1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * inet_diag.c Module for monitoring INET transport protocols sockets. 4 * 5 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> 6 */ 7 8 #include <linux/kernel.h> 9 #include <linux/module.h> 10 #include <linux/types.h> 11 #include <linux/fcntl.h> 12 #include <linux/random.h> 13 #include <linux/slab.h> 14 #include <linux/cache.h> 15 #include <linux/init.h> 16 #include <linux/time.h> 17 18 #include <net/icmp.h> 19 #include <net/tcp.h> 20 #include <net/ipv6.h> 21 #include <net/inet_common.h> 22 #include <net/inet_connection_sock.h> 23 #include <net/inet_hashtables.h> 24 #include <net/inet_timewait_sock.h> 25 #include <net/inet6_hashtables.h> 26 #include <net/netlink.h> 27 28 #include <linux/inet.h> 29 #include <linux/stddef.h> 30 31 #include <linux/inet_diag.h> 32 #include <linux/sock_diag.h> 33 34 static const struct inet_diag_handler **inet_diag_table; 35 36 struct inet_diag_entry { 37 const __be32 *saddr; 38 const __be32 *daddr; 39 u16 sport; 40 u16 dport; 41 u16 family; 42 u16 userlocks; 43 u32 ifindex; 44 u32 mark; 45 }; 46 47 static DEFINE_MUTEX(inet_diag_table_mutex); 48 49 static const struct inet_diag_handler *inet_diag_lock_handler(int proto) 50 { 51 if (!inet_diag_table[proto]) 52 sock_load_diag_module(AF_INET, proto); 53 54 mutex_lock(&inet_diag_table_mutex); 55 if (!inet_diag_table[proto]) 56 return ERR_PTR(-ENOENT); 57 58 return inet_diag_table[proto]; 59 } 60 61 static void inet_diag_unlock_handler(const struct inet_diag_handler *handler) 62 { 63 mutex_unlock(&inet_diag_table_mutex); 64 } 65 66 void inet_diag_msg_common_fill(struct inet_diag_msg *r, struct sock *sk) 67 { 68 r->idiag_family = sk->sk_family; 69 70 r->id.idiag_sport = htons(sk->sk_num); 71 r->id.idiag_dport = sk->sk_dport; 72 r->id.idiag_if = sk->sk_bound_dev_if; 73 sock_diag_save_cookie(sk, r->id.idiag_cookie); 74 75 #if IS_ENABLED(CONFIG_IPV6) 76 if (sk->sk_family == AF_INET6) { 77 *(struct in6_addr *)r->id.idiag_src = sk->sk_v6_rcv_saddr; 78 *(struct in6_addr *)r->id.idiag_dst = sk->sk_v6_daddr; 79 } else 80 #endif 81 { 82 memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src)); 83 memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst)); 84 85 r->id.idiag_src[0] = sk->sk_rcv_saddr; 86 r->id.idiag_dst[0] = sk->sk_daddr; 87 } 88 } 89 EXPORT_SYMBOL_GPL(inet_diag_msg_common_fill); 90 91 static size_t inet_sk_attr_size(struct sock *sk, 92 const struct inet_diag_req_v2 *req, 93 bool net_admin) 94 { 95 const struct inet_diag_handler *handler; 96 size_t aux = 0; 97 98 handler = inet_diag_table[req->sdiag_protocol]; 99 if (handler && handler->idiag_get_aux_size) 100 aux = handler->idiag_get_aux_size(sk, net_admin); 101 102 return nla_total_size(sizeof(struct tcp_info)) 103 + nla_total_size(sizeof(struct inet_diag_msg)) 104 + inet_diag_msg_attrs_size() 105 + nla_total_size(sizeof(struct inet_diag_meminfo)) 106 + nla_total_size(SK_MEMINFO_VARS * sizeof(u32)) 107 + nla_total_size(TCP_CA_NAME_MAX) 108 + nla_total_size(sizeof(struct tcpvegas_info)) 109 + aux 110 + 64; 111 } 112 113 int inet_diag_msg_attrs_fill(struct sock *sk, struct sk_buff *skb, 114 struct inet_diag_msg *r, int ext, 115 struct user_namespace *user_ns, 116 bool net_admin) 117 { 118 const struct inet_sock *inet = inet_sk(sk); 119 120 if (nla_put_u8(skb, INET_DIAG_SHUTDOWN, sk->sk_shutdown)) 121 goto errout; 122 123 /* IPv6 dual-stack sockets use inet->tos for IPv4 connections, 124 * hence this needs to be included regardless of socket family. 125 */ 126 if (ext & (1 << (INET_DIAG_TOS - 1))) 127 if (nla_put_u8(skb, INET_DIAG_TOS, inet->tos) < 0) 128 goto errout; 129 130 #if IS_ENABLED(CONFIG_IPV6) 131 if (r->idiag_family == AF_INET6) { 132 if (ext & (1 << (INET_DIAG_TCLASS - 1))) 133 if (nla_put_u8(skb, INET_DIAG_TCLASS, 134 inet6_sk(sk)->tclass) < 0) 135 goto errout; 136 137 if (((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) && 138 nla_put_u8(skb, INET_DIAG_SKV6ONLY, ipv6_only_sock(sk))) 139 goto errout; 140 } 141 #endif 142 143 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK, sk->sk_mark)) 144 goto errout; 145 146 if (ext & (1 << (INET_DIAG_CLASS_ID - 1)) || 147 ext & (1 << (INET_DIAG_TCLASS - 1))) { 148 u32 classid = 0; 149 150 #ifdef CONFIG_SOCK_CGROUP_DATA 151 classid = sock_cgroup_classid(&sk->sk_cgrp_data); 152 #endif 153 /* Fallback to socket priority if class id isn't set. 154 * Classful qdiscs use it as direct reference to class. 155 * For cgroup2 classid is always zero. 156 */ 157 if (!classid) 158 classid = sk->sk_priority; 159 160 if (nla_put_u32(skb, INET_DIAG_CLASS_ID, classid)) 161 goto errout; 162 } 163 164 r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk)); 165 r->idiag_inode = sock_i_ino(sk); 166 167 return 0; 168 errout: 169 return 1; 170 } 171 EXPORT_SYMBOL_GPL(inet_diag_msg_attrs_fill); 172 173 int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk, 174 struct sk_buff *skb, const struct inet_diag_req_v2 *req, 175 struct user_namespace *user_ns, 176 u32 portid, u32 seq, u16 nlmsg_flags, 177 const struct nlmsghdr *unlh, 178 bool net_admin) 179 { 180 const struct tcp_congestion_ops *ca_ops; 181 const struct inet_diag_handler *handler; 182 int ext = req->idiag_ext; 183 struct inet_diag_msg *r; 184 struct nlmsghdr *nlh; 185 struct nlattr *attr; 186 void *info = NULL; 187 188 handler = inet_diag_table[req->sdiag_protocol]; 189 BUG_ON(!handler); 190 191 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r), 192 nlmsg_flags); 193 if (!nlh) 194 return -EMSGSIZE; 195 196 r = nlmsg_data(nlh); 197 BUG_ON(!sk_fullsock(sk)); 198 199 inet_diag_msg_common_fill(r, sk); 200 r->idiag_state = sk->sk_state; 201 r->idiag_timer = 0; 202 r->idiag_retrans = 0; 203 204 if (inet_diag_msg_attrs_fill(sk, skb, r, ext, user_ns, net_admin)) 205 goto errout; 206 207 if (ext & (1 << (INET_DIAG_MEMINFO - 1))) { 208 struct inet_diag_meminfo minfo = { 209 .idiag_rmem = sk_rmem_alloc_get(sk), 210 .idiag_wmem = READ_ONCE(sk->sk_wmem_queued), 211 .idiag_fmem = sk->sk_forward_alloc, 212 .idiag_tmem = sk_wmem_alloc_get(sk), 213 }; 214 215 if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0) 216 goto errout; 217 } 218 219 if (ext & (1 << (INET_DIAG_SKMEMINFO - 1))) 220 if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO)) 221 goto errout; 222 223 /* 224 * RAW sockets might have user-defined protocols assigned, 225 * so report the one supplied on socket creation. 226 */ 227 if (sk->sk_type == SOCK_RAW) { 228 if (nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol)) 229 goto errout; 230 } 231 232 if (!icsk) { 233 handler->idiag_get_info(sk, r, NULL); 234 goto out; 235 } 236 237 if (icsk->icsk_pending == ICSK_TIME_RETRANS || 238 icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT || 239 icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) { 240 r->idiag_timer = 1; 241 r->idiag_retrans = icsk->icsk_retransmits; 242 r->idiag_expires = 243 jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies); 244 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { 245 r->idiag_timer = 4; 246 r->idiag_retrans = icsk->icsk_probes_out; 247 r->idiag_expires = 248 jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies); 249 } else if (timer_pending(&sk->sk_timer)) { 250 r->idiag_timer = 2; 251 r->idiag_retrans = icsk->icsk_probes_out; 252 r->idiag_expires = 253 jiffies_delta_to_msecs(sk->sk_timer.expires - jiffies); 254 } else { 255 r->idiag_timer = 0; 256 r->idiag_expires = 0; 257 } 258 259 if ((ext & (1 << (INET_DIAG_INFO - 1))) && handler->idiag_info_size) { 260 attr = nla_reserve_64bit(skb, INET_DIAG_INFO, 261 handler->idiag_info_size, 262 INET_DIAG_PAD); 263 if (!attr) 264 goto errout; 265 266 info = nla_data(attr); 267 } 268 269 if (ext & (1 << (INET_DIAG_CONG - 1))) { 270 int err = 0; 271 272 rcu_read_lock(); 273 ca_ops = READ_ONCE(icsk->icsk_ca_ops); 274 if (ca_ops) 275 err = nla_put_string(skb, INET_DIAG_CONG, ca_ops->name); 276 rcu_read_unlock(); 277 if (err < 0) 278 goto errout; 279 } 280 281 handler->idiag_get_info(sk, r, info); 282 283 if (ext & (1 << (INET_DIAG_INFO - 1)) && handler->idiag_get_aux) 284 if (handler->idiag_get_aux(sk, net_admin, skb) < 0) 285 goto errout; 286 287 if (sk->sk_state < TCP_TIME_WAIT) { 288 union tcp_cc_info info; 289 size_t sz = 0; 290 int attr; 291 292 rcu_read_lock(); 293 ca_ops = READ_ONCE(icsk->icsk_ca_ops); 294 if (ca_ops && ca_ops->get_info) 295 sz = ca_ops->get_info(sk, ext, &attr, &info); 296 rcu_read_unlock(); 297 if (sz && nla_put(skb, attr, sz, &info) < 0) 298 goto errout; 299 } 300 301 out: 302 nlmsg_end(skb, nlh); 303 return 0; 304 305 errout: 306 nlmsg_cancel(skb, nlh); 307 return -EMSGSIZE; 308 } 309 EXPORT_SYMBOL_GPL(inet_sk_diag_fill); 310 311 static int inet_csk_diag_fill(struct sock *sk, 312 struct sk_buff *skb, 313 const struct inet_diag_req_v2 *req, 314 struct user_namespace *user_ns, 315 u32 portid, u32 seq, u16 nlmsg_flags, 316 const struct nlmsghdr *unlh, 317 bool net_admin) 318 { 319 return inet_sk_diag_fill(sk, inet_csk(sk), skb, req, user_ns, 320 portid, seq, nlmsg_flags, unlh, net_admin); 321 } 322 323 static int inet_twsk_diag_fill(struct sock *sk, 324 struct sk_buff *skb, 325 u32 portid, u32 seq, u16 nlmsg_flags, 326 const struct nlmsghdr *unlh) 327 { 328 struct inet_timewait_sock *tw = inet_twsk(sk); 329 struct inet_diag_msg *r; 330 struct nlmsghdr *nlh; 331 long tmo; 332 333 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r), 334 nlmsg_flags); 335 if (!nlh) 336 return -EMSGSIZE; 337 338 r = nlmsg_data(nlh); 339 BUG_ON(tw->tw_state != TCP_TIME_WAIT); 340 341 inet_diag_msg_common_fill(r, sk); 342 r->idiag_retrans = 0; 343 344 r->idiag_state = tw->tw_substate; 345 r->idiag_timer = 3; 346 tmo = tw->tw_timer.expires - jiffies; 347 r->idiag_expires = jiffies_delta_to_msecs(tmo); 348 r->idiag_rqueue = 0; 349 r->idiag_wqueue = 0; 350 r->idiag_uid = 0; 351 r->idiag_inode = 0; 352 353 nlmsg_end(skb, nlh); 354 return 0; 355 } 356 357 static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb, 358 u32 portid, u32 seq, u16 nlmsg_flags, 359 const struct nlmsghdr *unlh, bool net_admin) 360 { 361 struct request_sock *reqsk = inet_reqsk(sk); 362 struct inet_diag_msg *r; 363 struct nlmsghdr *nlh; 364 long tmo; 365 366 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r), 367 nlmsg_flags); 368 if (!nlh) 369 return -EMSGSIZE; 370 371 r = nlmsg_data(nlh); 372 inet_diag_msg_common_fill(r, sk); 373 r->idiag_state = TCP_SYN_RECV; 374 r->idiag_timer = 1; 375 r->idiag_retrans = reqsk->num_retrans; 376 377 BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) != 378 offsetof(struct sock, sk_cookie)); 379 380 tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies; 381 r->idiag_expires = jiffies_delta_to_msecs(tmo); 382 r->idiag_rqueue = 0; 383 r->idiag_wqueue = 0; 384 r->idiag_uid = 0; 385 r->idiag_inode = 0; 386 387 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK, 388 inet_rsk(reqsk)->ir_mark)) 389 return -EMSGSIZE; 390 391 nlmsg_end(skb, nlh); 392 return 0; 393 } 394 395 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb, 396 const struct inet_diag_req_v2 *r, 397 struct user_namespace *user_ns, 398 u32 portid, u32 seq, u16 nlmsg_flags, 399 const struct nlmsghdr *unlh, bool net_admin) 400 { 401 if (sk->sk_state == TCP_TIME_WAIT) 402 return inet_twsk_diag_fill(sk, skb, portid, seq, 403 nlmsg_flags, unlh); 404 405 if (sk->sk_state == TCP_NEW_SYN_RECV) 406 return inet_req_diag_fill(sk, skb, portid, seq, 407 nlmsg_flags, unlh, net_admin); 408 409 return inet_csk_diag_fill(sk, skb, r, user_ns, portid, seq, 410 nlmsg_flags, unlh, net_admin); 411 } 412 413 struct sock *inet_diag_find_one_icsk(struct net *net, 414 struct inet_hashinfo *hashinfo, 415 const struct inet_diag_req_v2 *req) 416 { 417 struct sock *sk; 418 419 rcu_read_lock(); 420 if (req->sdiag_family == AF_INET) 421 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0], 422 req->id.idiag_dport, req->id.idiag_src[0], 423 req->id.idiag_sport, req->id.idiag_if); 424 #if IS_ENABLED(CONFIG_IPV6) 425 else if (req->sdiag_family == AF_INET6) { 426 if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) && 427 ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src)) 428 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3], 429 req->id.idiag_dport, req->id.idiag_src[3], 430 req->id.idiag_sport, req->id.idiag_if); 431 else 432 sk = inet6_lookup(net, hashinfo, NULL, 0, 433 (struct in6_addr *)req->id.idiag_dst, 434 req->id.idiag_dport, 435 (struct in6_addr *)req->id.idiag_src, 436 req->id.idiag_sport, 437 req->id.idiag_if); 438 } 439 #endif 440 else { 441 rcu_read_unlock(); 442 return ERR_PTR(-EINVAL); 443 } 444 rcu_read_unlock(); 445 if (!sk) 446 return ERR_PTR(-ENOENT); 447 448 if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) { 449 sock_gen_put(sk); 450 return ERR_PTR(-ENOENT); 451 } 452 453 return sk; 454 } 455 EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk); 456 457 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo, 458 struct sk_buff *in_skb, 459 const struct nlmsghdr *nlh, 460 const struct inet_diag_req_v2 *req) 461 { 462 bool net_admin = netlink_net_capable(in_skb, CAP_NET_ADMIN); 463 struct net *net = sock_net(in_skb->sk); 464 struct sk_buff *rep; 465 struct sock *sk; 466 int err; 467 468 sk = inet_diag_find_one_icsk(net, hashinfo, req); 469 if (IS_ERR(sk)) 470 return PTR_ERR(sk); 471 472 rep = nlmsg_new(inet_sk_attr_size(sk, req, net_admin), GFP_KERNEL); 473 if (!rep) { 474 err = -ENOMEM; 475 goto out; 476 } 477 478 err = sk_diag_fill(sk, rep, req, 479 sk_user_ns(NETLINK_CB(in_skb).sk), 480 NETLINK_CB(in_skb).portid, 481 nlh->nlmsg_seq, 0, nlh, net_admin); 482 if (err < 0) { 483 WARN_ON(err == -EMSGSIZE); 484 nlmsg_free(rep); 485 goto out; 486 } 487 err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid, 488 MSG_DONTWAIT); 489 if (err > 0) 490 err = 0; 491 492 out: 493 if (sk) 494 sock_gen_put(sk); 495 496 return err; 497 } 498 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk); 499 500 static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb, 501 const struct nlmsghdr *nlh, 502 const struct inet_diag_req_v2 *req) 503 { 504 const struct inet_diag_handler *handler; 505 int err; 506 507 handler = inet_diag_lock_handler(req->sdiag_protocol); 508 if (IS_ERR(handler)) 509 err = PTR_ERR(handler); 510 else if (cmd == SOCK_DIAG_BY_FAMILY) 511 err = handler->dump_one(in_skb, nlh, req); 512 else if (cmd == SOCK_DESTROY && handler->destroy) 513 err = handler->destroy(in_skb, req); 514 else 515 err = -EOPNOTSUPP; 516 inet_diag_unlock_handler(handler); 517 518 return err; 519 } 520 521 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits) 522 { 523 int words = bits >> 5; 524 525 bits &= 0x1f; 526 527 if (words) { 528 if (memcmp(a1, a2, words << 2)) 529 return 0; 530 } 531 if (bits) { 532 __be32 w1, w2; 533 __be32 mask; 534 535 w1 = a1[words]; 536 w2 = a2[words]; 537 538 mask = htonl((0xffffffff) << (32 - bits)); 539 540 if ((w1 ^ w2) & mask) 541 return 0; 542 } 543 544 return 1; 545 } 546 547 static int inet_diag_bc_run(const struct nlattr *_bc, 548 const struct inet_diag_entry *entry) 549 { 550 const void *bc = nla_data(_bc); 551 int len = nla_len(_bc); 552 553 while (len > 0) { 554 int yes = 1; 555 const struct inet_diag_bc_op *op = bc; 556 557 switch (op->code) { 558 case INET_DIAG_BC_NOP: 559 break; 560 case INET_DIAG_BC_JMP: 561 yes = 0; 562 break; 563 case INET_DIAG_BC_S_EQ: 564 yes = entry->sport == op[1].no; 565 break; 566 case INET_DIAG_BC_S_GE: 567 yes = entry->sport >= op[1].no; 568 break; 569 case INET_DIAG_BC_S_LE: 570 yes = entry->sport <= op[1].no; 571 break; 572 case INET_DIAG_BC_D_EQ: 573 yes = entry->dport == op[1].no; 574 break; 575 case INET_DIAG_BC_D_GE: 576 yes = entry->dport >= op[1].no; 577 break; 578 case INET_DIAG_BC_D_LE: 579 yes = entry->dport <= op[1].no; 580 break; 581 case INET_DIAG_BC_AUTO: 582 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK); 583 break; 584 case INET_DIAG_BC_S_COND: 585 case INET_DIAG_BC_D_COND: { 586 const struct inet_diag_hostcond *cond; 587 const __be32 *addr; 588 589 cond = (const struct inet_diag_hostcond *)(op + 1); 590 if (cond->port != -1 && 591 cond->port != (op->code == INET_DIAG_BC_S_COND ? 592 entry->sport : entry->dport)) { 593 yes = 0; 594 break; 595 } 596 597 if (op->code == INET_DIAG_BC_S_COND) 598 addr = entry->saddr; 599 else 600 addr = entry->daddr; 601 602 if (cond->family != AF_UNSPEC && 603 cond->family != entry->family) { 604 if (entry->family == AF_INET6 && 605 cond->family == AF_INET) { 606 if (addr[0] == 0 && addr[1] == 0 && 607 addr[2] == htonl(0xffff) && 608 bitstring_match(addr + 3, 609 cond->addr, 610 cond->prefix_len)) 611 break; 612 } 613 yes = 0; 614 break; 615 } 616 617 if (cond->prefix_len == 0) 618 break; 619 if (bitstring_match(addr, cond->addr, 620 cond->prefix_len)) 621 break; 622 yes = 0; 623 break; 624 } 625 case INET_DIAG_BC_DEV_COND: { 626 u32 ifindex; 627 628 ifindex = *((const u32 *)(op + 1)); 629 if (ifindex != entry->ifindex) 630 yes = 0; 631 break; 632 } 633 case INET_DIAG_BC_MARK_COND: { 634 struct inet_diag_markcond *cond; 635 636 cond = (struct inet_diag_markcond *)(op + 1); 637 if ((entry->mark & cond->mask) != cond->mark) 638 yes = 0; 639 break; 640 } 641 } 642 643 if (yes) { 644 len -= op->yes; 645 bc += op->yes; 646 } else { 647 len -= op->no; 648 bc += op->no; 649 } 650 } 651 return len == 0; 652 } 653 654 /* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV) 655 */ 656 static void entry_fill_addrs(struct inet_diag_entry *entry, 657 const struct sock *sk) 658 { 659 #if IS_ENABLED(CONFIG_IPV6) 660 if (sk->sk_family == AF_INET6) { 661 entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32; 662 entry->daddr = sk->sk_v6_daddr.s6_addr32; 663 } else 664 #endif 665 { 666 entry->saddr = &sk->sk_rcv_saddr; 667 entry->daddr = &sk->sk_daddr; 668 } 669 } 670 671 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk) 672 { 673 struct inet_sock *inet = inet_sk(sk); 674 struct inet_diag_entry entry; 675 676 if (!bc) 677 return 1; 678 679 entry.family = sk->sk_family; 680 entry_fill_addrs(&entry, sk); 681 entry.sport = inet->inet_num; 682 entry.dport = ntohs(inet->inet_dport); 683 entry.ifindex = sk->sk_bound_dev_if; 684 entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0; 685 if (sk_fullsock(sk)) 686 entry.mark = sk->sk_mark; 687 else if (sk->sk_state == TCP_NEW_SYN_RECV) 688 entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark; 689 else 690 entry.mark = 0; 691 692 return inet_diag_bc_run(bc, &entry); 693 } 694 EXPORT_SYMBOL_GPL(inet_diag_bc_sk); 695 696 static int valid_cc(const void *bc, int len, int cc) 697 { 698 while (len >= 0) { 699 const struct inet_diag_bc_op *op = bc; 700 701 if (cc > len) 702 return 0; 703 if (cc == len) 704 return 1; 705 if (op->yes < 4 || op->yes & 3) 706 return 0; 707 len -= op->yes; 708 bc += op->yes; 709 } 710 return 0; 711 } 712 713 /* data is u32 ifindex */ 714 static bool valid_devcond(const struct inet_diag_bc_op *op, int len, 715 int *min_len) 716 { 717 /* Check ifindex space. */ 718 *min_len += sizeof(u32); 719 if (len < *min_len) 720 return false; 721 722 return true; 723 } 724 /* Validate an inet_diag_hostcond. */ 725 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len, 726 int *min_len) 727 { 728 struct inet_diag_hostcond *cond; 729 int addr_len; 730 731 /* Check hostcond space. */ 732 *min_len += sizeof(struct inet_diag_hostcond); 733 if (len < *min_len) 734 return false; 735 cond = (struct inet_diag_hostcond *)(op + 1); 736 737 /* Check address family and address length. */ 738 switch (cond->family) { 739 case AF_UNSPEC: 740 addr_len = 0; 741 break; 742 case AF_INET: 743 addr_len = sizeof(struct in_addr); 744 break; 745 case AF_INET6: 746 addr_len = sizeof(struct in6_addr); 747 break; 748 default: 749 return false; 750 } 751 *min_len += addr_len; 752 if (len < *min_len) 753 return false; 754 755 /* Check prefix length (in bits) vs address length (in bytes). */ 756 if (cond->prefix_len > 8 * addr_len) 757 return false; 758 759 return true; 760 } 761 762 /* Validate a port comparison operator. */ 763 static bool valid_port_comparison(const struct inet_diag_bc_op *op, 764 int len, int *min_len) 765 { 766 /* Port comparisons put the port in a follow-on inet_diag_bc_op. */ 767 *min_len += sizeof(struct inet_diag_bc_op); 768 if (len < *min_len) 769 return false; 770 return true; 771 } 772 773 static bool valid_markcond(const struct inet_diag_bc_op *op, int len, 774 int *min_len) 775 { 776 *min_len += sizeof(struct inet_diag_markcond); 777 return len >= *min_len; 778 } 779 780 static int inet_diag_bc_audit(const struct nlattr *attr, 781 const struct sk_buff *skb) 782 { 783 bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN); 784 const void *bytecode, *bc; 785 int bytecode_len, len; 786 787 if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op)) 788 return -EINVAL; 789 790 bytecode = bc = nla_data(attr); 791 len = bytecode_len = nla_len(attr); 792 793 while (len > 0) { 794 int min_len = sizeof(struct inet_diag_bc_op); 795 const struct inet_diag_bc_op *op = bc; 796 797 switch (op->code) { 798 case INET_DIAG_BC_S_COND: 799 case INET_DIAG_BC_D_COND: 800 if (!valid_hostcond(bc, len, &min_len)) 801 return -EINVAL; 802 break; 803 case INET_DIAG_BC_DEV_COND: 804 if (!valid_devcond(bc, len, &min_len)) 805 return -EINVAL; 806 break; 807 case INET_DIAG_BC_S_EQ: 808 case INET_DIAG_BC_S_GE: 809 case INET_DIAG_BC_S_LE: 810 case INET_DIAG_BC_D_EQ: 811 case INET_DIAG_BC_D_GE: 812 case INET_DIAG_BC_D_LE: 813 if (!valid_port_comparison(bc, len, &min_len)) 814 return -EINVAL; 815 break; 816 case INET_DIAG_BC_MARK_COND: 817 if (!net_admin) 818 return -EPERM; 819 if (!valid_markcond(bc, len, &min_len)) 820 return -EINVAL; 821 break; 822 case INET_DIAG_BC_AUTO: 823 case INET_DIAG_BC_JMP: 824 case INET_DIAG_BC_NOP: 825 break; 826 default: 827 return -EINVAL; 828 } 829 830 if (op->code != INET_DIAG_BC_NOP) { 831 if (op->no < min_len || op->no > len + 4 || op->no & 3) 832 return -EINVAL; 833 if (op->no < len && 834 !valid_cc(bytecode, bytecode_len, len - op->no)) 835 return -EINVAL; 836 } 837 838 if (op->yes < min_len || op->yes > len + 4 || op->yes & 3) 839 return -EINVAL; 840 bc += op->yes; 841 len -= op->yes; 842 } 843 return len == 0 ? 0 : -EINVAL; 844 } 845 846 static int inet_csk_diag_dump(struct sock *sk, 847 struct sk_buff *skb, 848 struct netlink_callback *cb, 849 const struct inet_diag_req_v2 *r, 850 const struct nlattr *bc, 851 bool net_admin) 852 { 853 if (!inet_diag_bc_sk(bc, sk)) 854 return 0; 855 856 return inet_csk_diag_fill(sk, skb, r, 857 sk_user_ns(NETLINK_CB(cb->skb).sk), 858 NETLINK_CB(cb->skb).portid, 859 cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh, 860 net_admin); 861 } 862 863 static void twsk_build_assert(void) 864 { 865 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) != 866 offsetof(struct sock, sk_family)); 867 868 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) != 869 offsetof(struct inet_sock, inet_num)); 870 871 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) != 872 offsetof(struct inet_sock, inet_dport)); 873 874 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) != 875 offsetof(struct inet_sock, inet_rcv_saddr)); 876 877 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) != 878 offsetof(struct inet_sock, inet_daddr)); 879 880 #if IS_ENABLED(CONFIG_IPV6) 881 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) != 882 offsetof(struct sock, sk_v6_rcv_saddr)); 883 884 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) != 885 offsetof(struct sock, sk_v6_daddr)); 886 #endif 887 } 888 889 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb, 890 struct netlink_callback *cb, 891 const struct inet_diag_req_v2 *r, struct nlattr *bc) 892 { 893 bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN); 894 struct net *net = sock_net(skb->sk); 895 u32 idiag_states = r->idiag_states; 896 int i, num, s_i, s_num; 897 struct sock *sk; 898 899 if (idiag_states & TCPF_SYN_RECV) 900 idiag_states |= TCPF_NEW_SYN_RECV; 901 s_i = cb->args[1]; 902 s_num = num = cb->args[2]; 903 904 if (cb->args[0] == 0) { 905 if (!(idiag_states & TCPF_LISTEN) || r->id.idiag_dport) 906 goto skip_listen_ht; 907 908 for (i = s_i; i < INET_LHTABLE_SIZE; i++) { 909 struct inet_listen_hashbucket *ilb; 910 struct hlist_nulls_node *node; 911 912 num = 0; 913 ilb = &hashinfo->listening_hash[i]; 914 spin_lock(&ilb->lock); 915 sk_nulls_for_each(sk, node, &ilb->nulls_head) { 916 struct inet_sock *inet = inet_sk(sk); 917 918 if (!net_eq(sock_net(sk), net)) 919 continue; 920 921 if (num < s_num) { 922 num++; 923 continue; 924 } 925 926 if (r->sdiag_family != AF_UNSPEC && 927 sk->sk_family != r->sdiag_family) 928 goto next_listen; 929 930 if (r->id.idiag_sport != inet->inet_sport && 931 r->id.idiag_sport) 932 goto next_listen; 933 934 if (inet_csk_diag_dump(sk, skb, cb, r, 935 bc, net_admin) < 0) { 936 spin_unlock(&ilb->lock); 937 goto done; 938 } 939 940 next_listen: 941 ++num; 942 } 943 spin_unlock(&ilb->lock); 944 945 s_num = 0; 946 } 947 skip_listen_ht: 948 cb->args[0] = 1; 949 s_i = num = s_num = 0; 950 } 951 952 if (!(idiag_states & ~TCPF_LISTEN)) 953 goto out; 954 955 #define SKARR_SZ 16 956 for (i = s_i; i <= hashinfo->ehash_mask; i++) { 957 struct inet_ehash_bucket *head = &hashinfo->ehash[i]; 958 spinlock_t *lock = inet_ehash_lockp(hashinfo, i); 959 struct hlist_nulls_node *node; 960 struct sock *sk_arr[SKARR_SZ]; 961 int num_arr[SKARR_SZ]; 962 int idx, accum, res; 963 964 if (hlist_nulls_empty(&head->chain)) 965 continue; 966 967 if (i > s_i) 968 s_num = 0; 969 970 next_chunk: 971 num = 0; 972 accum = 0; 973 spin_lock_bh(lock); 974 sk_nulls_for_each(sk, node, &head->chain) { 975 int state; 976 977 if (!net_eq(sock_net(sk), net)) 978 continue; 979 if (num < s_num) 980 goto next_normal; 981 state = (sk->sk_state == TCP_TIME_WAIT) ? 982 inet_twsk(sk)->tw_substate : sk->sk_state; 983 if (!(idiag_states & (1 << state))) 984 goto next_normal; 985 if (r->sdiag_family != AF_UNSPEC && 986 sk->sk_family != r->sdiag_family) 987 goto next_normal; 988 if (r->id.idiag_sport != htons(sk->sk_num) && 989 r->id.idiag_sport) 990 goto next_normal; 991 if (r->id.idiag_dport != sk->sk_dport && 992 r->id.idiag_dport) 993 goto next_normal; 994 twsk_build_assert(); 995 996 if (!inet_diag_bc_sk(bc, sk)) 997 goto next_normal; 998 999 if (!refcount_inc_not_zero(&sk->sk_refcnt)) 1000 goto next_normal; 1001 1002 num_arr[accum] = num; 1003 sk_arr[accum] = sk; 1004 if (++accum == SKARR_SZ) 1005 break; 1006 next_normal: 1007 ++num; 1008 } 1009 spin_unlock_bh(lock); 1010 res = 0; 1011 for (idx = 0; idx < accum; idx++) { 1012 if (res >= 0) { 1013 res = sk_diag_fill(sk_arr[idx], skb, r, 1014 sk_user_ns(NETLINK_CB(cb->skb).sk), 1015 NETLINK_CB(cb->skb).portid, 1016 cb->nlh->nlmsg_seq, NLM_F_MULTI, 1017 cb->nlh, net_admin); 1018 if (res < 0) 1019 num = num_arr[idx]; 1020 } 1021 sock_gen_put(sk_arr[idx]); 1022 } 1023 if (res < 0) 1024 break; 1025 cond_resched(); 1026 if (accum == SKARR_SZ) { 1027 s_num = num + 1; 1028 goto next_chunk; 1029 } 1030 } 1031 1032 done: 1033 cb->args[1] = i; 1034 cb->args[2] = num; 1035 out: 1036 ; 1037 } 1038 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk); 1039 1040 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb, 1041 const struct inet_diag_req_v2 *r, 1042 struct nlattr *bc) 1043 { 1044 const struct inet_diag_handler *handler; 1045 int err = 0; 1046 1047 handler = inet_diag_lock_handler(r->sdiag_protocol); 1048 if (!IS_ERR(handler)) 1049 handler->dump(skb, cb, r, bc); 1050 else 1051 err = PTR_ERR(handler); 1052 inet_diag_unlock_handler(handler); 1053 1054 return err ? : skb->len; 1055 } 1056 1057 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb) 1058 { 1059 int hdrlen = sizeof(struct inet_diag_req_v2); 1060 struct nlattr *bc = NULL; 1061 1062 if (nlmsg_attrlen(cb->nlh, hdrlen)) 1063 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE); 1064 1065 return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh), bc); 1066 } 1067 1068 static int inet_diag_type2proto(int type) 1069 { 1070 switch (type) { 1071 case TCPDIAG_GETSOCK: 1072 return IPPROTO_TCP; 1073 case DCCPDIAG_GETSOCK: 1074 return IPPROTO_DCCP; 1075 default: 1076 return 0; 1077 } 1078 } 1079 1080 static int inet_diag_dump_compat(struct sk_buff *skb, 1081 struct netlink_callback *cb) 1082 { 1083 struct inet_diag_req *rc = nlmsg_data(cb->nlh); 1084 int hdrlen = sizeof(struct inet_diag_req); 1085 struct inet_diag_req_v2 req; 1086 struct nlattr *bc = NULL; 1087 1088 req.sdiag_family = AF_UNSPEC; /* compatibility */ 1089 req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type); 1090 req.idiag_ext = rc->idiag_ext; 1091 req.idiag_states = rc->idiag_states; 1092 req.id = rc->id; 1093 1094 if (nlmsg_attrlen(cb->nlh, hdrlen)) 1095 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE); 1096 1097 return __inet_diag_dump(skb, cb, &req, bc); 1098 } 1099 1100 static int inet_diag_get_exact_compat(struct sk_buff *in_skb, 1101 const struct nlmsghdr *nlh) 1102 { 1103 struct inet_diag_req *rc = nlmsg_data(nlh); 1104 struct inet_diag_req_v2 req; 1105 1106 req.sdiag_family = rc->idiag_family; 1107 req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type); 1108 req.idiag_ext = rc->idiag_ext; 1109 req.idiag_states = rc->idiag_states; 1110 req.id = rc->id; 1111 1112 return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh, &req); 1113 } 1114 1115 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh) 1116 { 1117 int hdrlen = sizeof(struct inet_diag_req); 1118 struct net *net = sock_net(skb->sk); 1119 1120 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX || 1121 nlmsg_len(nlh) < hdrlen) 1122 return -EINVAL; 1123 1124 if (nlh->nlmsg_flags & NLM_F_DUMP) { 1125 if (nlmsg_attrlen(nlh, hdrlen)) { 1126 struct nlattr *attr; 1127 int err; 1128 1129 attr = nlmsg_find_attr(nlh, hdrlen, 1130 INET_DIAG_REQ_BYTECODE); 1131 err = inet_diag_bc_audit(attr, skb); 1132 if (err) 1133 return err; 1134 } 1135 { 1136 struct netlink_dump_control c = { 1137 .dump = inet_diag_dump_compat, 1138 }; 1139 return netlink_dump_start(net->diag_nlsk, skb, nlh, &c); 1140 } 1141 } 1142 1143 return inet_diag_get_exact_compat(skb, nlh); 1144 } 1145 1146 static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h) 1147 { 1148 int hdrlen = sizeof(struct inet_diag_req_v2); 1149 struct net *net = sock_net(skb->sk); 1150 1151 if (nlmsg_len(h) < hdrlen) 1152 return -EINVAL; 1153 1154 if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY && 1155 h->nlmsg_flags & NLM_F_DUMP) { 1156 if (nlmsg_attrlen(h, hdrlen)) { 1157 struct nlattr *attr; 1158 int err; 1159 1160 attr = nlmsg_find_attr(h, hdrlen, 1161 INET_DIAG_REQ_BYTECODE); 1162 err = inet_diag_bc_audit(attr, skb); 1163 if (err) 1164 return err; 1165 } 1166 { 1167 struct netlink_dump_control c = { 1168 .dump = inet_diag_dump, 1169 }; 1170 return netlink_dump_start(net->diag_nlsk, skb, h, &c); 1171 } 1172 } 1173 1174 return inet_diag_cmd_exact(h->nlmsg_type, skb, h, nlmsg_data(h)); 1175 } 1176 1177 static 1178 int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk) 1179 { 1180 const struct inet_diag_handler *handler; 1181 struct nlmsghdr *nlh; 1182 struct nlattr *attr; 1183 struct inet_diag_msg *r; 1184 void *info = NULL; 1185 int err = 0; 1186 1187 nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0); 1188 if (!nlh) 1189 return -ENOMEM; 1190 1191 r = nlmsg_data(nlh); 1192 memset(r, 0, sizeof(*r)); 1193 inet_diag_msg_common_fill(r, sk); 1194 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM) 1195 r->id.idiag_sport = inet_sk(sk)->inet_sport; 1196 r->idiag_state = sk->sk_state; 1197 1198 if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) { 1199 nlmsg_cancel(skb, nlh); 1200 return err; 1201 } 1202 1203 handler = inet_diag_lock_handler(sk->sk_protocol); 1204 if (IS_ERR(handler)) { 1205 inet_diag_unlock_handler(handler); 1206 nlmsg_cancel(skb, nlh); 1207 return PTR_ERR(handler); 1208 } 1209 1210 attr = handler->idiag_info_size 1211 ? nla_reserve_64bit(skb, INET_DIAG_INFO, 1212 handler->idiag_info_size, 1213 INET_DIAG_PAD) 1214 : NULL; 1215 if (attr) 1216 info = nla_data(attr); 1217 1218 handler->idiag_get_info(sk, r, info); 1219 inet_diag_unlock_handler(handler); 1220 1221 nlmsg_end(skb, nlh); 1222 return 0; 1223 } 1224 1225 static const struct sock_diag_handler inet_diag_handler = { 1226 .family = AF_INET, 1227 .dump = inet_diag_handler_cmd, 1228 .get_info = inet_diag_handler_get_info, 1229 .destroy = inet_diag_handler_cmd, 1230 }; 1231 1232 static const struct sock_diag_handler inet6_diag_handler = { 1233 .family = AF_INET6, 1234 .dump = inet_diag_handler_cmd, 1235 .get_info = inet_diag_handler_get_info, 1236 .destroy = inet_diag_handler_cmd, 1237 }; 1238 1239 int inet_diag_register(const struct inet_diag_handler *h) 1240 { 1241 const __u16 type = h->idiag_type; 1242 int err = -EINVAL; 1243 1244 if (type >= IPPROTO_MAX) 1245 goto out; 1246 1247 mutex_lock(&inet_diag_table_mutex); 1248 err = -EEXIST; 1249 if (!inet_diag_table[type]) { 1250 inet_diag_table[type] = h; 1251 err = 0; 1252 } 1253 mutex_unlock(&inet_diag_table_mutex); 1254 out: 1255 return err; 1256 } 1257 EXPORT_SYMBOL_GPL(inet_diag_register); 1258 1259 void inet_diag_unregister(const struct inet_diag_handler *h) 1260 { 1261 const __u16 type = h->idiag_type; 1262 1263 if (type >= IPPROTO_MAX) 1264 return; 1265 1266 mutex_lock(&inet_diag_table_mutex); 1267 inet_diag_table[type] = NULL; 1268 mutex_unlock(&inet_diag_table_mutex); 1269 } 1270 EXPORT_SYMBOL_GPL(inet_diag_unregister); 1271 1272 static int __init inet_diag_init(void) 1273 { 1274 const int inet_diag_table_size = (IPPROTO_MAX * 1275 sizeof(struct inet_diag_handler *)); 1276 int err = -ENOMEM; 1277 1278 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL); 1279 if (!inet_diag_table) 1280 goto out; 1281 1282 err = sock_diag_register(&inet_diag_handler); 1283 if (err) 1284 goto out_free_nl; 1285 1286 err = sock_diag_register(&inet6_diag_handler); 1287 if (err) 1288 goto out_free_inet; 1289 1290 sock_diag_register_inet_compat(inet_diag_rcv_msg_compat); 1291 out: 1292 return err; 1293 1294 out_free_inet: 1295 sock_diag_unregister(&inet_diag_handler); 1296 out_free_nl: 1297 kfree(inet_diag_table); 1298 goto out; 1299 } 1300 1301 static void __exit inet_diag_exit(void) 1302 { 1303 sock_diag_unregister(&inet6_diag_handler); 1304 sock_diag_unregister(&inet_diag_handler); 1305 sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat); 1306 kfree(inet_diag_table); 1307 } 1308 1309 module_init(inet_diag_init); 1310 module_exit(inet_diag_exit); 1311 MODULE_LICENSE("GPL"); 1312 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */); 1313 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */); 1314