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