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 union tcp_cc_info info; 228 size_t sz = 0; 229 int attr; 230 231 rcu_read_lock(); 232 ca_ops = READ_ONCE(icsk->icsk_ca_ops); 233 if (ca_ops && ca_ops->get_info) 234 sz = ca_ops->get_info(sk, ext, &attr, &info); 235 rcu_read_unlock(); 236 if (sz && nla_put(skb, attr, sz, &info) < 0) 237 goto errout; 238 } 239 240 out: 241 nlmsg_end(skb, nlh); 242 return 0; 243 244 errout: 245 nlmsg_cancel(skb, nlh); 246 return -EMSGSIZE; 247 } 248 EXPORT_SYMBOL_GPL(inet_sk_diag_fill); 249 250 static int inet_csk_diag_fill(struct sock *sk, 251 struct sk_buff *skb, 252 const struct inet_diag_req_v2 *req, 253 struct user_namespace *user_ns, 254 u32 portid, u32 seq, u16 nlmsg_flags, 255 const struct nlmsghdr *unlh) 256 { 257 return inet_sk_diag_fill(sk, inet_csk(sk), skb, req, 258 user_ns, portid, seq, nlmsg_flags, unlh); 259 } 260 261 static int inet_twsk_diag_fill(struct sock *sk, 262 struct sk_buff *skb, 263 u32 portid, u32 seq, u16 nlmsg_flags, 264 const struct nlmsghdr *unlh) 265 { 266 struct inet_timewait_sock *tw = inet_twsk(sk); 267 struct inet_diag_msg *r; 268 struct nlmsghdr *nlh; 269 long tmo; 270 271 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r), 272 nlmsg_flags); 273 if (!nlh) 274 return -EMSGSIZE; 275 276 r = nlmsg_data(nlh); 277 BUG_ON(tw->tw_state != TCP_TIME_WAIT); 278 279 tmo = tw->tw_timer.expires - jiffies; 280 if (tmo < 0) 281 tmo = 0; 282 283 inet_diag_msg_common_fill(r, sk); 284 r->idiag_retrans = 0; 285 286 r->idiag_state = tw->tw_substate; 287 r->idiag_timer = 3; 288 r->idiag_expires = jiffies_to_msecs(tmo); 289 r->idiag_rqueue = 0; 290 r->idiag_wqueue = 0; 291 r->idiag_uid = 0; 292 r->idiag_inode = 0; 293 294 nlmsg_end(skb, nlh); 295 return 0; 296 } 297 298 static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb, 299 u32 portid, u32 seq, u16 nlmsg_flags, 300 const struct nlmsghdr *unlh) 301 { 302 struct inet_diag_msg *r; 303 struct nlmsghdr *nlh; 304 long tmo; 305 306 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r), 307 nlmsg_flags); 308 if (!nlh) 309 return -EMSGSIZE; 310 311 r = nlmsg_data(nlh); 312 inet_diag_msg_common_fill(r, sk); 313 r->idiag_state = TCP_SYN_RECV; 314 r->idiag_timer = 1; 315 r->idiag_retrans = inet_reqsk(sk)->num_retrans; 316 317 BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) != 318 offsetof(struct sock, sk_cookie)); 319 320 tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies; 321 r->idiag_expires = (tmo >= 0) ? jiffies_to_msecs(tmo) : 0; 322 r->idiag_rqueue = 0; 323 r->idiag_wqueue = 0; 324 r->idiag_uid = 0; 325 r->idiag_inode = 0; 326 327 nlmsg_end(skb, nlh); 328 return 0; 329 } 330 331 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb, 332 const struct inet_diag_req_v2 *r, 333 struct user_namespace *user_ns, 334 u32 portid, u32 seq, u16 nlmsg_flags, 335 const struct nlmsghdr *unlh) 336 { 337 if (sk->sk_state == TCP_TIME_WAIT) 338 return inet_twsk_diag_fill(sk, skb, portid, seq, 339 nlmsg_flags, unlh); 340 341 if (sk->sk_state == TCP_NEW_SYN_RECV) 342 return inet_req_diag_fill(sk, skb, portid, seq, 343 nlmsg_flags, unlh); 344 345 return inet_csk_diag_fill(sk, skb, r, user_ns, portid, seq, 346 nlmsg_flags, unlh); 347 } 348 349 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo, 350 struct sk_buff *in_skb, 351 const struct nlmsghdr *nlh, 352 const struct inet_diag_req_v2 *req) 353 { 354 struct net *net = sock_net(in_skb->sk); 355 struct sk_buff *rep; 356 struct sock *sk; 357 int err; 358 359 err = -EINVAL; 360 if (req->sdiag_family == AF_INET) 361 sk = inet_lookup(net, hashinfo, req->id.idiag_dst[0], 362 req->id.idiag_dport, req->id.idiag_src[0], 363 req->id.idiag_sport, req->id.idiag_if); 364 #if IS_ENABLED(CONFIG_IPV6) 365 else if (req->sdiag_family == AF_INET6) 366 sk = inet6_lookup(net, hashinfo, 367 (struct in6_addr *)req->id.idiag_dst, 368 req->id.idiag_dport, 369 (struct in6_addr *)req->id.idiag_src, 370 req->id.idiag_sport, 371 req->id.idiag_if); 372 #endif 373 else 374 goto out_nosk; 375 376 err = -ENOENT; 377 if (!sk) 378 goto out_nosk; 379 380 err = sock_diag_check_cookie(sk, req->id.idiag_cookie); 381 if (err) 382 goto out; 383 384 rep = nlmsg_new(inet_sk_attr_size(), GFP_KERNEL); 385 if (!rep) { 386 err = -ENOMEM; 387 goto out; 388 } 389 390 err = sk_diag_fill(sk, rep, req, 391 sk_user_ns(NETLINK_CB(in_skb).sk), 392 NETLINK_CB(in_skb).portid, 393 nlh->nlmsg_seq, 0, nlh); 394 if (err < 0) { 395 WARN_ON(err == -EMSGSIZE); 396 nlmsg_free(rep); 397 goto out; 398 } 399 err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid, 400 MSG_DONTWAIT); 401 if (err > 0) 402 err = 0; 403 404 out: 405 if (sk) 406 sock_gen_put(sk); 407 408 out_nosk: 409 return err; 410 } 411 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk); 412 413 static int inet_diag_get_exact(struct sk_buff *in_skb, 414 const struct nlmsghdr *nlh, 415 const struct inet_diag_req_v2 *req) 416 { 417 const struct inet_diag_handler *handler; 418 int err; 419 420 handler = inet_diag_lock_handler(req->sdiag_protocol); 421 if (IS_ERR(handler)) 422 err = PTR_ERR(handler); 423 else 424 err = handler->dump_one(in_skb, nlh, req); 425 inet_diag_unlock_handler(handler); 426 427 return err; 428 } 429 430 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits) 431 { 432 int words = bits >> 5; 433 434 bits &= 0x1f; 435 436 if (words) { 437 if (memcmp(a1, a2, words << 2)) 438 return 0; 439 } 440 if (bits) { 441 __be32 w1, w2; 442 __be32 mask; 443 444 w1 = a1[words]; 445 w2 = a2[words]; 446 447 mask = htonl((0xffffffff) << (32 - bits)); 448 449 if ((w1 ^ w2) & mask) 450 return 0; 451 } 452 453 return 1; 454 } 455 456 static int inet_diag_bc_run(const struct nlattr *_bc, 457 const struct inet_diag_entry *entry) 458 { 459 const void *bc = nla_data(_bc); 460 int len = nla_len(_bc); 461 462 while (len > 0) { 463 int yes = 1; 464 const struct inet_diag_bc_op *op = bc; 465 466 switch (op->code) { 467 case INET_DIAG_BC_NOP: 468 break; 469 case INET_DIAG_BC_JMP: 470 yes = 0; 471 break; 472 case INET_DIAG_BC_S_GE: 473 yes = entry->sport >= op[1].no; 474 break; 475 case INET_DIAG_BC_S_LE: 476 yes = entry->sport <= op[1].no; 477 break; 478 case INET_DIAG_BC_D_GE: 479 yes = entry->dport >= op[1].no; 480 break; 481 case INET_DIAG_BC_D_LE: 482 yes = entry->dport <= op[1].no; 483 break; 484 case INET_DIAG_BC_AUTO: 485 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK); 486 break; 487 case INET_DIAG_BC_S_COND: 488 case INET_DIAG_BC_D_COND: { 489 const struct inet_diag_hostcond *cond; 490 const __be32 *addr; 491 492 cond = (const struct inet_diag_hostcond *)(op + 1); 493 if (cond->port != -1 && 494 cond->port != (op->code == INET_DIAG_BC_S_COND ? 495 entry->sport : entry->dport)) { 496 yes = 0; 497 break; 498 } 499 500 if (op->code == INET_DIAG_BC_S_COND) 501 addr = entry->saddr; 502 else 503 addr = entry->daddr; 504 505 if (cond->family != AF_UNSPEC && 506 cond->family != entry->family) { 507 if (entry->family == AF_INET6 && 508 cond->family == AF_INET) { 509 if (addr[0] == 0 && addr[1] == 0 && 510 addr[2] == htonl(0xffff) && 511 bitstring_match(addr + 3, 512 cond->addr, 513 cond->prefix_len)) 514 break; 515 } 516 yes = 0; 517 break; 518 } 519 520 if (cond->prefix_len == 0) 521 break; 522 if (bitstring_match(addr, cond->addr, 523 cond->prefix_len)) 524 break; 525 yes = 0; 526 break; 527 } 528 } 529 530 if (yes) { 531 len -= op->yes; 532 bc += op->yes; 533 } else { 534 len -= op->no; 535 bc += op->no; 536 } 537 } 538 return len == 0; 539 } 540 541 /* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV) 542 */ 543 static void entry_fill_addrs(struct inet_diag_entry *entry, 544 const struct sock *sk) 545 { 546 #if IS_ENABLED(CONFIG_IPV6) 547 if (sk->sk_family == AF_INET6) { 548 entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32; 549 entry->daddr = sk->sk_v6_daddr.s6_addr32; 550 } else 551 #endif 552 { 553 entry->saddr = &sk->sk_rcv_saddr; 554 entry->daddr = &sk->sk_daddr; 555 } 556 } 557 558 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk) 559 { 560 struct inet_sock *inet = inet_sk(sk); 561 struct inet_diag_entry entry; 562 563 if (!bc) 564 return 1; 565 566 entry.family = sk->sk_family; 567 entry_fill_addrs(&entry, sk); 568 entry.sport = inet->inet_num; 569 entry.dport = ntohs(inet->inet_dport); 570 entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0; 571 572 return inet_diag_bc_run(bc, &entry); 573 } 574 EXPORT_SYMBOL_GPL(inet_diag_bc_sk); 575 576 static int valid_cc(const void *bc, int len, int cc) 577 { 578 while (len >= 0) { 579 const struct inet_diag_bc_op *op = bc; 580 581 if (cc > len) 582 return 0; 583 if (cc == len) 584 return 1; 585 if (op->yes < 4 || op->yes & 3) 586 return 0; 587 len -= op->yes; 588 bc += op->yes; 589 } 590 return 0; 591 } 592 593 /* Validate an inet_diag_hostcond. */ 594 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len, 595 int *min_len) 596 { 597 struct inet_diag_hostcond *cond; 598 int addr_len; 599 600 /* Check hostcond space. */ 601 *min_len += sizeof(struct inet_diag_hostcond); 602 if (len < *min_len) 603 return false; 604 cond = (struct inet_diag_hostcond *)(op + 1); 605 606 /* Check address family and address length. */ 607 switch (cond->family) { 608 case AF_UNSPEC: 609 addr_len = 0; 610 break; 611 case AF_INET: 612 addr_len = sizeof(struct in_addr); 613 break; 614 case AF_INET6: 615 addr_len = sizeof(struct in6_addr); 616 break; 617 default: 618 return false; 619 } 620 *min_len += addr_len; 621 if (len < *min_len) 622 return false; 623 624 /* Check prefix length (in bits) vs address length (in bytes). */ 625 if (cond->prefix_len > 8 * addr_len) 626 return false; 627 628 return true; 629 } 630 631 /* Validate a port comparison operator. */ 632 static bool valid_port_comparison(const struct inet_diag_bc_op *op, 633 int len, int *min_len) 634 { 635 /* Port comparisons put the port in a follow-on inet_diag_bc_op. */ 636 *min_len += sizeof(struct inet_diag_bc_op); 637 if (len < *min_len) 638 return false; 639 return true; 640 } 641 642 static int inet_diag_bc_audit(const void *bytecode, int bytecode_len) 643 { 644 const void *bc = bytecode; 645 int len = bytecode_len; 646 647 while (len > 0) { 648 int min_len = sizeof(struct inet_diag_bc_op); 649 const struct inet_diag_bc_op *op = bc; 650 651 switch (op->code) { 652 case INET_DIAG_BC_S_COND: 653 case INET_DIAG_BC_D_COND: 654 if (!valid_hostcond(bc, len, &min_len)) 655 return -EINVAL; 656 break; 657 case INET_DIAG_BC_S_GE: 658 case INET_DIAG_BC_S_LE: 659 case INET_DIAG_BC_D_GE: 660 case INET_DIAG_BC_D_LE: 661 if (!valid_port_comparison(bc, len, &min_len)) 662 return -EINVAL; 663 break; 664 case INET_DIAG_BC_AUTO: 665 case INET_DIAG_BC_JMP: 666 case INET_DIAG_BC_NOP: 667 break; 668 default: 669 return -EINVAL; 670 } 671 672 if (op->code != INET_DIAG_BC_NOP) { 673 if (op->no < min_len || op->no > len + 4 || op->no & 3) 674 return -EINVAL; 675 if (op->no < len && 676 !valid_cc(bytecode, bytecode_len, len - op->no)) 677 return -EINVAL; 678 } 679 680 if (op->yes < min_len || op->yes > len + 4 || op->yes & 3) 681 return -EINVAL; 682 bc += op->yes; 683 len -= op->yes; 684 } 685 return len == 0 ? 0 : -EINVAL; 686 } 687 688 static int inet_csk_diag_dump(struct sock *sk, 689 struct sk_buff *skb, 690 struct netlink_callback *cb, 691 const struct inet_diag_req_v2 *r, 692 const struct nlattr *bc) 693 { 694 if (!inet_diag_bc_sk(bc, sk)) 695 return 0; 696 697 return inet_csk_diag_fill(sk, skb, r, 698 sk_user_ns(NETLINK_CB(cb->skb).sk), 699 NETLINK_CB(cb->skb).portid, 700 cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh); 701 } 702 703 static void twsk_build_assert(void) 704 { 705 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) != 706 offsetof(struct sock, sk_family)); 707 708 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) != 709 offsetof(struct inet_sock, inet_num)); 710 711 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) != 712 offsetof(struct inet_sock, inet_dport)); 713 714 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) != 715 offsetof(struct inet_sock, inet_rcv_saddr)); 716 717 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) != 718 offsetof(struct inet_sock, inet_daddr)); 719 720 #if IS_ENABLED(CONFIG_IPV6) 721 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) != 722 offsetof(struct sock, sk_v6_rcv_saddr)); 723 724 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) != 725 offsetof(struct sock, sk_v6_daddr)); 726 #endif 727 } 728 729 static int inet_diag_dump_reqs(struct sk_buff *skb, struct sock *sk, 730 struct netlink_callback *cb, 731 const struct inet_diag_req_v2 *r, 732 const struct nlattr *bc) 733 { 734 struct inet_connection_sock *icsk = inet_csk(sk); 735 struct inet_sock *inet = inet_sk(sk); 736 struct inet_diag_entry entry; 737 int j, s_j, reqnum, s_reqnum; 738 struct listen_sock *lopt; 739 int err = 0; 740 741 s_j = cb->args[3]; 742 s_reqnum = cb->args[4]; 743 744 if (s_j > 0) 745 s_j--; 746 747 entry.family = sk->sk_family; 748 749 spin_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock); 750 751 lopt = icsk->icsk_accept_queue.listen_opt; 752 if (!lopt || !listen_sock_qlen(lopt)) 753 goto out; 754 755 if (bc) { 756 entry.sport = inet->inet_num; 757 entry.userlocks = sk->sk_userlocks; 758 } 759 760 for (j = s_j; j < lopt->nr_table_entries; j++) { 761 struct request_sock *req, *head = lopt->syn_table[j]; 762 763 reqnum = 0; 764 for (req = head; req; reqnum++, req = req->dl_next) { 765 struct inet_request_sock *ireq = inet_rsk(req); 766 767 if (reqnum < s_reqnum) 768 continue; 769 if (r->id.idiag_dport != ireq->ir_rmt_port && 770 r->id.idiag_dport) 771 continue; 772 773 if (bc) { 774 /* Note: entry.sport and entry.userlocks are already set */ 775 entry_fill_addrs(&entry, req_to_sk(req)); 776 entry.dport = ntohs(ireq->ir_rmt_port); 777 778 if (!inet_diag_bc_run(bc, &entry)) 779 continue; 780 } 781 782 err = inet_req_diag_fill(req_to_sk(req), skb, 783 NETLINK_CB(cb->skb).portid, 784 cb->nlh->nlmsg_seq, 785 NLM_F_MULTI, cb->nlh); 786 if (err < 0) { 787 cb->args[3] = j + 1; 788 cb->args[4] = reqnum; 789 goto out; 790 } 791 } 792 793 s_reqnum = 0; 794 } 795 796 out: 797 spin_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); 798 799 return err; 800 } 801 802 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb, 803 struct netlink_callback *cb, 804 const struct inet_diag_req_v2 *r, struct nlattr *bc) 805 { 806 struct net *net = sock_net(skb->sk); 807 int i, num, s_i, s_num; 808 809 s_i = cb->args[1]; 810 s_num = num = cb->args[2]; 811 812 if (cb->args[0] == 0) { 813 if (!(r->idiag_states & (TCPF_LISTEN | TCPF_SYN_RECV))) 814 goto skip_listen_ht; 815 816 for (i = s_i; i < INET_LHTABLE_SIZE; i++) { 817 struct inet_listen_hashbucket *ilb; 818 struct hlist_nulls_node *node; 819 struct sock *sk; 820 821 num = 0; 822 ilb = &hashinfo->listening_hash[i]; 823 spin_lock_bh(&ilb->lock); 824 sk_nulls_for_each(sk, node, &ilb->head) { 825 struct inet_sock *inet = inet_sk(sk); 826 827 if (!net_eq(sock_net(sk), net)) 828 continue; 829 830 if (num < s_num) { 831 num++; 832 continue; 833 } 834 835 if (r->sdiag_family != AF_UNSPEC && 836 sk->sk_family != r->sdiag_family) 837 goto next_listen; 838 839 if (r->id.idiag_sport != inet->inet_sport && 840 r->id.idiag_sport) 841 goto next_listen; 842 843 if (!(r->idiag_states & TCPF_LISTEN) || 844 r->id.idiag_dport || 845 cb->args[3] > 0) 846 goto syn_recv; 847 848 if (inet_csk_diag_dump(sk, skb, cb, r, bc) < 0) { 849 spin_unlock_bh(&ilb->lock); 850 goto done; 851 } 852 853 syn_recv: 854 if (!(r->idiag_states & TCPF_SYN_RECV)) 855 goto next_listen; 856 857 if (inet_diag_dump_reqs(skb, sk, cb, r, bc) < 0) { 858 spin_unlock_bh(&ilb->lock); 859 goto done; 860 } 861 862 next_listen: 863 cb->args[3] = 0; 864 cb->args[4] = 0; 865 ++num; 866 } 867 spin_unlock_bh(&ilb->lock); 868 869 s_num = 0; 870 cb->args[3] = 0; 871 cb->args[4] = 0; 872 } 873 skip_listen_ht: 874 cb->args[0] = 1; 875 s_i = num = s_num = 0; 876 } 877 878 if (!(r->idiag_states & ~(TCPF_LISTEN | TCPF_SYN_RECV))) 879 goto out; 880 881 for (i = s_i; i <= hashinfo->ehash_mask; i++) { 882 struct inet_ehash_bucket *head = &hashinfo->ehash[i]; 883 spinlock_t *lock = inet_ehash_lockp(hashinfo, i); 884 struct hlist_nulls_node *node; 885 struct sock *sk; 886 887 num = 0; 888 889 if (hlist_nulls_empty(&head->chain)) 890 continue; 891 892 if (i > s_i) 893 s_num = 0; 894 895 spin_lock_bh(lock); 896 sk_nulls_for_each(sk, node, &head->chain) { 897 int state, res; 898 899 if (!net_eq(sock_net(sk), net)) 900 continue; 901 if (num < s_num) 902 goto next_normal; 903 state = (sk->sk_state == TCP_TIME_WAIT) ? 904 inet_twsk(sk)->tw_substate : sk->sk_state; 905 if (!(r->idiag_states & (1 << state))) 906 goto next_normal; 907 if (r->sdiag_family != AF_UNSPEC && 908 sk->sk_family != r->sdiag_family) 909 goto next_normal; 910 if (r->id.idiag_sport != htons(sk->sk_num) && 911 r->id.idiag_sport) 912 goto next_normal; 913 if (r->id.idiag_dport != sk->sk_dport && 914 r->id.idiag_dport) 915 goto next_normal; 916 twsk_build_assert(); 917 918 if (!inet_diag_bc_sk(bc, sk)) 919 goto next_normal; 920 921 res = sk_diag_fill(sk, skb, r, 922 sk_user_ns(NETLINK_CB(cb->skb).sk), 923 NETLINK_CB(cb->skb).portid, 924 cb->nlh->nlmsg_seq, NLM_F_MULTI, 925 cb->nlh); 926 if (res < 0) { 927 spin_unlock_bh(lock); 928 goto done; 929 } 930 next_normal: 931 ++num; 932 } 933 934 spin_unlock_bh(lock); 935 } 936 937 done: 938 cb->args[1] = i; 939 cb->args[2] = num; 940 out: 941 ; 942 } 943 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk); 944 945 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb, 946 const struct inet_diag_req_v2 *r, 947 struct nlattr *bc) 948 { 949 const struct inet_diag_handler *handler; 950 int err = 0; 951 952 handler = inet_diag_lock_handler(r->sdiag_protocol); 953 if (!IS_ERR(handler)) 954 handler->dump(skb, cb, r, bc); 955 else 956 err = PTR_ERR(handler); 957 inet_diag_unlock_handler(handler); 958 959 return err ? : skb->len; 960 } 961 962 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb) 963 { 964 int hdrlen = sizeof(struct inet_diag_req_v2); 965 struct nlattr *bc = NULL; 966 967 if (nlmsg_attrlen(cb->nlh, hdrlen)) 968 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE); 969 970 return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh), bc); 971 } 972 973 static int inet_diag_type2proto(int type) 974 { 975 switch (type) { 976 case TCPDIAG_GETSOCK: 977 return IPPROTO_TCP; 978 case DCCPDIAG_GETSOCK: 979 return IPPROTO_DCCP; 980 default: 981 return 0; 982 } 983 } 984 985 static int inet_diag_dump_compat(struct sk_buff *skb, 986 struct netlink_callback *cb) 987 { 988 struct inet_diag_req *rc = nlmsg_data(cb->nlh); 989 int hdrlen = sizeof(struct inet_diag_req); 990 struct inet_diag_req_v2 req; 991 struct nlattr *bc = NULL; 992 993 req.sdiag_family = AF_UNSPEC; /* compatibility */ 994 req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type); 995 req.idiag_ext = rc->idiag_ext; 996 req.idiag_states = rc->idiag_states; 997 req.id = rc->id; 998 999 if (nlmsg_attrlen(cb->nlh, hdrlen)) 1000 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE); 1001 1002 return __inet_diag_dump(skb, cb, &req, bc); 1003 } 1004 1005 static int inet_diag_get_exact_compat(struct sk_buff *in_skb, 1006 const struct nlmsghdr *nlh) 1007 { 1008 struct inet_diag_req *rc = nlmsg_data(nlh); 1009 struct inet_diag_req_v2 req; 1010 1011 req.sdiag_family = rc->idiag_family; 1012 req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type); 1013 req.idiag_ext = rc->idiag_ext; 1014 req.idiag_states = rc->idiag_states; 1015 req.id = rc->id; 1016 1017 return inet_diag_get_exact(in_skb, nlh, &req); 1018 } 1019 1020 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh) 1021 { 1022 int hdrlen = sizeof(struct inet_diag_req); 1023 struct net *net = sock_net(skb->sk); 1024 1025 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX || 1026 nlmsg_len(nlh) < hdrlen) 1027 return -EINVAL; 1028 1029 if (nlh->nlmsg_flags & NLM_F_DUMP) { 1030 if (nlmsg_attrlen(nlh, hdrlen)) { 1031 struct nlattr *attr; 1032 1033 attr = nlmsg_find_attr(nlh, hdrlen, 1034 INET_DIAG_REQ_BYTECODE); 1035 if (!attr || 1036 nla_len(attr) < sizeof(struct inet_diag_bc_op) || 1037 inet_diag_bc_audit(nla_data(attr), nla_len(attr))) 1038 return -EINVAL; 1039 } 1040 { 1041 struct netlink_dump_control c = { 1042 .dump = inet_diag_dump_compat, 1043 }; 1044 return netlink_dump_start(net->diag_nlsk, skb, nlh, &c); 1045 } 1046 } 1047 1048 return inet_diag_get_exact_compat(skb, nlh); 1049 } 1050 1051 static int inet_diag_handler_dump(struct sk_buff *skb, struct nlmsghdr *h) 1052 { 1053 int hdrlen = sizeof(struct inet_diag_req_v2); 1054 struct net *net = sock_net(skb->sk); 1055 1056 if (nlmsg_len(h) < hdrlen) 1057 return -EINVAL; 1058 1059 if (h->nlmsg_flags & NLM_F_DUMP) { 1060 if (nlmsg_attrlen(h, hdrlen)) { 1061 struct nlattr *attr; 1062 1063 attr = nlmsg_find_attr(h, hdrlen, 1064 INET_DIAG_REQ_BYTECODE); 1065 if (!attr || 1066 nla_len(attr) < sizeof(struct inet_diag_bc_op) || 1067 inet_diag_bc_audit(nla_data(attr), nla_len(attr))) 1068 return -EINVAL; 1069 } 1070 { 1071 struct netlink_dump_control c = { 1072 .dump = inet_diag_dump, 1073 }; 1074 return netlink_dump_start(net->diag_nlsk, skb, h, &c); 1075 } 1076 } 1077 1078 return inet_diag_get_exact(skb, h, nlmsg_data(h)); 1079 } 1080 1081 static const struct sock_diag_handler inet_diag_handler = { 1082 .family = AF_INET, 1083 .dump = inet_diag_handler_dump, 1084 }; 1085 1086 static const struct sock_diag_handler inet6_diag_handler = { 1087 .family = AF_INET6, 1088 .dump = inet_diag_handler_dump, 1089 }; 1090 1091 int inet_diag_register(const struct inet_diag_handler *h) 1092 { 1093 const __u16 type = h->idiag_type; 1094 int err = -EINVAL; 1095 1096 if (type >= IPPROTO_MAX) 1097 goto out; 1098 1099 mutex_lock(&inet_diag_table_mutex); 1100 err = -EEXIST; 1101 if (!inet_diag_table[type]) { 1102 inet_diag_table[type] = h; 1103 err = 0; 1104 } 1105 mutex_unlock(&inet_diag_table_mutex); 1106 out: 1107 return err; 1108 } 1109 EXPORT_SYMBOL_GPL(inet_diag_register); 1110 1111 void inet_diag_unregister(const struct inet_diag_handler *h) 1112 { 1113 const __u16 type = h->idiag_type; 1114 1115 if (type >= IPPROTO_MAX) 1116 return; 1117 1118 mutex_lock(&inet_diag_table_mutex); 1119 inet_diag_table[type] = NULL; 1120 mutex_unlock(&inet_diag_table_mutex); 1121 } 1122 EXPORT_SYMBOL_GPL(inet_diag_unregister); 1123 1124 static int __init inet_diag_init(void) 1125 { 1126 const int inet_diag_table_size = (IPPROTO_MAX * 1127 sizeof(struct inet_diag_handler *)); 1128 int err = -ENOMEM; 1129 1130 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL); 1131 if (!inet_diag_table) 1132 goto out; 1133 1134 err = sock_diag_register(&inet_diag_handler); 1135 if (err) 1136 goto out_free_nl; 1137 1138 err = sock_diag_register(&inet6_diag_handler); 1139 if (err) 1140 goto out_free_inet; 1141 1142 sock_diag_register_inet_compat(inet_diag_rcv_msg_compat); 1143 out: 1144 return err; 1145 1146 out_free_inet: 1147 sock_diag_unregister(&inet_diag_handler); 1148 out_free_nl: 1149 kfree(inet_diag_table); 1150 goto out; 1151 } 1152 1153 static void __exit inet_diag_exit(void) 1154 { 1155 sock_diag_unregister(&inet6_diag_handler); 1156 sock_diag_unregister(&inet_diag_handler); 1157 sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat); 1158 kfree(inet_diag_table); 1159 } 1160 1161 module_init(inet_diag_init); 1162 module_exit(inet_diag_exit); 1163 MODULE_LICENSE("GPL"); 1164 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */); 1165 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */); 1166