1 /* 2 * tcp_diag.c Module for monitoring TCP sockets. 3 * 4 * Version: $Id: tcp_diag.c,v 1.3 2002/02/01 22:01:04 davem Exp $ 5 * 6 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> 7 * 8 * This program is free software; you can redistribute it and/or 9 * modify it under the terms of the GNU General Public License 10 * as published by the Free Software Foundation; either version 11 * 2 of the License, or (at your option) any later version. 12 */ 13 14 #include <linux/config.h> 15 #include <linux/module.h> 16 #include <linux/types.h> 17 #include <linux/fcntl.h> 18 #include <linux/random.h> 19 #include <linux/cache.h> 20 #include <linux/init.h> 21 #include <linux/time.h> 22 23 #include <net/icmp.h> 24 #include <net/tcp.h> 25 #include <net/ipv6.h> 26 #include <net/inet_common.h> 27 28 #include <linux/inet.h> 29 #include <linux/stddef.h> 30 31 #include <linux/tcp_diag.h> 32 33 struct tcpdiag_entry 34 { 35 u32 *saddr; 36 u32 *daddr; 37 u16 sport; 38 u16 dport; 39 u16 family; 40 u16 userlocks; 41 }; 42 43 static struct sock *tcpnl; 44 45 #define TCPDIAG_PUT(skb, attrtype, attrlen) \ 46 RTA_DATA(__RTA_PUT(skb, attrtype, attrlen)) 47 48 static int tcpdiag_fill(struct sk_buff *skb, struct sock *sk, 49 int ext, u32 pid, u32 seq, u16 nlmsg_flags) 50 { 51 struct inet_sock *inet = inet_sk(sk); 52 struct tcp_sock *tp = tcp_sk(sk); 53 struct tcpdiagmsg *r; 54 struct nlmsghdr *nlh; 55 struct tcp_info *info = NULL; 56 struct tcpdiag_meminfo *minfo = NULL; 57 unsigned char *b = skb->tail; 58 59 nlh = NLMSG_PUT(skb, pid, seq, TCPDIAG_GETSOCK, sizeof(*r)); 60 nlh->nlmsg_flags = nlmsg_flags; 61 r = NLMSG_DATA(nlh); 62 if (sk->sk_state != TCP_TIME_WAIT) { 63 if (ext & (1<<(TCPDIAG_MEMINFO-1))) 64 minfo = TCPDIAG_PUT(skb, TCPDIAG_MEMINFO, sizeof(*minfo)); 65 if (ext & (1<<(TCPDIAG_INFO-1))) 66 info = TCPDIAG_PUT(skb, TCPDIAG_INFO, sizeof(*info)); 67 68 if (ext & (1<<(TCPDIAG_CONG-1))) { 69 size_t len = strlen(tp->ca_ops->name); 70 strcpy(TCPDIAG_PUT(skb, TCPDIAG_CONG, len+1), 71 tp->ca_ops->name); 72 } 73 } 74 r->tcpdiag_family = sk->sk_family; 75 r->tcpdiag_state = sk->sk_state; 76 r->tcpdiag_timer = 0; 77 r->tcpdiag_retrans = 0; 78 79 r->id.tcpdiag_if = sk->sk_bound_dev_if; 80 r->id.tcpdiag_cookie[0] = (u32)(unsigned long)sk; 81 r->id.tcpdiag_cookie[1] = (u32)(((unsigned long)sk >> 31) >> 1); 82 83 if (r->tcpdiag_state == TCP_TIME_WAIT) { 84 struct tcp_tw_bucket *tw = (struct tcp_tw_bucket*)sk; 85 long tmo = tw->tw_ttd - jiffies; 86 if (tmo < 0) 87 tmo = 0; 88 89 r->id.tcpdiag_sport = tw->tw_sport; 90 r->id.tcpdiag_dport = tw->tw_dport; 91 r->id.tcpdiag_src[0] = tw->tw_rcv_saddr; 92 r->id.tcpdiag_dst[0] = tw->tw_daddr; 93 r->tcpdiag_state = tw->tw_substate; 94 r->tcpdiag_timer = 3; 95 r->tcpdiag_expires = (tmo*1000+HZ-1)/HZ; 96 r->tcpdiag_rqueue = 0; 97 r->tcpdiag_wqueue = 0; 98 r->tcpdiag_uid = 0; 99 r->tcpdiag_inode = 0; 100 #ifdef CONFIG_IP_TCPDIAG_IPV6 101 if (r->tcpdiag_family == AF_INET6) { 102 ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_src, 103 &tw->tw_v6_rcv_saddr); 104 ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_dst, 105 &tw->tw_v6_daddr); 106 } 107 #endif 108 nlh->nlmsg_len = skb->tail - b; 109 return skb->len; 110 } 111 112 r->id.tcpdiag_sport = inet->sport; 113 r->id.tcpdiag_dport = inet->dport; 114 r->id.tcpdiag_src[0] = inet->rcv_saddr; 115 r->id.tcpdiag_dst[0] = inet->daddr; 116 117 #ifdef CONFIG_IP_TCPDIAG_IPV6 118 if (r->tcpdiag_family == AF_INET6) { 119 struct ipv6_pinfo *np = inet6_sk(sk); 120 121 ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_src, 122 &np->rcv_saddr); 123 ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_dst, 124 &np->daddr); 125 } 126 #endif 127 128 #define EXPIRES_IN_MS(tmo) ((tmo-jiffies)*1000+HZ-1)/HZ 129 130 if (tp->pending == TCP_TIME_RETRANS) { 131 r->tcpdiag_timer = 1; 132 r->tcpdiag_retrans = tp->retransmits; 133 r->tcpdiag_expires = EXPIRES_IN_MS(tp->timeout); 134 } else if (tp->pending == TCP_TIME_PROBE0) { 135 r->tcpdiag_timer = 4; 136 r->tcpdiag_retrans = tp->probes_out; 137 r->tcpdiag_expires = EXPIRES_IN_MS(tp->timeout); 138 } else if (timer_pending(&sk->sk_timer)) { 139 r->tcpdiag_timer = 2; 140 r->tcpdiag_retrans = tp->probes_out; 141 r->tcpdiag_expires = EXPIRES_IN_MS(sk->sk_timer.expires); 142 } else { 143 r->tcpdiag_timer = 0; 144 r->tcpdiag_expires = 0; 145 } 146 #undef EXPIRES_IN_MS 147 148 r->tcpdiag_rqueue = tp->rcv_nxt - tp->copied_seq; 149 r->tcpdiag_wqueue = tp->write_seq - tp->snd_una; 150 r->tcpdiag_uid = sock_i_uid(sk); 151 r->tcpdiag_inode = sock_i_ino(sk); 152 153 if (minfo) { 154 minfo->tcpdiag_rmem = atomic_read(&sk->sk_rmem_alloc); 155 minfo->tcpdiag_wmem = sk->sk_wmem_queued; 156 minfo->tcpdiag_fmem = sk->sk_forward_alloc; 157 minfo->tcpdiag_tmem = atomic_read(&sk->sk_wmem_alloc); 158 } 159 160 if (info) 161 tcp_get_info(sk, info); 162 163 if (sk->sk_state < TCP_TIME_WAIT && tp->ca_ops->get_info) 164 tp->ca_ops->get_info(tp, ext, skb); 165 166 nlh->nlmsg_len = skb->tail - b; 167 return skb->len; 168 169 rtattr_failure: 170 nlmsg_failure: 171 skb_trim(skb, b - skb->data); 172 return -1; 173 } 174 175 extern struct sock *tcp_v4_lookup(u32 saddr, u16 sport, u32 daddr, u16 dport, 176 int dif); 177 #ifdef CONFIG_IP_TCPDIAG_IPV6 178 extern struct sock *tcp_v6_lookup(struct in6_addr *saddr, u16 sport, 179 struct in6_addr *daddr, u16 dport, 180 int dif); 181 #else 182 static inline struct sock *tcp_v6_lookup(struct in6_addr *saddr, u16 sport, 183 struct in6_addr *daddr, u16 dport, 184 int dif) 185 { 186 return NULL; 187 } 188 #endif 189 190 static int tcpdiag_get_exact(struct sk_buff *in_skb, const struct nlmsghdr *nlh) 191 { 192 int err; 193 struct sock *sk; 194 struct tcpdiagreq *req = NLMSG_DATA(nlh); 195 struct sk_buff *rep; 196 197 if (req->tcpdiag_family == AF_INET) { 198 sk = tcp_v4_lookup(req->id.tcpdiag_dst[0], req->id.tcpdiag_dport, 199 req->id.tcpdiag_src[0], req->id.tcpdiag_sport, 200 req->id.tcpdiag_if); 201 } 202 #ifdef CONFIG_IP_TCPDIAG_IPV6 203 else if (req->tcpdiag_family == AF_INET6) { 204 sk = tcp_v6_lookup((struct in6_addr*)req->id.tcpdiag_dst, req->id.tcpdiag_dport, 205 (struct in6_addr*)req->id.tcpdiag_src, req->id.tcpdiag_sport, 206 req->id.tcpdiag_if); 207 } 208 #endif 209 else { 210 return -EINVAL; 211 } 212 213 if (sk == NULL) 214 return -ENOENT; 215 216 err = -ESTALE; 217 if ((req->id.tcpdiag_cookie[0] != TCPDIAG_NOCOOKIE || 218 req->id.tcpdiag_cookie[1] != TCPDIAG_NOCOOKIE) && 219 ((u32)(unsigned long)sk != req->id.tcpdiag_cookie[0] || 220 (u32)((((unsigned long)sk) >> 31) >> 1) != req->id.tcpdiag_cookie[1])) 221 goto out; 222 223 err = -ENOMEM; 224 rep = alloc_skb(NLMSG_SPACE(sizeof(struct tcpdiagmsg)+ 225 sizeof(struct tcpdiag_meminfo)+ 226 sizeof(struct tcp_info)+64), GFP_KERNEL); 227 if (!rep) 228 goto out; 229 230 if (tcpdiag_fill(rep, sk, req->tcpdiag_ext, 231 NETLINK_CB(in_skb).pid, 232 nlh->nlmsg_seq, 0) <= 0) 233 BUG(); 234 235 err = netlink_unicast(tcpnl, rep, NETLINK_CB(in_skb).pid, MSG_DONTWAIT); 236 if (err > 0) 237 err = 0; 238 239 out: 240 if (sk) { 241 if (sk->sk_state == TCP_TIME_WAIT) 242 tcp_tw_put((struct tcp_tw_bucket*)sk); 243 else 244 sock_put(sk); 245 } 246 return err; 247 } 248 249 static int bitstring_match(const u32 *a1, const u32 *a2, int bits) 250 { 251 int words = bits >> 5; 252 253 bits &= 0x1f; 254 255 if (words) { 256 if (memcmp(a1, a2, words << 2)) 257 return 0; 258 } 259 if (bits) { 260 __u32 w1, w2; 261 __u32 mask; 262 263 w1 = a1[words]; 264 w2 = a2[words]; 265 266 mask = htonl((0xffffffff) << (32 - bits)); 267 268 if ((w1 ^ w2) & mask) 269 return 0; 270 } 271 272 return 1; 273 } 274 275 276 static int tcpdiag_bc_run(const void *bc, int len, 277 const struct tcpdiag_entry *entry) 278 { 279 while (len > 0) { 280 int yes = 1; 281 const struct tcpdiag_bc_op *op = bc; 282 283 switch (op->code) { 284 case TCPDIAG_BC_NOP: 285 break; 286 case TCPDIAG_BC_JMP: 287 yes = 0; 288 break; 289 case TCPDIAG_BC_S_GE: 290 yes = entry->sport >= op[1].no; 291 break; 292 case TCPDIAG_BC_S_LE: 293 yes = entry->dport <= op[1].no; 294 break; 295 case TCPDIAG_BC_D_GE: 296 yes = entry->dport >= op[1].no; 297 break; 298 case TCPDIAG_BC_D_LE: 299 yes = entry->dport <= op[1].no; 300 break; 301 case TCPDIAG_BC_AUTO: 302 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK); 303 break; 304 case TCPDIAG_BC_S_COND: 305 case TCPDIAG_BC_D_COND: 306 { 307 struct tcpdiag_hostcond *cond = (struct tcpdiag_hostcond*)(op+1); 308 u32 *addr; 309 310 if (cond->port != -1 && 311 cond->port != (op->code == TCPDIAG_BC_S_COND ? 312 entry->sport : entry->dport)) { 313 yes = 0; 314 break; 315 } 316 317 if (cond->prefix_len == 0) 318 break; 319 320 if (op->code == TCPDIAG_BC_S_COND) 321 addr = entry->saddr; 322 else 323 addr = entry->daddr; 324 325 if (bitstring_match(addr, cond->addr, cond->prefix_len)) 326 break; 327 if (entry->family == AF_INET6 && 328 cond->family == AF_INET) { 329 if (addr[0] == 0 && addr[1] == 0 && 330 addr[2] == htonl(0xffff) && 331 bitstring_match(addr+3, cond->addr, cond->prefix_len)) 332 break; 333 } 334 yes = 0; 335 break; 336 } 337 } 338 339 if (yes) { 340 len -= op->yes; 341 bc += op->yes; 342 } else { 343 len -= op->no; 344 bc += op->no; 345 } 346 } 347 return (len == 0); 348 } 349 350 static int valid_cc(const void *bc, int len, int cc) 351 { 352 while (len >= 0) { 353 const struct tcpdiag_bc_op *op = bc; 354 355 if (cc > len) 356 return 0; 357 if (cc == len) 358 return 1; 359 if (op->yes < 4) 360 return 0; 361 len -= op->yes; 362 bc += op->yes; 363 } 364 return 0; 365 } 366 367 static int tcpdiag_bc_audit(const void *bytecode, int bytecode_len) 368 { 369 const unsigned char *bc = bytecode; 370 int len = bytecode_len; 371 372 while (len > 0) { 373 struct tcpdiag_bc_op *op = (struct tcpdiag_bc_op*)bc; 374 375 //printk("BC: %d %d %d {%d} / %d\n", op->code, op->yes, op->no, op[1].no, len); 376 switch (op->code) { 377 case TCPDIAG_BC_AUTO: 378 case TCPDIAG_BC_S_COND: 379 case TCPDIAG_BC_D_COND: 380 case TCPDIAG_BC_S_GE: 381 case TCPDIAG_BC_S_LE: 382 case TCPDIAG_BC_D_GE: 383 case TCPDIAG_BC_D_LE: 384 if (op->yes < 4 || op->yes > len+4) 385 return -EINVAL; 386 case TCPDIAG_BC_JMP: 387 if (op->no < 4 || op->no > len+4) 388 return -EINVAL; 389 if (op->no < len && 390 !valid_cc(bytecode, bytecode_len, len-op->no)) 391 return -EINVAL; 392 break; 393 case TCPDIAG_BC_NOP: 394 if (op->yes < 4 || op->yes > len+4) 395 return -EINVAL; 396 break; 397 default: 398 return -EINVAL; 399 } 400 bc += op->yes; 401 len -= op->yes; 402 } 403 return len == 0 ? 0 : -EINVAL; 404 } 405 406 static int tcpdiag_dump_sock(struct sk_buff *skb, struct sock *sk, 407 struct netlink_callback *cb) 408 { 409 struct tcpdiagreq *r = NLMSG_DATA(cb->nlh); 410 411 if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) { 412 struct tcpdiag_entry entry; 413 struct rtattr *bc = (struct rtattr *)(r + 1); 414 struct inet_sock *inet = inet_sk(sk); 415 416 entry.family = sk->sk_family; 417 #ifdef CONFIG_IP_TCPDIAG_IPV6 418 if (entry.family == AF_INET6) { 419 struct ipv6_pinfo *np = inet6_sk(sk); 420 421 entry.saddr = np->rcv_saddr.s6_addr32; 422 entry.daddr = np->daddr.s6_addr32; 423 } else 424 #endif 425 { 426 entry.saddr = &inet->rcv_saddr; 427 entry.daddr = &inet->daddr; 428 } 429 entry.sport = inet->num; 430 entry.dport = ntohs(inet->dport); 431 entry.userlocks = sk->sk_userlocks; 432 433 if (!tcpdiag_bc_run(RTA_DATA(bc), RTA_PAYLOAD(bc), &entry)) 434 return 0; 435 } 436 437 return tcpdiag_fill(skb, sk, r->tcpdiag_ext, NETLINK_CB(cb->skb).pid, 438 cb->nlh->nlmsg_seq, NLM_F_MULTI); 439 } 440 441 static int tcpdiag_fill_req(struct sk_buff *skb, struct sock *sk, 442 struct request_sock *req, 443 u32 pid, u32 seq) 444 { 445 const struct inet_request_sock *ireq = inet_rsk(req); 446 struct inet_sock *inet = inet_sk(sk); 447 unsigned char *b = skb->tail; 448 struct tcpdiagmsg *r; 449 struct nlmsghdr *nlh; 450 long tmo; 451 452 nlh = NLMSG_PUT(skb, pid, seq, TCPDIAG_GETSOCK, sizeof(*r)); 453 nlh->nlmsg_flags = NLM_F_MULTI; 454 r = NLMSG_DATA(nlh); 455 456 r->tcpdiag_family = sk->sk_family; 457 r->tcpdiag_state = TCP_SYN_RECV; 458 r->tcpdiag_timer = 1; 459 r->tcpdiag_retrans = req->retrans; 460 461 r->id.tcpdiag_if = sk->sk_bound_dev_if; 462 r->id.tcpdiag_cookie[0] = (u32)(unsigned long)req; 463 r->id.tcpdiag_cookie[1] = (u32)(((unsigned long)req >> 31) >> 1); 464 465 tmo = req->expires - jiffies; 466 if (tmo < 0) 467 tmo = 0; 468 469 r->id.tcpdiag_sport = inet->sport; 470 r->id.tcpdiag_dport = ireq->rmt_port; 471 r->id.tcpdiag_src[0] = ireq->loc_addr; 472 r->id.tcpdiag_dst[0] = ireq->rmt_addr; 473 r->tcpdiag_expires = jiffies_to_msecs(tmo), 474 r->tcpdiag_rqueue = 0; 475 r->tcpdiag_wqueue = 0; 476 r->tcpdiag_uid = sock_i_uid(sk); 477 r->tcpdiag_inode = 0; 478 #ifdef CONFIG_IP_TCPDIAG_IPV6 479 if (r->tcpdiag_family == AF_INET6) { 480 ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_src, 481 &tcp6_rsk(req)->loc_addr); 482 ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_dst, 483 &tcp6_rsk(req)->rmt_addr); 484 } 485 #endif 486 nlh->nlmsg_len = skb->tail - b; 487 488 return skb->len; 489 490 nlmsg_failure: 491 skb_trim(skb, b - skb->data); 492 return -1; 493 } 494 495 static int tcpdiag_dump_reqs(struct sk_buff *skb, struct sock *sk, 496 struct netlink_callback *cb) 497 { 498 struct tcpdiag_entry entry; 499 struct tcpdiagreq *r = NLMSG_DATA(cb->nlh); 500 struct tcp_sock *tp = tcp_sk(sk); 501 struct listen_sock *lopt; 502 struct rtattr *bc = NULL; 503 struct inet_sock *inet = inet_sk(sk); 504 int j, s_j; 505 int reqnum, s_reqnum; 506 int err = 0; 507 508 s_j = cb->args[3]; 509 s_reqnum = cb->args[4]; 510 511 if (s_j > 0) 512 s_j--; 513 514 entry.family = sk->sk_family; 515 516 read_lock_bh(&tp->accept_queue.syn_wait_lock); 517 518 lopt = tp->accept_queue.listen_opt; 519 if (!lopt || !lopt->qlen) 520 goto out; 521 522 if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) { 523 bc = (struct rtattr *)(r + 1); 524 entry.sport = inet->num; 525 entry.userlocks = sk->sk_userlocks; 526 } 527 528 for (j = s_j; j < TCP_SYNQ_HSIZE; j++) { 529 struct request_sock *req, *head = lopt->syn_table[j]; 530 531 reqnum = 0; 532 for (req = head; req; reqnum++, req = req->dl_next) { 533 struct inet_request_sock *ireq = inet_rsk(req); 534 535 if (reqnum < s_reqnum) 536 continue; 537 if (r->id.tcpdiag_dport != ireq->rmt_port && 538 r->id.tcpdiag_dport) 539 continue; 540 541 if (bc) { 542 entry.saddr = 543 #ifdef CONFIG_IP_TCPDIAG_IPV6 544 (entry.family == AF_INET6) ? 545 tcp6_rsk(req)->loc_addr.s6_addr32 : 546 #endif 547 &ireq->loc_addr; 548 entry.daddr = 549 #ifdef CONFIG_IP_TCPDIAG_IPV6 550 (entry.family == AF_INET6) ? 551 tcp6_rsk(req)->rmt_addr.s6_addr32 : 552 #endif 553 &ireq->rmt_addr; 554 entry.dport = ntohs(ireq->rmt_port); 555 556 if (!tcpdiag_bc_run(RTA_DATA(bc), 557 RTA_PAYLOAD(bc), &entry)) 558 continue; 559 } 560 561 err = tcpdiag_fill_req(skb, sk, req, 562 NETLINK_CB(cb->skb).pid, 563 cb->nlh->nlmsg_seq); 564 if (err < 0) { 565 cb->args[3] = j + 1; 566 cb->args[4] = reqnum; 567 goto out; 568 } 569 } 570 571 s_reqnum = 0; 572 } 573 574 out: 575 read_unlock_bh(&tp->accept_queue.syn_wait_lock); 576 577 return err; 578 } 579 580 static int tcpdiag_dump(struct sk_buff *skb, struct netlink_callback *cb) 581 { 582 int i, num; 583 int s_i, s_num; 584 struct tcpdiagreq *r = NLMSG_DATA(cb->nlh); 585 586 s_i = cb->args[1]; 587 s_num = num = cb->args[2]; 588 589 if (cb->args[0] == 0) { 590 if (!(r->tcpdiag_states&(TCPF_LISTEN|TCPF_SYN_RECV))) 591 goto skip_listen_ht; 592 inet_listen_lock(&tcp_hashinfo); 593 for (i = s_i; i < INET_LHTABLE_SIZE; i++) { 594 struct sock *sk; 595 struct hlist_node *node; 596 597 num = 0; 598 sk_for_each(sk, node, &tcp_hashinfo.listening_hash[i]) { 599 struct inet_sock *inet = inet_sk(sk); 600 601 if (num < s_num) { 602 num++; 603 continue; 604 } 605 606 if (r->id.tcpdiag_sport != inet->sport && 607 r->id.tcpdiag_sport) 608 goto next_listen; 609 610 if (!(r->tcpdiag_states&TCPF_LISTEN) || 611 r->id.tcpdiag_dport || 612 cb->args[3] > 0) 613 goto syn_recv; 614 615 if (tcpdiag_dump_sock(skb, sk, cb) < 0) { 616 inet_listen_unlock(&tcp_hashinfo); 617 goto done; 618 } 619 620 syn_recv: 621 if (!(r->tcpdiag_states&TCPF_SYN_RECV)) 622 goto next_listen; 623 624 if (tcpdiag_dump_reqs(skb, sk, cb) < 0) { 625 inet_listen_unlock(&tcp_hashinfo); 626 goto done; 627 } 628 629 next_listen: 630 cb->args[3] = 0; 631 cb->args[4] = 0; 632 ++num; 633 } 634 635 s_num = 0; 636 cb->args[3] = 0; 637 cb->args[4] = 0; 638 } 639 inet_listen_unlock(&tcp_hashinfo); 640 skip_listen_ht: 641 cb->args[0] = 1; 642 s_i = num = s_num = 0; 643 } 644 645 if (!(r->tcpdiag_states&~(TCPF_LISTEN|TCPF_SYN_RECV))) 646 return skb->len; 647 648 for (i = s_i; i < tcp_hashinfo.ehash_size; i++) { 649 struct inet_ehash_bucket *head = &tcp_hashinfo.ehash[i]; 650 struct sock *sk; 651 struct hlist_node *node; 652 653 if (i > s_i) 654 s_num = 0; 655 656 read_lock_bh(&head->lock); 657 658 num = 0; 659 sk_for_each(sk, node, &head->chain) { 660 struct inet_sock *inet = inet_sk(sk); 661 662 if (num < s_num) 663 goto next_normal; 664 if (!(r->tcpdiag_states & (1 << sk->sk_state))) 665 goto next_normal; 666 if (r->id.tcpdiag_sport != inet->sport && 667 r->id.tcpdiag_sport) 668 goto next_normal; 669 if (r->id.tcpdiag_dport != inet->dport && r->id.tcpdiag_dport) 670 goto next_normal; 671 if (tcpdiag_dump_sock(skb, sk, cb) < 0) { 672 read_unlock_bh(&head->lock); 673 goto done; 674 } 675 next_normal: 676 ++num; 677 } 678 679 if (r->tcpdiag_states&TCPF_TIME_WAIT) { 680 sk_for_each(sk, node, 681 &tcp_hashinfo.ehash[i + tcp_hashinfo.ehash_size].chain) { 682 struct inet_sock *inet = inet_sk(sk); 683 684 if (num < s_num) 685 goto next_dying; 686 if (r->id.tcpdiag_sport != inet->sport && 687 r->id.tcpdiag_sport) 688 goto next_dying; 689 if (r->id.tcpdiag_dport != inet->dport && 690 r->id.tcpdiag_dport) 691 goto next_dying; 692 if (tcpdiag_dump_sock(skb, sk, cb) < 0) { 693 read_unlock_bh(&head->lock); 694 goto done; 695 } 696 next_dying: 697 ++num; 698 } 699 } 700 read_unlock_bh(&head->lock); 701 } 702 703 done: 704 cb->args[1] = i; 705 cb->args[2] = num; 706 return skb->len; 707 } 708 709 static int tcpdiag_dump_done(struct netlink_callback *cb) 710 { 711 return 0; 712 } 713 714 715 static __inline__ int 716 tcpdiag_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh) 717 { 718 if (!(nlh->nlmsg_flags&NLM_F_REQUEST)) 719 return 0; 720 721 if (nlh->nlmsg_type != TCPDIAG_GETSOCK) 722 goto err_inval; 723 724 if (NLMSG_LENGTH(sizeof(struct tcpdiagreq)) > skb->len) 725 goto err_inval; 726 727 if (nlh->nlmsg_flags&NLM_F_DUMP) { 728 if (nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(struct tcpdiagreq))) { 729 struct rtattr *rta = (struct rtattr*)(NLMSG_DATA(nlh) + sizeof(struct tcpdiagreq)); 730 if (rta->rta_type != TCPDIAG_REQ_BYTECODE || 731 rta->rta_len < 8 || 732 rta->rta_len > nlh->nlmsg_len - NLMSG_SPACE(sizeof(struct tcpdiagreq))) 733 goto err_inval; 734 if (tcpdiag_bc_audit(RTA_DATA(rta), RTA_PAYLOAD(rta))) 735 goto err_inval; 736 } 737 return netlink_dump_start(tcpnl, skb, nlh, 738 tcpdiag_dump, 739 tcpdiag_dump_done); 740 } else { 741 return tcpdiag_get_exact(skb, nlh); 742 } 743 744 err_inval: 745 return -EINVAL; 746 } 747 748 749 static inline void tcpdiag_rcv_skb(struct sk_buff *skb) 750 { 751 int err; 752 struct nlmsghdr * nlh; 753 754 if (skb->len >= NLMSG_SPACE(0)) { 755 nlh = (struct nlmsghdr *)skb->data; 756 if (nlh->nlmsg_len < sizeof(*nlh) || skb->len < nlh->nlmsg_len) 757 return; 758 err = tcpdiag_rcv_msg(skb, nlh); 759 if (err || nlh->nlmsg_flags & NLM_F_ACK) 760 netlink_ack(skb, nlh, err); 761 } 762 } 763 764 static void tcpdiag_rcv(struct sock *sk, int len) 765 { 766 struct sk_buff *skb; 767 unsigned int qlen = skb_queue_len(&sk->sk_receive_queue); 768 769 while (qlen-- && (skb = skb_dequeue(&sk->sk_receive_queue))) { 770 tcpdiag_rcv_skb(skb); 771 kfree_skb(skb); 772 } 773 } 774 775 static int __init tcpdiag_init(void) 776 { 777 tcpnl = netlink_kernel_create(NETLINK_TCPDIAG, tcpdiag_rcv, 778 THIS_MODULE); 779 if (tcpnl == NULL) 780 return -ENOMEM; 781 return 0; 782 } 783 784 static void __exit tcpdiag_exit(void) 785 { 786 sock_release(tcpnl->sk_socket); 787 } 788 789 module_init(tcpdiag_init); 790 module_exit(tcpdiag_exit); 791 MODULE_LICENSE("GPL"); 792