ipv4: fix multicast losses
[linux-2.6.git] / 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/slab.h>
18 #include <linux/cache.h>
19 #include <linux/init.h>
20 #include <linux/time.h>
21
22 #include <net/icmp.h>
23 #include <net/tcp.h>
24 #include <net/ipv6.h>
25 #include <net/inet_common.h>
26 #include <net/inet_connection_sock.h>
27 #include <net/inet_hashtables.h>
28 #include <net/inet_timewait_sock.h>
29 #include <net/inet6_hashtables.h>
30 #include <net/netlink.h>
31
32 #include <linux/inet.h>
33 #include <linux/stddef.h>
34
35 #include <linux/inet_diag.h>
36
37 static const struct inet_diag_handler **inet_diag_table;
38
39 struct inet_diag_entry {
40         __be32 *saddr;
41         __be32 *daddr;
42         u16 sport;
43         u16 dport;
44         u16 family;
45         u16 userlocks;
46 };
47
48 static struct sock *idiagnl;
49
50 #define INET_DIAG_PUT(skb, attrtype, attrlen) \
51         RTA_DATA(__RTA_PUT(skb, attrtype, attrlen))
52
53 static DEFINE_MUTEX(inet_diag_table_mutex);
54
55 static const struct inet_diag_handler *inet_diag_lock_handler(int type)
56 {
57         if (!inet_diag_table[type])
58                 request_module("net-pf-%d-proto-%d-type-%d", PF_NETLINK,
59                                NETLINK_INET_DIAG, type);
60
61         mutex_lock(&inet_diag_table_mutex);
62         if (!inet_diag_table[type])
63                 return ERR_PTR(-ENOENT);
64
65         return inet_diag_table[type];
66 }
67
68 static inline void inet_diag_unlock_handler(
69         const struct inet_diag_handler *handler)
70 {
71         mutex_unlock(&inet_diag_table_mutex);
72 }
73
74 static int inet_csk_diag_fill(struct sock *sk,
75                               struct sk_buff *skb,
76                               int ext, u32 pid, u32 seq, u16 nlmsg_flags,
77                               const struct nlmsghdr *unlh)
78 {
79         const struct inet_sock *inet = inet_sk(sk);
80         const struct inet_connection_sock *icsk = inet_csk(sk);
81         struct inet_diag_msg *r;
82         struct nlmsghdr  *nlh;
83         void *info = NULL;
84         struct inet_diag_meminfo  *minfo = NULL;
85         unsigned char    *b = skb_tail_pointer(skb);
86         const struct inet_diag_handler *handler;
87
88         handler = inet_diag_table[unlh->nlmsg_type];
89         BUG_ON(handler == NULL);
90
91         nlh = NLMSG_PUT(skb, pid, seq, unlh->nlmsg_type, sizeof(*r));
92         nlh->nlmsg_flags = nlmsg_flags;
93
94         r = NLMSG_DATA(nlh);
95         BUG_ON(sk->sk_state == TCP_TIME_WAIT);
96
97         if (ext & (1 << (INET_DIAG_MEMINFO - 1)))
98                 minfo = INET_DIAG_PUT(skb, INET_DIAG_MEMINFO, sizeof(*minfo));
99
100         if (ext & (1 << (INET_DIAG_INFO - 1)))
101                 info = INET_DIAG_PUT(skb, INET_DIAG_INFO,
102                                      handler->idiag_info_size);
103
104         if ((ext & (1 << (INET_DIAG_CONG - 1))) && icsk->icsk_ca_ops) {
105                 const size_t len = strlen(icsk->icsk_ca_ops->name);
106
107                 strcpy(INET_DIAG_PUT(skb, INET_DIAG_CONG, len + 1),
108                        icsk->icsk_ca_ops->name);
109         }
110
111         r->idiag_family = sk->sk_family;
112         r->idiag_state = sk->sk_state;
113         r->idiag_timer = 0;
114         r->idiag_retrans = 0;
115
116         r->id.idiag_if = sk->sk_bound_dev_if;
117         r->id.idiag_cookie[0] = (u32)(unsigned long)sk;
118         r->id.idiag_cookie[1] = (u32)(((unsigned long)sk >> 31) >> 1);
119
120         r->id.idiag_sport = inet->inet_sport;
121         r->id.idiag_dport = inet->inet_dport;
122         r->id.idiag_src[0] = inet->inet_rcv_saddr;
123         r->id.idiag_dst[0] = inet->inet_daddr;
124
125 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
126         if (r->idiag_family == AF_INET6) {
127                 const struct ipv6_pinfo *np = inet6_sk(sk);
128
129                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
130                                &np->rcv_saddr);
131                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
132                                &np->daddr);
133         }
134 #endif
135
136 #define EXPIRES_IN_MS(tmo)  DIV_ROUND_UP((tmo - jiffies) * 1000, HZ)
137
138         if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
139                 r->idiag_timer = 1;
140                 r->idiag_retrans = icsk->icsk_retransmits;
141                 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
142         } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
143                 r->idiag_timer = 4;
144                 r->idiag_retrans = icsk->icsk_probes_out;
145                 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
146         } else if (timer_pending(&sk->sk_timer)) {
147                 r->idiag_timer = 2;
148                 r->idiag_retrans = icsk->icsk_probes_out;
149                 r->idiag_expires = EXPIRES_IN_MS(sk->sk_timer.expires);
150         } else {
151                 r->idiag_timer = 0;
152                 r->idiag_expires = 0;
153         }
154 #undef EXPIRES_IN_MS
155
156         r->idiag_uid = sock_i_uid(sk);
157         r->idiag_inode = sock_i_ino(sk);
158
159         if (minfo) {
160                 minfo->idiag_rmem = sk_rmem_alloc_get(sk);
161                 minfo->idiag_wmem = sk->sk_wmem_queued;
162                 minfo->idiag_fmem = sk->sk_forward_alloc;
163                 minfo->idiag_tmem = sk_wmem_alloc_get(sk);
164         }
165
166         handler->idiag_get_info(sk, r, info);
167
168         if (sk->sk_state < TCP_TIME_WAIT &&
169             icsk->icsk_ca_ops && icsk->icsk_ca_ops->get_info)
170                 icsk->icsk_ca_ops->get_info(sk, ext, skb);
171
172         nlh->nlmsg_len = skb_tail_pointer(skb) - b;
173         return skb->len;
174
175 rtattr_failure:
176 nlmsg_failure:
177         nlmsg_trim(skb, b);
178         return -EMSGSIZE;
179 }
180
181 static int inet_twsk_diag_fill(struct inet_timewait_sock *tw,
182                                struct sk_buff *skb, int ext, u32 pid,
183                                u32 seq, u16 nlmsg_flags,
184                                const struct nlmsghdr *unlh)
185 {
186         long tmo;
187         struct inet_diag_msg *r;
188         const unsigned char *previous_tail = skb_tail_pointer(skb);
189         struct nlmsghdr *nlh = NLMSG_PUT(skb, pid, seq,
190                                          unlh->nlmsg_type, sizeof(*r));
191
192         r = NLMSG_DATA(nlh);
193         BUG_ON(tw->tw_state != TCP_TIME_WAIT);
194
195         nlh->nlmsg_flags = nlmsg_flags;
196
197         tmo = tw->tw_ttd - jiffies;
198         if (tmo < 0)
199                 tmo = 0;
200
201         r->idiag_family       = tw->tw_family;
202         r->idiag_retrans      = 0;
203         r->id.idiag_if        = tw->tw_bound_dev_if;
204         r->id.idiag_cookie[0] = (u32)(unsigned long)tw;
205         r->id.idiag_cookie[1] = (u32)(((unsigned long)tw >> 31) >> 1);
206         r->id.idiag_sport     = tw->tw_sport;
207         r->id.idiag_dport     = tw->tw_dport;
208         r->id.idiag_src[0]    = tw->tw_rcv_saddr;
209         r->id.idiag_dst[0]    = tw->tw_daddr;
210         r->idiag_state        = tw->tw_substate;
211         r->idiag_timer        = 3;
212         r->idiag_expires      = DIV_ROUND_UP(tmo * 1000, HZ);
213         r->idiag_rqueue       = 0;
214         r->idiag_wqueue       = 0;
215         r->idiag_uid          = 0;
216         r->idiag_inode        = 0;
217 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
218         if (tw->tw_family == AF_INET6) {
219                 const struct inet6_timewait_sock *tw6 =
220                                                 inet6_twsk((struct sock *)tw);
221
222                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
223                                &tw6->tw_v6_rcv_saddr);
224                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
225                                &tw6->tw_v6_daddr);
226         }
227 #endif
228         nlh->nlmsg_len = skb_tail_pointer(skb) - previous_tail;
229         return skb->len;
230 nlmsg_failure:
231         nlmsg_trim(skb, previous_tail);
232         return -EMSGSIZE;
233 }
234
235 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
236                         int ext, u32 pid, u32 seq, u16 nlmsg_flags,
237                         const struct nlmsghdr *unlh)
238 {
239         if (sk->sk_state == TCP_TIME_WAIT)
240                 return inet_twsk_diag_fill((struct inet_timewait_sock *)sk,
241                                            skb, ext, pid, seq, nlmsg_flags,
242                                            unlh);
243         return inet_csk_diag_fill(sk, skb, ext, pid, seq, nlmsg_flags, unlh);
244 }
245
246 static int inet_diag_get_exact(struct sk_buff *in_skb,
247                                const struct nlmsghdr *nlh)
248 {
249         int err;
250         struct sock *sk;
251         struct inet_diag_req *req = NLMSG_DATA(nlh);
252         struct sk_buff *rep;
253         struct inet_hashinfo *hashinfo;
254         const struct inet_diag_handler *handler;
255
256         handler = inet_diag_lock_handler(nlh->nlmsg_type);
257         if (IS_ERR(handler)) {
258                 err = PTR_ERR(handler);
259                 goto unlock;
260         }
261
262         hashinfo = handler->idiag_hashinfo;
263         err = -EINVAL;
264
265         if (req->idiag_family == AF_INET) {
266                 sk = inet_lookup(&init_net, hashinfo, req->id.idiag_dst[0],
267                                  req->id.idiag_dport, req->id.idiag_src[0],
268                                  req->id.idiag_sport, req->id.idiag_if);
269         }
270 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
271         else if (req->idiag_family == AF_INET6) {
272                 sk = inet6_lookup(&init_net, hashinfo,
273                                   (struct in6_addr *)req->id.idiag_dst,
274                                   req->id.idiag_dport,
275                                   (struct in6_addr *)req->id.idiag_src,
276                                   req->id.idiag_sport,
277                                   req->id.idiag_if);
278         }
279 #endif
280         else {
281                 goto unlock;
282         }
283
284         err = -ENOENT;
285         if (sk == NULL)
286                 goto unlock;
287
288         err = -ESTALE;
289         if ((req->id.idiag_cookie[0] != INET_DIAG_NOCOOKIE ||
290              req->id.idiag_cookie[1] != INET_DIAG_NOCOOKIE) &&
291             ((u32)(unsigned long)sk != req->id.idiag_cookie[0] ||
292              (u32)((((unsigned long)sk) >> 31) >> 1) != req->id.idiag_cookie[1]))
293                 goto out;
294
295         err = -ENOMEM;
296         rep = alloc_skb(NLMSG_SPACE((sizeof(struct inet_diag_msg) +
297                                      sizeof(struct inet_diag_meminfo) +
298                                      handler->idiag_info_size + 64)),
299                         GFP_KERNEL);
300         if (!rep)
301                 goto out;
302
303         err = sk_diag_fill(sk, rep, req->idiag_ext,
304                            NETLINK_CB(in_skb).pid,
305                            nlh->nlmsg_seq, 0, nlh);
306         if (err < 0) {
307                 WARN_ON(err == -EMSGSIZE);
308                 kfree_skb(rep);
309                 goto out;
310         }
311         err = netlink_unicast(idiagnl, rep, NETLINK_CB(in_skb).pid,
312                               MSG_DONTWAIT);
313         if (err > 0)
314                 err = 0;
315
316 out:
317         if (sk) {
318                 if (sk->sk_state == TCP_TIME_WAIT)
319                         inet_twsk_put((struct inet_timewait_sock *)sk);
320                 else
321                         sock_put(sk);
322         }
323 unlock:
324         inet_diag_unlock_handler(handler);
325         return err;
326 }
327
328 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
329 {
330         int words = bits >> 5;
331
332         bits &= 0x1f;
333
334         if (words) {
335                 if (memcmp(a1, a2, words << 2))
336                         return 0;
337         }
338         if (bits) {
339                 __be32 w1, w2;
340                 __be32 mask;
341
342                 w1 = a1[words];
343                 w2 = a2[words];
344
345                 mask = htonl((0xffffffff) << (32 - bits));
346
347                 if ((w1 ^ w2) & mask)
348                         return 0;
349         }
350
351         return 1;
352 }
353
354
355 static int inet_diag_bc_run(const void *bc, int len,
356                             const struct inet_diag_entry *entry)
357 {
358         while (len > 0) {
359                 int yes = 1;
360                 const struct inet_diag_bc_op *op = bc;
361
362                 switch (op->code) {
363                 case INET_DIAG_BC_NOP:
364                         break;
365                 case INET_DIAG_BC_JMP:
366                         yes = 0;
367                         break;
368                 case INET_DIAG_BC_S_GE:
369                         yes = entry->sport >= op[1].no;
370                         break;
371                 case INET_DIAG_BC_S_LE:
372                         yes = entry->sport <= op[1].no;
373                         break;
374                 case INET_DIAG_BC_D_GE:
375                         yes = entry->dport >= op[1].no;
376                         break;
377                 case INET_DIAG_BC_D_LE:
378                         yes = entry->dport <= op[1].no;
379                         break;
380                 case INET_DIAG_BC_AUTO:
381                         yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
382                         break;
383                 case INET_DIAG_BC_S_COND:
384                 case INET_DIAG_BC_D_COND: {
385                         struct inet_diag_hostcond *cond;
386                         __be32 *addr;
387
388                         cond = (struct inet_diag_hostcond *)(op + 1);
389                         if (cond->port != -1 &&
390                             cond->port != (op->code == INET_DIAG_BC_S_COND ?
391                                              entry->sport : entry->dport)) {
392                                 yes = 0;
393                                 break;
394                         }
395
396                         if (cond->prefix_len == 0)
397                                 break;
398
399                         if (op->code == INET_DIAG_BC_S_COND)
400                                 addr = entry->saddr;
401                         else
402                                 addr = entry->daddr;
403
404                         if (bitstring_match(addr, cond->addr,
405                                             cond->prefix_len))
406                                 break;
407                         if (entry->family == AF_INET6 &&
408                             cond->family == AF_INET) {
409                                 if (addr[0] == 0 && addr[1] == 0 &&
410                                     addr[2] == htonl(0xffff) &&
411                                     bitstring_match(addr + 3, cond->addr,
412                                                     cond->prefix_len))
413                                         break;
414                         }
415                         yes = 0;
416                         break;
417                 }
418                 }
419
420                 if (yes) {
421                         len -= op->yes;
422                         bc += op->yes;
423                 } else {
424                         len -= op->no;
425                         bc += op->no;
426                 }
427         }
428         return len == 0;
429 }
430
431 static int valid_cc(const void *bc, int len, int cc)
432 {
433         while (len >= 0) {
434                 const struct inet_diag_bc_op *op = bc;
435
436                 if (cc > len)
437                         return 0;
438                 if (cc == len)
439                         return 1;
440                 if (op->yes < 4 || op->yes & 3)
441                         return 0;
442                 len -= op->yes;
443                 bc  += op->yes;
444         }
445         return 0;
446 }
447
448 static int inet_diag_bc_audit(const void *bytecode, int bytecode_len)
449 {
450         const void *bc = bytecode;
451         int  len = bytecode_len;
452
453         while (len > 0) {
454                 const struct inet_diag_bc_op *op = bc;
455
456 //printk("BC: %d %d %d {%d} / %d\n", op->code, op->yes, op->no, op[1].no, len);
457                 switch (op->code) {
458                 case INET_DIAG_BC_AUTO:
459                 case INET_DIAG_BC_S_COND:
460                 case INET_DIAG_BC_D_COND:
461                 case INET_DIAG_BC_S_GE:
462                 case INET_DIAG_BC_S_LE:
463                 case INET_DIAG_BC_D_GE:
464                 case INET_DIAG_BC_D_LE:
465                 case INET_DIAG_BC_JMP:
466                         if (op->no < 4 || op->no > len + 4 || op->no & 3)
467                                 return -EINVAL;
468                         if (op->no < len &&
469                             !valid_cc(bytecode, bytecode_len, len - op->no))
470                                 return -EINVAL;
471                         break;
472                 case INET_DIAG_BC_NOP:
473                         break;
474                 default:
475                         return -EINVAL;
476                 }
477                 if (op->yes < 4 || op->yes > len + 4 || op->yes & 3)
478                         return -EINVAL;
479                 bc  += op->yes;
480                 len -= op->yes;
481         }
482         return len == 0 ? 0 : -EINVAL;
483 }
484
485 static int inet_csk_diag_dump(struct sock *sk,
486                               struct sk_buff *skb,
487                               struct netlink_callback *cb)
488 {
489         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
490
491         if (nlmsg_attrlen(cb->nlh, sizeof(*r))) {
492                 struct inet_diag_entry entry;
493                 const struct nlattr *bc = nlmsg_find_attr(cb->nlh,
494                                                           sizeof(*r),
495                                                           INET_DIAG_REQ_BYTECODE);
496                 struct inet_sock *inet = inet_sk(sk);
497
498                 entry.family = sk->sk_family;
499 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
500                 if (entry.family == AF_INET6) {
501                         struct ipv6_pinfo *np = inet6_sk(sk);
502
503                         entry.saddr = np->rcv_saddr.s6_addr32;
504                         entry.daddr = np->daddr.s6_addr32;
505                 } else
506 #endif
507                 {
508                         entry.saddr = &inet->inet_rcv_saddr;
509                         entry.daddr = &inet->inet_daddr;
510                 }
511                 entry.sport = inet->inet_num;
512                 entry.dport = ntohs(inet->inet_dport);
513                 entry.userlocks = sk->sk_userlocks;
514
515                 if (!inet_diag_bc_run(nla_data(bc), nla_len(bc), &entry))
516                         return 0;
517         }
518
519         return inet_csk_diag_fill(sk, skb, r->idiag_ext,
520                                   NETLINK_CB(cb->skb).pid,
521                                   cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
522 }
523
524 static int inet_twsk_diag_dump(struct inet_timewait_sock *tw,
525                                struct sk_buff *skb,
526                                struct netlink_callback *cb)
527 {
528         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
529
530         if (nlmsg_attrlen(cb->nlh, sizeof(*r))) {
531                 struct inet_diag_entry entry;
532                 const struct nlattr *bc = nlmsg_find_attr(cb->nlh,
533                                                           sizeof(*r),
534                                                           INET_DIAG_REQ_BYTECODE);
535
536                 entry.family = tw->tw_family;
537 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
538                 if (tw->tw_family == AF_INET6) {
539                         struct inet6_timewait_sock *tw6 =
540                                                 inet6_twsk((struct sock *)tw);
541                         entry.saddr = tw6->tw_v6_rcv_saddr.s6_addr32;
542                         entry.daddr = tw6->tw_v6_daddr.s6_addr32;
543                 } else
544 #endif
545                 {
546                         entry.saddr = &tw->tw_rcv_saddr;
547                         entry.daddr = &tw->tw_daddr;
548                 }
549                 entry.sport = tw->tw_num;
550                 entry.dport = ntohs(tw->tw_dport);
551                 entry.userlocks = 0;
552
553                 if (!inet_diag_bc_run(nla_data(bc), nla_len(bc), &entry))
554                         return 0;
555         }
556
557         return inet_twsk_diag_fill(tw, skb, r->idiag_ext,
558                                    NETLINK_CB(cb->skb).pid,
559                                    cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
560 }
561
562 static int inet_diag_fill_req(struct sk_buff *skb, struct sock *sk,
563                               struct request_sock *req, u32 pid, u32 seq,
564                               const struct nlmsghdr *unlh)
565 {
566         const struct inet_request_sock *ireq = inet_rsk(req);
567         struct inet_sock *inet = inet_sk(sk);
568         unsigned char *b = skb_tail_pointer(skb);
569         struct inet_diag_msg *r;
570         struct nlmsghdr *nlh;
571         long tmo;
572
573         nlh = NLMSG_PUT(skb, pid, seq, unlh->nlmsg_type, sizeof(*r));
574         nlh->nlmsg_flags = NLM_F_MULTI;
575         r = NLMSG_DATA(nlh);
576
577         r->idiag_family = sk->sk_family;
578         r->idiag_state = TCP_SYN_RECV;
579         r->idiag_timer = 1;
580         r->idiag_retrans = req->retrans;
581
582         r->id.idiag_if = sk->sk_bound_dev_if;
583         r->id.idiag_cookie[0] = (u32)(unsigned long)req;
584         r->id.idiag_cookie[1] = (u32)(((unsigned long)req >> 31) >> 1);
585
586         tmo = req->expires - jiffies;
587         if (tmo < 0)
588                 tmo = 0;
589
590         r->id.idiag_sport = inet->inet_sport;
591         r->id.idiag_dport = ireq->rmt_port;
592         r->id.idiag_src[0] = ireq->loc_addr;
593         r->id.idiag_dst[0] = ireq->rmt_addr;
594         r->idiag_expires = jiffies_to_msecs(tmo);
595         r->idiag_rqueue = 0;
596         r->idiag_wqueue = 0;
597         r->idiag_uid = sock_i_uid(sk);
598         r->idiag_inode = 0;
599 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
600         if (r->idiag_family == AF_INET6) {
601                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
602                                &inet6_rsk(req)->loc_addr);
603                 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
604                                &inet6_rsk(req)->rmt_addr);
605         }
606 #endif
607         nlh->nlmsg_len = skb_tail_pointer(skb) - b;
608
609         return skb->len;
610
611 nlmsg_failure:
612         nlmsg_trim(skb, b);
613         return -1;
614 }
615
616 static int inet_diag_dump_reqs(struct sk_buff *skb, struct sock *sk,
617                                struct netlink_callback *cb)
618 {
619         struct inet_diag_entry entry;
620         struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
621         struct inet_connection_sock *icsk = inet_csk(sk);
622         struct listen_sock *lopt;
623         const struct nlattr *bc = NULL;
624         struct inet_sock *inet = inet_sk(sk);
625         int j, s_j;
626         int reqnum, s_reqnum;
627         int err = 0;
628
629         s_j = cb->args[3];
630         s_reqnum = cb->args[4];
631
632         if (s_j > 0)
633                 s_j--;
634
635         entry.family = sk->sk_family;
636
637         read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
638
639         lopt = icsk->icsk_accept_queue.listen_opt;
640         if (!lopt || !lopt->qlen)
641                 goto out;
642
643         if (nlmsg_attrlen(cb->nlh, sizeof(*r))) {
644                 bc = nlmsg_find_attr(cb->nlh, sizeof(*r),
645                                      INET_DIAG_REQ_BYTECODE);
646                 entry.sport = inet->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(nla_data(bc),
679                                                       nla_len(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                 for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
724                         struct sock *sk;
725                         struct hlist_nulls_node *node;
726                         struct inet_listen_hashbucket *ilb;
727
728                         num = 0;
729                         ilb = &hashinfo->listening_hash[i];
730                         spin_lock_bh(&ilb->lock);
731                         sk_nulls_for_each(sk, node, &ilb->head) {
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->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                                         spin_unlock_bh(&ilb->lock);
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                                         spin_unlock_bh(&ilb->lock);
759                                         goto done;
760                                 }
761
762 next_listen:
763                                 cb->args[3] = 0;
764                                 cb->args[4] = 0;
765                                 ++num;
766                         }
767                         spin_unlock_bh(&ilb->lock);
768
769                         s_num = 0;
770                         cb->args[3] = 0;
771                         cb->args[4] = 0;
772                 }
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_mask; i++) {
782                 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
783                 spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
784                 struct sock *sk;
785                 struct hlist_nulls_node *node;
786
787                 num = 0;
788
789                 if (hlist_nulls_empty(&head->chain) &&
790                         hlist_nulls_empty(&head->twchain))
791                         continue;
792
793                 if (i > s_i)
794                         s_num = 0;
795
796                 spin_lock_bh(lock);
797                 sk_nulls_for_each(sk, node, &head->chain) {
798                         struct inet_sock *inet = inet_sk(sk);
799
800                         if (num < s_num)
801                                 goto next_normal;
802                         if (!(r->idiag_states & (1 << sk->sk_state)))
803                                 goto next_normal;
804                         if (r->id.idiag_sport != inet->inet_sport &&
805                             r->id.idiag_sport)
806                                 goto next_normal;
807                         if (r->id.idiag_dport != inet->inet_dport &&
808                             r->id.idiag_dport)
809                                 goto next_normal;
810                         if (inet_csk_diag_dump(sk, skb, cb) < 0) {
811                                 spin_unlock_bh(lock);
812                                 goto done;
813                         }
814 next_normal:
815                         ++num;
816                 }
817
818                 if (r->idiag_states & TCPF_TIME_WAIT) {
819                         struct inet_timewait_sock *tw;
820
821                         inet_twsk_for_each(tw, node,
822                                     &head->twchain) {
823
824                                 if (num < s_num)
825                                         goto next_dying;
826                                 if (r->id.idiag_sport != tw->tw_sport &&
827                                     r->id.idiag_sport)
828                                         goto next_dying;
829                                 if (r->id.idiag_dport != tw->tw_dport &&
830                                     r->id.idiag_dport)
831                                         goto next_dying;
832                                 if (inet_twsk_diag_dump(tw, skb, cb) < 0) {
833                                         spin_unlock_bh(lock);
834                                         goto done;
835                                 }
836 next_dying:
837                                 ++num;
838                         }
839                 }
840                 spin_unlock_bh(lock);
841         }
842
843 done:
844         cb->args[1] = i;
845         cb->args[2] = num;
846 unlock:
847         inet_diag_unlock_handler(handler);
848         return skb->len;
849 }
850
851 static int inet_diag_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
852 {
853         int hdrlen = sizeof(struct inet_diag_req);
854
855         if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
856             nlmsg_len(nlh) < hdrlen)
857                 return -EINVAL;
858
859         if (nlh->nlmsg_flags & NLM_F_DUMP) {
860                 if (nlmsg_attrlen(nlh, hdrlen)) {
861                         struct nlattr *attr;
862
863                         attr = nlmsg_find_attr(nlh, hdrlen,
864                                                INET_DIAG_REQ_BYTECODE);
865                         if (attr == NULL ||
866                             nla_len(attr) < sizeof(struct inet_diag_bc_op) ||
867                             inet_diag_bc_audit(nla_data(attr), nla_len(attr)))
868                                 return -EINVAL;
869                 }
870
871                 return netlink_dump_start(idiagnl, skb, nlh,
872                                           inet_diag_dump, NULL);
873         }
874
875         return inet_diag_get_exact(skb, nlh);
876 }
877
878 static DEFINE_MUTEX(inet_diag_mutex);
879
880 static void inet_diag_rcv(struct sk_buff *skb)
881 {
882         mutex_lock(&inet_diag_mutex);
883         netlink_rcv_skb(skb, &inet_diag_rcv_msg);
884         mutex_unlock(&inet_diag_mutex);
885 }
886
887 int inet_diag_register(const struct inet_diag_handler *h)
888 {
889         const __u16 type = h->idiag_type;
890         int err = -EINVAL;
891
892         if (type >= INET_DIAG_GETSOCK_MAX)
893                 goto out;
894
895         mutex_lock(&inet_diag_table_mutex);
896         err = -EEXIST;
897         if (inet_diag_table[type] == NULL) {
898                 inet_diag_table[type] = h;
899                 err = 0;
900         }
901         mutex_unlock(&inet_diag_table_mutex);
902 out:
903         return err;
904 }
905 EXPORT_SYMBOL_GPL(inet_diag_register);
906
907 void inet_diag_unregister(const struct inet_diag_handler *h)
908 {
909         const __u16 type = h->idiag_type;
910
911         if (type >= INET_DIAG_GETSOCK_MAX)
912                 return;
913
914         mutex_lock(&inet_diag_table_mutex);
915         inet_diag_table[type] = NULL;
916         mutex_unlock(&inet_diag_table_mutex);
917 }
918 EXPORT_SYMBOL_GPL(inet_diag_unregister);
919
920 static int __init inet_diag_init(void)
921 {
922         const int inet_diag_table_size = (INET_DIAG_GETSOCK_MAX *
923                                           sizeof(struct inet_diag_handler *));
924         int err = -ENOMEM;
925
926         inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
927         if (!inet_diag_table)
928                 goto out;
929
930         idiagnl = netlink_kernel_create(&init_net, NETLINK_INET_DIAG, 0,
931                                         inet_diag_rcv, NULL, THIS_MODULE);
932         if (idiagnl == NULL)
933                 goto out_free_table;
934         err = 0;
935 out:
936         return err;
937 out_free_table:
938         kfree(inet_diag_table);
939         goto out;
940 }
941
942 static void __exit inet_diag_exit(void)
943 {
944         netlink_kernel_release(idiagnl);
945         kfree(inet_diag_table);
946 }
947
948 module_init(inet_diag_init);
949 module_exit(inet_diag_exit);
950 MODULE_LICENSE("GPL");
951 MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK, NETLINK_INET_DIAG);