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            stale_loss_cnt;
  50         unsigned int            add_addr_signal_max;
  51         unsigned int            add_addr_accept_max;
  52         unsigned int            local_addr_max;
  53         unsigned int            subflows_max;
  54         unsigned int            next_id;
  55         unsigned long           id_bitmap[BITMAP_SZ];
  56 };
  57
  58 #define MPTCP_PM_ADDR_MAX       8
  59 #define ADD_ADDR_RETRANS_MAX    3
  60
  61 static bool addresses_equal(const struct mptcp_addr_info *a,
  62                             struct mptcp_addr_info *b, bool use_port)
  63 {
  64         bool addr_equals = false;
  65
  66         if (a->family == b->family) {
  67                 if (a->family == AF_INET)
  68                         addr_equals = a->addr.s_addr == b->addr.s_addr;
  69 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
  70                 else
  71                         addr_equals = !ipv6_addr_cmp(&a->addr6, &b->addr6);
  72         } else if (a->family == AF_INET) {
  73                 if (ipv6_addr_v4mapped(&b->addr6))
  74                         addr_equals = a->addr.s_addr == b->addr6.s6_addr32[3];
  75         } else if (b->family == AF_INET) {
  76                 if (ipv6_addr_v4mapped(&a->addr6))
  77                         addr_equals = a->addr6.s6_addr32[3] == b->addr.s_addr;
  78 #endif
  79         }
  80
  81         if (!addr_equals)
  82                 return false;
  83         if (!use_port)
  84                 return true;
  85
  86         return a->port == b->port;
  87 }
  88
  89 static bool address_zero(const struct mptcp_addr_info *addr)
  90 {
  91         struct mptcp_addr_info zero;
  92
  93         memset(&zero, 0, sizeof(zero));
  94         zero.family = addr->family;
  95
  96         return addresses_equal(addr, &zero, true);
  97 }
  98
  99 static void local_address(const struct sock_common *skc,
 100                           struct mptcp_addr_info *addr)
 101 {
 102         addr->family = skc->skc_family;
 103         addr->port = htons(skc->skc_num);
 104         if (addr->family == AF_INET)
 105                 addr->addr.s_addr = skc->skc_rcv_saddr;
 106 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 107         else if (addr->family == AF_INET6)
 108                 addr->addr6 = skc->skc_v6_rcv_saddr;
 109 #endif
 110 }
 111
 112 static void remote_address(const struct sock_common *skc,
 113                            struct mptcp_addr_info *addr)
 114 {
 115         addr->family = skc->skc_family;
 116         addr->port = skc->skc_dport;
 117         if (addr->family == AF_INET)
 118                 addr->addr.s_addr = skc->skc_daddr;
 119 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 120         else if (addr->family == AF_INET6)
 121                 addr->addr6 = skc->skc_v6_daddr;
 122 #endif
 123 }
 124
 125 static bool lookup_subflow_by_saddr(const struct list_head *list,
 126                                     struct mptcp_addr_info *saddr)
 127 {
 128         struct mptcp_subflow_context *subflow;
 129         struct mptcp_addr_info cur;
 130         struct sock_common *skc;
 131
 132         list_for_each_entry(subflow, list, node) {
 133                 skc = (struct sock_common *)mptcp_subflow_tcp_sock(subflow);
 134
 135                 local_address(skc, &cur);
 136                 if (addresses_equal(&cur, saddr, saddr->port))
 137                         return true;
 138         }
 139
 140         return false;
 141 }
 142
 143 static bool lookup_subflow_by_daddr(const struct list_head *list,
 144                                     struct mptcp_addr_info *daddr)
 145 {
 146         struct mptcp_subflow_context *subflow;
 147         struct mptcp_addr_info cur;
 148         struct sock_common *skc;
 149
 150         list_for_each_entry(subflow, list, node) {
 151                 skc = (struct sock_common *)mptcp_subflow_tcp_sock(subflow);
 152
 153                 remote_address(skc, &cur);
 154                 if (addresses_equal(&cur, daddr, daddr->port))
 155                         return true;
 156         }
 157
 158         return false;
 159 }
 160
 161 static struct mptcp_pm_addr_entry *
 162 select_local_address(const struct pm_nl_pernet *pernet,
 163                      struct mptcp_sock *msk)
 164 {
 165         struct mptcp_pm_addr_entry *entry, *ret = NULL;
 166         struct sock *sk = (struct sock *)msk;
 167
 168         msk_owned_by_me(msk);
 169
 170         rcu_read_lock();
 171         __mptcp_flush_join_list(msk);
 172         list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
 173                 if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW))
 174                         continue;
 175
 176                 if (entry->addr.family != sk->sk_family) {
 177 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 178                         if ((entry->addr.family == AF_INET &&
 179                              !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) ||
 180                             (sk->sk_family == AF_INET &&
 181                              !ipv6_addr_v4mapped(&entry->addr.addr6)))
 182 #endif
 183                                 continue;
 184                 }
 185
 186                 /* avoid any address already in use by subflows and
 187                  * pending join
 188                  */
 189                 if (!lookup_subflow_by_saddr(&msk->conn_list, &entry->addr)) {
 190                         ret = entry;
 191                         break;
 192                 }
 193         }
 194         rcu_read_unlock();
 195         return ret;
 196 }
 197
 198 static struct mptcp_pm_addr_entry *
 199 select_signal_address(struct pm_nl_pernet *pernet, unsigned int pos)
 200 {
 201         struct mptcp_pm_addr_entry *entry, *ret = NULL;
 202         int i = 0;
 203
 204         rcu_read_lock();
 205         /* do not keep any additional per socket state, just signal
 206          * the address list in order.
 207          * Note: removal from the local address list during the msk life-cycle
 208          * can lead to additional addresses not being announced.
 209          */
 210         list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
 211                 if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL))
 212                         continue;
 213                 if (i++ == pos) {
 214                         ret = entry;
 215                         break;
 216                 }
 217         }
 218         rcu_read_unlock();
 219         return ret;
 220 }
 221
 222 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk)
 223 {
 224         struct pm_nl_pernet *pernet;
 225
 226         pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
 227         return READ_ONCE(pernet->add_addr_signal_max);
 228 }
 229 EXPORT_SYMBOL_GPL(mptcp_pm_get_add_addr_signal_max);
 230
 231 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk)
 232 {
 233         struct pm_nl_pernet *pernet;
 234
 235         pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
 236         return READ_ONCE(pernet->add_addr_accept_max);
 237 }
 238 EXPORT_SYMBOL_GPL(mptcp_pm_get_add_addr_accept_max);
 239
 240 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk)
 241 {
 242         struct pm_nl_pernet *pernet;
 243
 244         pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
 245         return READ_ONCE(pernet->subflows_max);
 246 }
 247 EXPORT_SYMBOL_GPL(mptcp_pm_get_subflows_max);
 248
 249 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk)
 250 {
 251         struct pm_nl_pernet *pernet;
 252
 253         pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
 254         return READ_ONCE(pernet->local_addr_max);
 255 }
 256 EXPORT_SYMBOL_GPL(mptcp_pm_get_local_addr_max);
 257
 258 static void check_work_pending(struct mptcp_sock *msk)
 259 {
 260         if (msk->pm.add_addr_signaled == mptcp_pm_get_add_addr_signal_max(msk) &&
 261             (msk->pm.local_addr_used == mptcp_pm_get_local_addr_max(msk) ||
 262              msk->pm.subflows == mptcp_pm_get_subflows_max(msk)))
 263                 WRITE_ONCE(msk->pm.work_pending, false);
 264 }
 265
 266 struct mptcp_pm_add_entry *
 267 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk,
 268                                 struct mptcp_addr_info *addr)
 269 {
 270         struct mptcp_pm_add_entry *entry;
 271
 272         lockdep_assert_held(&msk->pm.lock);
 273
 274         list_for_each_entry(entry, &msk->pm.anno_list, list) {
 275                 if (addresses_equal(&entry->addr, addr, true))
 276                         return entry;
 277         }
 278
 279         return NULL;
 280 }
 281
 282 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk)
 283 {
 284         struct mptcp_pm_add_entry *entry;
 285         struct mptcp_addr_info saddr;
 286         bool ret = false;
 287
 288         local_address((struct sock_common *)sk, &saddr);
 289
 290         spin_lock_bh(&msk->pm.lock);
 291         list_for_each_entry(entry, &msk->pm.anno_list, list) {
 292                 if (addresses_equal(&entry->addr, &saddr, true)) {
 293                         ret = true;
 294                         goto out;
 295                 }
 296         }
 297
 298 out:
 299         spin_unlock_bh(&msk->pm.lock);
 300         return ret;
 301 }
 302
 303 static void mptcp_pm_add_timer(struct timer_list *timer)
 304 {
 305         struct mptcp_pm_add_entry *entry = from_timer(entry, timer, add_timer);
 306         struct mptcp_sock *msk = entry->sock;
 307         struct sock *sk = (struct sock *)msk;
 308
 309         pr_debug("msk=%p", msk);
 310
 311         if (!msk)
 312                 return;
 313
 314         if (inet_sk_state_load(sk) == TCP_CLOSE)
 315                 return;
 316
 317         if (!entry->addr.id)
 318                 return;
 319
 320         if (mptcp_pm_should_add_signal(msk)) {
 321                 sk_reset_timer(sk, timer, jiffies + TCP_RTO_MAX / 8);
 322                 goto out;
 323         }
 324
 325         spin_lock_bh(&msk->pm.lock);
 326
 327         if (!mptcp_pm_should_add_signal(msk)) {
 328                 pr_debug("retransmit ADD_ADDR id=%d", entry->addr.id);
 329                 mptcp_pm_announce_addr(msk, &entry->addr, false);
 330                 mptcp_pm_add_addr_send_ack(msk);
 331                 entry->retrans_times++;
 332         }
 333
 334         if (entry->retrans_times < ADD_ADDR_RETRANS_MAX)
 335                 sk_reset_timer(sk, timer,
 336                                jiffies + mptcp_get_add_addr_timeout(sock_net(sk)));
 337
 338         spin_unlock_bh(&msk->pm.lock);
 339
 340         if (entry->retrans_times == ADD_ADDR_RETRANS_MAX)
 341                 mptcp_pm_subflow_established(msk);
 342
 343 out:
 344         __sock_put(sk);
 345 }
 346
 347 struct mptcp_pm_add_entry *
 348 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
 349                        struct mptcp_addr_info *addr, bool check_id)
 350 {
 351         struct mptcp_pm_add_entry *entry;
 352         struct sock *sk = (struct sock *)msk;
 353
 354         spin_lock_bh(&msk->pm.lock);
 355         entry = mptcp_lookup_anno_list_by_saddr(msk, addr);
 356         if (entry && (!check_id || entry->addr.id == addr->id))
 357                 entry->retrans_times = ADD_ADDR_RETRANS_MAX;
 358         spin_unlock_bh(&msk->pm.lock);
 359
 360         if (entry && (!check_id || entry->addr.id == addr->id))
 361                 sk_stop_timer_sync(sk, &entry->add_timer);
 362
 363         return entry;
 364 }
 365
 366 static bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
 367                                      struct mptcp_pm_addr_entry *entry)
 368 {
 369         struct mptcp_pm_add_entry *add_entry = NULL;
 370         struct sock *sk = (struct sock *)msk;
 371         struct net *net = sock_net(sk);
 372
 373         lockdep_assert_held(&msk->pm.lock);
 374
 375         if (mptcp_lookup_anno_list_by_saddr(msk, &entry->addr))
 376                 return false;
 377
 378         add_entry = kmalloc(sizeof(*add_entry), GFP_ATOMIC);
 379         if (!add_entry)
 380                 return false;
 381
 382         list_add(&add_entry->list, &msk->pm.anno_list);
 383
 384         add_entry->addr = entry->addr;
 385         add_entry->sock = msk;
 386         add_entry->retrans_times = 0;
 387
 388         timer_setup(&add_entry->add_timer, mptcp_pm_add_timer, 0);
 389         sk_reset_timer(sk, &add_entry->add_timer,
 390                        jiffies + mptcp_get_add_addr_timeout(net));
 391
 392         return true;
 393 }
 394
 395 void mptcp_pm_free_anno_list(struct mptcp_sock *msk)
 396 {
 397         struct mptcp_pm_add_entry *entry, *tmp;
 398         struct sock *sk = (struct sock *)msk;
 399         LIST_HEAD(free_list);
 400
 401         pr_debug("msk=%p", msk);
 402
 403         spin_lock_bh(&msk->pm.lock);
 404         list_splice_init(&msk->pm.anno_list, &free_list);
 405         spin_unlock_bh(&msk->pm.lock);
 406
 407         list_for_each_entry_safe(entry, tmp, &free_list, list) {
 408                 sk_stop_timer_sync(sk, &entry->add_timer);
 409                 kfree(entry);
 410         }
 411 }
 412
 413 static void mptcp_pm_create_subflow_or_signal_addr(struct mptcp_sock *msk)
 414 {
 415         struct sock *sk = (struct sock *)msk;
 416         struct mptcp_pm_addr_entry *local;
 417         unsigned int add_addr_signal_max;
 418         unsigned int local_addr_max;
 419         struct pm_nl_pernet *pernet;
 420         unsigned int subflows_max;
 421
 422         pernet = net_generic(sock_net(sk), pm_nl_pernet_id);
 423
 424         add_addr_signal_max = mptcp_pm_get_add_addr_signal_max(msk);
 425         local_addr_max = mptcp_pm_get_local_addr_max(msk);
 426         subflows_max = mptcp_pm_get_subflows_max(msk);
 427
 428         pr_debug("local %d:%d signal %d:%d subflows %d:%d\n",
 429                  msk->pm.local_addr_used, local_addr_max,
 430                  msk->pm.add_addr_signaled, add_addr_signal_max,
 431                  msk->pm.subflows, subflows_max);
 432
 433         /* check first for announce */
 434         if (msk->pm.add_addr_signaled < add_addr_signal_max) {
 435                 local = select_signal_address(pernet,
 436                                               msk->pm.add_addr_signaled);
 437
 438                 if (local) {
 439                         if (mptcp_pm_alloc_anno_list(msk, local)) {
 440                                 msk->pm.add_addr_signaled++;
 441                                 mptcp_pm_announce_addr(msk, &local->addr, false);
 442                                 mptcp_pm_nl_addr_send_ack(msk);
 443                         }
 444                 } else {
 445                         /* pick failed, avoid fourther attempts later */
 446                         msk->pm.local_addr_used = add_addr_signal_max;
 447                 }
 448
 449                 check_work_pending(msk);
 450         }
 451
 452         /* check if should create a new subflow */
 453         if (msk->pm.local_addr_used < local_addr_max &&
 454             msk->pm.subflows < subflows_max &&
 455             !READ_ONCE(msk->pm.remote_deny_join_id0)) {
 456                 local = select_local_address(pernet, msk);
 457                 if (local) {
 458                         struct mptcp_addr_info remote = { 0 };
 459
 460                         msk->pm.local_addr_used++;
 461                         msk->pm.subflows++;
 462                         check_work_pending(msk);
 463                         remote_address((struct sock_common *)sk, &remote);
 464                         spin_unlock_bh(&msk->pm.lock);
 465                         __mptcp_subflow_connect(sk, &local->addr, &remote,
 466                                                 local->flags, local->ifindex);
 467                         spin_lock_bh(&msk->pm.lock);
 468                         return;
 469                 }
 470
 471                 /* lookup failed, avoid fourther attempts later */
 472                 msk->pm.local_addr_used = local_addr_max;
 473                 check_work_pending(msk);
 474         }
 475 }
 476
 477 static void mptcp_pm_nl_fully_established(struct mptcp_sock *msk)
 478 {
 479         mptcp_pm_create_subflow_or_signal_addr(msk);
 480 }
 481
 482 static void mptcp_pm_nl_subflow_established(struct mptcp_sock *msk)
 483 {
 484         mptcp_pm_create_subflow_or_signal_addr(msk);
 485 }
 486
 487 static void mptcp_pm_nl_add_addr_received(struct mptcp_sock *msk)
 488 {
 489         struct sock *sk = (struct sock *)msk;
 490         unsigned int add_addr_accept_max;
 491         struct mptcp_addr_info remote;
 492         struct mptcp_addr_info local;
 493         unsigned int subflows_max;
 494
 495         add_addr_accept_max = mptcp_pm_get_add_addr_accept_max(msk);
 496         subflows_max = mptcp_pm_get_subflows_max(msk);
 497
 498         pr_debug("accepted %d:%d remote family %d",
 499                  msk->pm.add_addr_accepted, add_addr_accept_max,
 500                  msk->pm.remote.family);
 501
 502         if (lookup_subflow_by_daddr(&msk->conn_list, &msk->pm.remote))
 503                 goto add_addr_echo;
 504
 505         msk->pm.add_addr_accepted++;
 506         msk->pm.subflows++;
 507         if (msk->pm.add_addr_accepted >= add_addr_accept_max ||
 508             msk->pm.subflows >= subflows_max)
 509                 WRITE_ONCE(msk->pm.accept_addr, false);
 510
 511         /* connect to the specified remote address, using whatever
 512          * local address the routing configuration will pick.
 513          */
 514         remote = msk->pm.remote;
 515         if (!remote.port)
 516                 remote.port = sk->sk_dport;
 517         memset(&local, 0, sizeof(local));
 518         local.family = remote.family;
 519
 520         spin_unlock_bh(&msk->pm.lock);
 521         __mptcp_subflow_connect(sk, &local, &remote, 0, 0);
 522         spin_lock_bh(&msk->pm.lock);
 523
 524 add_addr_echo:
 525         mptcp_pm_announce_addr(msk, &msk->pm.remote, true);
 526         mptcp_pm_nl_addr_send_ack(msk);
 527 }
 528
 529 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk)
 530 {
 531         struct mptcp_subflow_context *subflow;
 532
 533         msk_owned_by_me(msk);
 534         lockdep_assert_held(&msk->pm.lock);
 535
 536         if (!mptcp_pm_should_add_signal(msk) &&
 537             !mptcp_pm_should_rm_signal(msk))
 538                 return;
 539
 540         __mptcp_flush_join_list(msk);
 541         subflow = list_first_entry_or_null(&msk->conn_list, typeof(*subflow), node);
 542         if (subflow) {
 543                 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
 544                 bool slow;
 545
 546                 spin_unlock_bh(&msk->pm.lock);
 547                 pr_debug("send ack for %s%s%s",
 548                          mptcp_pm_should_add_signal(msk) ? "add_addr" : "rm_addr",
 549                          mptcp_pm_should_add_signal_ipv6(msk) ? " [ipv6]" : "",
 550                          mptcp_pm_should_add_signal_port(msk) ? " [port]" : "");
 551
 552                 slow = lock_sock_fast(ssk);
 553                 tcp_send_ack(ssk);
 554                 unlock_sock_fast(ssk, slow);
 555                 spin_lock_bh(&msk->pm.lock);
 556         }
 557 }
 558
 559 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
 560                                  struct mptcp_addr_info *addr,
 561                                  u8 bkup)
 562 {
 563         struct mptcp_subflow_context *subflow;
 564
 565         pr_debug("bkup=%d", bkup);
 566
 567         mptcp_for_each_subflow(msk, subflow) {
 568                 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
 569                 struct sock *sk = (struct sock *)msk;
 570                 struct mptcp_addr_info local;
 571                 bool slow;
 572
 573                 local_address((struct sock_common *)ssk, &local);
 574                 if (!addresses_equal(&local, addr, addr->port))
 575                         continue;
 576
 577                 subflow->backup = bkup;
 578                 subflow->send_mp_prio = 1;
 579                 subflow->request_bkup = bkup;
 580                 __MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPPRIOTX);
 581
 582                 spin_unlock_bh(&msk->pm.lock);
 583                 pr_debug("send ack for mp_prio");
 584                 slow = lock_sock_fast(ssk);
 585                 tcp_send_ack(ssk);
 586                 unlock_sock_fast(ssk, slow);
 587                 spin_lock_bh(&msk->pm.lock);
 588
 589                 return 0;
 590         }
 591
 592         return -EINVAL;
 593 }
 594
 595 static void mptcp_pm_nl_rm_addr_or_subflow(struct mptcp_sock *msk,
 596                                            const struct mptcp_rm_list *rm_list,
 597                                            enum linux_mptcp_mib_field rm_type)
 598 {
 599         struct mptcp_subflow_context *subflow, *tmp;
 600         struct sock *sk = (struct sock *)msk;
 601         u8 i;
 602
 603         pr_debug("%s rm_list_nr %d",
 604                  rm_type == MPTCP_MIB_RMADDR ? "address" : "subflow", rm_list->nr);
 605
 606         msk_owned_by_me(msk);
 607
 608         if (!rm_list->nr)
 609                 return;
 610
 611         if (list_empty(&msk->conn_list))
 612                 return;
 613
 614         for (i = 0; i < rm_list->nr; i++) {
 615                 list_for_each_entry_safe(subflow, tmp, &msk->conn_list, node) {
 616                         struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
 617                         int how = RCV_SHUTDOWN | SEND_SHUTDOWN;
 618                         u8 id = subflow->local_id;
 619
 620                         if (rm_type == MPTCP_MIB_RMADDR)
 621                                 id = subflow->remote_id;
 622
 623                         if (rm_list->ids[i] != id)
 624                                 continue;
 625
 626                         pr_debug(" -> %s rm_list_ids[%d]=%u local_id=%u remote_id=%u",
 627                                  rm_type == MPTCP_MIB_RMADDR ? "address" : "subflow",
 628                                  i, rm_list->ids[i], subflow->local_id, subflow->remote_id);
 629                         spin_unlock_bh(&msk->pm.lock);
 630                         mptcp_subflow_shutdown(sk, ssk, how);
 631                         mptcp_close_ssk(sk, ssk, subflow);
 632                         spin_lock_bh(&msk->pm.lock);
 633
 634                         if (rm_type == MPTCP_MIB_RMADDR) {
 635                                 msk->pm.add_addr_accepted--;
 636                                 WRITE_ONCE(msk->pm.accept_addr, true);
 637                         } else if (rm_type == MPTCP_MIB_RMSUBFLOW) {
 638                                 msk->pm.local_addr_used--;
 639                         }
 640                         msk->pm.subflows--;
 641                         __MPTCP_INC_STATS(sock_net(sk), rm_type);
 642                 }
 643         }
 644 }
 645
 646 static void mptcp_pm_nl_rm_addr_received(struct mptcp_sock *msk)
 647 {
 648         mptcp_pm_nl_rm_addr_or_subflow(msk, &msk->pm.rm_list_rx, MPTCP_MIB_RMADDR);
 649 }
 650
 651 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
 652                                      const struct mptcp_rm_list *rm_list)
 653 {
 654         mptcp_pm_nl_rm_addr_or_subflow(msk, rm_list, MPTCP_MIB_RMSUBFLOW);
 655 }
 656
 657 void mptcp_pm_nl_work(struct mptcp_sock *msk)
 658 {
 659         struct mptcp_pm_data *pm = &msk->pm;
 660
 661         msk_owned_by_me(msk);
 662
 663         spin_lock_bh(&msk->pm.lock);
 664
 665         pr_debug("msk=%p status=%x", msk, pm->status);
 666         if (pm->status & BIT(MPTCP_PM_ADD_ADDR_RECEIVED)) {
 667                 pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_RECEIVED);
 668                 mptcp_pm_nl_add_addr_received(msk);
 669         }
 670         if (pm->status & BIT(MPTCP_PM_ADD_ADDR_SEND_ACK)) {
 671                 pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_SEND_ACK);
 672                 mptcp_pm_nl_addr_send_ack(msk);
 673         }
 674         if (pm->status & BIT(MPTCP_PM_RM_ADDR_RECEIVED)) {
 675                 pm->status &= ~BIT(MPTCP_PM_RM_ADDR_RECEIVED);
 676                 mptcp_pm_nl_rm_addr_received(msk);
 677         }
 678         if (pm->status & BIT(MPTCP_PM_ESTABLISHED)) {
 679                 pm->status &= ~BIT(MPTCP_PM_ESTABLISHED);
 680                 mptcp_pm_nl_fully_established(msk);
 681         }
 682         if (pm->status & BIT(MPTCP_PM_SUBFLOW_ESTABLISHED)) {
 683                 pm->status &= ~BIT(MPTCP_PM_SUBFLOW_ESTABLISHED);
 684                 mptcp_pm_nl_subflow_established(msk);
 685         }
 686
 687         spin_unlock_bh(&msk->pm.lock);
 688 }
 689
 690 static bool address_use_port(struct mptcp_pm_addr_entry *entry)
 691 {
 692         return (entry->flags &
 693                 (MPTCP_PM_ADDR_FLAG_SIGNAL | MPTCP_PM_ADDR_FLAG_SUBFLOW)) ==
 694                 MPTCP_PM_ADDR_FLAG_SIGNAL;
 695 }
 696
 697 static int mptcp_pm_nl_append_new_local_addr(struct pm_nl_pernet *pernet,
 698                                              struct mptcp_pm_addr_entry *entry)
 699 {
 700         struct mptcp_pm_addr_entry *cur;
 701         unsigned int addr_max;
 702         int ret = -EINVAL;
 703
 704         spin_lock_bh(&pernet->lock);
 705         /* to keep the code simple, don't do IDR-like allocation for address ID,
 706          * just bail when we exceed limits
 707          */
 708         if (pernet->next_id == MAX_ADDR_ID)
 709                 pernet->next_id = 1;
 710         if (pernet->addrs >= MPTCP_PM_ADDR_MAX)
 711                 goto out;
 712         if (test_bit(entry->addr.id, pernet->id_bitmap))
 713                 goto out;
 714
 715         /* do not insert duplicate address, differentiate on port only
 716          * singled addresses
 717          */
 718         list_for_each_entry(cur, &pernet->local_addr_list, list) {
 719                 if (addresses_equal(&cur->addr, &entry->addr,
 720                                     address_use_port(entry) &&
 721                                     address_use_port(cur)))
 722                         goto out;
 723         }
 724
 725         if (!entry->addr.id) {
 726 find_next:
 727                 entry->addr.id = find_next_zero_bit(pernet->id_bitmap,
 728                                                     MAX_ADDR_ID + 1,
 729                                                     pernet->next_id);
 730                 if ((!entry->addr.id || entry->addr.id > MAX_ADDR_ID) &&
 731                     pernet->next_id != 1) {
 732                         pernet->next_id = 1;
 733                         goto find_next;
 734                 }
 735         }
 736
 737         if (!entry->addr.id || entry->addr.id > MAX_ADDR_ID)
 738                 goto out;
 739
 740         __set_bit(entry->addr.id, pernet->id_bitmap);
 741         if (entry->addr.id > pernet->next_id)
 742                 pernet->next_id = entry->addr.id;
 743
 744         if (entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL) {
 745                 addr_max = pernet->add_addr_signal_max;
 746                 WRITE_ONCE(pernet->add_addr_signal_max, addr_max + 1);
 747         }
 748         if (entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW) {
 749                 addr_max = pernet->local_addr_max;
 750                 WRITE_ONCE(pernet->local_addr_max, addr_max + 1);
 751         }
 752
 753         pernet->addrs++;
 754         list_add_tail_rcu(&entry->list, &pernet->local_addr_list);
 755         ret = entry->addr.id;
 756
 757 out:
 758         spin_unlock_bh(&pernet->lock);
 759         return ret;
 760 }
 761
 762 static int mptcp_pm_nl_create_listen_socket(struct sock *sk,
 763                                             struct mptcp_pm_addr_entry *entry)
 764 {
 765         struct sockaddr_storage addr;
 766         struct mptcp_sock *msk;
 767         struct socket *ssock;
 768         int backlog = 1024;
 769         int err;
 770
 771         err = sock_create_kern(sock_net(sk), entry->addr.family,
 772                                SOCK_STREAM, IPPROTO_MPTCP, &entry->lsk);
 773         if (err)
 774                 return err;
 775
 776         msk = mptcp_sk(entry->lsk->sk);
 777         if (!msk) {
 778                 err = -EINVAL;
 779                 goto out;
 780         }
 781
 782         ssock = __mptcp_nmpc_socket(msk);
 783         if (!ssock) {
 784                 err = -EINVAL;
 785                 goto out;
 786         }
 787
 788         mptcp_info2sockaddr(&entry->addr, &addr, entry->addr.family);
 789         err = kernel_bind(ssock, (struct sockaddr *)&addr,
 790                           sizeof(struct sockaddr_in));
 791         if (err) {
 792                 pr_warn("kernel_bind error, err=%d", err);
 793                 goto out;
 794         }
 795
 796         err = kernel_listen(ssock, backlog);
 797         if (err) {
 798                 pr_warn("kernel_listen error, err=%d", err);
 799                 goto out;
 800         }
 801
 802         return 0;
 803
 804 out:
 805         sock_release(entry->lsk);
 806         return err;
 807 }
 808
 809 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc)
 810 {
 811         struct mptcp_pm_addr_entry *entry;
 812         struct mptcp_addr_info skc_local;
 813         struct mptcp_addr_info msk_local;
 814         struct pm_nl_pernet *pernet;
 815         int ret = -1;
 816
 817         if (WARN_ON_ONCE(!msk))
 818                 return -1;
 819
 820         /* The 0 ID mapping is defined by the first subflow, copied into the msk
 821          * addr
 822          */
 823         local_address((struct sock_common *)msk, &msk_local);
 824         local_address((struct sock_common *)skc, &skc_local);
 825         if (addresses_equal(&msk_local, &skc_local, false))
 826                 return 0;
 827
 828         if (address_zero(&skc_local))
 829                 return 0;
 830
 831         pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
 832
 833         rcu_read_lock();
 834         list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
 835                 if (addresses_equal(&entry->addr, &skc_local, entry->addr.port)) {
 836                         ret = entry->addr.id;
 837                         break;
 838                 }
 839         }
 840         rcu_read_unlock();
 841         if (ret >= 0)
 842                 return ret;
 843
 844         /* address not found, add to local list */
 845         entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
 846         if (!entry)
 847                 return -ENOMEM;
 848
 849         entry->addr = skc_local;
 850         entry->addr.id = 0;
 851         entry->addr.port = 0;
 852         entry->ifindex = 0;
 853         entry->flags = 0;
 854         entry->lsk = NULL;
 855         ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
 856         if (ret < 0)
 857                 kfree(entry);
 858
 859         return ret;
 860 }
 861
 862 void mptcp_pm_nl_data_init(struct mptcp_sock *msk)
 863 {
 864         struct mptcp_pm_data *pm = &msk->pm;
 865         bool subflows;
 866
 867         subflows = !!mptcp_pm_get_subflows_max(msk);
 868         WRITE_ONCE(pm->work_pending, (!!mptcp_pm_get_local_addr_max(msk) && subflows) ||
 869                    !!mptcp_pm_get_add_addr_signal_max(msk));
 870         WRITE_ONCE(pm->accept_addr, !!mptcp_pm_get_add_addr_accept_max(msk) && subflows);
 871         WRITE_ONCE(pm->accept_subflow, subflows);
 872 }
 873
 874 #define MPTCP_PM_CMD_GRP_OFFSET       0
 875 #define MPTCP_PM_EV_GRP_OFFSET        1
 876
 877 static const struct genl_multicast_group mptcp_pm_mcgrps[] = {
 878         [MPTCP_PM_CMD_GRP_OFFSET]       = { .name = MPTCP_PM_CMD_GRP_NAME, },
 879         [MPTCP_PM_EV_GRP_OFFSET]        = { .name = MPTCP_PM_EV_GRP_NAME,
 880                                             .flags = GENL_UNS_ADMIN_PERM,
 881                                           },
 882 };
 883
 884 static const struct nla_policy
 885 mptcp_pm_addr_policy[MPTCP_PM_ADDR_ATTR_MAX + 1] = {
 886         [MPTCP_PM_ADDR_ATTR_FAMILY]     = { .type       = NLA_U16,      },
 887         [MPTCP_PM_ADDR_ATTR_ID]         = { .type       = NLA_U8,       },
 888         [MPTCP_PM_ADDR_ATTR_ADDR4]      = { .type       = NLA_U32,      },
 889         [MPTCP_PM_ADDR_ATTR_ADDR6]      =
 890                 NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
 891         [MPTCP_PM_ADDR_ATTR_PORT]       = { .type       = NLA_U16       },
 892         [MPTCP_PM_ADDR_ATTR_FLAGS]      = { .type       = NLA_U32       },
 893         [MPTCP_PM_ADDR_ATTR_IF_IDX]     = { .type       = NLA_S32       },
 894 };
 895
 896 static const struct nla_policy mptcp_pm_policy[MPTCP_PM_ATTR_MAX + 1] = {
 897         [MPTCP_PM_ATTR_ADDR]            =
 898                                         NLA_POLICY_NESTED(mptcp_pm_addr_policy),
 899         [MPTCP_PM_ATTR_RCV_ADD_ADDRS]   = { .type       = NLA_U32,      },
 900         [MPTCP_PM_ATTR_SUBFLOWS]        = { .type       = NLA_U32,      },
 901 };
 902
 903 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk)
 904 {
 905         struct mptcp_subflow_context *iter, *subflow = mptcp_subflow_ctx(ssk);
 906         struct sock *sk = (struct sock *)msk;
 907         unsigned int active_max_loss_cnt;
 908         struct net *net = sock_net(sk);
 909         unsigned int stale_loss_cnt;
 910         bool slow;
 911
 912         stale_loss_cnt = mptcp_stale_loss_cnt(net);
 913         if (subflow->stale || !stale_loss_cnt || subflow->stale_count <= stale_loss_cnt)
 914                 return;
 915
 916         /* look for another available subflow not in loss state */
 917         active_max_loss_cnt = max_t(int, stale_loss_cnt - 1, 1);
 918         mptcp_for_each_subflow(msk, iter) {
 919                 if (iter != subflow && mptcp_subflow_active(iter) &&
 920                     iter->stale_count < active_max_loss_cnt) {
 921                         /* we have some alternatives, try to mark this subflow as idle ...*/
 922                         slow = lock_sock_fast(ssk);
 923                         if (!tcp_rtx_and_write_queues_empty(ssk)) {
 924                                 subflow->stale = 1;
 925                                 __mptcp_retransmit_pending_data(sk);
 926                                 MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_SUBFLOWSTALE);
 927                         }
 928                         unlock_sock_fast(ssk, slow);
 929
 930                         /* always try to push the pending data regarless of re-injections:
 931                          * we can possibly use backup subflows now, and subflow selection
 932                          * is cheap under the msk socket lock
 933                          */
 934                         __mptcp_push_pending(sk, 0);
 935                         return;
 936                 }
 937         }
 938 }
 939
 940 static int mptcp_pm_family_to_addr(int family)
 941 {
 942 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 943         if (family == AF_INET6)
 944                 return MPTCP_PM_ADDR_ATTR_ADDR6;
 945 #endif
 946         return MPTCP_PM_ADDR_ATTR_ADDR4;
 947 }
 948
 949 static int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
 950                                bool require_family,
 951                                struct mptcp_pm_addr_entry *entry)
 952 {
 953         struct nlattr *tb[MPTCP_PM_ADDR_ATTR_MAX + 1];
 954         int err, addr_addr;
 955
 956         if (!attr) {
 957                 GENL_SET_ERR_MSG(info, "missing address info");
 958                 return -EINVAL;
 959         }
 960
 961         /* no validation needed - was already done via nested policy */
 962         err = nla_parse_nested_deprecated(tb, MPTCP_PM_ADDR_ATTR_MAX, attr,
 963                                           mptcp_pm_addr_policy, info->extack);
 964         if (err)
 965                 return err;
 966
 967         memset(entry, 0, sizeof(*entry));
 968         if (!tb[MPTCP_PM_ADDR_ATTR_FAMILY]) {
 969                 if (!require_family)
 970                         goto skip_family;
 971
 972                 NL_SET_ERR_MSG_ATTR(info->extack, attr,
 973                                     "missing family");
 974                 return -EINVAL;
 975         }
 976
 977         entry->addr.family = nla_get_u16(tb[MPTCP_PM_ADDR_ATTR_FAMILY]);
 978         if (entry->addr.family != AF_INET
 979 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 980             && entry->addr.family != AF_INET6
 981 #endif
 982             ) {
 983                 NL_SET_ERR_MSG_ATTR(info->extack, attr,
 984                                     "unknown address family");
 985                 return -EINVAL;
 986         }
 987         addr_addr = mptcp_pm_family_to_addr(entry->addr.family);
 988         if (!tb[addr_addr]) {
 989                 NL_SET_ERR_MSG_ATTR(info->extack, attr,
 990                                     "missing address data");
 991                 return -EINVAL;
 992         }
 993
 994 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 995         if (entry->addr.family == AF_INET6)
 996                 entry->addr.addr6 = nla_get_in6_addr(tb[addr_addr]);
 997         else
 998 #endif
 999                 entry->addr.addr.s_addr = nla_get_in_addr(tb[addr_addr]);
1000
1001 skip_family:
1002         if (tb[MPTCP_PM_ADDR_ATTR_IF_IDX]) {
1003                 u32 val = nla_get_s32(tb[MPTCP_PM_ADDR_ATTR_IF_IDX]);
1004
1005                 entry->ifindex = val;
1006         }
1007
1008         if (tb[MPTCP_PM_ADDR_ATTR_ID])
1009                 entry->addr.id = nla_get_u8(tb[MPTCP_PM_ADDR_ATTR_ID]);
1010
1011         if (tb[MPTCP_PM_ADDR_ATTR_FLAGS])
1012                 entry->flags = nla_get_u32(tb[MPTCP_PM_ADDR_ATTR_FLAGS]);
1013
1014         if (tb[MPTCP_PM_ADDR_ATTR_PORT]) {
1015                 if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL)) {
1016                         NL_SET_ERR_MSG_ATTR(info->extack, attr,
1017                                             "flags must have signal when using port");
1018                         return -EINVAL;
1019                 }
1020                 entry->addr.port = htons(nla_get_u16(tb[MPTCP_PM_ADDR_ATTR_PORT]));
1021         }
1022
1023         return 0;
1024 }
1025
1026 static struct pm_nl_pernet *genl_info_pm_nl(struct genl_info *info)
1027 {
1028         return net_generic(genl_info_net(info), pm_nl_pernet_id);
1029 }
1030
1031 static int mptcp_nl_add_subflow_or_signal_addr(struct net *net)
1032 {
1033         struct mptcp_sock *msk;
1034         long s_slot = 0, s_num = 0;
1035
1036         while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1037                 struct sock *sk = (struct sock *)msk;
1038
1039                 if (!READ_ONCE(msk->fully_established))
1040                         goto next;
1041
1042                 lock_sock(sk);
1043                 spin_lock_bh(&msk->pm.lock);
1044                 mptcp_pm_create_subflow_or_signal_addr(msk);
1045                 spin_unlock_bh(&msk->pm.lock);
1046                 release_sock(sk);
1047
1048 next:
1049                 sock_put(sk);
1050                 cond_resched();
1051         }
1052
1053         return 0;
1054 }
1055
1056 static int mptcp_nl_cmd_add_addr(struct sk_buff *skb, struct genl_info *info)
1057 {
1058         struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1059         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1060         struct mptcp_pm_addr_entry addr, *entry;
1061         int ret;
1062
1063         ret = mptcp_pm_parse_addr(attr, info, true, &addr);
1064         if (ret < 0)
1065                 return ret;
1066
1067         entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1068         if (!entry) {
1069                 GENL_SET_ERR_MSG(info, "can't allocate addr");
1070                 return -ENOMEM;
1071         }
1072
1073         *entry = addr;
1074         if (entry->addr.port) {
1075                 ret = mptcp_pm_nl_create_listen_socket(skb->sk, entry);
1076                 if (ret) {
1077                         GENL_SET_ERR_MSG(info, "create listen socket error");
1078                         kfree(entry);
1079                         return ret;
1080                 }
1081         }
1082         ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
1083         if (ret < 0) {
1084                 GENL_SET_ERR_MSG(info, "too many addresses or duplicate one");
1085                 if (entry->lsk)
1086                         sock_release(entry->lsk);
1087                 kfree(entry);
1088                 return ret;
1089         }
1090
1091         mptcp_nl_add_subflow_or_signal_addr(sock_net(skb->sk));
1092
1093         return 0;
1094 }
1095
1096 static struct mptcp_pm_addr_entry *
1097 __lookup_addr_by_id(struct pm_nl_pernet *pernet, unsigned int id)
1098 {
1099         struct mptcp_pm_addr_entry *entry;
1100
1101         list_for_each_entry(entry, &pernet->local_addr_list, list) {
1102                 if (entry->addr.id == id)
1103                         return entry;
1104         }
1105         return NULL;
1106 }
1107
1108 static bool remove_anno_list_by_saddr(struct mptcp_sock *msk,
1109                                       struct mptcp_addr_info *addr)
1110 {
1111         struct mptcp_pm_add_entry *entry;
1112
1113         entry = mptcp_pm_del_add_timer(msk, addr, false);
1114         if (entry) {
1115                 list_del(&entry->list);
1116                 kfree(entry);
1117                 return true;
1118         }
1119
1120         return false;
1121 }
1122
1123 static bool mptcp_pm_remove_anno_addr(struct mptcp_sock *msk,
1124                                       struct mptcp_addr_info *addr,
1125                                       bool force)
1126 {
1127         struct mptcp_rm_list list = { .nr = 0 };
1128         bool ret;
1129
1130         list.ids[list.nr++] = addr->id;
1131
1132         ret = remove_anno_list_by_saddr(msk, addr);
1133         if (ret || force) {
1134                 spin_lock_bh(&msk->pm.lock);
1135                 mptcp_pm_remove_addr(msk, &list);
1136                 spin_unlock_bh(&msk->pm.lock);
1137         }
1138         return ret;
1139 }
1140
1141 static int mptcp_nl_remove_subflow_and_signal_addr(struct net *net,
1142                                                    struct mptcp_addr_info *addr)
1143 {
1144         struct mptcp_sock *msk;
1145         long s_slot = 0, s_num = 0;
1146         struct mptcp_rm_list list = { .nr = 0 };
1147
1148         pr_debug("remove_id=%d", addr->id);
1149
1150         list.ids[list.nr++] = addr->id;
1151
1152         while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1153                 struct sock *sk = (struct sock *)msk;
1154                 bool remove_subflow;
1155
1156                 if (list_empty(&msk->conn_list)) {
1157                         mptcp_pm_remove_anno_addr(msk, addr, false);
1158                         goto next;
1159                 }
1160
1161                 lock_sock(sk);
1162                 remove_subflow = lookup_subflow_by_saddr(&msk->conn_list, addr);
1163                 mptcp_pm_remove_anno_addr(msk, addr, remove_subflow);
1164                 if (remove_subflow)
1165                         mptcp_pm_remove_subflow(msk, &list);
1166                 release_sock(sk);
1167
1168 next:
1169                 sock_put(sk);
1170                 cond_resched();
1171         }
1172
1173         return 0;
1174 }
1175
1176 struct addr_entry_release_work {
1177         struct rcu_work rwork;
1178         struct mptcp_pm_addr_entry *entry;
1179 };
1180
1181 static void mptcp_pm_release_addr_entry(struct work_struct *work)
1182 {
1183         struct addr_entry_release_work *w;
1184         struct mptcp_pm_addr_entry *entry;
1185
1186         w = container_of(to_rcu_work(work), struct addr_entry_release_work, rwork);
1187         entry = w->entry;
1188         if (entry) {
1189                 if (entry->lsk)
1190                         sock_release(entry->lsk);
1191                 kfree(entry);
1192         }
1193         kfree(w);
1194 }
1195
1196 static void mptcp_pm_free_addr_entry(struct mptcp_pm_addr_entry *entry)
1197 {
1198         struct addr_entry_release_work *w;
1199
1200         w = kmalloc(sizeof(*w), GFP_ATOMIC);
1201         if (w) {
1202                 INIT_RCU_WORK(&w->rwork, mptcp_pm_release_addr_entry);
1203                 w->entry = entry;
1204                 queue_rcu_work(system_wq, &w->rwork);
1205         }
1206 }
1207
1208 static int mptcp_nl_remove_id_zero_address(struct net *net,
1209                                            struct mptcp_addr_info *addr)
1210 {
1211         struct mptcp_rm_list list = { .nr = 0 };
1212         long s_slot = 0, s_num = 0;
1213         struct mptcp_sock *msk;
1214
1215         list.ids[list.nr++] = 0;
1216
1217         while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1218                 struct sock *sk = (struct sock *)msk;
1219                 struct mptcp_addr_info msk_local;
1220
1221                 if (list_empty(&msk->conn_list))
1222                         goto next;
1223
1224                 local_address((struct sock_common *)msk, &msk_local);
1225                 if (!addresses_equal(&msk_local, addr, addr->port))
1226                         goto next;
1227
1228                 lock_sock(sk);
1229                 spin_lock_bh(&msk->pm.lock);
1230                 mptcp_pm_remove_addr(msk, &list);
1231                 mptcp_pm_nl_rm_subflow_received(msk, &list);
1232                 spin_unlock_bh(&msk->pm.lock);
1233                 release_sock(sk);
1234
1235 next:
1236                 sock_put(sk);
1237                 cond_resched();
1238         }
1239
1240         return 0;
1241 }
1242
1243 static int mptcp_nl_cmd_del_addr(struct sk_buff *skb, struct genl_info *info)
1244 {
1245         struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1246         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1247         struct mptcp_pm_addr_entry addr, *entry;
1248         unsigned int addr_max;
1249         int ret;
1250
1251         ret = mptcp_pm_parse_addr(attr, info, false, &addr);
1252         if (ret < 0)
1253                 return ret;
1254
1255         /* the zero id address is special: the first address used by the msk
1256          * always gets such an id, so different subflows can have different zero
1257          * id addresses. Additionally zero id is not accounted for in id_bitmap.
1258          * Let's use an 'mptcp_rm_list' instead of the common remove code.
1259          */
1260         if (addr.addr.id == 0)
1261                 return mptcp_nl_remove_id_zero_address(sock_net(skb->sk), &addr.addr);
1262
1263         spin_lock_bh(&pernet->lock);
1264         entry = __lookup_addr_by_id(pernet, addr.addr.id);
1265         if (!entry) {
1266                 GENL_SET_ERR_MSG(info, "address not found");
1267                 spin_unlock_bh(&pernet->lock);
1268                 return -EINVAL;
1269         }
1270         if (entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL) {
1271                 addr_max = pernet->add_addr_signal_max;
1272                 WRITE_ONCE(pernet->add_addr_signal_max, addr_max - 1);
1273         }
1274         if (entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW) {
1275                 addr_max = pernet->local_addr_max;
1276                 WRITE_ONCE(pernet->local_addr_max, addr_max - 1);
1277         }
1278
1279         pernet->addrs--;
1280         list_del_rcu(&entry->list);
1281         __clear_bit(entry->addr.id, pernet->id_bitmap);
1282         spin_unlock_bh(&pernet->lock);
1283
1284         mptcp_nl_remove_subflow_and_signal_addr(sock_net(skb->sk), &entry->addr);
1285         mptcp_pm_free_addr_entry(entry);
1286
1287         return ret;
1288 }
1289
1290 static void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
1291                                                struct list_head *rm_list)
1292 {
1293         struct mptcp_rm_list alist = { .nr = 0 }, slist = { .nr = 0 };
1294         struct mptcp_pm_addr_entry *entry;
1295
1296         list_for_each_entry(entry, rm_list, list) {
1297                 if (lookup_subflow_by_saddr(&msk->conn_list, &entry->addr) &&
1298                     alist.nr < MPTCP_RM_IDS_MAX &&
1299                     slist.nr < MPTCP_RM_IDS_MAX) {
1300                         alist.ids[alist.nr++] = entry->addr.id;
1301                         slist.ids[slist.nr++] = entry->addr.id;
1302                 } else if (remove_anno_list_by_saddr(msk, &entry->addr) &&
1303                          alist.nr < MPTCP_RM_IDS_MAX) {
1304                         alist.ids[alist.nr++] = entry->addr.id;
1305                 }
1306         }
1307
1308         if (alist.nr) {
1309                 spin_lock_bh(&msk->pm.lock);
1310                 mptcp_pm_remove_addr(msk, &alist);
1311                 spin_unlock_bh(&msk->pm.lock);
1312         }
1313         if (slist.nr)
1314                 mptcp_pm_remove_subflow(msk, &slist);
1315 }
1316
1317 static void mptcp_nl_remove_addrs_list(struct net *net,
1318                                        struct list_head *rm_list)
1319 {
1320         long s_slot = 0, s_num = 0;
1321         struct mptcp_sock *msk;
1322
1323         if (list_empty(rm_list))
1324                 return;
1325
1326         while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1327                 struct sock *sk = (struct sock *)msk;
1328
1329                 lock_sock(sk);
1330                 mptcp_pm_remove_addrs_and_subflows(msk, rm_list);
1331                 release_sock(sk);
1332
1333                 sock_put(sk);
1334                 cond_resched();
1335         }
1336 }
1337
1338 static void __flush_addrs(struct list_head *list)
1339 {
1340         while (!list_empty(list)) {
1341                 struct mptcp_pm_addr_entry *cur;
1342
1343                 cur = list_entry(list->next,
1344                                  struct mptcp_pm_addr_entry, list);
1345                 list_del_rcu(&cur->list);
1346                 mptcp_pm_free_addr_entry(cur);
1347         }
1348 }
1349
1350 static void __reset_counters(struct pm_nl_pernet *pernet)
1351 {
1352         WRITE_ONCE(pernet->add_addr_signal_max, 0);
1353         WRITE_ONCE(pernet->add_addr_accept_max, 0);
1354         WRITE_ONCE(pernet->local_addr_max, 0);
1355         pernet->addrs = 0;
1356 }
1357
1358 static int mptcp_nl_cmd_flush_addrs(struct sk_buff *skb, struct genl_info *info)
1359 {
1360         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1361         LIST_HEAD(free_list);
1362
1363         spin_lock_bh(&pernet->lock);
1364         list_splice_init(&pernet->local_addr_list, &free_list);
1365         __reset_counters(pernet);
1366         pernet->next_id = 1;
1367         bitmap_zero(pernet->id_bitmap, MAX_ADDR_ID + 1);
1368         spin_unlock_bh(&pernet->lock);
1369         mptcp_nl_remove_addrs_list(sock_net(skb->sk), &free_list);
1370         __flush_addrs(&free_list);
1371         return 0;
1372 }
1373
1374 static int mptcp_nl_fill_addr(struct sk_buff *skb,
1375                               struct mptcp_pm_addr_entry *entry)
1376 {
1377         struct mptcp_addr_info *addr = &entry->addr;
1378         struct nlattr *attr;
1379
1380         attr = nla_nest_start(skb, MPTCP_PM_ATTR_ADDR);
1381         if (!attr)
1382                 return -EMSGSIZE;
1383
1384         if (nla_put_u16(skb, MPTCP_PM_ADDR_ATTR_FAMILY, addr->family))
1385                 goto nla_put_failure;
1386         if (nla_put_u16(skb, MPTCP_PM_ADDR_ATTR_PORT, ntohs(addr->port)))
1387                 goto nla_put_failure;
1388         if (nla_put_u8(skb, MPTCP_PM_ADDR_ATTR_ID, addr->id))
1389                 goto nla_put_failure;
1390         if (nla_put_u32(skb, MPTCP_PM_ADDR_ATTR_FLAGS, entry->flags))
1391                 goto nla_put_failure;
1392         if (entry->ifindex &&
1393             nla_put_s32(skb, MPTCP_PM_ADDR_ATTR_IF_IDX, entry->ifindex))
1394                 goto nla_put_failure;
1395
1396         if (addr->family == AF_INET &&
1397             nla_put_in_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR4,
1398                             addr->addr.s_addr))
1399                 goto nla_put_failure;
1400 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1401         else if (addr->family == AF_INET6 &&
1402                  nla_put_in6_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR6, &addr->addr6))
1403                 goto nla_put_failure;
1404 #endif
1405         nla_nest_end(skb, attr);
1406         return 0;
1407
1408 nla_put_failure:
1409         nla_nest_cancel(skb, attr);
1410         return -EMSGSIZE;
1411 }
1412
1413 static int mptcp_nl_cmd_get_addr(struct sk_buff *skb, struct genl_info *info)
1414 {
1415         struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1416         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1417         struct mptcp_pm_addr_entry addr, *entry;
1418         struct sk_buff *msg;
1419         void *reply;
1420         int ret;
1421
1422         ret = mptcp_pm_parse_addr(attr, info, false, &addr);
1423         if (ret < 0)
1424                 return ret;
1425
1426         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1427         if (!msg)
1428                 return -ENOMEM;
1429
1430         reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
1431                                   info->genlhdr->cmd);
1432         if (!reply) {
1433                 GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
1434                 ret = -EMSGSIZE;
1435                 goto fail;
1436         }
1437
1438         spin_lock_bh(&pernet->lock);
1439         entry = __lookup_addr_by_id(pernet, addr.addr.id);
1440         if (!entry) {
1441                 GENL_SET_ERR_MSG(info, "address not found");
1442                 ret = -EINVAL;
1443                 goto unlock_fail;
1444         }
1445
1446         ret = mptcp_nl_fill_addr(msg, entry);
1447         if (ret)
1448                 goto unlock_fail;
1449
1450         genlmsg_end(msg, reply);
1451         ret = genlmsg_reply(msg, info);
1452         spin_unlock_bh(&pernet->lock);
1453         return ret;
1454
1455 unlock_fail:
1456         spin_unlock_bh(&pernet->lock);
1457
1458 fail:
1459         nlmsg_free(msg);
1460         return ret;
1461 }
1462
1463 static int mptcp_nl_cmd_dump_addrs(struct sk_buff *msg,
1464                                    struct netlink_callback *cb)
1465 {
1466         struct net *net = sock_net(msg->sk);
1467         struct mptcp_pm_addr_entry *entry;
1468         struct pm_nl_pernet *pernet;
1469         int id = cb->args[0];
1470         void *hdr;
1471         int i;
1472
1473         pernet = net_generic(net, pm_nl_pernet_id);
1474
1475         spin_lock_bh(&pernet->lock);
1476         for (i = id; i < MAX_ADDR_ID + 1; i++) {
1477                 if (test_bit(i, pernet->id_bitmap)) {
1478                         entry = __lookup_addr_by_id(pernet, i);
1479                         if (!entry)
1480                                 break;
1481
1482                         if (entry->addr.id <= id)
1483                                 continue;
1484
1485                         hdr = genlmsg_put(msg, NETLINK_CB(cb->skb).portid,
1486                                           cb->nlh->nlmsg_seq, &mptcp_genl_family,
1487                                           NLM_F_MULTI, MPTCP_PM_CMD_GET_ADDR);
1488                         if (!hdr)
1489                                 break;
1490
1491                         if (mptcp_nl_fill_addr(msg, entry) < 0) {
1492                                 genlmsg_cancel(msg, hdr);
1493                                 break;
1494                         }
1495
1496                         id = entry->addr.id;
1497                         genlmsg_end(msg, hdr);
1498                 }
1499         }
1500         spin_unlock_bh(&pernet->lock);
1501
1502         cb->args[0] = id;
1503         return msg->len;
1504 }
1505
1506 static int parse_limit(struct genl_info *info, int id, unsigned int *limit)
1507 {
1508         struct nlattr *attr = info->attrs[id];
1509
1510         if (!attr)
1511                 return 0;
1512
1513         *limit = nla_get_u32(attr);
1514         if (*limit > MPTCP_PM_ADDR_MAX) {
1515                 GENL_SET_ERR_MSG(info, "limit greater than maximum");
1516                 return -EINVAL;
1517         }
1518         return 0;
1519 }
1520
1521 static int
1522 mptcp_nl_cmd_set_limits(struct sk_buff *skb, struct genl_info *info)
1523 {
1524         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1525         unsigned int rcv_addrs, subflows;
1526         int ret;
1527
1528         spin_lock_bh(&pernet->lock);
1529         rcv_addrs = pernet->add_addr_accept_max;
1530         ret = parse_limit(info, MPTCP_PM_ATTR_RCV_ADD_ADDRS, &rcv_addrs);
1531         if (ret)
1532                 goto unlock;
1533
1534         subflows = pernet->subflows_max;
1535         ret = parse_limit(info, MPTCP_PM_ATTR_SUBFLOWS, &subflows);
1536         if (ret)
1537                 goto unlock;
1538
1539         WRITE_ONCE(pernet->add_addr_accept_max, rcv_addrs);
1540         WRITE_ONCE(pernet->subflows_max, subflows);
1541
1542 unlock:
1543         spin_unlock_bh(&pernet->lock);
1544         return ret;
1545 }
1546
1547 static int
1548 mptcp_nl_cmd_get_limits(struct sk_buff *skb, struct genl_info *info)
1549 {
1550         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1551         struct sk_buff *msg;
1552         void *reply;
1553
1554         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1555         if (!msg)
1556                 return -ENOMEM;
1557
1558         reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
1559                                   MPTCP_PM_CMD_GET_LIMITS);
1560         if (!reply)
1561                 goto fail;
1562
1563         if (nla_put_u32(msg, MPTCP_PM_ATTR_RCV_ADD_ADDRS,
1564                         READ_ONCE(pernet->add_addr_accept_max)))
1565                 goto fail;
1566
1567         if (nla_put_u32(msg, MPTCP_PM_ATTR_SUBFLOWS,
1568                         READ_ONCE(pernet->subflows_max)))
1569                 goto fail;
1570
1571         genlmsg_end(msg, reply);
1572         return genlmsg_reply(msg, info);
1573
1574 fail:
1575         GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
1576         nlmsg_free(msg);
1577         return -EMSGSIZE;
1578 }
1579
1580 static int mptcp_nl_addr_backup(struct net *net,
1581                                 struct mptcp_addr_info *addr,
1582                                 u8 bkup)
1583 {
1584         long s_slot = 0, s_num = 0;
1585         struct mptcp_sock *msk;
1586         int ret = -EINVAL;
1587
1588         while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1589                 struct sock *sk = (struct sock *)msk;
1590
1591                 if (list_empty(&msk->conn_list))
1592                         goto next;
1593
1594                 lock_sock(sk);
1595                 spin_lock_bh(&msk->pm.lock);
1596                 ret = mptcp_pm_nl_mp_prio_send_ack(msk, addr, bkup);
1597                 spin_unlock_bh(&msk->pm.lock);
1598                 release_sock(sk);
1599
1600 next:
1601                 sock_put(sk);
1602                 cond_resched();
1603         }
1604
1605         return ret;
1606 }
1607
1608 static int mptcp_nl_cmd_set_flags(struct sk_buff *skb, struct genl_info *info)
1609 {
1610         struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1611         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1612         struct mptcp_pm_addr_entry addr, *entry;
1613         struct net *net = sock_net(skb->sk);
1614         u8 bkup = 0;
1615         int ret;
1616
1617         ret = mptcp_pm_parse_addr(attr, info, true, &addr);
1618         if (ret < 0)
1619                 return ret;
1620
1621         if (addr.flags & MPTCP_PM_ADDR_FLAG_BACKUP)
1622                 bkup = 1;
1623
1624         list_for_each_entry(entry, &pernet->local_addr_list, list) {
1625                 if (addresses_equal(&entry->addr, &addr.addr, true)) {
1626                         ret = mptcp_nl_addr_backup(net, &entry->addr, bkup);
1627                         if (ret)
1628                                 return ret;
1629
1630                         if (bkup)
1631                                 entry->flags |= MPTCP_PM_ADDR_FLAG_BACKUP;
1632                         else
1633                                 entry->flags &= ~MPTCP_PM_ADDR_FLAG_BACKUP;
1634                 }
1635         }
1636
1637         return 0;
1638 }
1639
1640 static void mptcp_nl_mcast_send(struct net *net, struct sk_buff *nlskb, gfp_t gfp)
1641 {
1642         genlmsg_multicast_netns(&mptcp_genl_family, net,
1643                                 nlskb, 0, MPTCP_PM_EV_GRP_OFFSET, gfp);
1644 }
1645
1646 static int mptcp_event_add_subflow(struct sk_buff *skb, const struct sock *ssk)
1647 {
1648         const struct inet_sock *issk = inet_sk(ssk);
1649         const struct mptcp_subflow_context *sf;
1650
1651         if (nla_put_u16(skb, MPTCP_ATTR_FAMILY, ssk->sk_family))
1652                 return -EMSGSIZE;
1653
1654         switch (ssk->sk_family) {
1655         case AF_INET:
1656                 if (nla_put_in_addr(skb, MPTCP_ATTR_SADDR4, issk->inet_saddr))
1657                         return -EMSGSIZE;
1658                 if (nla_put_in_addr(skb, MPTCP_ATTR_DADDR4, issk->inet_daddr))
1659                         return -EMSGSIZE;
1660                 break;
1661 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1662         case AF_INET6: {
1663                 const struct ipv6_pinfo *np = inet6_sk(ssk);
1664
1665                 if (nla_put_in6_addr(skb, MPTCP_ATTR_SADDR6, &np->saddr))
1666                         return -EMSGSIZE;
1667                 if (nla_put_in6_addr(skb, MPTCP_ATTR_DADDR6, &ssk->sk_v6_daddr))
1668                         return -EMSGSIZE;
1669                 break;
1670         }
1671 #endif
1672         default:
1673                 WARN_ON_ONCE(1);
1674                 return -EMSGSIZE;
1675         }
1676
1677         if (nla_put_be16(skb, MPTCP_ATTR_SPORT, issk->inet_sport))
1678                 return -EMSGSIZE;
1679         if (nla_put_be16(skb, MPTCP_ATTR_DPORT, issk->inet_dport))
1680                 return -EMSGSIZE;
1681
1682         sf = mptcp_subflow_ctx(ssk);
1683         if (WARN_ON_ONCE(!sf))
1684                 return -EINVAL;
1685
1686         if (nla_put_u8(skb, MPTCP_ATTR_LOC_ID, sf->local_id))
1687                 return -EMSGSIZE;
1688
1689         if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, sf->remote_id))
1690                 return -EMSGSIZE;
1691
1692         return 0;
1693 }
1694
1695 static int mptcp_event_put_token_and_ssk(struct sk_buff *skb,
1696                                          const struct mptcp_sock *msk,
1697                                          const struct sock *ssk)
1698 {
1699         const struct sock *sk = (const struct sock *)msk;
1700         const struct mptcp_subflow_context *sf;
1701         u8 sk_err;
1702
1703         if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1704                 return -EMSGSIZE;
1705
1706         if (mptcp_event_add_subflow(skb, ssk))
1707                 return -EMSGSIZE;
1708
1709         sf = mptcp_subflow_ctx(ssk);
1710         if (WARN_ON_ONCE(!sf))
1711                 return -EINVAL;
1712
1713         if (nla_put_u8(skb, MPTCP_ATTR_BACKUP, sf->backup))
1714                 return -EMSGSIZE;
1715
1716         if (ssk->sk_bound_dev_if &&
1717             nla_put_s32(skb, MPTCP_ATTR_IF_IDX, ssk->sk_bound_dev_if))
1718                 return -EMSGSIZE;
1719
1720         sk_err = ssk->sk_err;
1721         if (sk_err && sk->sk_state == TCP_ESTABLISHED &&
1722             nla_put_u8(skb, MPTCP_ATTR_ERROR, sk_err))
1723                 return -EMSGSIZE;
1724
1725         return 0;
1726 }
1727
1728 static int mptcp_event_sub_established(struct sk_buff *skb,
1729                                        const struct mptcp_sock *msk,
1730                                        const struct sock *ssk)
1731 {
1732         return mptcp_event_put_token_and_ssk(skb, msk, ssk);
1733 }
1734
1735 static int mptcp_event_sub_closed(struct sk_buff *skb,
1736                                   const struct mptcp_sock *msk,
1737                                   const struct sock *ssk)
1738 {
1739         const struct mptcp_subflow_context *sf;
1740
1741         if (mptcp_event_put_token_and_ssk(skb, msk, ssk))
1742                 return -EMSGSIZE;
1743
1744         sf = mptcp_subflow_ctx(ssk);
1745         if (!sf->reset_seen)
1746                 return 0;
1747
1748         if (nla_put_u32(skb, MPTCP_ATTR_RESET_REASON, sf->reset_reason))
1749                 return -EMSGSIZE;
1750
1751         if (nla_put_u32(skb, MPTCP_ATTR_RESET_FLAGS, sf->reset_transient))
1752                 return -EMSGSIZE;
1753
1754         return 0;
1755 }
1756
1757 static int mptcp_event_created(struct sk_buff *skb,
1758                                const struct mptcp_sock *msk,
1759                                const struct sock *ssk)
1760 {
1761         int err = nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token);
1762
1763         if (err)
1764                 return err;
1765
1766         return mptcp_event_add_subflow(skb, ssk);
1767 }
1768
1769 void mptcp_event_addr_removed(const struct mptcp_sock *msk, uint8_t id)
1770 {
1771         struct net *net = sock_net((const struct sock *)msk);
1772         struct nlmsghdr *nlh;
1773         struct sk_buff *skb;
1774
1775         if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1776                 return;
1777
1778         skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1779         if (!skb)
1780                 return;
1781
1782         nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0, MPTCP_EVENT_REMOVED);
1783         if (!nlh)
1784                 goto nla_put_failure;
1785
1786         if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1787                 goto nla_put_failure;
1788
1789         if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, id))
1790                 goto nla_put_failure;
1791
1792         genlmsg_end(skb, nlh);
1793         mptcp_nl_mcast_send(net, skb, GFP_ATOMIC);
1794         return;
1795
1796 nla_put_failure:
1797         kfree_skb(skb);
1798 }
1799
1800 void mptcp_event_addr_announced(const struct mptcp_sock *msk,
1801                                 const struct mptcp_addr_info *info)
1802 {
1803         struct net *net = sock_net((const struct sock *)msk);
1804         struct nlmsghdr *nlh;
1805         struct sk_buff *skb;
1806
1807         if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1808                 return;
1809
1810         skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1811         if (!skb)
1812                 return;
1813
1814         nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0,
1815                           MPTCP_EVENT_ANNOUNCED);
1816         if (!nlh)
1817                 goto nla_put_failure;
1818
1819         if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1820                 goto nla_put_failure;
1821
1822         if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, info->id))
1823                 goto nla_put_failure;
1824
1825         if (nla_put_be16(skb, MPTCP_ATTR_DPORT, info->port))
1826                 goto nla_put_failure;
1827
1828         switch (info->family) {
1829         case AF_INET:
1830                 if (nla_put_in_addr(skb, MPTCP_ATTR_DADDR4, info->addr.s_addr))
1831                         goto nla_put_failure;
1832                 break;
1833 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1834         case AF_INET6:
1835                 if (nla_put_in6_addr(skb, MPTCP_ATTR_DADDR6, &info->addr6))
1836                         goto nla_put_failure;
1837                 break;
1838 #endif
1839         default:
1840                 WARN_ON_ONCE(1);
1841                 goto nla_put_failure;
1842         }
1843
1844         genlmsg_end(skb, nlh);
1845         mptcp_nl_mcast_send(net, skb, GFP_ATOMIC);
1846         return;
1847
1848 nla_put_failure:
1849         kfree_skb(skb);
1850 }
1851
1852 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
1853                  const struct sock *ssk, gfp_t gfp)
1854 {
1855         struct net *net = sock_net((const struct sock *)msk);
1856         struct nlmsghdr *nlh;
1857         struct sk_buff *skb;
1858
1859         if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1860                 return;
1861
1862         skb = nlmsg_new(NLMSG_DEFAULT_SIZE, gfp);
1863         if (!skb)
1864                 return;
1865
1866         nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0, type);
1867         if (!nlh)
1868                 goto nla_put_failure;
1869
1870         switch (type) {
1871         case MPTCP_EVENT_UNSPEC:
1872                 WARN_ON_ONCE(1);
1873                 break;
1874         case MPTCP_EVENT_CREATED:
1875         case MPTCP_EVENT_ESTABLISHED:
1876                 if (mptcp_event_created(skb, msk, ssk) < 0)
1877                         goto nla_put_failure;
1878                 break;
1879         case MPTCP_EVENT_CLOSED:
1880                 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token) < 0)
1881                         goto nla_put_failure;
1882                 break;
1883         case MPTCP_EVENT_ANNOUNCED:
1884         case MPTCP_EVENT_REMOVED:
1885                 /* call mptcp_event_addr_announced()/removed instead */
1886                 WARN_ON_ONCE(1);
1887                 break;
1888         case MPTCP_EVENT_SUB_ESTABLISHED:
1889         case MPTCP_EVENT_SUB_PRIORITY:
1890                 if (mptcp_event_sub_established(skb, msk, ssk) < 0)
1891                         goto nla_put_failure;
1892                 break;
1893         case MPTCP_EVENT_SUB_CLOSED:
1894                 if (mptcp_event_sub_closed(skb, msk, ssk) < 0)
1895                         goto nla_put_failure;
1896                 break;
1897         }
1898
1899         genlmsg_end(skb, nlh);
1900         mptcp_nl_mcast_send(net, skb, gfp);
1901         return;
1902
1903 nla_put_failure:
1904         kfree_skb(skb);
1905 }
1906
1907 static const struct genl_small_ops mptcp_pm_ops[] = {
1908         {
1909                 .cmd    = MPTCP_PM_CMD_ADD_ADDR,
1910                 .doit   = mptcp_nl_cmd_add_addr,
1911                 .flags  = GENL_ADMIN_PERM,
1912         },
1913         {
1914                 .cmd    = MPTCP_PM_CMD_DEL_ADDR,
1915                 .doit   = mptcp_nl_cmd_del_addr,
1916                 .flags  = GENL_ADMIN_PERM,
1917         },
1918         {
1919                 .cmd    = MPTCP_PM_CMD_FLUSH_ADDRS,
1920                 .doit   = mptcp_nl_cmd_flush_addrs,
1921                 .flags  = GENL_ADMIN_PERM,
1922         },
1923         {
1924                 .cmd    = MPTCP_PM_CMD_GET_ADDR,
1925                 .doit   = mptcp_nl_cmd_get_addr,
1926                 .dumpit   = mptcp_nl_cmd_dump_addrs,
1927         },
1928         {
1929                 .cmd    = MPTCP_PM_CMD_SET_LIMITS,
1930                 .doit   = mptcp_nl_cmd_set_limits,
1931                 .flags  = GENL_ADMIN_PERM,
1932         },
1933         {
1934                 .cmd    = MPTCP_PM_CMD_GET_LIMITS,
1935                 .doit   = mptcp_nl_cmd_get_limits,
1936         },
1937         {
1938                 .cmd    = MPTCP_PM_CMD_SET_FLAGS,
1939                 .doit   = mptcp_nl_cmd_set_flags,
1940                 .flags  = GENL_ADMIN_PERM,
1941         },
1942 };
1943
1944 static struct genl_family mptcp_genl_family __ro_after_init = {
1945         .name           = MPTCP_PM_NAME,
1946         .version        = MPTCP_PM_VER,
1947         .maxattr        = MPTCP_PM_ATTR_MAX,
1948         .policy         = mptcp_pm_policy,
1949         .netnsok        = true,
1950         .module         = THIS_MODULE,
1951         .small_ops      = mptcp_pm_ops,
1952         .n_small_ops    = ARRAY_SIZE(mptcp_pm_ops),
1953         .mcgrps         = mptcp_pm_mcgrps,
1954         .n_mcgrps       = ARRAY_SIZE(mptcp_pm_mcgrps),
1955 };
1956
1957 static int __net_init pm_nl_init_net(struct net *net)
1958 {
1959         struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
1960
1961         INIT_LIST_HEAD_RCU(&pernet->local_addr_list);
1962         pernet->next_id = 1;
1963         pernet->stale_loss_cnt = 4;
1964         spin_lock_init(&pernet->lock);
1965
1966         /* No need to initialize other pernet fields, the struct is zeroed at
1967          * allocation time.
1968          */
1969
1970         return 0;
1971 }
1972
1973 static void __net_exit pm_nl_exit_net(struct list_head *net_list)
1974 {
1975         struct net *net;
1976
1977         list_for_each_entry(net, net_list, exit_list) {
1978                 struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
1979
1980                 /* net is removed from namespace list, can't race with
1981                  * other modifiers
1982                  */
1983                 __flush_addrs(&pernet->local_addr_list);
1984         }
1985 }
1986
1987 static struct pernet_operations mptcp_pm_pernet_ops = {
1988         .init = pm_nl_init_net,
1989         .exit_batch = pm_nl_exit_net,
1990         .id = &pm_nl_pernet_id,
1991         .size = sizeof(struct pm_nl_pernet),
1992 };
1993
1994 void __init mptcp_pm_nl_init(void)
1995 {
1996         if (register_pernet_subsys(&mptcp_pm_pernet_ops) < 0)
1997                 panic("Failed to register MPTCP PM pernet subsystem.\n");
1998
1999         if (genl_register_family(&mptcp_genl_family))
2000                 panic("Failed to register MPTCP PM netlink family\n");
2001 }