Merge branch 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
[linux-2.6] / net / ipv4 / inet_diag.c
1 /*
2  * inet_diag.c  Module for monitoring INET transport protocols sockets.
3  *
4  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
5  *
6  *      This program is free software; you can redistribute it and/or
7  *      modify it under the terms of the GNU General Public License
8  *      as published by the Free Software Foundation; either version
9  *      2 of the License, or (at your option) any later version.
10  */
11
12 #include <linux/kernel.h>
13 #include <linux/module.h>
14 #include <linux/types.h>
15 #include <linux/fcntl.h>
16 #include <linux/random.h>
17 #include <linux/cache.h>
18 #include <linux/init.h>
19 #include <linux/time.h>
20
21 #include <net/icmp.h>
22 #include <net/tcp.h>
23 #include <net/ipv6.h>
24 #include <net/inet_common.h>
25 #include <net/inet_connection_sock.h>
26 #include <net/inet_hashtables.h>
27 #include <net/inet_timewait_sock.h>
28 #include <net/inet6_hashtables.h>
29 #include <net/netlink.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 DEFINE_MUTEX(inet_diag_table_mutex);
53
54 static const struct inet_diag_handler *inet_diag_lock_handler(int type)
55 {
56         if (!inet_diag_table[type])
57                 request_module("net-pf-%d-proto-%d-type-%d", PF_NETLINK,
58                                NETLINK_INET_DIAG, type);
59
60         mutex_lock(&inet_diag_table_mutex);
61         if (!inet_diag_table[type])
62                 return ERR_PTR(-ENOENT);
63
64         return inet_diag_table[type];
65 }
66
67 static inline void inet_diag_unlock_handler(
68         const struct inet_diag_handler *handler)
69 {
70         mutex_unlock(&inet_diag_table_mutex);
71 }
72
73 static int inet_csk_diag_fill(struct sock *sk,
74                               struct sk_buff *skb,
75                               int ext, u32 pid, u32 seq, u16 nlmsg_flags,
76                               const struct nlmsghdr *unlh)
77 {
78         const struct inet_sock *inet = inet_sk(sk);
79         const struct inet_connection_sock *icsk = inet_csk(sk);
80         struct inet_diag_msg *r;
81         struct nlmsghdr  *nlh;
82         void *info = NULL;
83         struct inet_diag_meminfo  *minfo = NULL;
84         unsigned char    *b = skb_tail_pointer(skb);
85         const struct inet_diag_handler *handler;
86
87         handler = inet_diag_table[unlh->nlmsg_type];
88         BUG_ON(handler == NULL);
89
90         nlh = NLMSG_PUT(skb, pid, seq, unlh->nlmsg_type, sizeof(*r));
91         nlh->nlmsg_flags = nlmsg_flags;
92
93         r = NLMSG_DATA(nlh);
94         BUG_ON(sk->sk_state == TCP_TIME_WAIT);
95
96         if (ext & (1 << (INET_DIAG_MEMINFO - 1)))
97                 minfo = INET_DIAG_PUT(skb, INET_DIAG_MEMINFO, sizeof(*minfo));
98
99         if (ext & (1 << (INET_DIAG_INFO - 1)))
100                 info = INET_DIAG_PUT(skb, INET_DIAG_INFO,
101                                      handler->idiag_info_size);
102
103         if ((ext & (1 << (INET_DIAG_CONG - 1))) && icsk->icsk_ca_ops) {
104                 const size_t len = strlen(icsk->icsk_ca_ops->name);
105
106                 strcpy(INET_DIAG_PUT(skb, INET_DIAG_CONG, len + 1),
107                        icsk->icsk_ca_ops->name);
108         }
109
110         r->idiag_family = sk->sk_family;
111         r->idiag_state = sk->sk_state;
112         r->idiag_timer = 0;
113         r->idiag_retrans = 0;
114
115         r->id.idiag_if = sk->sk_bound_dev_if;
116         r->id.idiag_cookie[0] = (u32)(unsigned long)sk;
117         r->id.idiag_cookie[1] = (u32)(((unsigned long)sk >> 31) >> 1);
118
119         r->id.idiag_sport = inet->sport;
120         r->id.idiag_dport = inet->dport;
121         r->id.idiag_src[0] = inet->rcv_saddr;
122         r->id.idiag_dst[0] = inet->daddr;
123
124 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
125         if (r->idiag_family == AF_INET6) {
126                 struct ipv6_pinfo *np = inet6_sk(sk);
127
128                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
129                                &np->rcv_saddr);
130                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
131                                &np->daddr);
132         }
133 #endif
134
135 #define EXPIRES_IN_MS(tmo)  DIV_ROUND_UP((tmo - jiffies) * 1000, HZ)
136
137         if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
138                 r->idiag_timer = 1;
139                 r->idiag_retrans = icsk->icsk_retransmits;
140                 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
141         } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
142                 r->idiag_timer = 4;
143                 r->idiag_retrans = icsk->icsk_probes_out;
144                 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
145         } else if (timer_pending(&sk->sk_timer)) {
146                 r->idiag_timer = 2;
147                 r->idiag_retrans = icsk->icsk_probes_out;
148                 r->idiag_expires = EXPIRES_IN_MS(sk->sk_timer.expires);
149         } else {
150                 r->idiag_timer = 0;
151                 r->idiag_expires = 0;
152         }
153 #undef EXPIRES_IN_MS
154
155         r->idiag_uid = sock_i_uid(sk);
156         r->idiag_inode = sock_i_ino(sk);
157
158         if (minfo) {
159                 minfo->idiag_rmem = atomic_read(&sk->sk_rmem_alloc);
160                 minfo->idiag_wmem = sk->sk_wmem_queued;
161                 minfo->idiag_fmem = sk->sk_forward_alloc;
162                 minfo->idiag_tmem = atomic_read(&sk->sk_wmem_alloc);
163         }
164
165         handler->idiag_get_info(sk, r, info);
166
167         if (sk->sk_state < TCP_TIME_WAIT &&
168             icsk->icsk_ca_ops && icsk->icsk_ca_ops->get_info)
169                 icsk->icsk_ca_ops->get_info(sk, ext, skb);
170
171         nlh->nlmsg_len = skb_tail_pointer(skb) - b;
172         return skb->len;
173
174 rtattr_failure:
175 nlmsg_failure:
176         nlmsg_trim(skb, b);
177         return -EMSGSIZE;
178 }
179
180 static int inet_twsk_diag_fill(struct inet_timewait_sock *tw,
181                                struct sk_buff *skb, int ext, u32 pid,
182                                u32 seq, u16 nlmsg_flags,
183                                const struct nlmsghdr *unlh)
184 {
185         long tmo;
186         struct inet_diag_msg *r;
187         const unsigned char *previous_tail = skb_tail_pointer(skb);
188         struct nlmsghdr *nlh = NLMSG_PUT(skb, pid, seq,
189                                          unlh->nlmsg_type, sizeof(*r));
190
191         r = NLMSG_DATA(nlh);
192         BUG_ON(tw->tw_state != TCP_TIME_WAIT);
193
194         nlh->nlmsg_flags = nlmsg_flags;
195
196         tmo = tw->tw_ttd - jiffies;
197         if (tmo < 0)
198                 tmo = 0;
199
200         r->idiag_family       = tw->tw_family;
201         r->idiag_state        = tw->tw_state;
202         r->idiag_timer        = 0;
203         r->idiag_retrans      = 0;
204         r->id.idiag_if        = tw->tw_bound_dev_if;
205         r->id.idiag_cookie[0] = (u32)(unsigned long)tw;
206         r->id.idiag_cookie[1] = (u32)(((unsigned long)tw >> 31) >> 1);
207         r->id.idiag_sport     = tw->tw_sport;
208         r->id.idiag_dport     = tw->tw_dport;
209         r->id.idiag_src[0]    = tw->tw_rcv_saddr;
210         r->id.idiag_dst[0]    = tw->tw_daddr;
211         r->idiag_state        = tw->tw_substate;
212         r->idiag_timer        = 3;
213         r->idiag_expires      = DIV_ROUND_UP(tmo * 1000, HZ);
214         r->idiag_rqueue       = 0;
215         r->idiag_wqueue       = 0;
216         r->idiag_uid          = 0;
217         r->idiag_inode        = 0;
218 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
219         if (tw->tw_family == AF_INET6) {
220                 const struct inet6_timewait_sock *tw6 =
221                                                 inet6_twsk((struct sock *)tw);
222
223                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
224                                &tw6->tw_v6_rcv_saddr);
225                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
226                                &tw6->tw_v6_daddr);
227         }
228 #endif
229         nlh->nlmsg_len = skb_tail_pointer(skb) - previous_tail;
230         return skb->len;
231 nlmsg_failure:
232         nlmsg_trim(skb, previous_tail);
233         return -EMSGSIZE;
234 }
235
236 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
237                         int ext, u32 pid, u32 seq, u16 nlmsg_flags,
238                         const struct nlmsghdr *unlh)
239 {
240         if (sk->sk_state == TCP_TIME_WAIT)
241                 return inet_twsk_diag_fill((struct inet_timewait_sock *)sk,
242                                            skb, ext, pid, seq, nlmsg_flags,
243                                            unlh);
244         return inet_csk_diag_fill(sk, skb, ext, pid, seq, nlmsg_flags, unlh);
245 }
246
247 static int inet_diag_get_exact(struct sk_buff *in_skb,
248                                const struct nlmsghdr *nlh)
249 {
250         int err;
251         struct sock *sk;
252         struct inet_diag_req *req = NLMSG_DATA(nlh);
253         struct sk_buff *rep;
254         struct inet_hashinfo *hashinfo;
255         const struct inet_diag_handler *handler;
256
257         handler = inet_diag_lock_handler(nlh->nlmsg_type);
258         if (IS_ERR(handler)) {
259                 err = PTR_ERR(handler);
260                 goto unlock;
261         }
262
263         hashinfo = handler->idiag_hashinfo;
264         err = -EINVAL;
265
266         if (req->idiag_family == AF_INET) {
267                 sk = inet_lookup(&init_net, hashinfo, req->id.idiag_dst[0],
268                                  req->id.idiag_dport, req->id.idiag_src[0],
269                                  req->id.idiag_sport, req->id.idiag_if);
270         }
271 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
272         else if (req->idiag_family == AF_INET6) {
273                 sk = inet6_lookup(&init_net, hashinfo,
274                                   (struct in6_addr *)req->id.idiag_dst,
275                                   req->id.idiag_dport,
276                                   (struct in6_addr *)req->id.idiag_src,
277                                   req->id.idiag_sport,
278                                   req->id.idiag_if);
279         }
280 #endif
281         else {
282                 goto unlock;
283         }
284
285         err = -ENOENT;
286         if (sk == NULL)
287                 goto unlock;
288
289         err = -ESTALE;
290         if ((req->id.idiag_cookie[0] != INET_DIAG_NOCOOKIE ||
291              req->id.idiag_cookie[1] != INET_DIAG_NOCOOKIE) &&
292             ((u32)(unsigned long)sk != req->id.idiag_cookie[0] ||
293              (u32)((((unsigned long)sk) >> 31) >> 1) != req->id.idiag_cookie[1]))
294                 goto out;
295
296         err = -ENOMEM;
297         rep = alloc_skb(NLMSG_SPACE((sizeof(struct inet_diag_msg) +
298                                      sizeof(struct inet_diag_meminfo) +
299                                      handler->idiag_info_size + 64)),
300                         GFP_KERNEL);
301         if (!rep)
302                 goto out;
303
304         err = sk_diag_fill(sk, rep, req->idiag_ext,
305                            NETLINK_CB(in_skb).pid,
306                            nlh->nlmsg_seq, 0, nlh);
307         if (err < 0) {
308                 WARN_ON(err == -EMSGSIZE);
309                 kfree_skb(rep);
310                 goto out;
311         }
312         err = netlink_unicast(idiagnl, rep, NETLINK_CB(in_skb).pid,
313                               MSG_DONTWAIT);
314         if (err > 0)
315                 err = 0;
316
317 out:
318         if (sk) {
319                 if (sk->sk_state == TCP_TIME_WAIT)
320                         inet_twsk_put((struct inet_timewait_sock *)sk);
321                 else
322                         sock_put(sk);
323         }
324 unlock:
325         inet_diag_unlock_handler(handler);
326         return err;
327 }
328
329 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
330 {
331         int words = bits >> 5;
332
333         bits &= 0x1f;
334
335         if (words) {
336                 if (memcmp(a1, a2, words << 2))
337                         return 0;
338         }
339         if (bits) {
340                 __be32 w1, w2;
341                 __be32 mask;
342
343                 w1 = a1[words];
344                 w2 = a2[words];
345
346                 mask = htonl((0xffffffff) << (32 - bits));
347
348                 if ((w1 ^ w2) & mask)
349                         return 0;
350         }
351
352         return 1;
353 }
354
355
356 static int inet_diag_bc_run(const void *bc, int len,
357                             const struct inet_diag_entry *entry)
358 {
359         while (len > 0) {
360                 int yes = 1;
361                 const struct inet_diag_bc_op *op = bc;
362
363                 switch (op->code) {
364                 case INET_DIAG_BC_NOP:
365                         break;
366                 case INET_DIAG_BC_JMP:
367                         yes = 0;
368                         break;
369                 case INET_DIAG_BC_S_GE:
370                         yes = entry->sport >= op[1].no;
371                         break;
372                 case INET_DIAG_BC_S_LE:
373                         yes = entry->dport <= op[1].no;
374                         break;
375                 case INET_DIAG_BC_D_GE:
376                         yes = entry->dport >= op[1].no;
377                         break;
378                 case INET_DIAG_BC_D_LE:
379                         yes = entry->dport <= op[1].no;
380                         break;
381                 case INET_DIAG_BC_AUTO:
382                         yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
383                         break;
384                 case INET_DIAG_BC_S_COND:
385                 case INET_DIAG_BC_D_COND: {
386                         struct inet_diag_hostcond *cond;
387                         __be32 *addr;
388
389                         cond = (struct inet_diag_hostcond *)(op + 1);
390                         if (cond->port != -1 &&
391                             cond->port != (op->code == INET_DIAG_BC_S_COND ?
392                                              entry->sport : entry->dport)) {
393                                 yes = 0;
394                                 break;
395                         }
396
397                         if (cond->prefix_len == 0)
398                                 break;
399
400                         if (op->code == INET_DIAG_BC_S_COND)
401                                 addr = entry->saddr;
402                         else
403                                 addr = entry->daddr;
404
405                         if (bitstring_match(addr, cond->addr,
406                                             cond->prefix_len))
407                                 break;
408                         if (entry->family == AF_INET6 &&
409                             cond->family == AF_INET) {
410                                 if (addr[0] == 0 && addr[1] == 0 &&
411                                     addr[2] == htonl(0xffff) &&
412                                     bitstring_match(addr + 3, cond->addr,
413                                                     cond->prefix_len))
414                                         break;
415                         }
416                         yes = 0;
417                         break;
418                 }
419                 }
420
421                 if (yes) {
422                         len -= op->yes;
423                         bc += op->yes;
424                 } else {
425                         len -= op->no;
426                         bc += op->no;
427                 }
428         }
429         return (len == 0);
430 }
431
432 static int valid_cc(const void *bc, int len, int cc)
433 {
434         while (len >= 0) {
435                 const struct inet_diag_bc_op *op = bc;
436
437                 if (cc > len)
438                         return 0;
439                 if (cc == len)
440                         return 1;
441                 if (op->yes < 4)
442                         return 0;
443                 len -= op->yes;
444                 bc  += op->yes;
445         }
446         return 0;
447 }
448
449 static int inet_diag_bc_audit(const void *bytecode, int bytecode_len)
450 {
451         const unsigned char *bc = bytecode;
452         int  len = bytecode_len;
453
454         while (len > 0) {
455                 struct inet_diag_bc_op *op = (struct inet_diag_bc_op *)bc;
456
457 //printk("BC: %d %d %d {%d} / %d\n", op->code, op->yes, op->no, op[1].no, len);
458                 switch (op->code) {
459                 case INET_DIAG_BC_AUTO:
460                 case INET_DIAG_BC_S_COND:
461                 case INET_DIAG_BC_D_COND:
462                 case INET_DIAG_BC_S_GE:
463                 case INET_DIAG_BC_S_LE:
464                 case INET_DIAG_BC_D_GE:
465                 case INET_DIAG_BC_D_LE:
466                         if (op->yes < 4 || op->yes > len + 4)
467                                 return -EINVAL;
468                 case INET_DIAG_BC_JMP:
469                         if (op->no < 4 || op->no > len + 4)
470                                 return -EINVAL;
471                         if (op->no < len &&
472                             !valid_cc(bytecode, bytecode_len, len - op->no))
473                                 return -EINVAL;
474                         break;
475                 case INET_DIAG_BC_NOP:
476                         if (op->yes < 4 || op->yes > len + 4)
477                                 return -EINVAL;
478                         break;
479                 default:
480                         return -EINVAL;
481                 }
482                 bc  += op->yes;
483                 len -= op->yes;
484         }
485         return len == 0 ? 0 : -EINVAL;
486 }
487
488 static int inet_csk_diag_dump(struct sock *sk,
489                               struct sk_buff *skb,
490                               struct netlink_callback *cb)
491 {
492         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
493
494         if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) {
495                 struct inet_diag_entry entry;
496                 struct rtattr *bc = (struct rtattr *)(r + 1);
497                 struct inet_sock *inet = inet_sk(sk);
498
499                 entry.family = sk->sk_family;
500 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
501                 if (entry.family == AF_INET6) {
502                         struct ipv6_pinfo *np = inet6_sk(sk);
503
504                         entry.saddr = np->rcv_saddr.s6_addr32;
505                         entry.daddr = np->daddr.s6_addr32;
506                 } else
507 #endif
508                 {
509                         entry.saddr = &inet->rcv_saddr;
510                         entry.daddr = &inet->daddr;
511                 }
512                 entry.sport = inet->num;
513                 entry.dport = ntohs(inet->dport);
514                 entry.userlocks = sk->sk_userlocks;
515
516                 if (!inet_diag_bc_run(RTA_DATA(bc), RTA_PAYLOAD(bc), &entry))
517                         return 0;
518         }
519
520         return inet_csk_diag_fill(sk, skb, r->idiag_ext,
521                                   NETLINK_CB(cb->skb).pid,
522                                   cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
523 }
524
525 static int inet_twsk_diag_dump(struct inet_timewait_sock *tw,
526                                struct sk_buff *skb,
527                                struct netlink_callback *cb)
528 {
529         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
530
531         if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) {
532                 struct inet_diag_entry entry;
533                 struct rtattr *bc = (struct rtattr *)(r + 1);
534
535                 entry.family = tw->tw_family;
536 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
537                 if (tw->tw_family == AF_INET6) {
538                         struct inet6_timewait_sock *tw6 =
539                                                 inet6_twsk((struct sock *)tw);
540                         entry.saddr = tw6->tw_v6_rcv_saddr.s6_addr32;
541                         entry.daddr = tw6->tw_v6_daddr.s6_addr32;
542                 } else
543 #endif
544                 {
545                         entry.saddr = &tw->tw_rcv_saddr;
546                         entry.daddr = &tw->tw_daddr;
547                 }
548                 entry.sport = tw->tw_num;
549                 entry.dport = ntohs(tw->tw_dport);
550                 entry.userlocks = 0;
551
552                 if (!inet_diag_bc_run(RTA_DATA(bc), RTA_PAYLOAD(bc), &entry))
553                         return 0;
554         }
555
556         return inet_twsk_diag_fill(tw, skb, r->idiag_ext,
557                                    NETLINK_CB(cb->skb).pid,
558                                    cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
559 }
560
561 static int inet_diag_fill_req(struct sk_buff *skb, struct sock *sk,
562                               struct request_sock *req, u32 pid, u32 seq,
563                               const struct nlmsghdr *unlh)
564 {
565         const struct inet_request_sock *ireq = inet_rsk(req);
566         struct inet_sock *inet = inet_sk(sk);
567         unsigned char *b = skb_tail_pointer(skb);
568         struct inet_diag_msg *r;
569         struct nlmsghdr *nlh;
570         long tmo;
571
572         nlh = NLMSG_PUT(skb, pid, seq, unlh->nlmsg_type, sizeof(*r));
573         nlh->nlmsg_flags = NLM_F_MULTI;
574         r = NLMSG_DATA(nlh);
575
576         r->idiag_family = sk->sk_family;
577         r->idiag_state = TCP_SYN_RECV;
578         r->idiag_timer = 1;
579         r->idiag_retrans = req->retrans;
580
581         r->id.idiag_if = sk->sk_bound_dev_if;
582         r->id.idiag_cookie[0] = (u32)(unsigned long)req;
583         r->id.idiag_cookie[1] = (u32)(((unsigned long)req >> 31) >> 1);
584
585         tmo = req->expires - jiffies;
586         if (tmo < 0)
587                 tmo = 0;
588
589         r->id.idiag_sport = inet->sport;
590         r->id.idiag_dport = ireq->rmt_port;
591         r->id.idiag_src[0] = ireq->loc_addr;
592         r->id.idiag_dst[0] = ireq->rmt_addr;
593         r->idiag_expires = jiffies_to_msecs(tmo);
594         r->idiag_rqueue = 0;
595         r->idiag_wqueue = 0;
596         r->idiag_uid = sock_i_uid(sk);
597         r->idiag_inode = 0;
598 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
599         if (r->idiag_family == AF_INET6) {
600                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
601                                &inet6_rsk(req)->loc_addr);
602                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
603                                &inet6_rsk(req)->rmt_addr);
604         }
605 #endif
606         nlh->nlmsg_len = skb_tail_pointer(skb) - b;
607
608         return skb->len;
609
610 nlmsg_failure:
611         nlmsg_trim(skb, b);
612         return -1;
613 }
614
615 static int inet_diag_dump_reqs(struct sk_buff *skb, struct sock *sk,
616                                struct netlink_callback *cb)
617 {
618         struct inet_diag_entry entry;
619         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
620         struct inet_connection_sock *icsk = inet_csk(sk);
621         struct listen_sock *lopt;
622         struct rtattr *bc = NULL;
623         struct inet_sock *inet = inet_sk(sk);
624         int j, s_j;
625         int reqnum, s_reqnum;
626         int err = 0;
627
628         s_j = cb->args[3];
629         s_reqnum = cb->args[4];
630
631         if (s_j > 0)
632                 s_j--;
633
634         entry.family = sk->sk_family;
635
636         read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
637
638         lopt = icsk->icsk_accept_queue.listen_opt;
639         if (!lopt || !lopt->qlen)
640                 goto out;
641
642         if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) {
643                 bc = (struct rtattr *)(r + 1);
644                 entry.sport = inet->num;
645                 entry.userlocks = sk->sk_userlocks;
646         }
647
648         for (j = s_j; j < lopt->nr_table_entries; j++) {
649                 struct request_sock *req, *head = lopt->syn_table[j];
650
651                 reqnum = 0;
652                 for (req = head; req; reqnum++, req = req->dl_next) {
653                         struct inet_request_sock *ireq = inet_rsk(req);
654
655                         if (reqnum < s_reqnum)
656                                 continue;
657                         if (r->id.idiag_dport != ireq->rmt_port &&
658                             r->id.idiag_dport)
659                                 continue;
660
661                         if (bc) {
662                                 entry.saddr =
663 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
664                                         (entry.family == AF_INET6) ?
665                                         inet6_rsk(req)->loc_addr.s6_addr32 :
666 #endif
667                                         &ireq->loc_addr;
668                                 entry.daddr =
669 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
670                                         (entry.family == AF_INET6) ?
671                                         inet6_rsk(req)->rmt_addr.s6_addr32 :
672 #endif
673                                         &ireq->rmt_addr;
674                                 entry.dport = ntohs(ireq->rmt_port);
675
676                                 if (!inet_diag_bc_run(RTA_DATA(bc),
677                                                     RTA_PAYLOAD(bc), &entry))
678                                         continue;
679                         }
680
681                         err = inet_diag_fill_req(skb, sk, req,
682                                                NETLINK_CB(cb->skb).pid,
683                                                cb->nlh->nlmsg_seq, cb->nlh);
684                         if (err < 0) {
685                                 cb->args[3] = j + 1;
686                                 cb->args[4] = reqnum;
687                                 goto out;
688                         }
689                 }
690
691                 s_reqnum = 0;
692         }
693
694 out:
695         read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
696
697         return err;
698 }
699
700 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
701 {
702         int i, num;
703         int s_i, s_num;
704         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
705         const struct inet_diag_handler *handler;
706         struct inet_hashinfo *hashinfo;
707
708         handler = inet_diag_lock_handler(cb->nlh->nlmsg_type);
709         if (IS_ERR(handler))
710                 goto unlock;
711
712         hashinfo = handler->idiag_hashinfo;
713
714         s_i = cb->args[1];
715         s_num = num = cb->args[2];
716
717         if (cb->args[0] == 0) {
718                 if (!(r->idiag_states & (TCPF_LISTEN | TCPF_SYN_RECV)))
719                         goto skip_listen_ht;
720
721                 inet_listen_lock(hashinfo);
722                 for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
723                         struct sock *sk;
724                         struct hlist_node *node;
725
726                         num = 0;
727                         sk_for_each(sk, node, &hashinfo->listening_hash[i]) {
728                                 struct inet_sock *inet = inet_sk(sk);
729
730                                 if (num < s_num) {
731                                         num++;
732                                         continue;
733                                 }
734
735                                 if (r->id.idiag_sport != inet->sport &&
736                                     r->id.idiag_sport)
737                                         goto next_listen;
738
739                                 if (!(r->idiag_states & TCPF_LISTEN) ||
740                                     r->id.idiag_dport ||
741                                     cb->args[3] > 0)
742                                         goto syn_recv;
743
744                                 if (inet_csk_diag_dump(sk, skb, cb) < 0) {
745                                         inet_listen_unlock(hashinfo);
746                                         goto done;
747                                 }
748
749 syn_recv:
750                                 if (!(r->idiag_states & TCPF_SYN_RECV))
751                                         goto next_listen;
752
753                                 if (inet_diag_dump_reqs(skb, sk, cb) < 0) {
754                                         inet_listen_unlock(hashinfo);
755                                         goto done;
756                                 }
757
758 next_listen:
759                                 cb->args[3] = 0;
760                                 cb->args[4] = 0;
761                                 ++num;
762                         }
763
764                         s_num = 0;
765                         cb->args[3] = 0;
766                         cb->args[4] = 0;
767                 }
768                 inet_listen_unlock(hashinfo);
769 skip_listen_ht:
770                 cb->args[0] = 1;
771                 s_i = num = s_num = 0;
772         }
773
774         if (!(r->idiag_states & ~(TCPF_LISTEN | TCPF_SYN_RECV)))
775                 goto unlock;
776
777         for (i = s_i; i < hashinfo->ehash_size; i++) {
778                 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
779                 rwlock_t *lock = inet_ehash_lockp(hashinfo, i);
780                 struct sock *sk;
781                 struct hlist_node *node;
782
783                 num = 0;
784
785                 if (hlist_empty(&head->chain) && hlist_empty(&head->twchain))
786                         continue;
787
788                 if (i > s_i)
789                         s_num = 0;
790
791                 read_lock_bh(lock);
792                 sk_for_each(sk, node, &head->chain) {
793                         struct inet_sock *inet = inet_sk(sk);
794
795                         if (num < s_num)
796                                 goto next_normal;
797                         if (!(r->idiag_states & (1 << sk->sk_state)))
798                                 goto next_normal;
799                         if (r->id.idiag_sport != inet->sport &&
800                             r->id.idiag_sport)
801                                 goto next_normal;
802                         if (r->id.idiag_dport != inet->dport &&
803                             r->id.idiag_dport)
804                                 goto next_normal;
805                         if (inet_csk_diag_dump(sk, skb, cb) < 0) {
806                                 read_unlock_bh(lock);
807                                 goto done;
808                         }
809 next_normal:
810                         ++num;
811                 }
812
813                 if (r->idiag_states & TCPF_TIME_WAIT) {
814                         struct inet_timewait_sock *tw;
815
816                         inet_twsk_for_each(tw, node,
817                                     &head->twchain) {
818
819                                 if (num < s_num)
820                                         goto next_dying;
821                                 if (r->id.idiag_sport != tw->tw_sport &&
822                                     r->id.idiag_sport)
823                                         goto next_dying;
824                                 if (r->id.idiag_dport != tw->tw_dport &&
825                                     r->id.idiag_dport)
826                                         goto next_dying;
827                                 if (inet_twsk_diag_dump(tw, skb, cb) < 0) {
828                                         read_unlock_bh(lock);
829                                         goto done;
830                                 }
831 next_dying:
832                                 ++num;
833                         }
834                 }
835                 read_unlock_bh(lock);
836         }
837
838 done:
839         cb->args[1] = i;
840         cb->args[2] = num;
841 unlock:
842         inet_diag_unlock_handler(handler);
843         return skb->len;
844 }
845
846 static int inet_diag_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
847 {
848         int hdrlen = sizeof(struct inet_diag_req);
849
850         if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
851             nlmsg_len(nlh) < hdrlen)
852                 return -EINVAL;
853
854         if (nlh->nlmsg_flags & NLM_F_DUMP) {
855                 if (nlmsg_attrlen(nlh, hdrlen)) {
856                         struct nlattr *attr;
857
858                         attr = nlmsg_find_attr(nlh, hdrlen,
859                                                INET_DIAG_REQ_BYTECODE);
860                         if (attr == NULL ||
861                             nla_len(attr) < sizeof(struct inet_diag_bc_op) ||
862                             inet_diag_bc_audit(nla_data(attr), nla_len(attr)))
863                                 return -EINVAL;
864                 }
865
866                 return netlink_dump_start(idiagnl, skb, nlh,
867                                           inet_diag_dump, NULL);
868         }
869
870         return inet_diag_get_exact(skb, nlh);
871 }
872
873 static DEFINE_MUTEX(inet_diag_mutex);
874
875 static void inet_diag_rcv(struct sk_buff *skb)
876 {
877         mutex_lock(&inet_diag_mutex);
878         netlink_rcv_skb(skb, &inet_diag_rcv_msg);
879         mutex_unlock(&inet_diag_mutex);
880 }
881
882 int inet_diag_register(const struct inet_diag_handler *h)
883 {
884         const __u16 type = h->idiag_type;
885         int err = -EINVAL;
886
887         if (type >= INET_DIAG_GETSOCK_MAX)
888                 goto out;
889
890         mutex_lock(&inet_diag_table_mutex);
891         err = -EEXIST;
892         if (inet_diag_table[type] == NULL) {
893                 inet_diag_table[type] = h;
894                 err = 0;
895         }
896         mutex_unlock(&inet_diag_table_mutex);
897 out:
898         return err;
899 }
900 EXPORT_SYMBOL_GPL(inet_diag_register);
901
902 void inet_diag_unregister(const struct inet_diag_handler *h)
903 {
904         const __u16 type = h->idiag_type;
905
906         if (type >= INET_DIAG_GETSOCK_MAX)
907                 return;
908
909         mutex_lock(&inet_diag_table_mutex);
910         inet_diag_table[type] = NULL;
911         mutex_unlock(&inet_diag_table_mutex);
912 }
913 EXPORT_SYMBOL_GPL(inet_diag_unregister);
914
915 static int __init inet_diag_init(void)
916 {
917         const int inet_diag_table_size = (INET_DIAG_GETSOCK_MAX *
918                                           sizeof(struct inet_diag_handler *));
919         int err = -ENOMEM;
920
921         inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
922         if (!inet_diag_table)
923                 goto out;
924
925         idiagnl = netlink_kernel_create(&init_net, NETLINK_INET_DIAG, 0,
926                                         inet_diag_rcv, NULL, THIS_MODULE);
927         if (idiagnl == NULL)
928                 goto out_free_table;
929         err = 0;
930 out:
931         return err;
932 out_free_table:
933         kfree(inet_diag_table);
934         goto out;
935 }
936
937 static void __exit inet_diag_exit(void)
938 {
939         netlink_kernel_release(idiagnl);
940         kfree(inet_diag_table);
941 }
942
943 module_init(inet_diag_init);
944 module_exit(inet_diag_exit);
945 MODULE_LICENSE("GPL");
946 MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK, NETLINK_INET_DIAG);