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 #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 void inet_diag_unlock_handler(const struct inet_diag_handler *handler) 69 { 70 mutex_unlock(&inet_diag_table_mutex); 71 } 72 73 int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk, 74 struct sk_buff *skb, struct inet_diag_req_v2 *req, 75 struct user_namespace *user_ns, 76 u32 portid, u32 seq, u16 nlmsg_flags, 77 const struct nlmsghdr *unlh) 78 { 79 const struct inet_sock *inet = inet_sk(sk); 80 const struct inet_diag_handler *handler; 81 int ext = req->idiag_ext; 82 struct inet_diag_msg *r; 83 struct nlmsghdr *nlh; 84 struct nlattr *attr; 85 void *info = NULL; 86 87 handler = inet_diag_table[req->sdiag_protocol]; 88 BUG_ON(!handler); 89 90 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r), 91 nlmsg_flags); 92 if (!nlh) 93 return -EMSGSIZE; 94 95 r = nlmsg_data(nlh); 96 BUG_ON(sk->sk_state == TCP_TIME_WAIT); 97 98 r->idiag_family = sk->sk_family; 99 r->idiag_state = sk->sk_state; 100 r->idiag_timer = 0; 101 r->idiag_retrans = 0; 102 103 r->id.idiag_if = sk->sk_bound_dev_if; 104 sock_diag_save_cookie(sk, r->id.idiag_cookie); 105 106 r->id.idiag_sport = inet->inet_sport; 107 r->id.idiag_dport = inet->inet_dport; 108 109 memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src)); 110 memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst)); 111 112 r->id.idiag_src[0] = inet->inet_rcv_saddr; 113 r->id.idiag_dst[0] = inet->inet_daddr; 114 115 if (nla_put_u8(skb, INET_DIAG_SHUTDOWN, sk->sk_shutdown)) 116 goto errout; 117 118 /* IPv6 dual-stack sockets use inet->tos for IPv4 connections, 119 * hence this needs to be included regardless of socket family. 120 */ 121 if (ext & (1 << (INET_DIAG_TOS - 1))) 122 if (nla_put_u8(skb, INET_DIAG_TOS, inet->tos) < 0) 123 goto errout; 124 125 #if IS_ENABLED(CONFIG_IPV6) 126 if (r->idiag_family == AF_INET6) { 127 *(struct in6_addr *)r->id.idiag_src = sk->sk_v6_rcv_saddr; 128 *(struct in6_addr *)r->id.idiag_dst = sk->sk_v6_daddr; 129 130 if (ext & (1 << (INET_DIAG_TCLASS - 1))) 131 if (nla_put_u8(skb, INET_DIAG_TCLASS, 132 inet6_sk(sk)->tclass) < 0) 133 goto errout; 134 } 135 #endif 136 137 r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk)); 138 r->idiag_inode = sock_i_ino(sk); 139 140 if (ext & (1 << (INET_DIAG_MEMINFO - 1))) { 141 struct inet_diag_meminfo minfo = { 142 .idiag_rmem = sk_rmem_alloc_get(sk), 143 .idiag_wmem = sk->sk_wmem_queued, 144 .idiag_fmem = sk->sk_forward_alloc, 145 .idiag_tmem = sk_wmem_alloc_get(sk), 146 }; 147 148 if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0) 149 goto errout; 150 } 151 152 if (ext & (1 << (INET_DIAG_SKMEMINFO - 1))) 153 if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO)) 154 goto errout; 155 156 if (!icsk) { 157 handler->idiag_get_info(sk, r, NULL); 158 goto out; 159 } 160 161 #define EXPIRES_IN_MS(tmo) DIV_ROUND_UP((tmo - jiffies) * 1000, HZ) 162 163 if (icsk->icsk_pending == ICSK_TIME_RETRANS || 164 icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS || 165 icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) { 166 r->idiag_timer = 1; 167 r->idiag_retrans = icsk->icsk_retransmits; 168 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout); 169 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { 170 r->idiag_timer = 4; 171 r->idiag_retrans = icsk->icsk_probes_out; 172 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout); 173 } else if (timer_pending(&sk->sk_timer)) { 174 r->idiag_timer = 2; 175 r->idiag_retrans = icsk->icsk_probes_out; 176 r->idiag_expires = EXPIRES_IN_MS(sk->sk_timer.expires); 177 } else { 178 r->idiag_timer = 0; 179 r->idiag_expires = 0; 180 } 181 #undef EXPIRES_IN_MS 182 183 if (ext & (1 << (INET_DIAG_INFO - 1))) { 184 attr = nla_reserve(skb, INET_DIAG_INFO, 185 sizeof(struct tcp_info)); 186 if (!attr) 187 goto errout; 188 189 info = nla_data(attr); 190 } 191 192 if ((ext & (1 << (INET_DIAG_CONG - 1))) && icsk->icsk_ca_ops) 193 if (nla_put_string(skb, INET_DIAG_CONG, 194 icsk->icsk_ca_ops->name) < 0) 195 goto errout; 196 197 handler->idiag_get_info(sk, r, info); 198 199 if (sk->sk_state < TCP_TIME_WAIT && 200 icsk->icsk_ca_ops && icsk->icsk_ca_ops->get_info) 201 icsk->icsk_ca_ops->get_info(sk, ext, skb); 202 203 out: 204 nlmsg_end(skb, nlh); 205 return 0; 206 207 errout: 208 nlmsg_cancel(skb, nlh); 209 return -EMSGSIZE; 210 } 211 EXPORT_SYMBOL_GPL(inet_sk_diag_fill); 212 213 static int inet_csk_diag_fill(struct sock *sk, 214 struct sk_buff *skb, 215 struct inet_diag_req_v2 *req, 216 struct user_namespace *user_ns, 217 u32 portid, u32 seq, u16 nlmsg_flags, 218 const struct nlmsghdr *unlh) 219 { 220 return inet_sk_diag_fill(sk, inet_csk(sk), skb, req, 221 user_ns, portid, seq, nlmsg_flags, unlh); 222 } 223 224 static int inet_twsk_diag_fill(struct inet_timewait_sock *tw, 225 struct sk_buff *skb, 226 struct inet_diag_req_v2 *req, 227 u32 portid, u32 seq, u16 nlmsg_flags, 228 const struct nlmsghdr *unlh) 229 { 230 struct inet_diag_msg *r; 231 struct nlmsghdr *nlh; 232 s32 tmo; 233 234 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r), 235 nlmsg_flags); 236 if (!nlh) 237 return -EMSGSIZE; 238 239 r = nlmsg_data(nlh); 240 BUG_ON(tw->tw_state != TCP_TIME_WAIT); 241 242 tmo = tw->tw_ttd - inet_tw_time_stamp(); 243 if (tmo < 0) 244 tmo = 0; 245 246 r->idiag_family = tw->tw_family; 247 r->idiag_retrans = 0; 248 249 r->id.idiag_if = tw->tw_bound_dev_if; 250 sock_diag_save_cookie(tw, r->id.idiag_cookie); 251 252 r->id.idiag_sport = tw->tw_sport; 253 r->id.idiag_dport = tw->tw_dport; 254 255 memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src)); 256 memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst)); 257 258 r->id.idiag_src[0] = tw->tw_rcv_saddr; 259 r->id.idiag_dst[0] = tw->tw_daddr; 260 261 r->idiag_state = tw->tw_substate; 262 r->idiag_timer = 3; 263 r->idiag_expires = jiffies_to_msecs(tmo); 264 r->idiag_rqueue = 0; 265 r->idiag_wqueue = 0; 266 r->idiag_uid = 0; 267 r->idiag_inode = 0; 268 #if IS_ENABLED(CONFIG_IPV6) 269 if (tw->tw_family == AF_INET6) { 270 *(struct in6_addr *)r->id.idiag_src = tw->tw_v6_rcv_saddr; 271 *(struct in6_addr *)r->id.idiag_dst = tw->tw_v6_daddr; 272 } 273 #endif 274 275 nlmsg_end(skb, nlh); 276 return 0; 277 } 278 279 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb, 280 struct inet_diag_req_v2 *r, 281 struct user_namespace *user_ns, 282 u32 portid, u32 seq, u16 nlmsg_flags, 283 const struct nlmsghdr *unlh) 284 { 285 if (sk->sk_state == TCP_TIME_WAIT) 286 return inet_twsk_diag_fill(inet_twsk(sk), skb, r, portid, seq, 287 nlmsg_flags, unlh); 288 289 return inet_csk_diag_fill(sk, skb, r, user_ns, portid, seq, 290 nlmsg_flags, unlh); 291 } 292 293 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo, 294 struct sk_buff *in_skb, 295 const struct nlmsghdr *nlh, 296 struct inet_diag_req_v2 *req) 297 { 298 struct net *net = sock_net(in_skb->sk); 299 struct sk_buff *rep; 300 struct sock *sk; 301 int err; 302 303 err = -EINVAL; 304 if (req->sdiag_family == AF_INET) 305 sk = inet_lookup(net, hashinfo, req->id.idiag_dst[0], 306 req->id.idiag_dport, req->id.idiag_src[0], 307 req->id.idiag_sport, req->id.idiag_if); 308 #if IS_ENABLED(CONFIG_IPV6) 309 else if (req->sdiag_family == AF_INET6) 310 sk = inet6_lookup(net, hashinfo, 311 (struct in6_addr *)req->id.idiag_dst, 312 req->id.idiag_dport, 313 (struct in6_addr *)req->id.idiag_src, 314 req->id.idiag_sport, 315 req->id.idiag_if); 316 #endif 317 else 318 goto out_nosk; 319 320 err = -ENOENT; 321 if (!sk) 322 goto out_nosk; 323 324 err = sock_diag_check_cookie(sk, req->id.idiag_cookie); 325 if (err) 326 goto out; 327 328 rep = nlmsg_new(sizeof(struct inet_diag_msg) + 329 sizeof(struct inet_diag_meminfo) + 330 sizeof(struct tcp_info) + 64, GFP_KERNEL); 331 if (!rep) { 332 err = -ENOMEM; 333 goto out; 334 } 335 336 err = sk_diag_fill(sk, rep, req, 337 sk_user_ns(NETLINK_CB(in_skb).sk), 338 NETLINK_CB(in_skb).portid, 339 nlh->nlmsg_seq, 0, nlh); 340 if (err < 0) { 341 WARN_ON(err == -EMSGSIZE); 342 nlmsg_free(rep); 343 goto out; 344 } 345 err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid, 346 MSG_DONTWAIT); 347 if (err > 0) 348 err = 0; 349 350 out: 351 if (sk) 352 sock_gen_put(sk); 353 354 out_nosk: 355 return err; 356 } 357 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk); 358 359 static int inet_diag_get_exact(struct sk_buff *in_skb, 360 const struct nlmsghdr *nlh, 361 struct inet_diag_req_v2 *req) 362 { 363 const struct inet_diag_handler *handler; 364 int err; 365 366 handler = inet_diag_lock_handler(req->sdiag_protocol); 367 if (IS_ERR(handler)) 368 err = PTR_ERR(handler); 369 else 370 err = handler->dump_one(in_skb, nlh, req); 371 inet_diag_unlock_handler(handler); 372 373 return err; 374 } 375 376 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits) 377 { 378 int words = bits >> 5; 379 380 bits &= 0x1f; 381 382 if (words) { 383 if (memcmp(a1, a2, words << 2)) 384 return 0; 385 } 386 if (bits) { 387 __be32 w1, w2; 388 __be32 mask; 389 390 w1 = a1[words]; 391 w2 = a2[words]; 392 393 mask = htonl((0xffffffff) << (32 - bits)); 394 395 if ((w1 ^ w2) & mask) 396 return 0; 397 } 398 399 return 1; 400 } 401 402 static int inet_diag_bc_run(const struct nlattr *_bc, 403 const struct inet_diag_entry *entry) 404 { 405 const void *bc = nla_data(_bc); 406 int len = nla_len(_bc); 407 408 while (len > 0) { 409 int yes = 1; 410 const struct inet_diag_bc_op *op = bc; 411 412 switch (op->code) { 413 case INET_DIAG_BC_NOP: 414 break; 415 case INET_DIAG_BC_JMP: 416 yes = 0; 417 break; 418 case INET_DIAG_BC_S_GE: 419 yes = entry->sport >= op[1].no; 420 break; 421 case INET_DIAG_BC_S_LE: 422 yes = entry->sport <= op[1].no; 423 break; 424 case INET_DIAG_BC_D_GE: 425 yes = entry->dport >= op[1].no; 426 break; 427 case INET_DIAG_BC_D_LE: 428 yes = entry->dport <= op[1].no; 429 break; 430 case INET_DIAG_BC_AUTO: 431 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK); 432 break; 433 case INET_DIAG_BC_S_COND: 434 case INET_DIAG_BC_D_COND: { 435 const struct inet_diag_hostcond *cond; 436 const __be32 *addr; 437 438 cond = (const struct inet_diag_hostcond *)(op + 1); 439 if (cond->port != -1 && 440 cond->port != (op->code == INET_DIAG_BC_S_COND ? 441 entry->sport : entry->dport)) { 442 yes = 0; 443 break; 444 } 445 446 if (op->code == INET_DIAG_BC_S_COND) 447 addr = entry->saddr; 448 else 449 addr = entry->daddr; 450 451 if (cond->family != AF_UNSPEC && 452 cond->family != entry->family) { 453 if (entry->family == AF_INET6 && 454 cond->family == AF_INET) { 455 if (addr[0] == 0 && addr[1] == 0 && 456 addr[2] == htonl(0xffff) && 457 bitstring_match(addr + 3, 458 cond->addr, 459 cond->prefix_len)) 460 break; 461 } 462 yes = 0; 463 break; 464 } 465 466 if (cond->prefix_len == 0) 467 break; 468 if (bitstring_match(addr, cond->addr, 469 cond->prefix_len)) 470 break; 471 yes = 0; 472 break; 473 } 474 } 475 476 if (yes) { 477 len -= op->yes; 478 bc += op->yes; 479 } else { 480 len -= op->no; 481 bc += op->no; 482 } 483 } 484 return len == 0; 485 } 486 487 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk) 488 { 489 struct inet_sock *inet = inet_sk(sk); 490 struct inet_diag_entry entry; 491 492 if (!bc) 493 return 1; 494 495 entry.family = sk->sk_family; 496 #if IS_ENABLED(CONFIG_IPV6) 497 if (entry.family == AF_INET6) { 498 entry.saddr = sk->sk_v6_rcv_saddr.s6_addr32; 499 entry.daddr = sk->sk_v6_daddr.s6_addr32; 500 } else 501 #endif 502 { 503 entry.saddr = &inet->inet_rcv_saddr; 504 entry.daddr = &inet->inet_daddr; 505 } 506 entry.sport = inet->inet_num; 507 entry.dport = ntohs(inet->inet_dport); 508 entry.userlocks = (sk->sk_state != TCP_TIME_WAIT) ? sk->sk_userlocks : 0; 509 510 return inet_diag_bc_run(bc, &entry); 511 } 512 EXPORT_SYMBOL_GPL(inet_diag_bc_sk); 513 514 static int valid_cc(const void *bc, int len, int cc) 515 { 516 while (len >= 0) { 517 const struct inet_diag_bc_op *op = bc; 518 519 if (cc > len) 520 return 0; 521 if (cc == len) 522 return 1; 523 if (op->yes < 4 || op->yes & 3) 524 return 0; 525 len -= op->yes; 526 bc += op->yes; 527 } 528 return 0; 529 } 530 531 /* Validate an inet_diag_hostcond. */ 532 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len, 533 int *min_len) 534 { 535 struct inet_diag_hostcond *cond; 536 int addr_len; 537 538 /* Check hostcond space. */ 539 *min_len += sizeof(struct inet_diag_hostcond); 540 if (len < *min_len) 541 return false; 542 cond = (struct inet_diag_hostcond *)(op + 1); 543 544 /* Check address family and address length. */ 545 switch (cond->family) { 546 case AF_UNSPEC: 547 addr_len = 0; 548 break; 549 case AF_INET: 550 addr_len = sizeof(struct in_addr); 551 break; 552 case AF_INET6: 553 addr_len = sizeof(struct in6_addr); 554 break; 555 default: 556 return false; 557 } 558 *min_len += addr_len; 559 if (len < *min_len) 560 return false; 561 562 /* Check prefix length (in bits) vs address length (in bytes). */ 563 if (cond->prefix_len > 8 * addr_len) 564 return false; 565 566 return true; 567 } 568 569 /* Validate a port comparison operator. */ 570 static bool valid_port_comparison(const struct inet_diag_bc_op *op, 571 int len, int *min_len) 572 { 573 /* Port comparisons put the port in a follow-on inet_diag_bc_op. */ 574 *min_len += sizeof(struct inet_diag_bc_op); 575 if (len < *min_len) 576 return false; 577 return true; 578 } 579 580 static int inet_diag_bc_audit(const void *bytecode, int bytecode_len) 581 { 582 const void *bc = bytecode; 583 int len = bytecode_len; 584 585 while (len > 0) { 586 int min_len = sizeof(struct inet_diag_bc_op); 587 const struct inet_diag_bc_op *op = bc; 588 589 switch (op->code) { 590 case INET_DIAG_BC_S_COND: 591 case INET_DIAG_BC_D_COND: 592 if (!valid_hostcond(bc, len, &min_len)) 593 return -EINVAL; 594 break; 595 case INET_DIAG_BC_S_GE: 596 case INET_DIAG_BC_S_LE: 597 case INET_DIAG_BC_D_GE: 598 case INET_DIAG_BC_D_LE: 599 if (!valid_port_comparison(bc, len, &min_len)) 600 return -EINVAL; 601 break; 602 case INET_DIAG_BC_AUTO: 603 case INET_DIAG_BC_JMP: 604 case INET_DIAG_BC_NOP: 605 break; 606 default: 607 return -EINVAL; 608 } 609 610 if (op->code != INET_DIAG_BC_NOP) { 611 if (op->no < min_len || op->no > len + 4 || op->no & 3) 612 return -EINVAL; 613 if (op->no < len && 614 !valid_cc(bytecode, bytecode_len, len - op->no)) 615 return -EINVAL; 616 } 617 618 if (op->yes < min_len || op->yes > len + 4 || op->yes & 3) 619 return -EINVAL; 620 bc += op->yes; 621 len -= op->yes; 622 } 623 return len == 0 ? 0 : -EINVAL; 624 } 625 626 static int inet_csk_diag_dump(struct sock *sk, 627 struct sk_buff *skb, 628 struct netlink_callback *cb, 629 struct inet_diag_req_v2 *r, 630 const struct nlattr *bc) 631 { 632 if (!inet_diag_bc_sk(bc, sk)) 633 return 0; 634 635 return inet_csk_diag_fill(sk, skb, r, 636 sk_user_ns(NETLINK_CB(cb->skb).sk), 637 NETLINK_CB(cb->skb).portid, 638 cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh); 639 } 640 641 static void twsk_build_assert(void) 642 { 643 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) != 644 offsetof(struct sock, sk_family)); 645 646 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) != 647 offsetof(struct inet_sock, inet_num)); 648 649 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) != 650 offsetof(struct inet_sock, inet_dport)); 651 652 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) != 653 offsetof(struct inet_sock, inet_rcv_saddr)); 654 655 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) != 656 offsetof(struct inet_sock, inet_daddr)); 657 658 #if IS_ENABLED(CONFIG_IPV6) 659 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) != 660 offsetof(struct sock, sk_v6_rcv_saddr)); 661 662 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) != 663 offsetof(struct sock, sk_v6_daddr)); 664 #endif 665 } 666 667 static int inet_twsk_diag_dump(struct sock *sk, 668 struct sk_buff *skb, 669 struct netlink_callback *cb, 670 struct inet_diag_req_v2 *r, 671 const struct nlattr *bc) 672 { 673 twsk_build_assert(); 674 675 if (!inet_diag_bc_sk(bc, sk)) 676 return 0; 677 678 return inet_twsk_diag_fill(inet_twsk(sk), skb, r, 679 NETLINK_CB(cb->skb).portid, 680 cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh); 681 } 682 683 /* Get the IPv4, IPv6, or IPv4-mapped-IPv6 local and remote addresses 684 * from a request_sock. For IPv4-mapped-IPv6 we must map IPv4 to IPv6. 685 */ 686 static void inet_diag_req_addrs(const struct sock *sk, 687 const struct request_sock *req, 688 struct inet_diag_entry *entry) 689 { 690 const struct inet_request_sock *ireq = inet_rsk(req); 691 692 #if IS_ENABLED(CONFIG_IPV6) 693 if (sk->sk_family == AF_INET6) { 694 if (req->rsk_ops->family == AF_INET6) { 695 entry->saddr = ireq->ir_v6_loc_addr.s6_addr32; 696 entry->daddr = ireq->ir_v6_rmt_addr.s6_addr32; 697 } else if (req->rsk_ops->family == AF_INET) { 698 ipv6_addr_set_v4mapped(ireq->ir_loc_addr, 699 &entry->saddr_storage); 700 ipv6_addr_set_v4mapped(ireq->ir_rmt_addr, 701 &entry->daddr_storage); 702 entry->saddr = entry->saddr_storage.s6_addr32; 703 entry->daddr = entry->daddr_storage.s6_addr32; 704 } 705 } else 706 #endif 707 { 708 entry->saddr = &ireq->ir_loc_addr; 709 entry->daddr = &ireq->ir_rmt_addr; 710 } 711 } 712 713 static int inet_diag_fill_req(struct sk_buff *skb, struct sock *sk, 714 struct request_sock *req, 715 struct user_namespace *user_ns, 716 u32 portid, u32 seq, 717 const struct nlmsghdr *unlh) 718 { 719 const struct inet_request_sock *ireq = inet_rsk(req); 720 struct inet_sock *inet = inet_sk(sk); 721 struct inet_diag_msg *r; 722 struct nlmsghdr *nlh; 723 long tmo; 724 725 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r), 726 NLM_F_MULTI); 727 if (!nlh) 728 return -EMSGSIZE; 729 730 r = nlmsg_data(nlh); 731 r->idiag_family = sk->sk_family; 732 r->idiag_state = TCP_SYN_RECV; 733 r->idiag_timer = 1; 734 r->idiag_retrans = req->num_retrans; 735 736 r->id.idiag_if = sk->sk_bound_dev_if; 737 sock_diag_save_cookie(req, r->id.idiag_cookie); 738 739 tmo = req->expires - jiffies; 740 if (tmo < 0) 741 tmo = 0; 742 743 r->id.idiag_sport = inet->inet_sport; 744 r->id.idiag_dport = ireq->ir_rmt_port; 745 746 memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src)); 747 memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst)); 748 749 r->id.idiag_src[0] = ireq->ir_loc_addr; 750 r->id.idiag_dst[0] = ireq->ir_rmt_addr; 751 752 r->idiag_expires = jiffies_to_msecs(tmo); 753 r->idiag_rqueue = 0; 754 r->idiag_wqueue = 0; 755 r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk)); 756 r->idiag_inode = 0; 757 #if IS_ENABLED(CONFIG_IPV6) 758 if (r->idiag_family == AF_INET6) { 759 struct inet_diag_entry entry; 760 761 inet_diag_req_addrs(sk, req, &entry); 762 memcpy(r->id.idiag_src, entry.saddr, sizeof(struct in6_addr)); 763 memcpy(r->id.idiag_dst, entry.daddr, sizeof(struct in6_addr)); 764 } 765 #endif 766 767 nlmsg_end(skb, nlh); 768 return 0; 769 } 770 771 static int inet_diag_dump_reqs(struct sk_buff *skb, struct sock *sk, 772 struct netlink_callback *cb, 773 struct inet_diag_req_v2 *r, 774 const struct nlattr *bc) 775 { 776 struct inet_connection_sock *icsk = inet_csk(sk); 777 struct inet_sock *inet = inet_sk(sk); 778 struct inet_diag_entry entry; 779 int j, s_j, reqnum, s_reqnum; 780 struct listen_sock *lopt; 781 int err = 0; 782 783 s_j = cb->args[3]; 784 s_reqnum = cb->args[4]; 785 786 if (s_j > 0) 787 s_j--; 788 789 entry.family = sk->sk_family; 790 791 read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock); 792 793 lopt = icsk->icsk_accept_queue.listen_opt; 794 if (!lopt || !lopt->qlen) 795 goto out; 796 797 if (bc) { 798 entry.sport = inet->inet_num; 799 entry.userlocks = sk->sk_userlocks; 800 } 801 802 for (j = s_j; j < lopt->nr_table_entries; j++) { 803 struct request_sock *req, *head = lopt->syn_table[j]; 804 805 reqnum = 0; 806 for (req = head; req; reqnum++, req = req->dl_next) { 807 struct inet_request_sock *ireq = inet_rsk(req); 808 809 if (reqnum < s_reqnum) 810 continue; 811 if (r->id.idiag_dport != ireq->ir_rmt_port && 812 r->id.idiag_dport) 813 continue; 814 815 if (bc) { 816 inet_diag_req_addrs(sk, req, &entry); 817 entry.dport = ntohs(ireq->ir_rmt_port); 818 819 if (!inet_diag_bc_run(bc, &entry)) 820 continue; 821 } 822 823 err = inet_diag_fill_req(skb, sk, req, 824 sk_user_ns(NETLINK_CB(cb->skb).sk), 825 NETLINK_CB(cb->skb).portid, 826 cb->nlh->nlmsg_seq, cb->nlh); 827 if (err < 0) { 828 cb->args[3] = j + 1; 829 cb->args[4] = reqnum; 830 goto out; 831 } 832 } 833 834 s_reqnum = 0; 835 } 836 837 out: 838 read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); 839 840 return err; 841 } 842 843 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb, 844 struct netlink_callback *cb, 845 struct inet_diag_req_v2 *r, struct nlattr *bc) 846 { 847 struct net *net = sock_net(skb->sk); 848 int i, num, s_i, s_num; 849 850 s_i = cb->args[1]; 851 s_num = num = cb->args[2]; 852 853 if (cb->args[0] == 0) { 854 if (!(r->idiag_states & (TCPF_LISTEN | TCPF_SYN_RECV))) 855 goto skip_listen_ht; 856 857 for (i = s_i; i < INET_LHTABLE_SIZE; i++) { 858 struct inet_listen_hashbucket *ilb; 859 struct hlist_nulls_node *node; 860 struct sock *sk; 861 862 num = 0; 863 ilb = &hashinfo->listening_hash[i]; 864 spin_lock_bh(&ilb->lock); 865 sk_nulls_for_each(sk, node, &ilb->head) { 866 struct inet_sock *inet = inet_sk(sk); 867 868 if (!net_eq(sock_net(sk), net)) 869 continue; 870 871 if (num < s_num) { 872 num++; 873 continue; 874 } 875 876 if (r->sdiag_family != AF_UNSPEC && 877 sk->sk_family != r->sdiag_family) 878 goto next_listen; 879 880 if (r->id.idiag_sport != inet->inet_sport && 881 r->id.idiag_sport) 882 goto next_listen; 883 884 if (!(r->idiag_states & TCPF_LISTEN) || 885 r->id.idiag_dport || 886 cb->args[3] > 0) 887 goto syn_recv; 888 889 if (inet_csk_diag_dump(sk, skb, cb, r, bc) < 0) { 890 spin_unlock_bh(&ilb->lock); 891 goto done; 892 } 893 894 syn_recv: 895 if (!(r->idiag_states & TCPF_SYN_RECV)) 896 goto next_listen; 897 898 if (inet_diag_dump_reqs(skb, sk, cb, r, bc) < 0) { 899 spin_unlock_bh(&ilb->lock); 900 goto done; 901 } 902 903 next_listen: 904 cb->args[3] = 0; 905 cb->args[4] = 0; 906 ++num; 907 } 908 spin_unlock_bh(&ilb->lock); 909 910 s_num = 0; 911 cb->args[3] = 0; 912 cb->args[4] = 0; 913 } 914 skip_listen_ht: 915 cb->args[0] = 1; 916 s_i = num = s_num = 0; 917 } 918 919 if (!(r->idiag_states & ~(TCPF_LISTEN | TCPF_SYN_RECV))) 920 goto out; 921 922 for (i = s_i; i <= hashinfo->ehash_mask; i++) { 923 struct inet_ehash_bucket *head = &hashinfo->ehash[i]; 924 spinlock_t *lock = inet_ehash_lockp(hashinfo, i); 925 struct hlist_nulls_node *node; 926 struct sock *sk; 927 928 num = 0; 929 930 if (hlist_nulls_empty(&head->chain)) 931 continue; 932 933 if (i > s_i) 934 s_num = 0; 935 936 spin_lock_bh(lock); 937 sk_nulls_for_each(sk, node, &head->chain) { 938 int state, res; 939 940 if (!net_eq(sock_net(sk), net)) 941 continue; 942 if (num < s_num) 943 goto next_normal; 944 state = (sk->sk_state == TCP_TIME_WAIT) ? 945 inet_twsk(sk)->tw_substate : sk->sk_state; 946 if (!(r->idiag_states & (1 << state))) 947 goto next_normal; 948 if (r->sdiag_family != AF_UNSPEC && 949 sk->sk_family != r->sdiag_family) 950 goto next_normal; 951 if (r->id.idiag_sport != htons(sk->sk_num) && 952 r->id.idiag_sport) 953 goto next_normal; 954 if (r->id.idiag_dport != sk->sk_dport && 955 r->id.idiag_dport) 956 goto next_normal; 957 if (sk->sk_state == TCP_TIME_WAIT) 958 res = inet_twsk_diag_dump(sk, skb, cb, r, bc); 959 else 960 res = inet_csk_diag_dump(sk, skb, cb, r, bc); 961 if (res < 0) { 962 spin_unlock_bh(lock); 963 goto done; 964 } 965 next_normal: 966 ++num; 967 } 968 969 spin_unlock_bh(lock); 970 } 971 972 done: 973 cb->args[1] = i; 974 cb->args[2] = num; 975 out: 976 ; 977 } 978 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk); 979 980 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb, 981 struct inet_diag_req_v2 *r, 982 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 int hdrlen = sizeof(struct inet_diag_req_v2); 1000 struct nlattr *bc = NULL; 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 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, 1021 struct netlink_callback *cb) 1022 { 1023 struct inet_diag_req *rc = nlmsg_data(cb->nlh); 1024 int hdrlen = sizeof(struct inet_diag_req); 1025 struct inet_diag_req_v2 req; 1026 struct nlattr *bc = NULL; 1027 1028 req.sdiag_family = AF_UNSPEC; /* compatibility */ 1029 req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type); 1030 req.idiag_ext = rc->idiag_ext; 1031 req.idiag_states = rc->idiag_states; 1032 req.id = rc->id; 1033 1034 if (nlmsg_attrlen(cb->nlh, hdrlen)) 1035 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE); 1036 1037 return __inet_diag_dump(skb, cb, &req, bc); 1038 } 1039 1040 static int inet_diag_get_exact_compat(struct sk_buff *in_skb, 1041 const struct nlmsghdr *nlh) 1042 { 1043 struct inet_diag_req *rc = nlmsg_data(nlh); 1044 struct inet_diag_req_v2 req; 1045 1046 req.sdiag_family = rc->idiag_family; 1047 req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type); 1048 req.idiag_ext = rc->idiag_ext; 1049 req.idiag_states = rc->idiag_states; 1050 req.id = rc->id; 1051 1052 return inet_diag_get_exact(in_skb, nlh, &req); 1053 } 1054 1055 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh) 1056 { 1057 int hdrlen = sizeof(struct inet_diag_req); 1058 struct net *net = sock_net(skb->sk); 1059 1060 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX || 1061 nlmsg_len(nlh) < hdrlen) 1062 return -EINVAL; 1063 1064 if (nlh->nlmsg_flags & NLM_F_DUMP) { 1065 if (nlmsg_attrlen(nlh, hdrlen)) { 1066 struct nlattr *attr; 1067 1068 attr = nlmsg_find_attr(nlh, hdrlen, 1069 INET_DIAG_REQ_BYTECODE); 1070 if (!attr || 1071 nla_len(attr) < sizeof(struct inet_diag_bc_op) || 1072 inet_diag_bc_audit(nla_data(attr), nla_len(attr))) 1073 return -EINVAL; 1074 } 1075 { 1076 struct netlink_dump_control c = { 1077 .dump = inet_diag_dump_compat, 1078 }; 1079 return netlink_dump_start(net->diag_nlsk, skb, nlh, &c); 1080 } 1081 } 1082 1083 return inet_diag_get_exact_compat(skb, nlh); 1084 } 1085 1086 static int inet_diag_handler_dump(struct sk_buff *skb, struct nlmsghdr *h) 1087 { 1088 int hdrlen = sizeof(struct inet_diag_req_v2); 1089 struct net *net = sock_net(skb->sk); 1090 1091 if (nlmsg_len(h) < hdrlen) 1092 return -EINVAL; 1093 1094 if (h->nlmsg_flags & NLM_F_DUMP) { 1095 if (nlmsg_attrlen(h, hdrlen)) { 1096 struct nlattr *attr; 1097 1098 attr = nlmsg_find_attr(h, hdrlen, 1099 INET_DIAG_REQ_BYTECODE); 1100 if (!attr || 1101 nla_len(attr) < sizeof(struct inet_diag_bc_op) || 1102 inet_diag_bc_audit(nla_data(attr), nla_len(attr))) 1103 return -EINVAL; 1104 } 1105 { 1106 struct netlink_dump_control c = { 1107 .dump = inet_diag_dump, 1108 }; 1109 return netlink_dump_start(net->diag_nlsk, skb, h, &c); 1110 } 1111 } 1112 1113 return inet_diag_get_exact(skb, h, nlmsg_data(h)); 1114 } 1115 1116 static const struct sock_diag_handler inet_diag_handler = { 1117 .family = AF_INET, 1118 .dump = inet_diag_handler_dump, 1119 }; 1120 1121 static const struct sock_diag_handler inet6_diag_handler = { 1122 .family = AF_INET6, 1123 .dump = inet_diag_handler_dump, 1124 }; 1125 1126 int inet_diag_register(const struct inet_diag_handler *h) 1127 { 1128 const __u16 type = h->idiag_type; 1129 int err = -EINVAL; 1130 1131 if (type >= IPPROTO_MAX) 1132 goto out; 1133 1134 mutex_lock(&inet_diag_table_mutex); 1135 err = -EEXIST; 1136 if (!inet_diag_table[type]) { 1137 inet_diag_table[type] = h; 1138 err = 0; 1139 } 1140 mutex_unlock(&inet_diag_table_mutex); 1141 out: 1142 return err; 1143 } 1144 EXPORT_SYMBOL_GPL(inet_diag_register); 1145 1146 void inet_diag_unregister(const struct inet_diag_handler *h) 1147 { 1148 const __u16 type = h->idiag_type; 1149 1150 if (type >= IPPROTO_MAX) 1151 return; 1152 1153 mutex_lock(&inet_diag_table_mutex); 1154 inet_diag_table[type] = NULL; 1155 mutex_unlock(&inet_diag_table_mutex); 1156 } 1157 EXPORT_SYMBOL_GPL(inet_diag_unregister); 1158 1159 static int __init inet_diag_init(void) 1160 { 1161 const int inet_diag_table_size = (IPPROTO_MAX * 1162 sizeof(struct inet_diag_handler *)); 1163 int err = -ENOMEM; 1164 1165 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL); 1166 if (!inet_diag_table) 1167 goto out; 1168 1169 err = sock_diag_register(&inet_diag_handler); 1170 if (err) 1171 goto out_free_nl; 1172 1173 err = sock_diag_register(&inet6_diag_handler); 1174 if (err) 1175 goto out_free_inet; 1176 1177 sock_diag_register_inet_compat(inet_diag_rcv_msg_compat); 1178 out: 1179 return err; 1180 1181 out_free_inet: 1182 sock_diag_unregister(&inet_diag_handler); 1183 out_free_nl: 1184 kfree(inet_diag_table); 1185 goto out; 1186 } 1187 1188 static void __exit inet_diag_exit(void) 1189 { 1190 sock_diag_unregister(&inet6_diag_handler); 1191 sock_diag_unregister(&inet_diag_handler); 1192 sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat); 1193 kfree(inet_diag_table); 1194 } 1195 1196 module_init(inet_diag_init); 1197 module_exit(inet_diag_exit); 1198 MODULE_LICENSE("GPL"); 1199 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */); 1200 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */); 1201