied into the msk
 820          * addr
 821          */
 822         local_address((struct sock_common *)msk, &msk_local);
 823         local_address((struct sock_common *)skc, &skc_local);
 824         if (addresses_equal(&msk_local, &skc_local, false))
 825                 return 0;
 826
 827         if (address_zero(&skc_local))
 828                 return 0;
 829
 830         pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
 831
 832         rcu_read_lock();
 833         list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
 834                 if (addresses_equal(&entry->addr, &skc_local, entry->addr.port)) {
 835                         ret = entry->addr.id;
 836                         break;
 837                 }
 838         }
 839         rcu_read_unlock();
 840         if (ret >= 0)
 841                 return ret;
 842
 843         /* address not found, add to local list */
 844         entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
 845         if (!entry)
 846                 return -ENOMEM;
 847
 848         entry->addr = skc_local;
 849         entry->addr.id = 0;
 850         entry->addr.port = 0;
 851         entry->ifindex = 0;
 852         entry->flags = 0;
 853         entry->lsk = NULL;
 854         ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
 855         if (ret < 0)
 856                 kfree(entry);
 857
 858         return ret;
 859 }
 860
 861 void mptcp_pm_nl_data_init(struct mptcp_sock *msk)
 862 {
 863         struct mptcp_pm_data *pm = &msk->pm;
 864         bool subflows;
 865
 866         subflows = !!mptcp_pm_get_subflows_max(msk);
 867         WRITE_ONCE(pm->work_pending, (!!mptcp_pm_get_local_addr_max(msk) && subflows) ||
 868                    !!mptcp_pm_get_add_addr_signal_max(msk));
 869         WRITE_ONCE(pm->accept_addr, !!mptcp_pm_get_add_addr_accept_max(msk) && subflows);
 870         WRITE_ONCE(pm->accept_subflow, subflows);
 871 }
 872
 873 #define MPTCP_PM_CMD_GRP_OFFSET       0
 874 #define MPTCP_PM_EV_GRP_OFFSET        1
 875
 876 static const struct genl_multicast_group mptcp_pm_mcgrps[] = {
 877         [MPTCP_PM_CMD_GRP_OFFSET]       = { .name = MPTCP_PM_CMD_GRP_NAME, },
 878         [MPTCP_PM_EV_GRP_OFFSET]        = { .name = MPTCP_PM_EV_GRP_NAME,
 879                                             .flags = GENL_UNS_ADMIN_PERM,
 880                                           },
 881 };
 882
 883 static const struct nla_policy
 884 mptcp_pm_addr_policy[MPTCP_PM_ADDR_ATTR_MAX + 1] = {
 885         [MPTCP_PM_ADDR_ATTR_FAMILY]     = { .type       = NLA_U16,      },
 886         [MPTCP_PM_ADDR_ATTR_ID]         = { .type       = NLA_U8,       },
 887         [MPTCP_PM_ADDR_ATTR_ADDR4]      = { .type       = NLA_U32,      },
 888         [MPTCP_PM_ADDR_ATTR_ADDR6]      =
 889                 NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
 890         [MPTCP_PM_ADDR_ATTR_PORT]       = { .type       = NLA_U16       },
 891         [MPTCP_PM_ADDR_ATTR_FLAGS]      = { .type       = NLA_U32       },
 892         [MPTCP_PM_ADDR_ATTR_IF_IDX]     = { .type       = NLA_S32       },
 893 };
 894
 895 static const struct nla_policy mptcp_pm_policy[MPTCP_PM_ATTR_MAX + 1] = {
 896         [MPTCP_PM_ATTR_ADDR]            =
 897                                         NLA_POLICY_NESTED(mptcp_pm_addr_policy),
 898         [MPTCP_PM_ATTR_RCV_ADD_ADDRS]   = { .type       = NLA_U32,      },
 899         [MPTCP_PM_ATTR_SUBFLOWS]        = { .type       = NLA_U32,      },
 900 };
 901
 902 static int mptcp_pm_family_to_addr(int family)
 903 {
 904 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 905         if (family == AF_INET6)
 906                 return MPTCP_PM_ADDR_ATTR_ADDR6;
 907 #endif
 908         return MPTCP_PM_ADDR_ATTR_ADDR4;
 909 }
 910
 911 static int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
 912                                bool require_family,
 913                                struct mptcp_pm_addr_entry *entry)
 914 {
 915         struct nlattr *tb[MPTCP_PM_ADDR_ATTR_MAX + 1];
 916         int err, addr_addr;
 917
 918         if (!attr) {
 919                 GENL_SET_ERR_MSG(info, "missing address info");
 920                 return -EINVAL;
 921         }
 922
 923         /* no validation needed - was already done via nested policy */
 924         err = nla_parse_nested_deprecated(tb, MPTCP_PM_ADDR_ATTR_MAX, attr,
 925                                           mptcp_pm_addr_policy, info->extack);
 926         if (err)
 927                 return err;
 928
 929         memset(entry, 0, sizeof(*entry));
 930         if (!tb[MPTCP_PM_ADDR_ATTR_FAMILY]) {
 931                 if (!require_family)
 932                         goto skip_family;
 933
 934                 NL_SET_ERR_MSG_ATTR(info->extack, attr,
 935                                     "missing family");
 936                 return -EINVAL;
 937         }
 938
 939         entry->addr.family = nla_get_u16(tb[MPTCP_PM_ADDR_ATTR_FAMILY]);
 940         if (entry->addr.family != AF_INET
 941 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 942             && entry->addr.family != AF_INET6
 943 #endif
 944             ) {
 945                 NL_SET_ERR_MSG_ATTR(info->extack, attr,
 946                                     "unknown address family");
 947                 return -EINVAL;
 948         }
 949         addr_addr = mptcp_pm_family_to_addr(entry->addr.family);
 950         if (!tb[addr_addr]) {
 951                 NL_SET_ERR_MSG_ATTR(info->extack, attr,
 952                                     "missing address data");
 953                 return -EINVAL;
 954         }
 955
 956 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 957         if (entry->addr.family == AF_INET6)
 958                 entry->addr.addr6 = nla_get_in6_addr(tb[addr_addr]);
 959         else
 960 #endif
 961                 entry->addr.addr.s_addr = nla_get_in_addr(tb[addr_addr]);
 962
 963 skip_family:
 964         if (tb[MPTCP_PM_ADDR_ATTR_IF_IDX]) {
 965                 u32 val = nla_get_s32(tb[MPTCP_PM_ADDR_ATTR_IF_IDX]);
 966
 967                 entry->ifindex = val;
 968         }
 969
 970         if (tb[MPTCP_PM_ADDR_ATTR_ID])
 971                 entry->addr.id = nla_get_u8(tb[MPTCP_PM_ADDR_ATTR_ID]);
 972
 973         if (tb[MPTCP_PM_ADDR_ATTR_FLAGS])
 974                 entry->flags = nla_get_u32(tb[MPTCP_PM_ADDR_ATTR_FLAGS]);
 975
 976         if (tb[MPTCP_PM_ADDR_ATTR_PORT]) {
 977                 if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL)) {
 978                         NL_SET_ERR_MSG_ATTR(info->extack, attr,
 979                                             "flags must have signal when using port");
 980                         return -EINVAL;
 981                 }
 982                 entry->addr.port = htons(nla_get_u16(tb[MPTCP_PM_ADDR_ATTR_PORT]));
 983         }
 984
 985         return 0;
 986 }
 987
 988 static struct pm_nl_pernet *genl_info_pm_nl(struct genl_info *info)
 989 {
 990         return net_generic(genl_info_net(info), pm_nl_pernet_id);
 991 }
 992
 993 static int mptcp_nl_add_subflow_or_signal_addr(struct net *net)
 994 {
 995         struct mptcp_sock *msk;
 996         long s_slot = 0, s_num = 0;
 997
 998         while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
 999                 struct sock *sk = (struct sock *)msk;
1000
1001                 if (!READ_ONCE(msk->fully_established))
1002                         goto next;
1003
1004                 lock_sock(sk);
1005                 spin_lock_bh(&msk->pm.lock);
1006                 mptcp_pm_create_subflow_or_signal_addr(msk);
1007                 spin_unlock_bh(&msk->pm.lock);
1008                 release_sock(sk);
1009
1010 next:
1011                 sock_put(sk);
1012                 cond_resched();
1013         }
1014
1015         return 0;
1016 }
1017
1018 static int mptcp_nl_cmd_add_addr(struct sk_buff *skb, struct genl_info *info)
1019 {
1020         struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1021         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1022         struct mptcp_pm_addr_entry addr, *entry;
1023         int ret;
1024
1025         ret = mptcp_pm_parse_addr(attr, info, true, &addr);
1026         if (ret < 0)
1027                 return ret;
1028
1029         entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1030         if (!entry) {
1031                 GENL_SET_ERR_MSG(info, "can't allocate addr");
1032                 return -ENOMEM;
1033         }
1034
1035         *entry = addr;
1036         if (entry->addr.port) {
1037                 ret = mptcp_pm_nl_create_listen_socket(skb->sk, entry);
1038                 if (ret) {
1039                         GENL_SET_ERR_MSG(info, "create listen socket error");
1040                         kfree(entry);
1041                         return ret;
1042                 }
1043         }
1044         ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
1045         if (ret < 0) {
1046                 GENL_SET_ERR_MSG(info, "too many addresses or duplicate one");
1047                 if (entry->lsk)
1048                         sock_release(entry->lsk);
1049                 kfree(entry);
1050                 return ret;
1051         }
1052
1053         mptcp_nl_add_subflow_or_signal_addr(sock_net(skb->sk));
1054
1055         return 0;
1056 }
1057
1058 static struct mptcp_pm_addr_entry *
1059 __lookup_addr_by_id(struct pm_nl_pernet *pernet, unsigned int id)
1060 {
1061         struct mptcp_pm_addr_entry *entry;
1062
1063         list_for_each_entry(entry, &pernet->local_addr_list, list) {
1064                 if (entry->addr.id == id)
1065                         return entry;
1066         }
1067         return NULL;
1068 }
1069
1070 static bool remove_anno_list_by_saddr(struct mptcp_sock *msk,
1071                                       struct mptcp_addr_info *addr)
1072 {
1073         struct mptcp_pm_add_entry *entry;
1074
1075         entry = mptcp_pm_del_add_timer(msk, addr, false);
1076         if (entry) {
1077                 list_del(&entry->list);
1078                 kfree(entry);
1079                 return true;
1080         }
1081
1082         return false;
1083 }
1084
1085 static bool mptcp_pm_remove_anno_addr(struct mptcp_sock *msk,
1086                                       struct mptcp_addr_info *addr,
1087                                       bool force)
1088 {
1089         struct mptcp_rm_list list = { .nr = 0 };
1090         bool ret;
1091
1092         list.ids[list.nr++] = addr->id;
1093
1094         ret = remove_anno_list_by_saddr(msk, addr);
1095         if (ret || force) {
1096                 spin_lock_bh(&msk->pm.lock);
1097                 mptcp_pm_remove_addr(msk, &list);
1098                 spin_unlock_bh(&msk->pm.lock);
1099         }
1100         return ret;
1101 }
1102
1103 static int mptcp_nl_remove_subflow_and_signal_addr(struct net *net,
1104                                                    struct mptcp_addr_info *addr)
1105 {
1106         struct mptcp_sock *msk;
1107         long s_slot = 0, s_num = 0;
1108         struct mptcp_rm_list list = { .nr = 0 };
1109
1110         pr_debug("remove_id=%d", addr->id);
1111
1112         list.ids[list.nr++] = addr->id;
1113
1114         while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1115                 struct sock *sk = (struct sock *)msk;
1116                 bool remove_subflow;
1117
1118                 if (list_empty(&msk->conn_list)) {
1119                         mptcp_pm_remove_anno_addr(msk, addr, false);
1120                         goto next;
1121                 }
1122
1123                 lock_sock(sk);
1124                 remove_subflow = lookup_subflow_by_saddr(&msk->conn_list, addr);
1125                 mptcp_pm_remove_anno_addr(msk, addr, remove_subflow);
1126                 if (remove_subflow)
1127                         mptcp_pm_remove_subflow(msk, &list);
1128                 release_sock(sk);
1129
1130 next:
1131                 sock_put(sk);
1132                 cond_resched();
1133         }
1134
1135         return 0;
1136 }
1137
1138 /* caller must ensure the RCU grace period is already elapsed */
1139 static void __mptcp_pm_release_addr_entry(struct mptcp_pm_addr_entry *entry)
1140 {
1141         if (entry->lsk)
1142                 sock_release(entry->lsk);
1143         kfree(entry);
1144 }
1145
1146 static int mptcp_nl_remove_id_zero_address(struct net *net,
1147                                            struct mptcp_addr_info *addr)
1148 {
1149         struct mptcp_rm_list list = { .nr = 0 };
1150         long s_slot = 0, s_num = 0;
1151         struct mptcp_sock *msk;
1152
1153         list.ids[list.nr++] = 0;
1154
1155         while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1156                 struct sock *sk = (struct sock *)msk;
1157                 struct mptcp_addr_info msk_local;
1158
1159                 if (list_empty(&msk->conn_list))
1160                         goto next;
1161
1162                 local_address((struct sock_common *)msk, &msk_local);
1163                 if (!addresses_equal(&msk_local, addr, addr->port))
1164                         goto next;
1165
1166                 lock_sock(sk);
1167                 spin_lock_bh(&msk->pm.lock);
1168                 mptcp_pm_remove_addr(msk, &list);
1169                 mptcp_pm_nl_rm_subflow_received(msk, &list);
1170                 spin_unlock_bh(&msk->pm.lock);
1171                 release_sock(sk);
1172
1173 next:
1174                 sock_put(sk);
1175                 cond_resched();
1176         }
1177
1178         return 0;
1179 }
1180
1181 static int mptcp_nl_cmd_del_addr(struct sk_buff *skb, struct genl_info *info)
1182 {
1183         struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1184         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1185         struct mptcp_pm_addr_entry addr, *entry;
1186         unsigned int addr_max;
1187         int ret;
1188
1189         ret = mptcp_pm_parse_addr(attr, info, false, &addr);
1190         if (ret < 0)
1191                 return ret;
1192
1193         /* the zero id address is special: the first address used by the msk
1194          * always gets such an id, so different subflows can have different zero
1195          * id addresses. Additionally zero id is not accounted for in id_bitmap.
1196          * Let's use an 'mptcp_rm_list' instead of the common remove code.
1197          */
1198         if (addr.addr.id == 0)
1199                 return mptcp_nl_remove_id_zero_address(sock_net(skb->sk), &addr.addr);
1200
1201         spin_lock_bh(&pernet->lock);
1202         entry = __lookup_addr_by_id(pernet, addr.addr.id);
1203         if (!entry) {
1204                 GENL_SET_ERR_MSG(info, "address not found");
1205                 spin_unlock_bh(&pernet->lock);
1206                 return -EINVAL;
1207         }
1208         if (entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL) {
1209                 addr_max = pernet->add_addr_signal_max;
1210                 WRITE_ONCE(pernet->add_addr_signal_max, addr_max - 1);
1211         }
1212         if (entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW) {
1213                 addr_max = pernet->local_addr_max;
1214                 WRITE_ONCE(pernet->local_addr_max, addr_max - 1);
1215         }
1216
1217         pernet->addrs--;
1218         list_del_rcu(&entry->list);
1219         __clear_bit(entry->addr.id, pernet->id_bitmap);
1220         spin_unlock_bh(&pernet->lock);
1221
1222         mptcp_nl_remove_subflow_and_signal_addr(sock_net(skb->sk), &entry->addr);
1223         synchronize_rcu();
1224         __mptcp_pm_release_addr_entry(entry);
1225
1226         return ret;
1227 }
1228
1229 static void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
1230                                                struct list_head *rm_list)
1231 {
1232         struct mptcp_rm_list alist = { .nr = 0 }, slist = { .nr = 0 };
1233         struct mptcp_pm_addr_entry *entry;
1234
1235         list_for_each_entry(entry, rm_list, list) {
1236                 if (lookup_subflow_by_saddr(&msk->conn_list, &entry->addr) &&
1237                     alist.nr < MPTCP_RM_IDS_MAX &&
1238                     slist.nr < MPTCP_RM_IDS_MAX) {
1239                         alist.ids[alist.nr++] = entry->addr.id;
1240                         slist.ids[slist.nr++] = entry->addr.id;
1241                 } else if (remove_anno_list_by_saddr(msk, &entry->addr) &&
1242                          alist.nr < MPTCP_RM_IDS_MAX) {
1243                         alist.ids[alist.nr++] = entry->addr.id;
1244                 }
1245         }
1246
1247         if (alist.nr) {
1248                 spin_lock_bh(&msk->pm.lock);
1249                 mptcp_pm_remove_addr(msk, &alist);
1250                 spin_unlock_bh(&msk->pm.lock);
1251         }
1252         if (slist.nr)
1253                 mptcp_pm_remove_subflow(msk, &slist);
1254 }
1255
1256 static void mptcp_nl_remove_addrs_list(struct net *net,
1257                                        struct list_head *rm_list)
1258 {
1259         long s_slot = 0, s_num = 0;
1260         struct mptcp_sock *msk;
1261
1262         if (list_empty(rm_list))
1263                 return;
1264
1265         while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1266                 struct sock *sk = (struct sock *)msk;
1267
1268                 lock_sock(sk);
1269                 mptcp_pm_remove_addrs_and_subflows(msk, rm_list);
1270                 release_sock(sk);
1271
1272                 sock_put(sk);
1273                 cond_resched();
1274         }
1275 }
1276
1277 /* caller must ensure the RCU grace period is already elapsed */
1278 static void __flush_addrs(struct list_head *list)
1279 {
1280         while (!list_empty(list)) {
1281                 struct mptcp_pm_addr_entry *cur;
1282
1283                 cur = list_entry(list->next,
1284                                  struct mptcp_pm_addr_entry, list);
1285                 list_del_rcu(&cur->list);
1286                 __mptcp_pm_release_addr_entry(cur);
1287         }
1288 }
1289
1290 static void __reset_counters(struct pm_nl_pernet *pernet)
1291 {
1292         WRITE_ONCE(pernet->add_addr_signal_max, 0);
1293         WRITE_ONCE(pernet->add_addr_accept_max, 0);
1294         WRITE_ONCE(pernet->local_addr_max, 0);
1295         pernet->addrs = 0;
1296 }
1297
1298 static int mptcp_nl_cmd_flush_addrs(struct sk_buff *skb, struct genl_info *info)
1299 {
1300         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1301         LIST_HEAD(free_list);
1302
1303         spin_lock_bh(&pernet->lock);
1304         list_splice_init(&pernet->local_addr_list, &free_list);
1305         __reset_counters(pernet);
1306         pernet->next_id = 1;
1307         bitmap_zero(pernet->id_bitmap, MAX_ADDR_ID + 1);
1308         spin_unlock_bh(&pernet->lock);
1309         mptcp_nl_remove_addrs_list(sock_net(skb->sk), &free_list);
1310         synchronize_rcu();
1311         __flush_addrs(&free_list);
1312         return 0;
1313 }
1314
1315 static int mptcp_nl_fill_addr(struct sk_buff *skb,
1316                               struct mptcp_pm_addr_entry *entry)
1317 {
1318         struct mptcp_addr_info *addr = &entry->addr;
1319         struct nlattr *attr;
1320
1321         attr = nla_nest_start(skb, MPTCP_PM_ATTR_ADDR);
1322         if (!attr)
1323                 return -EMSGSIZE;
1324
1325         if (nla_put_u16(skb, MPTCP_PM_ADDR_ATTR_FAMILY, addr->family))
1326                 goto nla_put_failure;
1327         if (nla_put_u16(skb, MPTCP_PM_ADDR_ATTR_PORT, ntohs(addr->port)))
1328                 goto nla_put_failure;
1329         if (nla_put_u8(skb, MPTCP_PM_ADDR_ATTR_ID, addr->id))
1330                 goto nla_put_failure;
1331         if (nla_put_u32(skb, MPTCP_PM_ADDR_ATTR_FLAGS, entry->flags))
1332                 goto nla_put_failure;
1333         if (entry->ifindex &&
1334             nla_put_s32(skb, MPTCP_PM_ADDR_ATTR_IF_IDX, entry->ifindex))
1335                 goto nla_put_failure;
1336
1337         if (addr->family == AF_INET &&
1338             nla_put_in_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR4,
1339                             addr->addr.s_addr))
1340                 goto nla_put_failure;
1341 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1342         else if (addr->family == AF_INET6 &&
1343                  nla_put_in6_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR6, &addr->addr6))
1344                 goto nla_put_failure;
1345 #endif
1346         nla_nest_end(skb, attr);
1347         return 0;
1348
1349 nla_put_failure:
1350         nla_nest_cancel(skb, attr);
1351         return -EMSGSIZE;
1352 }
1353
1354 static int mptcp_nl_cmd_get_addr(struct sk_buff *skb, struct genl_info *info)
1355 {
1356         struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1357         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1358         struct mptcp_pm_addr_entry addr, *entry;
1359         struct sk_buff *msg;
1360         void *reply;
1361         int ret;
1362
1363         ret = mptcp_pm_parse_addr(attr, info, false, &addr);
1364         if (ret < 0)
1365                 return ret;
1366
1367         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1368         if (!msg)
1369                 return -ENOMEM;
1370
1371         reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
1372                                   info->genlhdr->cmd);
1373         if (!reply) {
1374                 GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
1375                 ret = -EMSGSIZE;
1376                 goto fail;
1377         }
1378
1379         spin_lock_bh(&pernet->lock);
1380         entry = __lookup_addr_by_id(pernet, addr.addr.id);
1381         if (!entry) {
1382                 GENL_SET_ERR_MSG(info, "address not found");
1383                 ret = -EINVAL;
1384                 goto unlock_fail;
1385         }
1386
1387         ret = mptcp_nl_fill_addr(msg, entry);
1388         if (ret)
1389                 goto unlock_fail;
1390
1391         genlmsg_end(msg, reply);
1392         ret = genlmsg_reply(msg, info);
1393         spin_unlock_bh(&pernet->lock);
1394         return ret;
1395
1396 unlock_fail:
1397         spin_unlock_bh(&pernet->lock);
1398
1399 fail:
1400         nlmsg_free(msg);
1401         return ret;
1402 }
1403
1404 static int mptcp_nl_cmd_dump_addrs(struct sk_buff *msg,
1405                                    struct netlink_callback *cb)
1406 {
1407         struct net *net = sock_net(msg->sk);
1408         struct mptcp_pm_addr_entry *entry;
1409         struct pm_nl_pernet *pernet;
1410         int id = cb->args[0];
1411         void *hdr;
1412         int i;
1413
1414         pernet = net_generic(net, pm_nl_pernet_id);
1415
1416         spin_lock_bh(&pernet->lock);
1417         for (i = id; i < MAX_ADDR_ID + 1; i++) {
1418                 if (test_bit(i, pernet->id_bitmap)) {
1419                         entry = __lookup_addr_by_id(pernet, i);
1420                         if (!entry)
1421                                 break;
1422
1423                         if (entry->addr.id <= id)
1424                                 continue;
1425
1426                         hdr = genlmsg_put(msg, NETLINK_CB(cb->skb).portid,
1427                                           cb->nlh->nlmsg_seq, &mptcp_genl_family,
1428                                           NLM_F_MULTI, MPTCP_PM_CMD_GET_ADDR);
1429                         if (!hdr)
1430                                 break;
1431
1432                         if (mptcp_nl_fill_addr(msg, entry) < 0) {
1433                                 genlmsg_cancel(msg, hdr);
1434                                 break;
1435                         }
1436
1437                         id = entry->addr.id;
1438                         genlmsg_end(msg, hdr);
1439                 }
1440         }
1441         spin_unlock_bh(&pernet->lock);
1442
1443         cb->args[0] = id;
1444         return msg->len;
1445 }
1446
1447 static int parse_limit(struct genl_info *info, int id, unsigned int *limit)
1448 {
1449         struct nlattr *attr = info->attrs[id];
1450
1451         if (!attr)
1452                 return 0;
1453
1454         *limit = nla_get_u32(attr);
1455         if (*limit > MPTCP_PM_ADDR_MAX) {
1456                 GENL_SET_ERR_MSG(info, "limit greater than maximum");
1457                 return -EINVAL;
1458         }
1459         return 0;
1460 }
1461
1462 static int
1463 mptcp_nl_cmd_set_limits(struct sk_buff *skb, struct genl_info *info)
1464 {
1465         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1466         unsigned int rcv_addrs, subflows;
1467         int ret;
1468
1469         spin_lock_bh(&pernet->lock);
1470         rcv_addrs = pernet->add_addr_accept_max;
1471         ret = parse_limit(info, MPTCP_PM_ATTR_RCV_ADD_ADDRS, &rcv_addrs);
1472         if (ret)
1473                 goto unlock;
1474
1475         subflows = pernet->subflows_max;
1476         ret = parse_limit(info, MPTCP_PM_ATTR_SUBFLOWS, &subflows);
1477         if (ret)
1478                 goto unlock;
1479
1480         WRITE_ONCE(pernet->add_addr_accept_max, rcv_addrs);
1481         WRITE_ONCE(pernet->subflows_max, subflows);
1482
1483 unlock:
1484         spin_unlock_bh(&pernet->lock);
1485         return ret;
1486 }
1487
1488 static int
1489 mptcp_nl_cmd_get_limits(struct sk_buff *skb, struct genl_info *info)
1490 {
1491         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1492         struct sk_buff *msg;
1493         void *reply;
1494
1495         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1496         if (!msg)
1497                 return -ENOMEM;
1498
1499         reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
1500                                   MPTCP_PM_CMD_GET_LIMITS);
1501         if (!reply)
1502                 goto fail;
1503
1504         if (nla_put_u32(msg, MPTCP_PM_ATTR_RCV_ADD_ADDRS,
1505                         READ_ONCE(pernet->add_addr_accept_max)))
1506                 goto fail;
1507
1508         if (nla_put_u32(msg, MPTCP_PM_ATTR_SUBFLOWS,
1509                         READ_ONCE(pernet->subflows_max)))
1510                 goto fail;
1511
1512         genlmsg_end(msg, reply);
1513         return genlmsg_reply(msg, info);
1514
1515 fail:
1516         GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
1517         nlmsg_free(msg);
1518         return -EMSGSIZE;
1519 }
1520
1521 static int mptcp_nl_addr_backup(struct net *net,
1522                                 struct mptcp_addr_info *addr,
1523                                 u8 bkup)
1524 {
1525         long s_slot = 0, s_num = 0;
1526         struct mptcp_sock *msk;
1527         int ret = -EINVAL;
1528
1529         while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1530                 struct sock *sk = (struct sock *)msk;
1531
1532                 if (list_empty(&msk->conn_list))
1533                         goto next;
1534
1535                 lock_sock(sk);
1536                 spin_lock_bh(&msk->pm.lock);
1537                 ret = mptcp_pm_nl_mp_prio_send_ack(msk, addr, bkup);
1538                 spin_unlock_bh(&msk->pm.lock);
1539                 release_sock(sk);
1540
1541 next:
1542                 sock_put(sk);
1543                 cond_resched();
1544         }
1545
1546         return ret;
1547 }
1548
1549 static int mptcp_nl_cmd_set_flags(struct sk_buff *skb, struct genl_info *info)
1550 {
1551         struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1552         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1553         struct mptcp_pm_addr_entry addr, *entry;
1554         struct net *net = sock_net(skb->sk);
1555         u8 bkup = 0;
1556         int ret;
1557
1558         ret = mptcp_pm_parse_addr(attr, info, true, &addr);
1559         if (ret < 0)
1560                 return ret;
1561
1562         if (addr.flags & MPTCP_PM_ADDR_FLAG_BACKUP)
1563                 bkup = 1;
1564
1565         list_for_each_entry(entry, &pernet->local_addr_list, list) {
1566                 if (addresses_equal(&entry->addr, &addr.addr, true)) {
1567                         ret = mptcp_nl_addr_backup(net, &entry->addr, bkup);
1568                         if (ret)
1569                                 return ret;
1570
1571                         if (bkup)
1572                                 entry->flags |= MPTCP_PM_ADDR_FLAG_BACKUP;
1573                         else
1574                                 entry->flags &= ~MPTCP_PM_ADDR_FLAG_BACKUP;
1575                 }
1576         }
1577
1578         return 0;
1579 }
1580
1581 static void mptcp_nl_mcast_send(struct net *net, struct sk_buff *nlskb, gfp_t gfp)
1582 {
1583         genlmsg_multicast_netns(&mptcp_genl_family, net,
1584                                 nlskb, 0, MPTCP_PM_EV_GRP_OFFSET, gfp);
1585 }
1586
1587 static int mptcp_event_add_subflow(struct sk_buff *skb, const struct sock *ssk)
1588 {
1589         const struct inet_sock *issk = inet_sk(ssk);
1590         const struct mptcp_subflow_context *sf;
1591
1592         if (nla_put_u16(skb, MPTCP_ATTR_FAMILY, ssk->sk_family))
1593                 return -EMSGSIZE;
1594
1595         switch (ssk->sk_family) {
1596         case AF_INET:
1597                 if (nla_put_in_addr(skb, MPTCP_ATTR_SADDR4, issk->inet_saddr))
1598                         return -EMSGSIZE;
1599                 if (nla_put_in_addr(skb, MPTCP_ATTR_DADDR4, issk->inet_daddr))
1600                         return -EMSGSIZE;
1601                 break;
1602 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1603         case AF_INET6: {
1604                 const struct ipv6_pinfo *np = inet6_sk(ssk);
1605
1606                 if (nla_put_in6_addr(skb, MPTCP_ATTR_SADDR6, &np->saddr))
1607                         return -EMSGSIZE;
1608                 if (nla_put_in6_addr(skb, MPTCP_ATTR_DADDR6, &ssk->sk_v6_daddr))
1609                         return -EMSGSIZE;
1610                 break;
1611         }
1612 #endif
1613         default:
1614                 WARN_ON_ONCE(1);
1615                 return -EMSGSIZE;
1616         }
1617
1618         if (nla_put_be16(skb, MPTCP_ATTR_SPORT, issk->inet_sport))
1619                 return -EMSGSIZE;
1620         if (nla_put_be16(skb, MPTCP_ATTR_DPORT, issk->inet_dport))
1621                 return -EMSGSIZE;
1622
1623         sf = mptcp_subflow_ctx(ssk);
1624         if (WARN_ON_ONCE(!sf))
1625                 return -EINVAL;
1626
1627         if (nla_put_u8(skb, MPTCP_ATTR_LOC_ID, sf->local_id))
1628                 return -EMSGSIZE;
1629
1630         if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, sf->remote_id))
1631                 return -EMSGSIZE;
1632
1633         return 0;
1634 }
1635
1636 static int mptcp_event_put_token_and_ssk(struct sk_buff *skb,
1637                                          const struct mptcp_sock *msk,
1638                                          const struct sock *ssk)
1639 {
1640         const struct sock *sk = (const struct sock *)msk;
1641         const struct mptcp_subflow_context *sf;
1642         u8 sk_err;
1643
1644         if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1645                 return -EMSGSIZE;
1646
1647         if (mptcp_event_add_subflow(skb, ssk))
1648                 return -EMSGSIZE;
1649
1650         sf = mptcp_subflow_ctx(ssk);
1651         if (WARN_ON_ONCE(!sf))
1652                 return -EINVAL;
1653
1654         if (nla_put_u8(skb, MPTCP_ATTR_BACKUP, sf->backup))
1655                 return -EMSGSIZE;
1656
1657         if (ssk->sk_bound_dev_if &&
1658             nla_put_s32(skb, MPTCP_ATTR_IF_IDX, ssk->sk_bound_dev_if))
1659                 return -EMSGSIZE;
1660
1661         sk_err = ssk->sk_err;
1662         if (sk_err && sk->sk_state == TCP_ESTABLISHED &&
1663             nla_put_u8(skb, MPTCP_ATTR_ERROR, sk_err))
1664                 return -EMSGSIZE;
1665
1666         return 0;
1667 }
1668
1669 static int mptcp_event_sub_established(struct sk_buff *skb,
1670                                        const struct mptcp_sock *msk,
1671                                        const struct sock *ssk)
1672 {
1673         return mptcp_event_put_token_and_ssk(skb, msk, ssk);
1674 }
1675
1676 static int mptcp_event_sub_closed(struct sk_buff *skb,
1677                                   const struct mptcp_sock *msk,
1678                                   const struct sock *ssk)
1679 {
1680         const struct mptcp_subflow_context *sf;
1681
1682         if (mptcp_event_put_token_and_ssk(skb, msk, ssk))
1683                 return -EMSGSIZE;
1684
1685         sf = mptcp_subflow_ctx(ssk);
1686         if (!sf->reset_seen)
1687                 return 0;
1688
1689         if (nla_put_u32(skb, MPTCP_ATTR_RESET_REASON, sf->reset_reason))
1690                 return -EMSGSIZE;
1691
1692         if (nla_put_u32(skb, MPTCP_ATTR_RESET_FLAGS, sf->reset_transient))
1693                 return -EMSGSIZE;
1694
1695         return 0;
1696 }
1697
1698 static int mptcp_event_created(struct sk_buff *skb,
1699                                const struct mptcp_sock *msk,
1700                                const struct sock *ssk)
1701 {
1702         int err = nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token);
1703
1704         if (err)
1705                 return err;
1706
1707         return mptcp_event_add_subflow(skb, ssk);
1708 }
1709
1710 void mptcp_event_addr_removed(const struct mptcp_sock *msk, uint8_t id)
1711 {
1712         struct net *net = sock_net((const struct sock *)msk);
1713         struct nlmsghdr *nlh;
1714         struct sk_buff *skb;
1715
1716         if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1717                 return;
1718
1719         skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1720         if (!skb)
1721                 return;
1722
1723         nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0, MPTCP_EVENT_REMOVED);
1724         if (!nlh)
1725                 goto nla_put_failure;
1726
1727         if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1728                 goto nla_put_failure;
1729
1730         if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, id))
1731                 goto nla_put_failure;
1732
1733         genlmsg_end(skb, nlh);
1734         mptcp_nl_mcast_send(net, skb, GFP_ATOMIC);
1735         return;
1736
1737 nla_put_failure:
1738         kfree_skb(skb);
1739 }
1740
1741 void mptcp_event_addr_announced(const struct mptcp_sock *msk,
1742                                 const struct mptcp_addr_info *info)
1743 {
1744         struct net *net = sock_net((const struct sock *)msk);
1745         struct nlmsghdr *nlh;
1746         struct sk_buff *skb;
1747
1748         if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1749                 return;
1750
1751         skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1752         if (!skb)
1753                 return;
1754
1755         nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0,
1756                           MPTCP_EVENT_ANNOUNCED);
1757         if (!nlh)
1758                 goto nla_put_failure;
1759
1760         if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1761                 goto nla_put_failure;
1762
1763         if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, info->id))
1764                 goto nla_put_failure;
1765
1766         if (nla_put_be16(skb, MPTCP_ATTR_DPORT, info->port))
1767                 goto nla_put_failure;
1768
1769         switch (info->family) {
1770         case AF_INET:
1771                 if (nla_put_in_addr(skb, MPTCP_ATTR_DADDR4, info->addr.s_addr))
1772                         goto nla_put_failure;
1773                 break;
1774 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1775         case AF_INET6:
1776                 if (nla_put_in6_addr(skb, MPTCP_ATTR_DADDR6, &info->addr6))
1777                         goto nla_put_failure;
1778                 break;
1779 #endif
1780         default:
1781                 WARN_ON_ONCE(1);
1782                 goto nla_put_failure;
1783         }
1784
1785         genlmsg_end(skb, nlh);
1786         mptcp_nl_mcast_send(net, skb, GFP_ATOMIC);
1787         return;
1788
1789 nla_put_failure:
1790         kfree_skb(skb);
1791 }
1792
1793 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
1794                  const struct sock *ssk, gfp_t gfp)
1795 {
1796         struct net *net = sock_net((const struct sock *)msk);
1797         struct nlmsghdr *nlh;
1798         struct sk_buff *skb;
1799
1800         if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1801                 return;
1802
1803         skb = nlmsg_new(NLMSG_DEFAULT_SIZE, gfp);
1804         if (!skb)
1805                 return;
1806
1807         nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0, type);
1808         if (!nlh)
1809                 goto nla_put_failure;
1810
1811         switch (type) {
1812         case MPTCP_EVENT_UNSPEC:
1813                 WARN_ON_ONCE(1);
1814                 break;
1815         case MPTCP_EVENT_CREATED:
1816         case MPTCP_EVENT_ESTABLISHED:
1817                 if (mptcp_event_created(skb, msk, ssk) < 0)
1818                         goto nla_put_failure;
1819                 break;
1820         case MPTCP_EVENT_CLOSED:
1821                 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token) < 0)
1822                         goto nla_put_failure;
1823                 break;
1824         case MPTCP_EVENT_ANNOUNCED:
1825         case MPTCP_EVENT_REMOVED:
1826                 /* call mptcp_event_addr_announced()/removed instead */
1827                 WARN_ON_ONCE(1);
1828                 break;
1829         case MPTCP_EVENT_SUB_ESTABLISHED:
1830         case MPTCP_EVENT_SUB_PRIORITY:
1831                 if (mptcp_event_sub_established(skb, msk, ssk) < 0)
1832                         goto nla_put_failure;
1833                 break;
1834         case MPTCP_EVENT_SUB_CLOSED:
1835                 if (mptcp_event_sub_closed(skb, msk, ssk) < 0)
1836                         goto nla_put_failure;
1837                 break;
1838         }
1839
1840         genlmsg_end(skb, nlh);
1841         mptcp_nl_mcast_send(net, skb, gfp);
1842         return;
1843
1844 nla_put_failure:
1845         kfree_skb(skb);
1846 }
1847
1848 static const struct genl_small_ops mptcp_pm_ops[] = {
1849         {
1850                 .cmd    = MPTCP_PM_CMD_ADD_ADDR,
1851                 .doit   = mptcp_nl_cmd_add_addr,
1852                 .flags  = GENL_ADMIN_PERM,
1853         },
1854         {
1855                 .cmd    = MPTCP_PM_CMD_DEL_ADDR,
1856                 .doit   = mptcp_nl_cmd_del_addr,
1857                 .flags  = GENL_ADMIN_PERM,
1858         },
1859         {
1860                 .cmd    = MPTCP_PM_CMD_FLUSH_ADDRS,
1861                 .doit   = mptcp_nl_cmd_flush_addrs,
1862                 .flags  = GENL_ADMIN_PERM,
1863         },
1864         {
1865                 .cmd    = MPTCP_PM_CMD_GET_ADDR,
1866                 .doit   = mptcp_nl_cmd_get_addr,
1867                 .dumpit   = mptcp_nl_cmd_dump_addrs,
1868         },
1869         {
1870                 .cmd    = MPTCP_PM_CMD_SET_LIMITS,
1871                 .doit   = mptcp_nl_cmd_set_limits,
1872                 .flags  = GENL_ADMIN_PERM,
1873         },
1874         {
1875                 .cmd    = MPTCP_PM_CMD_GET_LIMITS,
1876                 .doit   = mptcp_nl_cmd_get_limits,
1877         },
1878         {
1879                 .cmd    = MPTCP_PM_CMD_SET_FLAGS,
1880                 .doit   = mptcp_nl_cmd_set_flags,
1881                 .flags  = GENL_ADMIN_PERM,
1882         },
1883 };
1884
1885 static struct genl_family mptcp_genl_family __ro_after_init = {
1886         .name           = MPTCP_PM_NAME,
1887         .version        = MPTCP_PM_VER,
1888         .maxattr        = MPTCP_PM_ATTR_MAX,
1889         .policy         = mptcp_pm_policy,
1890         .netnsok        = true,
1891         .module         = THIS_MODULE,
1892         .small_ops      = mptcp_pm_ops,
1893         .n_small_ops    = ARRAY_SIZE(mptcp_pm_ops),
1894         .mcgrps         = mptcp_pm_mcgrps,
1895         .n_mcgrps       = ARRAY_SIZE(mptcp_pm_mcgrps),
1896 };
1897
1898 static int __net_init pm_nl_init_net(struct net *net)
1899 {
1900         struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
1901
1902         INIT_LIST_HEAD_RCU(&pernet->local_addr_list);
1903         pernet->next_id = 1;
1904         spin_lock_init(&pernet->lock);
1905
1906         /* No need to initialize other pernet fields, the struct is zeroed at
1907          * allocation time.
1908          */
1909
1910         return 0;
1911 }
1912
1913 static void __net_exit pm_nl_exit_net(struct list_head *net_list)
1914 {
1915         struct net *net;
1916
1917         list_for_each_entry(net, net_list, exit_list) {
1918                 struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
1919
1920                 /* net is removed from namespace list, can't race with
1921                  * other modifiers, also netns core already waited for a
1922                  * RCU grace period.
1923                  */
1924                 __flush_addrs(&pernet->local_addr_list);
1925         }
1926 }
1927
1928 static struct pernet_operations mptcp_pm_pernet_ops = {
1929         .init = pm_nl_init_net,
1930         .exit_batch = pm_nl_exit_net,
1931         .id = &pm_nl_pernet_id,
1932         .size = sizeof(struct pm_nl_pernet),
1933 };
1934
1935 void __init mptcp_pm_nl_init(void)
1936 {
1937         if (register_pernet_subsys(&mptcp_pm_pernet_ops) < 0)
1938                 panic("Failed to register MPTCP PM pernet subsystem.\n");
1939
1940         if (genl_register_family(&mptcp_genl_family))
1941                 panic("Failed to register MPTCP PM netlink family\n");
1942 }