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