f5fc84aaa9b42079a72e4a13d4762bee5953a60b
[linux-2.6.git] / net / ipv4 / tcp_diag.c
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 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
49 extern struct inet_hashinfo dccp_hashinfo;
50 #endif
51
52 static int tcpdiag_fill(struct sk_buff *skb, struct sock *sk,
53                         int ext, u32 pid, u32 seq, u16 nlmsg_flags,
54                         const struct nlmsghdr *unlh)
55 {
56         const struct inet_sock *inet = inet_sk(sk);
57         const struct inet_connection_sock *icsk = inet_csk(sk);
58         struct tcpdiagmsg *r;
59         struct nlmsghdr  *nlh;
60         struct tcp_info  *info = NULL;
61         struct tcpdiag_meminfo  *minfo = NULL;
62         unsigned char    *b = skb->tail;
63
64         nlh = NLMSG_PUT(skb, pid, seq, unlh->nlmsg_type, sizeof(*r));
65         nlh->nlmsg_flags = nlmsg_flags;
66         r = NLMSG_DATA(nlh);
67         if (sk->sk_state != TCP_TIME_WAIT) {
68                 if (ext & (1<<(TCPDIAG_MEMINFO-1)))
69                         minfo = TCPDIAG_PUT(skb, TCPDIAG_MEMINFO, sizeof(*minfo));
70                 if (ext & (1<<(TCPDIAG_INFO-1)))
71                         info = TCPDIAG_PUT(skb, TCPDIAG_INFO, sizeof(*info));
72                 
73                 if ((ext & (1 << (TCPDIAG_CONG - 1))) && icsk->icsk_ca_ops) {
74                         size_t len = strlen(icsk->icsk_ca_ops->name);
75                         strcpy(TCPDIAG_PUT(skb, TCPDIAG_CONG, len+1),
76                                icsk->icsk_ca_ops->name);
77                 }
78         }
79         r->tcpdiag_family = sk->sk_family;
80         r->tcpdiag_state = sk->sk_state;
81         r->tcpdiag_timer = 0;
82         r->tcpdiag_retrans = 0;
83
84         r->id.tcpdiag_if = sk->sk_bound_dev_if;
85         r->id.tcpdiag_cookie[0] = (u32)(unsigned long)sk;
86         r->id.tcpdiag_cookie[1] = (u32)(((unsigned long)sk >> 31) >> 1);
87
88         if (r->tcpdiag_state == TCP_TIME_WAIT) {
89                 const struct inet_timewait_sock *tw = inet_twsk(sk);
90                 long tmo = tw->tw_ttd - jiffies;
91                 if (tmo < 0)
92                         tmo = 0;
93
94                 r->id.tcpdiag_sport = tw->tw_sport;
95                 r->id.tcpdiag_dport = tw->tw_dport;
96                 r->id.tcpdiag_src[0] = tw->tw_rcv_saddr;
97                 r->id.tcpdiag_dst[0] = tw->tw_daddr;
98                 r->tcpdiag_state = tw->tw_substate;
99                 r->tcpdiag_timer = 3;
100                 r->tcpdiag_expires = (tmo*1000+HZ-1)/HZ;
101                 r->tcpdiag_rqueue = 0;
102                 r->tcpdiag_wqueue = 0;
103                 r->tcpdiag_uid = 0;
104                 r->tcpdiag_inode = 0;
105 #ifdef CONFIG_IP_TCPDIAG_IPV6
106                 if (r->tcpdiag_family == AF_INET6) {
107                         const struct tcp6_timewait_sock *tcp6tw = tcp6_twsk(sk);
108
109                         ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_src,
110                                        &tcp6tw->tw_v6_rcv_saddr);
111                         ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_dst,
112                                        &tcp6tw->tw_v6_daddr);
113                 }
114 #endif
115                 nlh->nlmsg_len = skb->tail - b;
116                 return skb->len;
117         }
118
119         r->id.tcpdiag_sport = inet->sport;
120         r->id.tcpdiag_dport = inet->dport;
121         r->id.tcpdiag_src[0] = inet->rcv_saddr;
122         r->id.tcpdiag_dst[0] = inet->daddr;
123
124 #ifdef CONFIG_IP_TCPDIAG_IPV6
125         if (r->tcpdiag_family == AF_INET6) {
126                 struct ipv6_pinfo *np = inet6_sk(sk);
127
128                 ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_src,
129                                &np->rcv_saddr);
130                 ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_dst,
131                                &np->daddr);
132         }
133 #endif
134
135 #define EXPIRES_IN_MS(tmo)  ((tmo-jiffies)*1000+HZ-1)/HZ
136
137         if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
138                 r->tcpdiag_timer = 1;
139                 r->tcpdiag_retrans = icsk->icsk_retransmits;
140                 r->tcpdiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
141         } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
142                 r->tcpdiag_timer = 4;
143                 r->tcpdiag_retrans = icsk->icsk_probes_out;
144                 r->tcpdiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
145         } else if (timer_pending(&sk->sk_timer)) {
146                 r->tcpdiag_timer = 2;
147                 r->tcpdiag_retrans = icsk->icsk_probes_out;
148                 r->tcpdiag_expires = EXPIRES_IN_MS(sk->sk_timer.expires);
149         } else {
150                 r->tcpdiag_timer = 0;
151                 r->tcpdiag_expires = 0;
152         }
153 #undef EXPIRES_IN_MS
154         /*
155          * Ahem... for now we'll have some knowledge about TCP -acme
156          * But this is just one of two small exceptions, both in this
157          * function, so lets close our eyes for some 15 lines or so... 8)
158          * -acme
159          */
160         if (sk->sk_protocol == IPPROTO_TCP) {
161                 const struct tcp_sock *tp = tcp_sk(sk);
162
163                 r->tcpdiag_rqueue = tp->rcv_nxt - tp->copied_seq;
164                 r->tcpdiag_wqueue = tp->write_seq - tp->snd_una;
165         } else
166                 r->tcpdiag_rqueue = r->tcpdiag_wqueue = 0;
167
168         r->tcpdiag_uid = sock_i_uid(sk);
169         r->tcpdiag_inode = sock_i_ino(sk);
170
171         if (minfo) {
172                 minfo->tcpdiag_rmem = atomic_read(&sk->sk_rmem_alloc);
173                 minfo->tcpdiag_wmem = sk->sk_wmem_queued;
174                 minfo->tcpdiag_fmem = sk->sk_forward_alloc;
175                 minfo->tcpdiag_tmem = atomic_read(&sk->sk_wmem_alloc);
176         }
177
178         /* Ahem... for now we'll have some knowledge about TCP -acme */
179         if (info) {
180                 if (sk->sk_protocol == IPPROTO_TCP) 
181                         tcp_get_info(sk, info);
182                 else
183                         memset(info, 0, sizeof(*info));
184         }
185
186         if (sk->sk_state < TCP_TIME_WAIT &&
187             icsk->icsk_ca_ops && icsk->icsk_ca_ops->get_info)
188                 icsk->icsk_ca_ops->get_info(sk, ext, skb);
189
190         nlh->nlmsg_len = skb->tail - b;
191         return skb->len;
192
193 rtattr_failure:
194 nlmsg_failure:
195         skb_trim(skb, b - skb->data);
196         return -1;
197 }
198
199 #ifdef CONFIG_IP_TCPDIAG_IPV6
200 extern struct sock *tcp_v6_lookup(struct in6_addr *saddr, u16 sport,
201                                   struct in6_addr *daddr, u16 dport,
202                                   int dif);
203 #else
204 static inline struct sock *tcp_v6_lookup(struct in6_addr *saddr, u16 sport,
205                                          struct in6_addr *daddr, u16 dport,
206                                          int dif)
207 {
208         return NULL;
209 }
210 #endif
211
212 static int tcpdiag_get_exact(struct sk_buff *in_skb, const struct nlmsghdr *nlh)
213 {
214         int err;
215         struct sock *sk;
216         struct tcpdiagreq *req = NLMSG_DATA(nlh);
217         struct sk_buff *rep;
218         struct inet_hashinfo *hashinfo = &tcp_hashinfo;
219 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
220         if (nlh->nlmsg_type == DCCPDIAG_GETSOCK)
221                 hashinfo = &dccp_hashinfo;
222 #endif
223         if (req->tcpdiag_family == AF_INET) {
224                 sk = inet_lookup(hashinfo, req->id.tcpdiag_dst[0],
225                                  req->id.tcpdiag_dport, req->id.tcpdiag_src[0],
226                                  req->id.tcpdiag_sport, req->id.tcpdiag_if);
227         }
228 #ifdef CONFIG_IP_TCPDIAG_IPV6
229         else if (req->tcpdiag_family == AF_INET6) {
230                 sk = tcp_v6_lookup((struct in6_addr*)req->id.tcpdiag_dst, req->id.tcpdiag_dport,
231                                    (struct in6_addr*)req->id.tcpdiag_src, req->id.tcpdiag_sport,
232                                    req->id.tcpdiag_if);
233         }
234 #endif
235         else {
236                 return -EINVAL;
237         }
238
239         if (sk == NULL)
240                 return -ENOENT;
241
242         err = -ESTALE;
243         if ((req->id.tcpdiag_cookie[0] != TCPDIAG_NOCOOKIE ||
244              req->id.tcpdiag_cookie[1] != TCPDIAG_NOCOOKIE) &&
245             ((u32)(unsigned long)sk != req->id.tcpdiag_cookie[0] ||
246              (u32)((((unsigned long)sk) >> 31) >> 1) != req->id.tcpdiag_cookie[1]))
247                 goto out;
248
249         err = -ENOMEM;
250         rep = alloc_skb(NLMSG_SPACE(sizeof(struct tcpdiagmsg)+
251                                     sizeof(struct tcpdiag_meminfo)+
252                                     sizeof(struct tcp_info)+64), GFP_KERNEL);
253         if (!rep)
254                 goto out;
255
256         if (tcpdiag_fill(rep, sk, req->tcpdiag_ext,
257                          NETLINK_CB(in_skb).pid,
258                          nlh->nlmsg_seq, 0, nlh) <= 0)
259                 BUG();
260
261         err = netlink_unicast(tcpnl, rep, NETLINK_CB(in_skb).pid, MSG_DONTWAIT);
262         if (err > 0)
263                 err = 0;
264
265 out:
266         if (sk) {
267                 if (sk->sk_state == TCP_TIME_WAIT)
268                         inet_twsk_put((struct inet_timewait_sock *)sk);
269                 else
270                         sock_put(sk);
271         }
272         return err;
273 }
274
275 static int bitstring_match(const u32 *a1, const u32 *a2, int bits)
276 {
277         int words = bits >> 5;
278
279         bits &= 0x1f;
280
281         if (words) {
282                 if (memcmp(a1, a2, words << 2))
283                         return 0;
284         }
285         if (bits) {
286                 __u32 w1, w2;
287                 __u32 mask;
288
289                 w1 = a1[words];
290                 w2 = a2[words];
291
292                 mask = htonl((0xffffffff) << (32 - bits));
293
294                 if ((w1 ^ w2) & mask)
295                         return 0;
296         }
297
298         return 1;
299 }
300
301
302 static int tcpdiag_bc_run(const void *bc, int len,
303                           const struct tcpdiag_entry *entry)
304 {
305         while (len > 0) {
306                 int yes = 1;
307                 const struct tcpdiag_bc_op *op = bc;
308
309                 switch (op->code) {
310                 case TCPDIAG_BC_NOP:
311                         break;
312                 case TCPDIAG_BC_JMP:
313                         yes = 0;
314                         break;
315                 case TCPDIAG_BC_S_GE:
316                         yes = entry->sport >= op[1].no;
317                         break;
318                 case TCPDIAG_BC_S_LE:
319                         yes = entry->dport <= op[1].no;
320                         break;
321                 case TCPDIAG_BC_D_GE:
322                         yes = entry->dport >= op[1].no;
323                         break;
324                 case TCPDIAG_BC_D_LE:
325                         yes = entry->dport <= op[1].no;
326                         break;
327                 case TCPDIAG_BC_AUTO:
328                         yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
329                         break;
330                 case TCPDIAG_BC_S_COND:
331                 case TCPDIAG_BC_D_COND:
332                 {
333                         struct tcpdiag_hostcond *cond = (struct tcpdiag_hostcond*)(op+1);
334                         u32 *addr;
335
336                         if (cond->port != -1 &&
337                             cond->port != (op->code == TCPDIAG_BC_S_COND ?
338                                              entry->sport : entry->dport)) {
339                                 yes = 0;
340                                 break;
341                         }
342                         
343                         if (cond->prefix_len == 0)
344                                 break;
345
346                         if (op->code == TCPDIAG_BC_S_COND)
347                                 addr = entry->saddr;
348                         else
349                                 addr = entry->daddr;
350
351                         if (bitstring_match(addr, cond->addr, cond->prefix_len))
352                                 break;
353                         if (entry->family == AF_INET6 &&
354                             cond->family == AF_INET) {
355                                 if (addr[0] == 0 && addr[1] == 0 &&
356                                     addr[2] == htonl(0xffff) &&
357                                     bitstring_match(addr+3, cond->addr, cond->prefix_len))
358                                         break;
359                         }
360                         yes = 0;
361                         break;
362                 }
363                 }
364
365                 if (yes) { 
366                         len -= op->yes;
367                         bc += op->yes;
368                 } else {
369                         len -= op->no;
370                         bc += op->no;
371                 }
372         }
373         return (len == 0);
374 }
375
376 static int valid_cc(const void *bc, int len, int cc)
377 {
378         while (len >= 0) {
379                 const struct tcpdiag_bc_op *op = bc;
380
381                 if (cc > len)
382                         return 0;
383                 if (cc == len)
384                         return 1;
385                 if (op->yes < 4)
386                         return 0;
387                 len -= op->yes;
388                 bc  += op->yes;
389         }
390         return 0;
391 }
392
393 static int tcpdiag_bc_audit(const void *bytecode, int bytecode_len)
394 {
395         const unsigned char *bc = bytecode;
396         int  len = bytecode_len;
397
398         while (len > 0) {
399                 struct tcpdiag_bc_op *op = (struct tcpdiag_bc_op*)bc;
400
401 //printk("BC: %d %d %d {%d} / %d\n", op->code, op->yes, op->no, op[1].no, len);
402                 switch (op->code) {
403                 case TCPDIAG_BC_AUTO:
404                 case TCPDIAG_BC_S_COND:
405                 case TCPDIAG_BC_D_COND:
406                 case TCPDIAG_BC_S_GE:
407                 case TCPDIAG_BC_S_LE:
408                 case TCPDIAG_BC_D_GE:
409                 case TCPDIAG_BC_D_LE:
410                         if (op->yes < 4 || op->yes > len+4)
411                                 return -EINVAL;
412                 case TCPDIAG_BC_JMP:
413                         if (op->no < 4 || op->no > len+4)
414                                 return -EINVAL;
415                         if (op->no < len &&
416                             !valid_cc(bytecode, bytecode_len, len-op->no))
417                                 return -EINVAL;
418                         break;
419                 case TCPDIAG_BC_NOP:
420                         if (op->yes < 4 || op->yes > len+4)
421                                 return -EINVAL;
422                         break;
423                 default:
424                         return -EINVAL;
425                 }
426                 bc += op->yes;
427                 len -= op->yes;
428         }
429         return len == 0 ? 0 : -EINVAL;
430 }
431
432 static int tcpdiag_dump_sock(struct sk_buff *skb, struct sock *sk,
433                              struct netlink_callback *cb)
434 {
435         struct tcpdiagreq *r = NLMSG_DATA(cb->nlh);
436
437         if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) {
438                 struct tcpdiag_entry entry;
439                 struct rtattr *bc = (struct rtattr *)(r + 1);
440                 struct inet_sock *inet = inet_sk(sk);
441
442                 entry.family = sk->sk_family;
443 #ifdef CONFIG_IP_TCPDIAG_IPV6
444                 if (entry.family == AF_INET6) {
445                         struct ipv6_pinfo *np = inet6_sk(sk);
446
447                         entry.saddr = np->rcv_saddr.s6_addr32;
448                         entry.daddr = np->daddr.s6_addr32;
449                 } else
450 #endif
451                 {
452                         entry.saddr = &inet->rcv_saddr;
453                         entry.daddr = &inet->daddr;
454                 }
455                 entry.sport = inet->num;
456                 entry.dport = ntohs(inet->dport);
457                 entry.userlocks = sk->sk_userlocks;
458
459                 if (!tcpdiag_bc_run(RTA_DATA(bc), RTA_PAYLOAD(bc), &entry))
460                         return 0;
461         }
462
463         return tcpdiag_fill(skb, sk, r->tcpdiag_ext, NETLINK_CB(cb->skb).pid,
464                             cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
465 }
466
467 static int tcpdiag_fill_req(struct sk_buff *skb, struct sock *sk,
468                             struct request_sock *req,
469                             u32 pid, u32 seq,
470                             const struct nlmsghdr *unlh)
471 {
472         const struct inet_request_sock *ireq = inet_rsk(req);
473         struct inet_sock *inet = inet_sk(sk);
474         unsigned char *b = skb->tail;
475         struct tcpdiagmsg *r;
476         struct nlmsghdr *nlh;
477         long tmo;
478
479         nlh = NLMSG_PUT(skb, pid, seq, unlh->nlmsg_type, sizeof(*r));
480         nlh->nlmsg_flags = NLM_F_MULTI;
481         r = NLMSG_DATA(nlh);
482
483         r->tcpdiag_family = sk->sk_family;
484         r->tcpdiag_state = TCP_SYN_RECV;
485         r->tcpdiag_timer = 1;
486         r->tcpdiag_retrans = req->retrans;
487
488         r->id.tcpdiag_if = sk->sk_bound_dev_if;
489         r->id.tcpdiag_cookie[0] = (u32)(unsigned long)req;
490         r->id.tcpdiag_cookie[1] = (u32)(((unsigned long)req >> 31) >> 1);
491
492         tmo = req->expires - jiffies;
493         if (tmo < 0)
494                 tmo = 0;
495
496         r->id.tcpdiag_sport = inet->sport;
497         r->id.tcpdiag_dport = ireq->rmt_port;
498         r->id.tcpdiag_src[0] = ireq->loc_addr;
499         r->id.tcpdiag_dst[0] = ireq->rmt_addr;
500         r->tcpdiag_expires = jiffies_to_msecs(tmo),
501         r->tcpdiag_rqueue = 0;
502         r->tcpdiag_wqueue = 0;
503         r->tcpdiag_uid = sock_i_uid(sk);
504         r->tcpdiag_inode = 0;
505 #ifdef CONFIG_IP_TCPDIAG_IPV6
506         if (r->tcpdiag_family == AF_INET6) {
507                 ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_src,
508                                &tcp6_rsk(req)->loc_addr);
509                 ipv6_addr_copy((struct in6_addr *)r->id.tcpdiag_dst,
510                                &tcp6_rsk(req)->rmt_addr);
511         }
512 #endif
513         nlh->nlmsg_len = skb->tail - b;
514
515         return skb->len;
516
517 nlmsg_failure:
518         skb_trim(skb, b - skb->data);
519         return -1;
520 }
521
522 static int tcpdiag_dump_reqs(struct sk_buff *skb, struct sock *sk,
523                              struct netlink_callback *cb)
524 {
525         struct tcpdiag_entry entry;
526         struct tcpdiagreq *r = NLMSG_DATA(cb->nlh);
527         struct inet_connection_sock *icsk = inet_csk(sk);
528         struct listen_sock *lopt;
529         struct rtattr *bc = NULL;
530         struct inet_sock *inet = inet_sk(sk);
531         int j, s_j;
532         int reqnum, s_reqnum;
533         int err = 0;
534
535         s_j = cb->args[3];
536         s_reqnum = cb->args[4];
537
538         if (s_j > 0)
539                 s_j--;
540
541         entry.family = sk->sk_family;
542
543         read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
544
545         lopt = icsk->icsk_accept_queue.listen_opt;
546         if (!lopt || !lopt->qlen)
547                 goto out;
548
549         if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) {
550                 bc = (struct rtattr *)(r + 1);
551                 entry.sport = inet->num;
552                 entry.userlocks = sk->sk_userlocks;
553         }
554
555         for (j = s_j; j < lopt->nr_table_entries; j++) {
556                 struct request_sock *req, *head = lopt->syn_table[j];
557
558                 reqnum = 0;
559                 for (req = head; req; reqnum++, req = req->dl_next) {
560                         struct inet_request_sock *ireq = inet_rsk(req);
561
562                         if (reqnum < s_reqnum)
563                                 continue;
564                         if (r->id.tcpdiag_dport != ireq->rmt_port &&
565                             r->id.tcpdiag_dport)
566                                 continue;
567
568                         if (bc) {
569                                 entry.saddr =
570 #ifdef CONFIG_IP_TCPDIAG_IPV6
571                                         (entry.family == AF_INET6) ?
572                                         tcp6_rsk(req)->loc_addr.s6_addr32 :
573 #endif
574                                         &ireq->loc_addr;
575                                 entry.daddr = 
576 #ifdef CONFIG_IP_TCPDIAG_IPV6
577                                         (entry.family == AF_INET6) ?
578                                         tcp6_rsk(req)->rmt_addr.s6_addr32 :
579 #endif
580                                         &ireq->rmt_addr;
581                                 entry.dport = ntohs(ireq->rmt_port);
582
583                                 if (!tcpdiag_bc_run(RTA_DATA(bc),
584                                                     RTA_PAYLOAD(bc), &entry))
585                                         continue;
586                         }
587
588                         err = tcpdiag_fill_req(skb, sk, req,
589                                                NETLINK_CB(cb->skb).pid,
590                                                cb->nlh->nlmsg_seq, cb->nlh);
591                         if (err < 0) {
592                                 cb->args[3] = j + 1;
593                                 cb->args[4] = reqnum;
594                                 goto out;
595                         }
596                 }
597
598                 s_reqnum = 0;
599         }
600
601 out:
602         read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
603
604         return err;
605 }
606
607 static int tcpdiag_dump(struct sk_buff *skb, struct netlink_callback *cb)
608 {
609         int i, num;
610         int s_i, s_num;
611         struct tcpdiagreq *r = NLMSG_DATA(cb->nlh);
612         struct inet_hashinfo *hashinfo;
613
614         s_i = cb->args[1];
615         s_num = num = cb->args[2];
616                 hashinfo = &tcp_hashinfo;
617 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
618         if (cb->nlh->nlmsg_type == DCCPDIAG_GETSOCK)
619                 hashinfo = &dccp_hashinfo;
620 #endif
621         if (cb->args[0] == 0) {
622                 if (!(r->tcpdiag_states&(TCPF_LISTEN|TCPF_SYN_RECV)))
623                         goto skip_listen_ht;
624
625                 inet_listen_lock(hashinfo);
626                 for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
627                         struct sock *sk;
628                         struct hlist_node *node;
629
630                         num = 0;
631                         sk_for_each(sk, node, &hashinfo->listening_hash[i]) {
632                                 struct inet_sock *inet = inet_sk(sk);
633
634                                 if (num < s_num) {
635                                         num++;
636                                         continue;
637                                 }
638
639                                 if (r->id.tcpdiag_sport != inet->sport &&
640                                     r->id.tcpdiag_sport)
641                                         goto next_listen;
642
643                                 if (!(r->tcpdiag_states&TCPF_LISTEN) ||
644                                     r->id.tcpdiag_dport ||
645                                     cb->args[3] > 0)
646                                         goto syn_recv;
647
648                                 if (tcpdiag_dump_sock(skb, sk, cb) < 0) {
649                                         inet_listen_unlock(hashinfo);
650                                         goto done;
651                                 }
652
653 syn_recv:
654                                 if (!(r->tcpdiag_states&TCPF_SYN_RECV))
655                                         goto next_listen;
656
657                                 if (tcpdiag_dump_reqs(skb, sk, cb) < 0) {
658                                         inet_listen_unlock(hashinfo);
659                                         goto done;
660                                 }
661
662 next_listen:
663                                 cb->args[3] = 0;
664                                 cb->args[4] = 0;
665                                 ++num;
666                         }
667
668                         s_num = 0;
669                         cb->args[3] = 0;
670                         cb->args[4] = 0;
671                 }
672                 inet_listen_unlock(hashinfo);
673 skip_listen_ht:
674                 cb->args[0] = 1;
675                 s_i = num = s_num = 0;
676         }
677
678         if (!(r->tcpdiag_states&~(TCPF_LISTEN|TCPF_SYN_RECV)))
679                 return skb->len;
680
681         for (i = s_i; i < hashinfo->ehash_size; i++) {
682                 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
683                 struct sock *sk;
684                 struct hlist_node *node;
685
686                 if (i > s_i)
687                         s_num = 0;
688
689                 read_lock_bh(&head->lock);
690
691                 num = 0;
692                 sk_for_each(sk, node, &head->chain) {
693                         struct inet_sock *inet = inet_sk(sk);
694
695                         if (num < s_num)
696                                 goto next_normal;
697                         if (!(r->tcpdiag_states & (1 << sk->sk_state)))
698                                 goto next_normal;
699                         if (r->id.tcpdiag_sport != inet->sport &&
700                             r->id.tcpdiag_sport)
701                                 goto next_normal;
702                         if (r->id.tcpdiag_dport != inet->dport && r->id.tcpdiag_dport)
703                                 goto next_normal;
704                         if (tcpdiag_dump_sock(skb, sk, cb) < 0) {
705                                 read_unlock_bh(&head->lock);
706                                 goto done;
707                         }
708 next_normal:
709                         ++num;
710                 }
711
712                 if (r->tcpdiag_states&TCPF_TIME_WAIT) {
713                         sk_for_each(sk, node,
714                                     &hashinfo->ehash[i + hashinfo->ehash_size].chain) {
715                                 struct inet_sock *inet = inet_sk(sk);
716
717                                 if (num < s_num)
718                                         goto next_dying;
719                                 if (r->id.tcpdiag_sport != inet->sport &&
720                                     r->id.tcpdiag_sport)
721                                         goto next_dying;
722                                 if (r->id.tcpdiag_dport != inet->dport &&
723                                     r->id.tcpdiag_dport)
724                                         goto next_dying;
725                                 if (tcpdiag_dump_sock(skb, sk, cb) < 0) {
726                                         read_unlock_bh(&head->lock);
727                                         goto done;
728                                 }
729 next_dying:
730                                 ++num;
731                         }
732                 }
733                 read_unlock_bh(&head->lock);
734         }
735
736 done:
737         cb->args[1] = i;
738         cb->args[2] = num;
739         return skb->len;
740 }
741
742 static int tcpdiag_dump_done(struct netlink_callback *cb)
743 {
744         return 0;
745 }
746
747
748 static __inline__ int
749 tcpdiag_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
750 {
751         if (!(nlh->nlmsg_flags&NLM_F_REQUEST))
752                 return 0;
753
754         if (nlh->nlmsg_type != TCPDIAG_GETSOCK
755 #if defined(CONFIG_IP_DCCP) || defined(CONFIG_IP_DCCP_MODULE)
756             && nlh->nlmsg_type != DCCPDIAG_GETSOCK
757 #endif
758            )
759                 goto err_inval;
760
761         if (NLMSG_LENGTH(sizeof(struct tcpdiagreq)) > skb->len)
762                 goto err_inval;
763
764         if (nlh->nlmsg_flags&NLM_F_DUMP) {
765                 if (nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(struct tcpdiagreq))) {
766                         struct rtattr *rta = (struct rtattr*)(NLMSG_DATA(nlh) + sizeof(struct tcpdiagreq));
767                         if (rta->rta_type != TCPDIAG_REQ_BYTECODE ||
768                             rta->rta_len < 8 ||
769                             rta->rta_len > nlh->nlmsg_len - NLMSG_SPACE(sizeof(struct tcpdiagreq)))
770                                 goto err_inval;
771                         if (tcpdiag_bc_audit(RTA_DATA(rta), RTA_PAYLOAD(rta)))
772                                 goto err_inval;
773                 }
774                 return netlink_dump_start(tcpnl, skb, nlh,
775                                           tcpdiag_dump,
776                                           tcpdiag_dump_done);
777         } else {
778                 return tcpdiag_get_exact(skb, nlh);
779         }
780
781 err_inval:
782         return -EINVAL;
783 }
784
785
786 static inline void tcpdiag_rcv_skb(struct sk_buff *skb)
787 {
788         int err;
789         struct nlmsghdr * nlh;
790
791         if (skb->len >= NLMSG_SPACE(0)) {
792                 nlh = (struct nlmsghdr *)skb->data;
793                 if (nlh->nlmsg_len < sizeof(*nlh) || skb->len < nlh->nlmsg_len)
794                         return;
795                 err = tcpdiag_rcv_msg(skb, nlh);
796                 if (err || nlh->nlmsg_flags & NLM_F_ACK) 
797                         netlink_ack(skb, nlh, err);
798         }
799 }
800
801 static void tcpdiag_rcv(struct sock *sk, int len)
802 {
803         struct sk_buff *skb;
804         unsigned int qlen = skb_queue_len(&sk->sk_receive_queue);
805
806         while (qlen-- && (skb = skb_dequeue(&sk->sk_receive_queue))) {
807                 tcpdiag_rcv_skb(skb);
808                 kfree_skb(skb);
809         }
810 }
811
812 static int __init tcpdiag_init(void)
813 {
814         tcpnl = netlink_kernel_create(NETLINK_TCPDIAG, tcpdiag_rcv,
815                                       THIS_MODULE);
816         if (tcpnl == NULL)
817                 return -ENOMEM;
818         return 0;
819 }
820
821 static void __exit tcpdiag_exit(void)
822 {
823         sock_release(tcpnl->sk_socket);
824 }
825
826 module_init(tcpdiag_init);
827 module_exit(tcpdiag_exit);
828 MODULE_LICENSE("GPL");