x86/smpboot: Init apic mapping before usage
[cascardo/linux.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 #include <linux/sock_diag.h>
37
38 static const struct inet_diag_handler **inet_diag_table;
39
40 struct inet_diag_entry {
41         const __be32 *saddr;
42         const __be32 *daddr;
43         u16 sport;
44         u16 dport;
45         u16 family;
46         u16 userlocks;
47         u32 ifindex;
48         u32 mark;
49 };
50
51 static DEFINE_MUTEX(inet_diag_table_mutex);
52
53 static const struct inet_diag_handler *inet_diag_lock_handler(int proto)
54 {
55         if (!inet_diag_table[proto])
56                 request_module("net-pf-%d-proto-%d-type-%d-%d", PF_NETLINK,
57                                NETLINK_SOCK_DIAG, AF_INET, proto);
58
59         mutex_lock(&inet_diag_table_mutex);
60         if (!inet_diag_table[proto])
61                 return ERR_PTR(-ENOENT);
62
63         return inet_diag_table[proto];
64 }
65
66 static void inet_diag_unlock_handler(const struct inet_diag_handler *handler)
67 {
68         mutex_unlock(&inet_diag_table_mutex);
69 }
70
71 void inet_diag_msg_common_fill(struct inet_diag_msg *r, struct sock *sk)
72 {
73         r->idiag_family = sk->sk_family;
74
75         r->id.idiag_sport = htons(sk->sk_num);
76         r->id.idiag_dport = sk->sk_dport;
77         r->id.idiag_if = sk->sk_bound_dev_if;
78         sock_diag_save_cookie(sk, r->id.idiag_cookie);
79
80 #if IS_ENABLED(CONFIG_IPV6)
81         if (sk->sk_family == AF_INET6) {
82                 *(struct in6_addr *)r->id.idiag_src = sk->sk_v6_rcv_saddr;
83                 *(struct in6_addr *)r->id.idiag_dst = sk->sk_v6_daddr;
84         } else
85 #endif
86         {
87         memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src));
88         memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst));
89
90         r->id.idiag_src[0] = sk->sk_rcv_saddr;
91         r->id.idiag_dst[0] = sk->sk_daddr;
92         }
93 }
94 EXPORT_SYMBOL_GPL(inet_diag_msg_common_fill);
95
96 static size_t inet_sk_attr_size(void)
97 {
98         return    nla_total_size(sizeof(struct tcp_info))
99                 + nla_total_size(1) /* INET_DIAG_SHUTDOWN */
100                 + nla_total_size(1) /* INET_DIAG_TOS */
101                 + nla_total_size(1) /* INET_DIAG_TCLASS */
102                 + nla_total_size(4) /* INET_DIAG_MARK */
103                 + nla_total_size(sizeof(struct inet_diag_meminfo))
104                 + nla_total_size(sizeof(struct inet_diag_msg))
105                 + nla_total_size(SK_MEMINFO_VARS * sizeof(u32))
106                 + nla_total_size(TCP_CA_NAME_MAX)
107                 + nla_total_size(sizeof(struct tcpvegas_info))
108                 + 64;
109 }
110
111 int inet_diag_msg_attrs_fill(struct sock *sk, struct sk_buff *skb,
112                              struct inet_diag_msg *r, int ext,
113                              struct user_namespace *user_ns,
114                              bool net_admin)
115 {
116         const struct inet_sock *inet = inet_sk(sk);
117
118         if (nla_put_u8(skb, INET_DIAG_SHUTDOWN, sk->sk_shutdown))
119                 goto errout;
120
121         /* IPv6 dual-stack sockets use inet->tos for IPv4 connections,
122          * hence this needs to be included regardless of socket family.
123          */
124         if (ext & (1 << (INET_DIAG_TOS - 1)))
125                 if (nla_put_u8(skb, INET_DIAG_TOS, inet->tos) < 0)
126                         goto errout;
127
128 #if IS_ENABLED(CONFIG_IPV6)
129         if (r->idiag_family == AF_INET6) {
130                 if (ext & (1 << (INET_DIAG_TCLASS - 1)))
131                         if (nla_put_u8(skb, INET_DIAG_TCLASS,
132                                        inet6_sk(sk)->tclass) < 0)
133                                 goto errout;
134
135                 if (((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
136                     nla_put_u8(skb, INET_DIAG_SKV6ONLY, ipv6_only_sock(sk)))
137                         goto errout;
138         }
139 #endif
140
141         if (net_admin && nla_put_u32(skb, INET_DIAG_MARK, sk->sk_mark))
142                 goto errout;
143
144         r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk));
145         r->idiag_inode = sock_i_ino(sk);
146
147         return 0;
148 errout:
149         return 1;
150 }
151 EXPORT_SYMBOL_GPL(inet_diag_msg_attrs_fill);
152
153 int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk,
154                       struct sk_buff *skb, const struct inet_diag_req_v2 *req,
155                       struct user_namespace *user_ns,
156                       u32 portid, u32 seq, u16 nlmsg_flags,
157                       const struct nlmsghdr *unlh,
158                       bool net_admin)
159 {
160         const struct tcp_congestion_ops *ca_ops;
161         const struct inet_diag_handler *handler;
162         int ext = req->idiag_ext;
163         struct inet_diag_msg *r;
164         struct nlmsghdr  *nlh;
165         struct nlattr *attr;
166         void *info = NULL;
167
168         handler = inet_diag_table[req->sdiag_protocol];
169         BUG_ON(!handler);
170
171         nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
172                         nlmsg_flags);
173         if (!nlh)
174                 return -EMSGSIZE;
175
176         r = nlmsg_data(nlh);
177         BUG_ON(!sk_fullsock(sk));
178
179         inet_diag_msg_common_fill(r, sk);
180         r->idiag_state = sk->sk_state;
181         r->idiag_timer = 0;
182         r->idiag_retrans = 0;
183
184         if (inet_diag_msg_attrs_fill(sk, skb, r, ext, user_ns, net_admin))
185                 goto errout;
186
187         if (ext & (1 << (INET_DIAG_MEMINFO - 1))) {
188                 struct inet_diag_meminfo minfo = {
189                         .idiag_rmem = sk_rmem_alloc_get(sk),
190                         .idiag_wmem = sk->sk_wmem_queued,
191                         .idiag_fmem = sk->sk_forward_alloc,
192                         .idiag_tmem = sk_wmem_alloc_get(sk),
193                 };
194
195                 if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0)
196                         goto errout;
197         }
198
199         if (ext & (1 << (INET_DIAG_SKMEMINFO - 1)))
200                 if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO))
201                         goto errout;
202
203         if (!icsk) {
204                 handler->idiag_get_info(sk, r, NULL);
205                 goto out;
206         }
207
208         if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
209             icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS ||
210             icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
211                 r->idiag_timer = 1;
212                 r->idiag_retrans = icsk->icsk_retransmits;
213                 r->idiag_expires =
214                         jiffies_to_msecs(icsk->icsk_timeout - jiffies);
215         } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
216                 r->idiag_timer = 4;
217                 r->idiag_retrans = icsk->icsk_probes_out;
218                 r->idiag_expires =
219                         jiffies_to_msecs(icsk->icsk_timeout - jiffies);
220         } else if (timer_pending(&sk->sk_timer)) {
221                 r->idiag_timer = 2;
222                 r->idiag_retrans = icsk->icsk_probes_out;
223                 r->idiag_expires =
224                         jiffies_to_msecs(sk->sk_timer.expires - jiffies);
225         } else {
226                 r->idiag_timer = 0;
227                 r->idiag_expires = 0;
228         }
229
230         if ((ext & (1 << (INET_DIAG_INFO - 1))) && handler->idiag_info_size) {
231                 attr = nla_reserve_64bit(skb, INET_DIAG_INFO,
232                                          handler->idiag_info_size,
233                                          INET_DIAG_PAD);
234                 if (!attr)
235                         goto errout;
236
237                 info = nla_data(attr);
238         }
239
240         if (ext & (1 << (INET_DIAG_CONG - 1))) {
241                 int err = 0;
242
243                 rcu_read_lock();
244                 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
245                 if (ca_ops)
246                         err = nla_put_string(skb, INET_DIAG_CONG, ca_ops->name);
247                 rcu_read_unlock();
248                 if (err < 0)
249                         goto errout;
250         }
251
252         handler->idiag_get_info(sk, r, info);
253
254         if (sk->sk_state < TCP_TIME_WAIT) {
255                 union tcp_cc_info info;
256                 size_t sz = 0;
257                 int attr;
258
259                 rcu_read_lock();
260                 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
261                 if (ca_ops && ca_ops->get_info)
262                         sz = ca_ops->get_info(sk, ext, &attr, &info);
263                 rcu_read_unlock();
264                 if (sz && nla_put(skb, attr, sz, &info) < 0)
265                         goto errout;
266         }
267
268 out:
269         nlmsg_end(skb, nlh);
270         return 0;
271
272 errout:
273         nlmsg_cancel(skb, nlh);
274         return -EMSGSIZE;
275 }
276 EXPORT_SYMBOL_GPL(inet_sk_diag_fill);
277
278 static int inet_csk_diag_fill(struct sock *sk,
279                               struct sk_buff *skb,
280                               const struct inet_diag_req_v2 *req,
281                               struct user_namespace *user_ns,
282                               u32 portid, u32 seq, u16 nlmsg_flags,
283                               const struct nlmsghdr *unlh,
284                               bool net_admin)
285 {
286         return inet_sk_diag_fill(sk, inet_csk(sk), skb, req, user_ns,
287                                  portid, seq, nlmsg_flags, unlh, net_admin);
288 }
289
290 static int inet_twsk_diag_fill(struct sock *sk,
291                                struct sk_buff *skb,
292                                u32 portid, u32 seq, u16 nlmsg_flags,
293                                const struct nlmsghdr *unlh)
294 {
295         struct inet_timewait_sock *tw = inet_twsk(sk);
296         struct inet_diag_msg *r;
297         struct nlmsghdr *nlh;
298         long tmo;
299
300         nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
301                         nlmsg_flags);
302         if (!nlh)
303                 return -EMSGSIZE;
304
305         r = nlmsg_data(nlh);
306         BUG_ON(tw->tw_state != TCP_TIME_WAIT);
307
308         tmo = tw->tw_timer.expires - jiffies;
309         if (tmo < 0)
310                 tmo = 0;
311
312         inet_diag_msg_common_fill(r, sk);
313         r->idiag_retrans      = 0;
314
315         r->idiag_state        = tw->tw_substate;
316         r->idiag_timer        = 3;
317         r->idiag_expires      = jiffies_to_msecs(tmo);
318         r->idiag_rqueue       = 0;
319         r->idiag_wqueue       = 0;
320         r->idiag_uid          = 0;
321         r->idiag_inode        = 0;
322
323         nlmsg_end(skb, nlh);
324         return 0;
325 }
326
327 static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb,
328                               u32 portid, u32 seq, u16 nlmsg_flags,
329                               const struct nlmsghdr *unlh, bool net_admin)
330 {
331         struct request_sock *reqsk = inet_reqsk(sk);
332         struct inet_diag_msg *r;
333         struct nlmsghdr *nlh;
334         long tmo;
335
336         nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
337                         nlmsg_flags);
338         if (!nlh)
339                 return -EMSGSIZE;
340
341         r = nlmsg_data(nlh);
342         inet_diag_msg_common_fill(r, sk);
343         r->idiag_state = TCP_SYN_RECV;
344         r->idiag_timer = 1;
345         r->idiag_retrans = reqsk->num_retrans;
346
347         BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) !=
348                      offsetof(struct sock, sk_cookie));
349
350         tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies;
351         r->idiag_expires = (tmo >= 0) ? jiffies_to_msecs(tmo) : 0;
352         r->idiag_rqueue = 0;
353         r->idiag_wqueue = 0;
354         r->idiag_uid    = 0;
355         r->idiag_inode  = 0;
356
357         if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
358                                      inet_rsk(reqsk)->ir_mark))
359                 return -EMSGSIZE;
360
361         nlmsg_end(skb, nlh);
362         return 0;
363 }
364
365 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
366                         const struct inet_diag_req_v2 *r,
367                         struct user_namespace *user_ns,
368                         u32 portid, u32 seq, u16 nlmsg_flags,
369                         const struct nlmsghdr *unlh, bool net_admin)
370 {
371         if (sk->sk_state == TCP_TIME_WAIT)
372                 return inet_twsk_diag_fill(sk, skb, portid, seq,
373                                            nlmsg_flags, unlh);
374
375         if (sk->sk_state == TCP_NEW_SYN_RECV)
376                 return inet_req_diag_fill(sk, skb, portid, seq,
377                                           nlmsg_flags, unlh, net_admin);
378
379         return inet_csk_diag_fill(sk, skb, r, user_ns, portid, seq,
380                                   nlmsg_flags, unlh, net_admin);
381 }
382
383 struct sock *inet_diag_find_one_icsk(struct net *net,
384                                      struct inet_hashinfo *hashinfo,
385                                      const struct inet_diag_req_v2 *req)
386 {
387         struct sock *sk;
388
389         rcu_read_lock();
390         if (req->sdiag_family == AF_INET)
391                 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0],
392                                  req->id.idiag_dport, req->id.idiag_src[0],
393                                  req->id.idiag_sport, req->id.idiag_if);
394 #if IS_ENABLED(CONFIG_IPV6)
395         else if (req->sdiag_family == AF_INET6) {
396                 if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) &&
397                     ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src))
398                         sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3],
399                                          req->id.idiag_dport, req->id.idiag_src[3],
400                                          req->id.idiag_sport, req->id.idiag_if);
401                 else
402                         sk = inet6_lookup(net, hashinfo, NULL, 0,
403                                           (struct in6_addr *)req->id.idiag_dst,
404                                           req->id.idiag_dport,
405                                           (struct in6_addr *)req->id.idiag_src,
406                                           req->id.idiag_sport,
407                                           req->id.idiag_if);
408         }
409 #endif
410         else {
411                 rcu_read_unlock();
412                 return ERR_PTR(-EINVAL);
413         }
414         rcu_read_unlock();
415         if (!sk)
416                 return ERR_PTR(-ENOENT);
417
418         if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) {
419                 sock_gen_put(sk);
420                 return ERR_PTR(-ENOENT);
421         }
422
423         return sk;
424 }
425 EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk);
426
427 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo,
428                             struct sk_buff *in_skb,
429                             const struct nlmsghdr *nlh,
430                             const struct inet_diag_req_v2 *req)
431 {
432         struct net *net = sock_net(in_skb->sk);
433         struct sk_buff *rep;
434         struct sock *sk;
435         int err;
436
437         sk = inet_diag_find_one_icsk(net, hashinfo, req);
438         if (IS_ERR(sk))
439                 return PTR_ERR(sk);
440
441         rep = nlmsg_new(inet_sk_attr_size(), GFP_KERNEL);
442         if (!rep) {
443                 err = -ENOMEM;
444                 goto out;
445         }
446
447         err = sk_diag_fill(sk, rep, req,
448                            sk_user_ns(NETLINK_CB(in_skb).sk),
449                            NETLINK_CB(in_skb).portid,
450                            nlh->nlmsg_seq, 0, nlh,
451                            netlink_net_capable(in_skb, CAP_NET_ADMIN));
452         if (err < 0) {
453                 WARN_ON(err == -EMSGSIZE);
454                 nlmsg_free(rep);
455                 goto out;
456         }
457         err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid,
458                               MSG_DONTWAIT);
459         if (err > 0)
460                 err = 0;
461
462 out:
463         if (sk)
464                 sock_gen_put(sk);
465
466         return err;
467 }
468 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
469
470 static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb,
471                                const struct nlmsghdr *nlh,
472                                const struct inet_diag_req_v2 *req)
473 {
474         const struct inet_diag_handler *handler;
475         int err;
476
477         handler = inet_diag_lock_handler(req->sdiag_protocol);
478         if (IS_ERR(handler))
479                 err = PTR_ERR(handler);
480         else if (cmd == SOCK_DIAG_BY_FAMILY)
481                 err = handler->dump_one(in_skb, nlh, req);
482         else if (cmd == SOCK_DESTROY && handler->destroy)
483                 err = handler->destroy(in_skb, req);
484         else
485                 err = -EOPNOTSUPP;
486         inet_diag_unlock_handler(handler);
487
488         return err;
489 }
490
491 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
492 {
493         int words = bits >> 5;
494
495         bits &= 0x1f;
496
497         if (words) {
498                 if (memcmp(a1, a2, words << 2))
499                         return 0;
500         }
501         if (bits) {
502                 __be32 w1, w2;
503                 __be32 mask;
504
505                 w1 = a1[words];
506                 w2 = a2[words];
507
508                 mask = htonl((0xffffffff) << (32 - bits));
509
510                 if ((w1 ^ w2) & mask)
511                         return 0;
512         }
513
514         return 1;
515 }
516
517 static int inet_diag_bc_run(const struct nlattr *_bc,
518                             const struct inet_diag_entry *entry)
519 {
520         const void *bc = nla_data(_bc);
521         int len = nla_len(_bc);
522
523         while (len > 0) {
524                 int yes = 1;
525                 const struct inet_diag_bc_op *op = bc;
526
527                 switch (op->code) {
528                 case INET_DIAG_BC_NOP:
529                         break;
530                 case INET_DIAG_BC_JMP:
531                         yes = 0;
532                         break;
533                 case INET_DIAG_BC_S_GE:
534                         yes = entry->sport >= op[1].no;
535                         break;
536                 case INET_DIAG_BC_S_LE:
537                         yes = entry->sport <= op[1].no;
538                         break;
539                 case INET_DIAG_BC_D_GE:
540                         yes = entry->dport >= op[1].no;
541                         break;
542                 case INET_DIAG_BC_D_LE:
543                         yes = entry->dport <= op[1].no;
544                         break;
545                 case INET_DIAG_BC_AUTO:
546                         yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
547                         break;
548                 case INET_DIAG_BC_S_COND:
549                 case INET_DIAG_BC_D_COND: {
550                         const struct inet_diag_hostcond *cond;
551                         const __be32 *addr;
552
553                         cond = (const struct inet_diag_hostcond *)(op + 1);
554                         if (cond->port != -1 &&
555                             cond->port != (op->code == INET_DIAG_BC_S_COND ?
556                                              entry->sport : entry->dport)) {
557                                 yes = 0;
558                                 break;
559                         }
560
561                         if (op->code == INET_DIAG_BC_S_COND)
562                                 addr = entry->saddr;
563                         else
564                                 addr = entry->daddr;
565
566                         if (cond->family != AF_UNSPEC &&
567                             cond->family != entry->family) {
568                                 if (entry->family == AF_INET6 &&
569                                     cond->family == AF_INET) {
570                                         if (addr[0] == 0 && addr[1] == 0 &&
571                                             addr[2] == htonl(0xffff) &&
572                                             bitstring_match(addr + 3,
573                                                             cond->addr,
574                                                             cond->prefix_len))
575                                                 break;
576                                 }
577                                 yes = 0;
578                                 break;
579                         }
580
581                         if (cond->prefix_len == 0)
582                                 break;
583                         if (bitstring_match(addr, cond->addr,
584                                             cond->prefix_len))
585                                 break;
586                         yes = 0;
587                         break;
588                 }
589                 case INET_DIAG_BC_DEV_COND: {
590                         u32 ifindex;
591
592                         ifindex = *((const u32 *)(op + 1));
593                         if (ifindex != entry->ifindex)
594                                 yes = 0;
595                         break;
596                 }
597                 case INET_DIAG_BC_MARK_COND: {
598                         struct inet_diag_markcond *cond;
599
600                         cond = (struct inet_diag_markcond *)(op + 1);
601                         if ((entry->mark & cond->mask) != cond->mark)
602                                 yes = 0;
603                         break;
604                 }
605                 }
606
607                 if (yes) {
608                         len -= op->yes;
609                         bc += op->yes;
610                 } else {
611                         len -= op->no;
612                         bc += op->no;
613                 }
614         }
615         return len == 0;
616 }
617
618 /* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV)
619  */
620 static void entry_fill_addrs(struct inet_diag_entry *entry,
621                              const struct sock *sk)
622 {
623 #if IS_ENABLED(CONFIG_IPV6)
624         if (sk->sk_family == AF_INET6) {
625                 entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32;
626                 entry->daddr = sk->sk_v6_daddr.s6_addr32;
627         } else
628 #endif
629         {
630                 entry->saddr = &sk->sk_rcv_saddr;
631                 entry->daddr = &sk->sk_daddr;
632         }
633 }
634
635 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
636 {
637         struct inet_sock *inet = inet_sk(sk);
638         struct inet_diag_entry entry;
639
640         if (!bc)
641                 return 1;
642
643         entry.family = sk->sk_family;
644         entry_fill_addrs(&entry, sk);
645         entry.sport = inet->inet_num;
646         entry.dport = ntohs(inet->inet_dport);
647         entry.ifindex = sk->sk_bound_dev_if;
648         entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0;
649         if (sk_fullsock(sk))
650                 entry.mark = sk->sk_mark;
651         else if (sk->sk_state == TCP_NEW_SYN_RECV)
652                 entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark;
653         else
654                 entry.mark = 0;
655
656         return inet_diag_bc_run(bc, &entry);
657 }
658 EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
659
660 static int valid_cc(const void *bc, int len, int cc)
661 {
662         while (len >= 0) {
663                 const struct inet_diag_bc_op *op = bc;
664
665                 if (cc > len)
666                         return 0;
667                 if (cc == len)
668                         return 1;
669                 if (op->yes < 4 || op->yes & 3)
670                         return 0;
671                 len -= op->yes;
672                 bc  += op->yes;
673         }
674         return 0;
675 }
676
677 /* data is u32 ifindex */
678 static bool valid_devcond(const struct inet_diag_bc_op *op, int len,
679                           int *min_len)
680 {
681         /* Check ifindex space. */
682         *min_len += sizeof(u32);
683         if (len < *min_len)
684                 return false;
685
686         return true;
687 }
688 /* Validate an inet_diag_hostcond. */
689 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
690                            int *min_len)
691 {
692         struct inet_diag_hostcond *cond;
693         int addr_len;
694
695         /* Check hostcond space. */
696         *min_len += sizeof(struct inet_diag_hostcond);
697         if (len < *min_len)
698                 return false;
699         cond = (struct inet_diag_hostcond *)(op + 1);
700
701         /* Check address family and address length. */
702         switch (cond->family) {
703         case AF_UNSPEC:
704                 addr_len = 0;
705                 break;
706         case AF_INET:
707                 addr_len = sizeof(struct in_addr);
708                 break;
709         case AF_INET6:
710                 addr_len = sizeof(struct in6_addr);
711                 break;
712         default:
713                 return false;
714         }
715         *min_len += addr_len;
716         if (len < *min_len)
717                 return false;
718
719         /* Check prefix length (in bits) vs address length (in bytes). */
720         if (cond->prefix_len > 8 * addr_len)
721                 return false;
722
723         return true;
724 }
725
726 /* Validate a port comparison operator. */
727 static bool valid_port_comparison(const struct inet_diag_bc_op *op,
728                                   int len, int *min_len)
729 {
730         /* Port comparisons put the port in a follow-on inet_diag_bc_op. */
731         *min_len += sizeof(struct inet_diag_bc_op);
732         if (len < *min_len)
733                 return false;
734         return true;
735 }
736
737 static bool valid_markcond(const struct inet_diag_bc_op *op, int len,
738                            int *min_len)
739 {
740         *min_len += sizeof(struct inet_diag_markcond);
741         return len >= *min_len;
742 }
743
744 static int inet_diag_bc_audit(const struct nlattr *attr,
745                               const struct sk_buff *skb)
746 {
747         bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN);
748         const void *bytecode, *bc;
749         int bytecode_len, len;
750
751         if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op))
752                 return -EINVAL;
753
754         bytecode = bc = nla_data(attr);
755         len = bytecode_len = nla_len(attr);
756
757         while (len > 0) {
758                 int min_len = sizeof(struct inet_diag_bc_op);
759                 const struct inet_diag_bc_op *op = bc;
760
761                 switch (op->code) {
762                 case INET_DIAG_BC_S_COND:
763                 case INET_DIAG_BC_D_COND:
764                         if (!valid_hostcond(bc, len, &min_len))
765                                 return -EINVAL;
766                         break;
767                 case INET_DIAG_BC_DEV_COND:
768                         if (!valid_devcond(bc, len, &min_len))
769                                 return -EINVAL;
770                         break;
771                 case INET_DIAG_BC_S_GE:
772                 case INET_DIAG_BC_S_LE:
773                 case INET_DIAG_BC_D_GE:
774                 case INET_DIAG_BC_D_LE:
775                         if (!valid_port_comparison(bc, len, &min_len))
776                                 return -EINVAL;
777                         break;
778                 case INET_DIAG_BC_MARK_COND:
779                         if (!net_admin)
780                                 return -EPERM;
781                         if (!valid_markcond(bc, len, &min_len))
782                                 return -EINVAL;
783                         break;
784                 case INET_DIAG_BC_AUTO:
785                 case INET_DIAG_BC_JMP:
786                 case INET_DIAG_BC_NOP:
787                         break;
788                 default:
789                         return -EINVAL;
790                 }
791
792                 if (op->code != INET_DIAG_BC_NOP) {
793                         if (op->no < min_len || op->no > len + 4 || op->no & 3)
794                                 return -EINVAL;
795                         if (op->no < len &&
796                             !valid_cc(bytecode, bytecode_len, len - op->no))
797                                 return -EINVAL;
798                 }
799
800                 if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
801                         return -EINVAL;
802                 bc  += op->yes;
803                 len -= op->yes;
804         }
805         return len == 0 ? 0 : -EINVAL;
806 }
807
808 static int inet_csk_diag_dump(struct sock *sk,
809                               struct sk_buff *skb,
810                               struct netlink_callback *cb,
811                               const struct inet_diag_req_v2 *r,
812                               const struct nlattr *bc,
813                               bool net_admin)
814 {
815         if (!inet_diag_bc_sk(bc, sk))
816                 return 0;
817
818         return inet_csk_diag_fill(sk, skb, r,
819                                   sk_user_ns(NETLINK_CB(cb->skb).sk),
820                                   NETLINK_CB(cb->skb).portid,
821                                   cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh,
822                                   net_admin);
823 }
824
825 static void twsk_build_assert(void)
826 {
827         BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) !=
828                      offsetof(struct sock, sk_family));
829
830         BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) !=
831                      offsetof(struct inet_sock, inet_num));
832
833         BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) !=
834                      offsetof(struct inet_sock, inet_dport));
835
836         BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) !=
837                      offsetof(struct inet_sock, inet_rcv_saddr));
838
839         BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) !=
840                      offsetof(struct inet_sock, inet_daddr));
841
842 #if IS_ENABLED(CONFIG_IPV6)
843         BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) !=
844                      offsetof(struct sock, sk_v6_rcv_saddr));
845
846         BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) !=
847                      offsetof(struct sock, sk_v6_daddr));
848 #endif
849 }
850
851 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
852                          struct netlink_callback *cb,
853                          const struct inet_diag_req_v2 *r, struct nlattr *bc)
854 {
855         struct net *net = sock_net(skb->sk);
856         int i, num, s_i, s_num;
857         u32 idiag_states = r->idiag_states;
858         bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN);
859
860         if (idiag_states & TCPF_SYN_RECV)
861                 idiag_states |= TCPF_NEW_SYN_RECV;
862         s_i = cb->args[1];
863         s_num = num = cb->args[2];
864
865         if (cb->args[0] == 0) {
866                 if (!(idiag_states & TCPF_LISTEN))
867                         goto skip_listen_ht;
868
869                 for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
870                         struct inet_listen_hashbucket *ilb;
871                         struct sock *sk;
872
873                         num = 0;
874                         ilb = &hashinfo->listening_hash[i];
875                         spin_lock_bh(&ilb->lock);
876                         sk_for_each(sk, &ilb->head) {
877                                 struct inet_sock *inet = inet_sk(sk);
878
879                                 if (!net_eq(sock_net(sk), net))
880                                         continue;
881
882                                 if (num < s_num) {
883                                         num++;
884                                         continue;
885                                 }
886
887                                 if (r->sdiag_family != AF_UNSPEC &&
888                                     sk->sk_family != r->sdiag_family)
889                                         goto next_listen;
890
891                                 if (r->id.idiag_sport != inet->inet_sport &&
892                                     r->id.idiag_sport)
893                                         goto next_listen;
894
895                                 if (r->id.idiag_dport ||
896                                     cb->args[3] > 0)
897                                         goto next_listen;
898
899                                 if (inet_csk_diag_dump(sk, skb, cb, r,
900                                                        bc, net_admin) < 0) {
901                                         spin_unlock_bh(&ilb->lock);
902                                         goto done;
903                                 }
904
905 next_listen:
906                                 cb->args[3] = 0;
907                                 cb->args[4] = 0;
908                                 ++num;
909                         }
910                         spin_unlock_bh(&ilb->lock);
911
912                         s_num = 0;
913                         cb->args[3] = 0;
914                         cb->args[4] = 0;
915                 }
916 skip_listen_ht:
917                 cb->args[0] = 1;
918                 s_i = num = s_num = 0;
919         }
920
921         if (!(idiag_states & ~TCPF_LISTEN))
922                 goto out;
923
924         for (i = s_i; i <= hashinfo->ehash_mask; i++) {
925                 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
926                 spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
927                 struct hlist_nulls_node *node;
928                 struct sock *sk;
929
930                 num = 0;
931
932                 if (hlist_nulls_empty(&head->chain))
933                         continue;
934
935                 if (i > s_i)
936                         s_num = 0;
937
938                 spin_lock_bh(lock);
939                 sk_nulls_for_each(sk, node, &head->chain) {
940                         int state, res;
941
942                         if (!net_eq(sock_net(sk), net))
943                                 continue;
944                         if (num < s_num)
945                                 goto next_normal;
946                         state = (sk->sk_state == TCP_TIME_WAIT) ?
947                                 inet_twsk(sk)->tw_substate : sk->sk_state;
948                         if (!(idiag_states & (1 << state)))
949                                 goto next_normal;
950                         if (r->sdiag_family != AF_UNSPEC &&
951                             sk->sk_family != r->sdiag_family)
952                                 goto next_normal;
953                         if (r->id.idiag_sport != htons(sk->sk_num) &&
954                             r->id.idiag_sport)
955                                 goto next_normal;
956                         if (r->id.idiag_dport != sk->sk_dport &&
957                             r->id.idiag_dport)
958                                 goto next_normal;
959                         twsk_build_assert();
960
961                         if (!inet_diag_bc_sk(bc, sk))
962                                 goto next_normal;
963
964                         res = sk_diag_fill(sk, skb, r,
965                                            sk_user_ns(NETLINK_CB(cb->skb).sk),
966                                            NETLINK_CB(cb->skb).portid,
967                                            cb->nlh->nlmsg_seq, NLM_F_MULTI,
968                                            cb->nlh, net_admin);
969                         if (res < 0) {
970                                 spin_unlock_bh(lock);
971                                 goto done;
972                         }
973 next_normal:
974                         ++num;
975                 }
976
977                 spin_unlock_bh(lock);
978                 cond_resched();
979         }
980
981 done:
982         cb->args[1] = i;
983         cb->args[2] = num;
984 out:
985         ;
986 }
987 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
988
989 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
990                             const struct inet_diag_req_v2 *r,
991                             struct nlattr *bc)
992 {
993         const struct inet_diag_handler *handler;
994         int err = 0;
995
996         handler = inet_diag_lock_handler(r->sdiag_protocol);
997         if (!IS_ERR(handler))
998                 handler->dump(skb, cb, r, bc);
999         else
1000                 err = PTR_ERR(handler);
1001         inet_diag_unlock_handler(handler);
1002
1003         return err ? : skb->len;
1004 }
1005
1006 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
1007 {
1008         int hdrlen = sizeof(struct inet_diag_req_v2);
1009         struct nlattr *bc = NULL;
1010
1011         if (nlmsg_attrlen(cb->nlh, hdrlen))
1012                 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE);
1013
1014         return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh), bc);
1015 }
1016
1017 static int inet_diag_type2proto(int type)
1018 {
1019         switch (type) {
1020         case TCPDIAG_GETSOCK:
1021                 return IPPROTO_TCP;
1022         case DCCPDIAG_GETSOCK:
1023                 return IPPROTO_DCCP;
1024         default:
1025                 return 0;
1026         }
1027 }
1028
1029 static int inet_diag_dump_compat(struct sk_buff *skb,
1030                                  struct netlink_callback *cb)
1031 {
1032         struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1033         int hdrlen = sizeof(struct inet_diag_req);
1034         struct inet_diag_req_v2 req;
1035         struct nlattr *bc = NULL;
1036
1037         req.sdiag_family = AF_UNSPEC; /* compatibility */
1038         req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1039         req.idiag_ext = rc->idiag_ext;
1040         req.idiag_states = rc->idiag_states;
1041         req.id = rc->id;
1042
1043         if (nlmsg_attrlen(cb->nlh, hdrlen))
1044                 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE);
1045
1046         return __inet_diag_dump(skb, cb, &req, bc);
1047 }
1048
1049 static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1050                                       const struct nlmsghdr *nlh)
1051 {
1052         struct inet_diag_req *rc = nlmsg_data(nlh);
1053         struct inet_diag_req_v2 req;
1054
1055         req.sdiag_family = rc->idiag_family;
1056         req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1057         req.idiag_ext = rc->idiag_ext;
1058         req.idiag_states = rc->idiag_states;
1059         req.id = rc->id;
1060
1061         return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh, &req);
1062 }
1063
1064 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1065 {
1066         int hdrlen = sizeof(struct inet_diag_req);
1067         struct net *net = sock_net(skb->sk);
1068
1069         if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1070             nlmsg_len(nlh) < hdrlen)
1071                 return -EINVAL;
1072
1073         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1074                 if (nlmsg_attrlen(nlh, hdrlen)) {
1075                         struct nlattr *attr;
1076                         int err;
1077
1078                         attr = nlmsg_find_attr(nlh, hdrlen,
1079                                                INET_DIAG_REQ_BYTECODE);
1080                         err = inet_diag_bc_audit(attr, skb);
1081                         if (err)
1082                                 return err;
1083                 }
1084                 {
1085                         struct netlink_dump_control c = {
1086                                 .dump = inet_diag_dump_compat,
1087                         };
1088                         return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1089                 }
1090         }
1091
1092         return inet_diag_get_exact_compat(skb, nlh);
1093 }
1094
1095 static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h)
1096 {
1097         int hdrlen = sizeof(struct inet_diag_req_v2);
1098         struct net *net = sock_net(skb->sk);
1099
1100         if (nlmsg_len(h) < hdrlen)
1101                 return -EINVAL;
1102
1103         if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY &&
1104             h->nlmsg_flags & NLM_F_DUMP) {
1105                 if (nlmsg_attrlen(h, hdrlen)) {
1106                         struct nlattr *attr;
1107                         int err;
1108
1109                         attr = nlmsg_find_attr(h, hdrlen,
1110                                                INET_DIAG_REQ_BYTECODE);
1111                         err = inet_diag_bc_audit(attr, skb);
1112                         if (err)
1113                                 return err;
1114                 }
1115                 {
1116                         struct netlink_dump_control c = {
1117                                 .dump = inet_diag_dump,
1118                         };
1119                         return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1120                 }
1121         }
1122
1123         return inet_diag_cmd_exact(h->nlmsg_type, skb, h, nlmsg_data(h));
1124 }
1125
1126 static
1127 int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk)
1128 {
1129         const struct inet_diag_handler *handler;
1130         struct nlmsghdr *nlh;
1131         struct nlattr *attr;
1132         struct inet_diag_msg *r;
1133         void *info = NULL;
1134         int err = 0;
1135
1136         nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0);
1137         if (!nlh)
1138                 return -ENOMEM;
1139
1140         r = nlmsg_data(nlh);
1141         memset(r, 0, sizeof(*r));
1142         inet_diag_msg_common_fill(r, sk);
1143         if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM)
1144                 r->id.idiag_sport = inet_sk(sk)->inet_sport;
1145         r->idiag_state = sk->sk_state;
1146
1147         if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) {
1148                 nlmsg_cancel(skb, nlh);
1149                 return err;
1150         }
1151
1152         handler = inet_diag_lock_handler(sk->sk_protocol);
1153         if (IS_ERR(handler)) {
1154                 inet_diag_unlock_handler(handler);
1155                 nlmsg_cancel(skb, nlh);
1156                 return PTR_ERR(handler);
1157         }
1158
1159         attr = handler->idiag_info_size
1160                 ? nla_reserve_64bit(skb, INET_DIAG_INFO,
1161                                     handler->idiag_info_size,
1162                                     INET_DIAG_PAD)
1163                 : NULL;
1164         if (attr)
1165                 info = nla_data(attr);
1166
1167         handler->idiag_get_info(sk, r, info);
1168         inet_diag_unlock_handler(handler);
1169
1170         nlmsg_end(skb, nlh);
1171         return 0;
1172 }
1173
1174 static const struct sock_diag_handler inet_diag_handler = {
1175         .family = AF_INET,
1176         .dump = inet_diag_handler_cmd,
1177         .get_info = inet_diag_handler_get_info,
1178         .destroy = inet_diag_handler_cmd,
1179 };
1180
1181 static const struct sock_diag_handler inet6_diag_handler = {
1182         .family = AF_INET6,
1183         .dump = inet_diag_handler_cmd,
1184         .get_info = inet_diag_handler_get_info,
1185         .destroy = inet_diag_handler_cmd,
1186 };
1187
1188 int inet_diag_register(const struct inet_diag_handler *h)
1189 {
1190         const __u16 type = h->idiag_type;
1191         int err = -EINVAL;
1192
1193         if (type >= IPPROTO_MAX)
1194                 goto out;
1195
1196         mutex_lock(&inet_diag_table_mutex);
1197         err = -EEXIST;
1198         if (!inet_diag_table[type]) {
1199                 inet_diag_table[type] = h;
1200                 err = 0;
1201         }
1202         mutex_unlock(&inet_diag_table_mutex);
1203 out:
1204         return err;
1205 }
1206 EXPORT_SYMBOL_GPL(inet_diag_register);
1207
1208 void inet_diag_unregister(const struct inet_diag_handler *h)
1209 {
1210         const __u16 type = h->idiag_type;
1211
1212         if (type >= IPPROTO_MAX)
1213                 return;
1214
1215         mutex_lock(&inet_diag_table_mutex);
1216         inet_diag_table[type] = NULL;
1217         mutex_unlock(&inet_diag_table_mutex);
1218 }
1219 EXPORT_SYMBOL_GPL(inet_diag_unregister);
1220
1221 static int __init inet_diag_init(void)
1222 {
1223         const int inet_diag_table_size = (IPPROTO_MAX *
1224                                           sizeof(struct inet_diag_handler *));
1225         int err = -ENOMEM;
1226
1227         inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1228         if (!inet_diag_table)
1229                 goto out;
1230
1231         err = sock_diag_register(&inet_diag_handler);
1232         if (err)
1233                 goto out_free_nl;
1234
1235         err = sock_diag_register(&inet6_diag_handler);
1236         if (err)
1237                 goto out_free_inet;
1238
1239         sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1240 out:
1241         return err;
1242
1243 out_free_inet:
1244         sock_diag_unregister(&inet_diag_handler);
1245 out_free_nl:
1246         kfree(inet_diag_table);
1247         goto out;
1248 }
1249
1250 static void __exit inet_diag_exit(void)
1251 {
1252         sock_diag_unregister(&inet6_diag_handler);
1253         sock_diag_unregister(&inet_diag_handler);
1254         sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1255         kfree(inet_diag_table);
1256 }
1257
1258 module_init(inet_diag_init);
1259 module_exit(inet_diag_exit);
1260 MODULE_LICENSE("GPL");
1261 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1262 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);