Loading...
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
38static const struct inet_diag_handler **inet_diag_table;
39
40struct 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
51static DEFINE_MUTEX(inet_diag_table_mutex);
52
53static const struct inet_diag_handler *inet_diag_lock_handler(int proto)
54{
55 if (!inet_diag_table[proto])
56 sock_load_diag_module(AF_INET, proto);
57
58 mutex_lock(&inet_diag_table_mutex);
59 if (!inet_diag_table[proto])
60 return ERR_PTR(-ENOENT);
61
62 return inet_diag_table[proto];
63}
64
65static void inet_diag_unlock_handler(const struct inet_diag_handler *handler)
66{
67 mutex_unlock(&inet_diag_table_mutex);
68}
69
70void inet_diag_msg_common_fill(struct inet_diag_msg *r, struct sock *sk)
71{
72 r->idiag_family = sk->sk_family;
73
74 r->id.idiag_sport = htons(sk->sk_num);
75 r->id.idiag_dport = sk->sk_dport;
76 r->id.idiag_if = sk->sk_bound_dev_if;
77 sock_diag_save_cookie(sk, r->id.idiag_cookie);
78
79#if IS_ENABLED(CONFIG_IPV6)
80 if (sk->sk_family == AF_INET6) {
81 *(struct in6_addr *)r->id.idiag_src = sk->sk_v6_rcv_saddr;
82 *(struct in6_addr *)r->id.idiag_dst = sk->sk_v6_daddr;
83 } else
84#endif
85 {
86 memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src));
87 memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst));
88
89 r->id.idiag_src[0] = sk->sk_rcv_saddr;
90 r->id.idiag_dst[0] = sk->sk_daddr;
91 }
92}
93EXPORT_SYMBOL_GPL(inet_diag_msg_common_fill);
94
95static size_t inet_sk_attr_size(struct sock *sk,
96 const struct inet_diag_req_v2 *req,
97 bool net_admin)
98{
99 const struct inet_diag_handler *handler;
100 size_t aux = 0;
101
102 handler = inet_diag_table[req->sdiag_protocol];
103 if (handler && handler->idiag_get_aux_size)
104 aux = handler->idiag_get_aux_size(sk, net_admin);
105
106 return nla_total_size(sizeof(struct tcp_info))
107 + nla_total_size(1) /* INET_DIAG_SHUTDOWN */
108 + nla_total_size(1) /* INET_DIAG_TOS */
109 + nla_total_size(1) /* INET_DIAG_TCLASS */
110 + nla_total_size(4) /* INET_DIAG_MARK */
111 + nla_total_size(sizeof(struct inet_diag_meminfo))
112 + nla_total_size(sizeof(struct inet_diag_msg))
113 + nla_total_size(SK_MEMINFO_VARS * sizeof(u32))
114 + nla_total_size(TCP_CA_NAME_MAX)
115 + nla_total_size(sizeof(struct tcpvegas_info))
116 + aux
117 + 64;
118}
119
120int inet_diag_msg_attrs_fill(struct sock *sk, struct sk_buff *skb,
121 struct inet_diag_msg *r, int ext,
122 struct user_namespace *user_ns,
123 bool net_admin)
124{
125 const struct inet_sock *inet = inet_sk(sk);
126
127 if (nla_put_u8(skb, INET_DIAG_SHUTDOWN, sk->sk_shutdown))
128 goto errout;
129
130 /* IPv6 dual-stack sockets use inet->tos for IPv4 connections,
131 * hence this needs to be included regardless of socket family.
132 */
133 if (ext & (1 << (INET_DIAG_TOS - 1)))
134 if (nla_put_u8(skb, INET_DIAG_TOS, inet->tos) < 0)
135 goto errout;
136
137#if IS_ENABLED(CONFIG_IPV6)
138 if (r->idiag_family == AF_INET6) {
139 if (ext & (1 << (INET_DIAG_TCLASS - 1)))
140 if (nla_put_u8(skb, INET_DIAG_TCLASS,
141 inet6_sk(sk)->tclass) < 0)
142 goto errout;
143
144 if (((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
145 nla_put_u8(skb, INET_DIAG_SKV6ONLY, ipv6_only_sock(sk)))
146 goto errout;
147 }
148#endif
149
150 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK, sk->sk_mark))
151 goto errout;
152
153 r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk));
154 r->idiag_inode = sock_i_ino(sk);
155
156 return 0;
157errout:
158 return 1;
159}
160EXPORT_SYMBOL_GPL(inet_diag_msg_attrs_fill);
161
162int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk,
163 struct sk_buff *skb, const struct inet_diag_req_v2 *req,
164 struct user_namespace *user_ns,
165 u32 portid, u32 seq, u16 nlmsg_flags,
166 const struct nlmsghdr *unlh,
167 bool net_admin)
168{
169 const struct tcp_congestion_ops *ca_ops;
170 const struct inet_diag_handler *handler;
171 int ext = req->idiag_ext;
172 struct inet_diag_msg *r;
173 struct nlmsghdr *nlh;
174 struct nlattr *attr;
175 void *info = NULL;
176
177 handler = inet_diag_table[req->sdiag_protocol];
178 BUG_ON(!handler);
179
180 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
181 nlmsg_flags);
182 if (!nlh)
183 return -EMSGSIZE;
184
185 r = nlmsg_data(nlh);
186 BUG_ON(!sk_fullsock(sk));
187
188 inet_diag_msg_common_fill(r, sk);
189 r->idiag_state = sk->sk_state;
190 r->idiag_timer = 0;
191 r->idiag_retrans = 0;
192
193 if (inet_diag_msg_attrs_fill(sk, skb, r, ext, user_ns, net_admin))
194 goto errout;
195
196 if (ext & (1 << (INET_DIAG_MEMINFO - 1))) {
197 struct inet_diag_meminfo minfo = {
198 .idiag_rmem = sk_rmem_alloc_get(sk),
199 .idiag_wmem = sk->sk_wmem_queued,
200 .idiag_fmem = sk->sk_forward_alloc,
201 .idiag_tmem = sk_wmem_alloc_get(sk),
202 };
203
204 if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0)
205 goto errout;
206 }
207
208 if (ext & (1 << (INET_DIAG_SKMEMINFO - 1)))
209 if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO))
210 goto errout;
211
212 /*
213 * RAW sockets might have user-defined protocols assigned,
214 * so report the one supplied on socket creation.
215 */
216 if (sk->sk_type == SOCK_RAW) {
217 if (nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))
218 goto errout;
219 }
220
221 if (!icsk) {
222 handler->idiag_get_info(sk, r, NULL);
223 goto out;
224 }
225
226 if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
227 icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
228 icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
229 r->idiag_timer = 1;
230 r->idiag_retrans = icsk->icsk_retransmits;
231 r->idiag_expires =
232 jiffies_to_msecs(icsk->icsk_timeout - jiffies);
233 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
234 r->idiag_timer = 4;
235 r->idiag_retrans = icsk->icsk_probes_out;
236 r->idiag_expires =
237 jiffies_to_msecs(icsk->icsk_timeout - jiffies);
238 } else if (timer_pending(&sk->sk_timer)) {
239 r->idiag_timer = 2;
240 r->idiag_retrans = icsk->icsk_probes_out;
241 r->idiag_expires =
242 jiffies_to_msecs(sk->sk_timer.expires - jiffies);
243 } else {
244 r->idiag_timer = 0;
245 r->idiag_expires = 0;
246 }
247
248 if ((ext & (1 << (INET_DIAG_INFO - 1))) && handler->idiag_info_size) {
249 attr = nla_reserve_64bit(skb, INET_DIAG_INFO,
250 handler->idiag_info_size,
251 INET_DIAG_PAD);
252 if (!attr)
253 goto errout;
254
255 info = nla_data(attr);
256 }
257
258 if (ext & (1 << (INET_DIAG_CONG - 1))) {
259 int err = 0;
260
261 rcu_read_lock();
262 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
263 if (ca_ops)
264 err = nla_put_string(skb, INET_DIAG_CONG, ca_ops->name);
265 rcu_read_unlock();
266 if (err < 0)
267 goto errout;
268 }
269
270 handler->idiag_get_info(sk, r, info);
271
272 if (ext & (1 << (INET_DIAG_INFO - 1)) && handler->idiag_get_aux)
273 if (handler->idiag_get_aux(sk, net_admin, skb) < 0)
274 goto errout;
275
276 if (sk->sk_state < TCP_TIME_WAIT) {
277 union tcp_cc_info info;
278 size_t sz = 0;
279 int attr;
280
281 rcu_read_lock();
282 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
283 if (ca_ops && ca_ops->get_info)
284 sz = ca_ops->get_info(sk, ext, &attr, &info);
285 rcu_read_unlock();
286 if (sz && nla_put(skb, attr, sz, &info) < 0)
287 goto errout;
288 }
289
290 if (ext & (1 << (INET_DIAG_CLASS_ID - 1))) {
291 u32 classid = 0;
292
293#ifdef CONFIG_SOCK_CGROUP_DATA
294 classid = sock_cgroup_classid(&sk->sk_cgrp_data);
295#endif
296
297 if (nla_put_u32(skb, INET_DIAG_CLASS_ID, classid))
298 goto errout;
299 }
300
301out:
302 nlmsg_end(skb, nlh);
303 return 0;
304
305errout:
306 nlmsg_cancel(skb, nlh);
307 return -EMSGSIZE;
308}
309EXPORT_SYMBOL_GPL(inet_sk_diag_fill);
310
311static int inet_csk_diag_fill(struct sock *sk,
312 struct sk_buff *skb,
313 const struct inet_diag_req_v2 *req,
314 struct user_namespace *user_ns,
315 u32 portid, u32 seq, u16 nlmsg_flags,
316 const struct nlmsghdr *unlh,
317 bool net_admin)
318{
319 return inet_sk_diag_fill(sk, inet_csk(sk), skb, req, user_ns,
320 portid, seq, nlmsg_flags, unlh, net_admin);
321}
322
323static int inet_twsk_diag_fill(struct sock *sk,
324 struct sk_buff *skb,
325 u32 portid, u32 seq, u16 nlmsg_flags,
326 const struct nlmsghdr *unlh)
327{
328 struct inet_timewait_sock *tw = inet_twsk(sk);
329 struct inet_diag_msg *r;
330 struct nlmsghdr *nlh;
331 long tmo;
332
333 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
334 nlmsg_flags);
335 if (!nlh)
336 return -EMSGSIZE;
337
338 r = nlmsg_data(nlh);
339 BUG_ON(tw->tw_state != TCP_TIME_WAIT);
340
341 tmo = tw->tw_timer.expires - jiffies;
342 if (tmo < 0)
343 tmo = 0;
344
345 inet_diag_msg_common_fill(r, sk);
346 r->idiag_retrans = 0;
347
348 r->idiag_state = tw->tw_substate;
349 r->idiag_timer = 3;
350 r->idiag_expires = jiffies_to_msecs(tmo);
351 r->idiag_rqueue = 0;
352 r->idiag_wqueue = 0;
353 r->idiag_uid = 0;
354 r->idiag_inode = 0;
355
356 nlmsg_end(skb, nlh);
357 return 0;
358}
359
360static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb,
361 u32 portid, u32 seq, u16 nlmsg_flags,
362 const struct nlmsghdr *unlh, bool net_admin)
363{
364 struct request_sock *reqsk = inet_reqsk(sk);
365 struct inet_diag_msg *r;
366 struct nlmsghdr *nlh;
367 long tmo;
368
369 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
370 nlmsg_flags);
371 if (!nlh)
372 return -EMSGSIZE;
373
374 r = nlmsg_data(nlh);
375 inet_diag_msg_common_fill(r, sk);
376 r->idiag_state = TCP_SYN_RECV;
377 r->idiag_timer = 1;
378 r->idiag_retrans = reqsk->num_retrans;
379
380 BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) !=
381 offsetof(struct sock, sk_cookie));
382
383 tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies;
384 r->idiag_expires = (tmo >= 0) ? jiffies_to_msecs(tmo) : 0;
385 r->idiag_rqueue = 0;
386 r->idiag_wqueue = 0;
387 r->idiag_uid = 0;
388 r->idiag_inode = 0;
389
390 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
391 inet_rsk(reqsk)->ir_mark))
392 return -EMSGSIZE;
393
394 nlmsg_end(skb, nlh);
395 return 0;
396}
397
398static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
399 const struct inet_diag_req_v2 *r,
400 struct user_namespace *user_ns,
401 u32 portid, u32 seq, u16 nlmsg_flags,
402 const struct nlmsghdr *unlh, bool net_admin)
403{
404 if (sk->sk_state == TCP_TIME_WAIT)
405 return inet_twsk_diag_fill(sk, skb, portid, seq,
406 nlmsg_flags, unlh);
407
408 if (sk->sk_state == TCP_NEW_SYN_RECV)
409 return inet_req_diag_fill(sk, skb, portid, seq,
410 nlmsg_flags, unlh, net_admin);
411
412 return inet_csk_diag_fill(sk, skb, r, user_ns, portid, seq,
413 nlmsg_flags, unlh, net_admin);
414}
415
416struct sock *inet_diag_find_one_icsk(struct net *net,
417 struct inet_hashinfo *hashinfo,
418 const struct inet_diag_req_v2 *req)
419{
420 struct sock *sk;
421
422 rcu_read_lock();
423 if (req->sdiag_family == AF_INET)
424 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0],
425 req->id.idiag_dport, req->id.idiag_src[0],
426 req->id.idiag_sport, req->id.idiag_if);
427#if IS_ENABLED(CONFIG_IPV6)
428 else if (req->sdiag_family == AF_INET6) {
429 if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) &&
430 ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src))
431 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3],
432 req->id.idiag_dport, req->id.idiag_src[3],
433 req->id.idiag_sport, req->id.idiag_if);
434 else
435 sk = inet6_lookup(net, hashinfo, NULL, 0,
436 (struct in6_addr *)req->id.idiag_dst,
437 req->id.idiag_dport,
438 (struct in6_addr *)req->id.idiag_src,
439 req->id.idiag_sport,
440 req->id.idiag_if);
441 }
442#endif
443 else {
444 rcu_read_unlock();
445 return ERR_PTR(-EINVAL);
446 }
447 rcu_read_unlock();
448 if (!sk)
449 return ERR_PTR(-ENOENT);
450
451 if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) {
452 sock_gen_put(sk);
453 return ERR_PTR(-ENOENT);
454 }
455
456 return sk;
457}
458EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk);
459
460int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo,
461 struct sk_buff *in_skb,
462 const struct nlmsghdr *nlh,
463 const struct inet_diag_req_v2 *req)
464{
465 bool net_admin = netlink_net_capable(in_skb, CAP_NET_ADMIN);
466 struct net *net = sock_net(in_skb->sk);
467 struct sk_buff *rep;
468 struct sock *sk;
469 int err;
470
471 sk = inet_diag_find_one_icsk(net, hashinfo, req);
472 if (IS_ERR(sk))
473 return PTR_ERR(sk);
474
475 rep = nlmsg_new(inet_sk_attr_size(sk, req, net_admin), GFP_KERNEL);
476 if (!rep) {
477 err = -ENOMEM;
478 goto out;
479 }
480
481 err = sk_diag_fill(sk, rep, req,
482 sk_user_ns(NETLINK_CB(in_skb).sk),
483 NETLINK_CB(in_skb).portid,
484 nlh->nlmsg_seq, 0, nlh, net_admin);
485 if (err < 0) {
486 WARN_ON(err == -EMSGSIZE);
487 nlmsg_free(rep);
488 goto out;
489 }
490 err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid,
491 MSG_DONTWAIT);
492 if (err > 0)
493 err = 0;
494
495out:
496 if (sk)
497 sock_gen_put(sk);
498
499 return err;
500}
501EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
502
503static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb,
504 const struct nlmsghdr *nlh,
505 const struct inet_diag_req_v2 *req)
506{
507 const struct inet_diag_handler *handler;
508 int err;
509
510 handler = inet_diag_lock_handler(req->sdiag_protocol);
511 if (IS_ERR(handler))
512 err = PTR_ERR(handler);
513 else if (cmd == SOCK_DIAG_BY_FAMILY)
514 err = handler->dump_one(in_skb, nlh, req);
515 else if (cmd == SOCK_DESTROY && handler->destroy)
516 err = handler->destroy(in_skb, req);
517 else
518 err = -EOPNOTSUPP;
519 inet_diag_unlock_handler(handler);
520
521 return err;
522}
523
524static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
525{
526 int words = bits >> 5;
527
528 bits &= 0x1f;
529
530 if (words) {
531 if (memcmp(a1, a2, words << 2))
532 return 0;
533 }
534 if (bits) {
535 __be32 w1, w2;
536 __be32 mask;
537
538 w1 = a1[words];
539 w2 = a2[words];
540
541 mask = htonl((0xffffffff) << (32 - bits));
542
543 if ((w1 ^ w2) & mask)
544 return 0;
545 }
546
547 return 1;
548}
549
550static int inet_diag_bc_run(const struct nlattr *_bc,
551 const struct inet_diag_entry *entry)
552{
553 const void *bc = nla_data(_bc);
554 int len = nla_len(_bc);
555
556 while (len > 0) {
557 int yes = 1;
558 const struct inet_diag_bc_op *op = bc;
559
560 switch (op->code) {
561 case INET_DIAG_BC_NOP:
562 break;
563 case INET_DIAG_BC_JMP:
564 yes = 0;
565 break;
566 case INET_DIAG_BC_S_EQ:
567 yes = entry->sport == op[1].no;
568 break;
569 case INET_DIAG_BC_S_GE:
570 yes = entry->sport >= op[1].no;
571 break;
572 case INET_DIAG_BC_S_LE:
573 yes = entry->sport <= op[1].no;
574 break;
575 case INET_DIAG_BC_D_EQ:
576 yes = entry->dport == op[1].no;
577 break;
578 case INET_DIAG_BC_D_GE:
579 yes = entry->dport >= op[1].no;
580 break;
581 case INET_DIAG_BC_D_LE:
582 yes = entry->dport <= op[1].no;
583 break;
584 case INET_DIAG_BC_AUTO:
585 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
586 break;
587 case INET_DIAG_BC_S_COND:
588 case INET_DIAG_BC_D_COND: {
589 const struct inet_diag_hostcond *cond;
590 const __be32 *addr;
591
592 cond = (const struct inet_diag_hostcond *)(op + 1);
593 if (cond->port != -1 &&
594 cond->port != (op->code == INET_DIAG_BC_S_COND ?
595 entry->sport : entry->dport)) {
596 yes = 0;
597 break;
598 }
599
600 if (op->code == INET_DIAG_BC_S_COND)
601 addr = entry->saddr;
602 else
603 addr = entry->daddr;
604
605 if (cond->family != AF_UNSPEC &&
606 cond->family != entry->family) {
607 if (entry->family == AF_INET6 &&
608 cond->family == AF_INET) {
609 if (addr[0] == 0 && addr[1] == 0 &&
610 addr[2] == htonl(0xffff) &&
611 bitstring_match(addr + 3,
612 cond->addr,
613 cond->prefix_len))
614 break;
615 }
616 yes = 0;
617 break;
618 }
619
620 if (cond->prefix_len == 0)
621 break;
622 if (bitstring_match(addr, cond->addr,
623 cond->prefix_len))
624 break;
625 yes = 0;
626 break;
627 }
628 case INET_DIAG_BC_DEV_COND: {
629 u32 ifindex;
630
631 ifindex = *((const u32 *)(op + 1));
632 if (ifindex != entry->ifindex)
633 yes = 0;
634 break;
635 }
636 case INET_DIAG_BC_MARK_COND: {
637 struct inet_diag_markcond *cond;
638
639 cond = (struct inet_diag_markcond *)(op + 1);
640 if ((entry->mark & cond->mask) != cond->mark)
641 yes = 0;
642 break;
643 }
644 }
645
646 if (yes) {
647 len -= op->yes;
648 bc += op->yes;
649 } else {
650 len -= op->no;
651 bc += op->no;
652 }
653 }
654 return len == 0;
655}
656
657/* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV)
658 */
659static void entry_fill_addrs(struct inet_diag_entry *entry,
660 const struct sock *sk)
661{
662#if IS_ENABLED(CONFIG_IPV6)
663 if (sk->sk_family == AF_INET6) {
664 entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32;
665 entry->daddr = sk->sk_v6_daddr.s6_addr32;
666 } else
667#endif
668 {
669 entry->saddr = &sk->sk_rcv_saddr;
670 entry->daddr = &sk->sk_daddr;
671 }
672}
673
674int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
675{
676 struct inet_sock *inet = inet_sk(sk);
677 struct inet_diag_entry entry;
678
679 if (!bc)
680 return 1;
681
682 entry.family = sk->sk_family;
683 entry_fill_addrs(&entry, sk);
684 entry.sport = inet->inet_num;
685 entry.dport = ntohs(inet->inet_dport);
686 entry.ifindex = sk->sk_bound_dev_if;
687 entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0;
688 if (sk_fullsock(sk))
689 entry.mark = sk->sk_mark;
690 else if (sk->sk_state == TCP_NEW_SYN_RECV)
691 entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark;
692 else
693 entry.mark = 0;
694
695 return inet_diag_bc_run(bc, &entry);
696}
697EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
698
699static int valid_cc(const void *bc, int len, int cc)
700{
701 while (len >= 0) {
702 const struct inet_diag_bc_op *op = bc;
703
704 if (cc > len)
705 return 0;
706 if (cc == len)
707 return 1;
708 if (op->yes < 4 || op->yes & 3)
709 return 0;
710 len -= op->yes;
711 bc += op->yes;
712 }
713 return 0;
714}
715
716/* data is u32 ifindex */
717static bool valid_devcond(const struct inet_diag_bc_op *op, int len,
718 int *min_len)
719{
720 /* Check ifindex space. */
721 *min_len += sizeof(u32);
722 if (len < *min_len)
723 return false;
724
725 return true;
726}
727/* Validate an inet_diag_hostcond. */
728static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
729 int *min_len)
730{
731 struct inet_diag_hostcond *cond;
732 int addr_len;
733
734 /* Check hostcond space. */
735 *min_len += sizeof(struct inet_diag_hostcond);
736 if (len < *min_len)
737 return false;
738 cond = (struct inet_diag_hostcond *)(op + 1);
739
740 /* Check address family and address length. */
741 switch (cond->family) {
742 case AF_UNSPEC:
743 addr_len = 0;
744 break;
745 case AF_INET:
746 addr_len = sizeof(struct in_addr);
747 break;
748 case AF_INET6:
749 addr_len = sizeof(struct in6_addr);
750 break;
751 default:
752 return false;
753 }
754 *min_len += addr_len;
755 if (len < *min_len)
756 return false;
757
758 /* Check prefix length (in bits) vs address length (in bytes). */
759 if (cond->prefix_len > 8 * addr_len)
760 return false;
761
762 return true;
763}
764
765/* Validate a port comparison operator. */
766static bool valid_port_comparison(const struct inet_diag_bc_op *op,
767 int len, int *min_len)
768{
769 /* Port comparisons put the port in a follow-on inet_diag_bc_op. */
770 *min_len += sizeof(struct inet_diag_bc_op);
771 if (len < *min_len)
772 return false;
773 return true;
774}
775
776static bool valid_markcond(const struct inet_diag_bc_op *op, int len,
777 int *min_len)
778{
779 *min_len += sizeof(struct inet_diag_markcond);
780 return len >= *min_len;
781}
782
783static int inet_diag_bc_audit(const struct nlattr *attr,
784 const struct sk_buff *skb)
785{
786 bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN);
787 const void *bytecode, *bc;
788 int bytecode_len, len;
789
790 if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op))
791 return -EINVAL;
792
793 bytecode = bc = nla_data(attr);
794 len = bytecode_len = nla_len(attr);
795
796 while (len > 0) {
797 int min_len = sizeof(struct inet_diag_bc_op);
798 const struct inet_diag_bc_op *op = bc;
799
800 switch (op->code) {
801 case INET_DIAG_BC_S_COND:
802 case INET_DIAG_BC_D_COND:
803 if (!valid_hostcond(bc, len, &min_len))
804 return -EINVAL;
805 break;
806 case INET_DIAG_BC_DEV_COND:
807 if (!valid_devcond(bc, len, &min_len))
808 return -EINVAL;
809 break;
810 case INET_DIAG_BC_S_EQ:
811 case INET_DIAG_BC_S_GE:
812 case INET_DIAG_BC_S_LE:
813 case INET_DIAG_BC_D_EQ:
814 case INET_DIAG_BC_D_GE:
815 case INET_DIAG_BC_D_LE:
816 if (!valid_port_comparison(bc, len, &min_len))
817 return -EINVAL;
818 break;
819 case INET_DIAG_BC_MARK_COND:
820 if (!net_admin)
821 return -EPERM;
822 if (!valid_markcond(bc, len, &min_len))
823 return -EINVAL;
824 break;
825 case INET_DIAG_BC_AUTO:
826 case INET_DIAG_BC_JMP:
827 case INET_DIAG_BC_NOP:
828 break;
829 default:
830 return -EINVAL;
831 }
832
833 if (op->code != INET_DIAG_BC_NOP) {
834 if (op->no < min_len || op->no > len + 4 || op->no & 3)
835 return -EINVAL;
836 if (op->no < len &&
837 !valid_cc(bytecode, bytecode_len, len - op->no))
838 return -EINVAL;
839 }
840
841 if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
842 return -EINVAL;
843 bc += op->yes;
844 len -= op->yes;
845 }
846 return len == 0 ? 0 : -EINVAL;
847}
848
849static int inet_csk_diag_dump(struct sock *sk,
850 struct sk_buff *skb,
851 struct netlink_callback *cb,
852 const struct inet_diag_req_v2 *r,
853 const struct nlattr *bc,
854 bool net_admin)
855{
856 if (!inet_diag_bc_sk(bc, sk))
857 return 0;
858
859 return inet_csk_diag_fill(sk, skb, r,
860 sk_user_ns(NETLINK_CB(cb->skb).sk),
861 NETLINK_CB(cb->skb).portid,
862 cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh,
863 net_admin);
864}
865
866static void twsk_build_assert(void)
867{
868 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) !=
869 offsetof(struct sock, sk_family));
870
871 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) !=
872 offsetof(struct inet_sock, inet_num));
873
874 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) !=
875 offsetof(struct inet_sock, inet_dport));
876
877 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) !=
878 offsetof(struct inet_sock, inet_rcv_saddr));
879
880 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) !=
881 offsetof(struct inet_sock, inet_daddr));
882
883#if IS_ENABLED(CONFIG_IPV6)
884 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) !=
885 offsetof(struct sock, sk_v6_rcv_saddr));
886
887 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) !=
888 offsetof(struct sock, sk_v6_daddr));
889#endif
890}
891
892void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
893 struct netlink_callback *cb,
894 const struct inet_diag_req_v2 *r, struct nlattr *bc)
895{
896 bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN);
897 struct net *net = sock_net(skb->sk);
898 u32 idiag_states = r->idiag_states;
899 int i, num, s_i, s_num;
900 struct sock *sk;
901
902 if (idiag_states & TCPF_SYN_RECV)
903 idiag_states |= TCPF_NEW_SYN_RECV;
904 s_i = cb->args[1];
905 s_num = num = cb->args[2];
906
907 if (cb->args[0] == 0) {
908 if (!(idiag_states & TCPF_LISTEN) || r->id.idiag_dport)
909 goto skip_listen_ht;
910
911 for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
912 struct inet_listen_hashbucket *ilb;
913
914 num = 0;
915 ilb = &hashinfo->listening_hash[i];
916 spin_lock(&ilb->lock);
917 sk_for_each(sk, &ilb->head) {
918 struct inet_sock *inet = inet_sk(sk);
919
920 if (!net_eq(sock_net(sk), net))
921 continue;
922
923 if (num < s_num) {
924 num++;
925 continue;
926 }
927
928 if (r->sdiag_family != AF_UNSPEC &&
929 sk->sk_family != r->sdiag_family)
930 goto next_listen;
931
932 if (r->id.idiag_sport != inet->inet_sport &&
933 r->id.idiag_sport)
934 goto next_listen;
935
936 if (inet_csk_diag_dump(sk, skb, cb, r,
937 bc, net_admin) < 0) {
938 spin_unlock(&ilb->lock);
939 goto done;
940 }
941
942next_listen:
943 ++num;
944 }
945 spin_unlock(&ilb->lock);
946
947 s_num = 0;
948 }
949skip_listen_ht:
950 cb->args[0] = 1;
951 s_i = num = s_num = 0;
952 }
953
954 if (!(idiag_states & ~TCPF_LISTEN))
955 goto out;
956
957#define SKARR_SZ 16
958 for (i = s_i; i <= hashinfo->ehash_mask; i++) {
959 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
960 spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
961 struct hlist_nulls_node *node;
962 struct sock *sk_arr[SKARR_SZ];
963 int num_arr[SKARR_SZ];
964 int idx, accum, res;
965
966 if (hlist_nulls_empty(&head->chain))
967 continue;
968
969 if (i > s_i)
970 s_num = 0;
971
972next_chunk:
973 num = 0;
974 accum = 0;
975 spin_lock_bh(lock);
976 sk_nulls_for_each(sk, node, &head->chain) {
977 int state;
978
979 if (!net_eq(sock_net(sk), net))
980 continue;
981 if (num < s_num)
982 goto next_normal;
983 state = (sk->sk_state == TCP_TIME_WAIT) ?
984 inet_twsk(sk)->tw_substate : sk->sk_state;
985 if (!(idiag_states & (1 << state)))
986 goto next_normal;
987 if (r->sdiag_family != AF_UNSPEC &&
988 sk->sk_family != r->sdiag_family)
989 goto next_normal;
990 if (r->id.idiag_sport != htons(sk->sk_num) &&
991 r->id.idiag_sport)
992 goto next_normal;
993 if (r->id.idiag_dport != sk->sk_dport &&
994 r->id.idiag_dport)
995 goto next_normal;
996 twsk_build_assert();
997
998 if (!inet_diag_bc_sk(bc, sk))
999 goto next_normal;
1000
1001 sock_hold(sk);
1002 num_arr[accum] = num;
1003 sk_arr[accum] = sk;
1004 if (++accum == SKARR_SZ)
1005 break;
1006next_normal:
1007 ++num;
1008 }
1009 spin_unlock_bh(lock);
1010 res = 0;
1011 for (idx = 0; idx < accum; idx++) {
1012 if (res >= 0) {
1013 res = sk_diag_fill(sk_arr[idx], skb, r,
1014 sk_user_ns(NETLINK_CB(cb->skb).sk),
1015 NETLINK_CB(cb->skb).portid,
1016 cb->nlh->nlmsg_seq, NLM_F_MULTI,
1017 cb->nlh, net_admin);
1018 if (res < 0)
1019 num = num_arr[idx];
1020 }
1021 sock_gen_put(sk_arr[idx]);
1022 }
1023 if (res < 0)
1024 break;
1025 cond_resched();
1026 if (accum == SKARR_SZ) {
1027 s_num = num + 1;
1028 goto next_chunk;
1029 }
1030 }
1031
1032done:
1033 cb->args[1] = i;
1034 cb->args[2] = num;
1035out:
1036 ;
1037}
1038EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
1039
1040static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
1041 const struct inet_diag_req_v2 *r,
1042 struct nlattr *bc)
1043{
1044 const struct inet_diag_handler *handler;
1045 int err = 0;
1046
1047 handler = inet_diag_lock_handler(r->sdiag_protocol);
1048 if (!IS_ERR(handler))
1049 handler->dump(skb, cb, r, bc);
1050 else
1051 err = PTR_ERR(handler);
1052 inet_diag_unlock_handler(handler);
1053
1054 return err ? : skb->len;
1055}
1056
1057static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
1058{
1059 int hdrlen = sizeof(struct inet_diag_req_v2);
1060 struct nlattr *bc = NULL;
1061
1062 if (nlmsg_attrlen(cb->nlh, hdrlen))
1063 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE);
1064
1065 return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh), bc);
1066}
1067
1068static int inet_diag_type2proto(int type)
1069{
1070 switch (type) {
1071 case TCPDIAG_GETSOCK:
1072 return IPPROTO_TCP;
1073 case DCCPDIAG_GETSOCK:
1074 return IPPROTO_DCCP;
1075 default:
1076 return 0;
1077 }
1078}
1079
1080static int inet_diag_dump_compat(struct sk_buff *skb,
1081 struct netlink_callback *cb)
1082{
1083 struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1084 int hdrlen = sizeof(struct inet_diag_req);
1085 struct inet_diag_req_v2 req;
1086 struct nlattr *bc = NULL;
1087
1088 req.sdiag_family = AF_UNSPEC; /* compatibility */
1089 req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1090 req.idiag_ext = rc->idiag_ext;
1091 req.idiag_states = rc->idiag_states;
1092 req.id = rc->id;
1093
1094 if (nlmsg_attrlen(cb->nlh, hdrlen))
1095 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE);
1096
1097 return __inet_diag_dump(skb, cb, &req, bc);
1098}
1099
1100static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1101 const struct nlmsghdr *nlh)
1102{
1103 struct inet_diag_req *rc = nlmsg_data(nlh);
1104 struct inet_diag_req_v2 req;
1105
1106 req.sdiag_family = rc->idiag_family;
1107 req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1108 req.idiag_ext = rc->idiag_ext;
1109 req.idiag_states = rc->idiag_states;
1110 req.id = rc->id;
1111
1112 return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh, &req);
1113}
1114
1115static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1116{
1117 int hdrlen = sizeof(struct inet_diag_req);
1118 struct net *net = sock_net(skb->sk);
1119
1120 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1121 nlmsg_len(nlh) < hdrlen)
1122 return -EINVAL;
1123
1124 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1125 if (nlmsg_attrlen(nlh, hdrlen)) {
1126 struct nlattr *attr;
1127 int err;
1128
1129 attr = nlmsg_find_attr(nlh, hdrlen,
1130 INET_DIAG_REQ_BYTECODE);
1131 err = inet_diag_bc_audit(attr, skb);
1132 if (err)
1133 return err;
1134 }
1135 {
1136 struct netlink_dump_control c = {
1137 .dump = inet_diag_dump_compat,
1138 };
1139 return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1140 }
1141 }
1142
1143 return inet_diag_get_exact_compat(skb, nlh);
1144}
1145
1146static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h)
1147{
1148 int hdrlen = sizeof(struct inet_diag_req_v2);
1149 struct net *net = sock_net(skb->sk);
1150
1151 if (nlmsg_len(h) < hdrlen)
1152 return -EINVAL;
1153
1154 if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY &&
1155 h->nlmsg_flags & NLM_F_DUMP) {
1156 if (nlmsg_attrlen(h, hdrlen)) {
1157 struct nlattr *attr;
1158 int err;
1159
1160 attr = nlmsg_find_attr(h, hdrlen,
1161 INET_DIAG_REQ_BYTECODE);
1162 err = inet_diag_bc_audit(attr, skb);
1163 if (err)
1164 return err;
1165 }
1166 {
1167 struct netlink_dump_control c = {
1168 .dump = inet_diag_dump,
1169 };
1170 return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1171 }
1172 }
1173
1174 return inet_diag_cmd_exact(h->nlmsg_type, skb, h, nlmsg_data(h));
1175}
1176
1177static
1178int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk)
1179{
1180 const struct inet_diag_handler *handler;
1181 struct nlmsghdr *nlh;
1182 struct nlattr *attr;
1183 struct inet_diag_msg *r;
1184 void *info = NULL;
1185 int err = 0;
1186
1187 nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0);
1188 if (!nlh)
1189 return -ENOMEM;
1190
1191 r = nlmsg_data(nlh);
1192 memset(r, 0, sizeof(*r));
1193 inet_diag_msg_common_fill(r, sk);
1194 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM)
1195 r->id.idiag_sport = inet_sk(sk)->inet_sport;
1196 r->idiag_state = sk->sk_state;
1197
1198 if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) {
1199 nlmsg_cancel(skb, nlh);
1200 return err;
1201 }
1202
1203 handler = inet_diag_lock_handler(sk->sk_protocol);
1204 if (IS_ERR(handler)) {
1205 inet_diag_unlock_handler(handler);
1206 nlmsg_cancel(skb, nlh);
1207 return PTR_ERR(handler);
1208 }
1209
1210 attr = handler->idiag_info_size
1211 ? nla_reserve_64bit(skb, INET_DIAG_INFO,
1212 handler->idiag_info_size,
1213 INET_DIAG_PAD)
1214 : NULL;
1215 if (attr)
1216 info = nla_data(attr);
1217
1218 handler->idiag_get_info(sk, r, info);
1219 inet_diag_unlock_handler(handler);
1220
1221 nlmsg_end(skb, nlh);
1222 return 0;
1223}
1224
1225static const struct sock_diag_handler inet_diag_handler = {
1226 .family = AF_INET,
1227 .dump = inet_diag_handler_cmd,
1228 .get_info = inet_diag_handler_get_info,
1229 .destroy = inet_diag_handler_cmd,
1230};
1231
1232static const struct sock_diag_handler inet6_diag_handler = {
1233 .family = AF_INET6,
1234 .dump = inet_diag_handler_cmd,
1235 .get_info = inet_diag_handler_get_info,
1236 .destroy = inet_diag_handler_cmd,
1237};
1238
1239int inet_diag_register(const struct inet_diag_handler *h)
1240{
1241 const __u16 type = h->idiag_type;
1242 int err = -EINVAL;
1243
1244 if (type >= IPPROTO_MAX)
1245 goto out;
1246
1247 mutex_lock(&inet_diag_table_mutex);
1248 err = -EEXIST;
1249 if (!inet_diag_table[type]) {
1250 inet_diag_table[type] = h;
1251 err = 0;
1252 }
1253 mutex_unlock(&inet_diag_table_mutex);
1254out:
1255 return err;
1256}
1257EXPORT_SYMBOL_GPL(inet_diag_register);
1258
1259void inet_diag_unregister(const struct inet_diag_handler *h)
1260{
1261 const __u16 type = h->idiag_type;
1262
1263 if (type >= IPPROTO_MAX)
1264 return;
1265
1266 mutex_lock(&inet_diag_table_mutex);
1267 inet_diag_table[type] = NULL;
1268 mutex_unlock(&inet_diag_table_mutex);
1269}
1270EXPORT_SYMBOL_GPL(inet_diag_unregister);
1271
1272static int __init inet_diag_init(void)
1273{
1274 const int inet_diag_table_size = (IPPROTO_MAX *
1275 sizeof(struct inet_diag_handler *));
1276 int err = -ENOMEM;
1277
1278 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1279 if (!inet_diag_table)
1280 goto out;
1281
1282 err = sock_diag_register(&inet_diag_handler);
1283 if (err)
1284 goto out_free_nl;
1285
1286 err = sock_diag_register(&inet6_diag_handler);
1287 if (err)
1288 goto out_free_inet;
1289
1290 sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1291out:
1292 return err;
1293
1294out_free_inet:
1295 sock_diag_unregister(&inet_diag_handler);
1296out_free_nl:
1297 kfree(inet_diag_table);
1298 goto out;
1299}
1300
1301static void __exit inet_diag_exit(void)
1302{
1303 sock_diag_unregister(&inet6_diag_handler);
1304 sock_diag_unregister(&inet_diag_handler);
1305 sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1306 kfree(inet_diag_table);
1307}
1308
1309module_init(inet_diag_init);
1310module_exit(inet_diag_exit);
1311MODULE_LICENSE("GPL");
1312MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1313MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);
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
38static const struct inet_diag_handler **inet_diag_table;
39
40struct inet_diag_entry {
41 const __be32 *saddr;
42 const __be32 *daddr;
43 u16 sport;
44 u16 dport;
45 u16 family;
46 u16 userlocks;
47};
48
49static DEFINE_MUTEX(inet_diag_table_mutex);
50
51static const struct inet_diag_handler *inet_diag_lock_handler(int proto)
52{
53 if (!inet_diag_table[proto])
54 request_module("net-pf-%d-proto-%d-type-%d-%d", PF_NETLINK,
55 NETLINK_SOCK_DIAG, AF_INET, proto);
56
57 mutex_lock(&inet_diag_table_mutex);
58 if (!inet_diag_table[proto])
59 return ERR_PTR(-ENOENT);
60
61 return inet_diag_table[proto];
62}
63
64static void inet_diag_unlock_handler(const struct inet_diag_handler *handler)
65{
66 mutex_unlock(&inet_diag_table_mutex);
67}
68
69static void inet_diag_msg_common_fill(struct inet_diag_msg *r, struct sock *sk)
70{
71 r->idiag_family = sk->sk_family;
72
73 r->id.idiag_sport = htons(sk->sk_num);
74 r->id.idiag_dport = sk->sk_dport;
75 r->id.idiag_if = sk->sk_bound_dev_if;
76 sock_diag_save_cookie(sk, r->id.idiag_cookie);
77
78#if IS_ENABLED(CONFIG_IPV6)
79 if (sk->sk_family == AF_INET6) {
80 *(struct in6_addr *)r->id.idiag_src = sk->sk_v6_rcv_saddr;
81 *(struct in6_addr *)r->id.idiag_dst = sk->sk_v6_daddr;
82 } else
83#endif
84 {
85 memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src));
86 memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst));
87
88 r->id.idiag_src[0] = sk->sk_rcv_saddr;
89 r->id.idiag_dst[0] = sk->sk_daddr;
90 }
91}
92
93static size_t inet_sk_attr_size(void)
94{
95 return nla_total_size(sizeof(struct tcp_info))
96 + nla_total_size(1) /* INET_DIAG_SHUTDOWN */
97 + nla_total_size(1) /* INET_DIAG_TOS */
98 + nla_total_size(1) /* INET_DIAG_TCLASS */
99 + nla_total_size(sizeof(struct inet_diag_meminfo))
100 + nla_total_size(sizeof(struct inet_diag_msg))
101 + nla_total_size(SK_MEMINFO_VARS * sizeof(u32))
102 + nla_total_size(TCP_CA_NAME_MAX)
103 + nla_total_size(sizeof(struct tcpvegas_info))
104 + 64;
105}
106
107int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk,
108 struct sk_buff *skb, const struct inet_diag_req_v2 *req,
109 struct user_namespace *user_ns,
110 u32 portid, u32 seq, u16 nlmsg_flags,
111 const struct nlmsghdr *unlh)
112{
113 const struct inet_sock *inet = inet_sk(sk);
114 const struct tcp_congestion_ops *ca_ops;
115 const struct inet_diag_handler *handler;
116 int ext = req->idiag_ext;
117 struct inet_diag_msg *r;
118 struct nlmsghdr *nlh;
119 struct nlattr *attr;
120 void *info = NULL;
121
122 handler = inet_diag_table[req->sdiag_protocol];
123 BUG_ON(!handler);
124
125 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
126 nlmsg_flags);
127 if (!nlh)
128 return -EMSGSIZE;
129
130 r = nlmsg_data(nlh);
131 BUG_ON(!sk_fullsock(sk));
132
133 inet_diag_msg_common_fill(r, sk);
134 r->idiag_state = sk->sk_state;
135 r->idiag_timer = 0;
136 r->idiag_retrans = 0;
137
138 if (nla_put_u8(skb, INET_DIAG_SHUTDOWN, sk->sk_shutdown))
139 goto errout;
140
141 /* IPv6 dual-stack sockets use inet->tos for IPv4 connections,
142 * hence this needs to be included regardless of socket family.
143 */
144 if (ext & (1 << (INET_DIAG_TOS - 1)))
145 if (nla_put_u8(skb, INET_DIAG_TOS, inet->tos) < 0)
146 goto errout;
147
148#if IS_ENABLED(CONFIG_IPV6)
149 if (r->idiag_family == AF_INET6) {
150 if (ext & (1 << (INET_DIAG_TCLASS - 1)))
151 if (nla_put_u8(skb, INET_DIAG_TCLASS,
152 inet6_sk(sk)->tclass) < 0)
153 goto errout;
154
155 if (((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
156 nla_put_u8(skb, INET_DIAG_SKV6ONLY, ipv6_only_sock(sk)))
157 goto errout;
158 }
159#endif
160
161 r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk));
162 r->idiag_inode = sock_i_ino(sk);
163
164 if (ext & (1 << (INET_DIAG_MEMINFO - 1))) {
165 struct inet_diag_meminfo minfo = {
166 .idiag_rmem = sk_rmem_alloc_get(sk),
167 .idiag_wmem = sk->sk_wmem_queued,
168 .idiag_fmem = sk->sk_forward_alloc,
169 .idiag_tmem = sk_wmem_alloc_get(sk),
170 };
171
172 if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0)
173 goto errout;
174 }
175
176 if (ext & (1 << (INET_DIAG_SKMEMINFO - 1)))
177 if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO))
178 goto errout;
179
180 if (!icsk) {
181 handler->idiag_get_info(sk, r, NULL);
182 goto out;
183 }
184
185#define EXPIRES_IN_MS(tmo) DIV_ROUND_UP((tmo - jiffies) * 1000, HZ)
186
187 if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
188 icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS ||
189 icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
190 r->idiag_timer = 1;
191 r->idiag_retrans = icsk->icsk_retransmits;
192 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
193 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
194 r->idiag_timer = 4;
195 r->idiag_retrans = icsk->icsk_probes_out;
196 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
197 } else if (timer_pending(&sk->sk_timer)) {
198 r->idiag_timer = 2;
199 r->idiag_retrans = icsk->icsk_probes_out;
200 r->idiag_expires = EXPIRES_IN_MS(sk->sk_timer.expires);
201 } else {
202 r->idiag_timer = 0;
203 r->idiag_expires = 0;
204 }
205#undef EXPIRES_IN_MS
206
207 if ((ext & (1 << (INET_DIAG_INFO - 1))) && handler->idiag_info_size) {
208 attr = nla_reserve(skb, INET_DIAG_INFO,
209 handler->idiag_info_size);
210 if (!attr)
211 goto errout;
212
213 info = nla_data(attr);
214 }
215
216 if (ext & (1 << (INET_DIAG_CONG - 1))) {
217 int err = 0;
218
219 rcu_read_lock();
220 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
221 if (ca_ops)
222 err = nla_put_string(skb, INET_DIAG_CONG, ca_ops->name);
223 rcu_read_unlock();
224 if (err < 0)
225 goto errout;
226 }
227
228 handler->idiag_get_info(sk, r, info);
229
230 if (sk->sk_state < TCP_TIME_WAIT) {
231 union tcp_cc_info info;
232 size_t sz = 0;
233 int attr;
234
235 rcu_read_lock();
236 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
237 if (ca_ops && ca_ops->get_info)
238 sz = ca_ops->get_info(sk, ext, &attr, &info);
239 rcu_read_unlock();
240 if (sz && nla_put(skb, attr, sz, &info) < 0)
241 goto errout;
242 }
243
244out:
245 nlmsg_end(skb, nlh);
246 return 0;
247
248errout:
249 nlmsg_cancel(skb, nlh);
250 return -EMSGSIZE;
251}
252EXPORT_SYMBOL_GPL(inet_sk_diag_fill);
253
254static int inet_csk_diag_fill(struct sock *sk,
255 struct sk_buff *skb,
256 const struct inet_diag_req_v2 *req,
257 struct user_namespace *user_ns,
258 u32 portid, u32 seq, u16 nlmsg_flags,
259 const struct nlmsghdr *unlh)
260{
261 return inet_sk_diag_fill(sk, inet_csk(sk), skb, req,
262 user_ns, portid, seq, nlmsg_flags, unlh);
263}
264
265static int inet_twsk_diag_fill(struct sock *sk,
266 struct sk_buff *skb,
267 u32 portid, u32 seq, u16 nlmsg_flags,
268 const struct nlmsghdr *unlh)
269{
270 struct inet_timewait_sock *tw = inet_twsk(sk);
271 struct inet_diag_msg *r;
272 struct nlmsghdr *nlh;
273 long tmo;
274
275 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
276 nlmsg_flags);
277 if (!nlh)
278 return -EMSGSIZE;
279
280 r = nlmsg_data(nlh);
281 BUG_ON(tw->tw_state != TCP_TIME_WAIT);
282
283 tmo = tw->tw_timer.expires - jiffies;
284 if (tmo < 0)
285 tmo = 0;
286
287 inet_diag_msg_common_fill(r, sk);
288 r->idiag_retrans = 0;
289
290 r->idiag_state = tw->tw_substate;
291 r->idiag_timer = 3;
292 r->idiag_expires = jiffies_to_msecs(tmo);
293 r->idiag_rqueue = 0;
294 r->idiag_wqueue = 0;
295 r->idiag_uid = 0;
296 r->idiag_inode = 0;
297
298 nlmsg_end(skb, nlh);
299 return 0;
300}
301
302static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb,
303 u32 portid, u32 seq, u16 nlmsg_flags,
304 const struct nlmsghdr *unlh)
305{
306 struct inet_diag_msg *r;
307 struct nlmsghdr *nlh;
308 long tmo;
309
310 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
311 nlmsg_flags);
312 if (!nlh)
313 return -EMSGSIZE;
314
315 r = nlmsg_data(nlh);
316 inet_diag_msg_common_fill(r, sk);
317 r->idiag_state = TCP_SYN_RECV;
318 r->idiag_timer = 1;
319 r->idiag_retrans = inet_reqsk(sk)->num_retrans;
320
321 BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) !=
322 offsetof(struct sock, sk_cookie));
323
324 tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies;
325 r->idiag_expires = (tmo >= 0) ? jiffies_to_msecs(tmo) : 0;
326 r->idiag_rqueue = 0;
327 r->idiag_wqueue = 0;
328 r->idiag_uid = 0;
329 r->idiag_inode = 0;
330
331 nlmsg_end(skb, nlh);
332 return 0;
333}
334
335static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
336 const struct inet_diag_req_v2 *r,
337 struct user_namespace *user_ns,
338 u32 portid, u32 seq, u16 nlmsg_flags,
339 const struct nlmsghdr *unlh)
340{
341 if (sk->sk_state == TCP_TIME_WAIT)
342 return inet_twsk_diag_fill(sk, skb, portid, seq,
343 nlmsg_flags, unlh);
344
345 if (sk->sk_state == TCP_NEW_SYN_RECV)
346 return inet_req_diag_fill(sk, skb, portid, seq,
347 nlmsg_flags, unlh);
348
349 return inet_csk_diag_fill(sk, skb, r, user_ns, portid, seq,
350 nlmsg_flags, unlh);
351}
352
353struct sock *inet_diag_find_one_icsk(struct net *net,
354 struct inet_hashinfo *hashinfo,
355 const struct inet_diag_req_v2 *req)
356{
357 struct sock *sk;
358
359 if (req->sdiag_family == AF_INET)
360 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0],
361 req->id.idiag_dport, req->id.idiag_src[0],
362 req->id.idiag_sport, req->id.idiag_if);
363#if IS_ENABLED(CONFIG_IPV6)
364 else if (req->sdiag_family == AF_INET6) {
365 if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) &&
366 ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src))
367 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3],
368 req->id.idiag_dport, req->id.idiag_src[3],
369 req->id.idiag_sport, req->id.idiag_if);
370 else
371 sk = inet6_lookup(net, hashinfo, NULL, 0,
372 (struct in6_addr *)req->id.idiag_dst,
373 req->id.idiag_dport,
374 (struct in6_addr *)req->id.idiag_src,
375 req->id.idiag_sport,
376 req->id.idiag_if);
377 }
378#endif
379 else
380 return ERR_PTR(-EINVAL);
381
382 if (!sk)
383 return ERR_PTR(-ENOENT);
384
385 if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) {
386 sock_gen_put(sk);
387 return ERR_PTR(-ENOENT);
388 }
389
390 return sk;
391}
392EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk);
393
394int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo,
395 struct sk_buff *in_skb,
396 const struct nlmsghdr *nlh,
397 const struct inet_diag_req_v2 *req)
398{
399 struct net *net = sock_net(in_skb->sk);
400 struct sk_buff *rep;
401 struct sock *sk;
402 int err;
403
404 sk = inet_diag_find_one_icsk(net, hashinfo, req);
405 if (IS_ERR(sk))
406 return PTR_ERR(sk);
407
408 rep = nlmsg_new(inet_sk_attr_size(), GFP_KERNEL);
409 if (!rep) {
410 err = -ENOMEM;
411 goto out;
412 }
413
414 err = sk_diag_fill(sk, rep, req,
415 sk_user_ns(NETLINK_CB(in_skb).sk),
416 NETLINK_CB(in_skb).portid,
417 nlh->nlmsg_seq, 0, nlh);
418 if (err < 0) {
419 WARN_ON(err == -EMSGSIZE);
420 nlmsg_free(rep);
421 goto out;
422 }
423 err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid,
424 MSG_DONTWAIT);
425 if (err > 0)
426 err = 0;
427
428out:
429 if (sk)
430 sock_gen_put(sk);
431
432 return err;
433}
434EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
435
436static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb,
437 const struct nlmsghdr *nlh,
438 const struct inet_diag_req_v2 *req)
439{
440 const struct inet_diag_handler *handler;
441 int err;
442
443 handler = inet_diag_lock_handler(req->sdiag_protocol);
444 if (IS_ERR(handler))
445 err = PTR_ERR(handler);
446 else if (cmd == SOCK_DIAG_BY_FAMILY)
447 err = handler->dump_one(in_skb, nlh, req);
448 else if (cmd == SOCK_DESTROY && handler->destroy)
449 err = handler->destroy(in_skb, req);
450 else
451 err = -EOPNOTSUPP;
452 inet_diag_unlock_handler(handler);
453
454 return err;
455}
456
457static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
458{
459 int words = bits >> 5;
460
461 bits &= 0x1f;
462
463 if (words) {
464 if (memcmp(a1, a2, words << 2))
465 return 0;
466 }
467 if (bits) {
468 __be32 w1, w2;
469 __be32 mask;
470
471 w1 = a1[words];
472 w2 = a2[words];
473
474 mask = htonl((0xffffffff) << (32 - bits));
475
476 if ((w1 ^ w2) & mask)
477 return 0;
478 }
479
480 return 1;
481}
482
483static int inet_diag_bc_run(const struct nlattr *_bc,
484 const struct inet_diag_entry *entry)
485{
486 const void *bc = nla_data(_bc);
487 int len = nla_len(_bc);
488
489 while (len > 0) {
490 int yes = 1;
491 const struct inet_diag_bc_op *op = bc;
492
493 switch (op->code) {
494 case INET_DIAG_BC_NOP:
495 break;
496 case INET_DIAG_BC_JMP:
497 yes = 0;
498 break;
499 case INET_DIAG_BC_S_GE:
500 yes = entry->sport >= op[1].no;
501 break;
502 case INET_DIAG_BC_S_LE:
503 yes = entry->sport <= op[1].no;
504 break;
505 case INET_DIAG_BC_D_GE:
506 yes = entry->dport >= op[1].no;
507 break;
508 case INET_DIAG_BC_D_LE:
509 yes = entry->dport <= op[1].no;
510 break;
511 case INET_DIAG_BC_AUTO:
512 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
513 break;
514 case INET_DIAG_BC_S_COND:
515 case INET_DIAG_BC_D_COND: {
516 const struct inet_diag_hostcond *cond;
517 const __be32 *addr;
518
519 cond = (const struct inet_diag_hostcond *)(op + 1);
520 if (cond->port != -1 &&
521 cond->port != (op->code == INET_DIAG_BC_S_COND ?
522 entry->sport : entry->dport)) {
523 yes = 0;
524 break;
525 }
526
527 if (op->code == INET_DIAG_BC_S_COND)
528 addr = entry->saddr;
529 else
530 addr = entry->daddr;
531
532 if (cond->family != AF_UNSPEC &&
533 cond->family != entry->family) {
534 if (entry->family == AF_INET6 &&
535 cond->family == AF_INET) {
536 if (addr[0] == 0 && addr[1] == 0 &&
537 addr[2] == htonl(0xffff) &&
538 bitstring_match(addr + 3,
539 cond->addr,
540 cond->prefix_len))
541 break;
542 }
543 yes = 0;
544 break;
545 }
546
547 if (cond->prefix_len == 0)
548 break;
549 if (bitstring_match(addr, cond->addr,
550 cond->prefix_len))
551 break;
552 yes = 0;
553 break;
554 }
555 }
556
557 if (yes) {
558 len -= op->yes;
559 bc += op->yes;
560 } else {
561 len -= op->no;
562 bc += op->no;
563 }
564 }
565 return len == 0;
566}
567
568/* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV)
569 */
570static void entry_fill_addrs(struct inet_diag_entry *entry,
571 const struct sock *sk)
572{
573#if IS_ENABLED(CONFIG_IPV6)
574 if (sk->sk_family == AF_INET6) {
575 entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32;
576 entry->daddr = sk->sk_v6_daddr.s6_addr32;
577 } else
578#endif
579 {
580 entry->saddr = &sk->sk_rcv_saddr;
581 entry->daddr = &sk->sk_daddr;
582 }
583}
584
585int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
586{
587 struct inet_sock *inet = inet_sk(sk);
588 struct inet_diag_entry entry;
589
590 if (!bc)
591 return 1;
592
593 entry.family = sk->sk_family;
594 entry_fill_addrs(&entry, sk);
595 entry.sport = inet->inet_num;
596 entry.dport = ntohs(inet->inet_dport);
597 entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0;
598
599 return inet_diag_bc_run(bc, &entry);
600}
601EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
602
603static int valid_cc(const void *bc, int len, int cc)
604{
605 while (len >= 0) {
606 const struct inet_diag_bc_op *op = bc;
607
608 if (cc > len)
609 return 0;
610 if (cc == len)
611 return 1;
612 if (op->yes < 4 || op->yes & 3)
613 return 0;
614 len -= op->yes;
615 bc += op->yes;
616 }
617 return 0;
618}
619
620/* Validate an inet_diag_hostcond. */
621static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
622 int *min_len)
623{
624 struct inet_diag_hostcond *cond;
625 int addr_len;
626
627 /* Check hostcond space. */
628 *min_len += sizeof(struct inet_diag_hostcond);
629 if (len < *min_len)
630 return false;
631 cond = (struct inet_diag_hostcond *)(op + 1);
632
633 /* Check address family and address length. */
634 switch (cond->family) {
635 case AF_UNSPEC:
636 addr_len = 0;
637 break;
638 case AF_INET:
639 addr_len = sizeof(struct in_addr);
640 break;
641 case AF_INET6:
642 addr_len = sizeof(struct in6_addr);
643 break;
644 default:
645 return false;
646 }
647 *min_len += addr_len;
648 if (len < *min_len)
649 return false;
650
651 /* Check prefix length (in bits) vs address length (in bytes). */
652 if (cond->prefix_len > 8 * addr_len)
653 return false;
654
655 return true;
656}
657
658/* Validate a port comparison operator. */
659static bool valid_port_comparison(const struct inet_diag_bc_op *op,
660 int len, int *min_len)
661{
662 /* Port comparisons put the port in a follow-on inet_diag_bc_op. */
663 *min_len += sizeof(struct inet_diag_bc_op);
664 if (len < *min_len)
665 return false;
666 return true;
667}
668
669static int inet_diag_bc_audit(const void *bytecode, int bytecode_len)
670{
671 const void *bc = bytecode;
672 int len = bytecode_len;
673
674 while (len > 0) {
675 int min_len = sizeof(struct inet_diag_bc_op);
676 const struct inet_diag_bc_op *op = bc;
677
678 switch (op->code) {
679 case INET_DIAG_BC_S_COND:
680 case INET_DIAG_BC_D_COND:
681 if (!valid_hostcond(bc, len, &min_len))
682 return -EINVAL;
683 break;
684 case INET_DIAG_BC_S_GE:
685 case INET_DIAG_BC_S_LE:
686 case INET_DIAG_BC_D_GE:
687 case INET_DIAG_BC_D_LE:
688 if (!valid_port_comparison(bc, len, &min_len))
689 return -EINVAL;
690 break;
691 case INET_DIAG_BC_AUTO:
692 case INET_DIAG_BC_JMP:
693 case INET_DIAG_BC_NOP:
694 break;
695 default:
696 return -EINVAL;
697 }
698
699 if (op->code != INET_DIAG_BC_NOP) {
700 if (op->no < min_len || op->no > len + 4 || op->no & 3)
701 return -EINVAL;
702 if (op->no < len &&
703 !valid_cc(bytecode, bytecode_len, len - op->no))
704 return -EINVAL;
705 }
706
707 if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
708 return -EINVAL;
709 bc += op->yes;
710 len -= op->yes;
711 }
712 return len == 0 ? 0 : -EINVAL;
713}
714
715static int inet_csk_diag_dump(struct sock *sk,
716 struct sk_buff *skb,
717 struct netlink_callback *cb,
718 const struct inet_diag_req_v2 *r,
719 const struct nlattr *bc)
720{
721 if (!inet_diag_bc_sk(bc, sk))
722 return 0;
723
724 return inet_csk_diag_fill(sk, skb, r,
725 sk_user_ns(NETLINK_CB(cb->skb).sk),
726 NETLINK_CB(cb->skb).portid,
727 cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
728}
729
730static void twsk_build_assert(void)
731{
732 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) !=
733 offsetof(struct sock, sk_family));
734
735 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) !=
736 offsetof(struct inet_sock, inet_num));
737
738 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) !=
739 offsetof(struct inet_sock, inet_dport));
740
741 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) !=
742 offsetof(struct inet_sock, inet_rcv_saddr));
743
744 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) !=
745 offsetof(struct inet_sock, inet_daddr));
746
747#if IS_ENABLED(CONFIG_IPV6)
748 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) !=
749 offsetof(struct sock, sk_v6_rcv_saddr));
750
751 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) !=
752 offsetof(struct sock, sk_v6_daddr));
753#endif
754}
755
756void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
757 struct netlink_callback *cb,
758 const struct inet_diag_req_v2 *r, struct nlattr *bc)
759{
760 struct net *net = sock_net(skb->sk);
761 int i, num, s_i, s_num;
762 u32 idiag_states = r->idiag_states;
763
764 if (idiag_states & TCPF_SYN_RECV)
765 idiag_states |= TCPF_NEW_SYN_RECV;
766 s_i = cb->args[1];
767 s_num = num = cb->args[2];
768
769 if (cb->args[0] == 0) {
770 if (!(idiag_states & TCPF_LISTEN))
771 goto skip_listen_ht;
772
773 for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
774 struct inet_listen_hashbucket *ilb;
775 struct hlist_nulls_node *node;
776 struct sock *sk;
777
778 num = 0;
779 ilb = &hashinfo->listening_hash[i];
780 spin_lock_bh(&ilb->lock);
781 sk_nulls_for_each(sk, node, &ilb->head) {
782 struct inet_sock *inet = inet_sk(sk);
783
784 if (!net_eq(sock_net(sk), net))
785 continue;
786
787 if (num < s_num) {
788 num++;
789 continue;
790 }
791
792 if (r->sdiag_family != AF_UNSPEC &&
793 sk->sk_family != r->sdiag_family)
794 goto next_listen;
795
796 if (r->id.idiag_sport != inet->inet_sport &&
797 r->id.idiag_sport)
798 goto next_listen;
799
800 if (r->id.idiag_dport ||
801 cb->args[3] > 0)
802 goto next_listen;
803
804 if (inet_csk_diag_dump(sk, skb, cb, r, bc) < 0) {
805 spin_unlock_bh(&ilb->lock);
806 goto done;
807 }
808
809next_listen:
810 cb->args[3] = 0;
811 cb->args[4] = 0;
812 ++num;
813 }
814 spin_unlock_bh(&ilb->lock);
815
816 s_num = 0;
817 cb->args[3] = 0;
818 cb->args[4] = 0;
819 }
820skip_listen_ht:
821 cb->args[0] = 1;
822 s_i = num = s_num = 0;
823 }
824
825 if (!(idiag_states & ~TCPF_LISTEN))
826 goto out;
827
828 for (i = s_i; i <= hashinfo->ehash_mask; i++) {
829 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
830 spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
831 struct hlist_nulls_node *node;
832 struct sock *sk;
833
834 num = 0;
835
836 if (hlist_nulls_empty(&head->chain))
837 continue;
838
839 if (i > s_i)
840 s_num = 0;
841
842 spin_lock_bh(lock);
843 sk_nulls_for_each(sk, node, &head->chain) {
844 int state, res;
845
846 if (!net_eq(sock_net(sk), net))
847 continue;
848 if (num < s_num)
849 goto next_normal;
850 state = (sk->sk_state == TCP_TIME_WAIT) ?
851 inet_twsk(sk)->tw_substate : sk->sk_state;
852 if (!(idiag_states & (1 << state)))
853 goto next_normal;
854 if (r->sdiag_family != AF_UNSPEC &&
855 sk->sk_family != r->sdiag_family)
856 goto next_normal;
857 if (r->id.idiag_sport != htons(sk->sk_num) &&
858 r->id.idiag_sport)
859 goto next_normal;
860 if (r->id.idiag_dport != sk->sk_dport &&
861 r->id.idiag_dport)
862 goto next_normal;
863 twsk_build_assert();
864
865 if (!inet_diag_bc_sk(bc, sk))
866 goto next_normal;
867
868 res = sk_diag_fill(sk, skb, r,
869 sk_user_ns(NETLINK_CB(cb->skb).sk),
870 NETLINK_CB(cb->skb).portid,
871 cb->nlh->nlmsg_seq, NLM_F_MULTI,
872 cb->nlh);
873 if (res < 0) {
874 spin_unlock_bh(lock);
875 goto done;
876 }
877next_normal:
878 ++num;
879 }
880
881 spin_unlock_bh(lock);
882 cond_resched();
883 }
884
885done:
886 cb->args[1] = i;
887 cb->args[2] = num;
888out:
889 ;
890}
891EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
892
893static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
894 const struct inet_diag_req_v2 *r,
895 struct nlattr *bc)
896{
897 const struct inet_diag_handler *handler;
898 int err = 0;
899
900 handler = inet_diag_lock_handler(r->sdiag_protocol);
901 if (!IS_ERR(handler))
902 handler->dump(skb, cb, r, bc);
903 else
904 err = PTR_ERR(handler);
905 inet_diag_unlock_handler(handler);
906
907 return err ? : skb->len;
908}
909
910static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
911{
912 int hdrlen = sizeof(struct inet_diag_req_v2);
913 struct nlattr *bc = NULL;
914
915 if (nlmsg_attrlen(cb->nlh, hdrlen))
916 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE);
917
918 return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh), bc);
919}
920
921static int inet_diag_type2proto(int type)
922{
923 switch (type) {
924 case TCPDIAG_GETSOCK:
925 return IPPROTO_TCP;
926 case DCCPDIAG_GETSOCK:
927 return IPPROTO_DCCP;
928 default:
929 return 0;
930 }
931}
932
933static int inet_diag_dump_compat(struct sk_buff *skb,
934 struct netlink_callback *cb)
935{
936 struct inet_diag_req *rc = nlmsg_data(cb->nlh);
937 int hdrlen = sizeof(struct inet_diag_req);
938 struct inet_diag_req_v2 req;
939 struct nlattr *bc = NULL;
940
941 req.sdiag_family = AF_UNSPEC; /* compatibility */
942 req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
943 req.idiag_ext = rc->idiag_ext;
944 req.idiag_states = rc->idiag_states;
945 req.id = rc->id;
946
947 if (nlmsg_attrlen(cb->nlh, hdrlen))
948 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE);
949
950 return __inet_diag_dump(skb, cb, &req, bc);
951}
952
953static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
954 const struct nlmsghdr *nlh)
955{
956 struct inet_diag_req *rc = nlmsg_data(nlh);
957 struct inet_diag_req_v2 req;
958
959 req.sdiag_family = rc->idiag_family;
960 req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
961 req.idiag_ext = rc->idiag_ext;
962 req.idiag_states = rc->idiag_states;
963 req.id = rc->id;
964
965 return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh, &req);
966}
967
968static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
969{
970 int hdrlen = sizeof(struct inet_diag_req);
971 struct net *net = sock_net(skb->sk);
972
973 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
974 nlmsg_len(nlh) < hdrlen)
975 return -EINVAL;
976
977 if (nlh->nlmsg_flags & NLM_F_DUMP) {
978 if (nlmsg_attrlen(nlh, hdrlen)) {
979 struct nlattr *attr;
980
981 attr = nlmsg_find_attr(nlh, hdrlen,
982 INET_DIAG_REQ_BYTECODE);
983 if (!attr ||
984 nla_len(attr) < sizeof(struct inet_diag_bc_op) ||
985 inet_diag_bc_audit(nla_data(attr), nla_len(attr)))
986 return -EINVAL;
987 }
988 {
989 struct netlink_dump_control c = {
990 .dump = inet_diag_dump_compat,
991 };
992 return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
993 }
994 }
995
996 return inet_diag_get_exact_compat(skb, nlh);
997}
998
999static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h)
1000{
1001 int hdrlen = sizeof(struct inet_diag_req_v2);
1002 struct net *net = sock_net(skb->sk);
1003
1004 if (nlmsg_len(h) < hdrlen)
1005 return -EINVAL;
1006
1007 if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY &&
1008 h->nlmsg_flags & NLM_F_DUMP) {
1009 if (nlmsg_attrlen(h, hdrlen)) {
1010 struct nlattr *attr;
1011
1012 attr = nlmsg_find_attr(h, hdrlen,
1013 INET_DIAG_REQ_BYTECODE);
1014 if (!attr ||
1015 nla_len(attr) < sizeof(struct inet_diag_bc_op) ||
1016 inet_diag_bc_audit(nla_data(attr), nla_len(attr)))
1017 return -EINVAL;
1018 }
1019 {
1020 struct netlink_dump_control c = {
1021 .dump = inet_diag_dump,
1022 };
1023 return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1024 }
1025 }
1026
1027 return inet_diag_cmd_exact(h->nlmsg_type, skb, h, nlmsg_data(h));
1028}
1029
1030static
1031int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk)
1032{
1033 const struct inet_diag_handler *handler;
1034 struct nlmsghdr *nlh;
1035 struct nlattr *attr;
1036 struct inet_diag_msg *r;
1037 void *info = NULL;
1038 int err = 0;
1039
1040 nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0);
1041 if (!nlh)
1042 return -ENOMEM;
1043
1044 r = nlmsg_data(nlh);
1045 memset(r, 0, sizeof(*r));
1046 inet_diag_msg_common_fill(r, sk);
1047 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM)
1048 r->id.idiag_sport = inet_sk(sk)->inet_sport;
1049 r->idiag_state = sk->sk_state;
1050
1051 if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) {
1052 nlmsg_cancel(skb, nlh);
1053 return err;
1054 }
1055
1056 handler = inet_diag_lock_handler(sk->sk_protocol);
1057 if (IS_ERR(handler)) {
1058 inet_diag_unlock_handler(handler);
1059 nlmsg_cancel(skb, nlh);
1060 return PTR_ERR(handler);
1061 }
1062
1063 attr = handler->idiag_info_size
1064 ? nla_reserve(skb, INET_DIAG_INFO, handler->idiag_info_size)
1065 : NULL;
1066 if (attr)
1067 info = nla_data(attr);
1068
1069 handler->idiag_get_info(sk, r, info);
1070 inet_diag_unlock_handler(handler);
1071
1072 nlmsg_end(skb, nlh);
1073 return 0;
1074}
1075
1076static const struct sock_diag_handler inet_diag_handler = {
1077 .family = AF_INET,
1078 .dump = inet_diag_handler_cmd,
1079 .get_info = inet_diag_handler_get_info,
1080 .destroy = inet_diag_handler_cmd,
1081};
1082
1083static const struct sock_diag_handler inet6_diag_handler = {
1084 .family = AF_INET6,
1085 .dump = inet_diag_handler_cmd,
1086 .get_info = inet_diag_handler_get_info,
1087 .destroy = inet_diag_handler_cmd,
1088};
1089
1090int inet_diag_register(const struct inet_diag_handler *h)
1091{
1092 const __u16 type = h->idiag_type;
1093 int err = -EINVAL;
1094
1095 if (type >= IPPROTO_MAX)
1096 goto out;
1097
1098 mutex_lock(&inet_diag_table_mutex);
1099 err = -EEXIST;
1100 if (!inet_diag_table[type]) {
1101 inet_diag_table[type] = h;
1102 err = 0;
1103 }
1104 mutex_unlock(&inet_diag_table_mutex);
1105out:
1106 return err;
1107}
1108EXPORT_SYMBOL_GPL(inet_diag_register);
1109
1110void inet_diag_unregister(const struct inet_diag_handler *h)
1111{
1112 const __u16 type = h->idiag_type;
1113
1114 if (type >= IPPROTO_MAX)
1115 return;
1116
1117 mutex_lock(&inet_diag_table_mutex);
1118 inet_diag_table[type] = NULL;
1119 mutex_unlock(&inet_diag_table_mutex);
1120}
1121EXPORT_SYMBOL_GPL(inet_diag_unregister);
1122
1123static int __init inet_diag_init(void)
1124{
1125 const int inet_diag_table_size = (IPPROTO_MAX *
1126 sizeof(struct inet_diag_handler *));
1127 int err = -ENOMEM;
1128
1129 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1130 if (!inet_diag_table)
1131 goto out;
1132
1133 err = sock_diag_register(&inet_diag_handler);
1134 if (err)
1135 goto out_free_nl;
1136
1137 err = sock_diag_register(&inet6_diag_handler);
1138 if (err)
1139 goto out_free_inet;
1140
1141 sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1142out:
1143 return err;
1144
1145out_free_inet:
1146 sock_diag_unregister(&inet_diag_handler);
1147out_free_nl:
1148 kfree(inet_diag_table);
1149 goto out;
1150}
1151
1152static void __exit inet_diag_exit(void)
1153{
1154 sock_diag_unregister(&inet6_diag_handler);
1155 sock_diag_unregister(&inet_diag_handler);
1156 sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1157 kfree(inet_diag_table);
1158}
1159
1160module_init(inet_diag_init);
1161module_exit(inet_diag_exit);
1162MODULE_LICENSE("GPL");
1163MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1164MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);