Merge branch 'linus' into x86/xen
[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 #ifdef CONFIG_KMOD
57         if (!inet_diag_table[type])
58                 request_module("net-pf-%d-proto-%d-type-%d", PF_NETLINK,
59                                NETLINK_INET_DIAG, type);
60 #endif
61
62         mutex_lock(&inet_diag_table_mutex);
63         if (!inet_diag_table[type])
64                 return ERR_PTR(-ENOENT);
65
66         return inet_diag_table[type];
67 }
68
69 static inline void inet_diag_unlock_handler(
70         const struct inet_diag_handler *handler)
71 {
72         mutex_unlock(&inet_diag_table_mutex);
73 }
74
75 static int inet_csk_diag_fill(struct sock *sk,
76                               struct sk_buff *skb,
77                               int ext, u32 pid, u32 seq, u16 nlmsg_flags,
78                               const struct nlmsghdr *unlh)
79 {
80         const struct inet_sock *inet = inet_sk(sk);
81         const struct inet_connection_sock *icsk = inet_csk(sk);
82         struct inet_diag_msg *r;
83         struct nlmsghdr  *nlh;
84         void *info = NULL;
85         struct inet_diag_meminfo  *minfo = NULL;
86         unsigned char    *b = skb_tail_pointer(skb);
87         const struct inet_diag_handler *handler;
88
89         handler = inet_diag_table[unlh->nlmsg_type];
90         BUG_ON(handler == NULL);
91
92         nlh = NLMSG_PUT(skb, pid, seq, unlh->nlmsg_type, sizeof(*r));
93         nlh->nlmsg_flags = nlmsg_flags;
94
95         r = NLMSG_DATA(nlh);
96         BUG_ON(sk->sk_state == TCP_TIME_WAIT);
97
98         if (ext & (1 << (INET_DIAG_MEMINFO - 1)))
99                 minfo = INET_DIAG_PUT(skb, INET_DIAG_MEMINFO, sizeof(*minfo));
100
101         if (ext & (1 << (INET_DIAG_INFO - 1)))
102                 info = INET_DIAG_PUT(skb, INET_DIAG_INFO,
103                                      handler->idiag_info_size);
104
105         if ((ext & (1 << (INET_DIAG_CONG - 1))) && icsk->icsk_ca_ops) {
106                 const size_t len = strlen(icsk->icsk_ca_ops->name);
107
108                 strcpy(INET_DIAG_PUT(skb, INET_DIAG_CONG, len + 1),
109                        icsk->icsk_ca_ops->name);
110         }
111
112         r->idiag_family = sk->sk_family;
113         r->idiag_state = sk->sk_state;
114         r->idiag_timer = 0;
115         r->idiag_retrans = 0;
116
117         r->id.idiag_if = sk->sk_bound_dev_if;
118         r->id.idiag_cookie[0] = (u32)(unsigned long)sk;
119         r->id.idiag_cookie[1] = (u32)(((unsigned long)sk >> 31) >> 1);
120
121         r->id.idiag_sport = inet->sport;
122         r->id.idiag_dport = inet->dport;
123         r->id.idiag_src[0] = inet->rcv_saddr;
124         r->id.idiag_dst[0] = inet->daddr;
125
126 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
127         if (r->idiag_family == AF_INET6) {
128                 struct ipv6_pinfo *np = inet6_sk(sk);
129
130                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
131                                &np->rcv_saddr);
132                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
133                                &np->daddr);
134         }
135 #endif
136
137 #define EXPIRES_IN_MS(tmo)  DIV_ROUND_UP((tmo - jiffies) * 1000, HZ)
138
139         if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
140                 r->idiag_timer = 1;
141                 r->idiag_retrans = icsk->icsk_retransmits;
142                 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
143         } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
144                 r->idiag_timer = 4;
145                 r->idiag_retrans = icsk->icsk_probes_out;
146                 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
147         } else if (timer_pending(&sk->sk_timer)) {
148                 r->idiag_timer = 2;
149                 r->idiag_retrans = icsk->icsk_probes_out;
150                 r->idiag_expires = EXPIRES_IN_MS(sk->sk_timer.expires);
151         } else {
152                 r->idiag_timer = 0;
153                 r->idiag_expires = 0;
154         }
155 #undef EXPIRES_IN_MS
156
157         r->idiag_uid = sock_i_uid(sk);
158         r->idiag_inode = sock_i_ino(sk);
159
160         if (minfo) {
161                 minfo->idiag_rmem = atomic_read(&sk->sk_rmem_alloc);
162                 minfo->idiag_wmem = sk->sk_wmem_queued;
163                 minfo->idiag_fmem = sk->sk_forward_alloc;
164                 minfo->idiag_tmem = atomic_read(&sk->sk_wmem_alloc);
165         }
166
167         handler->idiag_get_info(sk, r, info);
168
169         if (sk->sk_state < TCP_TIME_WAIT &&
170             icsk->icsk_ca_ops && icsk->icsk_ca_ops->get_info)
171                 icsk->icsk_ca_ops->get_info(sk, ext, skb);
172
173         nlh->nlmsg_len = skb_tail_pointer(skb) - b;
174         return skb->len;
175
176 rtattr_failure:
177 nlmsg_failure:
178         nlmsg_trim(skb, b);
179         return -EMSGSIZE;
180 }
181
182 static int inet_twsk_diag_fill(struct inet_timewait_sock *tw,
183                                struct sk_buff *skb, int ext, u32 pid,
184                                u32 seq, u16 nlmsg_flags,
185                                const struct nlmsghdr *unlh)
186 {
187         long tmo;
188         struct inet_diag_msg *r;
189         const unsigned char *previous_tail = skb_tail_pointer(skb);
190         struct nlmsghdr *nlh = NLMSG_PUT(skb, pid, seq,
191                                          unlh->nlmsg_type, sizeof(*r));
192
193         r = NLMSG_DATA(nlh);
194         BUG_ON(tw->tw_state != TCP_TIME_WAIT);
195
196         nlh->nlmsg_flags = nlmsg_flags;
197
198         tmo = tw->tw_ttd - jiffies;
199         if (tmo < 0)
200                 tmo = 0;
201
202         r->idiag_family       = tw->tw_family;
203         r->idiag_state        = tw->tw_state;
204         r->idiag_timer        = 0;
205         r->idiag_retrans      = 0;
206         r->id.idiag_if        = tw->tw_bound_dev_if;
207         r->id.idiag_cookie[0] = (u32)(unsigned long)tw;
208         r->id.idiag_cookie[1] = (u32)(((unsigned long)tw >> 31) >> 1);
209         r->id.idiag_sport     = tw->tw_sport;
210         r->id.idiag_dport     = tw->tw_dport;
211         r->id.idiag_src[0]    = tw->tw_rcv_saddr;
212         r->id.idiag_dst[0]    = tw->tw_daddr;
213         r->idiag_state        = tw->tw_substate;
214         r->idiag_timer        = 3;
215         r->idiag_expires      = DIV_ROUND_UP(tmo * 1000, HZ);
216         r->idiag_rqueue       = 0;
217         r->idiag_wqueue       = 0;
218         r->idiag_uid          = 0;
219         r->idiag_inode        = 0;
220 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
221         if (tw->tw_family == AF_INET6) {
222                 const struct inet6_timewait_sock *tw6 =
223                                                 inet6_twsk((struct sock *)tw);
224
225                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
226                                &tw6->tw_v6_rcv_saddr);
227                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
228                                &tw6->tw_v6_daddr);
229         }
230 #endif
231         nlh->nlmsg_len = skb_tail_pointer(skb) - previous_tail;
232         return skb->len;
233 nlmsg_failure:
234         nlmsg_trim(skb, previous_tail);
235         return -EMSGSIZE;
236 }
237
238 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
239                         int ext, u32 pid, u32 seq, u16 nlmsg_flags,
240                         const struct nlmsghdr *unlh)
241 {
242         if (sk->sk_state == TCP_TIME_WAIT)
243                 return inet_twsk_diag_fill((struct inet_timewait_sock *)sk,
244                                            skb, ext, pid, seq, nlmsg_flags,
245                                            unlh);
246         return inet_csk_diag_fill(sk, skb, ext, pid, seq, nlmsg_flags, unlh);
247 }
248
249 static int inet_diag_get_exact(struct sk_buff *in_skb,
250                                const struct nlmsghdr *nlh)
251 {
252         int err;
253         struct sock *sk;
254         struct inet_diag_req *req = NLMSG_DATA(nlh);
255         struct sk_buff *rep;
256         struct inet_hashinfo *hashinfo;
257         const struct inet_diag_handler *handler;
258
259         handler = inet_diag_lock_handler(nlh->nlmsg_type);
260         if (IS_ERR(handler)) {
261                 err = PTR_ERR(handler);
262                 goto unlock;
263         }
264
265         hashinfo = handler->idiag_hashinfo;
266         err = -EINVAL;
267
268         if (req->idiag_family == AF_INET) {
269                 sk = inet_lookup(&init_net, hashinfo, req->id.idiag_dst[0],
270                                  req->id.idiag_dport, req->id.idiag_src[0],
271                                  req->id.idiag_sport, req->id.idiag_if);
272         }
273 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
274         else if (req->idiag_family == AF_INET6) {
275                 sk = inet6_lookup(&init_net, hashinfo,
276                                   (struct in6_addr *)req->id.idiag_dst,
277                                   req->id.idiag_dport,
278                                   (struct in6_addr *)req->id.idiag_src,
279                                   req->id.idiag_sport,
280                                   req->id.idiag_if);
281         }
282 #endif
283         else {
284                 goto unlock;
285         }
286
287         err = -ENOENT;
288         if (sk == NULL)
289                 goto unlock;
290
291         err = -ESTALE;
292         if ((req->id.idiag_cookie[0] != INET_DIAG_NOCOOKIE ||
293              req->id.idiag_cookie[1] != INET_DIAG_NOCOOKIE) &&
294             ((u32)(unsigned long)sk != req->id.idiag_cookie[0] ||
295              (u32)((((unsigned long)sk) >> 31) >> 1) != req->id.idiag_cookie[1]))
296                 goto out;
297
298         err = -ENOMEM;
299         rep = alloc_skb(NLMSG_SPACE((sizeof(struct inet_diag_msg) +
300                                      sizeof(struct inet_diag_meminfo) +
301                                      handler->idiag_info_size + 64)),
302                         GFP_KERNEL);
303         if (!rep)
304                 goto out;
305
306         err = sk_diag_fill(sk, rep, req->idiag_ext,
307                            NETLINK_CB(in_skb).pid,
308                            nlh->nlmsg_seq, 0, nlh);
309         if (err < 0) {
310                 WARN_ON(err == -EMSGSIZE);
311                 kfree_skb(rep);
312                 goto out;
313         }
314         err = netlink_unicast(idiagnl, rep, NETLINK_CB(in_skb).pid,
315                               MSG_DONTWAIT);
316         if (err > 0)
317                 err = 0;
318
319 out:
320         if (sk) {
321                 if (sk->sk_state == TCP_TIME_WAIT)
322                         inet_twsk_put((struct inet_timewait_sock *)sk);
323                 else
324                         sock_put(sk);
325         }
326 unlock:
327         inet_diag_unlock_handler(handler);
328         return err;
329 }
330
331 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
332 {
333         int words = bits >> 5;
334
335         bits &= 0x1f;
336
337         if (words) {
338                 if (memcmp(a1, a2, words << 2))
339                         return 0;
340         }
341         if (bits) {
342                 __be32 w1, w2;
343                 __be32 mask;
344
345                 w1 = a1[words];
346                 w2 = a2[words];
347
348                 mask = htonl((0xffffffff) << (32 - bits));
349
350                 if ((w1 ^ w2) & mask)
351                         return 0;
352         }
353
354         return 1;
355 }
356
357
358 static int inet_diag_bc_run(const void *bc, int len,
359                             const struct inet_diag_entry *entry)
360 {
361         while (len > 0) {
362                 int yes = 1;
363                 const struct inet_diag_bc_op *op = bc;
364
365                 switch (op->code) {
366                 case INET_DIAG_BC_NOP:
367                         break;
368                 case INET_DIAG_BC_JMP:
369                         yes = 0;
370                         break;
371                 case INET_DIAG_BC_S_GE:
372                         yes = entry->sport >= op[1].no;
373                         break;
374                 case INET_DIAG_BC_S_LE:
375                         yes = entry->dport <= op[1].no;
376                         break;
377                 case INET_DIAG_BC_D_GE:
378                         yes = entry->dport >= op[1].no;
379                         break;
380                 case INET_DIAG_BC_D_LE:
381                         yes = entry->dport <= op[1].no;
382                         break;
383                 case INET_DIAG_BC_AUTO:
384                         yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
385                         break;
386                 case INET_DIAG_BC_S_COND:
387                 case INET_DIAG_BC_D_COND: {
388                         struct inet_diag_hostcond *cond;
389                         __be32 *addr;
390
391                         cond = (struct inet_diag_hostcond *)(op + 1);
392                         if (cond->port != -1 &&
393                             cond->port != (op->code == INET_DIAG_BC_S_COND ?
394                                              entry->sport : entry->dport)) {
395                                 yes = 0;
396                                 break;
397                         }
398
399                         if (cond->prefix_len == 0)
400                                 break;
401
402                         if (op->code == INET_DIAG_BC_S_COND)
403                                 addr = entry->saddr;
404                         else
405                                 addr = entry->daddr;
406
407                         if (bitstring_match(addr, cond->addr,
408                                             cond->prefix_len))
409                                 break;
410                         if (entry->family == AF_INET6 &&
411                             cond->family == AF_INET) {
412                                 if (addr[0] == 0 && addr[1] == 0 &&
413                                     addr[2] == htonl(0xffff) &&
414                                     bitstring_match(addr + 3, cond->addr,
415                                                     cond->prefix_len))
416                                         break;
417                         }
418                         yes = 0;
419                         break;
420                 }
421                 }
422
423                 if (yes) {
424                         len -= op->yes;
425                         bc += op->yes;
426                 } else {
427                         len -= op->no;
428                         bc += op->no;
429                 }
430         }
431         return (len == 0);
432 }
433
434 static int valid_cc(const void *bc, int len, int cc)
435 {
436         while (len >= 0) {
437                 const struct inet_diag_bc_op *op = bc;
438
439                 if (cc > len)
440                         return 0;
441                 if (cc == len)
442                         return 1;
443                 if (op->yes < 4)
444                         return 0;
445                 len -= op->yes;
446                 bc  += op->yes;
447         }
448         return 0;
449 }
450
451 static int inet_diag_bc_audit(const void *bytecode, int bytecode_len)
452 {
453         const unsigned char *bc = bytecode;
454         int  len = bytecode_len;
455
456         while (len > 0) {
457                 struct inet_diag_bc_op *op = (struct inet_diag_bc_op *)bc;
458
459 //printk("BC: %d %d %d {%d} / %d\n", op->code, op->yes, op->no, op[1].no, len);
460                 switch (op->code) {
461                 case INET_DIAG_BC_AUTO:
462                 case INET_DIAG_BC_S_COND:
463                 case INET_DIAG_BC_D_COND:
464                 case INET_DIAG_BC_S_GE:
465                 case INET_DIAG_BC_S_LE:
466                 case INET_DIAG_BC_D_GE:
467                 case INET_DIAG_BC_D_LE:
468                         if (op->yes < 4 || op->yes > len + 4)
469                                 return -EINVAL;
470                 case INET_DIAG_BC_JMP:
471                         if (op->no < 4 || op->no > len + 4)
472                                 return -EINVAL;
473                         if (op->no < len &&
474                             !valid_cc(bytecode, bytecode_len, len - op->no))
475                                 return -EINVAL;
476                         break;
477                 case INET_DIAG_BC_NOP:
478                         if (op->yes < 4 || op->yes > len + 4)
479                                 return -EINVAL;
480                         break;
481                 default:
482                         return -EINVAL;
483                 }
484                 bc  += op->yes;
485                 len -= op->yes;
486         }
487         return len == 0 ? 0 : -EINVAL;
488 }
489
490 static int inet_csk_diag_dump(struct sock *sk,
491                               struct sk_buff *skb,
492                               struct netlink_callback *cb)
493 {
494         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
495
496         if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) {
497                 struct inet_diag_entry entry;
498                 struct rtattr *bc = (struct rtattr *)(r + 1);
499                 struct inet_sock *inet = inet_sk(sk);
500
501                 entry.family = sk->sk_family;
502 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
503                 if (entry.family == AF_INET6) {
504                         struct ipv6_pinfo *np = inet6_sk(sk);
505
506                         entry.saddr = np->rcv_saddr.s6_addr32;
507                         entry.daddr = np->daddr.s6_addr32;
508                 } else
509 #endif
510                 {
511                         entry.saddr = &inet->rcv_saddr;
512                         entry.daddr = &inet->daddr;
513                 }
514                 entry.sport = inet->num;
515                 entry.dport = ntohs(inet->dport);
516                 entry.userlocks = sk->sk_userlocks;
517
518                 if (!inet_diag_bc_run(RTA_DATA(bc), RTA_PAYLOAD(bc), &entry))
519                         return 0;
520         }
521
522         return inet_csk_diag_fill(sk, skb, r->idiag_ext,
523                                   NETLINK_CB(cb->skb).pid,
524                                   cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
525 }
526
527 static int inet_twsk_diag_dump(struct inet_timewait_sock *tw,
528                                struct sk_buff *skb,
529                                struct netlink_callback *cb)
530 {
531         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
532
533         if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) {
534                 struct inet_diag_entry entry;
535                 struct rtattr *bc = (struct rtattr *)(r + 1);
536
537                 entry.family = tw->tw_family;
538 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
539                 if (tw->tw_family == AF_INET6) {
540                         struct inet6_timewait_sock *tw6 =
541                                                 inet6_twsk((struct sock *)tw);
542                         entry.saddr = tw6->tw_v6_rcv_saddr.s6_addr32;
543                         entry.daddr = tw6->tw_v6_daddr.s6_addr32;
544                 } else
545 #endif
546                 {
547                         entry.saddr = &tw->tw_rcv_saddr;
548                         entry.daddr = &tw->tw_daddr;
549                 }
550                 entry.sport = tw->tw_num;
551                 entry.dport = ntohs(tw->tw_dport);
552                 entry.userlocks = 0;
553
554                 if (!inet_diag_bc_run(RTA_DATA(bc), RTA_PAYLOAD(bc), &entry))
555                         return 0;
556         }
557
558         return inet_twsk_diag_fill(tw, skb, r->idiag_ext,
559                                    NETLINK_CB(cb->skb).pid,
560                                    cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
561 }
562
563 static int inet_diag_fill_req(struct sk_buff *skb, struct sock *sk,
564                               struct request_sock *req, u32 pid, u32 seq,
565                               const struct nlmsghdr *unlh)
566 {
567         const struct inet_request_sock *ireq = inet_rsk(req);
568         struct inet_sock *inet = inet_sk(sk);
569         unsigned char *b = skb_tail_pointer(skb);
570         struct inet_diag_msg *r;
571         struct nlmsghdr *nlh;
572         long tmo;
573
574         nlh = NLMSG_PUT(skb, pid, seq, unlh->nlmsg_type, sizeof(*r));
575         nlh->nlmsg_flags = NLM_F_MULTI;
576         r = NLMSG_DATA(nlh);
577
578         r->idiag_family = sk->sk_family;
579         r->idiag_state = TCP_SYN_RECV;
580         r->idiag_timer = 1;
581         r->idiag_retrans = req->retrans;
582
583         r->id.idiag_if = sk->sk_bound_dev_if;
584         r->id.idiag_cookie[0] = (u32)(unsigned long)req;
585         r->id.idiag_cookie[1] = (u32)(((unsigned long)req >> 31) >> 1);
586
587         tmo = req->expires - jiffies;
588         if (tmo < 0)
589                 tmo = 0;
590
591         r->id.idiag_sport = inet->sport;
592         r->id.idiag_dport = ireq->rmt_port;
593         r->id.idiag_src[0] = ireq->loc_addr;
594         r->id.idiag_dst[0] = ireq->rmt_addr;
595         r->idiag_expires = jiffies_to_msecs(tmo);
596         r->idiag_rqueue = 0;
597         r->idiag_wqueue = 0;
598         r->idiag_uid = sock_i_uid(sk);
599         r->idiag_inode = 0;
600 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
601         if (r->idiag_family == AF_INET6) {
602                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
603                                &inet6_rsk(req)->loc_addr);
604                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
605                                &inet6_rsk(req)->rmt_addr);
606         }
607 #endif
608         nlh->nlmsg_len = skb_tail_pointer(skb) - b;
609
610         return skb->len;
611
612 nlmsg_failure:
613         nlmsg_trim(skb, b);
614         return -1;
615 }
616
617 static int inet_diag_dump_reqs(struct sk_buff *skb, struct sock *sk,
618                                struct netlink_callback *cb)
619 {
620         struct inet_diag_entry entry;
621         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
622         struct inet_connection_sock *icsk = inet_csk(sk);
623         struct listen_sock *lopt;
624         struct rtattr *bc = NULL;
625         struct inet_sock *inet = inet_sk(sk);
626         int j, s_j;
627         int reqnum, s_reqnum;
628         int err = 0;
629
630         s_j = cb->args[3];
631         s_reqnum = cb->args[4];
632
633         if (s_j > 0)
634                 s_j--;
635
636         entry.family = sk->sk_family;
637
638         read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
639
640         lopt = icsk->icsk_accept_queue.listen_opt;
641         if (!lopt || !lopt->qlen)
642                 goto out;
643
644         if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) {
645                 bc = (struct rtattr *)(r + 1);
646                 entry.sport = inet->num;
647                 entry.userlocks = sk->sk_userlocks;
648         }
649
650         for (j = s_j; j < lopt->nr_table_entries; j++) {
651                 struct request_sock *req, *head = lopt->syn_table[j];
652
653                 reqnum = 0;
654                 for (req = head; req; reqnum++, req = req->dl_next) {
655                         struct inet_request_sock *ireq = inet_rsk(req);
656
657                         if (reqnum < s_reqnum)
658                                 continue;
659                         if (r->id.idiag_dport != ireq->rmt_port &&
660                             r->id.idiag_dport)
661                                 continue;
662
663                         if (bc) {
664                                 entry.saddr =
665 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
666                                         (entry.family == AF_INET6) ?
667                                         inet6_rsk(req)->loc_addr.s6_addr32 :
668 #endif
669                                         &ireq->loc_addr;
670                                 entry.daddr =
671 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
672                                         (entry.family == AF_INET6) ?
673                                         inet6_rsk(req)->rmt_addr.s6_addr32 :
674 #endif
675                                         &ireq->rmt_addr;
676                                 entry.dport = ntohs(ireq->rmt_port);
677
678                                 if (!inet_diag_bc_run(RTA_DATA(bc),
679                                                     RTA_PAYLOAD(bc), &entry))
680                                         continue;
681                         }
682
683                         err = inet_diag_fill_req(skb, sk, req,
684                                                NETLINK_CB(cb->skb).pid,
685                                                cb->nlh->nlmsg_seq, cb->nlh);
686                         if (err < 0) {
687                                 cb->args[3] = j + 1;
688                                 cb->args[4] = reqnum;
689                                 goto out;
690                         }
691                 }
692
693                 s_reqnum = 0;
694         }
695
696 out:
697         read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
698
699         return err;
700 }
701
702 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
703 {
704         int i, num;
705         int s_i, s_num;
706         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
707         const struct inet_diag_handler *handler;
708         struct inet_hashinfo *hashinfo;
709
710         handler = inet_diag_lock_handler(cb->nlh->nlmsg_type);
711         if (IS_ERR(handler))
712                 goto unlock;
713
714         hashinfo = handler->idiag_hashinfo;
715
716         s_i = cb->args[1];
717         s_num = num = cb->args[2];
718
719         if (cb->args[0] == 0) {
720                 if (!(r->idiag_states & (TCPF_LISTEN | TCPF_SYN_RECV)))
721                         goto skip_listen_ht;
722
723                 inet_listen_lock(hashinfo);
724                 for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
725                         struct sock *sk;
726                         struct hlist_node *node;
727
728                         num = 0;
729                         sk_for_each(sk, node, &hashinfo->listening_hash[i]) {
730                                 struct inet_sock *inet = inet_sk(sk);
731
732                                 if (num < s_num) {
733                                         num++;
734                                         continue;
735                                 }
736
737                                 if (r->id.idiag_sport != inet->sport &&
738                                     r->id.idiag_sport)
739                                         goto next_listen;
740
741                                 if (!(r->idiag_states & TCPF_LISTEN) ||
742                                     r->id.idiag_dport ||
743                                     cb->args[3] > 0)
744                                         goto syn_recv;
745
746                                 if (inet_csk_diag_dump(sk, skb, cb) < 0) {
747                                         inet_listen_unlock(hashinfo);
748                                         goto done;
749                                 }
750
751 syn_recv:
752                                 if (!(r->idiag_states & TCPF_SYN_RECV))
753                                         goto next_listen;
754
755                                 if (inet_diag_dump_reqs(skb, sk, cb) < 0) {
756                                         inet_listen_unlock(hashinfo);
757                                         goto done;
758                                 }
759
760 next_listen:
761                                 cb->args[3] = 0;
762                                 cb->args[4] = 0;
763                                 ++num;
764                         }
765
766                         s_num = 0;
767                         cb->args[3] = 0;
768                         cb->args[4] = 0;
769                 }
770                 inet_listen_unlock(hashinfo);
771 skip_listen_ht:
772                 cb->args[0] = 1;
773                 s_i = num = s_num = 0;
774         }
775
776         if (!(r->idiag_states & ~(TCPF_LISTEN | TCPF_SYN_RECV)))
777                 goto unlock;
778
779         for (i = s_i; i < hashinfo->ehash_size; i++) {
780                 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
781                 rwlock_t *lock = inet_ehash_lockp(hashinfo, i);
782                 struct sock *sk;
783                 struct hlist_node *node;
784
785                 num = 0;
786
787                 if (hlist_empty(&head->chain) && hlist_empty(&head->twchain))
788                         continue;
789
790                 if (i > s_i)
791                         s_num = 0;
792
793                 read_lock_bh(lock);
794                 sk_for_each(sk, node, &head->chain) {
795                         struct inet_sock *inet = inet_sk(sk);
796
797                         if (num < s_num)
798                                 goto next_normal;
799                         if (!(r->idiag_states & (1 << sk->sk_state)))
800                                 goto next_normal;
801                         if (r->id.idiag_sport != inet->sport &&
802                             r->id.idiag_sport)
803                                 goto next_normal;
804                         if (r->id.idiag_dport != inet->dport &&
805                             r->id.idiag_dport)
806                                 goto next_normal;
807                         if (inet_csk_diag_dump(sk, skb, cb) < 0) {
808                                 read_unlock_bh(lock);
809                                 goto done;
810                         }
811 next_normal:
812                         ++num;
813                 }
814
815                 if (r->idiag_states & TCPF_TIME_WAIT) {
816                         struct inet_timewait_sock *tw;
817
818                         inet_twsk_for_each(tw, node,
819                                     &head->twchain) {
820
821                                 if (num < s_num)
822                                         goto next_dying;
823                                 if (r->id.idiag_sport != tw->tw_sport &&
824                                     r->id.idiag_sport)
825                                         goto next_dying;
826                                 if (r->id.idiag_dport != tw->tw_dport &&
827                                     r->id.idiag_dport)
828                                         goto next_dying;
829                                 if (inet_twsk_diag_dump(tw, skb, cb) < 0) {
830                                         read_unlock_bh(lock);
831                                         goto done;
832                                 }
833 next_dying:
834                                 ++num;
835                         }
836                 }
837                 read_unlock_bh(lock);
838         }
839
840 done:
841         cb->args[1] = i;
842         cb->args[2] = num;
843 unlock:
844         inet_diag_unlock_handler(handler);
845         return skb->len;
846 }
847
848 static int inet_diag_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
849 {
850         int hdrlen = sizeof(struct inet_diag_req);
851
852         if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
853             nlmsg_len(nlh) < hdrlen)
854                 return -EINVAL;
855
856         if (nlh->nlmsg_flags & NLM_F_DUMP) {
857                 if (nlmsg_attrlen(nlh, hdrlen)) {
858                         struct nlattr *attr;
859
860                         attr = nlmsg_find_attr(nlh, hdrlen,
861                                                INET_DIAG_REQ_BYTECODE);
862                         if (attr == NULL ||
863                             nla_len(attr) < sizeof(struct inet_diag_bc_op) ||
864                             inet_diag_bc_audit(nla_data(attr), nla_len(attr)))
865                                 return -EINVAL;
866                 }
867
868                 return netlink_dump_start(idiagnl, skb, nlh,
869                                           inet_diag_dump, NULL);
870         }
871
872         return inet_diag_get_exact(skb, nlh);
873 }
874
875 static DEFINE_MUTEX(inet_diag_mutex);
876
877 static void inet_diag_rcv(struct sk_buff *skb)
878 {
879         mutex_lock(&inet_diag_mutex);
880         netlink_rcv_skb(skb, &inet_diag_rcv_msg);
881         mutex_unlock(&inet_diag_mutex);
882 }
883
884 int inet_diag_register(const struct inet_diag_handler *h)
885 {
886         const __u16 type = h->idiag_type;
887         int err = -EINVAL;
888
889         if (type >= INET_DIAG_GETSOCK_MAX)
890                 goto out;
891
892         mutex_lock(&inet_diag_table_mutex);
893         err = -EEXIST;
894         if (inet_diag_table[type] == NULL) {
895                 inet_diag_table[type] = h;
896                 err = 0;
897         }
898         mutex_unlock(&inet_diag_table_mutex);
899 out:
900         return err;
901 }
902 EXPORT_SYMBOL_GPL(inet_diag_register);
903
904 void inet_diag_unregister(const struct inet_diag_handler *h)
905 {
906         const __u16 type = h->idiag_type;
907
908         if (type >= INET_DIAG_GETSOCK_MAX)
909                 return;
910
911         mutex_lock(&inet_diag_table_mutex);
912         inet_diag_table[type] = NULL;
913         mutex_unlock(&inet_diag_table_mutex);
914 }
915 EXPORT_SYMBOL_GPL(inet_diag_unregister);
916
917 static int __init inet_diag_init(void)
918 {
919         const int inet_diag_table_size = (INET_DIAG_GETSOCK_MAX *
920                                           sizeof(struct inet_diag_handler *));
921         int err = -ENOMEM;
922
923         inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
924         if (!inet_diag_table)
925                 goto out;
926
927         idiagnl = netlink_kernel_create(&init_net, NETLINK_INET_DIAG, 0,
928                                         inet_diag_rcv, NULL, THIS_MODULE);
929         if (idiagnl == NULL)
930                 goto out_free_table;
931         err = 0;
932 out:
933         return err;
934 out_free_table:
935         kfree(inet_diag_table);
936         goto out;
937 }
938
939 static void __exit inet_diag_exit(void)
940 {
941         netlink_kernel_release(idiagnl);
942         kfree(inet_diag_table);
943 }
944
945 module_init(inet_diag_init);
946 module_exit(inet_diag_exit);
947 MODULE_LICENSE("GPL");
948 MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK, NETLINK_INET_DIAG);