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 u32 ifindex; 48 u32 mark; 49 }; 50 51 static DEFINE_MUTEX(inet_diag_table_mutex); 52 53 static const struct inet_diag_handler *inet_diag_lock_handler(int proto) 54 { 55 if (!inet_diag_table[proto]) 56 sock_load_diag_module(AF_INET, proto); 57 58 mutex_lock(&inet_diag_table_mutex); 59 if (!inet_diag_table[proto]) 60 return ERR_PTR(-ENOENT); 61 62 return inet_diag_table[proto]; 63 } 64 65 static void inet_diag_unlock_handler(const struct inet_diag_handler *handler) 66 { 67 mutex_unlock(&inet_diag_table_mutex); 68 } 69 70 void inet_diag_msg_common_fill(struct inet_diag_msg *r, struct sock *sk) 71 { 72 r->idiag_family = sk->sk_family; 73 74 r->id.idiag_sport = htons(sk->sk_num); 75 r->id.idiag_dport = sk->sk_dport; 76 r->id.idiag_if = sk->sk_bound_dev_if; 77 sock_diag_save_cookie(sk, r->id.idiag_cookie); 78 79 #if IS_ENABLED(CONFIG_IPV6) 80 if (sk->sk_family == AF_INET6) { 81 *(struct in6_addr *)r->id.idiag_src = sk->sk_v6_rcv_saddr; 82 *(struct in6_addr *)r->id.idiag_dst = sk->sk_v6_daddr; 83 } else 84 #endif 85 { 86 memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src)); 87 memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst)); 88 89 r->id.idiag_src[0] = sk->sk_rcv_saddr; 90 r->id.idiag_dst[0] = sk->sk_daddr; 91 } 92 } 93 EXPORT_SYMBOL_GPL(inet_diag_msg_common_fill); 94 95 static size_t inet_sk_attr_size(struct sock *sk, 96 const struct inet_diag_req_v2 *req, 97 bool net_admin) 98 { 99 const struct inet_diag_handler *handler; 100 size_t aux = 0; 101 102 handler = inet_diag_table[req->sdiag_protocol]; 103 if (handler && handler->idiag_get_aux_size) 104 aux = handler->idiag_get_aux_size(sk, net_admin); 105 106 return nla_total_size(sizeof(struct tcp_info)) 107 + nla_total_size(1) /* INET_DIAG_SHUTDOWN */ 108 + nla_total_size(1) /* INET_DIAG_TOS */ 109 + nla_total_size(1) /* INET_DIAG_TCLASS */ 110 + nla_total_size(4) /* INET_DIAG_MARK */ 111 + nla_total_size(sizeof(struct inet_diag_meminfo)) 112 + nla_total_size(sizeof(struct inet_diag_msg)) 113 + nla_total_size(SK_MEMINFO_VARS * sizeof(u32)) 114 + nla_total_size(TCP_CA_NAME_MAX) 115 + nla_total_size(sizeof(struct tcpvegas_info)) 116 + aux 117 + 64; 118 } 119 120 int inet_diag_msg_attrs_fill(struct sock *sk, struct sk_buff *skb, 121 struct inet_diag_msg *r, int ext, 122 struct user_namespace *user_ns, 123 bool net_admin) 124 { 125 const struct inet_sock *inet = inet_sk(sk); 126 127 if (nla_put_u8(skb, INET_DIAG_SHUTDOWN, sk->sk_shutdown)) 128 goto errout; 129 130 /* IPv6 dual-stack sockets use inet->tos for IPv4 connections, 131 * hence this needs to be included regardless of socket family. 132 */ 133 if (ext & (1 << (INET_DIAG_TOS - 1))) 134 if (nla_put_u8(skb, INET_DIAG_TOS, inet->tos) < 0) 135 goto errout; 136 137 #if IS_ENABLED(CONFIG_IPV6) 138 if (r->idiag_family == AF_INET6) { 139 if (ext & (1 << (INET_DIAG_TCLASS - 1))) 140 if (nla_put_u8(skb, INET_DIAG_TCLASS, 141 inet6_sk(sk)->tclass) < 0) 142 goto errout; 143 144 if (((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) && 145 nla_put_u8(skb, INET_DIAG_SKV6ONLY, ipv6_only_sock(sk))) 146 goto errout; 147 } 148 #endif 149 150 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK, sk->sk_mark)) 151 goto errout; 152 153 r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk)); 154 r->idiag_inode = sock_i_ino(sk); 155 156 return 0; 157 errout: 158 return 1; 159 } 160 EXPORT_SYMBOL_GPL(inet_diag_msg_attrs_fill); 161 162 int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk, 163 struct sk_buff *skb, const struct inet_diag_req_v2 *req, 164 struct user_namespace *user_ns, 165 u32 portid, u32 seq, u16 nlmsg_flags, 166 const struct nlmsghdr *unlh, 167 bool net_admin) 168 { 169 const struct tcp_congestion_ops *ca_ops; 170 const struct inet_diag_handler *handler; 171 int ext = req->idiag_ext; 172 struct inet_diag_msg *r; 173 struct nlmsghdr *nlh; 174 struct nlattr *attr; 175 void *info = NULL; 176 177 handler = inet_diag_table[req->sdiag_protocol]; 178 BUG_ON(!handler); 179 180 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r), 181 nlmsg_flags); 182 if (!nlh) 183 return -EMSGSIZE; 184 185 r = nlmsg_data(nlh); 186 BUG_ON(!sk_fullsock(sk)); 187 188 inet_diag_msg_common_fill(r, sk); 189 r->idiag_state = sk->sk_state; 190 r->idiag_timer = 0; 191 r->idiag_retrans = 0; 192 193 if (inet_diag_msg_attrs_fill(sk, skb, r, ext, user_ns, net_admin)) 194 goto errout; 195 196 if (ext & (1 << (INET_DIAG_MEMINFO - 1))) { 197 struct inet_diag_meminfo minfo = { 198 .idiag_rmem = sk_rmem_alloc_get(sk), 199 .idiag_wmem = sk->sk_wmem_queued, 200 .idiag_fmem = sk->sk_forward_alloc, 201 .idiag_tmem = sk_wmem_alloc_get(sk), 202 }; 203 204 if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0) 205 goto errout; 206 } 207 208 if (ext & (1 << (INET_DIAG_SKMEMINFO - 1))) 209 if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO)) 210 goto errout; 211 212 /* 213 * RAW sockets might have user-defined protocols assigned, 214 * so report the one supplied on socket creation. 215 */ 216 if (sk->sk_type == SOCK_RAW) { 217 if (nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol)) 218 goto errout; 219 } 220 221 if (!icsk) { 222 handler->idiag_get_info(sk, r, NULL); 223 goto out; 224 } 225 226 if (icsk->icsk_pending == ICSK_TIME_RETRANS || 227 icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT || 228 icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) { 229 r->idiag_timer = 1; 230 r->idiag_retrans = icsk->icsk_retransmits; 231 r->idiag_expires = 232 jiffies_to_msecs(icsk->icsk_timeout - jiffies); 233 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { 234 r->idiag_timer = 4; 235 r->idiag_retrans = icsk->icsk_probes_out; 236 r->idiag_expires = 237 jiffies_to_msecs(icsk->icsk_timeout - jiffies); 238 } else if (timer_pending(&sk->sk_timer)) { 239 r->idiag_timer = 2; 240 r->idiag_retrans = icsk->icsk_probes_out; 241 r->idiag_expires = 242 jiffies_to_msecs(sk->sk_timer.expires - jiffies); 243 } else { 244 r->idiag_timer = 0; 245 r->idiag_expires = 0; 246 } 247 248 if ((ext & (1 << (INET_DIAG_INFO - 1))) && handler->idiag_info_size) { 249 attr = nla_reserve_64bit(skb, INET_DIAG_INFO, 250 handler->idiag_info_size, 251 INET_DIAG_PAD); 252 if (!attr) 253 goto errout; 254 255 info = nla_data(attr); 256 } 257 258 if (ext & (1 << (INET_DIAG_CONG - 1))) { 259 int err = 0; 260 261 rcu_read_lock(); 262 ca_ops = READ_ONCE(icsk->icsk_ca_ops); 263 if (ca_ops) 264 err = nla_put_string(skb, INET_DIAG_CONG, ca_ops->name); 265 rcu_read_unlock(); 266 if (err < 0) 267 goto errout; 268 } 269 270 handler->idiag_get_info(sk, r, info); 271 272 if (ext & (1 << (INET_DIAG_INFO - 1)) && handler->idiag_get_aux) 273 if (handler->idiag_get_aux(sk, net_admin, skb) < 0) 274 goto errout; 275 276 if (sk->sk_state < TCP_TIME_WAIT) { 277 union tcp_cc_info info; 278 size_t sz = 0; 279 int attr; 280 281 rcu_read_lock(); 282 ca_ops = READ_ONCE(icsk->icsk_ca_ops); 283 if (ca_ops && ca_ops->get_info) 284 sz = ca_ops->get_info(sk, ext, &attr, &info); 285 rcu_read_unlock(); 286 if (sz && nla_put(skb, attr, sz, &info) < 0) 287 goto errout; 288 } 289 290 if (ext & (1 << (INET_DIAG_CLASS_ID - 1))) { 291 u32 classid = 0; 292 293 #ifdef CONFIG_SOCK_CGROUP_DATA 294 classid = sock_cgroup_classid(&sk->sk_cgrp_data); 295 #endif 296 297 if (nla_put_u32(skb, INET_DIAG_CLASS_ID, classid)) 298 goto errout; 299 } 300 301 out: 302 nlmsg_end(skb, nlh); 303 return 0; 304 305 errout: 306 nlmsg_cancel(skb, nlh); 307 return -EMSGSIZE; 308 } 309 EXPORT_SYMBOL_GPL(inet_sk_diag_fill); 310 311 static int inet_csk_diag_fill(struct sock *sk, 312 struct sk_buff *skb, 313 const struct inet_diag_req_v2 *req, 314 struct user_namespace *user_ns, 315 u32 portid, u32 seq, u16 nlmsg_flags, 316 const struct nlmsghdr *unlh, 317 bool net_admin) 318 { 319 return inet_sk_diag_fill(sk, inet_csk(sk), skb, req, user_ns, 320 portid, seq, nlmsg_flags, unlh, net_admin); 321 } 322 323 static int inet_twsk_diag_fill(struct sock *sk, 324 struct sk_buff *skb, 325 u32 portid, u32 seq, u16 nlmsg_flags, 326 const struct nlmsghdr *unlh) 327 { 328 struct inet_timewait_sock *tw = inet_twsk(sk); 329 struct inet_diag_msg *r; 330 struct nlmsghdr *nlh; 331 long tmo; 332 333 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r), 334 nlmsg_flags); 335 if (!nlh) 336 return -EMSGSIZE; 337 338 r = nlmsg_data(nlh); 339 BUG_ON(tw->tw_state != TCP_TIME_WAIT); 340 341 tmo = tw->tw_timer.expires - jiffies; 342 if (tmo < 0) 343 tmo = 0; 344 345 inet_diag_msg_common_fill(r, sk); 346 r->idiag_retrans = 0; 347 348 r->idiag_state = tw->tw_substate; 349 r->idiag_timer = 3; 350 r->idiag_expires = jiffies_to_msecs(tmo); 351 r->idiag_rqueue = 0; 352 r->idiag_wqueue = 0; 353 r->idiag_uid = 0; 354 r->idiag_inode = 0; 355 356 nlmsg_end(skb, nlh); 357 return 0; 358 } 359 360 static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb, 361 u32 portid, u32 seq, u16 nlmsg_flags, 362 const struct nlmsghdr *unlh, bool net_admin) 363 { 364 struct request_sock *reqsk = inet_reqsk(sk); 365 struct inet_diag_msg *r; 366 struct nlmsghdr *nlh; 367 long tmo; 368 369 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r), 370 nlmsg_flags); 371 if (!nlh) 372 return -EMSGSIZE; 373 374 r = nlmsg_data(nlh); 375 inet_diag_msg_common_fill(r, sk); 376 r->idiag_state = TCP_SYN_RECV; 377 r->idiag_timer = 1; 378 r->idiag_retrans = reqsk->num_retrans; 379 380 BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) != 381 offsetof(struct sock, sk_cookie)); 382 383 tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies; 384 r->idiag_expires = (tmo >= 0) ? jiffies_to_msecs(tmo) : 0; 385 r->idiag_rqueue = 0; 386 r->idiag_wqueue = 0; 387 r->idiag_uid = 0; 388 r->idiag_inode = 0; 389 390 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK, 391 inet_rsk(reqsk)->ir_mark)) 392 return -EMSGSIZE; 393 394 nlmsg_end(skb, nlh); 395 return 0; 396 } 397 398 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb, 399 const struct inet_diag_req_v2 *r, 400 struct user_namespace *user_ns, 401 u32 portid, u32 seq, u16 nlmsg_flags, 402 const struct nlmsghdr *unlh, bool net_admin) 403 { 404 if (sk->sk_state == TCP_TIME_WAIT) 405 return inet_twsk_diag_fill(sk, skb, portid, seq, 406 nlmsg_flags, unlh); 407 408 if (sk->sk_state == TCP_NEW_SYN_RECV) 409 return inet_req_diag_fill(sk, skb, portid, seq, 410 nlmsg_flags, unlh, net_admin); 411 412 return inet_csk_diag_fill(sk, skb, r, user_ns, portid, seq, 413 nlmsg_flags, unlh, net_admin); 414 } 415 416 struct sock *inet_diag_find_one_icsk(struct net *net, 417 struct inet_hashinfo *hashinfo, 418 const struct inet_diag_req_v2 *req) 419 { 420 struct sock *sk; 421 422 rcu_read_lock(); 423 if (req->sdiag_family == AF_INET) 424 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0], 425 req->id.idiag_dport, req->id.idiag_src[0], 426 req->id.idiag_sport, req->id.idiag_if); 427 #if IS_ENABLED(CONFIG_IPV6) 428 else if (req->sdiag_family == AF_INET6) { 429 if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) && 430 ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src)) 431 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3], 432 req->id.idiag_dport, req->id.idiag_src[3], 433 req->id.idiag_sport, req->id.idiag_if); 434 else 435 sk = inet6_lookup(net, hashinfo, NULL, 0, 436 (struct in6_addr *)req->id.idiag_dst, 437 req->id.idiag_dport, 438 (struct in6_addr *)req->id.idiag_src, 439 req->id.idiag_sport, 440 req->id.idiag_if); 441 } 442 #endif 443 else { 444 rcu_read_unlock(); 445 return ERR_PTR(-EINVAL); 446 } 447 rcu_read_unlock(); 448 if (!sk) 449 return ERR_PTR(-ENOENT); 450 451 if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) { 452 sock_gen_put(sk); 453 return ERR_PTR(-ENOENT); 454 } 455 456 return sk; 457 } 458 EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk); 459 460 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo, 461 struct sk_buff *in_skb, 462 const struct nlmsghdr *nlh, 463 const struct inet_diag_req_v2 *req) 464 { 465 bool net_admin = netlink_net_capable(in_skb, CAP_NET_ADMIN); 466 struct net *net = sock_net(in_skb->sk); 467 struct sk_buff *rep; 468 struct sock *sk; 469 int err; 470 471 sk = inet_diag_find_one_icsk(net, hashinfo, req); 472 if (IS_ERR(sk)) 473 return PTR_ERR(sk); 474 475 rep = nlmsg_new(inet_sk_attr_size(sk, req, net_admin), GFP_KERNEL); 476 if (!rep) { 477 err = -ENOMEM; 478 goto out; 479 } 480 481 err = sk_diag_fill(sk, rep, req, 482 sk_user_ns(NETLINK_CB(in_skb).sk), 483 NETLINK_CB(in_skb).portid, 484 nlh->nlmsg_seq, 0, nlh, net_admin); 485 if (err < 0) { 486 WARN_ON(err == -EMSGSIZE); 487 nlmsg_free(rep); 488 goto out; 489 } 490 err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid, 491 MSG_DONTWAIT); 492 if (err > 0) 493 err = 0; 494 495 out: 496 if (sk) 497 sock_gen_put(sk); 498 499 return err; 500 } 501 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk); 502 503 static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb, 504 const struct nlmsghdr *nlh, 505 const struct inet_diag_req_v2 *req) 506 { 507 const struct inet_diag_handler *handler; 508 int err; 509 510 handler = inet_diag_lock_handler(req->sdiag_protocol); 511 if (IS_ERR(handler)) 512 err = PTR_ERR(handler); 513 else if (cmd == SOCK_DIAG_BY_FAMILY) 514 err = handler->dump_one(in_skb, nlh, req); 515 else if (cmd == SOCK_DESTROY && handler->destroy) 516 err = handler->destroy(in_skb, req); 517 else 518 err = -EOPNOTSUPP; 519 inet_diag_unlock_handler(handler); 520 521 return err; 522 } 523 524 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits) 525 { 526 int words = bits >> 5; 527 528 bits &= 0x1f; 529 530 if (words) { 531 if (memcmp(a1, a2, words << 2)) 532 return 0; 533 } 534 if (bits) { 535 __be32 w1, w2; 536 __be32 mask; 537 538 w1 = a1[words]; 539 w2 = a2[words]; 540 541 mask = htonl((0xffffffff) << (32 - bits)); 542 543 if ((w1 ^ w2) & mask) 544 return 0; 545 } 546 547 return 1; 548 } 549 550 static int inet_diag_bc_run(const struct nlattr *_bc, 551 const struct inet_diag_entry *entry) 552 { 553 const void *bc = nla_data(_bc); 554 int len = nla_len(_bc); 555 556 while (len > 0) { 557 int yes = 1; 558 const struct inet_diag_bc_op *op = bc; 559 560 switch (op->code) { 561 case INET_DIAG_BC_NOP: 562 break; 563 case INET_DIAG_BC_JMP: 564 yes = 0; 565 break; 566 case INET_DIAG_BC_S_EQ: 567 yes = entry->sport == op[1].no; 568 break; 569 case INET_DIAG_BC_S_GE: 570 yes = entry->sport >= op[1].no; 571 break; 572 case INET_DIAG_BC_S_LE: 573 yes = entry->sport <= op[1].no; 574 break; 575 case INET_DIAG_BC_D_EQ: 576 yes = entry->dport == op[1].no; 577 break; 578 case INET_DIAG_BC_D_GE: 579 yes = entry->dport >= op[1].no; 580 break; 581 case INET_DIAG_BC_D_LE: 582 yes = entry->dport <= op[1].no; 583 break; 584 case INET_DIAG_BC_AUTO: 585 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK); 586 break; 587 case INET_DIAG_BC_S_COND: 588 case INET_DIAG_BC_D_COND: { 589 const struct inet_diag_hostcond *cond; 590 const __be32 *addr; 591 592 cond = (const struct inet_diag_hostcond *)(op + 1); 593 if (cond->port != -1 && 594 cond->port != (op->code == INET_DIAG_BC_S_COND ? 595 entry->sport : entry->dport)) { 596 yes = 0; 597 break; 598 } 599 600 if (op->code == INET_DIAG_BC_S_COND) 601 addr = entry->saddr; 602 else 603 addr = entry->daddr; 604 605 if (cond->family != AF_UNSPEC && 606 cond->family != entry->family) { 607 if (entry->family == AF_INET6 && 608 cond->family == AF_INET) { 609 if (addr[0] == 0 && addr[1] == 0 && 610 addr[2] == htonl(0xffff) && 611 bitstring_match(addr + 3, 612 cond->addr, 613 cond->prefix_len)) 614 break; 615 } 616 yes = 0; 617 break; 618 } 619 620 if (cond->prefix_len == 0) 621 break; 622 if (bitstring_match(addr, cond->addr, 623 cond->prefix_len)) 624 break; 625 yes = 0; 626 break; 627 } 628 case INET_DIAG_BC_DEV_COND: { 629 u32 ifindex; 630 631 ifindex = *((const u32 *)(op + 1)); 632 if (ifindex != entry->ifindex) 633 yes = 0; 634 break; 635 } 636 case INET_DIAG_BC_MARK_COND: { 637 struct inet_diag_markcond *cond; 638 639 cond = (struct inet_diag_markcond *)(op + 1); 640 if ((entry->mark & cond->mask) != cond->mark) 641 yes = 0; 642 break; 643 } 644 } 645 646 if (yes) { 647 len -= op->yes; 648 bc += op->yes; 649 } else { 650 len -= op->no; 651 bc += op->no; 652 } 653 } 654 return len == 0; 655 } 656 657 /* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV) 658 */ 659 static void entry_fill_addrs(struct inet_diag_entry *entry, 660 const struct sock *sk) 661 { 662 #if IS_ENABLED(CONFIG_IPV6) 663 if (sk->sk_family == AF_INET6) { 664 entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32; 665 entry->daddr = sk->sk_v6_daddr.s6_addr32; 666 } else 667 #endif 668 { 669 entry->saddr = &sk->sk_rcv_saddr; 670 entry->daddr = &sk->sk_daddr; 671 } 672 } 673 674 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk) 675 { 676 struct inet_sock *inet = inet_sk(sk); 677 struct inet_diag_entry entry; 678 679 if (!bc) 680 return 1; 681 682 entry.family = sk->sk_family; 683 entry_fill_addrs(&entry, sk); 684 entry.sport = inet->inet_num; 685 entry.dport = ntohs(inet->inet_dport); 686 entry.ifindex = sk->sk_bound_dev_if; 687 entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0; 688 if (sk_fullsock(sk)) 689 entry.mark = sk->sk_mark; 690 else if (sk->sk_state == TCP_NEW_SYN_RECV) 691 entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark; 692 else 693 entry.mark = 0; 694 695 return inet_diag_bc_run(bc, &entry); 696 } 697 EXPORT_SYMBOL_GPL(inet_diag_bc_sk); 698 699 static int valid_cc(const void *bc, int len, int cc) 700 { 701 while (len >= 0) { 702 const struct inet_diag_bc_op *op = bc; 703 704 if (cc > len) 705 return 0; 706 if (cc == len) 707 return 1; 708 if (op->yes < 4 || op->yes & 3) 709 return 0; 710 len -= op->yes; 711 bc += op->yes; 712 } 713 return 0; 714 } 715 716 /* data is u32 ifindex */ 717 static bool valid_devcond(const struct inet_diag_bc_op *op, int len, 718 int *min_len) 719 { 720 /* Check ifindex space. */ 721 *min_len += sizeof(u32); 722 if (len < *min_len) 723 return false; 724 725 return true; 726 } 727 /* Validate an inet_diag_hostcond. */ 728 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len, 729 int *min_len) 730 { 731 struct inet_diag_hostcond *cond; 732 int addr_len; 733 734 /* Check hostcond space. */ 735 *min_len += sizeof(struct inet_diag_hostcond); 736 if (len < *min_len) 737 return false; 738 cond = (struct inet_diag_hostcond *)(op + 1); 739 740 /* Check address family and address length. */ 741 switch (cond->family) { 742 case AF_UNSPEC: 743 addr_len = 0; 744 break; 745 case AF_INET: 746 addr_len = sizeof(struct in_addr); 747 break; 748 case AF_INET6: 749 addr_len = sizeof(struct in6_addr); 750 break; 751 default: 752 return false; 753 } 754 *min_len += addr_len; 755 if (len < *min_len) 756 return false; 757 758 /* Check prefix length (in bits) vs address length (in bytes). */ 759 if (cond->prefix_len > 8 * addr_len) 760 return false; 761 762 return true; 763 } 764 765 /* Validate a port comparison operator. */ 766 static bool valid_port_comparison(const struct inet_diag_bc_op *op, 767 int len, int *min_len) 768 { 769 /* Port comparisons put the port in a follow-on inet_diag_bc_op. */ 770 *min_len += sizeof(struct inet_diag_bc_op); 771 if (len < *min_len) 772 return false; 773 return true; 774 } 775 776 static bool valid_markcond(const struct inet_diag_bc_op *op, int len, 777 int *min_len) 778 { 779 *min_len += sizeof(struct inet_diag_markcond); 780 return len >= *min_len; 781 } 782 783 static int inet_diag_bc_audit(const struct nlattr *attr, 784 const struct sk_buff *skb) 785 { 786 bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN); 787 const void *bytecode, *bc; 788 int bytecode_len, len; 789 790 if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op)) 791 return -EINVAL; 792 793 bytecode = bc = nla_data(attr); 794 len = bytecode_len = nla_len(attr); 795 796 while (len > 0) { 797 int min_len = sizeof(struct inet_diag_bc_op); 798 const struct inet_diag_bc_op *op = bc; 799 800 switch (op->code) { 801 case INET_DIAG_BC_S_COND: 802 case INET_DIAG_BC_D_COND: 803 if (!valid_hostcond(bc, len, &min_len)) 804 return -EINVAL; 805 break; 806 case INET_DIAG_BC_DEV_COND: 807 if (!valid_devcond(bc, len, &min_len)) 808 return -EINVAL; 809 break; 810 case INET_DIAG_BC_S_EQ: 811 case INET_DIAG_BC_S_GE: 812 case INET_DIAG_BC_S_LE: 813 case INET_DIAG_BC_D_EQ: 814 case INET_DIAG_BC_D_GE: 815 case INET_DIAG_BC_D_LE: 816 if (!valid_port_comparison(bc, len, &min_len)) 817 return -EINVAL; 818 break; 819 case INET_DIAG_BC_MARK_COND: 820 if (!net_admin) 821 return -EPERM; 822 if (!valid_markcond(bc, len, &min_len)) 823 return -EINVAL; 824 break; 825 case INET_DIAG_BC_AUTO: 826 case INET_DIAG_BC_JMP: 827 case INET_DIAG_BC_NOP: 828 break; 829 default: 830 return -EINVAL; 831 } 832 833 if (op->code != INET_DIAG_BC_NOP) { 834 if (op->no < min_len || op->no > len + 4 || op->no & 3) 835 return -EINVAL; 836 if (op->no < len && 837 !valid_cc(bytecode, bytecode_len, len - op->no)) 838 return -EINVAL; 839 } 840 841 if (op->yes < min_len || op->yes > len + 4 || op->yes & 3) 842 return -EINVAL; 843 bc += op->yes; 844 len -= op->yes; 845 } 846 return len == 0 ? 0 : -EINVAL; 847 } 848 849 static int inet_csk_diag_dump(struct sock *sk, 850 struct sk_buff *skb, 851 struct netlink_callback *cb, 852 const struct inet_diag_req_v2 *r, 853 const struct nlattr *bc, 854 bool net_admin) 855 { 856 if (!inet_diag_bc_sk(bc, sk)) 857 return 0; 858 859 return inet_csk_diag_fill(sk, skb, r, 860 sk_user_ns(NETLINK_CB(cb->skb).sk), 861 NETLINK_CB(cb->skb).portid, 862 cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh, 863 net_admin); 864 } 865 866 static void twsk_build_assert(void) 867 { 868 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) != 869 offsetof(struct sock, sk_family)); 870 871 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) != 872 offsetof(struct inet_sock, inet_num)); 873 874 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) != 875 offsetof(struct inet_sock, inet_dport)); 876 877 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) != 878 offsetof(struct inet_sock, inet_rcv_saddr)); 879 880 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) != 881 offsetof(struct inet_sock, inet_daddr)); 882 883 #if IS_ENABLED(CONFIG_IPV6) 884 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) != 885 offsetof(struct sock, sk_v6_rcv_saddr)); 886 887 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) != 888 offsetof(struct sock, sk_v6_daddr)); 889 #endif 890 } 891 892 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb, 893 struct netlink_callback *cb, 894 const struct inet_diag_req_v2 *r, struct nlattr *bc) 895 { 896 bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN); 897 struct net *net = sock_net(skb->sk); 898 u32 idiag_states = r->idiag_states; 899 int i, num, s_i, s_num; 900 struct sock *sk; 901 902 if (idiag_states & TCPF_SYN_RECV) 903 idiag_states |= TCPF_NEW_SYN_RECV; 904 s_i = cb->args[1]; 905 s_num = num = cb->args[2]; 906 907 if (cb->args[0] == 0) { 908 if (!(idiag_states & TCPF_LISTEN) || r->id.idiag_dport) 909 goto skip_listen_ht; 910 911 for (i = s_i; i < INET_LHTABLE_SIZE; i++) { 912 struct inet_listen_hashbucket *ilb; 913 914 num = 0; 915 ilb = &hashinfo->listening_hash[i]; 916 spin_lock(&ilb->lock); 917 sk_for_each(sk, &ilb->head) { 918 struct inet_sock *inet = inet_sk(sk); 919 920 if (!net_eq(sock_net(sk), net)) 921 continue; 922 923 if (num < s_num) { 924 num++; 925 continue; 926 } 927 928 if (r->sdiag_family != AF_UNSPEC && 929 sk->sk_family != r->sdiag_family) 930 goto next_listen; 931 932 if (r->id.idiag_sport != inet->inet_sport && 933 r->id.idiag_sport) 934 goto next_listen; 935 936 if (inet_csk_diag_dump(sk, skb, cb, r, 937 bc, net_admin) < 0) { 938 spin_unlock(&ilb->lock); 939 goto done; 940 } 941 942 next_listen: 943 ++num; 944 } 945 spin_unlock(&ilb->lock); 946 947 s_num = 0; 948 } 949 skip_listen_ht: 950 cb->args[0] = 1; 951 s_i = num = s_num = 0; 952 } 953 954 if (!(idiag_states & ~TCPF_LISTEN)) 955 goto out; 956 957 #define SKARR_SZ 16 958 for (i = s_i; i <= hashinfo->ehash_mask; i++) { 959 struct inet_ehash_bucket *head = &hashinfo->ehash[i]; 960 spinlock_t *lock = inet_ehash_lockp(hashinfo, i); 961 struct hlist_nulls_node *node; 962 struct sock *sk_arr[SKARR_SZ]; 963 int num_arr[SKARR_SZ]; 964 int idx, accum, res; 965 966 if (hlist_nulls_empty(&head->chain)) 967 continue; 968 969 if (i > s_i) 970 s_num = 0; 971 972 next_chunk: 973 num = 0; 974 accum = 0; 975 spin_lock_bh(lock); 976 sk_nulls_for_each(sk, node, &head->chain) { 977 int state; 978 979 if (!net_eq(sock_net(sk), net)) 980 continue; 981 if (num < s_num) 982 goto next_normal; 983 state = (sk->sk_state == TCP_TIME_WAIT) ? 984 inet_twsk(sk)->tw_substate : sk->sk_state; 985 if (!(idiag_states & (1 << state))) 986 goto next_normal; 987 if (r->sdiag_family != AF_UNSPEC && 988 sk->sk_family != r->sdiag_family) 989 goto next_normal; 990 if (r->id.idiag_sport != htons(sk->sk_num) && 991 r->id.idiag_sport) 992 goto next_normal; 993 if (r->id.idiag_dport != sk->sk_dport && 994 r->id.idiag_dport) 995 goto next_normal; 996 twsk_build_assert(); 997 998 if (!inet_diag_bc_sk(bc, sk)) 999 goto next_normal; 1000 1001 if (!refcount_inc_not_zero(&sk->sk_refcnt)) 1002 goto next_normal; 1003 1004 num_arr[accum] = num; 1005 sk_arr[accum] = sk; 1006 if (++accum == SKARR_SZ) 1007 break; 1008 next_normal: 1009 ++num; 1010 } 1011 spin_unlock_bh(lock); 1012 res = 0; 1013 for (idx = 0; idx < accum; idx++) { 1014 if (res >= 0) { 1015 res = sk_diag_fill(sk_arr[idx], skb, r, 1016 sk_user_ns(NETLINK_CB(cb->skb).sk), 1017 NETLINK_CB(cb->skb).portid, 1018 cb->nlh->nlmsg_seq, NLM_F_MULTI, 1019 cb->nlh, net_admin); 1020 if (res < 0) 1021 num = num_arr[idx]; 1022 } 1023 sock_gen_put(sk_arr[idx]); 1024 } 1025 if (res < 0) 1026 break; 1027 cond_resched(); 1028 if (accum == SKARR_SZ) { 1029 s_num = num + 1; 1030 goto next_chunk; 1031 } 1032 } 1033 1034 done: 1035 cb->args[1] = i; 1036 cb->args[2] = num; 1037 out: 1038 ; 1039 } 1040 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk); 1041 1042 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb, 1043 const struct inet_diag_req_v2 *r, 1044 struct nlattr *bc) 1045 { 1046 const struct inet_diag_handler *handler; 1047 int err = 0; 1048 1049 handler = inet_diag_lock_handler(r->sdiag_protocol); 1050 if (!IS_ERR(handler)) 1051 handler->dump(skb, cb, r, bc); 1052 else 1053 err = PTR_ERR(handler); 1054 inet_diag_unlock_handler(handler); 1055 1056 return err ? : skb->len; 1057 } 1058 1059 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb) 1060 { 1061 int hdrlen = sizeof(struct inet_diag_req_v2); 1062 struct nlattr *bc = NULL; 1063 1064 if (nlmsg_attrlen(cb->nlh, hdrlen)) 1065 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE); 1066 1067 return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh), bc); 1068 } 1069 1070 static int inet_diag_type2proto(int type) 1071 { 1072 switch (type) { 1073 case TCPDIAG_GETSOCK: 1074 return IPPROTO_TCP; 1075 case DCCPDIAG_GETSOCK: 1076 return IPPROTO_DCCP; 1077 default: 1078 return 0; 1079 } 1080 } 1081 1082 static int inet_diag_dump_compat(struct sk_buff *skb, 1083 struct netlink_callback *cb) 1084 { 1085 struct inet_diag_req *rc = nlmsg_data(cb->nlh); 1086 int hdrlen = sizeof(struct inet_diag_req); 1087 struct inet_diag_req_v2 req; 1088 struct nlattr *bc = NULL; 1089 1090 req.sdiag_family = AF_UNSPEC; /* compatibility */ 1091 req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type); 1092 req.idiag_ext = rc->idiag_ext; 1093 req.idiag_states = rc->idiag_states; 1094 req.id = rc->id; 1095 1096 if (nlmsg_attrlen(cb->nlh, hdrlen)) 1097 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE); 1098 1099 return __inet_diag_dump(skb, cb, &req, bc); 1100 } 1101 1102 static int inet_diag_get_exact_compat(struct sk_buff *in_skb, 1103 const struct nlmsghdr *nlh) 1104 { 1105 struct inet_diag_req *rc = nlmsg_data(nlh); 1106 struct inet_diag_req_v2 req; 1107 1108 req.sdiag_family = rc->idiag_family; 1109 req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type); 1110 req.idiag_ext = rc->idiag_ext; 1111 req.idiag_states = rc->idiag_states; 1112 req.id = rc->id; 1113 1114 return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh, &req); 1115 } 1116 1117 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh) 1118 { 1119 int hdrlen = sizeof(struct inet_diag_req); 1120 struct net *net = sock_net(skb->sk); 1121 1122 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX || 1123 nlmsg_len(nlh) < hdrlen) 1124 return -EINVAL; 1125 1126 if (nlh->nlmsg_flags & NLM_F_DUMP) { 1127 if (nlmsg_attrlen(nlh, hdrlen)) { 1128 struct nlattr *attr; 1129 int err; 1130 1131 attr = nlmsg_find_attr(nlh, hdrlen, 1132 INET_DIAG_REQ_BYTECODE); 1133 err = inet_diag_bc_audit(attr, skb); 1134 if (err) 1135 return err; 1136 } 1137 { 1138 struct netlink_dump_control c = { 1139 .dump = inet_diag_dump_compat, 1140 }; 1141 return netlink_dump_start(net->diag_nlsk, skb, nlh, &c); 1142 } 1143 } 1144 1145 return inet_diag_get_exact_compat(skb, nlh); 1146 } 1147 1148 static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h) 1149 { 1150 int hdrlen = sizeof(struct inet_diag_req_v2); 1151 struct net *net = sock_net(skb->sk); 1152 1153 if (nlmsg_len(h) < hdrlen) 1154 return -EINVAL; 1155 1156 if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY && 1157 h->nlmsg_flags & NLM_F_DUMP) { 1158 if (nlmsg_attrlen(h, hdrlen)) { 1159 struct nlattr *attr; 1160 int err; 1161 1162 attr = nlmsg_find_attr(h, hdrlen, 1163 INET_DIAG_REQ_BYTECODE); 1164 err = inet_diag_bc_audit(attr, skb); 1165 if (err) 1166 return err; 1167 } 1168 { 1169 struct netlink_dump_control c = { 1170 .dump = inet_diag_dump, 1171 }; 1172 return netlink_dump_start(net->diag_nlsk, skb, h, &c); 1173 } 1174 } 1175 1176 return inet_diag_cmd_exact(h->nlmsg_type, skb, h, nlmsg_data(h)); 1177 } 1178 1179 static 1180 int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk) 1181 { 1182 const struct inet_diag_handler *handler; 1183 struct nlmsghdr *nlh; 1184 struct nlattr *attr; 1185 struct inet_diag_msg *r; 1186 void *info = NULL; 1187 int err = 0; 1188 1189 nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0); 1190 if (!nlh) 1191 return -ENOMEM; 1192 1193 r = nlmsg_data(nlh); 1194 memset(r, 0, sizeof(*r)); 1195 inet_diag_msg_common_fill(r, sk); 1196 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM) 1197 r->id.idiag_sport = inet_sk(sk)->inet_sport; 1198 r->idiag_state = sk->sk_state; 1199 1200 if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) { 1201 nlmsg_cancel(skb, nlh); 1202 return err; 1203 } 1204 1205 handler = inet_diag_lock_handler(sk->sk_protocol); 1206 if (IS_ERR(handler)) { 1207 inet_diag_unlock_handler(handler); 1208 nlmsg_cancel(skb, nlh); 1209 return PTR_ERR(handler); 1210 } 1211 1212 attr = handler->idiag_info_size 1213 ? nla_reserve_64bit(skb, INET_DIAG_INFO, 1214 handler->idiag_info_size, 1215 INET_DIAG_PAD) 1216 : NULL; 1217 if (attr) 1218 info = nla_data(attr); 1219 1220 handler->idiag_get_info(sk, r, info); 1221 inet_diag_unlock_handler(handler); 1222 1223 nlmsg_end(skb, nlh); 1224 return 0; 1225 } 1226 1227 static const struct sock_diag_handler inet_diag_handler = { 1228 .family = AF_INET, 1229 .dump = inet_diag_handler_cmd, 1230 .get_info = inet_diag_handler_get_info, 1231 .destroy = inet_diag_handler_cmd, 1232 }; 1233 1234 static const struct sock_diag_handler inet6_diag_handler = { 1235 .family = AF_INET6, 1236 .dump = inet_diag_handler_cmd, 1237 .get_info = inet_diag_handler_get_info, 1238 .destroy = inet_diag_handler_cmd, 1239 }; 1240 1241 int inet_diag_register(const struct inet_diag_handler *h) 1242 { 1243 const __u16 type = h->idiag_type; 1244 int err = -EINVAL; 1245 1246 if (type >= IPPROTO_MAX) 1247 goto out; 1248 1249 mutex_lock(&inet_diag_table_mutex); 1250 err = -EEXIST; 1251 if (!inet_diag_table[type]) { 1252 inet_diag_table[type] = h; 1253 err = 0; 1254 } 1255 mutex_unlock(&inet_diag_table_mutex); 1256 out: 1257 return err; 1258 } 1259 EXPORT_SYMBOL_GPL(inet_diag_register); 1260 1261 void inet_diag_unregister(const struct inet_diag_handler *h) 1262 { 1263 const __u16 type = h->idiag_type; 1264 1265 if (type >= IPPROTO_MAX) 1266 return; 1267 1268 mutex_lock(&inet_diag_table_mutex); 1269 inet_diag_table[type] = NULL; 1270 mutex_unlock(&inet_diag_table_mutex); 1271 } 1272 EXPORT_SYMBOL_GPL(inet_diag_unregister); 1273 1274 static int __init inet_diag_init(void) 1275 { 1276 const int inet_diag_table_size = (IPPROTO_MAX * 1277 sizeof(struct inet_diag_handler *)); 1278 int err = -ENOMEM; 1279 1280 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL); 1281 if (!inet_diag_table) 1282 goto out; 1283 1284 err = sock_diag_register(&inet_diag_handler); 1285 if (err) 1286 goto out_free_nl; 1287 1288 err = sock_diag_register(&inet6_diag_handler); 1289 if (err) 1290 goto out_free_inet; 1291 1292 sock_diag_register_inet_compat(inet_diag_rcv_msg_compat); 1293 out: 1294 return err; 1295 1296 out_free_inet: 1297 sock_diag_unregister(&inet_diag_handler); 1298 out_free_nl: 1299 kfree(inet_diag_table); 1300 goto out; 1301 } 1302 1303 static void __exit inet_diag_exit(void) 1304 { 1305 sock_diag_unregister(&inet6_diag_handler); 1306 sock_diag_unregister(&inet_diag_handler); 1307 sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat); 1308 kfree(inet_diag_table); 1309 } 1310 1311 module_init(inet_diag_init); 1312 module_exit(inet_diag_exit); 1313 MODULE_LICENSE("GPL"); 1314 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */); 1315 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */); 1316