Merge git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6
[linux-2.6] / net / ipv4 / inet_diag.c
1 /*
2  * inet_diag.c  Module for monitoring INET transport protocols sockets.
3  *
4  * Version:     $Id: inet_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/module.h>
15 #include <linux/types.h>
16 #include <linux/fcntl.h>
17 #include <linux/random.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
31 #include <linux/inet.h>
32 #include <linux/stddef.h>
33
34 #include <linux/inet_diag.h>
35
36 static const struct inet_diag_handler **inet_diag_table;
37
38 struct inet_diag_entry {
39         __be32 *saddr;
40         __be32 *daddr;
41         u16 sport;
42         u16 dport;
43         u16 family;
44         u16 userlocks;
45 };
46
47 static struct sock *idiagnl;
48
49 #define INET_DIAG_PUT(skb, attrtype, attrlen) \
50         RTA_DATA(__RTA_PUT(skb, attrtype, attrlen))
51
52 static int inet_csk_diag_fill(struct sock *sk,
53                               struct sk_buff *skb,
54                               int ext, u32 pid, u32 seq, u16 nlmsg_flags,
55                               const struct nlmsghdr *unlh)
56 {
57         const struct inet_sock *inet = inet_sk(sk);
58         const struct inet_connection_sock *icsk = inet_csk(sk);
59         struct inet_diag_msg *r;
60         struct nlmsghdr  *nlh;
61         void *info = NULL;
62         struct inet_diag_meminfo  *minfo = NULL;
63         unsigned char    *b = skb->tail;
64         const struct inet_diag_handler *handler;
65
66         handler = inet_diag_table[unlh->nlmsg_type];
67         BUG_ON(handler == NULL);
68
69         nlh = NLMSG_PUT(skb, pid, seq, unlh->nlmsg_type, sizeof(*r));
70         nlh->nlmsg_flags = nlmsg_flags;
71
72         r = NLMSG_DATA(nlh);
73         BUG_ON(sk->sk_state == TCP_TIME_WAIT);
74
75         if (ext & (1 << (INET_DIAG_MEMINFO - 1)))
76                 minfo = INET_DIAG_PUT(skb, INET_DIAG_MEMINFO, sizeof(*minfo));
77
78         if (ext & (1 << (INET_DIAG_INFO - 1)))
79                 info = INET_DIAG_PUT(skb, INET_DIAG_INFO,
80                                      handler->idiag_info_size);
81
82         if ((ext & (1 << (INET_DIAG_CONG - 1))) && icsk->icsk_ca_ops) {
83                 const size_t len = strlen(icsk->icsk_ca_ops->name);
84
85                 strcpy(INET_DIAG_PUT(skb, INET_DIAG_CONG, len + 1),
86                        icsk->icsk_ca_ops->name);
87         }
88
89         r->idiag_family = sk->sk_family;
90         r->idiag_state = sk->sk_state;
91         r->idiag_timer = 0;
92         r->idiag_retrans = 0;
93
94         r->id.idiag_if = sk->sk_bound_dev_if;
95         r->id.idiag_cookie[0] = (u32)(unsigned long)sk;
96         r->id.idiag_cookie[1] = (u32)(((unsigned long)sk >> 31) >> 1);
97
98         r->id.idiag_sport = inet->sport;
99         r->id.idiag_dport = inet->dport;
100         r->id.idiag_src[0] = inet->rcv_saddr;
101         r->id.idiag_dst[0] = inet->daddr;
102
103 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
104         if (r->idiag_family == AF_INET6) {
105                 struct ipv6_pinfo *np = inet6_sk(sk);
106
107                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
108                                &np->rcv_saddr);
109                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
110                                &np->daddr);
111         }
112 #endif
113
114 #define EXPIRES_IN_MS(tmo)  ((tmo - jiffies) * 1000 + HZ - 1) / HZ
115
116         if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
117                 r->idiag_timer = 1;
118                 r->idiag_retrans = icsk->icsk_retransmits;
119                 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
120         } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
121                 r->idiag_timer = 4;
122                 r->idiag_retrans = icsk->icsk_probes_out;
123                 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
124         } else if (timer_pending(&sk->sk_timer)) {
125                 r->idiag_timer = 2;
126                 r->idiag_retrans = icsk->icsk_probes_out;
127                 r->idiag_expires = EXPIRES_IN_MS(sk->sk_timer.expires);
128         } else {
129                 r->idiag_timer = 0;
130                 r->idiag_expires = 0;
131         }
132 #undef EXPIRES_IN_MS
133
134         r->idiag_uid = sock_i_uid(sk);
135         r->idiag_inode = sock_i_ino(sk);
136
137         if (minfo) {
138                 minfo->idiag_rmem = atomic_read(&sk->sk_rmem_alloc);
139                 minfo->idiag_wmem = sk->sk_wmem_queued;
140                 minfo->idiag_fmem = sk->sk_forward_alloc;
141                 minfo->idiag_tmem = atomic_read(&sk->sk_wmem_alloc);
142         }
143
144         handler->idiag_get_info(sk, r, info);
145
146         if (sk->sk_state < TCP_TIME_WAIT &&
147             icsk->icsk_ca_ops && icsk->icsk_ca_ops->get_info)
148                 icsk->icsk_ca_ops->get_info(sk, ext, skb);
149
150         nlh->nlmsg_len = skb->tail - b;
151         return skb->len;
152
153 rtattr_failure:
154 nlmsg_failure:
155         skb_trim(skb, b - skb->data);
156         return -1;
157 }
158
159 static int inet_twsk_diag_fill(struct inet_timewait_sock *tw,
160                                struct sk_buff *skb, int ext, u32 pid,
161                                u32 seq, u16 nlmsg_flags,
162                                const struct nlmsghdr *unlh)
163 {
164         long tmo;
165         struct inet_diag_msg *r;
166         const unsigned char *previous_tail = skb->tail;
167         struct nlmsghdr *nlh = NLMSG_PUT(skb, pid, seq,
168                                          unlh->nlmsg_type, sizeof(*r));
169
170         r = NLMSG_DATA(nlh);
171         BUG_ON(tw->tw_state != TCP_TIME_WAIT);
172
173         nlh->nlmsg_flags = nlmsg_flags;
174
175         tmo = tw->tw_ttd - jiffies;
176         if (tmo < 0)
177                 tmo = 0;
178
179         r->idiag_family       = tw->tw_family;
180         r->idiag_state        = tw->tw_state;
181         r->idiag_timer        = 0;
182         r->idiag_retrans      = 0;
183         r->id.idiag_if        = tw->tw_bound_dev_if;
184         r->id.idiag_cookie[0] = (u32)(unsigned long)tw;
185         r->id.idiag_cookie[1] = (u32)(((unsigned long)tw >> 31) >> 1);
186         r->id.idiag_sport     = tw->tw_sport;
187         r->id.idiag_dport     = tw->tw_dport;
188         r->id.idiag_src[0]    = tw->tw_rcv_saddr;
189         r->id.idiag_dst[0]    = tw->tw_daddr;
190         r->idiag_state        = tw->tw_substate;
191         r->idiag_timer        = 3;
192         r->idiag_expires      = (tmo * 1000 + HZ - 1) / HZ;
193         r->idiag_rqueue       = 0;
194         r->idiag_wqueue       = 0;
195         r->idiag_uid          = 0;
196         r->idiag_inode        = 0;
197 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
198         if (tw->tw_family == AF_INET6) {
199                 const struct inet6_timewait_sock *tw6 =
200                                                 inet6_twsk((struct sock *)tw);
201
202                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
203                                &tw6->tw_v6_rcv_saddr);
204                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
205                                &tw6->tw_v6_daddr);
206         }
207 #endif
208         nlh->nlmsg_len = skb->tail - previous_tail;
209         return skb->len;
210 nlmsg_failure:
211         skb_trim(skb, previous_tail - skb->data);
212         return -1;
213 }
214
215 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
216                         int ext, u32 pid, u32 seq, u16 nlmsg_flags,
217                         const struct nlmsghdr *unlh)
218 {
219         if (sk->sk_state == TCP_TIME_WAIT)
220                 return inet_twsk_diag_fill((struct inet_timewait_sock *)sk,
221                                            skb, ext, pid, seq, nlmsg_flags,
222                                            unlh);
223         return inet_csk_diag_fill(sk, skb, ext, pid, seq, nlmsg_flags, unlh);
224 }
225
226 static int inet_diag_get_exact(struct sk_buff *in_skb,
227                                const struct nlmsghdr *nlh)
228 {
229         int err;
230         struct sock *sk;
231         struct inet_diag_req *req = NLMSG_DATA(nlh);
232         struct sk_buff *rep;
233         struct inet_hashinfo *hashinfo;
234         const struct inet_diag_handler *handler;
235
236         handler = inet_diag_table[nlh->nlmsg_type];
237         BUG_ON(handler == NULL);
238         hashinfo = handler->idiag_hashinfo;
239
240         if (req->idiag_family == AF_INET) {
241                 sk = inet_lookup(hashinfo, req->id.idiag_dst[0],
242                                  req->id.idiag_dport, req->id.idiag_src[0],
243                                  req->id.idiag_sport, req->id.idiag_if);
244         }
245 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
246         else if (req->idiag_family == AF_INET6) {
247                 sk = inet6_lookup(hashinfo,
248                                   (struct in6_addr *)req->id.idiag_dst,
249                                   req->id.idiag_dport,
250                                   (struct in6_addr *)req->id.idiag_src,
251                                   req->id.idiag_sport,
252                                   req->id.idiag_if);
253         }
254 #endif
255         else {
256                 return -EINVAL;
257         }
258
259         if (sk == NULL)
260                 return -ENOENT;
261
262         err = -ESTALE;
263         if ((req->id.idiag_cookie[0] != INET_DIAG_NOCOOKIE ||
264              req->id.idiag_cookie[1] != INET_DIAG_NOCOOKIE) &&
265             ((u32)(unsigned long)sk != req->id.idiag_cookie[0] ||
266              (u32)((((unsigned long)sk) >> 31) >> 1) != req->id.idiag_cookie[1]))
267                 goto out;
268
269         err = -ENOMEM;
270         rep = alloc_skb(NLMSG_SPACE((sizeof(struct inet_diag_msg) +
271                                      sizeof(struct inet_diag_meminfo) +
272                                      handler->idiag_info_size + 64)),
273                         GFP_KERNEL);
274         if (!rep)
275                 goto out;
276
277         if (sk_diag_fill(sk, rep, req->idiag_ext,
278                          NETLINK_CB(in_skb).pid,
279                          nlh->nlmsg_seq, 0, nlh) <= 0)
280                 BUG();
281
282         err = netlink_unicast(idiagnl, rep, NETLINK_CB(in_skb).pid,
283                               MSG_DONTWAIT);
284         if (err > 0)
285                 err = 0;
286
287 out:
288         if (sk) {
289                 if (sk->sk_state == TCP_TIME_WAIT)
290                         inet_twsk_put((struct inet_timewait_sock *)sk);
291                 else
292                         sock_put(sk);
293         }
294         return err;
295 }
296
297 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
298 {
299         int words = bits >> 5;
300
301         bits &= 0x1f;
302
303         if (words) {
304                 if (memcmp(a1, a2, words << 2))
305                         return 0;
306         }
307         if (bits) {
308                 __be32 w1, w2;
309                 __be32 mask;
310
311                 w1 = a1[words];
312                 w2 = a2[words];
313
314                 mask = htonl((0xffffffff) << (32 - bits));
315
316                 if ((w1 ^ w2) & mask)
317                         return 0;
318         }
319
320         return 1;
321 }
322
323
324 static int inet_diag_bc_run(const void *bc, int len,
325                             const struct inet_diag_entry *entry)
326 {
327         while (len > 0) {
328                 int yes = 1;
329                 const struct inet_diag_bc_op *op = bc;
330
331                 switch (op->code) {
332                 case INET_DIAG_BC_NOP:
333                         break;
334                 case INET_DIAG_BC_JMP:
335                         yes = 0;
336                         break;
337                 case INET_DIAG_BC_S_GE:
338                         yes = entry->sport >= op[1].no;
339                         break;
340                 case INET_DIAG_BC_S_LE:
341                         yes = entry->dport <= op[1].no;
342                         break;
343                 case INET_DIAG_BC_D_GE:
344                         yes = entry->dport >= op[1].no;
345                         break;
346                 case INET_DIAG_BC_D_LE:
347                         yes = entry->dport <= op[1].no;
348                         break;
349                 case INET_DIAG_BC_AUTO:
350                         yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
351                         break;
352                 case INET_DIAG_BC_S_COND:
353                 case INET_DIAG_BC_D_COND: {
354                         struct inet_diag_hostcond *cond;
355                         __be32 *addr;
356
357                         cond = (struct inet_diag_hostcond *)(op + 1);
358                         if (cond->port != -1 &&
359                             cond->port != (op->code == INET_DIAG_BC_S_COND ?
360                                              entry->sport : entry->dport)) {
361                                 yes = 0;
362                                 break;
363                         }
364
365                         if (cond->prefix_len == 0)
366                                 break;
367
368                         if (op->code == INET_DIAG_BC_S_COND)
369                                 addr = entry->saddr;
370                         else
371                                 addr = entry->daddr;
372
373                         if (bitstring_match(addr, cond->addr,
374                                             cond->prefix_len))
375                                 break;
376                         if (entry->family == AF_INET6 &&
377                             cond->family == AF_INET) {
378                                 if (addr[0] == 0 && addr[1] == 0 &&
379                                     addr[2] == htonl(0xffff) &&
380                                     bitstring_match(addr + 3, cond->addr,
381                                                     cond->prefix_len))
382                                         break;
383                         }
384                         yes = 0;
385                         break;
386                 }
387                 }
388
389                 if (yes) {
390                         len -= op->yes;
391                         bc += op->yes;
392                 } else {
393                         len -= op->no;
394                         bc += op->no;
395                 }
396         }
397         return (len == 0);
398 }
399
400 static int valid_cc(const void *bc, int len, int cc)
401 {
402         while (len >= 0) {
403                 const struct inet_diag_bc_op *op = bc;
404
405                 if (cc > len)
406                         return 0;
407                 if (cc == len)
408                         return 1;
409                 if (op->yes < 4)
410                         return 0;
411                 len -= op->yes;
412                 bc  += op->yes;
413         }
414         return 0;
415 }
416
417 static int inet_diag_bc_audit(const void *bytecode, int bytecode_len)
418 {
419         const unsigned char *bc = bytecode;
420         int  len = bytecode_len;
421
422         while (len > 0) {
423                 struct inet_diag_bc_op *op = (struct inet_diag_bc_op *)bc;
424
425 //printk("BC: %d %d %d {%d} / %d\n", op->code, op->yes, op->no, op[1].no, len);
426                 switch (op->code) {
427                 case INET_DIAG_BC_AUTO:
428                 case INET_DIAG_BC_S_COND:
429                 case INET_DIAG_BC_D_COND:
430                 case INET_DIAG_BC_S_GE:
431                 case INET_DIAG_BC_S_LE:
432                 case INET_DIAG_BC_D_GE:
433                 case INET_DIAG_BC_D_LE:
434                         if (op->yes < 4 || op->yes > len + 4)
435                                 return -EINVAL;
436                 case INET_DIAG_BC_JMP:
437                         if (op->no < 4 || op->no > len + 4)
438                                 return -EINVAL;
439                         if (op->no < len &&
440                             !valid_cc(bytecode, bytecode_len, len - op->no))
441                                 return -EINVAL;
442                         break;
443                 case INET_DIAG_BC_NOP:
444                         if (op->yes < 4 || op->yes > len + 4)
445                                 return -EINVAL;
446                         break;
447                 default:
448                         return -EINVAL;
449                 }
450                 bc  += op->yes;
451                 len -= op->yes;
452         }
453         return len == 0 ? 0 : -EINVAL;
454 }
455
456 static int inet_csk_diag_dump(struct sock *sk,
457                               struct sk_buff *skb,
458                               struct netlink_callback *cb)
459 {
460         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
461
462         if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) {
463                 struct inet_diag_entry entry;
464                 struct rtattr *bc = (struct rtattr *)(r + 1);
465                 struct inet_sock *inet = inet_sk(sk);
466
467                 entry.family = sk->sk_family;
468 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
469                 if (entry.family == AF_INET6) {
470                         struct ipv6_pinfo *np = inet6_sk(sk);
471
472                         entry.saddr = np->rcv_saddr.s6_addr32;
473                         entry.daddr = np->daddr.s6_addr32;
474                 } else
475 #endif
476                 {
477                         entry.saddr = &inet->rcv_saddr;
478                         entry.daddr = &inet->daddr;
479                 }
480                 entry.sport = inet->num;
481                 entry.dport = ntohs(inet->dport);
482                 entry.userlocks = sk->sk_userlocks;
483
484                 if (!inet_diag_bc_run(RTA_DATA(bc), RTA_PAYLOAD(bc), &entry))
485                         return 0;
486         }
487
488         return inet_csk_diag_fill(sk, skb, r->idiag_ext,
489                                   NETLINK_CB(cb->skb).pid,
490                                   cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
491 }
492
493 static int inet_twsk_diag_dump(struct inet_timewait_sock *tw,
494                                struct sk_buff *skb,
495                                struct netlink_callback *cb)
496 {
497         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
498
499         if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) {
500                 struct inet_diag_entry entry;
501                 struct rtattr *bc = (struct rtattr *)(r + 1);
502
503                 entry.family = tw->tw_family;
504 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
505                 if (tw->tw_family == AF_INET6) {
506                         struct inet6_timewait_sock *tw6 =
507                                                 inet6_twsk((struct sock *)tw);
508                         entry.saddr = tw6->tw_v6_rcv_saddr.s6_addr32;
509                         entry.daddr = tw6->tw_v6_daddr.s6_addr32;
510                 } else
511 #endif
512                 {
513                         entry.saddr = &tw->tw_rcv_saddr;
514                         entry.daddr = &tw->tw_daddr;
515                 }
516                 entry.sport = tw->tw_num;
517                 entry.dport = ntohs(tw->tw_dport);
518                 entry.userlocks = 0; 
519
520                 if (!inet_diag_bc_run(RTA_DATA(bc), RTA_PAYLOAD(bc), &entry))
521                         return 0;
522         }
523
524         return inet_twsk_diag_fill(tw, skb, r->idiag_ext,
525                                    NETLINK_CB(cb->skb).pid,
526                                    cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
527 }
528
529 static int inet_diag_fill_req(struct sk_buff *skb, struct sock *sk,
530                               struct request_sock *req, u32 pid, u32 seq,
531                               const struct nlmsghdr *unlh)
532 {
533         const struct inet_request_sock *ireq = inet_rsk(req);
534         struct inet_sock *inet = inet_sk(sk);
535         unsigned char *b = skb->tail;
536         struct inet_diag_msg *r;
537         struct nlmsghdr *nlh;
538         long tmo;
539
540         nlh = NLMSG_PUT(skb, pid, seq, unlh->nlmsg_type, sizeof(*r));
541         nlh->nlmsg_flags = NLM_F_MULTI;
542         r = NLMSG_DATA(nlh);
543
544         r->idiag_family = sk->sk_family;
545         r->idiag_state = TCP_SYN_RECV;
546         r->idiag_timer = 1;
547         r->idiag_retrans = req->retrans;
548
549         r->id.idiag_if = sk->sk_bound_dev_if;
550         r->id.idiag_cookie[0] = (u32)(unsigned long)req;
551         r->id.idiag_cookie[1] = (u32)(((unsigned long)req >> 31) >> 1);
552
553         tmo = req->expires - jiffies;
554         if (tmo < 0)
555                 tmo = 0;
556
557         r->id.idiag_sport = inet->sport;
558         r->id.idiag_dport = ireq->rmt_port;
559         r->id.idiag_src[0] = ireq->loc_addr;
560         r->id.idiag_dst[0] = ireq->rmt_addr;
561         r->idiag_expires = jiffies_to_msecs(tmo);
562         r->idiag_rqueue = 0;
563         r->idiag_wqueue = 0;
564         r->idiag_uid = sock_i_uid(sk);
565         r->idiag_inode = 0;
566 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
567         if (r->idiag_family == AF_INET6) {
568                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
569                                &inet6_rsk(req)->loc_addr);
570                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
571                                &inet6_rsk(req)->rmt_addr);
572         }
573 #endif
574         nlh->nlmsg_len = skb->tail - b;
575
576         return skb->len;
577
578 nlmsg_failure:
579         skb_trim(skb, b - skb->data);
580         return -1;
581 }
582
583 static int inet_diag_dump_reqs(struct sk_buff *skb, struct sock *sk,
584                                struct netlink_callback *cb)
585 {
586         struct inet_diag_entry entry;
587         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
588         struct inet_connection_sock *icsk = inet_csk(sk);
589         struct listen_sock *lopt;
590         struct rtattr *bc = NULL;
591         struct inet_sock *inet = inet_sk(sk);
592         int j, s_j;
593         int reqnum, s_reqnum;
594         int err = 0;
595
596         s_j = cb->args[3];
597         s_reqnum = cb->args[4];
598
599         if (s_j > 0)
600                 s_j--;
601
602         entry.family = sk->sk_family;
603
604         read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
605
606         lopt = icsk->icsk_accept_queue.listen_opt;
607         if (!lopt || !lopt->qlen)
608                 goto out;
609
610         if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) {
611                 bc = (struct rtattr *)(r + 1);
612                 entry.sport = inet->num;
613                 entry.userlocks = sk->sk_userlocks;
614         }
615
616         for (j = s_j; j < lopt->nr_table_entries; j++) {
617                 struct request_sock *req, *head = lopt->syn_table[j];
618
619                 reqnum = 0;
620                 for (req = head; req; reqnum++, req = req->dl_next) {
621                         struct inet_request_sock *ireq = inet_rsk(req);
622
623                         if (reqnum < s_reqnum)
624                                 continue;
625                         if (r->id.idiag_dport != ireq->rmt_port &&
626                             r->id.idiag_dport)
627                                 continue;
628
629                         if (bc) {
630                                 entry.saddr =
631 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
632                                         (entry.family == AF_INET6) ?
633                                         inet6_rsk(req)->loc_addr.s6_addr32 :
634 #endif
635                                         &ireq->loc_addr;
636                                 entry.daddr =
637 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
638                                         (entry.family == AF_INET6) ?
639                                         inet6_rsk(req)->rmt_addr.s6_addr32 :
640 #endif
641                                         &ireq->rmt_addr;
642                                 entry.dport = ntohs(ireq->rmt_port);
643
644                                 if (!inet_diag_bc_run(RTA_DATA(bc),
645                                                     RTA_PAYLOAD(bc), &entry))
646                                         continue;
647                         }
648
649                         err = inet_diag_fill_req(skb, sk, req,
650                                                NETLINK_CB(cb->skb).pid,
651                                                cb->nlh->nlmsg_seq, cb->nlh);
652                         if (err < 0) {
653                                 cb->args[3] = j + 1;
654                                 cb->args[4] = reqnum;
655                                 goto out;
656                         }
657                 }
658
659                 s_reqnum = 0;
660         }
661
662 out:
663         read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
664
665         return err;
666 }
667
668 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
669 {
670         int i, num;
671         int s_i, s_num;
672         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
673         const struct inet_diag_handler *handler;
674         struct inet_hashinfo *hashinfo;
675
676         handler = inet_diag_table[cb->nlh->nlmsg_type];
677         BUG_ON(handler == NULL);
678         hashinfo = handler->idiag_hashinfo;
679
680         s_i = cb->args[1];
681         s_num = num = cb->args[2];
682
683         if (cb->args[0] == 0) {
684                 if (!(r->idiag_states & (TCPF_LISTEN | TCPF_SYN_RECV)))
685                         goto skip_listen_ht;
686
687                 inet_listen_lock(hashinfo);
688                 for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
689                         struct sock *sk;
690                         struct hlist_node *node;
691
692                         num = 0;
693                         sk_for_each(sk, node, &hashinfo->listening_hash[i]) {
694                                 struct inet_sock *inet = inet_sk(sk);
695
696                                 if (num < s_num) {
697                                         num++;
698                                         continue;
699                                 }
700
701                                 if (r->id.idiag_sport != inet->sport &&
702                                     r->id.idiag_sport)
703                                         goto next_listen;
704
705                                 if (!(r->idiag_states & TCPF_LISTEN) ||
706                                     r->id.idiag_dport ||
707                                     cb->args[3] > 0)
708                                         goto syn_recv;
709
710                                 if (inet_csk_diag_dump(sk, skb, cb) < 0) {
711                                         inet_listen_unlock(hashinfo);
712                                         goto done;
713                                 }
714
715 syn_recv:
716                                 if (!(r->idiag_states & TCPF_SYN_RECV))
717                                         goto next_listen;
718
719                                 if (inet_diag_dump_reqs(skb, sk, cb) < 0) {
720                                         inet_listen_unlock(hashinfo);
721                                         goto done;
722                                 }
723
724 next_listen:
725                                 cb->args[3] = 0;
726                                 cb->args[4] = 0;
727                                 ++num;
728                         }
729
730                         s_num = 0;
731                         cb->args[3] = 0;
732                         cb->args[4] = 0;
733                 }
734                 inet_listen_unlock(hashinfo);
735 skip_listen_ht:
736                 cb->args[0] = 1;
737                 s_i = num = s_num = 0;
738         }
739
740         if (!(r->idiag_states & ~(TCPF_LISTEN | TCPF_SYN_RECV)))
741                 return skb->len;
742
743         for (i = s_i; i < hashinfo->ehash_size; i++) {
744                 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
745                 struct sock *sk;
746                 struct hlist_node *node;
747
748                 if (i > s_i)
749                         s_num = 0;
750
751                 read_lock_bh(&head->lock);
752                 num = 0;
753                 sk_for_each(sk, node, &head->chain) {
754                         struct inet_sock *inet = inet_sk(sk);
755
756                         if (num < s_num)
757                                 goto next_normal;
758                         if (!(r->idiag_states & (1 << sk->sk_state)))
759                                 goto next_normal;
760                         if (r->id.idiag_sport != inet->sport &&
761                             r->id.idiag_sport)
762                                 goto next_normal;
763                         if (r->id.idiag_dport != inet->dport &&
764                             r->id.idiag_dport)
765                                 goto next_normal;
766                         if (inet_csk_diag_dump(sk, skb, cb) < 0) {
767                                 read_unlock_bh(&head->lock);
768                                 goto done;
769                         }
770 next_normal:
771                         ++num;
772                 }
773
774                 if (r->idiag_states & TCPF_TIME_WAIT) {
775                         struct inet_timewait_sock *tw;
776
777                         inet_twsk_for_each(tw, node,
778                                     &hashinfo->ehash[i + hashinfo->ehash_size].chain) {
779
780                                 if (num < s_num)
781                                         goto next_dying;
782                                 if (r->id.idiag_sport != tw->tw_sport &&
783                                     r->id.idiag_sport)
784                                         goto next_dying;
785                                 if (r->id.idiag_dport != tw->tw_dport &&
786                                     r->id.idiag_dport)
787                                         goto next_dying;
788                                 if (inet_twsk_diag_dump(tw, skb, cb) < 0) {
789                                         read_unlock_bh(&head->lock);
790                                         goto done;
791                                 }
792 next_dying:
793                                 ++num;
794                         }
795                 }
796                 read_unlock_bh(&head->lock);
797         }
798
799 done:
800         cb->args[1] = i;
801         cb->args[2] = num;
802         return skb->len;
803 }
804
805 static inline int inet_diag_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
806 {
807         if (!(nlh->nlmsg_flags&NLM_F_REQUEST))
808                 return 0;
809
810         if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX)
811                 goto err_inval;
812
813         if (inet_diag_table[nlh->nlmsg_type] == NULL)
814                 return -ENOENT;
815
816         if (NLMSG_LENGTH(sizeof(struct inet_diag_req)) > skb->len)
817                 goto err_inval;
818
819         if (nlh->nlmsg_flags&NLM_F_DUMP) {
820                 if (nlh->nlmsg_len >
821                     (4 + NLMSG_SPACE(sizeof(struct inet_diag_req)))) {
822                         struct rtattr *rta = (void *)(NLMSG_DATA(nlh) +
823                                                  sizeof(struct inet_diag_req));
824                         if (rta->rta_type != INET_DIAG_REQ_BYTECODE ||
825                             rta->rta_len < 8 ||
826                             rta->rta_len >
827                             (nlh->nlmsg_len -
828                              NLMSG_SPACE(sizeof(struct inet_diag_req))))
829                                 goto err_inval;
830                         if (inet_diag_bc_audit(RTA_DATA(rta), RTA_PAYLOAD(rta)))
831                                 goto err_inval;
832                 }
833                 return netlink_dump_start(idiagnl, skb, nlh,
834                                           inet_diag_dump, NULL);
835         } else
836                 return inet_diag_get_exact(skb, nlh);
837
838 err_inval:
839         return -EINVAL;
840 }
841
842
843 static inline void inet_diag_rcv_skb(struct sk_buff *skb)
844 {
845         if (skb->len >= NLMSG_SPACE(0)) {
846                 int err;
847                 struct nlmsghdr *nlh = (struct nlmsghdr *)skb->data;
848
849                 if (nlh->nlmsg_len < sizeof(*nlh) ||
850                     skb->len < nlh->nlmsg_len)
851                         return;
852                 err = inet_diag_rcv_msg(skb, nlh);
853                 if (err || nlh->nlmsg_flags & NLM_F_ACK)
854                         netlink_ack(skb, nlh, err);
855         }
856 }
857
858 static void inet_diag_rcv(struct sock *sk, int len)
859 {
860         struct sk_buff *skb;
861         unsigned int qlen = skb_queue_len(&sk->sk_receive_queue);
862
863         while (qlen-- && (skb = skb_dequeue(&sk->sk_receive_queue))) {
864                 inet_diag_rcv_skb(skb);
865                 kfree_skb(skb);
866         }
867 }
868
869 static DEFINE_SPINLOCK(inet_diag_register_lock);
870
871 int inet_diag_register(const struct inet_diag_handler *h)
872 {
873         const __u16 type = h->idiag_type;
874         int err = -EINVAL;
875
876         if (type >= INET_DIAG_GETSOCK_MAX)
877                 goto out;
878
879         spin_lock(&inet_diag_register_lock);
880         err = -EEXIST;
881         if (inet_diag_table[type] == NULL) {
882                 inet_diag_table[type] = h;
883                 err = 0;
884         }
885         spin_unlock(&inet_diag_register_lock);
886 out:
887         return err;
888 }
889 EXPORT_SYMBOL_GPL(inet_diag_register);
890
891 void inet_diag_unregister(const struct inet_diag_handler *h)
892 {
893         const __u16 type = h->idiag_type;
894
895         if (type >= INET_DIAG_GETSOCK_MAX)
896                 return;
897
898         spin_lock(&inet_diag_register_lock);
899         inet_diag_table[type] = NULL;
900         spin_unlock(&inet_diag_register_lock);
901
902         synchronize_rcu();
903 }
904 EXPORT_SYMBOL_GPL(inet_diag_unregister);
905
906 static int __init inet_diag_init(void)
907 {
908         const int inet_diag_table_size = (INET_DIAG_GETSOCK_MAX *
909                                           sizeof(struct inet_diag_handler *));
910         int err = -ENOMEM;
911
912         inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
913         if (!inet_diag_table)
914                 goto out;
915
916         idiagnl = netlink_kernel_create(NETLINK_INET_DIAG, 0, inet_diag_rcv,
917                                         THIS_MODULE);
918         if (idiagnl == NULL)
919                 goto out_free_table;
920         err = 0;
921 out:
922         return err;
923 out_free_table:
924         kfree(inet_diag_table);
925         goto out;
926 }
927
928 static void __exit inet_diag_exit(void)
929 {
930         sock_release(idiagnl->sk_socket);
931         kfree(inet_diag_table);
932 }
933
934 module_init(inet_diag_init);
935 module_exit(inet_diag_exit);
936 MODULE_LICENSE("GPL");