inet: frags: remove inet_frag_maybe_warn_overflow()
[GitHub/LineageOS/android_kernel_motorola_exynos9610.git] / net / ipv4 / inet_diag.c
1 /*
2 * inet_diag.c Module for monitoring INET transport protocols sockets.
3 *
4 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 */
11
12 #include <linux/kernel.h>
13 #include <linux/module.h>
14 #include <linux/types.h>
15 #include <linux/fcntl.h>
16 #include <linux/random.h>
17 #include <linux/slab.h>
18 #include <linux/cache.h>
19 #include <linux/init.h>
20 #include <linux/time.h>
21
22 #include <net/icmp.h>
23 #include <net/tcp.h>
24 #include <net/ipv6.h>
25 #include <net/inet_common.h>
26 #include <net/inet_connection_sock.h>
27 #include <net/inet_hashtables.h>
28 #include <net/inet_timewait_sock.h>
29 #include <net/inet6_hashtables.h>
30 #include <net/netlink.h>
31
32 #include <linux/inet.h>
33 #include <linux/stddef.h>
34
35 #include <linux/inet_diag.h>
36 #include <linux/sock_diag.h>
37
38 static const struct inet_diag_handler **inet_diag_table;
39
40 struct inet_diag_entry {
41 const __be32 *saddr;
42 const __be32 *daddr;
43 u16 sport;
44 u16 dport;
45 u16 family;
46 u16 userlocks;
47 u32 ifindex;
48 u32 mark;
49 };
50
51 static DEFINE_MUTEX(inet_diag_table_mutex);
52
53 static const struct inet_diag_handler *inet_diag_lock_handler(int proto)
54 {
55 if (!inet_diag_table[proto])
56 request_module("net-pf-%d-proto-%d-type-%d-%d", PF_NETLINK,
57 NETLINK_SOCK_DIAG, AF_INET, proto);
58
59 mutex_lock(&inet_diag_table_mutex);
60 if (!inet_diag_table[proto])
61 return ERR_PTR(-ENOENT);
62
63 return inet_diag_table[proto];
64 }
65
66 static void inet_diag_unlock_handler(const struct inet_diag_handler *handler)
67 {
68 mutex_unlock(&inet_diag_table_mutex);
69 }
70
71 void inet_diag_msg_common_fill(struct inet_diag_msg *r, struct sock *sk)
72 {
73 r->idiag_family = sk->sk_family;
74
75 r->id.idiag_sport = htons(sk->sk_num);
76 r->id.idiag_dport = sk->sk_dport;
77 r->id.idiag_if = sk->sk_bound_dev_if;
78 sock_diag_save_cookie(sk, r->id.idiag_cookie);
79
80 #if IS_ENABLED(CONFIG_IPV6)
81 if (sk->sk_family == AF_INET6) {
82 *(struct in6_addr *)r->id.idiag_src = sk->sk_v6_rcv_saddr;
83 *(struct in6_addr *)r->id.idiag_dst = sk->sk_v6_daddr;
84 } else
85 #endif
86 {
87 memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src));
88 memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst));
89
90 r->id.idiag_src[0] = sk->sk_rcv_saddr;
91 r->id.idiag_dst[0] = sk->sk_daddr;
92 }
93 }
94 EXPORT_SYMBOL_GPL(inet_diag_msg_common_fill);
95
96 static size_t inet_sk_attr_size(struct sock *sk,
97 const struct inet_diag_req_v2 *req,
98 bool net_admin)
99 {
100 const struct inet_diag_handler *handler;
101 size_t aux = 0;
102
103 handler = inet_diag_table[req->sdiag_protocol];
104 if (handler && handler->idiag_get_aux_size)
105 aux = handler->idiag_get_aux_size(sk, net_admin);
106
107 return nla_total_size(sizeof(struct tcp_info))
108 + nla_total_size(1) /* INET_DIAG_SHUTDOWN */
109 + nla_total_size(1) /* INET_DIAG_TOS */
110 + nla_total_size(1) /* INET_DIAG_TCLASS */
111 + nla_total_size(4) /* INET_DIAG_MARK */
112 + nla_total_size(sizeof(struct inet_diag_meminfo))
113 + nla_total_size(sizeof(struct inet_diag_msg))
114 + nla_total_size(SK_MEMINFO_VARS * sizeof(u32))
115 + nla_total_size(TCP_CA_NAME_MAX)
116 + nla_total_size(sizeof(struct tcpvegas_info))
117 + aux
118 + 64;
119 }
120
121 int inet_diag_msg_attrs_fill(struct sock *sk, struct sk_buff *skb,
122 struct inet_diag_msg *r, int ext,
123 struct user_namespace *user_ns,
124 bool net_admin)
125 {
126 const struct inet_sock *inet = inet_sk(sk);
127
128 if (nla_put_u8(skb, INET_DIAG_SHUTDOWN, sk->sk_shutdown))
129 goto errout;
130
131 /* IPv6 dual-stack sockets use inet->tos for IPv4 connections,
132 * hence this needs to be included regardless of socket family.
133 */
134 if (ext & (1 << (INET_DIAG_TOS - 1)))
135 if (nla_put_u8(skb, INET_DIAG_TOS, inet->tos) < 0)
136 goto errout;
137
138 #if IS_ENABLED(CONFIG_IPV6)
139 if (r->idiag_family == AF_INET6) {
140 if (ext & (1 << (INET_DIAG_TCLASS - 1)))
141 if (nla_put_u8(skb, INET_DIAG_TCLASS,
142 inet6_sk(sk)->tclass) < 0)
143 goto errout;
144
145 if (((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
146 nla_put_u8(skb, INET_DIAG_SKV6ONLY, ipv6_only_sock(sk)))
147 goto errout;
148 }
149 #endif
150
151 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK, sk->sk_mark))
152 goto errout;
153
154 r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk));
155 r->idiag_inode = sock_i_ino(sk);
156
157 return 0;
158 errout:
159 return 1;
160 }
161 EXPORT_SYMBOL_GPL(inet_diag_msg_attrs_fill);
162
163 int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk,
164 struct sk_buff *skb, const struct inet_diag_req_v2 *req,
165 struct user_namespace *user_ns,
166 u32 portid, u32 seq, u16 nlmsg_flags,
167 const struct nlmsghdr *unlh,
168 bool net_admin)
169 {
170 const struct tcp_congestion_ops *ca_ops;
171 const struct inet_diag_handler *handler;
172 int ext = req->idiag_ext;
173 struct inet_diag_msg *r;
174 struct nlmsghdr *nlh;
175 struct nlattr *attr;
176 void *info = NULL;
177
178 handler = inet_diag_table[req->sdiag_protocol];
179 BUG_ON(!handler);
180
181 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
182 nlmsg_flags);
183 if (!nlh)
184 return -EMSGSIZE;
185
186 r = nlmsg_data(nlh);
187 BUG_ON(!sk_fullsock(sk));
188
189 inet_diag_msg_common_fill(r, sk);
190 r->idiag_state = sk->sk_state;
191 r->idiag_timer = 0;
192 r->idiag_retrans = 0;
193
194 if (inet_diag_msg_attrs_fill(sk, skb, r, ext, user_ns, net_admin))
195 goto errout;
196
197 if (ext & (1 << (INET_DIAG_MEMINFO - 1))) {
198 struct inet_diag_meminfo minfo = {
199 .idiag_rmem = sk_rmem_alloc_get(sk),
200 .idiag_wmem = sk->sk_wmem_queued,
201 .idiag_fmem = sk->sk_forward_alloc,
202 .idiag_tmem = sk_wmem_alloc_get(sk),
203 };
204
205 if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0)
206 goto errout;
207 }
208
209 if (ext & (1 << (INET_DIAG_SKMEMINFO - 1)))
210 if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO))
211 goto errout;
212
213 /*
214 * RAW sockets might have user-defined protocols assigned,
215 * so report the one supplied on socket creation.
216 */
217 if (sk->sk_type == SOCK_RAW) {
218 if (nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))
219 goto errout;
220 }
221
222 if (!icsk) {
223 handler->idiag_get_info(sk, r, NULL);
224 goto out;
225 }
226
227 if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
228 icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
229 icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
230 r->idiag_timer = 1;
231 r->idiag_retrans = icsk->icsk_retransmits;
232 r->idiag_expires =
233 jiffies_to_msecs(icsk->icsk_timeout - jiffies);
234 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
235 r->idiag_timer = 4;
236 r->idiag_retrans = icsk->icsk_probes_out;
237 r->idiag_expires =
238 jiffies_to_msecs(icsk->icsk_timeout - jiffies);
239 } else if (timer_pending(&sk->sk_timer)) {
240 r->idiag_timer = 2;
241 r->idiag_retrans = icsk->icsk_probes_out;
242 r->idiag_expires =
243 jiffies_to_msecs(sk->sk_timer.expires - jiffies);
244 } else {
245 r->idiag_timer = 0;
246 r->idiag_expires = 0;
247 }
248
249 if ((ext & (1 << (INET_DIAG_INFO - 1))) && handler->idiag_info_size) {
250 attr = nla_reserve_64bit(skb, INET_DIAG_INFO,
251 handler->idiag_info_size,
252 INET_DIAG_PAD);
253 if (!attr)
254 goto errout;
255
256 info = nla_data(attr);
257 }
258
259 if (ext & (1 << (INET_DIAG_CONG - 1))) {
260 int err = 0;
261
262 rcu_read_lock();
263 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
264 if (ca_ops)
265 err = nla_put_string(skb, INET_DIAG_CONG, ca_ops->name);
266 rcu_read_unlock();
267 if (err < 0)
268 goto errout;
269 }
270
271 handler->idiag_get_info(sk, r, info);
272
273 if (ext & (1 << (INET_DIAG_INFO - 1)) && handler->idiag_get_aux)
274 if (handler->idiag_get_aux(sk, net_admin, skb) < 0)
275 goto errout;
276
277 if (sk->sk_state < TCP_TIME_WAIT) {
278 union tcp_cc_info info;
279 size_t sz = 0;
280 int attr;
281
282 rcu_read_lock();
283 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
284 if (ca_ops && ca_ops->get_info)
285 sz = ca_ops->get_info(sk, ext, &attr, &info);
286 rcu_read_unlock();
287 if (sz && nla_put(skb, attr, sz, &info) < 0)
288 goto errout;
289 }
290
291 if (ext & (1 << (INET_DIAG_CLASS_ID - 1))) {
292 u32 classid = 0;
293
294 #ifdef CONFIG_SOCK_CGROUP_DATA
295 classid = sock_cgroup_classid(&sk->sk_cgrp_data);
296 #endif
297
298 if (nla_put_u32(skb, INET_DIAG_CLASS_ID, classid))
299 goto errout;
300 }
301
302 out:
303 nlmsg_end(skb, nlh);
304 return 0;
305
306 errout:
307 nlmsg_cancel(skb, nlh);
308 return -EMSGSIZE;
309 }
310 EXPORT_SYMBOL_GPL(inet_sk_diag_fill);
311
312 static int inet_csk_diag_fill(struct sock *sk,
313 struct sk_buff *skb,
314 const struct inet_diag_req_v2 *req,
315 struct user_namespace *user_ns,
316 u32 portid, u32 seq, u16 nlmsg_flags,
317 const struct nlmsghdr *unlh,
318 bool net_admin)
319 {
320 return inet_sk_diag_fill(sk, inet_csk(sk), skb, req, user_ns,
321 portid, seq, nlmsg_flags, unlh, net_admin);
322 }
323
324 static int inet_twsk_diag_fill(struct sock *sk,
325 struct sk_buff *skb,
326 u32 portid, u32 seq, u16 nlmsg_flags,
327 const struct nlmsghdr *unlh)
328 {
329 struct inet_timewait_sock *tw = inet_twsk(sk);
330 struct inet_diag_msg *r;
331 struct nlmsghdr *nlh;
332 long tmo;
333
334 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
335 nlmsg_flags);
336 if (!nlh)
337 return -EMSGSIZE;
338
339 r = nlmsg_data(nlh);
340 BUG_ON(tw->tw_state != TCP_TIME_WAIT);
341
342 tmo = tw->tw_timer.expires - jiffies;
343 if (tmo < 0)
344 tmo = 0;
345
346 inet_diag_msg_common_fill(r, sk);
347 r->idiag_retrans = 0;
348
349 r->idiag_state = tw->tw_substate;
350 r->idiag_timer = 3;
351 r->idiag_expires = jiffies_to_msecs(tmo);
352 r->idiag_rqueue = 0;
353 r->idiag_wqueue = 0;
354 r->idiag_uid = 0;
355 r->idiag_inode = 0;
356
357 nlmsg_end(skb, nlh);
358 return 0;
359 }
360
361 static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb,
362 u32 portid, u32 seq, u16 nlmsg_flags,
363 const struct nlmsghdr *unlh, bool net_admin)
364 {
365 struct request_sock *reqsk = inet_reqsk(sk);
366 struct inet_diag_msg *r;
367 struct nlmsghdr *nlh;
368 long tmo;
369
370 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
371 nlmsg_flags);
372 if (!nlh)
373 return -EMSGSIZE;
374
375 r = nlmsg_data(nlh);
376 inet_diag_msg_common_fill(r, sk);
377 r->idiag_state = TCP_SYN_RECV;
378 r->idiag_timer = 1;
379 r->idiag_retrans = reqsk->num_retrans;
380
381 BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) !=
382 offsetof(struct sock, sk_cookie));
383
384 tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies;
385 r->idiag_expires = (tmo >= 0) ? jiffies_to_msecs(tmo) : 0;
386 r->idiag_rqueue = 0;
387 r->idiag_wqueue = 0;
388 r->idiag_uid = 0;
389 r->idiag_inode = 0;
390
391 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
392 inet_rsk(reqsk)->ir_mark))
393 return -EMSGSIZE;
394
395 nlmsg_end(skb, nlh);
396 return 0;
397 }
398
399 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
400 const struct inet_diag_req_v2 *r,
401 struct user_namespace *user_ns,
402 u32 portid, u32 seq, u16 nlmsg_flags,
403 const struct nlmsghdr *unlh, bool net_admin)
404 {
405 if (sk->sk_state == TCP_TIME_WAIT)
406 return inet_twsk_diag_fill(sk, skb, portid, seq,
407 nlmsg_flags, unlh);
408
409 if (sk->sk_state == TCP_NEW_SYN_RECV)
410 return inet_req_diag_fill(sk, skb, portid, seq,
411 nlmsg_flags, unlh, net_admin);
412
413 return inet_csk_diag_fill(sk, skb, r, user_ns, portid, seq,
414 nlmsg_flags, unlh, net_admin);
415 }
416
417 struct sock *inet_diag_find_one_icsk(struct net *net,
418 struct inet_hashinfo *hashinfo,
419 const struct inet_diag_req_v2 *req)
420 {
421 struct sock *sk;
422
423 rcu_read_lock();
424 if (req->sdiag_family == AF_INET)
425 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0],
426 req->id.idiag_dport, req->id.idiag_src[0],
427 req->id.idiag_sport, req->id.idiag_if);
428 #if IS_ENABLED(CONFIG_IPV6)
429 else if (req->sdiag_family == AF_INET6) {
430 if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) &&
431 ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src))
432 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3],
433 req->id.idiag_dport, req->id.idiag_src[3],
434 req->id.idiag_sport, req->id.idiag_if);
435 else
436 sk = inet6_lookup(net, hashinfo, NULL, 0,
437 (struct in6_addr *)req->id.idiag_dst,
438 req->id.idiag_dport,
439 (struct in6_addr *)req->id.idiag_src,
440 req->id.idiag_sport,
441 req->id.idiag_if);
442 }
443 #endif
444 else {
445 rcu_read_unlock();
446 return ERR_PTR(-EINVAL);
447 }
448 rcu_read_unlock();
449 if (!sk)
450 return ERR_PTR(-ENOENT);
451
452 if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) {
453 sock_gen_put(sk);
454 return ERR_PTR(-ENOENT);
455 }
456
457 return sk;
458 }
459 EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk);
460
461 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo,
462 struct sk_buff *in_skb,
463 const struct nlmsghdr *nlh,
464 const struct inet_diag_req_v2 *req)
465 {
466 bool net_admin = netlink_net_capable(in_skb, CAP_NET_ADMIN);
467 struct net *net = sock_net(in_skb->sk);
468 struct sk_buff *rep;
469 struct sock *sk;
470 int err;
471
472 sk = inet_diag_find_one_icsk(net, hashinfo, req);
473 if (IS_ERR(sk))
474 return PTR_ERR(sk);
475
476 rep = nlmsg_new(inet_sk_attr_size(sk, req, net_admin), GFP_KERNEL);
477 if (!rep) {
478 err = -ENOMEM;
479 goto out;
480 }
481
482 err = sk_diag_fill(sk, rep, req,
483 sk_user_ns(NETLINK_CB(in_skb).sk),
484 NETLINK_CB(in_skb).portid,
485 nlh->nlmsg_seq, 0, nlh, net_admin);
486 if (err < 0) {
487 WARN_ON(err == -EMSGSIZE);
488 nlmsg_free(rep);
489 goto out;
490 }
491 err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid,
492 MSG_DONTWAIT);
493 if (err > 0)
494 err = 0;
495
496 out:
497 if (sk)
498 sock_gen_put(sk);
499
500 return err;
501 }
502 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
503
504 static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb,
505 const struct nlmsghdr *nlh,
506 const struct inet_diag_req_v2 *req)
507 {
508 const struct inet_diag_handler *handler;
509 int err;
510
511 handler = inet_diag_lock_handler(req->sdiag_protocol);
512 if (IS_ERR(handler))
513 err = PTR_ERR(handler);
514 else if (cmd == SOCK_DIAG_BY_FAMILY)
515 err = handler->dump_one(in_skb, nlh, req);
516 else if (cmd == SOCK_DESTROY && handler->destroy)
517 err = handler->destroy(in_skb, req);
518 else
519 err = -EOPNOTSUPP;
520 inet_diag_unlock_handler(handler);
521
522 return err;
523 }
524
525 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
526 {
527 int words = bits >> 5;
528
529 bits &= 0x1f;
530
531 if (words) {
532 if (memcmp(a1, a2, words << 2))
533 return 0;
534 }
535 if (bits) {
536 __be32 w1, w2;
537 __be32 mask;
538
539 w1 = a1[words];
540 w2 = a2[words];
541
542 mask = htonl((0xffffffff) << (32 - bits));
543
544 if ((w1 ^ w2) & mask)
545 return 0;
546 }
547
548 return 1;
549 }
550
551 static int inet_diag_bc_run(const struct nlattr *_bc,
552 const struct inet_diag_entry *entry)
553 {
554 const void *bc = nla_data(_bc);
555 int len = nla_len(_bc);
556
557 while (len > 0) {
558 int yes = 1;
559 const struct inet_diag_bc_op *op = bc;
560
561 switch (op->code) {
562 case INET_DIAG_BC_NOP:
563 break;
564 case INET_DIAG_BC_JMP:
565 yes = 0;
566 break;
567 case INET_DIAG_BC_S_GE:
568 yes = entry->sport >= op[1].no;
569 break;
570 case INET_DIAG_BC_S_LE:
571 yes = entry->sport <= op[1].no;
572 break;
573 case INET_DIAG_BC_D_GE:
574 yes = entry->dport >= op[1].no;
575 break;
576 case INET_DIAG_BC_D_LE:
577 yes = entry->dport <= op[1].no;
578 break;
579 case INET_DIAG_BC_AUTO:
580 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
581 break;
582 case INET_DIAG_BC_S_COND:
583 case INET_DIAG_BC_D_COND: {
584 const struct inet_diag_hostcond *cond;
585 const __be32 *addr;
586
587 cond = (const struct inet_diag_hostcond *)(op + 1);
588 if (cond->port != -1 &&
589 cond->port != (op->code == INET_DIAG_BC_S_COND ?
590 entry->sport : entry->dport)) {
591 yes = 0;
592 break;
593 }
594
595 if (op->code == INET_DIAG_BC_S_COND)
596 addr = entry->saddr;
597 else
598 addr = entry->daddr;
599
600 if (cond->family != AF_UNSPEC &&
601 cond->family != entry->family) {
602 if (entry->family == AF_INET6 &&
603 cond->family == AF_INET) {
604 if (addr[0] == 0 && addr[1] == 0 &&
605 addr[2] == htonl(0xffff) &&
606 bitstring_match(addr + 3,
607 cond->addr,
608 cond->prefix_len))
609 break;
610 }
611 yes = 0;
612 break;
613 }
614
615 if (cond->prefix_len == 0)
616 break;
617 if (bitstring_match(addr, cond->addr,
618 cond->prefix_len))
619 break;
620 yes = 0;
621 break;
622 }
623 case INET_DIAG_BC_DEV_COND: {
624 u32 ifindex;
625
626 ifindex = *((const u32 *)(op + 1));
627 if (ifindex != entry->ifindex)
628 yes = 0;
629 break;
630 }
631 case INET_DIAG_BC_MARK_COND: {
632 struct inet_diag_markcond *cond;
633
634 cond = (struct inet_diag_markcond *)(op + 1);
635 if ((entry->mark & cond->mask) != cond->mark)
636 yes = 0;
637 break;
638 }
639 }
640
641 if (yes) {
642 len -= op->yes;
643 bc += op->yes;
644 } else {
645 len -= op->no;
646 bc += op->no;
647 }
648 }
649 return len == 0;
650 }
651
652 /* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV)
653 */
654 static void entry_fill_addrs(struct inet_diag_entry *entry,
655 const struct sock *sk)
656 {
657 #if IS_ENABLED(CONFIG_IPV6)
658 if (sk->sk_family == AF_INET6) {
659 entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32;
660 entry->daddr = sk->sk_v6_daddr.s6_addr32;
661 } else
662 #endif
663 {
664 entry->saddr = &sk->sk_rcv_saddr;
665 entry->daddr = &sk->sk_daddr;
666 }
667 }
668
669 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
670 {
671 struct inet_sock *inet = inet_sk(sk);
672 struct inet_diag_entry entry;
673
674 if (!bc)
675 return 1;
676
677 entry.family = sk->sk_family;
678 entry_fill_addrs(&entry, sk);
679 entry.sport = inet->inet_num;
680 entry.dport = ntohs(inet->inet_dport);
681 entry.ifindex = sk->sk_bound_dev_if;
682 entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0;
683 if (sk_fullsock(sk))
684 entry.mark = sk->sk_mark;
685 else if (sk->sk_state == TCP_NEW_SYN_RECV)
686 entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark;
687 else
688 entry.mark = 0;
689
690 return inet_diag_bc_run(bc, &entry);
691 }
692 EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
693
694 static int valid_cc(const void *bc, int len, int cc)
695 {
696 while (len >= 0) {
697 const struct inet_diag_bc_op *op = bc;
698
699 if (cc > len)
700 return 0;
701 if (cc == len)
702 return 1;
703 if (op->yes < 4 || op->yes & 3)
704 return 0;
705 len -= op->yes;
706 bc += op->yes;
707 }
708 return 0;
709 }
710
711 /* data is u32 ifindex */
712 static bool valid_devcond(const struct inet_diag_bc_op *op, int len,
713 int *min_len)
714 {
715 /* Check ifindex space. */
716 *min_len += sizeof(u32);
717 if (len < *min_len)
718 return false;
719
720 return true;
721 }
722 /* Validate an inet_diag_hostcond. */
723 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
724 int *min_len)
725 {
726 struct inet_diag_hostcond *cond;
727 int addr_len;
728
729 /* Check hostcond space. */
730 *min_len += sizeof(struct inet_diag_hostcond);
731 if (len < *min_len)
732 return false;
733 cond = (struct inet_diag_hostcond *)(op + 1);
734
735 /* Check address family and address length. */
736 switch (cond->family) {
737 case AF_UNSPEC:
738 addr_len = 0;
739 break;
740 case AF_INET:
741 addr_len = sizeof(struct in_addr);
742 break;
743 case AF_INET6:
744 addr_len = sizeof(struct in6_addr);
745 break;
746 default:
747 return false;
748 }
749 *min_len += addr_len;
750 if (len < *min_len)
751 return false;
752
753 /* Check prefix length (in bits) vs address length (in bytes). */
754 if (cond->prefix_len > 8 * addr_len)
755 return false;
756
757 return true;
758 }
759
760 /* Validate a port comparison operator. */
761 static bool valid_port_comparison(const struct inet_diag_bc_op *op,
762 int len, int *min_len)
763 {
764 /* Port comparisons put the port in a follow-on inet_diag_bc_op. */
765 *min_len += sizeof(struct inet_diag_bc_op);
766 if (len < *min_len)
767 return false;
768 return true;
769 }
770
771 static bool valid_markcond(const struct inet_diag_bc_op *op, int len,
772 int *min_len)
773 {
774 *min_len += sizeof(struct inet_diag_markcond);
775 return len >= *min_len;
776 }
777
778 static int inet_diag_bc_audit(const struct nlattr *attr,
779 const struct sk_buff *skb)
780 {
781 bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN);
782 const void *bytecode, *bc;
783 int bytecode_len, len;
784
785 if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op))
786 return -EINVAL;
787
788 bytecode = bc = nla_data(attr);
789 len = bytecode_len = nla_len(attr);
790
791 while (len > 0) {
792 int min_len = sizeof(struct inet_diag_bc_op);
793 const struct inet_diag_bc_op *op = bc;
794
795 switch (op->code) {
796 case INET_DIAG_BC_S_COND:
797 case INET_DIAG_BC_D_COND:
798 if (!valid_hostcond(bc, len, &min_len))
799 return -EINVAL;
800 break;
801 case INET_DIAG_BC_DEV_COND:
802 if (!valid_devcond(bc, len, &min_len))
803 return -EINVAL;
804 break;
805 case INET_DIAG_BC_S_GE:
806 case INET_DIAG_BC_S_LE:
807 case INET_DIAG_BC_D_GE:
808 case INET_DIAG_BC_D_LE:
809 if (!valid_port_comparison(bc, len, &min_len))
810 return -EINVAL;
811 break;
812 case INET_DIAG_BC_MARK_COND:
813 if (!net_admin)
814 return -EPERM;
815 if (!valid_markcond(bc, len, &min_len))
816 return -EINVAL;
817 break;
818 case INET_DIAG_BC_AUTO:
819 case INET_DIAG_BC_JMP:
820 case INET_DIAG_BC_NOP:
821 break;
822 default:
823 return -EINVAL;
824 }
825
826 if (op->code != INET_DIAG_BC_NOP) {
827 if (op->no < min_len || op->no > len + 4 || op->no & 3)
828 return -EINVAL;
829 if (op->no < len &&
830 !valid_cc(bytecode, bytecode_len, len - op->no))
831 return -EINVAL;
832 }
833
834 if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
835 return -EINVAL;
836 bc += op->yes;
837 len -= op->yes;
838 }
839 return len == 0 ? 0 : -EINVAL;
840 }
841
842 static int inet_csk_diag_dump(struct sock *sk,
843 struct sk_buff *skb,
844 struct netlink_callback *cb,
845 const struct inet_diag_req_v2 *r,
846 const struct nlattr *bc,
847 bool net_admin)
848 {
849 if (!inet_diag_bc_sk(bc, sk))
850 return 0;
851
852 return inet_csk_diag_fill(sk, skb, r,
853 sk_user_ns(NETLINK_CB(cb->skb).sk),
854 NETLINK_CB(cb->skb).portid,
855 cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh,
856 net_admin);
857 }
858
859 static void twsk_build_assert(void)
860 {
861 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) !=
862 offsetof(struct sock, sk_family));
863
864 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) !=
865 offsetof(struct inet_sock, inet_num));
866
867 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) !=
868 offsetof(struct inet_sock, inet_dport));
869
870 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) !=
871 offsetof(struct inet_sock, inet_rcv_saddr));
872
873 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) !=
874 offsetof(struct inet_sock, inet_daddr));
875
876 #if IS_ENABLED(CONFIG_IPV6)
877 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) !=
878 offsetof(struct sock, sk_v6_rcv_saddr));
879
880 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) !=
881 offsetof(struct sock, sk_v6_daddr));
882 #endif
883 }
884
885 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
886 struct netlink_callback *cb,
887 const struct inet_diag_req_v2 *r, struct nlattr *bc)
888 {
889 bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN);
890 struct net *net = sock_net(skb->sk);
891 u32 idiag_states = r->idiag_states;
892 int i, num, s_i, s_num;
893 struct sock *sk;
894
895 if (idiag_states & TCPF_SYN_RECV)
896 idiag_states |= TCPF_NEW_SYN_RECV;
897 s_i = cb->args[1];
898 s_num = num = cb->args[2];
899
900 if (cb->args[0] == 0) {
901 if (!(idiag_states & TCPF_LISTEN) || r->id.idiag_dport)
902 goto skip_listen_ht;
903
904 for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
905 struct inet_listen_hashbucket *ilb;
906
907 num = 0;
908 ilb = &hashinfo->listening_hash[i];
909 spin_lock(&ilb->lock);
910 sk_for_each(sk, &ilb->head) {
911 struct inet_sock *inet = inet_sk(sk);
912
913 if (!net_eq(sock_net(sk), net))
914 continue;
915
916 if (num < s_num) {
917 num++;
918 continue;
919 }
920
921 if (r->sdiag_family != AF_UNSPEC &&
922 sk->sk_family != r->sdiag_family)
923 goto next_listen;
924
925 if (r->id.idiag_sport != inet->inet_sport &&
926 r->id.idiag_sport)
927 goto next_listen;
928
929 if (inet_csk_diag_dump(sk, skb, cb, r,
930 bc, net_admin) < 0) {
931 spin_unlock(&ilb->lock);
932 goto done;
933 }
934
935 next_listen:
936 ++num;
937 }
938 spin_unlock(&ilb->lock);
939
940 s_num = 0;
941 }
942 skip_listen_ht:
943 cb->args[0] = 1;
944 s_i = num = s_num = 0;
945 }
946
947 if (!(idiag_states & ~TCPF_LISTEN))
948 goto out;
949
950 #define SKARR_SZ 16
951 for (i = s_i; i <= hashinfo->ehash_mask; i++) {
952 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
953 spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
954 struct hlist_nulls_node *node;
955 struct sock *sk_arr[SKARR_SZ];
956 int num_arr[SKARR_SZ];
957 int idx, accum, res;
958
959 if (hlist_nulls_empty(&head->chain))
960 continue;
961
962 if (i > s_i)
963 s_num = 0;
964
965 next_chunk:
966 num = 0;
967 accum = 0;
968 spin_lock_bh(lock);
969 sk_nulls_for_each(sk, node, &head->chain) {
970 int state;
971
972 if (!net_eq(sock_net(sk), net))
973 continue;
974 if (num < s_num)
975 goto next_normal;
976 state = (sk->sk_state == TCP_TIME_WAIT) ?
977 inet_twsk(sk)->tw_substate : sk->sk_state;
978 if (!(idiag_states & (1 << state)))
979 goto next_normal;
980 if (r->sdiag_family != AF_UNSPEC &&
981 sk->sk_family != r->sdiag_family)
982 goto next_normal;
983 if (r->id.idiag_sport != htons(sk->sk_num) &&
984 r->id.idiag_sport)
985 goto next_normal;
986 if (r->id.idiag_dport != sk->sk_dport &&
987 r->id.idiag_dport)
988 goto next_normal;
989 twsk_build_assert();
990
991 if (!inet_diag_bc_sk(bc, sk))
992 goto next_normal;
993
994 sock_hold(sk);
995 num_arr[accum] = num;
996 sk_arr[accum] = sk;
997 if (++accum == SKARR_SZ)
998 break;
999 next_normal:
1000 ++num;
1001 }
1002 spin_unlock_bh(lock);
1003 res = 0;
1004 for (idx = 0; idx < accum; idx++) {
1005 if (res >= 0) {
1006 res = sk_diag_fill(sk_arr[idx], skb, r,
1007 sk_user_ns(NETLINK_CB(cb->skb).sk),
1008 NETLINK_CB(cb->skb).portid,
1009 cb->nlh->nlmsg_seq, NLM_F_MULTI,
1010 cb->nlh, net_admin);
1011 if (res < 0)
1012 num = num_arr[idx];
1013 }
1014 sock_gen_put(sk_arr[idx]);
1015 }
1016 if (res < 0)
1017 break;
1018 cond_resched();
1019 if (accum == SKARR_SZ) {
1020 s_num = num + 1;
1021 goto next_chunk;
1022 }
1023 }
1024
1025 done:
1026 cb->args[1] = i;
1027 cb->args[2] = num;
1028 out:
1029 ;
1030 }
1031 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
1032
1033 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
1034 const struct inet_diag_req_v2 *r,
1035 struct nlattr *bc)
1036 {
1037 const struct inet_diag_handler *handler;
1038 int err = 0;
1039
1040 handler = inet_diag_lock_handler(r->sdiag_protocol);
1041 if (!IS_ERR(handler))
1042 handler->dump(skb, cb, r, bc);
1043 else
1044 err = PTR_ERR(handler);
1045 inet_diag_unlock_handler(handler);
1046
1047 return err ? : skb->len;
1048 }
1049
1050 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
1051 {
1052 int hdrlen = sizeof(struct inet_diag_req_v2);
1053 struct nlattr *bc = NULL;
1054
1055 if (nlmsg_attrlen(cb->nlh, hdrlen))
1056 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE);
1057
1058 return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh), bc);
1059 }
1060
1061 static int inet_diag_type2proto(int type)
1062 {
1063 switch (type) {
1064 case TCPDIAG_GETSOCK:
1065 return IPPROTO_TCP;
1066 case DCCPDIAG_GETSOCK:
1067 return IPPROTO_DCCP;
1068 default:
1069 return 0;
1070 }
1071 }
1072
1073 static int inet_diag_dump_compat(struct sk_buff *skb,
1074 struct netlink_callback *cb)
1075 {
1076 struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1077 int hdrlen = sizeof(struct inet_diag_req);
1078 struct inet_diag_req_v2 req;
1079 struct nlattr *bc = NULL;
1080
1081 req.sdiag_family = AF_UNSPEC; /* compatibility */
1082 req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1083 req.idiag_ext = rc->idiag_ext;
1084 req.idiag_states = rc->idiag_states;
1085 req.id = rc->id;
1086
1087 if (nlmsg_attrlen(cb->nlh, hdrlen))
1088 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE);
1089
1090 return __inet_diag_dump(skb, cb, &req, bc);
1091 }
1092
1093 static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1094 const struct nlmsghdr *nlh)
1095 {
1096 struct inet_diag_req *rc = nlmsg_data(nlh);
1097 struct inet_diag_req_v2 req;
1098
1099 req.sdiag_family = rc->idiag_family;
1100 req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1101 req.idiag_ext = rc->idiag_ext;
1102 req.idiag_states = rc->idiag_states;
1103 req.id = rc->id;
1104
1105 return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh, &req);
1106 }
1107
1108 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1109 {
1110 int hdrlen = sizeof(struct inet_diag_req);
1111 struct net *net = sock_net(skb->sk);
1112
1113 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1114 nlmsg_len(nlh) < hdrlen)
1115 return -EINVAL;
1116
1117 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1118 if (nlmsg_attrlen(nlh, hdrlen)) {
1119 struct nlattr *attr;
1120 int err;
1121
1122 attr = nlmsg_find_attr(nlh, hdrlen,
1123 INET_DIAG_REQ_BYTECODE);
1124 err = inet_diag_bc_audit(attr, skb);
1125 if (err)
1126 return err;
1127 }
1128 {
1129 struct netlink_dump_control c = {
1130 .dump = inet_diag_dump_compat,
1131 };
1132 return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1133 }
1134 }
1135
1136 return inet_diag_get_exact_compat(skb, nlh);
1137 }
1138
1139 static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h)
1140 {
1141 int hdrlen = sizeof(struct inet_diag_req_v2);
1142 struct net *net = sock_net(skb->sk);
1143
1144 if (nlmsg_len(h) < hdrlen)
1145 return -EINVAL;
1146
1147 if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY &&
1148 h->nlmsg_flags & NLM_F_DUMP) {
1149 if (nlmsg_attrlen(h, hdrlen)) {
1150 struct nlattr *attr;
1151 int err;
1152
1153 attr = nlmsg_find_attr(h, hdrlen,
1154 INET_DIAG_REQ_BYTECODE);
1155 err = inet_diag_bc_audit(attr, skb);
1156 if (err)
1157 return err;
1158 }
1159 {
1160 struct netlink_dump_control c = {
1161 .dump = inet_diag_dump,
1162 };
1163 return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1164 }
1165 }
1166
1167 return inet_diag_cmd_exact(h->nlmsg_type, skb, h, nlmsg_data(h));
1168 }
1169
1170 static
1171 int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk)
1172 {
1173 const struct inet_diag_handler *handler;
1174 struct nlmsghdr *nlh;
1175 struct nlattr *attr;
1176 struct inet_diag_msg *r;
1177 void *info = NULL;
1178 int err = 0;
1179
1180 nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0);
1181 if (!nlh)
1182 return -ENOMEM;
1183
1184 r = nlmsg_data(nlh);
1185 memset(r, 0, sizeof(*r));
1186 inet_diag_msg_common_fill(r, sk);
1187 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM)
1188 r->id.idiag_sport = inet_sk(sk)->inet_sport;
1189 r->idiag_state = sk->sk_state;
1190
1191 if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) {
1192 nlmsg_cancel(skb, nlh);
1193 return err;
1194 }
1195
1196 handler = inet_diag_lock_handler(sk->sk_protocol);
1197 if (IS_ERR(handler)) {
1198 inet_diag_unlock_handler(handler);
1199 nlmsg_cancel(skb, nlh);
1200 return PTR_ERR(handler);
1201 }
1202
1203 attr = handler->idiag_info_size
1204 ? nla_reserve_64bit(skb, INET_DIAG_INFO,
1205 handler->idiag_info_size,
1206 INET_DIAG_PAD)
1207 : NULL;
1208 if (attr)
1209 info = nla_data(attr);
1210
1211 handler->idiag_get_info(sk, r, info);
1212 inet_diag_unlock_handler(handler);
1213
1214 nlmsg_end(skb, nlh);
1215 return 0;
1216 }
1217
1218 static const struct sock_diag_handler inet_diag_handler = {
1219 .family = AF_INET,
1220 .dump = inet_diag_handler_cmd,
1221 .get_info = inet_diag_handler_get_info,
1222 .destroy = inet_diag_handler_cmd,
1223 };
1224
1225 static const struct sock_diag_handler inet6_diag_handler = {
1226 .family = AF_INET6,
1227 .dump = inet_diag_handler_cmd,
1228 .get_info = inet_diag_handler_get_info,
1229 .destroy = inet_diag_handler_cmd,
1230 };
1231
1232 int inet_diag_register(const struct inet_diag_handler *h)
1233 {
1234 const __u16 type = h->idiag_type;
1235 int err = -EINVAL;
1236
1237 if (type >= IPPROTO_MAX)
1238 goto out;
1239
1240 mutex_lock(&inet_diag_table_mutex);
1241 err = -EEXIST;
1242 if (!inet_diag_table[type]) {
1243 inet_diag_table[type] = h;
1244 err = 0;
1245 }
1246 mutex_unlock(&inet_diag_table_mutex);
1247 out:
1248 return err;
1249 }
1250 EXPORT_SYMBOL_GPL(inet_diag_register);
1251
1252 void inet_diag_unregister(const struct inet_diag_handler *h)
1253 {
1254 const __u16 type = h->idiag_type;
1255
1256 if (type >= IPPROTO_MAX)
1257 return;
1258
1259 mutex_lock(&inet_diag_table_mutex);
1260 inet_diag_table[type] = NULL;
1261 mutex_unlock(&inet_diag_table_mutex);
1262 }
1263 EXPORT_SYMBOL_GPL(inet_diag_unregister);
1264
1265 static int __init inet_diag_init(void)
1266 {
1267 const int inet_diag_table_size = (IPPROTO_MAX *
1268 sizeof(struct inet_diag_handler *));
1269 int err = -ENOMEM;
1270
1271 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1272 if (!inet_diag_table)
1273 goto out;
1274
1275 err = sock_diag_register(&inet_diag_handler);
1276 if (err)
1277 goto out_free_nl;
1278
1279 err = sock_diag_register(&inet6_diag_handler);
1280 if (err)
1281 goto out_free_inet;
1282
1283 sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1284 out:
1285 return err;
1286
1287 out_free_inet:
1288 sock_diag_unregister(&inet_diag_handler);
1289 out_free_nl:
1290 kfree(inet_diag_table);
1291 goto out;
1292 }
1293
1294 static void __exit inet_diag_exit(void)
1295 {
1296 sock_diag_unregister(&inet6_diag_handler);
1297 sock_diag_unregister(&inet_diag_handler);
1298 sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1299 kfree(inet_diag_table);
1300 }
1301
1302 module_init(inet_diag_init);
1303 module_exit(inet_diag_exit);
1304 MODULE_LICENSE("GPL");
1305 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1306 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);