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 spinlock_t lock; 22 }; 23 24 #define SOCK_CREATE_FLAG_MASK \ 25 (BPF_F_NUMA_NODE | BPF_F_RDONLY | BPF_F_WRONLY) 26 27 /* This mutex is used to 28 * - protect race between prog/link attach/detach and link prog update, and 29 * - protect race between releasing and accessing map in bpf_link. 30 * A single global mutex lock is used since it is expected contention is low. 31 */ 32 static DEFINE_MUTEX(sockmap_mutex); 33 34 static int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog, 35 struct bpf_prog *old, struct bpf_link *link, 36 u32 which); 37 static struct sk_psock_progs *sock_map_progs(struct bpf_map *map); 38 39 static struct bpf_map *sock_map_alloc(union bpf_attr *attr) 40 { 41 struct bpf_stab *stab; 42 43 if (attr->max_entries == 0 || 44 attr->key_size != 4 || 45 (attr->value_size != sizeof(u32) && 46 attr->value_size != sizeof(u64)) || 47 attr->map_flags & ~SOCK_CREATE_FLAG_MASK) 48 return ERR_PTR(-EINVAL); 49 50 stab = bpf_map_area_alloc(sizeof(*stab), NUMA_NO_NODE); 51 if (!stab) 52 return ERR_PTR(-ENOMEM); 53 54 bpf_map_init_from_attr(&stab->map, attr); 55 spin_lock_init(&stab->lock); 56 57 stab->sks = bpf_map_area_alloc((u64) stab->map.max_entries * 58 sizeof(struct sock *), 59 stab->map.numa_node); 60 if (!stab->sks) { 61 bpf_map_area_free(stab); 62 return ERR_PTR(-ENOMEM); 63 } 64 65 return &stab->map; 66 } 67 68 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog) 69 { 70 struct bpf_map *map; 71 int ret; 72 73 if (attr->attach_flags || attr->replace_bpf_fd) 74 return -EINVAL; 75 76 CLASS(fd, f)(attr->target_fd); 77 map = __bpf_map_get(f); 78 if (IS_ERR(map)) 79 return PTR_ERR(map); 80 mutex_lock(&sockmap_mutex); 81 ret = sock_map_prog_update(map, prog, NULL, NULL, attr->attach_type); 82 mutex_unlock(&sockmap_mutex); 83 return ret; 84 } 85 86 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype) 87 { 88 struct bpf_prog *prog; 89 struct bpf_map *map; 90 int ret; 91 92 if (attr->attach_flags || attr->replace_bpf_fd) 93 return -EINVAL; 94 95 CLASS(fd, f)(attr->target_fd); 96 map = __bpf_map_get(f); 97 if (IS_ERR(map)) 98 return PTR_ERR(map); 99 100 prog = bpf_prog_get(attr->attach_bpf_fd); 101 if (IS_ERR(prog)) 102 return PTR_ERR(prog); 103 104 if (prog->type != ptype) { 105 ret = -EINVAL; 106 goto put_prog; 107 } 108 109 mutex_lock(&sockmap_mutex); 110 ret = sock_map_prog_update(map, NULL, prog, NULL, attr->attach_type); 111 mutex_unlock(&sockmap_mutex); 112 put_prog: 113 bpf_prog_put(prog); 114 return ret; 115 } 116 117 static void sock_map_sk_acquire(struct sock *sk) 118 __acquires(&sk->sk_lock.slock) 119 { 120 lock_sock(sk); 121 rcu_read_lock(); 122 } 123 124 static void sock_map_sk_release(struct sock *sk) 125 __releases(&sk->sk_lock.slock) 126 { 127 rcu_read_unlock(); 128 release_sock(sk); 129 } 130 131 static void sock_map_add_link(struct sk_psock *psock, 132 struct sk_psock_link *link, 133 struct bpf_map *map, void *link_raw) 134 { 135 link->link_raw = link_raw; 136 link->map = map; 137 spin_lock_bh(&psock->link_lock); 138 list_add_tail(&link->list, &psock->link); 139 spin_unlock_bh(&psock->link_lock); 140 } 141 142 static void sock_map_del_link(struct sock *sk, 143 struct sk_psock *psock, void *link_raw) 144 { 145 bool strp_stop = false, verdict_stop = false; 146 struct sk_psock_link *link, *tmp; 147 148 spin_lock_bh(&psock->link_lock); 149 list_for_each_entry_safe(link, tmp, &psock->link, list) { 150 if (link->link_raw == link_raw) { 151 struct bpf_map *map = link->map; 152 struct sk_psock_progs *progs = sock_map_progs(map); 153 154 if (psock->saved_data_ready && progs->stream_parser) 155 strp_stop = true; 156 if (psock->saved_data_ready && progs->stream_verdict) 157 verdict_stop = true; 158 if (psock->saved_data_ready && progs->skb_verdict) 159 verdict_stop = true; 160 list_del(&link->list); 161 sk_psock_free_link(link); 162 } 163 } 164 spin_unlock_bh(&psock->link_lock); 165 if (strp_stop || verdict_stop) { 166 write_lock_bh(&sk->sk_callback_lock); 167 if (strp_stop) 168 sk_psock_stop_strp(sk, psock); 169 if (verdict_stop) 170 sk_psock_stop_verdict(sk, psock); 171 172 if (psock->psock_update_sk_prot) 173 psock->psock_update_sk_prot(sk, psock, false); 174 write_unlock_bh(&sk->sk_callback_lock); 175 } 176 } 177 178 static void sock_map_unref(struct sock *sk, void *link_raw) 179 { 180 struct sk_psock *psock = sk_psock(sk); 181 182 if (likely(psock)) { 183 sock_map_del_link(sk, psock, link_raw); 184 sk_psock_put(sk, psock); 185 } 186 } 187 188 static int sock_map_init_proto(struct sock *sk, struct sk_psock *psock) 189 { 190 if (!sk->sk_prot->psock_update_sk_prot) 191 return -EINVAL; 192 psock->psock_update_sk_prot = sk->sk_prot->psock_update_sk_prot; 193 return sk->sk_prot->psock_update_sk_prot(sk, psock, false); 194 } 195 196 static struct sk_psock *sock_map_psock_get_checked(struct sock *sk) 197 { 198 struct sk_psock *psock; 199 200 rcu_read_lock(); 201 psock = sk_psock(sk); 202 if (psock) { 203 if (sk->sk_prot->close != sock_map_close) { 204 psock = ERR_PTR(-EBUSY); 205 goto out; 206 } 207 208 if (!refcount_inc_not_zero(&psock->refcnt)) 209 psock = ERR_PTR(-EBUSY); 210 } 211 out: 212 rcu_read_unlock(); 213 return psock; 214 } 215 216 static int sock_map_link(struct bpf_map *map, struct sock *sk) 217 { 218 struct sk_psock_progs *progs = sock_map_progs(map); 219 struct bpf_prog *stream_verdict = NULL; 220 struct bpf_prog *stream_parser = NULL; 221 struct bpf_prog *skb_verdict = NULL; 222 struct bpf_prog *msg_parser = NULL; 223 struct sk_psock *psock; 224 int ret; 225 226 stream_verdict = READ_ONCE(progs->stream_verdict); 227 if (stream_verdict) { 228 stream_verdict = bpf_prog_inc_not_zero(stream_verdict); 229 if (IS_ERR(stream_verdict)) 230 return PTR_ERR(stream_verdict); 231 } 232 233 stream_parser = READ_ONCE(progs->stream_parser); 234 if (stream_parser) { 235 stream_parser = bpf_prog_inc_not_zero(stream_parser); 236 if (IS_ERR(stream_parser)) { 237 ret = PTR_ERR(stream_parser); 238 goto out_put_stream_verdict; 239 } 240 } 241 242 msg_parser = READ_ONCE(progs->msg_parser); 243 if (msg_parser) { 244 msg_parser = bpf_prog_inc_not_zero(msg_parser); 245 if (IS_ERR(msg_parser)) { 246 ret = PTR_ERR(msg_parser); 247 goto out_put_stream_parser; 248 } 249 } 250 251 skb_verdict = READ_ONCE(progs->skb_verdict); 252 if (skb_verdict) { 253 skb_verdict = bpf_prog_inc_not_zero(skb_verdict); 254 if (IS_ERR(skb_verdict)) { 255 ret = PTR_ERR(skb_verdict); 256 goto out_put_msg_parser; 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 (stream_parser && READ_ONCE(psock->progs.stream_parser)) || 269 (skb_verdict && READ_ONCE(psock->progs.skb_verdict)) || 270 (skb_verdict && READ_ONCE(psock->progs.stream_verdict)) || 271 (stream_verdict && READ_ONCE(psock->progs.skb_verdict)) || 272 (stream_verdict && READ_ONCE(psock->progs.stream_verdict))) { 273 sk_psock_put(sk, psock); 274 ret = -EBUSY; 275 goto out_progs; 276 } 277 } else { 278 psock = sk_psock_init(sk, map->numa_node); 279 if (IS_ERR(psock)) { 280 ret = PTR_ERR(psock); 281 goto out_progs; 282 } 283 } 284 285 if (msg_parser) 286 psock_set_prog(&psock->progs.msg_parser, msg_parser); 287 if (stream_parser) 288 psock_set_prog(&psock->progs.stream_parser, stream_parser); 289 if (stream_verdict) 290 psock_set_prog(&psock->progs.stream_verdict, stream_verdict); 291 if (skb_verdict) 292 psock_set_prog(&psock->progs.skb_verdict, skb_verdict); 293 294 /* msg_* and stream_* programs references tracked in psock after this 295 * point. Reference dec and cleanup will occur through psock destructor 296 */ 297 ret = sock_map_init_proto(sk, psock); 298 if (ret < 0) { 299 sk_psock_put(sk, psock); 300 goto out; 301 } 302 303 write_lock_bh(&sk->sk_callback_lock); 304 if (stream_parser && stream_verdict && !psock->saved_data_ready) { 305 ret = sk_psock_init_strp(sk, psock); 306 if (ret) { 307 write_unlock_bh(&sk->sk_callback_lock); 308 sk_psock_put(sk, psock); 309 goto out; 310 } 311 sk_psock_start_strp(sk, psock); 312 } else if (!stream_parser && stream_verdict && !psock->saved_data_ready) { 313 sk_psock_start_verdict(sk,psock); 314 } else if (!stream_verdict && skb_verdict && !psock->saved_data_ready) { 315 sk_psock_start_verdict(sk, psock); 316 } 317 write_unlock_bh(&sk->sk_callback_lock); 318 return 0; 319 out_progs: 320 if (skb_verdict) 321 bpf_prog_put(skb_verdict); 322 out_put_msg_parser: 323 if (msg_parser) 324 bpf_prog_put(msg_parser); 325 out_put_stream_parser: 326 if (stream_parser) 327 bpf_prog_put(stream_parser); 328 out_put_stream_verdict: 329 if (stream_verdict) 330 bpf_prog_put(stream_verdict); 331 out: 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 sock_hold(sk); 352 lock_sock(sk); 353 rcu_read_lock(); 354 sock_map_unref(sk, psk); 355 rcu_read_unlock(); 356 release_sock(sk); 357 sock_put(sk); 358 } 359 } 360 361 /* wait for psock readers accessing its map link */ 362 synchronize_rcu(); 363 364 bpf_map_area_free(stab->sks); 365 bpf_map_area_free(stab); 366 } 367 368 static void sock_map_release_progs(struct bpf_map *map) 369 { 370 psock_progs_drop(&container_of(map, struct bpf_stab, map)->progs); 371 } 372 373 static struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key) 374 { 375 struct bpf_stab *stab = container_of(map, struct bpf_stab, map); 376 377 WARN_ON_ONCE(!rcu_read_lock_held()); 378 379 if (unlikely(key >= map->max_entries)) 380 return NULL; 381 return READ_ONCE(stab->sks[key]); 382 } 383 384 static void *sock_map_lookup(struct bpf_map *map, void *key) 385 { 386 struct sock *sk; 387 388 sk = __sock_map_lookup_elem(map, *(u32 *)key); 389 if (!sk) 390 return NULL; 391 if (sk_is_refcounted(sk) && !refcount_inc_not_zero(&sk->sk_refcnt)) 392 return NULL; 393 return sk; 394 } 395 396 static void *sock_map_lookup_sys(struct bpf_map *map, void *key) 397 { 398 struct sock *sk; 399 400 if (map->value_size != sizeof(u64)) 401 return ERR_PTR(-ENOSPC); 402 403 sk = __sock_map_lookup_elem(map, *(u32 *)key); 404 if (!sk) 405 return ERR_PTR(-ENOENT); 406 407 __sock_gen_cookie(sk); 408 return &sk->sk_cookie; 409 } 410 411 static int __sock_map_delete(struct bpf_stab *stab, struct sock *sk_test, 412 struct sock **psk) 413 { 414 struct sock *sk; 415 int err = 0; 416 417 spin_lock_bh(&stab->lock); 418 sk = *psk; 419 if (!sk_test || sk_test == sk) 420 sk = xchg(psk, NULL); 421 422 if (likely(sk)) 423 sock_map_unref(sk, psk); 424 else 425 err = -EINVAL; 426 427 spin_unlock_bh(&stab->lock); 428 return err; 429 } 430 431 static void sock_map_delete_from_link(struct bpf_map *map, struct sock *sk, 432 void *link_raw) 433 { 434 struct bpf_stab *stab = container_of(map, struct bpf_stab, map); 435 436 __sock_map_delete(stab, sk, link_raw); 437 } 438 439 static long sock_map_delete_elem(struct bpf_map *map, void *key) 440 { 441 struct bpf_stab *stab = container_of(map, struct bpf_stab, map); 442 u32 i = *(u32 *)key; 443 struct sock **psk; 444 445 if (unlikely(i >= map->max_entries)) 446 return -EINVAL; 447 448 psk = &stab->sks[i]; 449 return __sock_map_delete(stab, NULL, psk); 450 } 451 452 static int sock_map_get_next_key(struct bpf_map *map, void *key, void *next) 453 { 454 struct bpf_stab *stab = container_of(map, struct bpf_stab, map); 455 u32 i = key ? *(u32 *)key : U32_MAX; 456 u32 *key_next = next; 457 458 if (i == stab->map.max_entries - 1) 459 return -ENOENT; 460 if (i >= stab->map.max_entries) 461 *key_next = 0; 462 else 463 *key_next = i + 1; 464 return 0; 465 } 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; 481 482 link = sk_psock_init_link(); 483 if (!link) 484 return -ENOMEM; 485 486 ret = sock_map_link(map, sk); 487 if (ret < 0) 488 goto out_free; 489 490 psock = sk_psock(sk); 491 WARN_ON_ONCE(!psock); 492 493 spin_lock_bh(&stab->lock); 494 osk = stab->sks[idx]; 495 if (osk && flags == BPF_NOEXIST) { 496 ret = -EEXIST; 497 goto out_unlock; 498 } else if (!osk && flags == BPF_EXIST) { 499 ret = -ENOENT; 500 goto out_unlock; 501 } 502 503 sock_map_add_link(psock, link, map, &stab->sks[idx]); 504 stab->sks[idx] = sk; 505 if (osk) 506 sock_map_unref(osk, &stab->sks[idx]); 507 spin_unlock_bh(&stab->lock); 508 return 0; 509 out_unlock: 510 spin_unlock_bh(&stab->lock); 511 if (psock) 512 sk_psock_put(sk, psock); 513 out_free: 514 sk_psock_free_link(link); 515 return ret; 516 } 517 518 static bool sock_map_op_okay(const struct bpf_sock_ops_kern *ops) 519 { 520 return ops->op == BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB || 521 ops->op == BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB || 522 ops->op == BPF_SOCK_OPS_TCP_LISTEN_CB; 523 } 524 525 static bool sock_map_redirect_allowed(const struct sock *sk) 526 { 527 if (sk_is_tcp(sk)) 528 return sk->sk_state != TCP_LISTEN; 529 else 530 return sk->sk_state == TCP_ESTABLISHED; 531 } 532 533 static bool sock_map_sk_is_suitable(const struct sock *sk) 534 { 535 return !!sk->sk_prot->psock_update_sk_prot; 536 } 537 538 static bool sock_map_sk_state_allowed(const struct sock *sk) 539 { 540 if (sk_is_tcp(sk)) 541 return (1 << sk->sk_state) & (TCPF_ESTABLISHED | TCPF_LISTEN); 542 if (sk_is_stream_unix(sk)) 543 return (1 << sk->sk_state) & TCPF_ESTABLISHED; 544 return true; 545 } 546 547 static int sock_hash_update_common(struct bpf_map *map, void *key, 548 struct sock *sk, u64 flags); 549 550 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, 551 u64 flags) 552 { 553 struct socket *sock; 554 struct sock *sk; 555 int ret; 556 u64 ufd; 557 558 if (map->value_size == sizeof(u64)) 559 ufd = *(u64 *)value; 560 else 561 ufd = *(u32 *)value; 562 if (ufd > S32_MAX) 563 return -EINVAL; 564 565 sock = sockfd_lookup(ufd, &ret); 566 if (!sock) 567 return ret; 568 sk = sock->sk; 569 if (!sk) { 570 ret = -EINVAL; 571 goto out; 572 } 573 if (!sock_map_sk_is_suitable(sk)) { 574 ret = -EOPNOTSUPP; 575 goto out; 576 } 577 578 sock_map_sk_acquire(sk); 579 if (!sock_map_sk_state_allowed(sk)) 580 ret = -EOPNOTSUPP; 581 else if (map->map_type == BPF_MAP_TYPE_SOCKMAP) 582 ret = sock_map_update_common(map, *(u32 *)key, sk, flags); 583 else 584 ret = sock_hash_update_common(map, key, sk, flags); 585 sock_map_sk_release(sk); 586 out: 587 sockfd_put(sock); 588 return ret; 589 } 590 591 static long sock_map_update_elem(struct bpf_map *map, void *key, 592 void *value, u64 flags) 593 { 594 struct sock *sk = (struct sock *)value; 595 int ret; 596 597 if (unlikely(!sk || !sk_fullsock(sk))) 598 return -EINVAL; 599 600 if (!sock_map_sk_is_suitable(sk)) 601 return -EOPNOTSUPP; 602 603 local_bh_disable(); 604 bh_lock_sock(sk); 605 if (!sock_map_sk_state_allowed(sk)) 606 ret = -EOPNOTSUPP; 607 else if (map->map_type == BPF_MAP_TYPE_SOCKMAP) 608 ret = sock_map_update_common(map, *(u32 *)key, sk, flags); 609 else 610 ret = sock_hash_update_common(map, key, sk, flags); 611 bh_unlock_sock(sk); 612 local_bh_enable(); 613 return ret; 614 } 615 616 BPF_CALL_4(bpf_sock_map_update, struct bpf_sock_ops_kern *, sops, 617 struct bpf_map *, map, void *, key, u64, flags) 618 { 619 WARN_ON_ONCE(!rcu_read_lock_held()); 620 621 if (likely(sock_map_sk_is_suitable(sops->sk) && 622 sock_map_op_okay(sops))) 623 return sock_map_update_common(map, *(u32 *)key, sops->sk, 624 flags); 625 return -EOPNOTSUPP; 626 } 627 628 const struct bpf_func_proto bpf_sock_map_update_proto = { 629 .func = bpf_sock_map_update, 630 .gpl_only = false, 631 .pkt_access = true, 632 .ret_type = RET_INTEGER, 633 .arg1_type = ARG_PTR_TO_CTX, 634 .arg2_type = ARG_CONST_MAP_PTR, 635 .arg3_type = ARG_PTR_TO_MAP_KEY, 636 .arg4_type = ARG_ANYTHING, 637 }; 638 639 BPF_CALL_4(bpf_sk_redirect_map, struct sk_buff *, skb, 640 struct bpf_map *, map, u32, key, u64, flags) 641 { 642 struct sock *sk; 643 644 if (unlikely(flags & ~(BPF_F_INGRESS))) 645 return SK_DROP; 646 647 sk = __sock_map_lookup_elem(map, key); 648 if (unlikely(!sk || !sock_map_redirect_allowed(sk))) 649 return SK_DROP; 650 if ((flags & BPF_F_INGRESS) && sk_is_vsock(sk)) 651 return SK_DROP; 652 653 skb_bpf_set_redir(skb, sk, flags & BPF_F_INGRESS); 654 return SK_PASS; 655 } 656 657 const struct bpf_func_proto bpf_sk_redirect_map_proto = { 658 .func = bpf_sk_redirect_map, 659 .gpl_only = false, 660 .ret_type = RET_INTEGER, 661 .arg1_type = ARG_PTR_TO_CTX, 662 .arg2_type = ARG_CONST_MAP_PTR, 663 .arg3_type = ARG_ANYTHING, 664 .arg4_type = ARG_ANYTHING, 665 }; 666 667 BPF_CALL_4(bpf_msg_redirect_map, struct sk_msg *, msg, 668 struct bpf_map *, map, u32, key, u64, flags) 669 { 670 struct sock *sk; 671 672 if (unlikely(flags & ~(BPF_F_INGRESS))) 673 return SK_DROP; 674 675 sk = __sock_map_lookup_elem(map, key); 676 if (unlikely(!sk || !sock_map_redirect_allowed(sk))) 677 return SK_DROP; 678 if (!(flags & BPF_F_INGRESS) && !sk_is_tcp(sk)) 679 return SK_DROP; 680 if (sk_is_vsock(sk)) 681 return SK_DROP; 682 683 msg->flags = flags; 684 msg->sk_redir = sk; 685 return SK_PASS; 686 } 687 688 const struct bpf_func_proto bpf_msg_redirect_map_proto = { 689 .func = bpf_msg_redirect_map, 690 .gpl_only = false, 691 .ret_type = RET_INTEGER, 692 .arg1_type = ARG_PTR_TO_CTX, 693 .arg2_type = ARG_CONST_MAP_PTR, 694 .arg3_type = ARG_ANYTHING, 695 .arg4_type = ARG_ANYTHING, 696 }; 697 698 struct sock_map_seq_info { 699 struct bpf_map *map; 700 struct sock *sk; 701 u32 index; 702 }; 703 704 struct bpf_iter__sockmap { 705 __bpf_md_ptr(struct bpf_iter_meta *, meta); 706 __bpf_md_ptr(struct bpf_map *, map); 707 __bpf_md_ptr(void *, key); 708 __bpf_md_ptr(struct sock *, sk); 709 }; 710 711 DEFINE_BPF_ITER_FUNC(sockmap, struct bpf_iter_meta *meta, 712 struct bpf_map *map, void *key, 713 struct sock *sk) 714 715 static void *sock_map_seq_lookup_elem(struct sock_map_seq_info *info) 716 { 717 if (unlikely(info->index >= info->map->max_entries)) 718 return NULL; 719 720 info->sk = __sock_map_lookup_elem(info->map, info->index); 721 722 /* can't return sk directly, since that might be NULL */ 723 return info; 724 } 725 726 static void *sock_map_seq_start(struct seq_file *seq, loff_t *pos) 727 __acquires(rcu) 728 { 729 struct sock_map_seq_info *info = seq->private; 730 731 if (*pos == 0) 732 ++*pos; 733 734 /* pairs with sock_map_seq_stop */ 735 rcu_read_lock(); 736 return sock_map_seq_lookup_elem(info); 737 } 738 739 static void *sock_map_seq_next(struct seq_file *seq, void *v, loff_t *pos) 740 __must_hold(rcu) 741 { 742 struct sock_map_seq_info *info = seq->private; 743 744 ++*pos; 745 ++info->index; 746 747 return sock_map_seq_lookup_elem(info); 748 } 749 750 static int sock_map_seq_show(struct seq_file *seq, void *v) 751 __must_hold(rcu) 752 { 753 struct sock_map_seq_info *info = seq->private; 754 struct bpf_iter__sockmap ctx = {}; 755 struct bpf_iter_meta meta; 756 struct bpf_prog *prog; 757 758 meta.seq = seq; 759 prog = bpf_iter_get_info(&meta, !v); 760 if (!prog) 761 return 0; 762 763 ctx.meta = &meta; 764 ctx.map = info->map; 765 if (v) { 766 ctx.key = &info->index; 767 ctx.sk = info->sk; 768 } 769 770 return bpf_iter_run_prog(prog, &ctx); 771 } 772 773 static void sock_map_seq_stop(struct seq_file *seq, void *v) 774 __releases(rcu) 775 { 776 if (!v) 777 (void)sock_map_seq_show(seq, NULL); 778 779 /* pairs with sock_map_seq_start */ 780 rcu_read_unlock(); 781 } 782 783 static const struct seq_operations sock_map_seq_ops = { 784 .start = sock_map_seq_start, 785 .next = sock_map_seq_next, 786 .stop = sock_map_seq_stop, 787 .show = sock_map_seq_show, 788 }; 789 790 static int sock_map_init_seq_private(void *priv_data, 791 struct bpf_iter_aux_info *aux) 792 { 793 struct sock_map_seq_info *info = priv_data; 794 795 bpf_map_inc_with_uref(aux->map); 796 info->map = aux->map; 797 return 0; 798 } 799 800 static void sock_map_fini_seq_private(void *priv_data) 801 { 802 struct sock_map_seq_info *info = priv_data; 803 804 bpf_map_put_with_uref(info->map); 805 } 806 807 static u64 sock_map_mem_usage(const struct bpf_map *map) 808 { 809 u64 usage = sizeof(struct bpf_stab); 810 811 usage += (u64)map->max_entries * sizeof(struct sock *); 812 return usage; 813 } 814 815 static const struct bpf_iter_seq_info sock_map_iter_seq_info = { 816 .seq_ops = &sock_map_seq_ops, 817 .init_seq_private = sock_map_init_seq_private, 818 .fini_seq_private = sock_map_fini_seq_private, 819 .seq_priv_size = sizeof(struct sock_map_seq_info), 820 }; 821 822 BTF_ID_LIST_SINGLE(sock_map_btf_ids, struct, bpf_stab) 823 const struct bpf_map_ops sock_map_ops = { 824 .map_meta_equal = bpf_map_meta_equal, 825 .map_alloc = sock_map_alloc, 826 .map_free = sock_map_free, 827 .map_get_next_key = sock_map_get_next_key, 828 .map_lookup_elem_sys_only = sock_map_lookup_sys, 829 .map_update_elem = sock_map_update_elem, 830 .map_delete_elem = sock_map_delete_elem, 831 .map_lookup_elem = sock_map_lookup, 832 .map_release_uref = sock_map_release_progs, 833 .map_check_btf = map_check_no_btf, 834 .map_mem_usage = sock_map_mem_usage, 835 .map_btf_id = &sock_map_btf_ids[0], 836 .iter_seq_info = &sock_map_iter_seq_info, 837 }; 838 839 struct bpf_shtab_elem { 840 struct rcu_head rcu; 841 u32 hash; 842 struct sock *sk; 843 struct hlist_node node; 844 u8 key[]; 845 }; 846 847 struct bpf_shtab_bucket { 848 struct hlist_head head; 849 spinlock_t lock; 850 }; 851 852 struct bpf_shtab { 853 struct bpf_map map; 854 struct bpf_shtab_bucket *buckets; 855 u32 buckets_num; 856 u32 elem_size; 857 struct sk_psock_progs progs; 858 atomic_t count; 859 }; 860 861 static inline u32 sock_hash_bucket_hash(const void *key, u32 len) 862 { 863 return jhash(key, len, 0); 864 } 865 866 static struct bpf_shtab_bucket *sock_hash_select_bucket(struct bpf_shtab *htab, 867 u32 hash) 868 { 869 return &htab->buckets[hash & (htab->buckets_num - 1)]; 870 } 871 872 static struct bpf_shtab_elem * 873 sock_hash_lookup_elem_raw(struct hlist_head *head, u32 hash, void *key, 874 u32 key_size) 875 { 876 struct bpf_shtab_elem *elem; 877 878 hlist_for_each_entry_rcu(elem, head, node) { 879 if (elem->hash == hash && 880 !memcmp(&elem->key, key, key_size)) 881 return elem; 882 } 883 884 return NULL; 885 } 886 887 static struct sock *__sock_hash_lookup_elem(struct bpf_map *map, void *key) 888 { 889 struct bpf_shtab *htab = container_of(map, struct bpf_shtab, map); 890 u32 key_size = map->key_size, hash; 891 struct bpf_shtab_bucket *bucket; 892 struct bpf_shtab_elem *elem; 893 894 WARN_ON_ONCE(!rcu_read_lock_held()); 895 896 hash = sock_hash_bucket_hash(key, key_size); 897 bucket = sock_hash_select_bucket(htab, hash); 898 elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size); 899 900 return elem ? elem->sk : NULL; 901 } 902 903 static void sock_hash_free_elem(struct bpf_shtab *htab, 904 struct bpf_shtab_elem *elem) 905 { 906 atomic_dec(&htab->count); 907 kfree_rcu(elem, rcu); 908 } 909 910 static void sock_hash_delete_from_link(struct bpf_map *map, struct sock *sk, 911 void *link_raw) 912 { 913 struct bpf_shtab *htab = container_of(map, struct bpf_shtab, map); 914 struct bpf_shtab_elem *elem_probe, *elem = link_raw; 915 struct bpf_shtab_bucket *bucket; 916 917 WARN_ON_ONCE(!rcu_read_lock_held()); 918 bucket = sock_hash_select_bucket(htab, elem->hash); 919 920 /* elem may be deleted in parallel from the map, but access here 921 * is okay since it's going away only after RCU grace period. 922 * However, we need to check whether it's still present. 923 */ 924 spin_lock_bh(&bucket->lock); 925 elem_probe = sock_hash_lookup_elem_raw(&bucket->head, elem->hash, 926 elem->key, map->key_size); 927 if (elem_probe && elem_probe == elem) { 928 hlist_del_rcu(&elem->node); 929 sock_map_unref(elem->sk, elem); 930 sock_hash_free_elem(htab, elem); 931 } 932 spin_unlock_bh(&bucket->lock); 933 } 934 935 static long sock_hash_delete_elem(struct bpf_map *map, void *key) 936 { 937 struct bpf_shtab *htab = container_of(map, struct bpf_shtab, map); 938 u32 hash, key_size = map->key_size; 939 struct bpf_shtab_bucket *bucket; 940 struct bpf_shtab_elem *elem; 941 int ret = -ENOENT; 942 943 hash = sock_hash_bucket_hash(key, key_size); 944 bucket = sock_hash_select_bucket(htab, hash); 945 946 spin_lock_bh(&bucket->lock); 947 elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size); 948 if (elem) { 949 hlist_del_rcu(&elem->node); 950 sock_map_unref(elem->sk, elem); 951 sock_hash_free_elem(htab, elem); 952 ret = 0; 953 } 954 spin_unlock_bh(&bucket->lock); 955 return ret; 956 } 957 958 static struct bpf_shtab_elem *sock_hash_alloc_elem(struct bpf_shtab *htab, 959 void *key, u32 key_size, 960 u32 hash, struct sock *sk, 961 struct bpf_shtab_elem *old) 962 { 963 struct bpf_shtab_elem *new; 964 965 if (atomic_inc_return(&htab->count) > htab->map.max_entries) { 966 if (!old) { 967 atomic_dec(&htab->count); 968 return ERR_PTR(-E2BIG); 969 } 970 } 971 972 new = bpf_map_kmalloc_node(&htab->map, htab->elem_size, 973 GFP_ATOMIC | __GFP_NOWARN, 974 htab->map.numa_node); 975 if (!new) { 976 atomic_dec(&htab->count); 977 return ERR_PTR(-ENOMEM); 978 } 979 memcpy(new->key, key, key_size); 980 new->sk = sk; 981 new->hash = hash; 982 return new; 983 } 984 985 static int sock_hash_update_common(struct bpf_map *map, void *key, 986 struct sock *sk, u64 flags) 987 { 988 struct bpf_shtab *htab = container_of(map, struct bpf_shtab, map); 989 u32 key_size = map->key_size, hash; 990 struct bpf_shtab_elem *elem, *elem_new; 991 struct bpf_shtab_bucket *bucket; 992 struct sk_psock_link *link; 993 struct sk_psock *psock; 994 int ret; 995 996 WARN_ON_ONCE(!rcu_read_lock_held()); 997 if (unlikely(flags > BPF_EXIST)) 998 return -EINVAL; 999 1000 link = sk_psock_init_link(); 1001 if (!link) 1002 return -ENOMEM; 1003 1004 ret = sock_map_link(map, sk); 1005 if (ret < 0) 1006 goto out_free; 1007 1008 psock = sk_psock(sk); 1009 WARN_ON_ONCE(!psock); 1010 1011 hash = sock_hash_bucket_hash(key, key_size); 1012 bucket = sock_hash_select_bucket(htab, hash); 1013 1014 spin_lock_bh(&bucket->lock); 1015 elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size); 1016 if (elem && flags == BPF_NOEXIST) { 1017 ret = -EEXIST; 1018 goto out_unlock; 1019 } else if (!elem && flags == BPF_EXIST) { 1020 ret = -ENOENT; 1021 goto out_unlock; 1022 } 1023 1024 elem_new = sock_hash_alloc_elem(htab, key, key_size, hash, sk, elem); 1025 if (IS_ERR(elem_new)) { 1026 ret = PTR_ERR(elem_new); 1027 goto out_unlock; 1028 } 1029 1030 sock_map_add_link(psock, link, map, elem_new); 1031 /* Add new element to the head of the list, so that 1032 * concurrent search will find it before old elem. 1033 */ 1034 hlist_add_head_rcu(&elem_new->node, &bucket->head); 1035 if (elem) { 1036 hlist_del_rcu(&elem->node); 1037 sock_map_unref(elem->sk, elem); 1038 sock_hash_free_elem(htab, elem); 1039 } 1040 spin_unlock_bh(&bucket->lock); 1041 return 0; 1042 out_unlock: 1043 spin_unlock_bh(&bucket->lock); 1044 sk_psock_put(sk, psock); 1045 out_free: 1046 sk_psock_free_link(link); 1047 return ret; 1048 } 1049 1050 static int sock_hash_get_next_key(struct bpf_map *map, void *key, 1051 void *key_next) 1052 { 1053 struct bpf_shtab *htab = container_of(map, struct bpf_shtab, map); 1054 struct bpf_shtab_elem *elem, *elem_next; 1055 u32 hash, key_size = map->key_size; 1056 struct hlist_head *head; 1057 int i = 0; 1058 1059 if (!key) 1060 goto find_first_elem; 1061 hash = sock_hash_bucket_hash(key, key_size); 1062 head = &sock_hash_select_bucket(htab, hash)->head; 1063 elem = sock_hash_lookup_elem_raw(head, hash, key, key_size); 1064 if (!elem) 1065 goto find_first_elem; 1066 1067 elem_next = hlist_entry_safe(rcu_dereference(hlist_next_rcu(&elem->node)), 1068 struct bpf_shtab_elem, node); 1069 if (elem_next) { 1070 memcpy(key_next, elem_next->key, key_size); 1071 return 0; 1072 } 1073 1074 i = hash & (htab->buckets_num - 1); 1075 i++; 1076 find_first_elem: 1077 for (; i < htab->buckets_num; i++) { 1078 head = &sock_hash_select_bucket(htab, i)->head; 1079 elem_next = hlist_entry_safe(rcu_dereference(hlist_first_rcu(head)), 1080 struct bpf_shtab_elem, node); 1081 if (elem_next) { 1082 memcpy(key_next, elem_next->key, key_size); 1083 return 0; 1084 } 1085 } 1086 1087 return -ENOENT; 1088 } 1089 1090 static struct bpf_map *sock_hash_alloc(union bpf_attr *attr) 1091 { 1092 struct bpf_shtab *htab; 1093 int i, err; 1094 1095 if (attr->max_entries == 0 || 1096 attr->key_size == 0 || 1097 (attr->value_size != sizeof(u32) && 1098 attr->value_size != sizeof(u64)) || 1099 attr->map_flags & ~SOCK_CREATE_FLAG_MASK) 1100 return ERR_PTR(-EINVAL); 1101 if (attr->key_size > MAX_BPF_STACK) 1102 return ERR_PTR(-E2BIG); 1103 1104 htab = bpf_map_area_alloc(sizeof(*htab), NUMA_NO_NODE); 1105 if (!htab) 1106 return ERR_PTR(-ENOMEM); 1107 1108 bpf_map_init_from_attr(&htab->map, attr); 1109 1110 htab->buckets_num = roundup_pow_of_two(htab->map.max_entries); 1111 htab->elem_size = sizeof(struct bpf_shtab_elem) + 1112 round_up(htab->map.key_size, 8); 1113 if (htab->buckets_num == 0 || 1114 htab->buckets_num > U32_MAX / sizeof(struct bpf_shtab_bucket)) { 1115 err = -EINVAL; 1116 goto free_htab; 1117 } 1118 1119 htab->buckets = bpf_map_area_alloc(htab->buckets_num * 1120 sizeof(struct bpf_shtab_bucket), 1121 htab->map.numa_node); 1122 if (!htab->buckets) { 1123 err = -ENOMEM; 1124 goto free_htab; 1125 } 1126 1127 for (i = 0; i < htab->buckets_num; i++) { 1128 INIT_HLIST_HEAD(&htab->buckets[i].head); 1129 spin_lock_init(&htab->buckets[i].lock); 1130 } 1131 1132 return &htab->map; 1133 free_htab: 1134 bpf_map_area_free(htab); 1135 return ERR_PTR(err); 1136 } 1137 1138 static void sock_hash_free(struct bpf_map *map) 1139 { 1140 struct bpf_shtab *htab = container_of(map, struct bpf_shtab, map); 1141 struct bpf_shtab_bucket *bucket; 1142 struct hlist_head unlink_list; 1143 struct bpf_shtab_elem *elem; 1144 struct hlist_node *node; 1145 int i; 1146 1147 /* After the sync no updates or deletes will be in-flight so it 1148 * is safe to walk map and remove entries without risking a race 1149 * in EEXIST update case. 1150 */ 1151 synchronize_rcu(); 1152 for (i = 0; i < htab->buckets_num; i++) { 1153 bucket = sock_hash_select_bucket(htab, i); 1154 1155 /* We are racing with sock_hash_delete_from_link to 1156 * enter the spin-lock critical section. Every socket on 1157 * the list is still linked to sockhash. Since link 1158 * exists, psock exists and holds a ref to socket. That 1159 * lets us to grab a socket ref too. 1160 */ 1161 spin_lock_bh(&bucket->lock); 1162 hlist_for_each_entry(elem, &bucket->head, node) 1163 sock_hold(elem->sk); 1164 hlist_move_list(&bucket->head, &unlink_list); 1165 spin_unlock_bh(&bucket->lock); 1166 1167 /* Process removed entries out of atomic context to 1168 * block for socket lock before deleting the psock's 1169 * link to sockhash. 1170 */ 1171 hlist_for_each_entry_safe(elem, node, &unlink_list, node) { 1172 hlist_del(&elem->node); 1173 lock_sock(elem->sk); 1174 rcu_read_lock(); 1175 sock_map_unref(elem->sk, elem); 1176 rcu_read_unlock(); 1177 release_sock(elem->sk); 1178 sock_put(elem->sk); 1179 sock_hash_free_elem(htab, elem); 1180 } 1181 cond_resched(); 1182 } 1183 1184 /* wait for psock readers accessing its map link */ 1185 synchronize_rcu(); 1186 1187 bpf_map_area_free(htab->buckets); 1188 bpf_map_area_free(htab); 1189 } 1190 1191 static void *sock_hash_lookup_sys(struct bpf_map *map, void *key) 1192 { 1193 struct sock *sk; 1194 1195 if (map->value_size != sizeof(u64)) 1196 return ERR_PTR(-ENOSPC); 1197 1198 sk = __sock_hash_lookup_elem(map, key); 1199 if (!sk) 1200 return ERR_PTR(-ENOENT); 1201 1202 __sock_gen_cookie(sk); 1203 return &sk->sk_cookie; 1204 } 1205 1206 static void *sock_hash_lookup(struct bpf_map *map, void *key) 1207 { 1208 struct sock *sk; 1209 1210 sk = __sock_hash_lookup_elem(map, key); 1211 if (!sk) 1212 return NULL; 1213 if (sk_is_refcounted(sk) && !refcount_inc_not_zero(&sk->sk_refcnt)) 1214 return NULL; 1215 return sk; 1216 } 1217 1218 static void sock_hash_release_progs(struct bpf_map *map) 1219 { 1220 psock_progs_drop(&container_of(map, struct bpf_shtab, map)->progs); 1221 } 1222 1223 BPF_CALL_4(bpf_sock_hash_update, struct bpf_sock_ops_kern *, sops, 1224 struct bpf_map *, map, void *, key, u64, flags) 1225 { 1226 WARN_ON_ONCE(!rcu_read_lock_held()); 1227 1228 if (likely(sock_map_sk_is_suitable(sops->sk) && 1229 sock_map_op_okay(sops))) 1230 return sock_hash_update_common(map, key, sops->sk, flags); 1231 return -EOPNOTSUPP; 1232 } 1233 1234 const struct bpf_func_proto bpf_sock_hash_update_proto = { 1235 .func = bpf_sock_hash_update, 1236 .gpl_only = false, 1237 .pkt_access = true, 1238 .ret_type = RET_INTEGER, 1239 .arg1_type = ARG_PTR_TO_CTX, 1240 .arg2_type = ARG_CONST_MAP_PTR, 1241 .arg3_type = ARG_PTR_TO_MAP_KEY, 1242 .arg4_type = ARG_ANYTHING, 1243 }; 1244 1245 BPF_CALL_4(bpf_sk_redirect_hash, struct sk_buff *, skb, 1246 struct bpf_map *, map, void *, key, u64, flags) 1247 { 1248 struct sock *sk; 1249 1250 if (unlikely(flags & ~(BPF_F_INGRESS))) 1251 return SK_DROP; 1252 1253 sk = __sock_hash_lookup_elem(map, key); 1254 if (unlikely(!sk || !sock_map_redirect_allowed(sk))) 1255 return SK_DROP; 1256 if ((flags & BPF_F_INGRESS) && sk_is_vsock(sk)) 1257 return SK_DROP; 1258 1259 skb_bpf_set_redir(skb, sk, flags & BPF_F_INGRESS); 1260 return SK_PASS; 1261 } 1262 1263 const struct bpf_func_proto bpf_sk_redirect_hash_proto = { 1264 .func = bpf_sk_redirect_hash, 1265 .gpl_only = false, 1266 .ret_type = RET_INTEGER, 1267 .arg1_type = ARG_PTR_TO_CTX, 1268 .arg2_type = ARG_CONST_MAP_PTR, 1269 .arg3_type = ARG_PTR_TO_MAP_KEY, 1270 .arg4_type = ARG_ANYTHING, 1271 }; 1272 1273 BPF_CALL_4(bpf_msg_redirect_hash, struct sk_msg *, msg, 1274 struct bpf_map *, map, void *, key, u64, flags) 1275 { 1276 struct sock *sk; 1277 1278 if (unlikely(flags & ~(BPF_F_INGRESS))) 1279 return SK_DROP; 1280 1281 sk = __sock_hash_lookup_elem(map, key); 1282 if (unlikely(!sk || !sock_map_redirect_allowed(sk))) 1283 return SK_DROP; 1284 if (!(flags & BPF_F_INGRESS) && !sk_is_tcp(sk)) 1285 return SK_DROP; 1286 if (sk_is_vsock(sk)) 1287 return SK_DROP; 1288 1289 msg->flags = flags; 1290 msg->sk_redir = sk; 1291 return SK_PASS; 1292 } 1293 1294 const struct bpf_func_proto bpf_msg_redirect_hash_proto = { 1295 .func = bpf_msg_redirect_hash, 1296 .gpl_only = false, 1297 .ret_type = RET_INTEGER, 1298 .arg1_type = ARG_PTR_TO_CTX, 1299 .arg2_type = ARG_CONST_MAP_PTR, 1300 .arg3_type = ARG_PTR_TO_MAP_KEY, 1301 .arg4_type = ARG_ANYTHING, 1302 }; 1303 1304 struct sock_hash_seq_info { 1305 struct bpf_map *map; 1306 struct bpf_shtab *htab; 1307 u32 bucket_id; 1308 }; 1309 1310 static void *sock_hash_seq_find_next(struct sock_hash_seq_info *info, 1311 struct bpf_shtab_elem *prev_elem) 1312 { 1313 const struct bpf_shtab *htab = info->htab; 1314 struct bpf_shtab_bucket *bucket; 1315 struct bpf_shtab_elem *elem; 1316 struct hlist_node *node; 1317 1318 /* try to find next elem in the same bucket */ 1319 if (prev_elem) { 1320 node = rcu_dereference(hlist_next_rcu(&prev_elem->node)); 1321 elem = hlist_entry_safe(node, struct bpf_shtab_elem, node); 1322 if (elem) 1323 return elem; 1324 1325 /* no more elements, continue in the next bucket */ 1326 info->bucket_id++; 1327 } 1328 1329 for (; info->bucket_id < htab->buckets_num; info->bucket_id++) { 1330 bucket = &htab->buckets[info->bucket_id]; 1331 node = rcu_dereference(hlist_first_rcu(&bucket->head)); 1332 elem = hlist_entry_safe(node, struct bpf_shtab_elem, node); 1333 if (elem) 1334 return elem; 1335 } 1336 1337 return NULL; 1338 } 1339 1340 static void *sock_hash_seq_start(struct seq_file *seq, loff_t *pos) 1341 __acquires(rcu) 1342 { 1343 struct sock_hash_seq_info *info = seq->private; 1344 1345 if (*pos == 0) 1346 ++*pos; 1347 1348 /* pairs with sock_hash_seq_stop */ 1349 rcu_read_lock(); 1350 return sock_hash_seq_find_next(info, NULL); 1351 } 1352 1353 static void *sock_hash_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1354 __must_hold(rcu) 1355 { 1356 struct sock_hash_seq_info *info = seq->private; 1357 1358 ++*pos; 1359 return sock_hash_seq_find_next(info, v); 1360 } 1361 1362 static int sock_hash_seq_show(struct seq_file *seq, void *v) 1363 __must_hold(rcu) 1364 { 1365 struct sock_hash_seq_info *info = seq->private; 1366 struct bpf_iter__sockmap ctx = {}; 1367 struct bpf_shtab_elem *elem = v; 1368 struct bpf_iter_meta meta; 1369 struct bpf_prog *prog; 1370 1371 meta.seq = seq; 1372 prog = bpf_iter_get_info(&meta, !elem); 1373 if (!prog) 1374 return 0; 1375 1376 ctx.meta = &meta; 1377 ctx.map = info->map; 1378 if (elem) { 1379 ctx.key = elem->key; 1380 ctx.sk = elem->sk; 1381 } 1382 1383 return bpf_iter_run_prog(prog, &ctx); 1384 } 1385 1386 static void sock_hash_seq_stop(struct seq_file *seq, void *v) 1387 __releases(rcu) 1388 { 1389 if (!v) 1390 (void)sock_hash_seq_show(seq, NULL); 1391 1392 /* pairs with sock_hash_seq_start */ 1393 rcu_read_unlock(); 1394 } 1395 1396 static const struct seq_operations sock_hash_seq_ops = { 1397 .start = sock_hash_seq_start, 1398 .next = sock_hash_seq_next, 1399 .stop = sock_hash_seq_stop, 1400 .show = sock_hash_seq_show, 1401 }; 1402 1403 static int sock_hash_init_seq_private(void *priv_data, 1404 struct bpf_iter_aux_info *aux) 1405 { 1406 struct sock_hash_seq_info *info = priv_data; 1407 1408 bpf_map_inc_with_uref(aux->map); 1409 info->map = aux->map; 1410 info->htab = container_of(aux->map, struct bpf_shtab, map); 1411 return 0; 1412 } 1413 1414 static void sock_hash_fini_seq_private(void *priv_data) 1415 { 1416 struct sock_hash_seq_info *info = priv_data; 1417 1418 bpf_map_put_with_uref(info->map); 1419 } 1420 1421 static u64 sock_hash_mem_usage(const struct bpf_map *map) 1422 { 1423 struct bpf_shtab *htab = container_of(map, struct bpf_shtab, map); 1424 u64 usage = sizeof(*htab); 1425 1426 usage += htab->buckets_num * sizeof(struct bpf_shtab_bucket); 1427 usage += atomic_read(&htab->count) * (u64)htab->elem_size; 1428 return usage; 1429 } 1430 1431 static const struct bpf_iter_seq_info sock_hash_iter_seq_info = { 1432 .seq_ops = &sock_hash_seq_ops, 1433 .init_seq_private = sock_hash_init_seq_private, 1434 .fini_seq_private = sock_hash_fini_seq_private, 1435 .seq_priv_size = sizeof(struct sock_hash_seq_info), 1436 }; 1437 1438 BTF_ID_LIST_SINGLE(sock_hash_map_btf_ids, struct, bpf_shtab) 1439 const struct bpf_map_ops sock_hash_ops = { 1440 .map_meta_equal = bpf_map_meta_equal, 1441 .map_alloc = sock_hash_alloc, 1442 .map_free = sock_hash_free, 1443 .map_get_next_key = sock_hash_get_next_key, 1444 .map_update_elem = sock_map_update_elem, 1445 .map_delete_elem = sock_hash_delete_elem, 1446 .map_lookup_elem = sock_hash_lookup, 1447 .map_lookup_elem_sys_only = sock_hash_lookup_sys, 1448 .map_release_uref = sock_hash_release_progs, 1449 .map_check_btf = map_check_no_btf, 1450 .map_mem_usage = sock_hash_mem_usage, 1451 .map_btf_id = &sock_hash_map_btf_ids[0], 1452 .iter_seq_info = &sock_hash_iter_seq_info, 1453 }; 1454 1455 static struct sk_psock_progs *sock_map_progs(struct bpf_map *map) 1456 { 1457 switch (map->map_type) { 1458 case BPF_MAP_TYPE_SOCKMAP: 1459 return &container_of(map, struct bpf_stab, map)->progs; 1460 case BPF_MAP_TYPE_SOCKHASH: 1461 return &container_of(map, struct bpf_shtab, map)->progs; 1462 default: 1463 break; 1464 } 1465 1466 return NULL; 1467 } 1468 1469 static int sock_map_prog_link_lookup(struct bpf_map *map, struct bpf_prog ***pprog, 1470 struct bpf_link ***plink, u32 which) 1471 { 1472 struct sk_psock_progs *progs = sock_map_progs(map); 1473 struct bpf_prog **cur_pprog; 1474 struct bpf_link **cur_plink; 1475 1476 if (!progs) 1477 return -EOPNOTSUPP; 1478 1479 switch (which) { 1480 case BPF_SK_MSG_VERDICT: 1481 cur_pprog = &progs->msg_parser; 1482 cur_plink = &progs->msg_parser_link; 1483 break; 1484 #if IS_ENABLED(CONFIG_BPF_STREAM_PARSER) 1485 case BPF_SK_SKB_STREAM_PARSER: 1486 cur_pprog = &progs->stream_parser; 1487 cur_plink = &progs->stream_parser_link; 1488 break; 1489 #endif 1490 case BPF_SK_SKB_STREAM_VERDICT: 1491 if (progs->skb_verdict) 1492 return -EBUSY; 1493 cur_pprog = &progs->stream_verdict; 1494 cur_plink = &progs->stream_verdict_link; 1495 break; 1496 case BPF_SK_SKB_VERDICT: 1497 if (progs->stream_verdict) 1498 return -EBUSY; 1499 cur_pprog = &progs->skb_verdict; 1500 cur_plink = &progs->skb_verdict_link; 1501 break; 1502 default: 1503 return -EOPNOTSUPP; 1504 } 1505 1506 *pprog = cur_pprog; 1507 if (plink) 1508 *plink = cur_plink; 1509 return 0; 1510 } 1511 1512 /* Handle the following four cases: 1513 * prog_attach: prog != NULL, old == NULL, link == NULL 1514 * prog_detach: prog == NULL, old != NULL, link == NULL 1515 * link_attach: prog != NULL, old == NULL, link != NULL 1516 * link_detach: prog == NULL, old != NULL, link != NULL 1517 */ 1518 static int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog, 1519 struct bpf_prog *old, struct bpf_link *link, 1520 u32 which) 1521 { 1522 struct bpf_prog **pprog; 1523 struct bpf_link **plink; 1524 int ret; 1525 1526 ret = sock_map_prog_link_lookup(map, &pprog, &plink, which); 1527 if (ret) 1528 return ret; 1529 1530 /* for prog_attach/prog_detach/link_attach, return error if a bpf_link 1531 * exists for that prog. 1532 */ 1533 if ((!link || prog) && *plink) 1534 return -EBUSY; 1535 1536 if (old) { 1537 ret = psock_replace_prog(pprog, prog, old); 1538 if (!ret) 1539 *plink = NULL; 1540 } else { 1541 psock_set_prog(pprog, prog); 1542 if (link) 1543 *plink = link; 1544 } 1545 1546 return ret; 1547 } 1548 1549 int sock_map_bpf_prog_query(const union bpf_attr *attr, 1550 union bpf_attr __user *uattr) 1551 { 1552 __u32 __user *prog_ids = u64_to_user_ptr(attr->query.prog_ids); 1553 u32 prog_cnt = 0, flags = 0; 1554 struct bpf_prog **pprog; 1555 struct bpf_prog *prog; 1556 struct bpf_map *map; 1557 u32 id = 0; 1558 int ret; 1559 1560 if (attr->query.query_flags) 1561 return -EINVAL; 1562 1563 CLASS(fd, f)(attr->target_fd); 1564 map = __bpf_map_get(f); 1565 if (IS_ERR(map)) 1566 return PTR_ERR(map); 1567 1568 rcu_read_lock(); 1569 1570 ret = sock_map_prog_link_lookup(map, &pprog, NULL, attr->query.attach_type); 1571 if (ret) 1572 goto end; 1573 1574 prog = *pprog; 1575 prog_cnt = !prog ? 0 : 1; 1576 1577 if (!attr->query.prog_cnt || !prog_ids || !prog_cnt) 1578 goto end; 1579 1580 /* we do not hold the refcnt, the bpf prog may be released 1581 * asynchronously and the id would be set to 0. 1582 */ 1583 id = data_race(prog->aux->id); 1584 if (id == 0) 1585 prog_cnt = 0; 1586 1587 end: 1588 rcu_read_unlock(); 1589 1590 if (copy_to_user(&uattr->query.attach_flags, &flags, sizeof(flags)) || 1591 (id != 0 && copy_to_user(prog_ids, &id, sizeof(u32))) || 1592 copy_to_user(&uattr->query.prog_cnt, &prog_cnt, sizeof(prog_cnt))) 1593 ret = -EFAULT; 1594 1595 return ret; 1596 } 1597 1598 static void sock_map_unlink(struct sock *sk, struct sk_psock_link *link) 1599 { 1600 switch (link->map->map_type) { 1601 case BPF_MAP_TYPE_SOCKMAP: 1602 return sock_map_delete_from_link(link->map, sk, 1603 link->link_raw); 1604 case BPF_MAP_TYPE_SOCKHASH: 1605 return sock_hash_delete_from_link(link->map, sk, 1606 link->link_raw); 1607 default: 1608 break; 1609 } 1610 } 1611 1612 static void sock_map_remove_links(struct sock *sk, struct sk_psock *psock) 1613 { 1614 struct sk_psock_link *link; 1615 1616 while ((link = sk_psock_link_pop(psock))) { 1617 sock_map_unlink(sk, link); 1618 sk_psock_free_link(link); 1619 } 1620 } 1621 1622 void sock_map_unhash(struct sock *sk) 1623 { 1624 void (*saved_unhash)(struct sock *sk); 1625 struct sk_psock *psock; 1626 1627 rcu_read_lock(); 1628 psock = sk_psock(sk); 1629 if (unlikely(!psock)) { 1630 rcu_read_unlock(); 1631 saved_unhash = READ_ONCE(sk->sk_prot)->unhash; 1632 } else { 1633 saved_unhash = psock->saved_unhash; 1634 sock_map_remove_links(sk, psock); 1635 rcu_read_unlock(); 1636 } 1637 if (WARN_ON_ONCE(saved_unhash == sock_map_unhash)) 1638 return; 1639 if (saved_unhash) 1640 saved_unhash(sk); 1641 } 1642 EXPORT_SYMBOL_GPL(sock_map_unhash); 1643 1644 void sock_map_destroy(struct sock *sk) 1645 { 1646 void (*saved_destroy)(struct sock *sk); 1647 struct sk_psock *psock; 1648 1649 rcu_read_lock(); 1650 psock = sk_psock_get(sk); 1651 if (unlikely(!psock)) { 1652 rcu_read_unlock(); 1653 saved_destroy = READ_ONCE(sk->sk_prot)->destroy; 1654 } else { 1655 saved_destroy = psock->saved_destroy; 1656 sock_map_remove_links(sk, psock); 1657 rcu_read_unlock(); 1658 sk_psock_stop(psock); 1659 sk_psock_put(sk, psock); 1660 } 1661 if (WARN_ON_ONCE(saved_destroy == sock_map_destroy)) 1662 return; 1663 if (saved_destroy) 1664 saved_destroy(sk); 1665 } 1666 EXPORT_SYMBOL_GPL(sock_map_destroy); 1667 1668 void sock_map_close(struct sock *sk, long timeout) 1669 { 1670 void (*saved_close)(struct sock *sk, long timeout); 1671 struct sk_psock *psock; 1672 1673 lock_sock(sk); 1674 rcu_read_lock(); 1675 psock = sk_psock(sk); 1676 if (likely(psock)) { 1677 saved_close = psock->saved_close; 1678 sock_map_remove_links(sk, psock); 1679 psock = sk_psock_get(sk); 1680 if (unlikely(!psock)) 1681 goto no_psock; 1682 rcu_read_unlock(); 1683 sk_psock_stop(psock); 1684 release_sock(sk); 1685 cancel_delayed_work_sync(&psock->work); 1686 sk_psock_put(sk, psock); 1687 } else { 1688 saved_close = READ_ONCE(sk->sk_prot)->close; 1689 no_psock: 1690 rcu_read_unlock(); 1691 release_sock(sk); 1692 } 1693 1694 /* Make sure we do not recurse. This is a bug. 1695 * Leak the socket instead of crashing on a stack overflow. 1696 */ 1697 if (WARN_ON_ONCE(saved_close == sock_map_close)) 1698 return; 1699 saved_close(sk, timeout); 1700 } 1701 EXPORT_SYMBOL_GPL(sock_map_close); 1702 1703 struct sockmap_link { 1704 struct bpf_link link; 1705 struct bpf_map *map; 1706 enum bpf_attach_type attach_type; 1707 }; 1708 1709 static void sock_map_link_release(struct bpf_link *link) 1710 { 1711 struct sockmap_link *sockmap_link = container_of(link, struct sockmap_link, link); 1712 1713 mutex_lock(&sockmap_mutex); 1714 if (!sockmap_link->map) 1715 goto out; 1716 1717 WARN_ON_ONCE(sock_map_prog_update(sockmap_link->map, NULL, link->prog, link, 1718 sockmap_link->attach_type)); 1719 1720 bpf_map_put_with_uref(sockmap_link->map); 1721 sockmap_link->map = NULL; 1722 out: 1723 mutex_unlock(&sockmap_mutex); 1724 } 1725 1726 static int sock_map_link_detach(struct bpf_link *link) 1727 { 1728 sock_map_link_release(link); 1729 return 0; 1730 } 1731 1732 static void sock_map_link_dealloc(struct bpf_link *link) 1733 { 1734 kfree(link); 1735 } 1736 1737 /* Handle the following two cases: 1738 * case 1: link != NULL, prog != NULL, old != NULL 1739 * case 2: link != NULL, prog != NULL, old == NULL 1740 */ 1741 static int sock_map_link_update_prog(struct bpf_link *link, 1742 struct bpf_prog *prog, 1743 struct bpf_prog *old) 1744 { 1745 const struct sockmap_link *sockmap_link = container_of(link, struct sockmap_link, link); 1746 struct bpf_prog **pprog, *old_link_prog; 1747 struct bpf_link **plink; 1748 int ret = 0; 1749 1750 mutex_lock(&sockmap_mutex); 1751 1752 /* If old prog is not NULL, ensure old prog is the same as link->prog. */ 1753 if (old && link->prog != old) { 1754 ret = -EPERM; 1755 goto out; 1756 } 1757 /* Ensure link->prog has the same type/attach_type as the new prog. */ 1758 if (link->prog->type != prog->type || 1759 link->prog->expected_attach_type != prog->expected_attach_type) { 1760 ret = -EINVAL; 1761 goto out; 1762 } 1763 1764 ret = sock_map_prog_link_lookup(sockmap_link->map, &pprog, &plink, 1765 sockmap_link->attach_type); 1766 if (ret) 1767 goto out; 1768 1769 /* return error if the stored bpf_link does not match the incoming bpf_link. */ 1770 if (link != *plink) { 1771 ret = -EBUSY; 1772 goto out; 1773 } 1774 1775 if (old) { 1776 ret = psock_replace_prog(pprog, prog, old); 1777 if (ret) 1778 goto out; 1779 } else { 1780 psock_set_prog(pprog, prog); 1781 } 1782 1783 bpf_prog_inc(prog); 1784 old_link_prog = xchg(&link->prog, prog); 1785 bpf_prog_put(old_link_prog); 1786 1787 out: 1788 mutex_unlock(&sockmap_mutex); 1789 return ret; 1790 } 1791 1792 static u32 sock_map_link_get_map_id(const struct sockmap_link *sockmap_link) 1793 { 1794 u32 map_id = 0; 1795 1796 mutex_lock(&sockmap_mutex); 1797 if (sockmap_link->map) 1798 map_id = sockmap_link->map->id; 1799 mutex_unlock(&sockmap_mutex); 1800 return map_id; 1801 } 1802 1803 static int sock_map_link_fill_info(const struct bpf_link *link, 1804 struct bpf_link_info *info) 1805 { 1806 const struct sockmap_link *sockmap_link = container_of(link, struct sockmap_link, link); 1807 u32 map_id = sock_map_link_get_map_id(sockmap_link); 1808 1809 info->sockmap.map_id = map_id; 1810 info->sockmap.attach_type = sockmap_link->attach_type; 1811 return 0; 1812 } 1813 1814 static void sock_map_link_show_fdinfo(const struct bpf_link *link, 1815 struct seq_file *seq) 1816 { 1817 const struct sockmap_link *sockmap_link = container_of(link, struct sockmap_link, link); 1818 u32 map_id = sock_map_link_get_map_id(sockmap_link); 1819 1820 seq_printf(seq, "map_id:\t%u\n", map_id); 1821 seq_printf(seq, "attach_type:\t%u\n", sockmap_link->attach_type); 1822 } 1823 1824 static const struct bpf_link_ops sock_map_link_ops = { 1825 .release = sock_map_link_release, 1826 .dealloc = sock_map_link_dealloc, 1827 .detach = sock_map_link_detach, 1828 .update_prog = sock_map_link_update_prog, 1829 .fill_link_info = sock_map_link_fill_info, 1830 .show_fdinfo = sock_map_link_show_fdinfo, 1831 }; 1832 1833 int sock_map_link_create(const union bpf_attr *attr, struct bpf_prog *prog) 1834 { 1835 struct bpf_link_primer link_primer; 1836 struct sockmap_link *sockmap_link; 1837 enum bpf_attach_type attach_type; 1838 struct bpf_map *map; 1839 int ret; 1840 1841 if (attr->link_create.flags) 1842 return -EINVAL; 1843 1844 map = bpf_map_get_with_uref(attr->link_create.target_fd); 1845 if (IS_ERR(map)) 1846 return PTR_ERR(map); 1847 if (map->map_type != BPF_MAP_TYPE_SOCKMAP && map->map_type != BPF_MAP_TYPE_SOCKHASH) { 1848 ret = -EINVAL; 1849 goto out; 1850 } 1851 1852 sockmap_link = kzalloc(sizeof(*sockmap_link), GFP_USER); 1853 if (!sockmap_link) { 1854 ret = -ENOMEM; 1855 goto out; 1856 } 1857 1858 attach_type = attr->link_create.attach_type; 1859 bpf_link_init(&sockmap_link->link, BPF_LINK_TYPE_SOCKMAP, &sock_map_link_ops, prog); 1860 sockmap_link->map = map; 1861 sockmap_link->attach_type = attach_type; 1862 1863 ret = bpf_link_prime(&sockmap_link->link, &link_primer); 1864 if (ret) { 1865 kfree(sockmap_link); 1866 goto out; 1867 } 1868 1869 mutex_lock(&sockmap_mutex); 1870 ret = sock_map_prog_update(map, prog, NULL, &sockmap_link->link, attach_type); 1871 mutex_unlock(&sockmap_mutex); 1872 if (ret) { 1873 bpf_link_cleanup(&link_primer); 1874 goto out; 1875 } 1876 1877 /* Increase refcnt for the prog since when old prog is replaced with 1878 * psock_replace_prog() and psock_set_prog() its refcnt will be decreased. 1879 * 1880 * Actually, we do not need to increase refcnt for the prog since bpf_link 1881 * will hold a reference. But in order to have less complexity w.r.t. 1882 * replacing/setting prog, let us increase the refcnt to make things simpler. 1883 */ 1884 bpf_prog_inc(prog); 1885 1886 return bpf_link_settle(&link_primer); 1887 1888 out: 1889 bpf_map_put_with_uref(map); 1890 return ret; 1891 } 1892 1893 static int sock_map_iter_attach_target(struct bpf_prog *prog, 1894 union bpf_iter_link_info *linfo, 1895 struct bpf_iter_aux_info *aux) 1896 { 1897 struct bpf_map *map; 1898 int err = -EINVAL; 1899 1900 if (!linfo->map.map_fd) 1901 return -EBADF; 1902 1903 map = bpf_map_get_with_uref(linfo->map.map_fd); 1904 if (IS_ERR(map)) 1905 return PTR_ERR(map); 1906 1907 if (map->map_type != BPF_MAP_TYPE_SOCKMAP && 1908 map->map_type != BPF_MAP_TYPE_SOCKHASH) 1909 goto put_map; 1910 1911 if (prog->aux->max_rdonly_access > map->key_size) { 1912 err = -EACCES; 1913 goto put_map; 1914 } 1915 1916 aux->map = map; 1917 return 0; 1918 1919 put_map: 1920 bpf_map_put_with_uref(map); 1921 return err; 1922 } 1923 1924 static void sock_map_iter_detach_target(struct bpf_iter_aux_info *aux) 1925 { 1926 bpf_map_put_with_uref(aux->map); 1927 } 1928 1929 static struct bpf_iter_reg sock_map_iter_reg = { 1930 .target = "sockmap", 1931 .attach_target = sock_map_iter_attach_target, 1932 .detach_target = sock_map_iter_detach_target, 1933 .show_fdinfo = bpf_iter_map_show_fdinfo, 1934 .fill_link_info = bpf_iter_map_fill_link_info, 1935 .ctx_arg_info_size = 2, 1936 .ctx_arg_info = { 1937 { offsetof(struct bpf_iter__sockmap, key), 1938 PTR_TO_BUF | PTR_MAYBE_NULL | MEM_RDONLY }, 1939 { offsetof(struct bpf_iter__sockmap, sk), 1940 PTR_TO_BTF_ID_OR_NULL }, 1941 }, 1942 }; 1943 1944 static int __init bpf_sockmap_iter_init(void) 1945 { 1946 sock_map_iter_reg.ctx_arg_info[1].btf_id = 1947 btf_sock_ids[BTF_SOCK_TYPE_SOCK]; 1948 return bpf_iter_reg_target(&sock_map_iter_reg); 1949 } 1950 late_initcall(bpf_sockmap_iter_init); 1951