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 __be32 *saddr; 42 __be32 *daddr; 43 u16 sport; 44 u16 dport; 45 u16 family; 46 u16 userlocks; 47 #if IS_ENABLED(CONFIG_IPV6) 48 struct in6_addr saddr_storage; /* for IPv4-mapped-IPv6 addresses */ 49 struct in6_addr daddr_storage; /* for IPv4-mapped-IPv6 addresses */ 50 #endif 51 }; 52 53 static DEFINE_MUTEX(inet_diag_table_mutex); 54 55 static const struct inet_diag_handler *inet_diag_lock_handler(int proto) 56 { 57 if (!inet_diag_table[proto]) 58 request_module("net-pf-%d-proto-%d-type-%d-%d", PF_NETLINK, 59 NETLINK_SOCK_DIAG, AF_INET, proto); 60 61 mutex_lock(&inet_diag_table_mutex); 62 if (!inet_diag_table[proto]) 63 return ERR_PTR(-ENOENT); 64 65 return inet_diag_table[proto]; 66 } 67 68 static inline void inet_diag_unlock_handler( 69 const struct inet_diag_handler *handler) 70 { 71 mutex_unlock(&inet_diag_table_mutex); 72 } 73 74 int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk, 75 struct sk_buff *skb, struct inet_diag_req_v2 *req, 76 struct user_namespace *user_ns, 77 u32 portid, u32 seq, u16 nlmsg_flags, 78 const struct nlmsghdr *unlh) 79 { 80 const struct inet_sock *inet = inet_sk(sk); 81 struct inet_diag_msg *r; 82 struct nlmsghdr *nlh; 83 struct nlattr *attr; 84 void *info = NULL; 85 const struct inet_diag_handler *handler; 86 int ext = req->idiag_ext; 87 88 handler = inet_diag_table[req->sdiag_protocol]; 89 BUG_ON(handler == NULL); 90 91 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r), 92 nlmsg_flags); 93 if (!nlh) 94 return -EMSGSIZE; 95 96 r = nlmsg_data(nlh); 97 BUG_ON(sk->sk_state == TCP_TIME_WAIT); 98 99 r->idiag_family = sk->sk_family; 100 r->idiag_state = sk->sk_state; 101 r->idiag_timer = 0; 102 r->idiag_retrans = 0; 103 104 r->id.idiag_if = sk->sk_bound_dev_if; 105 sock_diag_save_cookie(sk, r->id.idiag_cookie); 106 107 r->id.idiag_sport = inet->inet_sport; 108 r->id.idiag_dport = inet->inet_dport; 109 r->id.idiag_src[0] = inet->inet_rcv_saddr; 110 r->id.idiag_dst[0] = inet->inet_daddr; 111 112 /* IPv6 dual-stack sockets use inet->tos for IPv4 connections, 113 * hence this needs to be included regardless of socket family. 114 */ 115 if (ext & (1 << (INET_DIAG_TOS - 1))) 116 if (nla_put_u8(skb, INET_DIAG_TOS, inet->tos) < 0) 117 goto errout; 118 119 #if IS_ENABLED(CONFIG_IPV6) 120 if (r->idiag_family == AF_INET6) { 121 const struct ipv6_pinfo *np = inet6_sk(sk); 122 123 *(struct in6_addr *)r->id.idiag_src = np->rcv_saddr; 124 *(struct in6_addr *)r->id.idiag_dst = np->daddr; 125 126 if (ext & (1 << (INET_DIAG_TCLASS - 1))) 127 if (nla_put_u8(skb, INET_DIAG_TCLASS, np->tclass) < 0) 128 goto errout; 129 } 130 #endif 131 132 r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk)); 133 r->idiag_inode = sock_i_ino(sk); 134 135 if (ext & (1 << (INET_DIAG_MEMINFO - 1))) { 136 struct inet_diag_meminfo minfo = { 137 .idiag_rmem = sk_rmem_alloc_get(sk), 138 .idiag_wmem = sk->sk_wmem_queued, 139 .idiag_fmem = sk->sk_forward_alloc, 140 .idiag_tmem = sk_wmem_alloc_get(sk), 141 }; 142 143 if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0) 144 goto errout; 145 } 146 147 if (ext & (1 << (INET_DIAG_SKMEMINFO - 1))) 148 if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO)) 149 goto errout; 150 151 if (icsk == NULL) { 152 handler->idiag_get_info(sk, r, NULL); 153 goto out; 154 } 155 156 #define EXPIRES_IN_MS(tmo) DIV_ROUND_UP((tmo - jiffies) * 1000, HZ) 157 158 if (icsk->icsk_pending == ICSK_TIME_RETRANS) { 159 r->idiag_timer = 1; 160 r->idiag_retrans = icsk->icsk_retransmits; 161 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout); 162 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { 163 r->idiag_timer = 4; 164 r->idiag_retrans = icsk->icsk_probes_out; 165 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout); 166 } else if (timer_pending(&sk->sk_timer)) { 167 r->idiag_timer = 2; 168 r->idiag_retrans = icsk->icsk_probes_out; 169 r->idiag_expires = EXPIRES_IN_MS(sk->sk_timer.expires); 170 } else { 171 r->idiag_timer = 0; 172 r->idiag_expires = 0; 173 } 174 #undef EXPIRES_IN_MS 175 176 if (ext & (1 << (INET_DIAG_INFO - 1))) { 177 attr = nla_reserve(skb, INET_DIAG_INFO, 178 sizeof(struct tcp_info)); 179 if (!attr) 180 goto errout; 181 182 info = nla_data(attr); 183 } 184 185 if ((ext & (1 << (INET_DIAG_CONG - 1))) && icsk->icsk_ca_ops) 186 if (nla_put_string(skb, INET_DIAG_CONG, 187 icsk->icsk_ca_ops->name) < 0) 188 goto errout; 189 190 handler->idiag_get_info(sk, r, info); 191 192 if (sk->sk_state < TCP_TIME_WAIT && 193 icsk->icsk_ca_ops && icsk->icsk_ca_ops->get_info) 194 icsk->icsk_ca_ops->get_info(sk, ext, skb); 195 196 out: 197 return nlmsg_end(skb, nlh); 198 199 errout: 200 nlmsg_cancel(skb, nlh); 201 return -EMSGSIZE; 202 } 203 EXPORT_SYMBOL_GPL(inet_sk_diag_fill); 204 205 static int inet_csk_diag_fill(struct sock *sk, 206 struct sk_buff *skb, struct inet_diag_req_v2 *req, 207 struct user_namespace *user_ns, 208 u32 portid, u32 seq, u16 nlmsg_flags, 209 const struct nlmsghdr *unlh) 210 { 211 return inet_sk_diag_fill(sk, inet_csk(sk), 212 skb, req, user_ns, portid, seq, nlmsg_flags, unlh); 213 } 214 215 static int inet_twsk_diag_fill(struct inet_timewait_sock *tw, 216 struct sk_buff *skb, struct inet_diag_req_v2 *req, 217 u32 portid, u32 seq, u16 nlmsg_flags, 218 const struct nlmsghdr *unlh) 219 { 220 long tmo; 221 struct inet_diag_msg *r; 222 struct nlmsghdr *nlh; 223 224 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r), 225 nlmsg_flags); 226 if (!nlh) 227 return -EMSGSIZE; 228 229 r = nlmsg_data(nlh); 230 BUG_ON(tw->tw_state != TCP_TIME_WAIT); 231 232 tmo = tw->tw_ttd - jiffies; 233 if (tmo < 0) 234 tmo = 0; 235 236 r->idiag_family = tw->tw_family; 237 r->idiag_retrans = 0; 238 r->id.idiag_if = tw->tw_bound_dev_if; 239 sock_diag_save_cookie(tw, r->id.idiag_cookie); 240 r->id.idiag_sport = tw->tw_sport; 241 r->id.idiag_dport = tw->tw_dport; 242 r->id.idiag_src[0] = tw->tw_rcv_saddr; 243 r->id.idiag_dst[0] = tw->tw_daddr; 244 r->idiag_state = tw->tw_substate; 245 r->idiag_timer = 3; 246 r->idiag_expires = DIV_ROUND_UP(tmo * 1000, HZ); 247 r->idiag_rqueue = 0; 248 r->idiag_wqueue = 0; 249 r->idiag_uid = 0; 250 r->idiag_inode = 0; 251 #if IS_ENABLED(CONFIG_IPV6) 252 if (tw->tw_family == AF_INET6) { 253 const struct inet6_timewait_sock *tw6 = 254 inet6_twsk((struct sock *)tw); 255 256 *(struct in6_addr *)r->id.idiag_src = tw6->tw_v6_rcv_saddr; 257 *(struct in6_addr *)r->id.idiag_dst = tw6->tw_v6_daddr; 258 } 259 #endif 260 261 return nlmsg_end(skb, nlh); 262 } 263 264 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb, 265 struct inet_diag_req_v2 *r, 266 struct user_namespace *user_ns, 267 u32 portid, u32 seq, u16 nlmsg_flags, 268 const struct nlmsghdr *unlh) 269 { 270 if (sk->sk_state == TCP_TIME_WAIT) 271 return inet_twsk_diag_fill((struct inet_timewait_sock *)sk, 272 skb, r, portid, seq, nlmsg_flags, 273 unlh); 274 return inet_csk_diag_fill(sk, skb, r, user_ns, portid, seq, nlmsg_flags, unlh); 275 } 276 277 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *in_skb, 278 const struct nlmsghdr *nlh, struct inet_diag_req_v2 *req) 279 { 280 int err; 281 struct sock *sk; 282 struct sk_buff *rep; 283 struct net *net = sock_net(in_skb->sk); 284 285 err = -EINVAL; 286 if (req->sdiag_family == AF_INET) { 287 sk = inet_lookup(net, hashinfo, req->id.idiag_dst[0], 288 req->id.idiag_dport, req->id.idiag_src[0], 289 req->id.idiag_sport, req->id.idiag_if); 290 } 291 #if IS_ENABLED(CONFIG_IPV6) 292 else if (req->sdiag_family == AF_INET6) { 293 sk = inet6_lookup(net, hashinfo, 294 (struct in6_addr *)req->id.idiag_dst, 295 req->id.idiag_dport, 296 (struct in6_addr *)req->id.idiag_src, 297 req->id.idiag_sport, 298 req->id.idiag_if); 299 } 300 #endif 301 else { 302 goto out_nosk; 303 } 304 305 err = -ENOENT; 306 if (sk == NULL) 307 goto out_nosk; 308 309 err = sock_diag_check_cookie(sk, req->id.idiag_cookie); 310 if (err) 311 goto out; 312 313 rep = nlmsg_new(sizeof(struct inet_diag_msg) + 314 sizeof(struct inet_diag_meminfo) + 315 sizeof(struct tcp_info) + 64, GFP_KERNEL); 316 if (!rep) { 317 err = -ENOMEM; 318 goto out; 319 } 320 321 err = sk_diag_fill(sk, rep, req, 322 sk_user_ns(NETLINK_CB(in_skb).ssk), 323 NETLINK_CB(in_skb).portid, 324 nlh->nlmsg_seq, 0, nlh); 325 if (err < 0) { 326 WARN_ON(err == -EMSGSIZE); 327 nlmsg_free(rep); 328 goto out; 329 } 330 err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid, 331 MSG_DONTWAIT); 332 if (err > 0) 333 err = 0; 334 335 out: 336 if (sk) { 337 if (sk->sk_state == TCP_TIME_WAIT) 338 inet_twsk_put((struct inet_timewait_sock *)sk); 339 else 340 sock_put(sk); 341 } 342 out_nosk: 343 return err; 344 } 345 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk); 346 347 static int inet_diag_get_exact(struct sk_buff *in_skb, 348 const struct nlmsghdr *nlh, 349 struct inet_diag_req_v2 *req) 350 { 351 const struct inet_diag_handler *handler; 352 int err; 353 354 handler = inet_diag_lock_handler(req->sdiag_protocol); 355 if (IS_ERR(handler)) 356 err = PTR_ERR(handler); 357 else 358 err = handler->dump_one(in_skb, nlh, req); 359 inet_diag_unlock_handler(handler); 360 361 return err; 362 } 363 364 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits) 365 { 366 int words = bits >> 5; 367 368 bits &= 0x1f; 369 370 if (words) { 371 if (memcmp(a1, a2, words << 2)) 372 return 0; 373 } 374 if (bits) { 375 __be32 w1, w2; 376 __be32 mask; 377 378 w1 = a1[words]; 379 w2 = a2[words]; 380 381 mask = htonl((0xffffffff) << (32 - bits)); 382 383 if ((w1 ^ w2) & mask) 384 return 0; 385 } 386 387 return 1; 388 } 389 390 391 static int inet_diag_bc_run(const struct nlattr *_bc, 392 const struct inet_diag_entry *entry) 393 { 394 const void *bc = nla_data(_bc); 395 int len = nla_len(_bc); 396 397 while (len > 0) { 398 int yes = 1; 399 const struct inet_diag_bc_op *op = bc; 400 401 switch (op->code) { 402 case INET_DIAG_BC_NOP: 403 break; 404 case INET_DIAG_BC_JMP: 405 yes = 0; 406 break; 407 case INET_DIAG_BC_S_GE: 408 yes = entry->sport >= op[1].no; 409 break; 410 case INET_DIAG_BC_S_LE: 411 yes = entry->sport <= op[1].no; 412 break; 413 case INET_DIAG_BC_D_GE: 414 yes = entry->dport >= op[1].no; 415 break; 416 case INET_DIAG_BC_D_LE: 417 yes = entry->dport <= op[1].no; 418 break; 419 case INET_DIAG_BC_AUTO: 420 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK); 421 break; 422 case INET_DIAG_BC_S_COND: 423 case INET_DIAG_BC_D_COND: { 424 struct inet_diag_hostcond *cond; 425 __be32 *addr; 426 427 cond = (struct inet_diag_hostcond *)(op + 1); 428 if (cond->port != -1 && 429 cond->port != (op->code == INET_DIAG_BC_S_COND ? 430 entry->sport : entry->dport)) { 431 yes = 0; 432 break; 433 } 434 435 if (op->code == INET_DIAG_BC_S_COND) 436 addr = entry->saddr; 437 else 438 addr = entry->daddr; 439 440 if (cond->family != AF_UNSPEC && 441 cond->family != entry->family) { 442 if (entry->family == AF_INET6 && 443 cond->family == AF_INET) { 444 if (addr[0] == 0 && addr[1] == 0 && 445 addr[2] == htonl(0xffff) && 446 bitstring_match(addr + 3, 447 cond->addr, 448 cond->prefix_len)) 449 break; 450 } 451 yes = 0; 452 break; 453 } 454 455 if (cond->prefix_len == 0) 456 break; 457 if (bitstring_match(addr, cond->addr, 458 cond->prefix_len)) 459 break; 460 yes = 0; 461 break; 462 } 463 } 464 465 if (yes) { 466 len -= op->yes; 467 bc += op->yes; 468 } else { 469 len -= op->no; 470 bc += op->no; 471 } 472 } 473 return len == 0; 474 } 475 476 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk) 477 { 478 struct inet_diag_entry entry; 479 struct inet_sock *inet = inet_sk(sk); 480 481 if (bc == NULL) 482 return 1; 483 484 entry.family = sk->sk_family; 485 #if IS_ENABLED(CONFIG_IPV6) 486 if (entry.family == AF_INET6) { 487 struct ipv6_pinfo *np = inet6_sk(sk); 488 489 entry.saddr = np->rcv_saddr.s6_addr32; 490 entry.daddr = np->daddr.s6_addr32; 491 } else 492 #endif 493 { 494 entry.saddr = &inet->inet_rcv_saddr; 495 entry.daddr = &inet->inet_daddr; 496 } 497 entry.sport = inet->inet_num; 498 entry.dport = ntohs(inet->inet_dport); 499 entry.userlocks = sk->sk_userlocks; 500 501 return inet_diag_bc_run(bc, &entry); 502 } 503 EXPORT_SYMBOL_GPL(inet_diag_bc_sk); 504 505 static int valid_cc(const void *bc, int len, int cc) 506 { 507 while (len >= 0) { 508 const struct inet_diag_bc_op *op = bc; 509 510 if (cc > len) 511 return 0; 512 if (cc == len) 513 return 1; 514 if (op->yes < 4 || op->yes & 3) 515 return 0; 516 len -= op->yes; 517 bc += op->yes; 518 } 519 return 0; 520 } 521 522 /* Validate an inet_diag_hostcond. */ 523 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len, 524 int *min_len) 525 { 526 int addr_len; 527 struct inet_diag_hostcond *cond; 528 529 /* Check hostcond space. */ 530 *min_len += sizeof(struct inet_diag_hostcond); 531 if (len < *min_len) 532 return false; 533 cond = (struct inet_diag_hostcond *)(op + 1); 534 535 /* Check address family and address length. */ 536 switch (cond->family) { 537 case AF_UNSPEC: 538 addr_len = 0; 539 break; 540 case AF_INET: 541 addr_len = sizeof(struct in_addr); 542 break; 543 case AF_INET6: 544 addr_len = sizeof(struct in6_addr); 545 break; 546 default: 547 return false; 548 } 549 *min_len += addr_len; 550 if (len < *min_len) 551 return false; 552 553 /* Check prefix length (in bits) vs address length (in bytes). */ 554 if (cond->prefix_len > 8 * addr_len) 555 return false; 556 557 return true; 558 } 559 560 /* Validate a port comparison operator. */ 561 static inline bool valid_port_comparison(const struct inet_diag_bc_op *op, 562 int len, int *min_len) 563 { 564 /* Port comparisons put the port in a follow-on inet_diag_bc_op. */ 565 *min_len += sizeof(struct inet_diag_bc_op); 566 if (len < *min_len) 567 return false; 568 return true; 569 } 570 571 static int inet_diag_bc_audit(const void *bytecode, int bytecode_len) 572 { 573 const void *bc = bytecode; 574 int len = bytecode_len; 575 576 while (len > 0) { 577 const struct inet_diag_bc_op *op = bc; 578 int min_len = sizeof(struct inet_diag_bc_op); 579 580 //printk("BC: %d %d %d {%d} / %d\n", op->code, op->yes, op->no, op[1].no, len); 581 switch (op->code) { 582 case INET_DIAG_BC_S_COND: 583 case INET_DIAG_BC_D_COND: 584 if (!valid_hostcond(bc, len, &min_len)) 585 return -EINVAL; 586 break; 587 case INET_DIAG_BC_S_GE: 588 case INET_DIAG_BC_S_LE: 589 case INET_DIAG_BC_D_GE: 590 case INET_DIAG_BC_D_LE: 591 if (!valid_port_comparison(bc, len, &min_len)) 592 return -EINVAL; 593 break; 594 case INET_DIAG_BC_AUTO: 595 case INET_DIAG_BC_JMP: 596 case INET_DIAG_BC_NOP: 597 break; 598 default: 599 return -EINVAL; 600 } 601 602 if (op->code != INET_DIAG_BC_NOP) { 603 if (op->no < min_len || op->no > len + 4 || op->no & 3) 604 return -EINVAL; 605 if (op->no < len && 606 !valid_cc(bytecode, bytecode_len, len - op->no)) 607 return -EINVAL; 608 } 609 610 if (op->yes < min_len || op->yes > len + 4 || op->yes & 3) 611 return -EINVAL; 612 bc += op->yes; 613 len -= op->yes; 614 } 615 return len == 0 ? 0 : -EINVAL; 616 } 617 618 static int inet_csk_diag_dump(struct sock *sk, 619 struct sk_buff *skb, 620 struct netlink_callback *cb, 621 struct inet_diag_req_v2 *r, 622 const struct nlattr *bc) 623 { 624 if (!inet_diag_bc_sk(bc, sk)) 625 return 0; 626 627 return inet_csk_diag_fill(sk, skb, r, 628 sk_user_ns(NETLINK_CB(cb->skb).ssk), 629 NETLINK_CB(cb->skb).portid, 630 cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh); 631 } 632 633 static int inet_twsk_diag_dump(struct inet_timewait_sock *tw, 634 struct sk_buff *skb, 635 struct netlink_callback *cb, 636 struct inet_diag_req_v2 *r, 637 const struct nlattr *bc) 638 { 639 if (bc != NULL) { 640 struct inet_diag_entry entry; 641 642 entry.family = tw->tw_family; 643 #if IS_ENABLED(CONFIG_IPV6) 644 if (tw->tw_family == AF_INET6) { 645 struct inet6_timewait_sock *tw6 = 646 inet6_twsk((struct sock *)tw); 647 entry.saddr = tw6->tw_v6_rcv_saddr.s6_addr32; 648 entry.daddr = tw6->tw_v6_daddr.s6_addr32; 649 } else 650 #endif 651 { 652 entry.saddr = &tw->tw_rcv_saddr; 653 entry.daddr = &tw->tw_daddr; 654 } 655 entry.sport = tw->tw_num; 656 entry.dport = ntohs(tw->tw_dport); 657 entry.userlocks = 0; 658 659 if (!inet_diag_bc_run(bc, &entry)) 660 return 0; 661 } 662 663 return inet_twsk_diag_fill(tw, skb, r, 664 NETLINK_CB(cb->skb).portid, 665 cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh); 666 } 667 668 /* Get the IPv4, IPv6, or IPv4-mapped-IPv6 local and remote addresses 669 * from a request_sock. For IPv4-mapped-IPv6 we must map IPv4 to IPv6. 670 */ 671 static inline void inet_diag_req_addrs(const struct sock *sk, 672 const struct request_sock *req, 673 struct inet_diag_entry *entry) 674 { 675 struct inet_request_sock *ireq = inet_rsk(req); 676 677 #if IS_ENABLED(CONFIG_IPV6) 678 if (sk->sk_family == AF_INET6) { 679 if (req->rsk_ops->family == AF_INET6) { 680 entry->saddr = inet6_rsk(req)->loc_addr.s6_addr32; 681 entry->daddr = inet6_rsk(req)->rmt_addr.s6_addr32; 682 } else if (req->rsk_ops->family == AF_INET) { 683 ipv6_addr_set_v4mapped(ireq->loc_addr, 684 &entry->saddr_storage); 685 ipv6_addr_set_v4mapped(ireq->rmt_addr, 686 &entry->daddr_storage); 687 entry->saddr = entry->saddr_storage.s6_addr32; 688 entry->daddr = entry->daddr_storage.s6_addr32; 689 } 690 } else 691 #endif 692 { 693 entry->saddr = &ireq->loc_addr; 694 entry->daddr = &ireq->rmt_addr; 695 } 696 } 697 698 static int inet_diag_fill_req(struct sk_buff *skb, struct sock *sk, 699 struct request_sock *req, 700 struct user_namespace *user_ns, 701 u32 portid, u32 seq, 702 const struct nlmsghdr *unlh) 703 { 704 const struct inet_request_sock *ireq = inet_rsk(req); 705 struct inet_sock *inet = inet_sk(sk); 706 struct inet_diag_msg *r; 707 struct nlmsghdr *nlh; 708 long tmo; 709 710 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r), 711 NLM_F_MULTI); 712 if (!nlh) 713 return -EMSGSIZE; 714 715 r = nlmsg_data(nlh); 716 r->idiag_family = sk->sk_family; 717 r->idiag_state = TCP_SYN_RECV; 718 r->idiag_timer = 1; 719 r->idiag_retrans = req->retrans; 720 721 r->id.idiag_if = sk->sk_bound_dev_if; 722 sock_diag_save_cookie(req, r->id.idiag_cookie); 723 724 tmo = req->expires - jiffies; 725 if (tmo < 0) 726 tmo = 0; 727 728 r->id.idiag_sport = inet->inet_sport; 729 r->id.idiag_dport = ireq->rmt_port; 730 r->id.idiag_src[0] = ireq->loc_addr; 731 r->id.idiag_dst[0] = ireq->rmt_addr; 732 r->idiag_expires = jiffies_to_msecs(tmo); 733 r->idiag_rqueue = 0; 734 r->idiag_wqueue = 0; 735 r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk)); 736 r->idiag_inode = 0; 737 #if IS_ENABLED(CONFIG_IPV6) 738 if (r->idiag_family == AF_INET6) { 739 struct inet_diag_entry entry; 740 inet_diag_req_addrs(sk, req, &entry); 741 memcpy(r->id.idiag_src, entry.saddr, sizeof(struct in6_addr)); 742 memcpy(r->id.idiag_dst, entry.daddr, sizeof(struct in6_addr)); 743 } 744 #endif 745 746 return nlmsg_end(skb, nlh); 747 } 748 749 static int inet_diag_dump_reqs(struct sk_buff *skb, struct sock *sk, 750 struct netlink_callback *cb, 751 struct inet_diag_req_v2 *r, 752 const struct nlattr *bc) 753 { 754 struct inet_diag_entry entry; 755 struct inet_connection_sock *icsk = inet_csk(sk); 756 struct listen_sock *lopt; 757 struct inet_sock *inet = inet_sk(sk); 758 int j, s_j; 759 int reqnum, s_reqnum; 760 int err = 0; 761 762 s_j = cb->args[3]; 763 s_reqnum = cb->args[4]; 764 765 if (s_j > 0) 766 s_j--; 767 768 entry.family = sk->sk_family; 769 770 read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock); 771 772 lopt = icsk->icsk_accept_queue.listen_opt; 773 if (!lopt || !lopt->qlen) 774 goto out; 775 776 if (bc != NULL) { 777 entry.sport = inet->inet_num; 778 entry.userlocks = sk->sk_userlocks; 779 } 780 781 for (j = s_j; j < lopt->nr_table_entries; j++) { 782 struct request_sock *req, *head = lopt->syn_table[j]; 783 784 reqnum = 0; 785 for (req = head; req; reqnum++, req = req->dl_next) { 786 struct inet_request_sock *ireq = inet_rsk(req); 787 788 if (reqnum < s_reqnum) 789 continue; 790 if (r->id.idiag_dport != ireq->rmt_port && 791 r->id.idiag_dport) 792 continue; 793 794 if (bc) { 795 inet_diag_req_addrs(sk, req, &entry); 796 entry.dport = ntohs(ireq->rmt_port); 797 798 if (!inet_diag_bc_run(bc, &entry)) 799 continue; 800 } 801 802 err = inet_diag_fill_req(skb, sk, req, 803 sk_user_ns(NETLINK_CB(cb->skb).ssk), 804 NETLINK_CB(cb->skb).portid, 805 cb->nlh->nlmsg_seq, cb->nlh); 806 if (err < 0) { 807 cb->args[3] = j + 1; 808 cb->args[4] = reqnum; 809 goto out; 810 } 811 } 812 813 s_reqnum = 0; 814 } 815 816 out: 817 read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); 818 819 return err; 820 } 821 822 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb, 823 struct netlink_callback *cb, struct inet_diag_req_v2 *r, struct nlattr *bc) 824 { 825 int i, num; 826 int s_i, s_num; 827 struct net *net = sock_net(skb->sk); 828 829 s_i = cb->args[1]; 830 s_num = num = cb->args[2]; 831 832 if (cb->args[0] == 0) { 833 if (!(r->idiag_states & (TCPF_LISTEN | TCPF_SYN_RECV))) 834 goto skip_listen_ht; 835 836 for (i = s_i; i < INET_LHTABLE_SIZE; i++) { 837 struct sock *sk; 838 struct hlist_nulls_node *node; 839 struct inet_listen_hashbucket *ilb; 840 841 num = 0; 842 ilb = &hashinfo->listening_hash[i]; 843 spin_lock_bh(&ilb->lock); 844 sk_nulls_for_each(sk, node, &ilb->head) { 845 struct inet_sock *inet = inet_sk(sk); 846 847 if (!net_eq(sock_net(sk), net)) 848 continue; 849 850 if (num < s_num) { 851 num++; 852 continue; 853 } 854 855 if (r->sdiag_family != AF_UNSPEC && 856 sk->sk_family != r->sdiag_family) 857 goto next_listen; 858 859 if (r->id.idiag_sport != inet->inet_sport && 860 r->id.idiag_sport) 861 goto next_listen; 862 863 if (!(r->idiag_states & TCPF_LISTEN) || 864 r->id.idiag_dport || 865 cb->args[3] > 0) 866 goto syn_recv; 867 868 if (inet_csk_diag_dump(sk, skb, cb, r, bc) < 0) { 869 spin_unlock_bh(&ilb->lock); 870 goto done; 871 } 872 873 syn_recv: 874 if (!(r->idiag_states & TCPF_SYN_RECV)) 875 goto next_listen; 876 877 if (inet_diag_dump_reqs(skb, sk, cb, r, bc) < 0) { 878 spin_unlock_bh(&ilb->lock); 879 goto done; 880 } 881 882 next_listen: 883 cb->args[3] = 0; 884 cb->args[4] = 0; 885 ++num; 886 } 887 spin_unlock_bh(&ilb->lock); 888 889 s_num = 0; 890 cb->args[3] = 0; 891 cb->args[4] = 0; 892 } 893 skip_listen_ht: 894 cb->args[0] = 1; 895 s_i = num = s_num = 0; 896 } 897 898 if (!(r->idiag_states & ~(TCPF_LISTEN | TCPF_SYN_RECV))) 899 goto out; 900 901 for (i = s_i; i <= hashinfo->ehash_mask; i++) { 902 struct inet_ehash_bucket *head = &hashinfo->ehash[i]; 903 spinlock_t *lock = inet_ehash_lockp(hashinfo, i); 904 struct sock *sk; 905 struct hlist_nulls_node *node; 906 907 num = 0; 908 909 if (hlist_nulls_empty(&head->chain) && 910 hlist_nulls_empty(&head->twchain)) 911 continue; 912 913 if (i > s_i) 914 s_num = 0; 915 916 spin_lock_bh(lock); 917 sk_nulls_for_each(sk, node, &head->chain) { 918 struct inet_sock *inet = inet_sk(sk); 919 920 if (!net_eq(sock_net(sk), net)) 921 continue; 922 if (num < s_num) 923 goto next_normal; 924 if (!(r->idiag_states & (1 << sk->sk_state))) 925 goto next_normal; 926 if (r->sdiag_family != AF_UNSPEC && 927 sk->sk_family != r->sdiag_family) 928 goto next_normal; 929 if (r->id.idiag_sport != inet->inet_sport && 930 r->id.idiag_sport) 931 goto next_normal; 932 if (r->id.idiag_dport != inet->inet_dport && 933 r->id.idiag_dport) 934 goto next_normal; 935 if (inet_csk_diag_dump(sk, skb, cb, r, bc) < 0) { 936 spin_unlock_bh(lock); 937 goto done; 938 } 939 next_normal: 940 ++num; 941 } 942 943 if (r->idiag_states & TCPF_TIME_WAIT) { 944 struct inet_timewait_sock *tw; 945 946 inet_twsk_for_each(tw, node, 947 &head->twchain) { 948 if (!net_eq(twsk_net(tw), net)) 949 continue; 950 951 if (num < s_num) 952 goto next_dying; 953 if (r->sdiag_family != AF_UNSPEC && 954 tw->tw_family != r->sdiag_family) 955 goto next_dying; 956 if (r->id.idiag_sport != tw->tw_sport && 957 r->id.idiag_sport) 958 goto next_dying; 959 if (r->id.idiag_dport != tw->tw_dport && 960 r->id.idiag_dport) 961 goto next_dying; 962 if (inet_twsk_diag_dump(tw, skb, cb, r, bc) < 0) { 963 spin_unlock_bh(lock); 964 goto done; 965 } 966 next_dying: 967 ++num; 968 } 969 } 970 spin_unlock_bh(lock); 971 } 972 973 done: 974 cb->args[1] = i; 975 cb->args[2] = num; 976 out: 977 ; 978 } 979 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk); 980 981 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb, 982 struct inet_diag_req_v2 *r, struct nlattr *bc) 983 { 984 const struct inet_diag_handler *handler; 985 int err = 0; 986 987 handler = inet_diag_lock_handler(r->sdiag_protocol); 988 if (!IS_ERR(handler)) 989 handler->dump(skb, cb, r, bc); 990 else 991 err = PTR_ERR(handler); 992 inet_diag_unlock_handler(handler); 993 994 return err ? : skb->len; 995 } 996 997 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb) 998 { 999 struct nlattr *bc = NULL; 1000 int hdrlen = sizeof(struct inet_diag_req_v2); 1001 1002 if (nlmsg_attrlen(cb->nlh, hdrlen)) 1003 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE); 1004 1005 return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh), bc); 1006 } 1007 1008 static inline int inet_diag_type2proto(int type) 1009 { 1010 switch (type) { 1011 case TCPDIAG_GETSOCK: 1012 return IPPROTO_TCP; 1013 case DCCPDIAG_GETSOCK: 1014 return IPPROTO_DCCP; 1015 default: 1016 return 0; 1017 } 1018 } 1019 1020 static int inet_diag_dump_compat(struct sk_buff *skb, struct netlink_callback *cb) 1021 { 1022 struct inet_diag_req *rc = nlmsg_data(cb->nlh); 1023 struct inet_diag_req_v2 req; 1024 struct nlattr *bc = NULL; 1025 int hdrlen = sizeof(struct inet_diag_req); 1026 1027 req.sdiag_family = AF_UNSPEC; /* compatibility */ 1028 req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type); 1029 req.idiag_ext = rc->idiag_ext; 1030 req.idiag_states = rc->idiag_states; 1031 req.id = rc->id; 1032 1033 if (nlmsg_attrlen(cb->nlh, hdrlen)) 1034 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE); 1035 1036 return __inet_diag_dump(skb, cb, &req, bc); 1037 } 1038 1039 static int inet_diag_get_exact_compat(struct sk_buff *in_skb, 1040 const struct nlmsghdr *nlh) 1041 { 1042 struct inet_diag_req *rc = nlmsg_data(nlh); 1043 struct inet_diag_req_v2 req; 1044 1045 req.sdiag_family = rc->idiag_family; 1046 req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type); 1047 req.idiag_ext = rc->idiag_ext; 1048 req.idiag_states = rc->idiag_states; 1049 req.id = rc->id; 1050 1051 return inet_diag_get_exact(in_skb, nlh, &req); 1052 } 1053 1054 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh) 1055 { 1056 int hdrlen = sizeof(struct inet_diag_req); 1057 struct net *net = sock_net(skb->sk); 1058 1059 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX || 1060 nlmsg_len(nlh) < hdrlen) 1061 return -EINVAL; 1062 1063 if (nlh->nlmsg_flags & NLM_F_DUMP) { 1064 if (nlmsg_attrlen(nlh, hdrlen)) { 1065 struct nlattr *attr; 1066 1067 attr = nlmsg_find_attr(nlh, hdrlen, 1068 INET_DIAG_REQ_BYTECODE); 1069 if (attr == NULL || 1070 nla_len(attr) < sizeof(struct inet_diag_bc_op) || 1071 inet_diag_bc_audit(nla_data(attr), nla_len(attr))) 1072 return -EINVAL; 1073 } 1074 { 1075 struct netlink_dump_control c = { 1076 .dump = inet_diag_dump_compat, 1077 }; 1078 return netlink_dump_start(net->diag_nlsk, skb, nlh, &c); 1079 } 1080 } 1081 1082 return inet_diag_get_exact_compat(skb, nlh); 1083 } 1084 1085 static int inet_diag_handler_dump(struct sk_buff *skb, struct nlmsghdr *h) 1086 { 1087 int hdrlen = sizeof(struct inet_diag_req_v2); 1088 struct net *net = sock_net(skb->sk); 1089 1090 if (nlmsg_len(h) < hdrlen) 1091 return -EINVAL; 1092 1093 if (h->nlmsg_flags & NLM_F_DUMP) { 1094 if (nlmsg_attrlen(h, hdrlen)) { 1095 struct nlattr *attr; 1096 attr = nlmsg_find_attr(h, hdrlen, 1097 INET_DIAG_REQ_BYTECODE); 1098 if (attr == NULL || 1099 nla_len(attr) < sizeof(struct inet_diag_bc_op) || 1100 inet_diag_bc_audit(nla_data(attr), nla_len(attr))) 1101 return -EINVAL; 1102 } 1103 { 1104 struct netlink_dump_control c = { 1105 .dump = inet_diag_dump, 1106 }; 1107 return netlink_dump_start(net->diag_nlsk, skb, h, &c); 1108 } 1109 } 1110 1111 return inet_diag_get_exact(skb, h, nlmsg_data(h)); 1112 } 1113 1114 static const struct sock_diag_handler inet_diag_handler = { 1115 .family = AF_INET, 1116 .dump = inet_diag_handler_dump, 1117 }; 1118 1119 static const struct sock_diag_handler inet6_diag_handler = { 1120 .family = AF_INET6, 1121 .dump = inet_diag_handler_dump, 1122 }; 1123 1124 int inet_diag_register(const struct inet_diag_handler *h) 1125 { 1126 const __u16 type = h->idiag_type; 1127 int err = -EINVAL; 1128 1129 if (type >= IPPROTO_MAX) 1130 goto out; 1131 1132 mutex_lock(&inet_diag_table_mutex); 1133 err = -EEXIST; 1134 if (inet_diag_table[type] == NULL) { 1135 inet_diag_table[type] = h; 1136 err = 0; 1137 } 1138 mutex_unlock(&inet_diag_table_mutex); 1139 out: 1140 return err; 1141 } 1142 EXPORT_SYMBOL_GPL(inet_diag_register); 1143 1144 void inet_diag_unregister(const struct inet_diag_handler *h) 1145 { 1146 const __u16 type = h->idiag_type; 1147 1148 if (type >= IPPROTO_MAX) 1149 return; 1150 1151 mutex_lock(&inet_diag_table_mutex); 1152 inet_diag_table[type] = NULL; 1153 mutex_unlock(&inet_diag_table_mutex); 1154 } 1155 EXPORT_SYMBOL_GPL(inet_diag_unregister); 1156 1157 static int __init inet_diag_init(void) 1158 { 1159 const int inet_diag_table_size = (IPPROTO_MAX * 1160 sizeof(struct inet_diag_handler *)); 1161 int err = -ENOMEM; 1162 1163 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL); 1164 if (!inet_diag_table) 1165 goto out; 1166 1167 err = sock_diag_register(&inet_diag_handler); 1168 if (err) 1169 goto out_free_nl; 1170 1171 err = sock_diag_register(&inet6_diag_handler); 1172 if (err) 1173 goto out_free_inet; 1174 1175 sock_diag_register_inet_compat(inet_diag_rcv_msg_compat); 1176 out: 1177 return err; 1178 1179 out_free_inet: 1180 sock_diag_unregister(&inet_diag_handler); 1181 out_free_nl: 1182 kfree(inet_diag_table); 1183 goto out; 1184 } 1185 1186 static void __exit inet_diag_exit(void) 1187 { 1188 sock_diag_unregister(&inet6_diag_handler); 1189 sock_diag_unregister(&inet_diag_handler); 1190 sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat); 1191 kfree(inet_diag_table); 1192 } 1193 1194 module_init(inet_diag_init); 1195 module_exit(inet_diag_exit); 1196 MODULE_LICENSE("GPL"); 1197 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */); 1198 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */); 1199