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