1 /* 2 * inet_diag.c Module for monitoring INET transport protocols sockets. 3 * 4 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> 5 * 6 * This program is free software; you can redistribute it and/or 7 * modify it under the terms of the GNU General Public License 8 * as published by the Free Software Foundation; either version 9 * 2 of the License, or (at your option) any later version. 10 */ 11 12 #include <linux/kernel.h> 13 #include <linux/module.h> 14 #include <linux/types.h> 15 #include <linux/fcntl.h> 16 #include <linux/random.h> 17 #include <linux/slab.h> 18 #include <linux/cache.h> 19 #include <linux/init.h> 20 #include <linux/time.h> 21 22 #include <net/icmp.h> 23 #include <net/tcp.h> 24 #include <net/ipv6.h> 25 #include <net/inet_common.h> 26 #include <net/inet_connection_sock.h> 27 #include <net/inet_hashtables.h> 28 #include <net/inet_timewait_sock.h> 29 #include <net/inet6_hashtables.h> 30 #include <net/netlink.h> 31 32 #include <linux/inet.h> 33 #include <linux/stddef.h> 34 35 #include <linux/inet_diag.h> 36 #include <linux/sock_diag.h> 37 38 static const struct inet_diag_handler **inet_diag_table; 39 40 struct inet_diag_entry { 41 __be32 *saddr; 42 __be32 *daddr; 43 u16 sport; 44 u16 dport; 45 u16 family; 46 u16 userlocks; 47 #if IS_ENABLED(CONFIG_IPV6) 48 struct in6_addr saddr_storage; /* for IPv4-mapped-IPv6 addresses */ 49 struct in6_addr daddr_storage; /* for IPv4-mapped-IPv6 addresses */ 50 #endif 51 }; 52 53 static DEFINE_MUTEX(inet_diag_table_mutex); 54 55 static const struct inet_diag_handler *inet_diag_lock_handler(int proto) 56 { 57 if (!inet_diag_table[proto]) 58 request_module("net-pf-%d-proto-%d-type-%d-%d", PF_NETLINK, 59 NETLINK_SOCK_DIAG, AF_INET, proto); 60 61 mutex_lock(&inet_diag_table_mutex); 62 if (!inet_diag_table[proto]) 63 return ERR_PTR(-ENOENT); 64 65 return inet_diag_table[proto]; 66 } 67 68 static inline void inet_diag_unlock_handler( 69 const struct inet_diag_handler *handler) 70 { 71 mutex_unlock(&inet_diag_table_mutex); 72 } 73 74 int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk, 75 struct sk_buff *skb, struct inet_diag_req_v2 *req, 76 struct user_namespace *user_ns, 77 u32 portid, u32 seq, u16 nlmsg_flags, 78 const struct nlmsghdr *unlh) 79 { 80 const struct inet_sock *inet = inet_sk(sk); 81 struct inet_diag_msg *r; 82 struct nlmsghdr *nlh; 83 struct nlattr *attr; 84 void *info = NULL; 85 const struct inet_diag_handler *handler; 86 int ext = req->idiag_ext; 87 88 handler = inet_diag_table[req->sdiag_protocol]; 89 BUG_ON(handler == NULL); 90 91 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r), 92 nlmsg_flags); 93 if (!nlh) 94 return -EMSGSIZE; 95 96 r = nlmsg_data(nlh); 97 BUG_ON(sk->sk_state == TCP_TIME_WAIT); 98 99 r->idiag_family = sk->sk_family; 100 r->idiag_state = sk->sk_state; 101 r->idiag_timer = 0; 102 r->idiag_retrans = 0; 103 104 r->id.idiag_if = sk->sk_bound_dev_if; 105 sock_diag_save_cookie(sk, r->id.idiag_cookie); 106 107 r->id.idiag_sport = inet->inet_sport; 108 r->id.idiag_dport = inet->inet_dport; 109 r->id.idiag_src[0] = inet->inet_rcv_saddr; 110 r->id.idiag_dst[0] = inet->inet_daddr; 111 112 /* IPv6 dual-stack sockets use inet->tos for IPv4 connections, 113 * hence this needs to be included regardless of socket family. 114 */ 115 if (ext & (1 << (INET_DIAG_TOS - 1))) 116 if (nla_put_u8(skb, INET_DIAG_TOS, inet->tos) < 0) 117 goto errout; 118 119 #if IS_ENABLED(CONFIG_IPV6) 120 if (r->idiag_family == AF_INET6) { 121 const struct ipv6_pinfo *np = inet6_sk(sk); 122 123 *(struct in6_addr *)r->id.idiag_src = np->rcv_saddr; 124 *(struct in6_addr *)r->id.idiag_dst = np->daddr; 125 126 if (ext & (1 << (INET_DIAG_TCLASS - 1))) 127 if (nla_put_u8(skb, INET_DIAG_TCLASS, np->tclass) < 0) 128 goto errout; 129 } 130 #endif 131 132 r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk)); 133 r->idiag_inode = sock_i_ino(sk); 134 135 if (ext & (1 << (INET_DIAG_MEMINFO - 1))) { 136 struct inet_diag_meminfo minfo = { 137 .idiag_rmem = sk_rmem_alloc_get(sk), 138 .idiag_wmem = sk->sk_wmem_queued, 139 .idiag_fmem = sk->sk_forward_alloc, 140 .idiag_tmem = sk_wmem_alloc_get(sk), 141 }; 142 143 if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0) 144 goto errout; 145 } 146 147 if (ext & (1 << (INET_DIAG_SKMEMINFO - 1))) 148 if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO)) 149 goto errout; 150 151 if (icsk == NULL) { 152 handler->idiag_get_info(sk, r, NULL); 153 goto out; 154 } 155 156 #define EXPIRES_IN_MS(tmo) DIV_ROUND_UP((tmo - jiffies) * 1000, HZ) 157 158 if (icsk->icsk_pending == ICSK_TIME_RETRANS) { 159 r->idiag_timer = 1; 160 r->idiag_retrans = icsk->icsk_retransmits; 161 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout); 162 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { 163 r->idiag_timer = 4; 164 r->idiag_retrans = icsk->icsk_probes_out; 165 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout); 166 } else if (timer_pending(&sk->sk_timer)) { 167 r->idiag_timer = 2; 168 r->idiag_retrans = icsk->icsk_probes_out; 169 r->idiag_expires = EXPIRES_IN_MS(sk->sk_timer.expires); 170 } else { 171 r->idiag_timer = 0; 172 r->idiag_expires = 0; 173 } 174 #undef EXPIRES_IN_MS 175 176 if (ext & (1 << (INET_DIAG_INFO - 1))) { 177 attr = nla_reserve(skb, INET_DIAG_INFO, 178 sizeof(struct tcp_info)); 179 if (!attr) 180 goto errout; 181 182 info = nla_data(attr); 183 } 184 185 if ((ext & (1 << (INET_DIAG_CONG - 1))) && icsk->icsk_ca_ops) 186 if (nla_put_string(skb, INET_DIAG_CONG, 187 icsk->icsk_ca_ops->name) < 0) 188 goto errout; 189 190 handler->idiag_get_info(sk, r, info); 191 192 if (sk->sk_state < TCP_TIME_WAIT && 193 icsk->icsk_ca_ops && icsk->icsk_ca_ops->get_info) 194 icsk->icsk_ca_ops->get_info(sk, ext, skb); 195 196 out: 197 return nlmsg_end(skb, nlh); 198 199 errout: 200 nlmsg_cancel(skb, nlh); 201 return -EMSGSIZE; 202 } 203 EXPORT_SYMBOL_GPL(inet_sk_diag_fill); 204 205 static int inet_csk_diag_fill(struct sock *sk, 206 struct sk_buff *skb, struct inet_diag_req_v2 *req, 207 struct user_namespace *user_ns, 208 u32 portid, u32 seq, u16 nlmsg_flags, 209 const struct nlmsghdr *unlh) 210 { 211 return inet_sk_diag_fill(sk, inet_csk(sk), 212 skb, req, user_ns, portid, seq, nlmsg_flags, unlh); 213 } 214 215 static int inet_twsk_diag_fill(struct inet_timewait_sock *tw, 216 struct sk_buff *skb, struct inet_diag_req_v2 *req, 217 u32 portid, u32 seq, u16 nlmsg_flags, 218 const struct nlmsghdr *unlh) 219 { 220 long tmo; 221 struct inet_diag_msg *r; 222 struct nlmsghdr *nlh; 223 224 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r), 225 nlmsg_flags); 226 if (!nlh) 227 return -EMSGSIZE; 228 229 r = nlmsg_data(nlh); 230 BUG_ON(tw->tw_state != TCP_TIME_WAIT); 231 232 tmo = tw->tw_ttd - jiffies; 233 if (tmo < 0) 234 tmo = 0; 235 236 r->idiag_family = tw->tw_family; 237 r->idiag_retrans = 0; 238 r->id.idiag_if = tw->tw_bound_dev_if; 239 sock_diag_save_cookie(tw, r->id.idiag_cookie); 240 r->id.idiag_sport = tw->tw_sport; 241 r->id.idiag_dport = tw->tw_dport; 242 r->id.idiag_src[0] = tw->tw_rcv_saddr; 243 r->id.idiag_dst[0] = tw->tw_daddr; 244 r->idiag_state = tw->tw_substate; 245 r->idiag_timer = 3; 246 r->idiag_expires = DIV_ROUND_UP(tmo * 1000, HZ); 247 r->idiag_rqueue = 0; 248 r->idiag_wqueue = 0; 249 r->idiag_uid = 0; 250 r->idiag_inode = 0; 251 #if IS_ENABLED(CONFIG_IPV6) 252 if (tw->tw_family == AF_INET6) { 253 const struct inet6_timewait_sock *tw6 = 254 inet6_twsk((struct sock *)tw); 255 256 *(struct in6_addr *)r->id.idiag_src = tw6->tw_v6_rcv_saddr; 257 *(struct in6_addr *)r->id.idiag_dst = tw6->tw_v6_daddr; 258 } 259 #endif 260 261 return nlmsg_end(skb, nlh); 262 } 263 264 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb, 265 struct inet_diag_req_v2 *r, 266 struct user_namespace *user_ns, 267 u32 portid, u32 seq, u16 nlmsg_flags, 268 const struct nlmsghdr *unlh) 269 { 270 if (sk->sk_state == TCP_TIME_WAIT) 271 return inet_twsk_diag_fill((struct inet_timewait_sock *)sk, 272 skb, r, portid, seq, nlmsg_flags, 273 unlh); 274 return inet_csk_diag_fill(sk, skb, r, user_ns, portid, seq, nlmsg_flags, unlh); 275 } 276 277 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *in_skb, 278 const struct nlmsghdr *nlh, struct inet_diag_req_v2 *req) 279 { 280 int err; 281 struct sock *sk; 282 struct sk_buff *rep; 283 struct net *net = sock_net(in_skb->sk); 284 285 err = -EINVAL; 286 if (req->sdiag_family == AF_INET) { 287 sk = inet_lookup(net, hashinfo, req->id.idiag_dst[0], 288 req->id.idiag_dport, req->id.idiag_src[0], 289 req->id.idiag_sport, req->id.idiag_if); 290 } 291 #if IS_ENABLED(CONFIG_IPV6) 292 else if (req->sdiag_family == AF_INET6) { 293 sk = inet6_lookup(net, hashinfo, 294 (struct in6_addr *)req->id.idiag_dst, 295 req->id.idiag_dport, 296 (struct in6_addr *)req->id.idiag_src, 297 req->id.idiag_sport, 298 req->id.idiag_if); 299 } 300 #endif 301 else { 302 goto out_nosk; 303 } 304 305 err = -ENOENT; 306 if (sk == NULL) 307 goto out_nosk; 308 309 err = sock_diag_check_cookie(sk, req->id.idiag_cookie); 310 if (err) 311 goto out; 312 313 rep = nlmsg_new(sizeof(struct inet_diag_msg) + 314 sizeof(struct inet_diag_meminfo) + 315 sizeof(struct tcp_info) + 64, GFP_KERNEL); 316 if (!rep) { 317 err = -ENOMEM; 318 goto out; 319 } 320 321 err = sk_diag_fill(sk, rep, req, 322 sk_user_ns(NETLINK_CB(in_skb).ssk), 323 NETLINK_CB(in_skb).portid, 324 nlh->nlmsg_seq, 0, nlh); 325 if (err < 0) { 326 WARN_ON(err == -EMSGSIZE); 327 nlmsg_free(rep); 328 goto out; 329 } 330 err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid, 331 MSG_DONTWAIT); 332 if (err > 0) 333 err = 0; 334 335 out: 336 if (sk) { 337 if (sk->sk_state == TCP_TIME_WAIT) 338 inet_twsk_put((struct inet_timewait_sock *)sk); 339 else 340 sock_put(sk); 341 } 342 out_nosk: 343 return err; 344 } 345 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk); 346 347 static int inet_diag_get_exact(struct sk_buff *in_skb, 348 const struct nlmsghdr *nlh, 349 struct inet_diag_req_v2 *req) 350 { 351 const struct inet_diag_handler *handler; 352 int err; 353 354 handler = inet_diag_lock_handler(req->sdiag_protocol); 355 if (IS_ERR(handler)) 356 err = PTR_ERR(handler); 357 else 358 err = handler->dump_one(in_skb, nlh, req); 359 inet_diag_unlock_handler(handler); 360 361 return err; 362 } 363 364 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits) 365 { 366 int words = bits >> 5; 367 368 bits &= 0x1f; 369 370 if (words) { 371 if (memcmp(a1, a2, words << 2)) 372 return 0; 373 } 374 if (bits) { 375 __be32 w1, w2; 376 __be32 mask; 377 378 w1 = a1[words]; 379 w2 = a2[words]; 380 381 mask = htonl((0xffffffff) << (32 - bits)); 382 383 if ((w1 ^ w2) & mask) 384 return 0; 385 } 386 387 return 1; 388 } 389 390 391 static int inet_diag_bc_run(const struct nlattr *_bc, 392 const struct inet_diag_entry *entry) 393 { 394 const void *bc = nla_data(_bc); 395 int len = nla_len(_bc); 396 397 while (len > 0) { 398 int yes = 1; 399 const struct inet_diag_bc_op *op = bc; 400 401 switch (op->code) { 402 case INET_DIAG_BC_NOP: 403 break; 404 case INET_DIAG_BC_JMP: 405 yes = 0; 406 break; 407 case INET_DIAG_BC_S_GE: 408 yes = entry->sport >= op[1].no; 409 break; 410 case INET_DIAG_BC_S_LE: 411 yes = entry->sport <= op[1].no; 412 break; 413 case INET_DIAG_BC_D_GE: 414 yes = entry->dport >= op[1].no; 415 break; 416 case INET_DIAG_BC_D_LE: 417 yes = entry->dport <= op[1].no; 418 break; 419 case INET_DIAG_BC_AUTO: 420 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK); 421 break; 422 case INET_DIAG_BC_S_COND: 423 case INET_DIAG_BC_D_COND: { 424 struct inet_diag_hostcond *cond; 425 __be32 *addr; 426 427 cond = (struct inet_diag_hostcond *)(op + 1); 428 if (cond->port != -1 && 429 cond->port != (op->code == INET_DIAG_BC_S_COND ? 430 entry->sport : entry->dport)) { 431 yes = 0; 432 break; 433 } 434 435 if (cond->prefix_len == 0) 436 break; 437 438 if (op->code == INET_DIAG_BC_S_COND) 439 addr = entry->saddr; 440 else 441 addr = entry->daddr; 442 443 if (bitstring_match(addr, cond->addr, 444 cond->prefix_len)) 445 break; 446 if (entry->family == AF_INET6 && 447 cond->family == AF_INET) { 448 if (addr[0] == 0 && addr[1] == 0 && 449 addr[2] == htonl(0xffff) && 450 bitstring_match(addr + 3, cond->addr, 451 cond->prefix_len)) 452 break; 453 } 454 yes = 0; 455 break; 456 } 457 } 458 459 if (yes) { 460 len -= op->yes; 461 bc += op->yes; 462 } else { 463 len -= op->no; 464 bc += op->no; 465 } 466 } 467 return len == 0; 468 } 469 470 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk) 471 { 472 struct inet_diag_entry entry; 473 struct inet_sock *inet = inet_sk(sk); 474 475 if (bc == NULL) 476 return 1; 477 478 entry.family = sk->sk_family; 479 #if IS_ENABLED(CONFIG_IPV6) 480 if (entry.family == AF_INET6) { 481 struct ipv6_pinfo *np = inet6_sk(sk); 482 483 entry.saddr = np->rcv_saddr.s6_addr32; 484 entry.daddr = np->daddr.s6_addr32; 485 } else 486 #endif 487 { 488 entry.saddr = &inet->inet_rcv_saddr; 489 entry.daddr = &inet->inet_daddr; 490 } 491 entry.sport = inet->inet_num; 492 entry.dport = ntohs(inet->inet_dport); 493 entry.userlocks = sk->sk_userlocks; 494 495 return inet_diag_bc_run(bc, &entry); 496 } 497 EXPORT_SYMBOL_GPL(inet_diag_bc_sk); 498 499 static int valid_cc(const void *bc, int len, int cc) 500 { 501 while (len >= 0) { 502 const struct inet_diag_bc_op *op = bc; 503 504 if (cc > len) 505 return 0; 506 if (cc == len) 507 return 1; 508 if (op->yes < 4 || op->yes & 3) 509 return 0; 510 len -= op->yes; 511 bc += op->yes; 512 } 513 return 0; 514 } 515 516 static int inet_diag_bc_audit(const void *bytecode, int bytecode_len) 517 { 518 const void *bc = bytecode; 519 int len = bytecode_len; 520 521 while (len > 0) { 522 const struct inet_diag_bc_op *op = bc; 523 524 //printk("BC: %d %d %d {%d} / %d\n", op->code, op->yes, op->no, op[1].no, len); 525 switch (op->code) { 526 case INET_DIAG_BC_AUTO: 527 case INET_DIAG_BC_S_COND: 528 case INET_DIAG_BC_D_COND: 529 case INET_DIAG_BC_S_GE: 530 case INET_DIAG_BC_S_LE: 531 case INET_DIAG_BC_D_GE: 532 case INET_DIAG_BC_D_LE: 533 case INET_DIAG_BC_JMP: 534 if (op->no < 4 || op->no > len + 4 || op->no & 3) 535 return -EINVAL; 536 if (op->no < len && 537 !valid_cc(bytecode, bytecode_len, len - op->no)) 538 return -EINVAL; 539 break; 540 case INET_DIAG_BC_NOP: 541 break; 542 default: 543 return -EINVAL; 544 } 545 if (op->yes < 4 || op->yes > len + 4 || op->yes & 3) 546 return -EINVAL; 547 bc += op->yes; 548 len -= op->yes; 549 } 550 return len == 0 ? 0 : -EINVAL; 551 } 552 553 static int inet_csk_diag_dump(struct sock *sk, 554 struct sk_buff *skb, 555 struct netlink_callback *cb, 556 struct inet_diag_req_v2 *r, 557 const struct nlattr *bc) 558 { 559 if (!inet_diag_bc_sk(bc, sk)) 560 return 0; 561 562 return inet_csk_diag_fill(sk, skb, r, 563 sk_user_ns(NETLINK_CB(cb->skb).ssk), 564 NETLINK_CB(cb->skb).portid, 565 cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh); 566 } 567 568 static int inet_twsk_diag_dump(struct inet_timewait_sock *tw, 569 struct sk_buff *skb, 570 struct netlink_callback *cb, 571 struct inet_diag_req_v2 *r, 572 const struct nlattr *bc) 573 { 574 if (bc != NULL) { 575 struct inet_diag_entry entry; 576 577 entry.family = tw->tw_family; 578 #if IS_ENABLED(CONFIG_IPV6) 579 if (tw->tw_family == AF_INET6) { 580 struct inet6_timewait_sock *tw6 = 581 inet6_twsk((struct sock *)tw); 582 entry.saddr = tw6->tw_v6_rcv_saddr.s6_addr32; 583 entry.daddr = tw6->tw_v6_daddr.s6_addr32; 584 } else 585 #endif 586 { 587 entry.saddr = &tw->tw_rcv_saddr; 588 entry.daddr = &tw->tw_daddr; 589 } 590 entry.sport = tw->tw_num; 591 entry.dport = ntohs(tw->tw_dport); 592 entry.userlocks = 0; 593 594 if (!inet_diag_bc_run(bc, &entry)) 595 return 0; 596 } 597 598 return inet_twsk_diag_fill(tw, skb, r, 599 NETLINK_CB(cb->skb).portid, 600 cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh); 601 } 602 603 /* Get the IPv4, IPv6, or IPv4-mapped-IPv6 local and remote addresses 604 * from a request_sock. For IPv4-mapped-IPv6 we must map IPv4 to IPv6. 605 */ 606 static inline void inet_diag_req_addrs(const struct sock *sk, 607 const struct request_sock *req, 608 struct inet_diag_entry *entry) 609 { 610 struct inet_request_sock *ireq = inet_rsk(req); 611 612 #if IS_ENABLED(CONFIG_IPV6) 613 if (sk->sk_family == AF_INET6) { 614 if (req->rsk_ops->family == AF_INET6) { 615 entry->saddr = inet6_rsk(req)->loc_addr.s6_addr32; 616 entry->daddr = inet6_rsk(req)->rmt_addr.s6_addr32; 617 } else if (req->rsk_ops->family == AF_INET) { 618 ipv6_addr_set_v4mapped(ireq->loc_addr, 619 &entry->saddr_storage); 620 ipv6_addr_set_v4mapped(ireq->rmt_addr, 621 &entry->daddr_storage); 622 entry->saddr = entry->saddr_storage.s6_addr32; 623 entry->daddr = entry->daddr_storage.s6_addr32; 624 } 625 } else 626 #endif 627 { 628 entry->saddr = &ireq->loc_addr; 629 entry->daddr = &ireq->rmt_addr; 630 } 631 } 632 633 static int inet_diag_fill_req(struct sk_buff *skb, struct sock *sk, 634 struct request_sock *req, 635 struct user_namespace *user_ns, 636 u32 portid, u32 seq, 637 const struct nlmsghdr *unlh) 638 { 639 const struct inet_request_sock *ireq = inet_rsk(req); 640 struct inet_sock *inet = inet_sk(sk); 641 struct inet_diag_msg *r; 642 struct nlmsghdr *nlh; 643 long tmo; 644 645 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r), 646 NLM_F_MULTI); 647 if (!nlh) 648 return -EMSGSIZE; 649 650 r = nlmsg_data(nlh); 651 r->idiag_family = sk->sk_family; 652 r->idiag_state = TCP_SYN_RECV; 653 r->idiag_timer = 1; 654 r->idiag_retrans = req->retrans; 655 656 r->id.idiag_if = sk->sk_bound_dev_if; 657 sock_diag_save_cookie(req, r->id.idiag_cookie); 658 659 tmo = req->expires - jiffies; 660 if (tmo < 0) 661 tmo = 0; 662 663 r->id.idiag_sport = inet->inet_sport; 664 r->id.idiag_dport = ireq->rmt_port; 665 r->id.idiag_src[0] = ireq->loc_addr; 666 r->id.idiag_dst[0] = ireq->rmt_addr; 667 r->idiag_expires = jiffies_to_msecs(tmo); 668 r->idiag_rqueue = 0; 669 r->idiag_wqueue = 0; 670 r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk)); 671 r->idiag_inode = 0; 672 #if IS_ENABLED(CONFIG_IPV6) 673 if (r->idiag_family == AF_INET6) { 674 struct inet_diag_entry entry; 675 inet_diag_req_addrs(sk, req, &entry); 676 memcpy(r->id.idiag_src, entry.saddr, sizeof(struct in6_addr)); 677 memcpy(r->id.idiag_dst, entry.daddr, sizeof(struct in6_addr)); 678 } 679 #endif 680 681 return nlmsg_end(skb, nlh); 682 } 683 684 static int inet_diag_dump_reqs(struct sk_buff *skb, struct sock *sk, 685 struct netlink_callback *cb, 686 struct inet_diag_req_v2 *r, 687 const struct nlattr *bc) 688 { 689 struct inet_diag_entry entry; 690 struct inet_connection_sock *icsk = inet_csk(sk); 691 struct listen_sock *lopt; 692 struct inet_sock *inet = inet_sk(sk); 693 int j, s_j; 694 int reqnum, s_reqnum; 695 int err = 0; 696 697 s_j = cb->args[3]; 698 s_reqnum = cb->args[4]; 699 700 if (s_j > 0) 701 s_j--; 702 703 entry.family = sk->sk_family; 704 705 read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock); 706 707 lopt = icsk->icsk_accept_queue.listen_opt; 708 if (!lopt || !lopt->qlen) 709 goto out; 710 711 if (bc != NULL) { 712 entry.sport = inet->inet_num; 713 entry.userlocks = sk->sk_userlocks; 714 } 715 716 for (j = s_j; j < lopt->nr_table_entries; j++) { 717 struct request_sock *req, *head = lopt->syn_table[j]; 718 719 reqnum = 0; 720 for (req = head; req; reqnum++, req = req->dl_next) { 721 struct inet_request_sock *ireq = inet_rsk(req); 722 723 if (reqnum < s_reqnum) 724 continue; 725 if (r->id.idiag_dport != ireq->rmt_port && 726 r->id.idiag_dport) 727 continue; 728 729 if (bc) { 730 inet_diag_req_addrs(sk, req, &entry); 731 entry.dport = ntohs(ireq->rmt_port); 732 733 if (!inet_diag_bc_run(bc, &entry)) 734 continue; 735 } 736 737 err = inet_diag_fill_req(skb, sk, req, 738 sk_user_ns(NETLINK_CB(cb->skb).ssk), 739 NETLINK_CB(cb->skb).portid, 740 cb->nlh->nlmsg_seq, cb->nlh); 741 if (err < 0) { 742 cb->args[3] = j + 1; 743 cb->args[4] = reqnum; 744 goto out; 745 } 746 } 747 748 s_reqnum = 0; 749 } 750 751 out: 752 read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock); 753 754 return err; 755 } 756 757 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb, 758 struct netlink_callback *cb, struct inet_diag_req_v2 *r, struct nlattr *bc) 759 { 760 int i, num; 761 int s_i, s_num; 762 struct net *net = sock_net(skb->sk); 763 764 s_i = cb->args[1]; 765 s_num = num = cb->args[2]; 766 767 if (cb->args[0] == 0) { 768 if (!(r->idiag_states & (TCPF_LISTEN | TCPF_SYN_RECV))) 769 goto skip_listen_ht; 770 771 for (i = s_i; i < INET_LHTABLE_SIZE; i++) { 772 struct sock *sk; 773 struct hlist_nulls_node *node; 774 struct inet_listen_hashbucket *ilb; 775 776 num = 0; 777 ilb = &hashinfo->listening_hash[i]; 778 spin_lock_bh(&ilb->lock); 779 sk_nulls_for_each(sk, node, &ilb->head) { 780 struct inet_sock *inet = inet_sk(sk); 781 782 if (!net_eq(sock_net(sk), net)) 783 continue; 784 785 if (num < s_num) { 786 num++; 787 continue; 788 } 789 790 if (r->sdiag_family != AF_UNSPEC && 791 sk->sk_family != r->sdiag_family) 792 goto next_listen; 793 794 if (r->id.idiag_sport != inet->inet_sport && 795 r->id.idiag_sport) 796 goto next_listen; 797 798 if (!(r->idiag_states & TCPF_LISTEN) || 799 r->id.idiag_dport || 800 cb->args[3] > 0) 801 goto syn_recv; 802 803 if (inet_csk_diag_dump(sk, skb, cb, r, bc) < 0) { 804 spin_unlock_bh(&ilb->lock); 805 goto done; 806 } 807 808 syn_recv: 809 if (!(r->idiag_states & TCPF_SYN_RECV)) 810 goto next_listen; 811 812 if (inet_diag_dump_reqs(skb, sk, cb, r, bc) < 0) { 813 spin_unlock_bh(&ilb->lock); 814 goto done; 815 } 816 817 next_listen: 818 cb->args[3] = 0; 819 cb->args[4] = 0; 820 ++num; 821 } 822 spin_unlock_bh(&ilb->lock); 823 824 s_num = 0; 825 cb->args[3] = 0; 826 cb->args[4] = 0; 827 } 828 skip_listen_ht: 829 cb->args[0] = 1; 830 s_i = num = s_num = 0; 831 } 832 833 if (!(r->idiag_states & ~(TCPF_LISTEN | TCPF_SYN_RECV))) 834 goto out; 835 836 for (i = s_i; i <= hashinfo->ehash_mask; i++) { 837 struct inet_ehash_bucket *head = &hashinfo->ehash[i]; 838 spinlock_t *lock = inet_ehash_lockp(hashinfo, i); 839 struct sock *sk; 840 struct hlist_nulls_node *node; 841 842 num = 0; 843 844 if (hlist_nulls_empty(&head->chain) && 845 hlist_nulls_empty(&head->twchain)) 846 continue; 847 848 if (i > s_i) 849 s_num = 0; 850 851 spin_lock_bh(lock); 852 sk_nulls_for_each(sk, node, &head->chain) { 853 struct inet_sock *inet = inet_sk(sk); 854 855 if (!net_eq(sock_net(sk), net)) 856 continue; 857 if (num < s_num) 858 goto next_normal; 859 if (!(r->idiag_states & (1 << sk->sk_state))) 860 goto next_normal; 861 if (r->sdiag_family != AF_UNSPEC && 862 sk->sk_family != r->sdiag_family) 863 goto next_normal; 864 if (r->id.idiag_sport != inet->inet_sport && 865 r->id.idiag_sport) 866 goto next_normal; 867 if (r->id.idiag_dport != inet->inet_dport && 868 r->id.idiag_dport) 869 goto next_normal; 870 if (inet_csk_diag_dump(sk, skb, cb, r, bc) < 0) { 871 spin_unlock_bh(lock); 872 goto done; 873 } 874 next_normal: 875 ++num; 876 } 877 878 if (r->idiag_states & TCPF_TIME_WAIT) { 879 struct inet_timewait_sock *tw; 880 881 inet_twsk_for_each(tw, node, 882 &head->twchain) { 883 if (!net_eq(twsk_net(tw), net)) 884 continue; 885 886 if (num < s_num) 887 goto next_dying; 888 if (r->sdiag_family != AF_UNSPEC && 889 tw->tw_family != r->sdiag_family) 890 goto next_dying; 891 if (r->id.idiag_sport != tw->tw_sport && 892 r->id.idiag_sport) 893 goto next_dying; 894 if (r->id.idiag_dport != tw->tw_dport && 895 r->id.idiag_dport) 896 goto next_dying; 897 if (inet_twsk_diag_dump(tw, skb, cb, r, bc) < 0) { 898 spin_unlock_bh(lock); 899 goto done; 900 } 901 next_dying: 902 ++num; 903 } 904 } 905 spin_unlock_bh(lock); 906 } 907 908 done: 909 cb->args[1] = i; 910 cb->args[2] = num; 911 out: 912 ; 913 } 914 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk); 915 916 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb, 917 struct inet_diag_req_v2 *r, struct nlattr *bc) 918 { 919 const struct inet_diag_handler *handler; 920 int err = 0; 921 922 handler = inet_diag_lock_handler(r->sdiag_protocol); 923 if (!IS_ERR(handler)) 924 handler->dump(skb, cb, r, bc); 925 else 926 err = PTR_ERR(handler); 927 inet_diag_unlock_handler(handler); 928 929 return err ? : skb->len; 930 } 931 932 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb) 933 { 934 struct nlattr *bc = NULL; 935 int hdrlen = sizeof(struct inet_diag_req_v2); 936 937 if (nlmsg_attrlen(cb->nlh, hdrlen)) 938 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE); 939 940 return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh), bc); 941 } 942 943 static inline int inet_diag_type2proto(int type) 944 { 945 switch (type) { 946 case TCPDIAG_GETSOCK: 947 return IPPROTO_TCP; 948 case DCCPDIAG_GETSOCK: 949 return IPPROTO_DCCP; 950 default: 951 return 0; 952 } 953 } 954 955 static int inet_diag_dump_compat(struct sk_buff *skb, struct netlink_callback *cb) 956 { 957 struct inet_diag_req *rc = nlmsg_data(cb->nlh); 958 struct inet_diag_req_v2 req; 959 struct nlattr *bc = NULL; 960 int hdrlen = sizeof(struct inet_diag_req); 961 962 req.sdiag_family = AF_UNSPEC; /* compatibility */ 963 req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type); 964 req.idiag_ext = rc->idiag_ext; 965 req.idiag_states = rc->idiag_states; 966 req.id = rc->id; 967 968 if (nlmsg_attrlen(cb->nlh, hdrlen)) 969 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE); 970 971 return __inet_diag_dump(skb, cb, &req, bc); 972 } 973 974 static int inet_diag_get_exact_compat(struct sk_buff *in_skb, 975 const struct nlmsghdr *nlh) 976 { 977 struct inet_diag_req *rc = nlmsg_data(nlh); 978 struct inet_diag_req_v2 req; 979 980 req.sdiag_family = rc->idiag_family; 981 req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type); 982 req.idiag_ext = rc->idiag_ext; 983 req.idiag_states = rc->idiag_states; 984 req.id = rc->id; 985 986 return inet_diag_get_exact(in_skb, nlh, &req); 987 } 988 989 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh) 990 { 991 int hdrlen = sizeof(struct inet_diag_req); 992 struct net *net = sock_net(skb->sk); 993 994 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX || 995 nlmsg_len(nlh) < hdrlen) 996 return -EINVAL; 997 998 if (nlh->nlmsg_flags & NLM_F_DUMP) { 999 if (nlmsg_attrlen(nlh, hdrlen)) { 1000 struct nlattr *attr; 1001 1002 attr = nlmsg_find_attr(nlh, hdrlen, 1003 INET_DIAG_REQ_BYTECODE); 1004 if (attr == NULL || 1005 nla_len(attr) < sizeof(struct inet_diag_bc_op) || 1006 inet_diag_bc_audit(nla_data(attr), nla_len(attr))) 1007 return -EINVAL; 1008 } 1009 { 1010 struct netlink_dump_control c = { 1011 .dump = inet_diag_dump_compat, 1012 }; 1013 return netlink_dump_start(net->diag_nlsk, skb, nlh, &c); 1014 } 1015 } 1016 1017 return inet_diag_get_exact_compat(skb, nlh); 1018 } 1019 1020 static int inet_diag_handler_dump(struct sk_buff *skb, struct nlmsghdr *h) 1021 { 1022 int hdrlen = sizeof(struct inet_diag_req_v2); 1023 struct net *net = sock_net(skb->sk); 1024 1025 if (nlmsg_len(h) < hdrlen) 1026 return -EINVAL; 1027 1028 if (h->nlmsg_flags & NLM_F_DUMP) { 1029 if (nlmsg_attrlen(h, hdrlen)) { 1030 struct nlattr *attr; 1031 attr = nlmsg_find_attr(h, hdrlen, 1032 INET_DIAG_REQ_BYTECODE); 1033 if (attr == NULL || 1034 nla_len(attr) < sizeof(struct inet_diag_bc_op) || 1035 inet_diag_bc_audit(nla_data(attr), nla_len(attr))) 1036 return -EINVAL; 1037 } 1038 { 1039 struct netlink_dump_control c = { 1040 .dump = inet_diag_dump, 1041 }; 1042 return netlink_dump_start(net->diag_nlsk, skb, h, &c); 1043 } 1044 } 1045 1046 return inet_diag_get_exact(skb, h, nlmsg_data(h)); 1047 } 1048 1049 static const struct sock_diag_handler inet_diag_handler = { 1050 .family = AF_INET, 1051 .dump = inet_diag_handler_dump, 1052 }; 1053 1054 static const struct sock_diag_handler inet6_diag_handler = { 1055 .family = AF_INET6, 1056 .dump = inet_diag_handler_dump, 1057 }; 1058 1059 int inet_diag_register(const struct inet_diag_handler *h) 1060 { 1061 const __u16 type = h->idiag_type; 1062 int err = -EINVAL; 1063 1064 if (type >= IPPROTO_MAX) 1065 goto out; 1066 1067 mutex_lock(&inet_diag_table_mutex); 1068 err = -EEXIST; 1069 if (inet_diag_table[type] == NULL) { 1070 inet_diag_table[type] = h; 1071 err = 0; 1072 } 1073 mutex_unlock(&inet_diag_table_mutex); 1074 out: 1075 return err; 1076 } 1077 EXPORT_SYMBOL_GPL(inet_diag_register); 1078 1079 void inet_diag_unregister(const struct inet_diag_handler *h) 1080 { 1081 const __u16 type = h->idiag_type; 1082 1083 if (type >= IPPROTO_MAX) 1084 return; 1085 1086 mutex_lock(&inet_diag_table_mutex); 1087 inet_diag_table[type] = NULL; 1088 mutex_unlock(&inet_diag_table_mutex); 1089 } 1090 EXPORT_SYMBOL_GPL(inet_diag_unregister); 1091 1092 static int __init inet_diag_init(void) 1093 { 1094 const int inet_diag_table_size = (IPPROTO_MAX * 1095 sizeof(struct inet_diag_handler *)); 1096 int err = -ENOMEM; 1097 1098 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL); 1099 if (!inet_diag_table) 1100 goto out; 1101 1102 err = sock_diag_register(&inet_diag_handler); 1103 if (err) 1104 goto out_free_nl; 1105 1106 err = sock_diag_register(&inet6_diag_handler); 1107 if (err) 1108 goto out_free_inet; 1109 1110 sock_diag_register_inet_compat(inet_diag_rcv_msg_compat); 1111 out: 1112 return err; 1113 1114 out_free_inet: 1115 sock_diag_unregister(&inet_diag_handler); 1116 out_free_nl: 1117 kfree(inet_diag_table); 1118 goto out; 1119 } 1120 1121 static void __exit inet_diag_exit(void) 1122 { 1123 sock_diag_unregister(&inet6_diag_handler); 1124 sock_diag_unregister(&inet_diag_handler); 1125 sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat); 1126 kfree(inet_diag_table); 1127 } 1128 1129 module_init(inet_diag_init); 1130 module_exit(inet_diag_exit); 1131 MODULE_LICENSE("GPL"); 1132 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */); 1133 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */); 1134