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// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * inet_diag.c Module for monitoring INET transport protocols sockets.
4 *
5 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
6 */
7
8#include <linux/kernel.h>
9#include <linux/module.h>
10#include <linux/types.h>
11#include <linux/fcntl.h>
12#include <linux/random.h>
13#include <linux/slab.h>
14#include <linux/cache.h>
15#include <linux/init.h>
16#include <linux/time.h>
17
18#include <net/icmp.h>
19#include <net/tcp.h>
20#include <net/ipv6.h>
21#include <net/inet_common.h>
22#include <net/inet_connection_sock.h>
23#include <net/inet_hashtables.h>
24#include <net/inet_timewait_sock.h>
25#include <net/inet6_hashtables.h>
26#include <net/bpf_sk_storage.h>
27#include <net/netlink.h>
28
29#include <linux/inet.h>
30#include <linux/stddef.h>
31
32#include <linux/inet_diag.h>
33#include <linux/sock_diag.h>
34
35static const struct inet_diag_handler **inet_diag_table;
36
37struct inet_diag_entry {
38 const __be32 *saddr;
39 const __be32 *daddr;
40 u16 sport;
41 u16 dport;
42 u16 family;
43 u16 userlocks;
44 u32 ifindex;
45 u32 mark;
46#ifdef CONFIG_SOCK_CGROUP_DATA
47 u64 cgroup_id;
48#endif
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 (proto < 0 || proto >= IPPROTO_MAX) {
56 mutex_lock(&inet_diag_table_mutex);
57 return ERR_PTR(-ENOENT);
58 }
59
60 if (!inet_diag_table[proto])
61 sock_load_diag_module(AF_INET, proto);
62
63 mutex_lock(&inet_diag_table_mutex);
64 if (!inet_diag_table[proto])
65 return ERR_PTR(-ENOENT);
66
67 return inet_diag_table[proto];
68}
69
70static void inet_diag_unlock_handler(const struct inet_diag_handler *handler)
71{
72 mutex_unlock(&inet_diag_table_mutex);
73}
74
75void inet_diag_msg_common_fill(struct inet_diag_msg *r, struct sock *sk)
76{
77 r->idiag_family = sk->sk_family;
78
79 r->id.idiag_sport = htons(sk->sk_num);
80 r->id.idiag_dport = sk->sk_dport;
81 r->id.idiag_if = sk->sk_bound_dev_if;
82 sock_diag_save_cookie(sk, r->id.idiag_cookie);
83
84#if IS_ENABLED(CONFIG_IPV6)
85 if (sk->sk_family == AF_INET6) {
86 *(struct in6_addr *)r->id.idiag_src = sk->sk_v6_rcv_saddr;
87 *(struct in6_addr *)r->id.idiag_dst = sk->sk_v6_daddr;
88 } else
89#endif
90 {
91 memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src));
92 memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst));
93
94 r->id.idiag_src[0] = sk->sk_rcv_saddr;
95 r->id.idiag_dst[0] = sk->sk_daddr;
96 }
97}
98EXPORT_SYMBOL_GPL(inet_diag_msg_common_fill);
99
100static size_t inet_sk_attr_size(struct sock *sk,
101 const struct inet_diag_req_v2 *req,
102 bool net_admin)
103{
104 const struct inet_diag_handler *handler;
105 size_t aux = 0;
106
107 handler = inet_diag_table[req->sdiag_protocol];
108 if (handler && handler->idiag_get_aux_size)
109 aux = handler->idiag_get_aux_size(sk, net_admin);
110
111 return nla_total_size(sizeof(struct tcp_info))
112 + nla_total_size(sizeof(struct inet_diag_msg))
113 + inet_diag_msg_attrs_size()
114 + nla_total_size(sizeof(struct inet_diag_meminfo))
115 + nla_total_size(SK_MEMINFO_VARS * sizeof(u32))
116 + nla_total_size(TCP_CA_NAME_MAX)
117 + nla_total_size(sizeof(struct tcpvegas_info))
118 + aux
119 + 64;
120}
121
122int inet_diag_msg_attrs_fill(struct sock *sk, struct sk_buff *skb,
123 struct inet_diag_msg *r, int ext,
124 struct user_namespace *user_ns,
125 bool net_admin)
126{
127 const struct inet_sock *inet = inet_sk(sk);
128 struct inet_diag_sockopt inet_sockopt;
129
130 if (nla_put_u8(skb, INET_DIAG_SHUTDOWN, sk->sk_shutdown))
131 goto errout;
132
133 /* IPv6 dual-stack sockets use inet->tos for IPv4 connections,
134 * hence this needs to be included regardless of socket family.
135 */
136 if (ext & (1 << (INET_DIAG_TOS - 1)))
137 if (nla_put_u8(skb, INET_DIAG_TOS, READ_ONCE(inet->tos)) < 0)
138 goto errout;
139
140#if IS_ENABLED(CONFIG_IPV6)
141 if (r->idiag_family == AF_INET6) {
142 if (ext & (1 << (INET_DIAG_TCLASS - 1)))
143 if (nla_put_u8(skb, INET_DIAG_TCLASS,
144 inet6_sk(sk)->tclass) < 0)
145 goto errout;
146
147 if (((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
148 nla_put_u8(skb, INET_DIAG_SKV6ONLY, ipv6_only_sock(sk)))
149 goto errout;
150 }
151#endif
152
153 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK, READ_ONCE(sk->sk_mark)))
154 goto errout;
155
156 if (ext & (1 << (INET_DIAG_CLASS_ID - 1)) ||
157 ext & (1 << (INET_DIAG_TCLASS - 1))) {
158 u32 classid = 0;
159
160#ifdef CONFIG_SOCK_CGROUP_DATA
161 classid = sock_cgroup_classid(&sk->sk_cgrp_data);
162#endif
163 /* Fallback to socket priority if class id isn't set.
164 * Classful qdiscs use it as direct reference to class.
165 * For cgroup2 classid is always zero.
166 */
167 if (!classid)
168 classid = READ_ONCE(sk->sk_priority);
169
170 if (nla_put_u32(skb, INET_DIAG_CLASS_ID, classid))
171 goto errout;
172 }
173
174#ifdef CONFIG_SOCK_CGROUP_DATA
175 if (nla_put_u64_64bit(skb, INET_DIAG_CGROUP_ID,
176 cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)),
177 INET_DIAG_PAD))
178 goto errout;
179#endif
180
181 r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk));
182 r->idiag_inode = sock_i_ino(sk);
183
184 memset(&inet_sockopt, 0, sizeof(inet_sockopt));
185 inet_sockopt.recverr = inet_test_bit(RECVERR, sk);
186 inet_sockopt.is_icsk = inet_test_bit(IS_ICSK, sk);
187 inet_sockopt.freebind = inet_test_bit(FREEBIND, sk);
188 inet_sockopt.hdrincl = inet_test_bit(HDRINCL, sk);
189 inet_sockopt.mc_loop = inet_test_bit(MC_LOOP, sk);
190 inet_sockopt.transparent = inet_test_bit(TRANSPARENT, sk);
191 inet_sockopt.mc_all = inet_test_bit(MC_ALL, sk);
192 inet_sockopt.nodefrag = inet_test_bit(NODEFRAG, sk);
193 inet_sockopt.bind_address_no_port = inet_test_bit(BIND_ADDRESS_NO_PORT, sk);
194 inet_sockopt.recverr_rfc4884 = inet_test_bit(RECVERR_RFC4884, sk);
195 inet_sockopt.defer_connect = inet_test_bit(DEFER_CONNECT, sk);
196 if (nla_put(skb, INET_DIAG_SOCKOPT, sizeof(inet_sockopt),
197 &inet_sockopt))
198 goto errout;
199
200 return 0;
201errout:
202 return 1;
203}
204EXPORT_SYMBOL_GPL(inet_diag_msg_attrs_fill);
205
206static int inet_diag_parse_attrs(const struct nlmsghdr *nlh, int hdrlen,
207 struct nlattr **req_nlas)
208{
209 struct nlattr *nla;
210 int remaining;
211
212 nlmsg_for_each_attr(nla, nlh, hdrlen, remaining) {
213 int type = nla_type(nla);
214
215 if (type == INET_DIAG_REQ_PROTOCOL && nla_len(nla) != sizeof(u32))
216 return -EINVAL;
217
218 if (type < __INET_DIAG_REQ_MAX)
219 req_nlas[type] = nla;
220 }
221 return 0;
222}
223
224static int inet_diag_get_protocol(const struct inet_diag_req_v2 *req,
225 const struct inet_diag_dump_data *data)
226{
227 if (data->req_nlas[INET_DIAG_REQ_PROTOCOL])
228 return nla_get_u32(data->req_nlas[INET_DIAG_REQ_PROTOCOL]);
229 return req->sdiag_protocol;
230}
231
232#define MAX_DUMP_ALLOC_SIZE (KMALLOC_MAX_SIZE - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
233
234int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk,
235 struct sk_buff *skb, struct netlink_callback *cb,
236 const struct inet_diag_req_v2 *req,
237 u16 nlmsg_flags, bool net_admin)
238{
239 const struct tcp_congestion_ops *ca_ops;
240 const struct inet_diag_handler *handler;
241 struct inet_diag_dump_data *cb_data;
242 int ext = req->idiag_ext;
243 struct inet_diag_msg *r;
244 struct nlmsghdr *nlh;
245 struct nlattr *attr;
246 void *info = NULL;
247
248 cb_data = cb->data;
249 handler = inet_diag_table[inet_diag_get_protocol(req, cb_data)];
250 BUG_ON(!handler);
251
252 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
253 cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
254 if (!nlh)
255 return -EMSGSIZE;
256
257 r = nlmsg_data(nlh);
258 BUG_ON(!sk_fullsock(sk));
259
260 inet_diag_msg_common_fill(r, sk);
261 r->idiag_state = sk->sk_state;
262 r->idiag_timer = 0;
263 r->idiag_retrans = 0;
264 r->idiag_expires = 0;
265
266 if (inet_diag_msg_attrs_fill(sk, skb, r, ext,
267 sk_user_ns(NETLINK_CB(cb->skb).sk),
268 net_admin))
269 goto errout;
270
271 if (ext & (1 << (INET_DIAG_MEMINFO - 1))) {
272 struct inet_diag_meminfo minfo = {
273 .idiag_rmem = sk_rmem_alloc_get(sk),
274 .idiag_wmem = READ_ONCE(sk->sk_wmem_queued),
275 .idiag_fmem = sk_forward_alloc_get(sk),
276 .idiag_tmem = sk_wmem_alloc_get(sk),
277 };
278
279 if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0)
280 goto errout;
281 }
282
283 if (ext & (1 << (INET_DIAG_SKMEMINFO - 1)))
284 if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO))
285 goto errout;
286
287 /*
288 * RAW sockets might have user-defined protocols assigned,
289 * so report the one supplied on socket creation.
290 */
291 if (sk->sk_type == SOCK_RAW) {
292 if (nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))
293 goto errout;
294 }
295
296 if (!icsk) {
297 handler->idiag_get_info(sk, r, NULL);
298 goto out;
299 }
300
301 if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
302 icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
303 icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
304 r->idiag_timer = 1;
305 r->idiag_retrans = icsk->icsk_retransmits;
306 r->idiag_expires =
307 jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
308 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
309 r->idiag_timer = 4;
310 r->idiag_retrans = icsk->icsk_probes_out;
311 r->idiag_expires =
312 jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
313 } else if (timer_pending(&sk->sk_timer)) {
314 r->idiag_timer = 2;
315 r->idiag_retrans = icsk->icsk_probes_out;
316 r->idiag_expires =
317 jiffies_delta_to_msecs(sk->sk_timer.expires - jiffies);
318 }
319
320 if ((ext & (1 << (INET_DIAG_INFO - 1))) && handler->idiag_info_size) {
321 attr = nla_reserve_64bit(skb, INET_DIAG_INFO,
322 handler->idiag_info_size,
323 INET_DIAG_PAD);
324 if (!attr)
325 goto errout;
326
327 info = nla_data(attr);
328 }
329
330 if (ext & (1 << (INET_DIAG_CONG - 1))) {
331 int err = 0;
332
333 rcu_read_lock();
334 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
335 if (ca_ops)
336 err = nla_put_string(skb, INET_DIAG_CONG, ca_ops->name);
337 rcu_read_unlock();
338 if (err < 0)
339 goto errout;
340 }
341
342 handler->idiag_get_info(sk, r, info);
343
344 if (ext & (1 << (INET_DIAG_INFO - 1)) && handler->idiag_get_aux)
345 if (handler->idiag_get_aux(sk, net_admin, skb) < 0)
346 goto errout;
347
348 if (sk->sk_state < TCP_TIME_WAIT) {
349 union tcp_cc_info info;
350 size_t sz = 0;
351 int attr;
352
353 rcu_read_lock();
354 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
355 if (ca_ops && ca_ops->get_info)
356 sz = ca_ops->get_info(sk, ext, &attr, &info);
357 rcu_read_unlock();
358 if (sz && nla_put(skb, attr, sz, &info) < 0)
359 goto errout;
360 }
361
362 /* Keep it at the end for potential retry with a larger skb,
363 * or else do best-effort fitting, which is only done for the
364 * first_nlmsg.
365 */
366 if (cb_data->bpf_stg_diag) {
367 bool first_nlmsg = ((unsigned char *)nlh == skb->data);
368 unsigned int prev_min_dump_alloc;
369 unsigned int total_nla_size = 0;
370 unsigned int msg_len;
371 int err;
372
373 msg_len = skb_tail_pointer(skb) - (unsigned char *)nlh;
374 err = bpf_sk_storage_diag_put(cb_data->bpf_stg_diag, sk, skb,
375 INET_DIAG_SK_BPF_STORAGES,
376 &total_nla_size);
377
378 if (!err)
379 goto out;
380
381 total_nla_size += msg_len;
382 prev_min_dump_alloc = cb->min_dump_alloc;
383 if (total_nla_size > prev_min_dump_alloc)
384 cb->min_dump_alloc = min_t(u32, total_nla_size,
385 MAX_DUMP_ALLOC_SIZE);
386
387 if (!first_nlmsg)
388 goto errout;
389
390 if (cb->min_dump_alloc > prev_min_dump_alloc)
391 /* Retry with pskb_expand_head() with
392 * __GFP_DIRECT_RECLAIM
393 */
394 goto errout;
395
396 WARN_ON_ONCE(total_nla_size <= prev_min_dump_alloc);
397
398 /* Send what we have for this sk
399 * and move on to the next sk in the following
400 * dump()
401 */
402 }
403
404out:
405 nlmsg_end(skb, nlh);
406 return 0;
407
408errout:
409 nlmsg_cancel(skb, nlh);
410 return -EMSGSIZE;
411}
412EXPORT_SYMBOL_GPL(inet_sk_diag_fill);
413
414static int inet_twsk_diag_fill(struct sock *sk,
415 struct sk_buff *skb,
416 struct netlink_callback *cb,
417 u16 nlmsg_flags, bool net_admin)
418{
419 struct inet_timewait_sock *tw = inet_twsk(sk);
420 struct inet_diag_msg *r;
421 struct nlmsghdr *nlh;
422 long tmo;
423
424 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
425 cb->nlh->nlmsg_seq, cb->nlh->nlmsg_type,
426 sizeof(*r), nlmsg_flags);
427 if (!nlh)
428 return -EMSGSIZE;
429
430 r = nlmsg_data(nlh);
431 BUG_ON(tw->tw_state != TCP_TIME_WAIT);
432
433 inet_diag_msg_common_fill(r, sk);
434 r->idiag_retrans = 0;
435
436 r->idiag_state = tw->tw_substate;
437 r->idiag_timer = 3;
438 tmo = tw->tw_timer.expires - jiffies;
439 r->idiag_expires = jiffies_delta_to_msecs(tmo);
440 r->idiag_rqueue = 0;
441 r->idiag_wqueue = 0;
442 r->idiag_uid = 0;
443 r->idiag_inode = 0;
444
445 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
446 tw->tw_mark)) {
447 nlmsg_cancel(skb, nlh);
448 return -EMSGSIZE;
449 }
450
451 nlmsg_end(skb, nlh);
452 return 0;
453}
454
455static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb,
456 struct netlink_callback *cb,
457 u16 nlmsg_flags, bool net_admin)
458{
459 struct request_sock *reqsk = inet_reqsk(sk);
460 struct inet_diag_msg *r;
461 struct nlmsghdr *nlh;
462 long tmo;
463
464 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
465 cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
466 if (!nlh)
467 return -EMSGSIZE;
468
469 r = nlmsg_data(nlh);
470 inet_diag_msg_common_fill(r, sk);
471 r->idiag_state = TCP_SYN_RECV;
472 r->idiag_timer = 1;
473 r->idiag_retrans = reqsk->num_retrans;
474
475 BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) !=
476 offsetof(struct sock, sk_cookie));
477
478 tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies;
479 r->idiag_expires = jiffies_delta_to_msecs(tmo);
480 r->idiag_rqueue = 0;
481 r->idiag_wqueue = 0;
482 r->idiag_uid = 0;
483 r->idiag_inode = 0;
484
485 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
486 inet_rsk(reqsk)->ir_mark)) {
487 nlmsg_cancel(skb, nlh);
488 return -EMSGSIZE;
489 }
490
491 nlmsg_end(skb, nlh);
492 return 0;
493}
494
495static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
496 struct netlink_callback *cb,
497 const struct inet_diag_req_v2 *r,
498 u16 nlmsg_flags, bool net_admin)
499{
500 if (sk->sk_state == TCP_TIME_WAIT)
501 return inet_twsk_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
502
503 if (sk->sk_state == TCP_NEW_SYN_RECV)
504 return inet_req_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
505
506 return inet_sk_diag_fill(sk, inet_csk(sk), skb, cb, r, nlmsg_flags,
507 net_admin);
508}
509
510struct sock *inet_diag_find_one_icsk(struct net *net,
511 struct inet_hashinfo *hashinfo,
512 const struct inet_diag_req_v2 *req)
513{
514 struct sock *sk;
515
516 rcu_read_lock();
517 if (req->sdiag_family == AF_INET)
518 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0],
519 req->id.idiag_dport, req->id.idiag_src[0],
520 req->id.idiag_sport, req->id.idiag_if);
521#if IS_ENABLED(CONFIG_IPV6)
522 else if (req->sdiag_family == AF_INET6) {
523 if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) &&
524 ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src))
525 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3],
526 req->id.idiag_dport, req->id.idiag_src[3],
527 req->id.idiag_sport, req->id.idiag_if);
528 else
529 sk = inet6_lookup(net, hashinfo, NULL, 0,
530 (struct in6_addr *)req->id.idiag_dst,
531 req->id.idiag_dport,
532 (struct in6_addr *)req->id.idiag_src,
533 req->id.idiag_sport,
534 req->id.idiag_if);
535 }
536#endif
537 else {
538 rcu_read_unlock();
539 return ERR_PTR(-EINVAL);
540 }
541 rcu_read_unlock();
542 if (!sk)
543 return ERR_PTR(-ENOENT);
544
545 if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) {
546 sock_gen_put(sk);
547 return ERR_PTR(-ENOENT);
548 }
549
550 return sk;
551}
552EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk);
553
554int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo,
555 struct netlink_callback *cb,
556 const struct inet_diag_req_v2 *req)
557{
558 struct sk_buff *in_skb = cb->skb;
559 bool net_admin = netlink_net_capable(in_skb, CAP_NET_ADMIN);
560 struct net *net = sock_net(in_skb->sk);
561 struct sk_buff *rep;
562 struct sock *sk;
563 int err;
564
565 sk = inet_diag_find_one_icsk(net, hashinfo, req);
566 if (IS_ERR(sk))
567 return PTR_ERR(sk);
568
569 rep = nlmsg_new(inet_sk_attr_size(sk, req, net_admin), GFP_KERNEL);
570 if (!rep) {
571 err = -ENOMEM;
572 goto out;
573 }
574
575 err = sk_diag_fill(sk, rep, cb, req, 0, net_admin);
576 if (err < 0) {
577 WARN_ON(err == -EMSGSIZE);
578 nlmsg_free(rep);
579 goto out;
580 }
581 err = nlmsg_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid);
582
583out:
584 if (sk)
585 sock_gen_put(sk);
586
587 return err;
588}
589EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
590
591static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb,
592 const struct nlmsghdr *nlh,
593 int hdrlen,
594 const struct inet_diag_req_v2 *req)
595{
596 const struct inet_diag_handler *handler;
597 struct inet_diag_dump_data dump_data;
598 int err, protocol;
599
600 memset(&dump_data, 0, sizeof(dump_data));
601 err = inet_diag_parse_attrs(nlh, hdrlen, dump_data.req_nlas);
602 if (err)
603 return err;
604
605 protocol = inet_diag_get_protocol(req, &dump_data);
606
607 handler = inet_diag_lock_handler(protocol);
608 if (IS_ERR(handler)) {
609 err = PTR_ERR(handler);
610 } else if (cmd == SOCK_DIAG_BY_FAMILY) {
611 struct netlink_callback cb = {
612 .nlh = nlh,
613 .skb = in_skb,
614 .data = &dump_data,
615 };
616 err = handler->dump_one(&cb, req);
617 } else if (cmd == SOCK_DESTROY && handler->destroy) {
618 err = handler->destroy(in_skb, req);
619 } else {
620 err = -EOPNOTSUPP;
621 }
622 inet_diag_unlock_handler(handler);
623
624 return err;
625}
626
627static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
628{
629 int words = bits >> 5;
630
631 bits &= 0x1f;
632
633 if (words) {
634 if (memcmp(a1, a2, words << 2))
635 return 0;
636 }
637 if (bits) {
638 __be32 w1, w2;
639 __be32 mask;
640
641 w1 = a1[words];
642 w2 = a2[words];
643
644 mask = htonl((0xffffffff) << (32 - bits));
645
646 if ((w1 ^ w2) & mask)
647 return 0;
648 }
649
650 return 1;
651}
652
653static int inet_diag_bc_run(const struct nlattr *_bc,
654 const struct inet_diag_entry *entry)
655{
656 const void *bc = nla_data(_bc);
657 int len = nla_len(_bc);
658
659 while (len > 0) {
660 int yes = 1;
661 const struct inet_diag_bc_op *op = bc;
662
663 switch (op->code) {
664 case INET_DIAG_BC_NOP:
665 break;
666 case INET_DIAG_BC_JMP:
667 yes = 0;
668 break;
669 case INET_DIAG_BC_S_EQ:
670 yes = entry->sport == op[1].no;
671 break;
672 case INET_DIAG_BC_S_GE:
673 yes = entry->sport >= op[1].no;
674 break;
675 case INET_DIAG_BC_S_LE:
676 yes = entry->sport <= op[1].no;
677 break;
678 case INET_DIAG_BC_D_EQ:
679 yes = entry->dport == op[1].no;
680 break;
681 case INET_DIAG_BC_D_GE:
682 yes = entry->dport >= op[1].no;
683 break;
684 case INET_DIAG_BC_D_LE:
685 yes = entry->dport <= op[1].no;
686 break;
687 case INET_DIAG_BC_AUTO:
688 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
689 break;
690 case INET_DIAG_BC_S_COND:
691 case INET_DIAG_BC_D_COND: {
692 const struct inet_diag_hostcond *cond;
693 const __be32 *addr;
694
695 cond = (const struct inet_diag_hostcond *)(op + 1);
696 if (cond->port != -1 &&
697 cond->port != (op->code == INET_DIAG_BC_S_COND ?
698 entry->sport : entry->dport)) {
699 yes = 0;
700 break;
701 }
702
703 if (op->code == INET_DIAG_BC_S_COND)
704 addr = entry->saddr;
705 else
706 addr = entry->daddr;
707
708 if (cond->family != AF_UNSPEC &&
709 cond->family != entry->family) {
710 if (entry->family == AF_INET6 &&
711 cond->family == AF_INET) {
712 if (addr[0] == 0 && addr[1] == 0 &&
713 addr[2] == htonl(0xffff) &&
714 bitstring_match(addr + 3,
715 cond->addr,
716 cond->prefix_len))
717 break;
718 }
719 yes = 0;
720 break;
721 }
722
723 if (cond->prefix_len == 0)
724 break;
725 if (bitstring_match(addr, cond->addr,
726 cond->prefix_len))
727 break;
728 yes = 0;
729 break;
730 }
731 case INET_DIAG_BC_DEV_COND: {
732 u32 ifindex;
733
734 ifindex = *((const u32 *)(op + 1));
735 if (ifindex != entry->ifindex)
736 yes = 0;
737 break;
738 }
739 case INET_DIAG_BC_MARK_COND: {
740 struct inet_diag_markcond *cond;
741
742 cond = (struct inet_diag_markcond *)(op + 1);
743 if ((entry->mark & cond->mask) != cond->mark)
744 yes = 0;
745 break;
746 }
747#ifdef CONFIG_SOCK_CGROUP_DATA
748 case INET_DIAG_BC_CGROUP_COND: {
749 u64 cgroup_id;
750
751 cgroup_id = get_unaligned((const u64 *)(op + 1));
752 if (cgroup_id != entry->cgroup_id)
753 yes = 0;
754 break;
755 }
756#endif
757 }
758
759 if (yes) {
760 len -= op->yes;
761 bc += op->yes;
762 } else {
763 len -= op->no;
764 bc += op->no;
765 }
766 }
767 return len == 0;
768}
769
770/* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV)
771 */
772static void entry_fill_addrs(struct inet_diag_entry *entry,
773 const struct sock *sk)
774{
775#if IS_ENABLED(CONFIG_IPV6)
776 if (sk->sk_family == AF_INET6) {
777 entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32;
778 entry->daddr = sk->sk_v6_daddr.s6_addr32;
779 } else
780#endif
781 {
782 entry->saddr = &sk->sk_rcv_saddr;
783 entry->daddr = &sk->sk_daddr;
784 }
785}
786
787int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
788{
789 struct inet_sock *inet = inet_sk(sk);
790 struct inet_diag_entry entry;
791
792 if (!bc)
793 return 1;
794
795 entry.family = sk->sk_family;
796 entry_fill_addrs(&entry, sk);
797 entry.sport = inet->inet_num;
798 entry.dport = ntohs(inet->inet_dport);
799 entry.ifindex = sk->sk_bound_dev_if;
800 entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0;
801 if (sk_fullsock(sk))
802 entry.mark = READ_ONCE(sk->sk_mark);
803 else if (sk->sk_state == TCP_NEW_SYN_RECV)
804 entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark;
805 else if (sk->sk_state == TCP_TIME_WAIT)
806 entry.mark = inet_twsk(sk)->tw_mark;
807 else
808 entry.mark = 0;
809#ifdef CONFIG_SOCK_CGROUP_DATA
810 entry.cgroup_id = sk_fullsock(sk) ?
811 cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)) : 0;
812#endif
813
814 return inet_diag_bc_run(bc, &entry);
815}
816EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
817
818static int valid_cc(const void *bc, int len, int cc)
819{
820 while (len >= 0) {
821 const struct inet_diag_bc_op *op = bc;
822
823 if (cc > len)
824 return 0;
825 if (cc == len)
826 return 1;
827 if (op->yes < 4 || op->yes & 3)
828 return 0;
829 len -= op->yes;
830 bc += op->yes;
831 }
832 return 0;
833}
834
835/* data is u32 ifindex */
836static bool valid_devcond(const struct inet_diag_bc_op *op, int len,
837 int *min_len)
838{
839 /* Check ifindex space. */
840 *min_len += sizeof(u32);
841 if (len < *min_len)
842 return false;
843
844 return true;
845}
846/* Validate an inet_diag_hostcond. */
847static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
848 int *min_len)
849{
850 struct inet_diag_hostcond *cond;
851 int addr_len;
852
853 /* Check hostcond space. */
854 *min_len += sizeof(struct inet_diag_hostcond);
855 if (len < *min_len)
856 return false;
857 cond = (struct inet_diag_hostcond *)(op + 1);
858
859 /* Check address family and address length. */
860 switch (cond->family) {
861 case AF_UNSPEC:
862 addr_len = 0;
863 break;
864 case AF_INET:
865 addr_len = sizeof(struct in_addr);
866 break;
867 case AF_INET6:
868 addr_len = sizeof(struct in6_addr);
869 break;
870 default:
871 return false;
872 }
873 *min_len += addr_len;
874 if (len < *min_len)
875 return false;
876
877 /* Check prefix length (in bits) vs address length (in bytes). */
878 if (cond->prefix_len > 8 * addr_len)
879 return false;
880
881 return true;
882}
883
884/* Validate a port comparison operator. */
885static bool valid_port_comparison(const struct inet_diag_bc_op *op,
886 int len, int *min_len)
887{
888 /* Port comparisons put the port in a follow-on inet_diag_bc_op. */
889 *min_len += sizeof(struct inet_diag_bc_op);
890 if (len < *min_len)
891 return false;
892 return true;
893}
894
895static bool valid_markcond(const struct inet_diag_bc_op *op, int len,
896 int *min_len)
897{
898 *min_len += sizeof(struct inet_diag_markcond);
899 return len >= *min_len;
900}
901
902#ifdef CONFIG_SOCK_CGROUP_DATA
903static bool valid_cgroupcond(const struct inet_diag_bc_op *op, int len,
904 int *min_len)
905{
906 *min_len += sizeof(u64);
907 return len >= *min_len;
908}
909#endif
910
911static int inet_diag_bc_audit(const struct nlattr *attr,
912 const struct sk_buff *skb)
913{
914 bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN);
915 const void *bytecode, *bc;
916 int bytecode_len, len;
917
918 if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op))
919 return -EINVAL;
920
921 bytecode = bc = nla_data(attr);
922 len = bytecode_len = nla_len(attr);
923
924 while (len > 0) {
925 int min_len = sizeof(struct inet_diag_bc_op);
926 const struct inet_diag_bc_op *op = bc;
927
928 switch (op->code) {
929 case INET_DIAG_BC_S_COND:
930 case INET_DIAG_BC_D_COND:
931 if (!valid_hostcond(bc, len, &min_len))
932 return -EINVAL;
933 break;
934 case INET_DIAG_BC_DEV_COND:
935 if (!valid_devcond(bc, len, &min_len))
936 return -EINVAL;
937 break;
938 case INET_DIAG_BC_S_EQ:
939 case INET_DIAG_BC_S_GE:
940 case INET_DIAG_BC_S_LE:
941 case INET_DIAG_BC_D_EQ:
942 case INET_DIAG_BC_D_GE:
943 case INET_DIAG_BC_D_LE:
944 if (!valid_port_comparison(bc, len, &min_len))
945 return -EINVAL;
946 break;
947 case INET_DIAG_BC_MARK_COND:
948 if (!net_admin)
949 return -EPERM;
950 if (!valid_markcond(bc, len, &min_len))
951 return -EINVAL;
952 break;
953#ifdef CONFIG_SOCK_CGROUP_DATA
954 case INET_DIAG_BC_CGROUP_COND:
955 if (!valid_cgroupcond(bc, len, &min_len))
956 return -EINVAL;
957 break;
958#endif
959 case INET_DIAG_BC_AUTO:
960 case INET_DIAG_BC_JMP:
961 case INET_DIAG_BC_NOP:
962 break;
963 default:
964 return -EINVAL;
965 }
966
967 if (op->code != INET_DIAG_BC_NOP) {
968 if (op->no < min_len || op->no > len + 4 || op->no & 3)
969 return -EINVAL;
970 if (op->no < len &&
971 !valid_cc(bytecode, bytecode_len, len - op->no))
972 return -EINVAL;
973 }
974
975 if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
976 return -EINVAL;
977 bc += op->yes;
978 len -= op->yes;
979 }
980 return len == 0 ? 0 : -EINVAL;
981}
982
983static void twsk_build_assert(void)
984{
985 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) !=
986 offsetof(struct sock, sk_family));
987
988 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) !=
989 offsetof(struct inet_sock, inet_num));
990
991 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) !=
992 offsetof(struct inet_sock, inet_dport));
993
994 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) !=
995 offsetof(struct inet_sock, inet_rcv_saddr));
996
997 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) !=
998 offsetof(struct inet_sock, inet_daddr));
999
1000#if IS_ENABLED(CONFIG_IPV6)
1001 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) !=
1002 offsetof(struct sock, sk_v6_rcv_saddr));
1003
1004 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) !=
1005 offsetof(struct sock, sk_v6_daddr));
1006#endif
1007}
1008
1009void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
1010 struct netlink_callback *cb,
1011 const struct inet_diag_req_v2 *r)
1012{
1013 bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN);
1014 struct inet_diag_dump_data *cb_data = cb->data;
1015 struct net *net = sock_net(skb->sk);
1016 u32 idiag_states = r->idiag_states;
1017 int i, num, s_i, s_num;
1018 struct nlattr *bc;
1019 struct sock *sk;
1020
1021 bc = cb_data->inet_diag_nla_bc;
1022 if (idiag_states & TCPF_SYN_RECV)
1023 idiag_states |= TCPF_NEW_SYN_RECV;
1024 s_i = cb->args[1];
1025 s_num = num = cb->args[2];
1026
1027 if (cb->args[0] == 0) {
1028 if (!(idiag_states & TCPF_LISTEN) || r->id.idiag_dport)
1029 goto skip_listen_ht;
1030
1031 for (i = s_i; i <= hashinfo->lhash2_mask; i++) {
1032 struct inet_listen_hashbucket *ilb;
1033 struct hlist_nulls_node *node;
1034
1035 num = 0;
1036 ilb = &hashinfo->lhash2[i];
1037
1038 spin_lock(&ilb->lock);
1039 sk_nulls_for_each(sk, node, &ilb->nulls_head) {
1040 struct inet_sock *inet = inet_sk(sk);
1041
1042 if (!net_eq(sock_net(sk), net))
1043 continue;
1044
1045 if (num < s_num) {
1046 num++;
1047 continue;
1048 }
1049
1050 if (r->sdiag_family != AF_UNSPEC &&
1051 sk->sk_family != r->sdiag_family)
1052 goto next_listen;
1053
1054 if (r->id.idiag_sport != inet->inet_sport &&
1055 r->id.idiag_sport)
1056 goto next_listen;
1057
1058 if (!inet_diag_bc_sk(bc, sk))
1059 goto next_listen;
1060
1061 if (inet_sk_diag_fill(sk, inet_csk(sk), skb,
1062 cb, r, NLM_F_MULTI,
1063 net_admin) < 0) {
1064 spin_unlock(&ilb->lock);
1065 goto done;
1066 }
1067
1068next_listen:
1069 ++num;
1070 }
1071 spin_unlock(&ilb->lock);
1072
1073 s_num = 0;
1074 }
1075skip_listen_ht:
1076 cb->args[0] = 1;
1077 s_i = num = s_num = 0;
1078 }
1079
1080/* Process a maximum of SKARR_SZ sockets at a time when walking hash buckets
1081 * with bh disabled.
1082 */
1083#define SKARR_SZ 16
1084
1085 /* Dump bound but inactive (not listening, connecting, etc.) sockets */
1086 if (cb->args[0] == 1) {
1087 if (!(idiag_states & TCPF_BOUND_INACTIVE))
1088 goto skip_bind_ht;
1089
1090 for (i = s_i; i < hashinfo->bhash_size; i++) {
1091 struct inet_bind_hashbucket *ibb;
1092 struct inet_bind2_bucket *tb2;
1093 struct sock *sk_arr[SKARR_SZ];
1094 int num_arr[SKARR_SZ];
1095 int idx, accum, res;
1096
1097resume_bind_walk:
1098 num = 0;
1099 accum = 0;
1100 ibb = &hashinfo->bhash2[i];
1101
1102 spin_lock_bh(&ibb->lock);
1103 inet_bind_bucket_for_each(tb2, &ibb->chain) {
1104 if (!net_eq(ib2_net(tb2), net))
1105 continue;
1106
1107 sk_for_each_bound(sk, &tb2->owners) {
1108 struct inet_sock *inet = inet_sk(sk);
1109
1110 if (num < s_num)
1111 goto next_bind;
1112
1113 if (sk->sk_state != TCP_CLOSE ||
1114 !inet->inet_num)
1115 goto next_bind;
1116
1117 if (r->sdiag_family != AF_UNSPEC &&
1118 r->sdiag_family != sk->sk_family)
1119 goto next_bind;
1120
1121 if (!inet_diag_bc_sk(bc, sk))
1122 goto next_bind;
1123
1124 sock_hold(sk);
1125 num_arr[accum] = num;
1126 sk_arr[accum] = sk;
1127 if (++accum == SKARR_SZ)
1128 goto pause_bind_walk;
1129next_bind:
1130 num++;
1131 }
1132 }
1133pause_bind_walk:
1134 spin_unlock_bh(&ibb->lock);
1135
1136 res = 0;
1137 for (idx = 0; idx < accum; idx++) {
1138 if (res >= 0) {
1139 res = inet_sk_diag_fill(sk_arr[idx],
1140 NULL, skb, cb,
1141 r, NLM_F_MULTI,
1142 net_admin);
1143 if (res < 0)
1144 num = num_arr[idx];
1145 }
1146 sock_put(sk_arr[idx]);
1147 }
1148 if (res < 0)
1149 goto done;
1150
1151 cond_resched();
1152
1153 if (accum == SKARR_SZ) {
1154 s_num = num + 1;
1155 goto resume_bind_walk;
1156 }
1157
1158 s_num = 0;
1159 }
1160skip_bind_ht:
1161 cb->args[0] = 2;
1162 s_i = num = s_num = 0;
1163 }
1164
1165 if (!(idiag_states & ~TCPF_LISTEN))
1166 goto out;
1167
1168 for (i = s_i; i <= hashinfo->ehash_mask; i++) {
1169 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
1170 spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
1171 struct hlist_nulls_node *node;
1172 struct sock *sk_arr[SKARR_SZ];
1173 int num_arr[SKARR_SZ];
1174 int idx, accum, res;
1175
1176 if (hlist_nulls_empty(&head->chain))
1177 continue;
1178
1179 if (i > s_i)
1180 s_num = 0;
1181
1182next_chunk:
1183 num = 0;
1184 accum = 0;
1185 spin_lock_bh(lock);
1186 sk_nulls_for_each(sk, node, &head->chain) {
1187 int state;
1188
1189 if (!net_eq(sock_net(sk), net))
1190 continue;
1191 if (num < s_num)
1192 goto next_normal;
1193 state = (sk->sk_state == TCP_TIME_WAIT) ?
1194 inet_twsk(sk)->tw_substate : sk->sk_state;
1195 if (!(idiag_states & (1 << state)))
1196 goto next_normal;
1197 if (r->sdiag_family != AF_UNSPEC &&
1198 sk->sk_family != r->sdiag_family)
1199 goto next_normal;
1200 if (r->id.idiag_sport != htons(sk->sk_num) &&
1201 r->id.idiag_sport)
1202 goto next_normal;
1203 if (r->id.idiag_dport != sk->sk_dport &&
1204 r->id.idiag_dport)
1205 goto next_normal;
1206 twsk_build_assert();
1207
1208 if (!inet_diag_bc_sk(bc, sk))
1209 goto next_normal;
1210
1211 if (!refcount_inc_not_zero(&sk->sk_refcnt))
1212 goto next_normal;
1213
1214 num_arr[accum] = num;
1215 sk_arr[accum] = sk;
1216 if (++accum == SKARR_SZ)
1217 break;
1218next_normal:
1219 ++num;
1220 }
1221 spin_unlock_bh(lock);
1222 res = 0;
1223 for (idx = 0; idx < accum; idx++) {
1224 if (res >= 0) {
1225 res = sk_diag_fill(sk_arr[idx], skb, cb, r,
1226 NLM_F_MULTI, net_admin);
1227 if (res < 0)
1228 num = num_arr[idx];
1229 }
1230 sock_gen_put(sk_arr[idx]);
1231 }
1232 if (res < 0)
1233 break;
1234 cond_resched();
1235 if (accum == SKARR_SZ) {
1236 s_num = num + 1;
1237 goto next_chunk;
1238 }
1239 }
1240
1241done:
1242 cb->args[1] = i;
1243 cb->args[2] = num;
1244out:
1245 ;
1246}
1247EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
1248
1249static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
1250 const struct inet_diag_req_v2 *r)
1251{
1252 struct inet_diag_dump_data *cb_data = cb->data;
1253 const struct inet_diag_handler *handler;
1254 u32 prev_min_dump_alloc;
1255 int protocol, err = 0;
1256
1257 protocol = inet_diag_get_protocol(r, cb_data);
1258
1259again:
1260 prev_min_dump_alloc = cb->min_dump_alloc;
1261 handler = inet_diag_lock_handler(protocol);
1262 if (!IS_ERR(handler))
1263 handler->dump(skb, cb, r);
1264 else
1265 err = PTR_ERR(handler);
1266 inet_diag_unlock_handler(handler);
1267
1268 /* The skb is not large enough to fit one sk info and
1269 * inet_sk_diag_fill() has requested for a larger skb.
1270 */
1271 if (!skb->len && cb->min_dump_alloc > prev_min_dump_alloc) {
1272 err = pskb_expand_head(skb, 0, cb->min_dump_alloc, GFP_KERNEL);
1273 if (!err)
1274 goto again;
1275 }
1276
1277 return err ? : skb->len;
1278}
1279
1280static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
1281{
1282 return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh));
1283}
1284
1285static int __inet_diag_dump_start(struct netlink_callback *cb, int hdrlen)
1286{
1287 const struct nlmsghdr *nlh = cb->nlh;
1288 struct inet_diag_dump_data *cb_data;
1289 struct sk_buff *skb = cb->skb;
1290 struct nlattr *nla;
1291 int err;
1292
1293 cb_data = kzalloc(sizeof(*cb_data), GFP_KERNEL);
1294 if (!cb_data)
1295 return -ENOMEM;
1296
1297 err = inet_diag_parse_attrs(nlh, hdrlen, cb_data->req_nlas);
1298 if (err) {
1299 kfree(cb_data);
1300 return err;
1301 }
1302 nla = cb_data->inet_diag_nla_bc;
1303 if (nla) {
1304 err = inet_diag_bc_audit(nla, skb);
1305 if (err) {
1306 kfree(cb_data);
1307 return err;
1308 }
1309 }
1310
1311 nla = cb_data->inet_diag_nla_bpf_stgs;
1312 if (nla) {
1313 struct bpf_sk_storage_diag *bpf_stg_diag;
1314
1315 bpf_stg_diag = bpf_sk_storage_diag_alloc(nla);
1316 if (IS_ERR(bpf_stg_diag)) {
1317 kfree(cb_data);
1318 return PTR_ERR(bpf_stg_diag);
1319 }
1320 cb_data->bpf_stg_diag = bpf_stg_diag;
1321 }
1322
1323 cb->data = cb_data;
1324 return 0;
1325}
1326
1327static int inet_diag_dump_start(struct netlink_callback *cb)
1328{
1329 return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req_v2));
1330}
1331
1332static int inet_diag_dump_start_compat(struct netlink_callback *cb)
1333{
1334 return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req));
1335}
1336
1337static int inet_diag_dump_done(struct netlink_callback *cb)
1338{
1339 struct inet_diag_dump_data *cb_data = cb->data;
1340
1341 bpf_sk_storage_diag_free(cb_data->bpf_stg_diag);
1342 kfree(cb->data);
1343
1344 return 0;
1345}
1346
1347static int inet_diag_type2proto(int type)
1348{
1349 switch (type) {
1350 case TCPDIAG_GETSOCK:
1351 return IPPROTO_TCP;
1352 case DCCPDIAG_GETSOCK:
1353 return IPPROTO_DCCP;
1354 default:
1355 return 0;
1356 }
1357}
1358
1359static int inet_diag_dump_compat(struct sk_buff *skb,
1360 struct netlink_callback *cb)
1361{
1362 struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1363 struct inet_diag_req_v2 req;
1364
1365 req.sdiag_family = AF_UNSPEC; /* compatibility */
1366 req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1367 req.idiag_ext = rc->idiag_ext;
1368 req.idiag_states = rc->idiag_states;
1369 req.id = rc->id;
1370
1371 return __inet_diag_dump(skb, cb, &req);
1372}
1373
1374static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1375 const struct nlmsghdr *nlh)
1376{
1377 struct inet_diag_req *rc = nlmsg_data(nlh);
1378 struct inet_diag_req_v2 req;
1379
1380 req.sdiag_family = rc->idiag_family;
1381 req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1382 req.idiag_ext = rc->idiag_ext;
1383 req.idiag_states = rc->idiag_states;
1384 req.id = rc->id;
1385
1386 return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh,
1387 sizeof(struct inet_diag_req), &req);
1388}
1389
1390static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1391{
1392 int hdrlen = sizeof(struct inet_diag_req);
1393 struct net *net = sock_net(skb->sk);
1394
1395 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1396 nlmsg_len(nlh) < hdrlen)
1397 return -EINVAL;
1398
1399 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1400 struct netlink_dump_control c = {
1401 .start = inet_diag_dump_start_compat,
1402 .done = inet_diag_dump_done,
1403 .dump = inet_diag_dump_compat,
1404 };
1405 return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1406 }
1407
1408 return inet_diag_get_exact_compat(skb, nlh);
1409}
1410
1411static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h)
1412{
1413 int hdrlen = sizeof(struct inet_diag_req_v2);
1414 struct net *net = sock_net(skb->sk);
1415
1416 if (nlmsg_len(h) < hdrlen)
1417 return -EINVAL;
1418
1419 if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY &&
1420 h->nlmsg_flags & NLM_F_DUMP) {
1421 struct netlink_dump_control c = {
1422 .start = inet_diag_dump_start,
1423 .done = inet_diag_dump_done,
1424 .dump = inet_diag_dump,
1425 };
1426 return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1427 }
1428
1429 return inet_diag_cmd_exact(h->nlmsg_type, skb, h, hdrlen,
1430 nlmsg_data(h));
1431}
1432
1433static
1434int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk)
1435{
1436 const struct inet_diag_handler *handler;
1437 struct nlmsghdr *nlh;
1438 struct nlattr *attr;
1439 struct inet_diag_msg *r;
1440 void *info = NULL;
1441 int err = 0;
1442
1443 nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0);
1444 if (!nlh)
1445 return -ENOMEM;
1446
1447 r = nlmsg_data(nlh);
1448 memset(r, 0, sizeof(*r));
1449 inet_diag_msg_common_fill(r, sk);
1450 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM)
1451 r->id.idiag_sport = inet_sk(sk)->inet_sport;
1452 r->idiag_state = sk->sk_state;
1453
1454 if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) {
1455 nlmsg_cancel(skb, nlh);
1456 return err;
1457 }
1458
1459 handler = inet_diag_lock_handler(sk->sk_protocol);
1460 if (IS_ERR(handler)) {
1461 inet_diag_unlock_handler(handler);
1462 nlmsg_cancel(skb, nlh);
1463 return PTR_ERR(handler);
1464 }
1465
1466 attr = handler->idiag_info_size
1467 ? nla_reserve_64bit(skb, INET_DIAG_INFO,
1468 handler->idiag_info_size,
1469 INET_DIAG_PAD)
1470 : NULL;
1471 if (attr)
1472 info = nla_data(attr);
1473
1474 handler->idiag_get_info(sk, r, info);
1475 inet_diag_unlock_handler(handler);
1476
1477 nlmsg_end(skb, nlh);
1478 return 0;
1479}
1480
1481static const struct sock_diag_handler inet_diag_handler = {
1482 .family = AF_INET,
1483 .dump = inet_diag_handler_cmd,
1484 .get_info = inet_diag_handler_get_info,
1485 .destroy = inet_diag_handler_cmd,
1486};
1487
1488static const struct sock_diag_handler inet6_diag_handler = {
1489 .family = AF_INET6,
1490 .dump = inet_diag_handler_cmd,
1491 .get_info = inet_diag_handler_get_info,
1492 .destroy = inet_diag_handler_cmd,
1493};
1494
1495int inet_diag_register(const struct inet_diag_handler *h)
1496{
1497 const __u16 type = h->idiag_type;
1498 int err = -EINVAL;
1499
1500 if (type >= IPPROTO_MAX)
1501 goto out;
1502
1503 mutex_lock(&inet_diag_table_mutex);
1504 err = -EEXIST;
1505 if (!inet_diag_table[type]) {
1506 inet_diag_table[type] = h;
1507 err = 0;
1508 }
1509 mutex_unlock(&inet_diag_table_mutex);
1510out:
1511 return err;
1512}
1513EXPORT_SYMBOL_GPL(inet_diag_register);
1514
1515void inet_diag_unregister(const struct inet_diag_handler *h)
1516{
1517 const __u16 type = h->idiag_type;
1518
1519 if (type >= IPPROTO_MAX)
1520 return;
1521
1522 mutex_lock(&inet_diag_table_mutex);
1523 inet_diag_table[type] = NULL;
1524 mutex_unlock(&inet_diag_table_mutex);
1525}
1526EXPORT_SYMBOL_GPL(inet_diag_unregister);
1527
1528static int __init inet_diag_init(void)
1529{
1530 const int inet_diag_table_size = (IPPROTO_MAX *
1531 sizeof(struct inet_diag_handler *));
1532 int err = -ENOMEM;
1533
1534 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1535 if (!inet_diag_table)
1536 goto out;
1537
1538 err = sock_diag_register(&inet_diag_handler);
1539 if (err)
1540 goto out_free_nl;
1541
1542 err = sock_diag_register(&inet6_diag_handler);
1543 if (err)
1544 goto out_free_inet;
1545
1546 sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1547out:
1548 return err;
1549
1550out_free_inet:
1551 sock_diag_unregister(&inet_diag_handler);
1552out_free_nl:
1553 kfree(inet_diag_table);
1554 goto out;
1555}
1556
1557static void __exit inet_diag_exit(void)
1558{
1559 sock_diag_unregister(&inet6_diag_handler);
1560 sock_diag_unregister(&inet_diag_handler);
1561 sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1562 kfree(inet_diag_table);
1563}
1564
1565module_init(inet_diag_init);
1566module_exit(inet_diag_exit);
1567MODULE_LICENSE("GPL");
1568MODULE_DESCRIPTION("INET/INET6: socket monitoring via SOCK_DIAG");
1569MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1570MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);