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 46 #define TCPDIAG_PUT(skb, attrtype, attrlen) \ 47 ({ int rtalen = RTA_LENGTH(attrlen); \ 48 struct rtattr *rta; \ 49 if (skb_tailroom(skb) < RTA_ALIGN(rtalen)) goto nlmsg_failure; \ 50 rta = (void*)__skb_put(skb, RTA_ALIGN(rtalen)); \ 51 rta->rta_type = attrtype; \ 52 rta->rta_len = rtalen; \ 53 RTA_DATA(rta); }) 54 55 static int tcpdiag_fill(struct sk_buff *skb, struct sock *sk, 56 int ext, u32 pid, u32 seq, u16 nlmsg_flags) 57 { 58 struct inet_sock *inet = inet_sk(sk); 59 struct tcp_sock *tp = tcp_sk(sk); 60 struct tcpdiagmsg *r; 61 struct nlmsghdr *nlh; 62 struct tcp_info *info = NULL; 63 struct tcpdiag_meminfo *minfo = NULL; 64 struct tcpvegas_info *vinfo = NULL; 65 unsigned char *b = skb->tail; 66 67 nlh = NLMSG_PUT(skb, pid, seq, TCPDIAG_GETSOCK, sizeof(*r)); 68 nlh->nlmsg_flags = nlmsg_flags; 69 r = NLMSG_DATA(nlh); 70 if (sk->sk_state != TCP_TIME_WAIT) { 71 if (ext & (1<<(TCPDIAG_MEMINFO-1))) 72 minfo = TCPDIAG_PUT(skb, TCPDIAG_MEMINFO, sizeof(*minfo)); 73 if (ext & (1<<(TCPDIAG_INFO-1))) 74 info = TCPDIAG_PUT(skb, TCPDIAG_INFO, sizeof(*info)); 75 76 if ((tcp_is_westwood(tp) || tcp_is_vegas(tp)) 77 && (ext & (1<<(TCPDIAG_VEGASINFO-1)))) 78 vinfo = TCPDIAG_PUT(skb, TCPDIAG_VEGASINFO, sizeof(*vinfo)); 79 } 80 r->tcpdiag_family = sk->sk_family; 81 r->tcpdiag_state = sk->sk_state; 82 r->tcpdiag_timer = 0; 83 r->tcpdiag_retrans = 0; 84 85 r->id.tcpdiag_if = sk->sk_bound_dev_if; 86 r->id.tcpdiag_cookie[0] = (u32)(unsigned long)sk; 87 r->id.tcpdiag_cookie[1] = (u32)(((unsigned long)sk >> 31) >> 1); 88 89 if (r->tcpdiag_state == TCP_TIME_WAIT) { 90 struct tcp_tw_bucket *tw = (struct tcp_tw_bucket*)sk; 91 long tmo = tw->tw_ttd - jiffies; 92 if (tmo < 0) 93 tmo = 0; 94 95 r->id.tcpdiag_sport = tw->tw_sport; 96 r->id.tcpdiag_dport = tw->tw_dport; 97 r->id.tcpdiag_src[0] = tw->tw_rcv_saddr; 98 r->id.tcpdiag_dst[0] = tw->tw_daddr; 99 r->tcpdiag_state = tw->tw_substate; 100 r->tcpdiag_timer = 3; 101 r->tcpdiag_expires = (tmo*1000+HZ-1)/HZ; 102 r->tcpdiag_rqueue = 0; 103 r->tcpdiag_wqueue = 0; 104 r->tcpdiag_uid = 0; 105 r->tcpdiag_inode = 0; 106 #ifdef CONFIG_IP_TCPDIAG_IPV6 107 if (r->tcpdiag_family == AF_INET6) { 108 ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_src, 109 &tw->tw_v6_rcv_saddr); 110 ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_dst, 111 &tw->tw_v6_daddr); 112 } 113 #endif 114 nlh->nlmsg_len = skb->tail - b; 115 return skb->len; 116 } 117 118 r->id.tcpdiag_sport = inet->sport; 119 r->id.tcpdiag_dport = inet->dport; 120 r->id.tcpdiag_src[0] = inet->rcv_saddr; 121 r->id.tcpdiag_dst[0] = inet->daddr; 122 123 #ifdef CONFIG_IP_TCPDIAG_IPV6 124 if (r->tcpdiag_family == AF_INET6) { 125 struct ipv6_pinfo *np = inet6_sk(sk); 126 127 ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_src, 128 &np->rcv_saddr); 129 ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_dst, 130 &np->daddr); 131 } 132 #endif 133 134 #define EXPIRES_IN_MS(tmo) ((tmo-jiffies)*1000+HZ-1)/HZ 135 136 if (tp->pending == TCP_TIME_RETRANS) { 137 r->tcpdiag_timer = 1; 138 r->tcpdiag_retrans = tp->retransmits; 139 r->tcpdiag_expires = EXPIRES_IN_MS(tp->timeout); 140 } else if (tp->pending == TCP_TIME_PROBE0) { 141 r->tcpdiag_timer = 4; 142 r->tcpdiag_retrans = tp->probes_out; 143 r->tcpdiag_expires = EXPIRES_IN_MS(tp->timeout); 144 } else if (timer_pending(&sk->sk_timer)) { 145 r->tcpdiag_timer = 2; 146 r->tcpdiag_retrans = tp->probes_out; 147 r->tcpdiag_expires = EXPIRES_IN_MS(sk->sk_timer.expires); 148 } else { 149 r->tcpdiag_timer = 0; 150 r->tcpdiag_expires = 0; 151 } 152 #undef EXPIRES_IN_MS 153 154 r->tcpdiag_rqueue = tp->rcv_nxt - tp->copied_seq; 155 r->tcpdiag_wqueue = tp->write_seq - tp->snd_una; 156 r->tcpdiag_uid = sock_i_uid(sk); 157 r->tcpdiag_inode = sock_i_ino(sk); 158 159 if (minfo) { 160 minfo->tcpdiag_rmem = atomic_read(&sk->sk_rmem_alloc); 161 minfo->tcpdiag_wmem = sk->sk_wmem_queued; 162 minfo->tcpdiag_fmem = sk->sk_forward_alloc; 163 minfo->tcpdiag_tmem = atomic_read(&sk->sk_wmem_alloc); 164 } 165 166 if (info) 167 tcp_get_info(sk, info); 168 169 if (vinfo) { 170 if (tcp_is_vegas(tp)) { 171 vinfo->tcpv_enabled = tp->vegas.doing_vegas_now; 172 vinfo->tcpv_rttcnt = tp->vegas.cntRTT; 173 vinfo->tcpv_rtt = jiffies_to_usecs(tp->vegas.baseRTT); 174 vinfo->tcpv_minrtt = jiffies_to_usecs(tp->vegas.minRTT); 175 } else { 176 vinfo->tcpv_enabled = 0; 177 vinfo->tcpv_rttcnt = 0; 178 vinfo->tcpv_rtt = jiffies_to_usecs(tp->westwood.rtt); 179 vinfo->tcpv_minrtt = jiffies_to_usecs(tp->westwood.rtt_min); 180 } 181 } 182 183 nlh->nlmsg_len = skb->tail - b; 184 return skb->len; 185 186 nlmsg_failure: 187 skb_trim(skb, b - skb->data); 188 return -1; 189 } 190 191 extern struct sock *tcp_v4_lookup(u32 saddr, u16 sport, u32 daddr, u16 dport, 192 int dif); 193 #ifdef CONFIG_IP_TCPDIAG_IPV6 194 extern struct sock *tcp_v6_lookup(struct in6_addr *saddr, u16 sport, 195 struct in6_addr *daddr, u16 dport, 196 int dif); 197 #else 198 static inline struct sock *tcp_v6_lookup(struct in6_addr *saddr, u16 sport, 199 struct in6_addr *daddr, u16 dport, 200 int dif) 201 { 202 return NULL; 203 } 204 #endif 205 206 static int tcpdiag_get_exact(struct sk_buff *in_skb, const struct nlmsghdr *nlh) 207 { 208 int err; 209 struct sock *sk; 210 struct tcpdiagreq *req = NLMSG_DATA(nlh); 211 struct sk_buff *rep; 212 213 if (req->tcpdiag_family == AF_INET) { 214 sk = tcp_v4_lookup(req->id.tcpdiag_dst[0], req->id.tcpdiag_dport, 215 req->id.tcpdiag_src[0], req->id.tcpdiag_sport, 216 req->id.tcpdiag_if); 217 } 218 #ifdef CONFIG_IP_TCPDIAG_IPV6 219 else if (req->tcpdiag_family == AF_INET6) { 220 sk = tcp_v6_lookup((struct in6_addr*)req->id.tcpdiag_dst, req->id.tcpdiag_dport, 221 (struct in6_addr*)req->id.tcpdiag_src, req->id.tcpdiag_sport, 222 req->id.tcpdiag_if); 223 } 224 #endif 225 else { 226 return -EINVAL; 227 } 228 229 if (sk == NULL) 230 return -ENOENT; 231 232 err = -ESTALE; 233 if ((req->id.tcpdiag_cookie[0] != TCPDIAG_NOCOOKIE || 234 req->id.tcpdiag_cookie[1] != TCPDIAG_NOCOOKIE) && 235 ((u32)(unsigned long)sk != req->id.tcpdiag_cookie[0] || 236 (u32)((((unsigned long)sk) >> 31) >> 1) != req->id.tcpdiag_cookie[1])) 237 goto out; 238 239 err = -ENOMEM; 240 rep = alloc_skb(NLMSG_SPACE(sizeof(struct tcpdiagmsg)+ 241 sizeof(struct tcpdiag_meminfo)+ 242 sizeof(struct tcp_info)+64), GFP_KERNEL); 243 if (!rep) 244 goto out; 245 246 if (tcpdiag_fill(rep, sk, req->tcpdiag_ext, 247 NETLINK_CB(in_skb).pid, 248 nlh->nlmsg_seq, 0) <= 0) 249 BUG(); 250 251 err = netlink_unicast(tcpnl, rep, NETLINK_CB(in_skb).pid, MSG_DONTWAIT); 252 if (err > 0) 253 err = 0; 254 255 out: 256 if (sk) { 257 if (sk->sk_state == TCP_TIME_WAIT) 258 tcp_tw_put((struct tcp_tw_bucket*)sk); 259 else 260 sock_put(sk); 261 } 262 return err; 263 } 264 265 static int bitstring_match(const u32 *a1, const u32 *a2, int bits) 266 { 267 int words = bits >> 5; 268 269 bits &= 0x1f; 270 271 if (words) { 272 if (memcmp(a1, a2, words << 2)) 273 return 0; 274 } 275 if (bits) { 276 __u32 w1, w2; 277 __u32 mask; 278 279 w1 = a1[words]; 280 w2 = a2[words]; 281 282 mask = htonl((0xffffffff) << (32 - bits)); 283 284 if ((w1 ^ w2) & mask) 285 return 0; 286 } 287 288 return 1; 289 } 290 291 292 static int tcpdiag_bc_run(const void *bc, int len, 293 const struct tcpdiag_entry *entry) 294 { 295 while (len > 0) { 296 int yes = 1; 297 const struct tcpdiag_bc_op *op = bc; 298 299 switch (op->code) { 300 case TCPDIAG_BC_NOP: 301 break; 302 case TCPDIAG_BC_JMP: 303 yes = 0; 304 break; 305 case TCPDIAG_BC_S_GE: 306 yes = entry->sport >= op[1].no; 307 break; 308 case TCPDIAG_BC_S_LE: 309 yes = entry->dport <= op[1].no; 310 break; 311 case TCPDIAG_BC_D_GE: 312 yes = entry->dport >= op[1].no; 313 break; 314 case TCPDIAG_BC_D_LE: 315 yes = entry->dport <= op[1].no; 316 break; 317 case TCPDIAG_BC_AUTO: 318 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK); 319 break; 320 case TCPDIAG_BC_S_COND: 321 case TCPDIAG_BC_D_COND: 322 { 323 struct tcpdiag_hostcond *cond = (struct tcpdiag_hostcond*)(op+1); 324 u32 *addr; 325 326 if (cond->port != -1 && 327 cond->port != (op->code == TCPDIAG_BC_S_COND ? 328 entry->sport : entry->dport)) { 329 yes = 0; 330 break; 331 } 332 333 if (cond->prefix_len == 0) 334 break; 335 336 if (op->code == TCPDIAG_BC_S_COND) 337 addr = entry->saddr; 338 else 339 addr = entry->daddr; 340 341 if (bitstring_match(addr, cond->addr, cond->prefix_len)) 342 break; 343 if (entry->family == AF_INET6 && 344 cond->family == AF_INET) { 345 if (addr[0] == 0 && addr[1] == 0 && 346 addr[2] == htonl(0xffff) && 347 bitstring_match(addr+3, cond->addr, cond->prefix_len)) 348 break; 349 } 350 yes = 0; 351 break; 352 } 353 } 354 355 if (yes) { 356 len -= op->yes; 357 bc += op->yes; 358 } else { 359 len -= op->no; 360 bc += op->no; 361 } 362 } 363 return (len == 0); 364 } 365 366 static int valid_cc(const void *bc, int len, int cc) 367 { 368 while (len >= 0) { 369 const struct tcpdiag_bc_op *op = bc; 370 371 if (cc > len) 372 return 0; 373 if (cc == len) 374 return 1; 375 if (op->yes < 4) 376 return 0; 377 len -= op->yes; 378 bc += op->yes; 379 } 380 return 0; 381 } 382 383 static int tcpdiag_bc_audit(const void *bytecode, int bytecode_len) 384 { 385 const unsigned char *bc = bytecode; 386 int len = bytecode_len; 387 388 while (len > 0) { 389 struct tcpdiag_bc_op *op = (struct tcpdiag_bc_op*)bc; 390 391 //printk("BC: %d %d %d {%d} / %d\n", op->code, op->yes, op->no, op[1].no, len); 392 switch (op->code) { 393 case TCPDIAG_BC_AUTO: 394 case TCPDIAG_BC_S_COND: 395 case TCPDIAG_BC_D_COND: 396 case TCPDIAG_BC_S_GE: 397 case TCPDIAG_BC_S_LE: 398 case TCPDIAG_BC_D_GE: 399 case TCPDIAG_BC_D_LE: 400 if (op->yes < 4 || op->yes > len+4) 401 return -EINVAL; 402 case TCPDIAG_BC_JMP: 403 if (op->no < 4 || op->no > len+4) 404 return -EINVAL; 405 if (op->no < len && 406 !valid_cc(bytecode, bytecode_len, len-op->no)) 407 return -EINVAL; 408 break; 409 case TCPDIAG_BC_NOP: 410 if (op->yes < 4 || op->yes > len+4) 411 return -EINVAL; 412 break; 413 default: 414 return -EINVAL; 415 } 416 bc += op->yes; 417 len -= op->yes; 418 } 419 return len == 0 ? 0 : -EINVAL; 420 } 421 422 static int tcpdiag_dump_sock(struct sk_buff *skb, struct sock *sk, 423 struct netlink_callback *cb) 424 { 425 struct tcpdiagreq *r = NLMSG_DATA(cb->nlh); 426 427 if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) { 428 struct tcpdiag_entry entry; 429 struct rtattr *bc = (struct rtattr *)(r + 1); 430 struct inet_sock *inet = inet_sk(sk); 431 432 entry.family = sk->sk_family; 433 #ifdef CONFIG_IP_TCPDIAG_IPV6 434 if (entry.family == AF_INET6) { 435 struct ipv6_pinfo *np = inet6_sk(sk); 436 437 entry.saddr = np->rcv_saddr.s6_addr32; 438 entry.daddr = np->daddr.s6_addr32; 439 } else 440 #endif 441 { 442 entry.saddr = &inet->rcv_saddr; 443 entry.daddr = &inet->daddr; 444 } 445 entry.sport = inet->num; 446 entry.dport = ntohs(inet->dport); 447 entry.userlocks = sk->sk_userlocks; 448 449 if (!tcpdiag_bc_run(RTA_DATA(bc), RTA_PAYLOAD(bc), &entry)) 450 return 0; 451 } 452 453 return tcpdiag_fill(skb, sk, r->tcpdiag_ext, NETLINK_CB(cb->skb).pid, 454 cb->nlh->nlmsg_seq, NLM_F_MULTI); 455 } 456 457 static int tcpdiag_fill_req(struct sk_buff *skb, struct sock *sk, 458 struct open_request *req, 459 u32 pid, u32 seq) 460 { 461 struct inet_sock *inet = inet_sk(sk); 462 unsigned char *b = skb->tail; 463 struct tcpdiagmsg *r; 464 struct nlmsghdr *nlh; 465 long tmo; 466 467 nlh = NLMSG_PUT(skb, pid, seq, TCPDIAG_GETSOCK, sizeof(*r)); 468 nlh->nlmsg_flags = NLM_F_MULTI; 469 r = NLMSG_DATA(nlh); 470 471 r->tcpdiag_family = sk->sk_family; 472 r->tcpdiag_state = TCP_SYN_RECV; 473 r->tcpdiag_timer = 1; 474 r->tcpdiag_retrans = req->retrans; 475 476 r->id.tcpdiag_if = sk->sk_bound_dev_if; 477 r->id.tcpdiag_cookie[0] = (u32)(unsigned long)req; 478 r->id.tcpdiag_cookie[1] = (u32)(((unsigned long)req >> 31) >> 1); 479 480 tmo = req->expires - jiffies; 481 if (tmo < 0) 482 tmo = 0; 483 484 r->id.tcpdiag_sport = inet->sport; 485 r->id.tcpdiag_dport = req->rmt_port; 486 r->id.tcpdiag_src[0] = req->af.v4_req.loc_addr; 487 r->id.tcpdiag_dst[0] = req->af.v4_req.rmt_addr; 488 r->tcpdiag_expires = jiffies_to_msecs(tmo), 489 r->tcpdiag_rqueue = 0; 490 r->tcpdiag_wqueue = 0; 491 r->tcpdiag_uid = sock_i_uid(sk); 492 r->tcpdiag_inode = 0; 493 #ifdef CONFIG_IP_TCPDIAG_IPV6 494 if (r->tcpdiag_family == AF_INET6) { 495 ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_src, 496 &req->af.v6_req.loc_addr); 497 ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_dst, 498 &req->af.v6_req.rmt_addr); 499 } 500 #endif 501 nlh->nlmsg_len = skb->tail - b; 502 503 return skb->len; 504 505 nlmsg_failure: 506 skb_trim(skb, b - skb->data); 507 return -1; 508 } 509 510 static int tcpdiag_dump_reqs(struct sk_buff *skb, struct sock *sk, 511 struct netlink_callback *cb) 512 { 513 struct tcpdiag_entry entry; 514 struct tcpdiagreq *r = NLMSG_DATA(cb->nlh); 515 struct tcp_sock *tp = tcp_sk(sk); 516 struct tcp_listen_opt *lopt; 517 struct rtattr *bc = NULL; 518 struct inet_sock *inet = inet_sk(sk); 519 int j, s_j; 520 int reqnum, s_reqnum; 521 int err = 0; 522 523 s_j = cb->args[3]; 524 s_reqnum = cb->args[4]; 525 526 if (s_j > 0) 527 s_j--; 528 529 entry.family = sk->sk_family; 530 531 read_lock_bh(&tp->syn_wait_lock); 532 533 lopt = tp->listen_opt; 534 if (!lopt || !lopt->qlen) 535 goto out; 536 537 if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) { 538 bc = (struct rtattr *)(r + 1); 539 entry.sport = inet->num; 540 entry.userlocks = sk->sk_userlocks; 541 } 542 543 for (j = s_j; j < TCP_SYNQ_HSIZE; j++) { 544 struct open_request *req, *head = lopt->syn_table[j]; 545 546 reqnum = 0; 547 for (req = head; req; reqnum++, req = req->dl_next) { 548 if (reqnum < s_reqnum) 549 continue; 550 if (r->id.tcpdiag_dport != req->rmt_port && 551 r->id.tcpdiag_dport) 552 continue; 553 554 if (bc) { 555 entry.saddr = 556 #ifdef CONFIG_IP_TCPDIAG_IPV6 557 (entry.family == AF_INET6) ? 558 req->af.v6_req.loc_addr.s6_addr32 : 559 #endif 560 &req->af.v4_req.loc_addr; 561 entry.daddr = 562 #ifdef CONFIG_IP_TCPDIAG_IPV6 563 (entry.family == AF_INET6) ? 564 req->af.v6_req.rmt_addr.s6_addr32 : 565 #endif 566 &req->af.v4_req.rmt_addr; 567 entry.dport = ntohs(req->rmt_port); 568 569 if (!tcpdiag_bc_run(RTA_DATA(bc), 570 RTA_PAYLOAD(bc), &entry)) 571 continue; 572 } 573 574 err = tcpdiag_fill_req(skb, sk, req, 575 NETLINK_CB(cb->skb).pid, 576 cb->nlh->nlmsg_seq); 577 if (err < 0) { 578 cb->args[3] = j + 1; 579 cb->args[4] = reqnum; 580 goto out; 581 } 582 } 583 584 s_reqnum = 0; 585 } 586 587 out: 588 read_unlock_bh(&tp->syn_wait_lock); 589 590 return err; 591 } 592 593 static int tcpdiag_dump(struct sk_buff *skb, struct netlink_callback *cb) 594 { 595 int i, num; 596 int s_i, s_num; 597 struct tcpdiagreq *r = NLMSG_DATA(cb->nlh); 598 599 s_i = cb->args[1]; 600 s_num = num = cb->args[2]; 601 602 if (cb->args[0] == 0) { 603 if (!(r->tcpdiag_states&(TCPF_LISTEN|TCPF_SYN_RECV))) 604 goto skip_listen_ht; 605 tcp_listen_lock(); 606 for (i = s_i; i < TCP_LHTABLE_SIZE; i++) { 607 struct sock *sk; 608 struct hlist_node *node; 609 610 num = 0; 611 sk_for_each(sk, node, &tcp_listening_hash[i]) { 612 struct inet_sock *inet = inet_sk(sk); 613 614 if (num < s_num) { 615 num++; 616 continue; 617 } 618 619 if (r->id.tcpdiag_sport != inet->sport && 620 r->id.tcpdiag_sport) 621 goto next_listen; 622 623 if (!(r->tcpdiag_states&TCPF_LISTEN) || 624 r->id.tcpdiag_dport || 625 cb->args[3] > 0) 626 goto syn_recv; 627 628 if (tcpdiag_dump_sock(skb, sk, cb) < 0) { 629 tcp_listen_unlock(); 630 goto done; 631 } 632 633 syn_recv: 634 if (!(r->tcpdiag_states&TCPF_SYN_RECV)) 635 goto next_listen; 636 637 if (tcpdiag_dump_reqs(skb, sk, cb) < 0) { 638 tcp_listen_unlock(); 639 goto done; 640 } 641 642 next_listen: 643 cb->args[3] = 0; 644 cb->args[4] = 0; 645 ++num; 646 } 647 648 s_num = 0; 649 cb->args[3] = 0; 650 cb->args[4] = 0; 651 } 652 tcp_listen_unlock(); 653 skip_listen_ht: 654 cb->args[0] = 1; 655 s_i = num = s_num = 0; 656 } 657 658 if (!(r->tcpdiag_states&~(TCPF_LISTEN|TCPF_SYN_RECV))) 659 return skb->len; 660 661 for (i = s_i; i < tcp_ehash_size; i++) { 662 struct tcp_ehash_bucket *head = &tcp_ehash[i]; 663 struct sock *sk; 664 struct hlist_node *node; 665 666 if (i > s_i) 667 s_num = 0; 668 669 read_lock_bh(&head->lock); 670 671 num = 0; 672 sk_for_each(sk, node, &head->chain) { 673 struct inet_sock *inet = inet_sk(sk); 674 675 if (num < s_num) 676 goto next_normal; 677 if (!(r->tcpdiag_states & (1 << sk->sk_state))) 678 goto next_normal; 679 if (r->id.tcpdiag_sport != inet->sport && 680 r->id.tcpdiag_sport) 681 goto next_normal; 682 if (r->id.tcpdiag_dport != inet->dport && r->id.tcpdiag_dport) 683 goto next_normal; 684 if (tcpdiag_dump_sock(skb, sk, cb) < 0) { 685 read_unlock_bh(&head->lock); 686 goto done; 687 } 688 next_normal: 689 ++num; 690 } 691 692 if (r->tcpdiag_states&TCPF_TIME_WAIT) { 693 sk_for_each(sk, node, 694 &tcp_ehash[i + tcp_ehash_size].chain) { 695 struct inet_sock *inet = inet_sk(sk); 696 697 if (num < s_num) 698 goto next_dying; 699 if (r->id.tcpdiag_sport != inet->sport && 700 r->id.tcpdiag_sport) 701 goto next_dying; 702 if (r->id.tcpdiag_dport != inet->dport && 703 r->id.tcpdiag_dport) 704 goto next_dying; 705 if (tcpdiag_dump_sock(skb, sk, cb) < 0) { 706 read_unlock_bh(&head->lock); 707 goto done; 708 } 709 next_dying: 710 ++num; 711 } 712 } 713 read_unlock_bh(&head->lock); 714 } 715 716 done: 717 cb->args[1] = i; 718 cb->args[2] = num; 719 return skb->len; 720 } 721 722 static int tcpdiag_dump_done(struct netlink_callback *cb) 723 { 724 return 0; 725 } 726 727 728 static __inline__ int 729 tcpdiag_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh) 730 { 731 if (!(nlh->nlmsg_flags&NLM_F_REQUEST)) 732 return 0; 733 734 if (nlh->nlmsg_type != TCPDIAG_GETSOCK) 735 goto err_inval; 736 737 if (NLMSG_LENGTH(sizeof(struct tcpdiagreq)) > skb->len) 738 goto err_inval; 739 740 if (nlh->nlmsg_flags&NLM_F_DUMP) { 741 if (nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(struct tcpdiagreq))) { 742 struct rtattr *rta = (struct rtattr*)(NLMSG_DATA(nlh) + sizeof(struct tcpdiagreq)); 743 if (rta->rta_type != TCPDIAG_REQ_BYTECODE || 744 rta->rta_len < 8 || 745 rta->rta_len > nlh->nlmsg_len - NLMSG_SPACE(sizeof(struct tcpdiagreq))) 746 goto err_inval; 747 if (tcpdiag_bc_audit(RTA_DATA(rta), RTA_PAYLOAD(rta))) 748 goto err_inval; 749 } 750 return netlink_dump_start(tcpnl, skb, nlh, 751 tcpdiag_dump, 752 tcpdiag_dump_done); 753 } else { 754 return tcpdiag_get_exact(skb, nlh); 755 } 756 757 err_inval: 758 return -EINVAL; 759 } 760 761 762 static inline void tcpdiag_rcv_skb(struct sk_buff *skb) 763 { 764 int err; 765 struct nlmsghdr * nlh; 766 767 if (skb->len >= NLMSG_SPACE(0)) { 768 nlh = (struct nlmsghdr *)skb->data; 769 if (nlh->nlmsg_len < sizeof(*nlh) || skb->len < nlh->nlmsg_len) 770 return; 771 err = tcpdiag_rcv_msg(skb, nlh); 772 if (err || nlh->nlmsg_flags & NLM_F_ACK) 773 netlink_ack(skb, nlh, err); 774 } 775 } 776 777 static void tcpdiag_rcv(struct sock *sk, int len) 778 { 779 struct sk_buff *skb; 780 781 while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) { 782 tcpdiag_rcv_skb(skb); 783 kfree_skb(skb); 784 } 785 } 786 787 static int __init tcpdiag_init(void) 788 { 789 tcpnl = netlink_kernel_create(NETLINK_TCPDIAG, tcpdiag_rcv); 790 if (tcpnl == NULL) 791 return -ENOMEM; 792 return 0; 793 } 794 795 static void __exit tcpdiag_exit(void) 796 { 797 sock_release(tcpnl->sk_socket); 798 } 799 800 module_init(tcpdiag_init); 801 module_exit(tcpdiag_exit); 802 MODULE_LICENSE("GPL"); 803