Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-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/kernel.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 #include <net/inet_connection_sock.h>
28 #include <net/inet_hashtables.h>
29 #include <net/inet_timewait_sock.h>
30 #include <net/inet6_hashtables.h>
31 #include <net/netlink.h>
32
33 #include <linux/inet.h>
34 #include <linux/stddef.h>
35
36 #include <linux/inet_diag.h>
37
38 static const struct inet_diag_handler **inet_diag_table;
39
40 struct inet_diag_entry {
41         __be32 *saddr;
42         __be32 *daddr;
43         u16 sport;
44         u16 dport;
45         u16 family;
46         u16 userlocks;
47 };
48
49 static struct sock *idiagnl;
50
51 #define INET_DIAG_PUT(skb, attrtype, attrlen) \
52         RTA_DATA(__RTA_PUT(skb, attrtype, attrlen))
53
54 static DEFINE_MUTEX(inet_diag_table_mutex);
55
56 static const struct inet_diag_handler *inet_diag_lock_handler(int type)
57 {
58 #ifdef CONFIG_KMOD
59         if (!inet_diag_table[type])
60                 request_module("net-pf-%d-proto-%d-type-%d", PF_NETLINK,
61                                NETLINK_INET_DIAG, type);
62 #endif
63
64         mutex_lock(&inet_diag_table_mutex);
65         if (!inet_diag_table[type])
66                 return ERR_PTR(-ENOENT);
67
68         return inet_diag_table[type];
69 }
70
71 static inline void inet_diag_unlock_handler(
72         const struct inet_diag_handler *handler)
73 {
74         mutex_unlock(&inet_diag_table_mutex);
75 }
76
77 static int inet_csk_diag_fill(struct sock *sk,
78                               struct sk_buff *skb,
79                               int ext, u32 pid, u32 seq, u16 nlmsg_flags,
80                               const struct nlmsghdr *unlh)
81 {
82         const struct inet_sock *inet = inet_sk(sk);
83         const struct inet_connection_sock *icsk = inet_csk(sk);
84         struct inet_diag_msg *r;
85         struct nlmsghdr  *nlh;
86         void *info = NULL;
87         struct inet_diag_meminfo  *minfo = NULL;
88         unsigned char    *b = skb_tail_pointer(skb);
89         const struct inet_diag_handler *handler;
90
91         handler = inet_diag_table[unlh->nlmsg_type];
92         BUG_ON(handler == NULL);
93
94         nlh = NLMSG_PUT(skb, pid, seq, unlh->nlmsg_type, sizeof(*r));
95         nlh->nlmsg_flags = nlmsg_flags;
96
97         r = NLMSG_DATA(nlh);
98         BUG_ON(sk->sk_state == TCP_TIME_WAIT);
99
100         if (ext & (1 << (INET_DIAG_MEMINFO - 1)))
101                 minfo = INET_DIAG_PUT(skb, INET_DIAG_MEMINFO, sizeof(*minfo));
102
103         if (ext & (1 << (INET_DIAG_INFO - 1)))
104                 info = INET_DIAG_PUT(skb, INET_DIAG_INFO,
105                                      handler->idiag_info_size);
106
107         if ((ext & (1 << (INET_DIAG_CONG - 1))) && icsk->icsk_ca_ops) {
108                 const size_t len = strlen(icsk->icsk_ca_ops->name);
109
110                 strcpy(INET_DIAG_PUT(skb, INET_DIAG_CONG, len + 1),
111                        icsk->icsk_ca_ops->name);
112         }
113
114         r->idiag_family = sk->sk_family;
115         r->idiag_state = sk->sk_state;
116         r->idiag_timer = 0;
117         r->idiag_retrans = 0;
118
119         r->id.idiag_if = sk->sk_bound_dev_if;
120         r->id.idiag_cookie[0] = (u32)(unsigned long)sk;
121         r->id.idiag_cookie[1] = (u32)(((unsigned long)sk >> 31) >> 1);
122
123         r->id.idiag_sport = inet->sport;
124         r->id.idiag_dport = inet->dport;
125         r->id.idiag_src[0] = inet->rcv_saddr;
126         r->id.idiag_dst[0] = inet->daddr;
127
128 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
129         if (r->idiag_family == AF_INET6) {
130                 struct ipv6_pinfo *np = inet6_sk(sk);
131
132                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
133                                &np->rcv_saddr);
134                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
135                                &np->daddr);
136         }
137 #endif
138
139 #define EXPIRES_IN_MS(tmo)  DIV_ROUND_UP((tmo - jiffies) * 1000, HZ)
140
141         if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
142                 r->idiag_timer = 1;
143                 r->idiag_retrans = icsk->icsk_retransmits;
144                 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
145         } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
146                 r->idiag_timer = 4;
147                 r->idiag_retrans = icsk->icsk_probes_out;
148                 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
149         } else if (timer_pending(&sk->sk_timer)) {
150                 r->idiag_timer = 2;
151                 r->idiag_retrans = icsk->icsk_probes_out;
152                 r->idiag_expires = EXPIRES_IN_MS(sk->sk_timer.expires);
153         } else {
154                 r->idiag_timer = 0;
155                 r->idiag_expires = 0;
156         }
157 #undef EXPIRES_IN_MS
158
159         r->idiag_uid = sock_i_uid(sk);
160         r->idiag_inode = sock_i_ino(sk);
161
162         if (minfo) {
163                 minfo->idiag_rmem = atomic_read(&sk->sk_rmem_alloc);
164                 minfo->idiag_wmem = sk->sk_wmem_queued;
165                 minfo->idiag_fmem = sk->sk_forward_alloc;
166                 minfo->idiag_tmem = atomic_read(&sk->sk_wmem_alloc);
167         }
168
169         handler->idiag_get_info(sk, r, info);
170
171         if (sk->sk_state < TCP_TIME_WAIT &&
172             icsk->icsk_ca_ops && icsk->icsk_ca_ops->get_info)
173                 icsk->icsk_ca_ops->get_info(sk, ext, skb);
174
175         nlh->nlmsg_len = skb_tail_pointer(skb) - b;
176         return skb->len;
177
178 rtattr_failure:
179 nlmsg_failure:
180         nlmsg_trim(skb, b);
181         return -EMSGSIZE;
182 }
183
184 static int inet_twsk_diag_fill(struct inet_timewait_sock *tw,
185                                struct sk_buff *skb, int ext, u32 pid,
186                                u32 seq, u16 nlmsg_flags,
187                                const struct nlmsghdr *unlh)
188 {
189         long tmo;
190         struct inet_diag_msg *r;
191         const unsigned char *previous_tail = skb_tail_pointer(skb);
192         struct nlmsghdr *nlh = NLMSG_PUT(skb, pid, seq,
193                                          unlh->nlmsg_type, sizeof(*r));
194
195         r = NLMSG_DATA(nlh);
196         BUG_ON(tw->tw_state != TCP_TIME_WAIT);
197
198         nlh->nlmsg_flags = nlmsg_flags;
199
200         tmo = tw->tw_ttd - jiffies;
201         if (tmo < 0)
202                 tmo = 0;
203
204         r->idiag_family       = tw->tw_family;
205         r->idiag_state        = tw->tw_state;
206         r->idiag_timer        = 0;
207         r->idiag_retrans      = 0;
208         r->id.idiag_if        = tw->tw_bound_dev_if;
209         r->id.idiag_cookie[0] = (u32)(unsigned long)tw;
210         r->id.idiag_cookie[1] = (u32)(((unsigned long)tw >> 31) >> 1);
211         r->id.idiag_sport     = tw->tw_sport;
212         r->id.idiag_dport     = tw->tw_dport;
213         r->id.idiag_src[0]    = tw->tw_rcv_saddr;
214         r->id.idiag_dst[0]    = tw->tw_daddr;
215         r->idiag_state        = tw->tw_substate;
216         r->idiag_timer        = 3;
217         r->idiag_expires      = DIV_ROUND_UP(tmo * 1000, HZ);
218         r->idiag_rqueue       = 0;
219         r->idiag_wqueue       = 0;
220         r->idiag_uid          = 0;
221         r->idiag_inode        = 0;
222 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
223         if (tw->tw_family == AF_INET6) {
224                 const struct inet6_timewait_sock *tw6 =
225                                                 inet6_twsk((struct sock *)tw);
226
227                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
228                                &tw6->tw_v6_rcv_saddr);
229                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
230                                &tw6->tw_v6_daddr);
231         }
232 #endif
233         nlh->nlmsg_len = skb_tail_pointer(skb) - previous_tail;
234         return skb->len;
235 nlmsg_failure:
236         nlmsg_trim(skb, previous_tail);
237         return -EMSGSIZE;
238 }
239
240 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
241                         int ext, u32 pid, u32 seq, u16 nlmsg_flags,
242                         const struct nlmsghdr *unlh)
243 {
244         if (sk->sk_state == TCP_TIME_WAIT)
245                 return inet_twsk_diag_fill((struct inet_timewait_sock *)sk,
246                                            skb, ext, pid, seq, nlmsg_flags,
247                                            unlh);
248         return inet_csk_diag_fill(sk, skb, ext, pid, seq, nlmsg_flags, unlh);
249 }
250
251 static int inet_diag_get_exact(struct sk_buff *in_skb,
252                                const struct nlmsghdr *nlh)
253 {
254         int err;
255         struct sock *sk;
256         struct inet_diag_req *req = NLMSG_DATA(nlh);
257         struct sk_buff *rep;
258         struct inet_hashinfo *hashinfo;
259         const struct inet_diag_handler *handler;
260
261         handler = inet_diag_lock_handler(nlh->nlmsg_type);
262         if (IS_ERR(handler)) {
263                 err = PTR_ERR(handler);
264                 goto unlock;
265         }
266
267         hashinfo = handler->idiag_hashinfo;
268         err = -EINVAL;
269
270         if (req->idiag_family == AF_INET) {
271                 sk = inet_lookup(&init_net, hashinfo, req->id.idiag_dst[0],
272                                  req->id.idiag_dport, req->id.idiag_src[0],
273                                  req->id.idiag_sport, req->id.idiag_if);
274         }
275 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
276         else if (req->idiag_family == AF_INET6) {
277                 sk = inet6_lookup(&init_net, hashinfo,
278                                   (struct in6_addr *)req->id.idiag_dst,
279                                   req->id.idiag_dport,
280                                   (struct in6_addr *)req->id.idiag_src,
281                                   req->id.idiag_sport,
282                                   req->id.idiag_if);
283         }
284 #endif
285         else {
286                 goto unlock;
287         }
288
289         err = -ENOENT;
290         if (sk == NULL)
291                 goto unlock;
292
293         err = -ESTALE;
294         if ((req->id.idiag_cookie[0] != INET_DIAG_NOCOOKIE ||
295              req->id.idiag_cookie[1] != INET_DIAG_NOCOOKIE) &&
296             ((u32)(unsigned long)sk != req->id.idiag_cookie[0] ||
297              (u32)((((unsigned long)sk) >> 31) >> 1) != req->id.idiag_cookie[1]))
298                 goto out;
299
300         err = -ENOMEM;
301         rep = alloc_skb(NLMSG_SPACE((sizeof(struct inet_diag_msg) +
302                                      sizeof(struct inet_diag_meminfo) +
303                                      handler->idiag_info_size + 64)),
304                         GFP_KERNEL);
305         if (!rep)
306                 goto out;
307
308         err = sk_diag_fill(sk, rep, req->idiag_ext,
309                            NETLINK_CB(in_skb).pid,
310                            nlh->nlmsg_seq, 0, nlh);
311         if (err < 0) {
312                 WARN_ON(err == -EMSGSIZE);
313                 kfree_skb(rep);
314                 goto out;
315         }
316         err = netlink_unicast(idiagnl, rep, NETLINK_CB(in_skb).pid,
317                               MSG_DONTWAIT);
318         if (err > 0)
319                 err = 0;
320
321 out:
322         if (sk) {
323                 if (sk->sk_state == TCP_TIME_WAIT)
324                         inet_twsk_put((struct inet_timewait_sock *)sk);
325                 else
326                         sock_put(sk);
327         }
328 unlock:
329         inet_diag_unlock_handler(handler);
330         return err;
331 }
332
333 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
334 {
335         int words = bits >> 5;
336
337         bits &= 0x1f;
338
339         if (words) {
340                 if (memcmp(a1, a2, words << 2))
341                         return 0;
342         }
343         if (bits) {
344                 __be32 w1, w2;
345                 __be32 mask;
346
347                 w1 = a1[words];
348                 w2 = a2[words];
349
350                 mask = htonl((0xffffffff) << (32 - bits));
351
352                 if ((w1 ^ w2) & mask)
353                         return 0;
354         }
355
356         return 1;
357 }
358
359
360 static int inet_diag_bc_run(const void *bc, int len,
361                             const struct inet_diag_entry *entry)
362 {
363         while (len > 0) {
364                 int yes = 1;
365                 const struct inet_diag_bc_op *op = bc;
366
367                 switch (op->code) {
368                 case INET_DIAG_BC_NOP:
369                         break;
370                 case INET_DIAG_BC_JMP:
371                         yes = 0;
372                         break;
373                 case INET_DIAG_BC_S_GE:
374                         yes = entry->sport >= op[1].no;
375                         break;
376                 case INET_DIAG_BC_S_LE:
377                         yes = entry->dport <= op[1].no;
378                         break;
379                 case INET_DIAG_BC_D_GE:
380                         yes = entry->dport >= op[1].no;
381                         break;
382                 case INET_DIAG_BC_D_LE:
383                         yes = entry->dport <= op[1].no;
384                         break;
385                 case INET_DIAG_BC_AUTO:
386                         yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
387                         break;
388                 case INET_DIAG_BC_S_COND:
389                 case INET_DIAG_BC_D_COND: {
390                         struct inet_diag_hostcond *cond;
391                         __be32 *addr;
392
393                         cond = (struct inet_diag_hostcond *)(op + 1);
394                         if (cond->port != -1 &&
395                             cond->port != (op->code == INET_DIAG_BC_S_COND ?
396                                              entry->sport : entry->dport)) {
397                                 yes = 0;
398                                 break;
399                         }
400
401                         if (cond->prefix_len == 0)
402                                 break;
403
404                         if (op->code == INET_DIAG_BC_S_COND)
405                                 addr = entry->saddr;
406                         else
407                                 addr = entry->daddr;
408
409                         if (bitstring_match(addr, cond->addr,
410                                             cond->prefix_len))
411                                 break;
412                         if (entry->family == AF_INET6 &&
413                             cond->family == AF_INET) {
414                                 if (addr[0] == 0 && addr[1] == 0 &&
415                                     addr[2] == htonl(0xffff) &&
416                                     bitstring_match(addr + 3, cond->addr,
417                                                     cond->prefix_len))
418                                         break;
419                         }
420                         yes = 0;
421                         break;
422                 }
423                 }
424
425                 if (yes) {
426                         len -= op->yes;
427                         bc += op->yes;
428                 } else {
429                         len -= op->no;
430                         bc += op->no;
431                 }
432         }
433         return (len == 0);
434 }
435
436 static int valid_cc(const void *bc, int len, int cc)
437 {
438         while (len >= 0) {
439                 const struct inet_diag_bc_op *op = bc;
440
441                 if (cc > len)
442                         return 0;
443                 if (cc == len)
444                         return 1;
445                 if (op->yes < 4)
446                         return 0;
447                 len -= op->yes;
448                 bc  += op->yes;
449         }
450         return 0;
451 }
452
453 static int inet_diag_bc_audit(const void *bytecode, int bytecode_len)
454 {
455         const unsigned char *bc = bytecode;
456         int  len = bytecode_len;
457
458         while (len > 0) {
459                 struct inet_diag_bc_op *op = (struct inet_diag_bc_op *)bc;
460
461 //printk("BC: %d %d %d {%d} / %d\n", op->code, op->yes, op->no, op[1].no, len);
462                 switch (op->code) {
463                 case INET_DIAG_BC_AUTO:
464                 case INET_DIAG_BC_S_COND:
465                 case INET_DIAG_BC_D_COND:
466                 case INET_DIAG_BC_S_GE:
467                 case INET_DIAG_BC_S_LE:
468                 case INET_DIAG_BC_D_GE:
469                 case INET_DIAG_BC_D_LE:
470                         if (op->yes < 4 || op->yes > len + 4)
471                                 return -EINVAL;
472                 case INET_DIAG_BC_JMP:
473                         if (op->no < 4 || op->no > len + 4)
474                                 return -EINVAL;
475                         if (op->no < len &&
476                             !valid_cc(bytecode, bytecode_len, len - op->no))
477                                 return -EINVAL;
478                         break;
479                 case INET_DIAG_BC_NOP:
480                         if (op->yes < 4 || op->yes > len + 4)
481                                 return -EINVAL;
482                         break;
483                 default:
484                         return -EINVAL;
485                 }
486                 bc  += op->yes;
487                 len -= op->yes;
488         }
489         return len == 0 ? 0 : -EINVAL;
490 }
491
492 static int inet_csk_diag_dump(struct sock *sk,
493                               struct sk_buff *skb,
494                               struct netlink_callback *cb)
495 {
496         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
497
498         if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) {
499                 struct inet_diag_entry entry;
500                 struct rtattr *bc = (struct rtattr *)(r + 1);
501                 struct inet_sock *inet = inet_sk(sk);
502
503                 entry.family = sk->sk_family;
504 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
505                 if (entry.family == AF_INET6) {
506                         struct ipv6_pinfo *np = inet6_sk(sk);
507
508                         entry.saddr = np->rcv_saddr.s6_addr32;
509                         entry.daddr = np->daddr.s6_addr32;
510                 } else
511 #endif
512                 {
513                         entry.saddr = &inet->rcv_saddr;
514                         entry.daddr = &inet->daddr;
515                 }
516                 entry.sport = inet->num;
517                 entry.dport = ntohs(inet->dport);
518                 entry.userlocks = sk->sk_userlocks;
519
520                 if (!inet_diag_bc_run(RTA_DATA(bc), RTA_PAYLOAD(bc), &entry))
521                         return 0;
522         }
523
524         return inet_csk_diag_fill(sk, 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_twsk_diag_dump(struct inet_timewait_sock *tw,
530                                struct sk_buff *skb,
531                                struct netlink_callback *cb)
532 {
533         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
534
535         if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) {
536                 struct inet_diag_entry entry;
537                 struct rtattr *bc = (struct rtattr *)(r + 1);
538
539                 entry.family = tw->tw_family;
540 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
541                 if (tw->tw_family == AF_INET6) {
542                         struct inet6_timewait_sock *tw6 =
543                                                 inet6_twsk((struct sock *)tw);
544                         entry.saddr = tw6->tw_v6_rcv_saddr.s6_addr32;
545                         entry.daddr = tw6->tw_v6_daddr.s6_addr32;
546                 } else
547 #endif
548                 {
549                         entry.saddr = &tw->tw_rcv_saddr;
550                         entry.daddr = &tw->tw_daddr;
551                 }
552                 entry.sport = tw->tw_num;
553                 entry.dport = ntohs(tw->tw_dport);
554                 entry.userlocks = 0;
555
556                 if (!inet_diag_bc_run(RTA_DATA(bc), RTA_PAYLOAD(bc), &entry))
557                         return 0;
558         }
559
560         return inet_twsk_diag_fill(tw, skb, r->idiag_ext,
561                                    NETLINK_CB(cb->skb).pid,
562                                    cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
563 }
564
565 static int inet_diag_fill_req(struct sk_buff *skb, struct sock *sk,
566                               struct request_sock *req, u32 pid, u32 seq,
567                               const struct nlmsghdr *unlh)
568 {
569         const struct inet_request_sock *ireq = inet_rsk(req);
570         struct inet_sock *inet = inet_sk(sk);
571         unsigned char *b = skb_tail_pointer(skb);
572         struct inet_diag_msg *r;
573         struct nlmsghdr *nlh;
574         long tmo;
575
576         nlh = NLMSG_PUT(skb, pid, seq, unlh->nlmsg_type, sizeof(*r));
577         nlh->nlmsg_flags = NLM_F_MULTI;
578         r = NLMSG_DATA(nlh);
579
580         r->idiag_family = sk->sk_family;
581         r->idiag_state = TCP_SYN_RECV;
582         r->idiag_timer = 1;
583         r->idiag_retrans = req->retrans;
584
585         r->id.idiag_if = sk->sk_bound_dev_if;
586         r->id.idiag_cookie[0] = (u32)(unsigned long)req;
587         r->id.idiag_cookie[1] = (u32)(((unsigned long)req >> 31) >> 1);
588
589         tmo = req->expires - jiffies;
590         if (tmo < 0)
591                 tmo = 0;
592
593         r->id.idiag_sport = inet->sport;
594         r->id.idiag_dport = ireq->rmt_port;
595         r->id.idiag_src[0] = ireq->loc_addr;
596         r->id.idiag_dst[0] = ireq->rmt_addr;
597         r->idiag_expires = jiffies_to_msecs(tmo);
598         r->idiag_rqueue = 0;
599         r->idiag_wqueue = 0;
600         r->idiag_uid = sock_i_uid(sk);
601         r->idiag_inode = 0;
602 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
603         if (r->idiag_family == AF_INET6) {
604                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
605                                &inet6_rsk(req)->loc_addr);
606                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
607                                &inet6_rsk(req)->rmt_addr);
608         }
609 #endif
610         nlh->nlmsg_len = skb_tail_pointer(skb) - b;
611
612         return skb->len;
613
614 nlmsg_failure:
615         nlmsg_trim(skb, b);
616         return -1;
617 }
618
619 static int inet_diag_dump_reqs(struct sk_buff *skb, struct sock *sk,
620                                struct netlink_callback *cb)
621 {
622         struct inet_diag_entry entry;
623         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
624         struct inet_connection_sock *icsk = inet_csk(sk);
625         struct listen_sock *lopt;
626         struct rtattr *bc = NULL;
627         struct inet_sock *inet = inet_sk(sk);
628         int j, s_j;
629         int reqnum, s_reqnum;
630         int err = 0;
631
632         s_j = cb->args[3];
633         s_reqnum = cb->args[4];
634
635         if (s_j > 0)
636                 s_j--;
637
638         entry.family = sk->sk_family;
639
640         read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
641
642         lopt = icsk->icsk_accept_queue.listen_opt;
643         if (!lopt || !lopt->qlen)
644                 goto out;
645
646         if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) {
647                 bc = (struct rtattr *)(r + 1);
648                 entry.sport = inet->num;
649                 entry.userlocks = sk->sk_userlocks;
650         }
651
652         for (j = s_j; j < lopt->nr_table_entries; j++) {
653                 struct request_sock *req, *head = lopt->syn_table[j];
654
655                 reqnum = 0;
656                 for (req = head; req; reqnum++, req = req->dl_next) {
657                         struct inet_request_sock *ireq = inet_rsk(req);
658
659                         if (reqnum < s_reqnum)
660                                 continue;
661                         if (r->id.idiag_dport != ireq->rmt_port &&
662                             r->id.idiag_dport)
663                                 continue;
664
665                         if (bc) {
666                                 entry.saddr =
667 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
668                                         (entry.family == AF_INET6) ?
669                                         inet6_rsk(req)->loc_addr.s6_addr32 :
670 #endif
671                                         &ireq->loc_addr;
672                                 entry.daddr =
673 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
674                                         (entry.family == AF_INET6) ?
675                                         inet6_rsk(req)->rmt_addr.s6_addr32 :
676 #endif
677                                         &ireq->rmt_addr;
678                                 entry.dport = ntohs(ireq->rmt_port);
679
680                                 if (!inet_diag_bc_run(RTA_DATA(bc),
681                                                     RTA_PAYLOAD(bc), &entry))
682                                         continue;
683                         }
684
685                         err = inet_diag_fill_req(skb, sk, req,
686                                                NETLINK_CB(cb->skb).pid,
687                                                cb->nlh->nlmsg_seq, cb->nlh);
688                         if (err < 0) {
689                                 cb->args[3] = j + 1;
690                                 cb->args[4] = reqnum;
691                                 goto out;
692                         }
693                 }
694
695                 s_reqnum = 0;
696         }
697
698 out:
699         read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
700
701         return err;
702 }
703
704 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
705 {
706         int i, num;
707         int s_i, s_num;
708         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
709         const struct inet_diag_handler *handler;
710         struct inet_hashinfo *hashinfo;
711
712         handler = inet_diag_lock_handler(cb->nlh->nlmsg_type);
713         if (IS_ERR(handler))
714                 goto unlock;
715
716         hashinfo = handler->idiag_hashinfo;
717
718         s_i = cb->args[1];
719         s_num = num = cb->args[2];
720
721         if (cb->args[0] == 0) {
722                 if (!(r->idiag_states & (TCPF_LISTEN | TCPF_SYN_RECV)))
723                         goto skip_listen_ht;
724
725                 inet_listen_lock(hashinfo);
726                 for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
727                         struct sock *sk;
728                         struct hlist_node *node;
729
730                         num = 0;
731                         sk_for_each(sk, node, &hashinfo->listening_hash[i]) {
732                                 struct inet_sock *inet = inet_sk(sk);
733
734                                 if (num < s_num) {
735                                         num++;
736                                         continue;
737                                 }
738
739                                 if (r->id.idiag_sport != inet->sport &&
740                                     r->id.idiag_sport)
741                                         goto next_listen;
742
743                                 if (!(r->idiag_states & TCPF_LISTEN) ||
744                                     r->id.idiag_dport ||
745                                     cb->args[3] > 0)
746                                         goto syn_recv;
747
748                                 if (inet_csk_diag_dump(sk, skb, cb) < 0) {
749                                         inet_listen_unlock(hashinfo);
750                                         goto done;
751                                 }
752
753 syn_recv:
754                                 if (!(r->idiag_states & TCPF_SYN_RECV))
755                                         goto next_listen;
756
757                                 if (inet_diag_dump_reqs(skb, sk, cb) < 0) {
758                                         inet_listen_unlock(hashinfo);
759                                         goto done;
760                                 }
761
762 next_listen:
763                                 cb->args[3] = 0;
764                                 cb->args[4] = 0;
765                                 ++num;
766                         }
767
768                         s_num = 0;
769                         cb->args[3] = 0;
770                         cb->args[4] = 0;
771                 }
772                 inet_listen_unlock(hashinfo);
773 skip_listen_ht:
774                 cb->args[0] = 1;
775                 s_i = num = s_num = 0;
776         }
777
778         if (!(r->idiag_states & ~(TCPF_LISTEN | TCPF_SYN_RECV)))
779                 goto unlock;
780
781         for (i = s_i; i < hashinfo->ehash_size; i++) {
782                 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
783                 rwlock_t *lock = inet_ehash_lockp(hashinfo, i);
784                 struct sock *sk;
785                 struct hlist_node *node;
786
787                 if (i > s_i)
788                         s_num = 0;
789
790                 read_lock_bh(lock);
791                 num = 0;
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);