net/core/sock_map.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /* Copyright (c) 2017 - 2018 Covalent IO, Inc. http://covalent.io */
   3
   4 #include <linux/bpf.h>
   5 #include <linux/btf_ids.h>
   6 #include <linux/filter.h>
   7 #include <linux/errno.h>
   8 #include <linux/file.h>
   9 #include <linux/net.h>
  10 #include <linux/workqueue.h>
  11 #include <linux/skmsg.h>
  12 #include <linux/list.h>
  13 #include <linux/jhash.h>
  14 #include <linux/sock_diag.h>
  15 #include <net/udp.h>
  16
  17 struct bpf_stab {
  18         struct bpf_map map;
  19         struct sock **sks;
  20         struct sk_psock_progs progs;
  21         raw_spinlock_t lock;
  22 };
  23
  24 #define SOCK_CREATE_FLAG_MASK                           \
  25         (BPF_F_NUMA_NODE | BPF_F_RDONLY | BPF_F_WRONLY)
  26
  27 static struct bpf_map *sock_map_alloc(union bpf_attr *attr)
  28 {
  29         struct bpf_stab *stab;
  30
  31         if (!capable(CAP_NET_ADMIN))
  32                 return ERR_PTR(-EPERM);
  33         if (attr->max_entries == 0 ||
  34             attr->key_size    != 4 ||
  35             (attr->value_size != sizeof(u32) &&
  36              attr->value_size != sizeof(u64)) ||
  37             attr->map_flags & ~SOCK_CREATE_FLAG_MASK)
  38                 return ERR_PTR(-EINVAL);
  39
  40         stab = kzalloc(sizeof(*stab), GFP_USER | __GFP_ACCOUNT);
  41         if (!stab)
  42                 return ERR_PTR(-ENOMEM);
  43
  44         bpf_map_init_from_attr(&stab->map, attr);
  45         raw_spin_lock_init(&stab->lock);
  46
  47         stab->sks = bpf_map_area_alloc(stab->map.max_entries *
  48                                        sizeof(struct sock *),
  49                                        stab->map.numa_node);
  50         if (!stab->sks) {
  51                 kfree(stab);
  52                 return ERR_PTR(-ENOMEM);
  53         }
  54
  55         return &stab->map;
  56 }
  57
  58 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog)
  59 {
  60         u32 ufd = attr->target_fd;
  61         struct bpf_map *map;
  62         struct fd f;
  63         int ret;
  64
  65         if (attr->attach_flags || attr->replace_bpf_fd)
  66                 return -EINVAL;
  67
  68         f = fdget(ufd);
  69         map = __bpf_map_get(f);
  70         if (IS_ERR(map))
  71                 return PTR_ERR(map);
  72         ret = sock_map_prog_update(map, prog, NULL, attr->attach_type);
  73         fdput(f);
  74         return ret;
  75 }
  76
  77 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype)
  78 {
  79         u32 ufd = attr->target_fd;
  80         struct bpf_prog *prog;
  81         struct bpf_map *map;
  82         struct fd f;
  83         int ret;
  84
  85         if (attr->attach_flags || attr->replace_bpf_fd)
  86                 return -EINVAL;
  87
  88         f = fdget(ufd);
  89         map = __bpf_map_get(f);
  90         if (IS_ERR(map))
  91                 return PTR_ERR(map);
  92
  93         prog = bpf_prog_get(attr->attach_bpf_fd);
  94         if (IS_ERR(prog)) {
  95                 ret = PTR_ERR(prog);
  96                 goto put_map;
  97         }
  98
  99         if (prog->type != ptype) {
 100                 ret = -EINVAL;
 101                 goto put_prog;
 102         }
 103
 104         ret = sock_map_prog_update(map, NULL, prog, attr->attach_type);
 105 put_prog:
 106         bpf_prog_put(prog);
 107 put_map:
 108         fdput(f);
 109         return ret;
 110 }
 111
 112 static void sock_map_sk_acquire(struct sock *sk)
 113         __acquires(&sk->sk_lock.slock)
 114 {
 115         lock_sock(sk);
 116         preempt_disable();
 117         rcu_read_lock();
 118 }
 119
 120 static void sock_map_sk_release(struct sock *sk)
 121         __releases(&sk->sk_lock.slock)
 122 {
 123         rcu_read_unlock();
 124         preempt_enable();
 125         release_sock(sk);
 126 }
 127
 128 static void sock_map_add_link(struct sk_psock *psock,
 129                               struct sk_psock_link *link,
 130                               struct bpf_map *map, void *link_raw)
 131 {
 132         link->link_raw = link_raw;
 133         link->map = map;
 134         spin_lock_bh(&psock->link_lock);
 135         list_add_tail(&link->list, &psock->link);
 136         spin_unlock_bh(&psock->link_lock);
 137 }
 138
 139 static void sock_map_del_link(struct sock *sk,
 140                               struct sk_psock *psock, void *link_raw)
 141 {
 142         bool strp_stop = false, verdict_stop = false;
 143         struct sk_psock_link *link, *tmp;
 144
 145         spin_lock_bh(&psock->link_lock);
 146         list_for_each_entry_safe(link, tmp, &psock->link, list) {
 147                 if (link->link_raw == link_raw) {
 148                         struct bpf_map *map = link->map;
 149                         struct bpf_stab *stab = container_of(map, struct bpf_stab,
 150                                                              map);
 151                         if (psock->parser.enabled && stab->progs.skb_parser)
 152                                 strp_stop = true;
 153                         if (psock->parser.enabled && stab->progs.skb_verdict)
 154                                 verdict_stop = true;
 155                         list_del(&link->list);
 156                         sk_psock_free_link(link);
 157                 }
 158         }
 159         spin_unlock_bh(&psock->link_lock);
 160         if (strp_stop || verdict_stop) {
 161                 write_lock_bh(&sk->sk_callback_lock);
 162                 if (strp_stop)
 163                         sk_psock_stop_strp(sk, psock);
 164                 else
 165                         sk_psock_stop_verdict(sk, psock);
 166                 write_unlock_bh(&sk->sk_callback_lock);
 167         }
 168 }
 169
 170 static void sock_map_unref(struct sock *sk, void *link_raw)
 171 {
 172         struct sk_psock *psock = sk_psock(sk);
 173
 174         if (likely(psock)) {
 175                 sock_map_del_link(sk, psock, link_raw);
 176                 sk_psock_put(sk, psock);
 177         }
 178 }
 179
 180 static int sock_map_init_proto(struct sock *sk, struct sk_psock *psock)
 181 {
 182         struct proto *prot;
 183
 184         switch (sk->sk_type) {
 185         case SOCK_STREAM:
 186                 prot = tcp_bpf_get_proto(sk, psock);
 187                 break;
 188
 189         case SOCK_DGRAM:
 190                 prot = udp_bpf_get_proto(sk, psock);
 191                 break;
 192
 193         default:
 194                 return -EINVAL;
 195         }
 196
 197         if (IS_ERR(prot))
 198                 return PTR_ERR(prot);
 199
 200         sk_psock_update_proto(sk, psock, prot);
 201         return 0;
 202 }
 203
 204 static struct sk_psock *sock_map_psock_get_checked(struct sock *sk)
 205 {
 206         struct sk_psock *psock;
 207
 208         rcu_read_lock();
 209         psock = sk_psock(sk);
 210         if (psock) {
 211                 if (sk->sk_prot->close != sock_map_close) {
 212                         psock = ERR_PTR(-EBUSY);
 213                         goto out;
 214                 }
 215
 216                 if (!refcount_inc_not_zero(&psock->refcnt))
 217                         psock = ERR_PTR(-EBUSY);
 218         }
 219 out:
 220         rcu_read_unlock();
 221         return psock;
 222 }
 223
 224 static int sock_map_link(struct bpf_map *map, struct sk_psock_progs *progs,
 225                          struct sock *sk)
 226 {
 227         struct bpf_prog *msg_parser, *skb_parser, *skb_verdict;
 228         struct sk_psock *psock;
 229         int ret;
 230
 231         skb_verdict = READ_ONCE(progs->skb_verdict);
 232         if (skb_verdict) {
 233                 skb_verdict = bpf_prog_inc_not_zero(skb_verdict);
 234                 if (IS_ERR(skb_verdict))
 235                         return PTR_ERR(skb_verdict);
 236         }
 237
 238         skb_parser = READ_ONCE(progs->skb_parser);
 239         if (skb_parser) {
 240                 skb_parser = bpf_prog_inc_not_zero(skb_parser);
 241                 if (IS_ERR(skb_parser)) {
 242                         ret = PTR_ERR(skb_parser);
 243                         goto out_put_skb_verdict;
 244                 }
 245         }
 246
 247         msg_parser = READ_ONCE(progs->msg_parser);
 248         if (msg_parser) {
 249                 msg_parser = bpf_prog_inc_not_zero(msg_parser);
 250                 if (IS_ERR(msg_parser)) {
 251                         ret = PTR_ERR(msg_parser);
 252                         goto out_put_skb_parser;
 253                 }
 254         }
 255
 256         psock = sock_map_psock_get_checked(sk);
 257         if (IS_ERR(psock)) {
 258                 ret = PTR_ERR(psock);
 259                 goto out_progs;
 260         }
 261
 262         if (psock) {
 263                 if ((msg_parser && READ_ONCE(psock->progs.msg_parser)) ||
 264                     (skb_parser  && READ_ONCE(psock->progs.skb_parser)) ||
 265                     (skb_verdict && READ_ONCE(psock->progs.skb_verdict))) {
 266                         sk_psock_put(sk, psock);
 267                         ret = -EBUSY;
 268                         goto out_progs;
 269                 }
 270         } else {
 271                 psock = sk_psock_init(sk, map->numa_node);
 272                 if (IS_ERR(psock)) {
 273                         ret = PTR_ERR(psock);
 274                         goto out_progs;
 275                 }
 276         }
 277
 278         if (msg_parser)
 279                 psock_set_prog(&psock->progs.msg_parser, msg_parser);
 280
 281         ret = sock_map_init_proto(sk, psock);
 282         if (ret < 0)
 283                 goto out_drop;
 284
 285         write_lock_bh(&sk->sk_callback_lock);
 286         if (skb_parser && skb_verdict && !psock->parser.enabled) {
 287                 ret = sk_psock_init_strp(sk, psock);
 288                 if (ret)
 289                         goto out_unlock_drop;
 290                 psock_set_prog(&psock->progs.skb_verdict, skb_verdict);
 291                 psock_set_prog(&psock->progs.skb_parser, skb_parser);
 292                 sk_psock_start_strp(sk, psock);
 293         } else if (!skb_parser && skb_verdict && !psock->parser.enabled) {
 294                 psock_set_prog(&psock->progs.skb_verdict, skb_verdict);
 295                 sk_psock_start_verdict(sk,psock);
 296         }
 297         write_unlock_bh(&sk->sk_callback_lock);
 298         return 0;
 299 out_unlock_drop:
 300         write_unlock_bh(&sk->sk_callback_lock);
 301 out_drop:
 302         sk_psock_put(sk, psock);
 303 out_progs:
 304         if (msg_parser)
 305                 bpf_prog_put(msg_parser);
 306 out_put_skb_parser:
 307         if (skb_parser)
 308                 bpf_prog_put(skb_parser);
 309 out_put_skb_verdict:
 310         if (skb_verdict)
 311                 bpf_prog_put(skb_verdict);
 312         return ret;
 313 }
 314
 315 static int sock_map_link_no_progs(struct bpf_map *map, struct sock *sk)
 316 {
 317         struct sk_psock *psock;
 318         int ret;
 319
 320         psock = sock_map_psock_get_checked(sk);
 321         if (IS_ERR(psock))
 322                 return PTR_ERR(psock);
 323
 324         if (!psock) {
 325                 psock = sk_psock_init(sk, map->numa_node);
 326                 if (IS_ERR(psock))
 327                         return PTR_ERR(psock);
 328         }
 329
 330         ret = sock_map_init_proto(sk, psock);
 331         if (ret < 0)
 332                 sk_psock_put(sk, psock);
 333         return ret;
 334 }
 335
 336 static void sock_map_free(struct bpf_map *map)
 337 {
 338         struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
 339         int i;
 340
 341         /* After the sync no updates or deletes will be in-flight so it
 342          * is safe to walk map and remove entries without risking a race
 343          * in EEXIST update case.
 344          */
 345         synchronize_rcu();
 346         for (i = 0; i < stab->map.max_entries; i++) {
 347                 struct sock **psk = &stab->sks[i];
 348                 struct sock *sk;
 349
 350                 sk = xchg(psk, NULL);
 351                 if (sk) {
 352                         lock_sock(sk);
 353                         rcu_read_lock();
 354                         sock_map_unref(sk, psk);
 355                         rcu_read_unlock();
 356                         release_sock(sk);
 357                 }
 358         }
 359
 360         /* wait for psock readers accessing its map link */
 361         synchronize_rcu();
 362
 363         bpf_map_area_free(stab->sks);
 364         kfree(stab);
 365 }
 366
 367 static void sock_map_release_progs(struct bpf_map *map)
 368 {
 369         psock_progs_drop(&container_of(map, struct bpf_stab, map)->progs);
 370 }
 371
 372 static struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key)
 373 {
 374         struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
 375
 376         WARN_ON_ONCE(!rcu_read_lock_held());
 377
 378         if (unlikely(key >= map->max_entries))
 379                 return NULL;
 380         return READ_ONCE(stab->sks[key]);
 381 }
 382
 383 static void *sock_map_lookup(struct bpf_map *map, void *key)
 384 {
 385         struct sock *sk;
 386
 387         sk = __sock_map_lookup_elem(map, *(u32 *)key);
 388         if (!sk)
 389                 return NULL;
 390         if (sk_is_refcounted(sk) && !refcount_inc_not_zero(&sk->sk_refcnt))
 391                 return NULL;
 392         return sk;
 393 }
 394
 395 static void *sock_map_lookup_sys(struct bpf_map *map, void *key)
 396 {
 397         struct sock *sk;
 398
 399         if (map->value_size != sizeof(u64))
 400                 return ERR_PTR(-ENOSPC);
 401
 402         sk = __sock_map_lookup_elem(map, *(u32 *)key);
 403         if (!sk)
 404                 return ERR_PTR(-ENOENT);
 405
 406         __sock_gen_cookie(sk);
 407         return &sk->sk_cookie;
 408 }
 409
 410 static int __sock_map_delete(struct bpf_stab *stab, struct sock *sk_test,
 411                              struct sock **psk)
 412 {
 413         struct sock *sk;
 414         int err = 0;
 415
 416         raw_spin_lock_bh(&stab->lock);
 417         sk = *psk;
 418         if (!sk_test || sk_test == sk)
 419                 sk = xchg(psk, NULL);
 420
 421         if (likely(sk))
 422                 sock_map_unref(sk, psk);
 423         else
 424                 err = -EINVAL;
 425
 426         raw_spin_unlock_bh(&stab->lock);
 427         return err;
 428 }
 429
 430 static void sock_map_delete_from_link(struct bpf_map *map, struct sock *sk,
 431                                       void *link_raw)
 432 {
 433         struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
 434
 435         __sock_map_delete(stab, sk, link_raw);
 436 }
 437
 438 static int sock_map_delete_elem(struct bpf_map *map, void *key)
 439 {
 440         struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
 441         u32 i = *(u32 *)key;
 442         struct sock **psk;
 443
 444         if (unlikely(i >= map->max_entries))
 445                 return -EINVAL;
 446
 447         psk = &stab->sks[i];
 448         return __sock_map_delete(stab, NULL, psk);
 449 }
 450
 451 static int sock_map_get_next_key(struct bpf_map *map, void *key, void *next)
 452 {
 453         struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
 454         u32 i = key ? *(u32 *)key : U32_MAX;
 455         u32 *key_next = next;
 456
 457         if (i == stab->map.max_entries - 1)
 458                 return -ENOENT;
 459         if (i >= stab->map.max_entries)
 460                 *key_next = 0;
 461         else
 462                 *key_next = i + 1;
 463         return 0;
 464 }
 465
 466 static bool sock_map_redirect_allowed(const struct sock *sk);
 467
 468 static int sock_map_update_common(struct bpf_map *map, u32 idx,
 469                                   struct sock *sk, u64 flags)
 470 {
 471         struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
 472         struct sk_psock_link *link;
 473         struct sk_psock *psock;
 474         struct sock *osk;
 475         int ret;
 476
 477         WARN_ON_ONCE(!rcu_read_lock_held());
 478         if (unlikely(flags > BPF_EXIST))
 479                 return -EINVAL;
 480         if (unlikely(idx >= map->max_entries))
 481                 return -E2BIG;
 482
 483         link = sk_psock_init_link();
 484         if (!link)
 485                 return -ENOMEM;
 486
 487         /* Only sockets we can redirect into/from in BPF need to hold
 488          * refs to parser/verdict progs and have their sk_data_ready
 489          * and sk_write_space callbacks overridden.
 490          */
 491         if (sock_map_redirect_allowed(sk))
 492                 ret = sock_map_link(map, &stab->progs, sk);
 493         else
 494                 ret = sock_map_link_no_progs(map, sk);
 495         if (ret < 0)
 496                 goto out_free;
 497
 498         psock = sk_psock(sk);
 499         WARN_ON_ONCE(!psock);
 500
 501         raw_spin_lock_bh(&stab->lock);
 502         osk = stab->sks[idx];
 503         if (osk && flags == BPF_NOEXIST) {
 504                 ret = -EEXIST;
 505                 goto out_unlock;
 506         } else if (!osk && flags == BPF_EXIST) {
 507                 ret = -ENOENT;
 508                 goto out_unlock;
 509         }
 510
 511         sock_map_add_link(psock, link, map, &stab->sks[idx]);
 512         stab->sks[idx] = sk;
 513         if (osk)
 514                 sock_map_unref(osk, &stab->sks[idx]);
 515         raw_spin_unlock_bh(&stab->lock);
 516         return 0;
 517 out_unlock:
 518         raw_spin_unlock_bh(&stab->lock);
 519         if (psock)
 520                 sk_psock_put(sk, psock);
 521 out_free:
 522         sk_psock_free_link(link);
 523         return ret;
 524 }
 525
 526 static bool sock_map_op_okay(const struct bpf_sock_ops_kern *ops)
 527 {
 528         return ops->op == BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB ||
 529                ops->op == BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB ||
 530                ops->op == BPF_SOCK_OPS_TCP_LISTEN_CB;
 531 }
 532
 533 static bool sk_is_tcp(const struct sock *sk)
 534 {
 535         return sk->sk_type == SOCK_STREAM &&
 536                sk->sk_protocol == IPPROTO_TCP;
 537 }
 538
 539 static bool sk_is_udp(const struct sock *sk)
 540 {
 541         return sk->sk_type == SOCK_DGRAM &&
 542                sk->sk_protocol == IPPROTO_UDP;
 543 }
 544
 545 static bool sock_map_redirect_allowed(const struct sock *sk)
 546 {
 547         return sk_is_tcp(sk) && sk->sk_state != TCP_LISTEN;
 548 }
 549
 550 static bool sock_map_sk_is_suitable(const struct sock *sk)
 551 {
 552         return sk_is_tcp(sk) || sk_is_udp(sk);
 553 }
 554
 555 static bool sock_map_sk_state_allowed(const struct sock *sk)
 556 {
 557         if (sk_is_tcp(sk))
 558                 return (1 << sk->sk_state) & (TCPF_ESTABLISHED | TCPF_LISTEN);
 559         else if (sk_is_udp(sk))
 560                 return sk_hashed(sk);
 561
 562         return false;
 563 }
 564
 565 static int sock_hash_update_common(struct bpf_map *map, void *key,
 566                                    struct sock *sk, u64 flags);
 567
 568 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value,
 569                              u64 flags)
 570 {
 571         struct socket *sock;
 572         struct sock *sk;
 573         int ret;
 574         u64 ufd;
 575
 576         if (map->value_size == sizeof(u64))
 577                 ufd = *(u64 *)value;
 578         else
 579                 ufd = *(u32 *)value;
 580         if (ufd > S32_MAX)
 581                 return -EINVAL;
 582
 583         sock = sockfd_lookup(ufd, &ret);
 584         if (!sock)
 585                 return ret;
 586         sk = sock->sk;
 587         if (!sk) {
 588                 ret = -EINVAL;
 589                 goto out;
 590         }
 591         if (!sock_map_sk_is_suitable(sk)) {
 592                 ret = -EOPNOTSUPP;
 593                 goto out;
 594         }
 595
 596         sock_map_sk_acquire(sk);
 597         if (!sock_map_sk_state_allowed(sk))
 598                 ret = -EOPNOTSUPP;
 599         else if (map->map_type == BPF_MAP_TYPE_SOCKMAP)
 600                 ret = sock_map_update_common(map, *(u32 *)key, sk, flags);
 601         else
 602                 ret = sock_hash_update_common(map, key, sk, flags);
 603         sock_map_sk_release(sk);
 604 out:
 605         sockfd_put(sock);
 606         return ret;
 607 }
 608
 609 static int sock_map_update_elem(struct bpf_map *map, void *key,
 610                                 void *value, u64 flags)
 611 {
 612         struct sock *sk = (struct sock *)value;
 613         int ret;
 614
 615         if (unlikely(!sk || !sk_fullsock(sk)))
 616                 return -EINVAL;
 617
 618         if (!sock_map_sk_is_suitable(sk))
 619                 return -EOPNOTSUPP;
 620
 621         local_bh_disable();
 622         bh_lock_sock(sk);
 623         if (!sock_map_sk_state_allowed(sk))
 624                 ret = -EOPNOTSUPP;
 625         else if (map->map_type == BPF_MAP_TYPE_SOCKMAP)
 626                 ret = sock_map_update_common(map, *(u32 *)key, sk, flags);
 627         else
 628                 ret = sock_hash_update_common(map, key, sk, flags);
 629         bh_unlock_sock(sk);
 630         local_bh_enable();
 631         return ret;
 632 }
 633
 634 BPF_CALL_4(bpf_sock_map_update, struct bpf_sock_ops_kern *, sops,
 635            struct bpf_map *, map, void *, key, u64, flags)
 636 {
 637         WARN_ON_ONCE(!rcu_read_lock_held());
 638
 639         if (likely(sock_map_sk_is_suitable(sops->sk) &&
 640                    sock_map_op_okay(sops)))
 641                 return sock_map_update_common(map, *(u32 *)key, sops->sk,
 642                                               flags);
 643         return -EOPNOTSUPP;
 644 }
 645
 646 const struct bpf_func_proto bpf_sock_map_update_proto = {
 647         .func           = bpf_sock_map_update,
 648         .gpl_only       = false,
 649         .pkt_access     = true,
 650         .ret_type       = RET_INTEGER,
 651         .arg1_type      = ARG_PTR_TO_CTX,
 652         .arg2_type      = ARG_CONST_MAP_PTR,
 653         .arg3_type      = ARG_PTR_TO_MAP_KEY,
 654         .arg4_type      = ARG_ANYTHING,
 655 };
 656
 657 BPF_CALL_4(bpf_sk_redirect_map, struct sk_buff *, skb,
 658            struct bpf_map *, map, u32, key, u64, flags)
 659 {
 660         struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
 661         struct sock *sk;
 662
 663         if (unlikely(flags & ~(BPF_F_INGRESS)))
 664                 return SK_DROP;
 665
 666         sk = __sock_map_lookup_elem(map, key);
 667         if (unlikely(!sk || !sock_map_redirect_allowed(sk)))
 668                 return SK_DROP;
 669
 670         tcb->bpf.flags = flags;
 671         tcb->bpf.sk_redir = sk;
 672         return SK_PASS;
 673 }
 674
 675 const struct bpf_func_proto bpf_sk_redirect_map_proto = {
 676         .func           = bpf_sk_redirect_map,
 677         .gpl_only       = false,
 678         .ret_type       = RET_INTEGER,
 679         .arg1_type      = ARG_PTR_TO_CTX,
 680         .arg2_type      = ARG_CONST_MAP_PTR,
 681         .arg3_type      = ARG_ANYTHING,
 682         .arg4_type      = ARG_ANYTHING,
 683 };
 684
 685 BPF_CALL_4(bpf_msg_redirect_map, struct sk_msg *, msg,
 686            struct bpf_map *, map, u32, key, u64, flags)
 687 {
 688         struct sock *sk;
 689
 690         if (unlikely(flags & ~(BPF_F_INGRESS)))
 691                 return SK_DROP;
 692
 693         sk = __sock_map_lookup_elem(map, key);
 694         if (unlikely(!sk || !sock_map_redirect_allowed(sk)))
 695                 return SK_DROP;
 696
 697         msg->flags = flags;
 698         msg->sk_redir = sk;
 699         return SK_PASS;
 700 }
 701
 702 const struct bpf_func_proto bpf_msg_redirect_map_proto = {
 703         .func           = bpf_msg_redirect_map,
 704         .gpl_only       = false,
 705         .ret_type       = RET_INTEGER,
 706         .arg1_type      = ARG_PTR_TO_CTX,
 707         .arg2_type      = ARG_CONST_MAP_PTR,
 708         .arg3_type      = ARG_ANYTHING,
 709         .arg4_type      = ARG_ANYTHING,
 710 };
 711
 712 struct sock_map_seq_info {
 713         struct bpf_map *map;
 714         struct sock *sk;
 715         u32 index;
 716 };
 717
 718 struct bpf_iter__sockmap {
 719         __bpf_md_ptr(struct bpf_iter_meta *, meta);
 720         __bpf_md_ptr(struct bpf_map *, map);
 721         __bpf_md_ptr(void *, key);
 722         __bpf_md_ptr(struct sock *, sk);
 723 };
 724
 725 DEFINE_BPF_ITER_FUNC(sockmap, struct bpf_iter_meta *meta,
 726                      struct bpf_map *map, void *key,
 727                      struct sock *sk)
 728
 729 static void *sock_map_seq_lookup_elem(struct sock_map_seq_info *info)
 730 {
 731         if (unlikely(info->index >= info->map->max_entries))
 732                 return NULL;
 733
 734         info->sk = __sock_map_lookup_elem(info->map, info->index);
 735
 736         /* can't return sk directly, since that might be NULL */
 737         return info;
 738 }
 739
 740 static void *sock_map_seq_start(struct seq_file *seq, loff_t *pos)
 741         __acquires(rcu)
 742 {
 743         struct sock_map_seq_info *info = seq->private;
 744
 745         if (*pos == 0)
 746                 ++*pos;
 747
 748         /* pairs with sock_map_seq_stop */
 749         rcu_read_lock();
 750         return sock_map_seq_lookup_elem(info);
 751 }
 752
 753 static void *sock_map_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 754         __must_hold(rcu)
 755 {
 756         struct sock_map_seq_info *info = seq->private;
 757
 758         ++*pos;
 759         ++info->index;
 760
 761         return sock_map_seq_lookup_elem(info);
 762 }
 763
 764 static int sock_map_seq_show(struct seq_file *seq, void *v)
 765         __must_hold(rcu)
 766 {
 767         struct sock_map_seq_info *info = seq->private;
 768         struct bpf_iter__sockmap ctx = {};
 769         struct bpf_iter_meta meta;
 770         struct bpf_prog *prog;
 771
 772         meta.seq = seq;
 773         prog = bpf_iter_get_info(&meta, !v);
 774         if (!prog)
 775                 return 0;
 776
 777         ctx.meta = &meta;
 778         ctx.map = info->map;
 779         if (v) {
 780                 ctx.key = &info->index;
 781                 ctx.sk = info->sk;
 782         }
 783
 784         return bpf_iter_run_prog(prog, &ctx);
 785 }
 786
 787 static void sock_map_seq_stop(struct seq_file *seq, void *v)
 788         __releases(rcu)
 789 {
 790         if (!v)
 791                 (void)sock_map_seq_show(seq, NULL);
 792
 793         /* pairs with sock_map_seq_start */
 794         rcu_read_unlock();
 795 }
 796
 797 static const struct seq_operations sock_map_seq_ops = {
 798         .start  = sock_map_seq_start,
 799         .next   = sock_map_seq_next,
 800         .stop   = sock_map_seq_stop,
 801         .show   = sock_map_seq_show,
 802 };
 803
 804 static int sock_map_init_seq_private(void *priv_data,
 805                                      struct bpf_iter_aux_info *aux)
 806 {
 807         struct sock_map_seq_info *info = priv_data;
 808
 809         info->map = aux->map;
 810         return 0;
 811 }
 812
 813 static const struct bpf_iter_seq_info sock_map_iter_seq_info = {
 814         .seq_ops                = &sock_map_seq_ops,
 815         .init_seq_private       = sock_map_init_seq_private,
 816         .seq_priv_size          = sizeof(struct sock_map_seq_info),
 817 };
 818
 819 static int sock_map_btf_id;
 820 const struct bpf_map_ops sock_map_ops = {
 821         .map_meta_equal         = bpf_map_meta_equal,
 822         .map_alloc              = sock_map_alloc,
 823         .map_free               = sock_map_free,
 824         .map_get_next_key       = sock_map_get_next_key,
 825         .map_lookup_elem_sys_only = sock_map_lookup_sys,
 826         .map_update_elem        = sock_map_update_elem,
 827         .map_delete_elem        = sock_map_delete_elem,
 828         .map_lookup_elem        = sock_map_lookup,
 829         .map_release_uref       = sock_map_release_progs,
 830         .map_check_btf          = map_check_no_btf,
 831         .map_btf_name           = "bpf_stab",
 832         .map_btf_id             = &sock_map_btf_id,
 833         .iter_seq_info          = &sock_map_iter_seq_info,
 834 };
 835
 836 struct bpf_shtab_elem {
 837         struct rcu_head rcu;
 838         u32 hash;
 839         struct sock *sk;
 840         struct hlist_node node;
 841         u8 key[];
 842 };
 843
 844 struct bpf_shtab_bucket {
 845         struct hlist_head head;
 846         raw_spinlock_t lock;
 847 };
 848
 849 struct bpf_shtab {
 850         struct bpf_map map;
 851         struct bpf_shtab_bucket *buckets;
 852         u32 buckets_num;
 853         u32 elem_size;
 854         struct sk_psock_progs progs;
 855         atomic_t count;
 856 };
 857
 858 static inline u32 sock_hash_bucket_hash(const void *key, u32 len)
 859 {
 860         return jhash(key, len, 0);
 861 }
 862
 863 static struct bpf_shtab_bucket *sock_hash_select_bucket(struct bpf_shtab *htab,
 864                                                         u32 hash)
 865 {
 866         return &htab->buckets[hash & (htab->buckets_num - 1)];
 867 }
 868
 869 static struct bpf_shtab_elem *
 870 sock_hash_lookup_elem_raw(struct hlist_head *head, u32 hash, void *key,
 871                           u32 key_size)
 872 {
 873         struct bpf_shtab_elem *elem;
 874
 875         hlist_for_each_entry_rcu(elem, head, node) {
 876                 if (elem->hash == hash &&
 877                     !memcmp(&elem->key, key, key_size))
 878                         return elem;
 879         }
 880
 881         return NULL;
 882 }
 883
 884 static struct sock *__sock_hash_lookup_elem(struct bpf_map *map, void *key)
 885 {
 886         struct bpf_shtab *htab = container_of(map, struct bpf_shtab, map);
 887         u32 key_size = map->key_size, hash;
 888         struct bpf_shtab_bucket *bucket;
 889         struct bpf_shtab_elem *elem;
 890
 891         WARN_ON_ONCE(!rcu_read_lock_held());
 892
 893         hash = sock_hash_bucket_hash(key, key_size);
 894         bucket = sock_hash_select_bucket(htab, hash);
 895         elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
 896
 897         return elem ? elem->sk : NULL;
 898 }
 899
 900 static void sock_hash_free_elem(struct bpf_shtab *htab,
 901                                 struct bpf_shtab_elem *elem)
 902 {
 903         atomic_dec(&htab->count);
 904         kfree_rcu(elem, rcu);
 905 }
 906
 907 static void sock_hash_delete_from_link(struct bpf_map *map, struct sock *sk,
 908                                        void *link_raw)
 909 {
 910         struct bpf_shtab *htab = container_of(map, struct bpf_shtab, map);
 911         struct bpf_shtab_elem *elem_probe, *elem = link_raw;
 912         struct bpf_shtab_bucket *bucket;
 913
 914         WARN_ON_ONCE(!rcu_read_lock_held());
 915         bucket = sock_hash_select_bucket(htab, elem->hash);
 916
 917         /* elem may be deleted in parallel from the map, but access here
 918          * is okay since it's going away only after RCU grace period.
 919          * However, we need to check whether it's still present.
 920          */
 921         raw_spin_lock_bh(&bucket->lock);
 922         elem_probe = sock_hash_lookup_elem_raw(&bucket->head, elem->hash,
 923                                                elem->key, map->key_size);
 924         if (elem_probe && elem_probe == elem) {
 925                 hlist_del_rcu(&elem->node);
 926                 sock_map_unref(elem->sk, elem);
 927                 sock_hash_free_elem(htab, elem);
 928         }
 929         raw_spin_unlock_bh(&bucket->lock);
 930 }
 931
 932 static int sock_hash_delete_elem(struct bpf_map *map, void *key)
 933 {
 934         struct bpf_shtab *htab = container_of(map, struct bpf_shtab, map);
 935         u32 hash, key_size = map->key_size;
 936         struct bpf_shtab_bucket *bucket;
 937         struct bpf_shtab_elem *elem;
 938         int ret = -ENOENT;
 939
 940         hash = sock_hash_bucket_hash(key, key_size);
 941         bucket = sock_hash_select_bucket(htab, hash);
 942
 943         raw_spin_lock_bh(&bucket->lock);
 944         elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
 945         if (elem) {
 946                 hlist_del_rcu(&elem->node);
 947                 sock_map_unref(elem->sk, elem);
 948                 sock_hash_free_elem(htab, elem);
 949                 ret = 0;
 950         }
 951         raw_spin_unlock_bh(&bucket->lock);
 952         return ret;
 953 }
 954
 955 static struct bpf_shtab_elem *sock_hash_alloc_elem(struct bpf_shtab *htab,
 956                                                    void *key, u32 key_size,
 957                                                    u32 hash, struct sock *sk,
 958                                                    struct bpf_shtab_elem *old)
 959 {
 960         struct bpf_shtab_elem *new;
 961
 962         if (atomic_inc_return(&htab->count) > htab->map.max_entries) {
 963                 if (!old) {
 964                         atomic_dec(&htab->count);
 965                         return ERR_PTR(-E2BIG);
 966                 }
 967         }
 968
 969         new = bpf_map_kmalloc_node(&htab->map, htab->elem_size,
 970                                    GFP_ATOMIC | __GFP_NOWARN,
 971                                    htab->map.numa_node);
 972         if (!new) {
 973                 atomic_dec(&htab->count);
 974                 return ERR_PTR(-ENOMEM);
 975         }
 976         memcpy(new->key, key, key_size);
 977         new->sk = sk;
 978         new->hash = hash;
 979         return new;
 980 }
 981
 982 static int sock_hash_update_common(struct bpf_map *map, void *key,
 983                                    struct sock *sk, u64 flags)
 984 {
 985         struct bpf_shtab *htab = container_of(map, struct bpf_shtab, map);
 986         u32 key_size = map->key_size, hash;
 987         struct bpf_shtab_elem *elem, *elem_new;
 988         struct bpf_shtab_bucket *bucket;
 989         struct sk_psock_link *link;
 990         struct sk_psock *psock;
 991         int ret;
 992
 993         WARN_ON_ONCE(!rcu_read_lock_held());
 994         if (unlikely(flags > BPF_EXIST))
 995                 return -EINVAL;
 996
 997         link = sk_psock_init_link();
 998         if (!link)
 999                 return -ENOMEM;
1000
1001         /* Only sockets we can redirect into/from in BPF need to hold
1002          * refs to parser/verdict progs and have their sk_data_ready
1003          * and sk_write_space callbacks overridden.
1004          */
1005         if (sock_map_redirect_allowed(sk))
1006                 ret = sock_map_link(map, &htab->progs, sk);
1007         else
1008                 ret = sock_map_link_no_progs(map, sk);
1009         if (ret < 0)
1010                 goto out_free;
1011
1012         psock = sk_psock(sk);
1013         WARN_ON_ONCE(!psock);
1014
1015         hash = sock_hash_bucket_hash(key, key_size);
1016         bucket = sock_hash_select_bucket(htab, hash);
1017
1018         raw_spin_lock_bh(&bucket->lock);
1019         elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
1020         if (elem && flags == BPF_NOEXIST) {
1021                 ret = -EEXIST;
1022                 goto out_unlock;
1023         } else if (!elem && flags == BPF_EXIST) {
1024                 ret = -ENOENT;
1025                 goto out_unlock;
1026         }
1027
1028         elem_new = sock_hash_alloc_elem(htab, key, key_size, hash, sk, elem);
1029         if (IS_ERR(elem_new)) {
1030                 ret = PTR_ERR(elem_new);
1031                 goto out_unlock;
1032         }
1033
1034         sock_map_add_link(psock, link, map, elem_new);
1035         /* Add new element to the head of the list, so that
1036          * concurrent search will find it before old elem.
1037          */
1038         hlist_add_head_rcu(&elem_new->node, &bucket->head);
1039         if (elem) {
1040                 hlist_del_rcu(&elem->node);
1041                 sock_map_unref(elem->sk, elem);
1042                 sock_hash_free_elem(htab, elem);
1043         }
1044         raw_spin_unlock_bh(&bucket->lock);
1045         return 0;
1046 out_unlock:
1047         raw_spin_unlock_bh(&bucket->lock);
1048         sk_psock_put(sk, psock);
1049 out_free:
1050         sk_psock_free_link(link);
1051         return ret;
1052 }
1053
1054 static int sock_hash_get_next_key(struct bpf_map *map, void *key,
1055                                   void *key_next)
1056 {
1057         struct bpf_shtab *htab = container_of(map, struct bpf_shtab, map);
1058         struct bpf_shtab_elem *elem, *elem_next;
1059         u32 hash, key_size = map->key_size;
1060         struct hlist_head *head;
1061         int i = 0;
1062
1063         if (!key)
1064                 goto find_first_elem;
1065         hash = sock_hash_bucket_hash(key, key_size);
1066         head = &sock_hash_select_bucket(htab, hash)->head;
1067         elem = sock_hash_lookup_elem_raw(head, hash, key, key_size);
1068         if (!elem)
1069                 goto find_first_elem;
1070
1071         elem_next = hlist_entry_safe(rcu_dereference(hlist_next_rcu(&elem->node)),
1072                                      struct bpf_shtab_elem, node);
1073         if (elem_next) {
1074                 memcpy(key_next, elem_next->key, key_size);
1075                 return 0;
1076         }
1077
1078         i = hash & (htab->buckets_num - 1);
1079         i++;
1080 find_first_elem:
1081         for (; i < htab->buckets_num; i++) {
1082                 head = &sock_hash_select_bucket(htab, i)->head;
1083                 elem_next = hlist_entry_safe(rcu_dereference(hlist_first_rcu(head)),
1084                                              struct bpf_shtab_elem, node);
1085                 if (elem_next) {
1086                         memcpy(key_next, elem_next->key, key_size);
1087                         return 0;
1088                 }
1089         }
1090
1091         return -ENOENT;
1092 }
1093
1094 static struct bpf_map *sock_hash_alloc(union bpf_attr *attr)
1095 {
1096         struct bpf_shtab *htab;
1097         int i, err;
1098
1099         if (!capable(CAP_NET_ADMIN))
1100                 return ERR_PTR(-EPERM);
1101         if (attr->max_entries == 0 ||
1102             attr->key_size    == 0 ||
1103             (attr->value_size != sizeof(u32) &&
1104              attr->value_size != sizeof(u64)) ||
1105             attr->map_flags & ~SOCK_CREATE_FLAG_MASK)
1106                 return ERR_PTR(-EINVAL);
1107         if (attr->key_size > MAX_BPF_STACK)
1108                 return ERR_PTR(-E2BIG);
1109
1110         htab = kzalloc(sizeof(*htab), GFP_USER | __GFP_ACCOUNT);
1111         if (!htab)
1112                 return ERR_PTR(-ENOMEM);
1113
1114         bpf_map_init_from_attr(&htab->map, attr);
1115
1116         htab->buckets_num = roundup_pow_of_two(htab->map.max_entries);
1117         htab->elem_size = sizeof(struct bpf_shtab_elem) +
1118                           round_up(htab->map.key_size, 8);
1119         if (htab->buckets_num == 0 ||
1120             htab->buckets_num > U32_MAX / sizeof(struct bpf_shtab_bucket)) {
1121                 err = -EINVAL;
1122                 goto free_htab;
1123         }
1124
1125         htab->buckets = bpf_map_area_alloc(htab->buckets_num *
1126                                            sizeof(struct bpf_shtab_bucket),
1127                                            htab->map.numa_node);
1128         if (!htab->buckets) {
1129                 err = -ENOMEM;
1130                 goto free_htab;
1131         }
1132
1133         for (i = 0; i < htab->buckets_num; i++) {
1134                 INIT_HLIST_HEAD(&htab->buckets[i].head);
1135                 raw_spin_lock_init(&htab->buckets[i].lock);
1136         }
1137
1138         return &htab->map;
1139 free_htab:
1140         kfree(htab);
1141         return ERR_PTR(err);
1142 }
1143
1144 static void sock_hash_free(struct bpf_map *map)
1145 {
1146         struct bpf_shtab *htab = container_of(map, struct bpf_shtab, map);
1147         struct bpf_shtab_bucket *bucket;
1148         struct hlist_head unlink_list;
1149         struct bpf_shtab_elem *elem;
1150         struct hlist_node *node;
1151         int i;
1152
1153         /* After the sync no updates or deletes will be in-flight so it
1154          * is safe to walk map and remove entries without risking a race
1155          * in EEXIST update case.
1156          */
1157         synchronize_rcu();
1158         for (i = 0; i < htab->buckets_num; i++) {
1159                 bucket = sock_hash_select_bucket(htab, i);
1160
1161                 /* We are racing with sock_hash_delete_from_link to
1162                  * enter the spin-lock critical section. Every socket on
1163                  * the list is still linked to sockhash. Since link
1164                  * exists, psock exists and holds a ref to socket. That
1165                  * lets us to grab a socket ref too.
1166                  */
1167                 raw_spin_lock_bh(&bucket->lock);
1168                 hlist_for_each_entry(elem, &bucket->head, node)
1169                         sock_hold(elem->sk);
1170                 hlist_move_list(&bucket->head, &unlink_list);
1171                 raw_spin_unlock_bh(&bucket->lock);
1172
1173                 /* Process removed entries out of atomic context to
1174                  * block for socket lock before deleting the psock's
1175                  * link to sockhash.
1176                  */
1177                 hlist_for_each_entry_safe(elem, node, &unlink_list, node) {
1178                         hlist_del(&elem->node);
1179                         lock_sock(elem->sk);
1180                         rcu_read_lock();
1181                         sock_map_unref(elem->sk, elem);
1182                         rcu_read_unlock();
1183                         release_sock(elem->sk);
1184                         sock_put(elem->sk);
1185                         sock_hash_free_elem(htab, elem);
1186                 }
1187         }
1188
1189         /* wait for psock readers accessing its map link */
1190         synchronize_rcu();
1191
1192         bpf_map_area_free(htab->buckets);
1193         kfree(htab);
1194 }
1195
1196 static void *sock_hash_lookup_sys(struct bpf_map *map, void *key)
1197 {
1198         struct sock *sk;
1199
1200         if (map->value_size != sizeof(u64))
1201                 return ERR_PTR(-ENOSPC);
1202
1203         sk = __sock_hash_lookup_elem(map, key);
1204         if (!sk)
1205                 return ERR_PTR(-ENOENT);
1206
1207         __sock_gen_cookie(sk);
1208         return &sk->sk_cookie;
1209 }
1210
1211 static void *sock_hash_lookup(struct bpf_map *map, void *key)
1212 {
1213         struct sock *sk;
1214
1215         sk = __sock_hash_lookup_elem(map, key);
1216         if (!sk)
1217                 return NULL;
1218         if (sk_is_refcounted(sk) && !refcount_inc_not_zero(&sk->sk_refcnt))
1219                 return NULL;
1220         return sk;
1221 }
1222
1223 static void sock_hash_release_progs(struct bpf_map *map)
1224 {
1225         psock_progs_drop(&container_of(map, struct bpf_shtab, map)->progs);
1226 }
1227
1228 BPF_CALL_4(bpf_sock_hash_update, struct bpf_sock_ops_kern *, sops,
1229            struct bpf_map *, map, void *, key, u64, flags)
1230 {
1231         WARN_ON_ONCE(!rcu_read_lock_held());
1232
1233         if (likely(sock_map_sk_is_suitable(sops->sk) &&
1234                    sock_map_op_okay(sops)))
1235                 return sock_hash_update_common(map, key, sops->sk, flags);
1236         return -EOPNOTSUPP;
1237 }
1238
1239 const struct bpf_func_proto bpf_sock_hash_update_proto = {
1240         .func           = bpf_sock_hash_update,
1241         .gpl_only       = false,
1242         .pkt_access     = true,
1243         .ret_type       = RET_INTEGER,
1244         .arg1_type      = ARG_PTR_TO_CTX,
1245         .arg2_type      = ARG_CONST_MAP_PTR,
1246         .arg3_type      = ARG_PTR_TO_MAP_KEY,
1247         .arg4_type      = ARG_ANYTHING,
1248 };
1249
1250 BPF_CALL_4(bpf_sk_redirect_hash, struct sk_buff *, skb,
1251            struct bpf_map *, map, void *, key, u64, flags)
1252 {
1253         struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
1254         struct sock *sk;
1255
1256         if (unlikely(flags & ~(BPF_F_INGRESS)))
1257                 return SK_DROP;
1258
1259         sk = __sock_hash_lookup_elem(map, key);
1260         if (unlikely(!sk || !sock_map_redirect_allowed(sk)))
1261                 return SK_DROP;
1262
1263         tcb->bpf.flags = flags;
1264         tcb->bpf.sk_redir = sk;
1265         return SK_PASS;
1266 }
1267
1268 const struct bpf_func_proto bpf_sk_redirect_hash_proto = {
1269         .func           = bpf_sk_redirect_hash,
1270         .gpl_only       = false,
1271         .ret_type       = RET_INTEGER,
1272         .arg1_type      = ARG_PTR_TO_CTX,
1273         .arg2_type      = ARG_CONST_MAP_PTR,
1274         .arg3_type      = ARG_PTR_TO_MAP_KEY,
1275         .arg4_type      = ARG_ANYTHING,
1276 };
1277
1278 BPF_CALL_4(bpf_msg_redirect_hash, struct sk_msg *, msg,
1279            struct bpf_map *, map, void *, key, u64, flags)
1280 {
1281         struct sock *sk;
1282
1283         if (unlikely(flags & ~(BPF_F_INGRESS)))
1284                 return SK_DROP;
1285
1286         sk = __sock_hash_lookup_elem(map, key);
1287         if (unlikely(!sk || !sock_map_redirect_allowed(sk)))
1288                 return SK_DROP;
1289
1290         msg->flags = flags;
1291         msg->sk_redir = sk;
1292         return SK_PASS;
1293 }
1294
1295 const struct bpf_func_proto bpf_msg_redirect_hash_proto = {
1296         .func           = bpf_msg_redirect_hash,
1297         .gpl_only       = false,
1298         .ret_type       = RET_INTEGER,
1299         .arg1_type      = ARG_PTR_TO_CTX,
1300         .arg2_type      = ARG_CONST_MAP_PTR,
1301         .arg3_type      = ARG_PTR_TO_MAP_KEY,
1302         .arg4_type      = ARG_ANYTHING,
1303 };
1304
1305 struct sock_hash_seq_info {
1306         struct bpf_map *map;
1307         struct bpf_shtab *htab;
1308         u32 bucket_id;
1309 };
1310
1311 static void *sock_hash_seq_find_next(struct sock_hash_seq_info *info,
1312                                      struct bpf_shtab_elem *prev_elem)
1313 {
1314         const struct bpf_shtab *htab = info->htab;
1315         struct bpf_shtab_bucket *bucket;
1316         struct bpf_shtab_elem *elem;
1317         struct hlist_node *node;
1318
1319         /* try to find next elem in the same bucket */
1320         if (prev_elem) {
1321                 node = rcu_dereference(hlist_next_rcu(&prev_elem->node));
1322                 elem = hlist_entry_safe(node, struct bpf_shtab_elem, node);
1323                 if (elem)
1324                         return elem;
1325
1326                 /* no more elements, continue in the next bucket */
1327                 info->bucket_id++;
1328         }
1329
1330         for (; info->bucket_id < htab->buckets_num; info->bucket_id++) {
1331                 bucket = &htab->buckets[info->bucket_id];
1332                 node = rcu_dereference(hlist_first_rcu(&bucket->head));
1333                 elem = hlist_entry_safe(node, struct bpf_shtab_elem, node);
1334                 if (elem)
1335                         return elem;
1336         }
1337
1338         return NULL;
1339 }
1340
1341 static void *sock_hash_seq_start(struct seq_file *seq, loff_t *pos)
1342         __acquires(rcu)
1343 {
1344         struct sock_hash_seq_info *info = seq->private;
1345
1346         if (*pos == 0)
1347                 ++*pos;
1348
1349         /* pairs with sock_hash_seq_stop */
1350         rcu_read_lock();
1351         return sock_hash_seq_find_next(info, NULL);
1352 }
1353
1354 static void *sock_hash_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1355         __must_hold(rcu)
1356 {
1357         struct sock_hash_seq_info *info = seq->private;
1358
1359         ++*pos;
1360         return sock_hash_seq_find_next(info, v);
1361 }
1362
1363 static int sock_hash_seq_show(struct seq_file *seq, void *v)
1364         __must_hold(rcu)
1365 {
1366         struct sock_hash_seq_info *info = seq->private;
1367         struct bpf_iter__sockmap ctx = {};
1368         struct bpf_shtab_elem *elem = v;
1369         struct bpf_iter_meta meta;
1370         struct bpf_prog *prog;
1371
1372         meta.seq = seq;
1373         prog = bpf_iter_get_info(&meta, !elem);
1374         if (!prog)
1375                 return 0;
1376
1377         ctx.meta = &meta;
1378         ctx.map = info->map;
1379         if (elem) {
1380                 ctx.key = elem->key;
1381                 ctx.sk = elem->sk;
1382         }
1383
1384         return bpf_iter_run_prog(prog, &ctx);
1385 }
1386
1387 static void sock_hash_seq_stop(struct seq_file *seq, void *v)
1388         __releases(rcu)
1389 {
1390         if (!v)
1391                 (void)sock_hash_seq_show(seq, NULL);
1392
1393         /* pairs with sock_hash_seq_start */
1394         rcu_read_unlock();
1395 }
1396
1397 static const struct seq_operations sock_hash_seq_ops = {
1398         .start  = sock_hash_seq_start,
1399         .next   = sock_hash_seq_next,
1400         .stop   = sock_hash_seq_stop,
1401         .show   = sock_hash_seq_show,
1402 };
1403
1404 static int sock_hash_init_seq_private(void *priv_data,
1405                                      struct bpf_iter_aux_info *aux)
1406 {
1407         struct sock_hash_seq_info *info = priv_data;
1408
1409         info->map = aux->map;
1410         info->htab = container_of(aux->map, struct bpf_shtab, map);
1411         return 0;
1412 }
1413
1414 static const struct bpf_iter_seq_info sock_hash_iter_seq_info = {
1415         .seq_ops                = &sock_hash_seq_ops,
1416         .init_seq_private       = sock_hash_init_seq_private,
1417         .seq_priv_size          = sizeof(struct sock_hash_seq_info),
1418 };
1419
1420 static int sock_hash_map_btf_id;
1421 const struct bpf_map_ops sock_hash_ops = {
1422         .map_meta_equal         = bpf_map_meta_equal,
1423         .map_alloc              = sock_hash_alloc,
1424         .map_free               = sock_hash_free,
1425         .map_get_next_key       = sock_hash_get_next_key,
1426         .map_update_elem        = sock_map_update_elem,
1427         .map_delete_elem        = sock_hash_delete_elem,
1428         .map_lookup_elem        = sock_hash_lookup,
1429         .map_lookup_elem_sys_only = sock_hash_lookup_sys,
1430         .map_release_uref       = sock_hash_release_progs,
1431         .map_check_btf          = map_check_no_btf,
1432         .map_btf_name           = "bpf_shtab",
1433         .map_btf_id             = &sock_hash_map_btf_id,
1434         .iter_seq_info          = &sock_hash_iter_seq_info,
1435 };
1436
1437 static struct sk_psock_progs *sock_map_progs(struct bpf_map *map)
1438 {
1439         switch (map->map_type) {
1440         case BPF_MAP_TYPE_SOCKMAP:
1441                 return &container_of(map, struct bpf_stab, map)->progs;
1442         case BPF_MAP_TYPE_SOCKHASH:
1443                 return &container_of(map, struct bpf_shtab, map)->progs;
1444         default:
1445                 break;
1446         }
1447
1448         return NULL;
1449 }
1450
1451 int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog,
1452                          struct bpf_prog *old, u32 which)
1453 {
1454         struct sk_psock_progs *progs = sock_map_progs(map);
1455         struct bpf_prog **pprog;
1456
1457         if (!progs)
1458                 return -EOPNOTSUPP;
1459
1460         switch (which) {
1461         case BPF_SK_MSG_VERDICT:
1462                 pprog = &progs->msg_parser;
1463                 break;
1464         case BPF_SK_SKB_STREAM_PARSER:
1465                 pprog = &progs->skb_parser;
1466                 break;
1467         case BPF_SK_SKB_STREAM_VERDICT:
1468                 pprog = &progs->skb_verdict;
1469                 break;
1470         default:
1471                 return -EOPNOTSUPP;
1472         }
1473
1474         if (old)
1475                 return psock_replace_prog(pprog, prog, old);
1476
1477         psock_set_prog(pprog, prog);
1478         return 0;
1479 }
1480
1481 static void sock_map_unlink(struct sock *sk, struct sk_psock_link *link)
1482 {
1483         switch (link->map->map_type) {
1484         case BPF_MAP_TYPE_SOCKMAP:
1485                 return sock_map_delete_from_link(link->map, sk,
1486                                                  link->link_raw);
1487         case BPF_MAP_TYPE_SOCKHASH:
1488                 return sock_hash_delete_from_link(link->map, sk,
1489                                                   link->link_raw);
1490         default:
1491                 break;
1492         }
1493 }
1494
1495 static void sock_map_remove_links(struct sock *sk, struct sk_psock *psock)
1496 {
1497         struct sk_psock_link *link;
1498
1499         while ((link = sk_psock_link_pop(psock))) {
1500                 sock_map_unlink(sk, link);
1501                 sk_psock_free_link(link);
1502         }
1503 }
1504
1505 void sock_map_unhash(struct sock *sk)
1506 {
1507         void (*saved_unhash)(struct sock *sk);
1508         struct sk_psock *psock;
1509
1510         rcu_read_lock();
1511         psock = sk_psock(sk);
1512         if (unlikely(!psock)) {
1513                 rcu_read_unlock();
1514                 if (sk->sk_prot->unhash)
1515                         sk->sk_prot->unhash(sk);
1516                 return;
1517         }
1518
1519         saved_unhash = psock->saved_unhash;
1520         sock_map_remove_links(sk, psock);
1521         rcu_read_unlock();
1522         saved_unhash(sk);
1523 }
1524
1525 void sock_map_close(struct sock *sk, long timeout)
1526 {
1527         void (*saved_close)(struct sock *sk, long timeout);
1528         struct sk_psock *psock;
1529
1530         lock_sock(sk);
1531         rcu_read_lock();
1532         psock = sk_psock(sk);
1533         if (unlikely(!psock)) {
1534                 rcu_read_unlock();
1535                 release_sock(sk);
1536                 return sk->sk_prot->close(sk, timeout);
1537         }
1538
1539         saved_close = psock->saved_close;
1540         sock_map_remove_links(sk, psock);
1541         rcu_read_unlock();
1542         release_sock(sk);
1543         saved_close(sk, timeout);
1544 }
1545
1546 static int sock_map_iter_attach_target(struct bpf_prog *prog,
1547                                        union bpf_iter_link_info *linfo,
1548                                        struct bpf_iter_aux_info *aux)
1549 {
1550         struct bpf_map *map;
1551         int err = -EINVAL;
1552
1553         if (!linfo->map.map_fd)
1554                 return -EBADF;
1555
1556         map = bpf_map_get_with_uref(linfo->map.map_fd);
1557         if (IS_ERR(map))
1558                 return PTR_ERR(map);
1559
1560         if (map->map_type != BPF_MAP_TYPE_SOCKMAP &&
1561             map->map_type != BPF_MAP_TYPE_SOCKHASH)
1562                 goto put_map;
1563
1564         if (prog->aux->max_rdonly_access > map->key_size) {
1565                 err = -EACCES;
1566                 goto put_map;
1567         }
1568
1569         aux->map = map;
1570         return 0;
1571
1572 put_map:
1573         bpf_map_put_with_uref(map);
1574         return err;
1575 }
1576
1577 static void sock_map_iter_detach_target(struct bpf_iter_aux_info *aux)
1578 {
1579         bpf_map_put_with_uref(aux->map);
1580 }
1581
1582 static struct bpf_iter_reg sock_map_iter_reg = {
1583         .target                 = "sockmap",
1584         .attach_target          = sock_map_iter_attach_target,
1585         .detach_target          = sock_map_iter_detach_target,
1586         .show_fdinfo            = bpf_iter_map_show_fdinfo,
1587         .fill_link_info         = bpf_iter_map_fill_link_info,
1588         .ctx_arg_info_size      = 2,
1589         .ctx_arg_info           = {
1590                 { offsetof(struct bpf_iter__sockmap, key),
1591                   PTR_TO_RDONLY_BUF_OR_NULL },
1592                 { offsetof(struct bpf_iter__sockmap, sk),
1593                   PTR_TO_BTF_ID_OR_NULL },
1594         },
1595 };
1596
1597 static int __init bpf_sockmap_iter_init(void)
1598 {
1599         sock_map_iter_reg.ctx_arg_info[1].btf_id =
1600                 btf_sock_ids[BTF_SOCK_TYPE_SOCK];
1601         return bpf_iter_reg_target(&sock_map_iter_reg);
1602 }
1603 late_initcall(bpf_sockmap_iter_init);