xref: /openbmc/linux/net/ipv4/inet_diag.c (revision 5e29a910)
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 	__be32 *saddr;
42 	__be32 *daddr;
43 	u16 sport;
44 	u16 dport;
45 	u16 family;
46 	u16 userlocks;
47 #if IS_ENABLED(CONFIG_IPV6)
48 	struct in6_addr saddr_storage;	/* for IPv4-mapped-IPv6 addresses */
49 	struct in6_addr daddr_storage;	/* for IPv4-mapped-IPv6 addresses */
50 #endif
51 };
52 
53 static DEFINE_MUTEX(inet_diag_table_mutex);
54 
55 static const struct inet_diag_handler *inet_diag_lock_handler(int proto)
56 {
57 	if (!inet_diag_table[proto])
58 		request_module("net-pf-%d-proto-%d-type-%d-%d", PF_NETLINK,
59 			       NETLINK_SOCK_DIAG, AF_INET, proto);
60 
61 	mutex_lock(&inet_diag_table_mutex);
62 	if (!inet_diag_table[proto])
63 		return ERR_PTR(-ENOENT);
64 
65 	return inet_diag_table[proto];
66 }
67 
68 static inline void inet_diag_unlock_handler(
69 	const struct inet_diag_handler *handler)
70 {
71 	mutex_unlock(&inet_diag_table_mutex);
72 }
73 
74 int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk,
75 			      struct sk_buff *skb, struct inet_diag_req_v2 *req,
76 			      struct user_namespace *user_ns,
77 			      u32 portid, u32 seq, u16 nlmsg_flags,
78 			      const struct nlmsghdr *unlh)
79 {
80 	const struct inet_sock *inet = inet_sk(sk);
81 	struct inet_diag_msg *r;
82 	struct nlmsghdr  *nlh;
83 	struct nlattr *attr;
84 	void *info = NULL;
85 	const struct inet_diag_handler *handler;
86 	int ext = req->idiag_ext;
87 
88 	handler = inet_diag_table[req->sdiag_protocol];
89 	BUG_ON(handler == NULL);
90 
91 	nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
92 			nlmsg_flags);
93 	if (!nlh)
94 		return -EMSGSIZE;
95 
96 	r = nlmsg_data(nlh);
97 	BUG_ON(sk->sk_state == TCP_TIME_WAIT);
98 
99 	r->idiag_family = sk->sk_family;
100 	r->idiag_state = sk->sk_state;
101 	r->idiag_timer = 0;
102 	r->idiag_retrans = 0;
103 
104 	r->id.idiag_if = sk->sk_bound_dev_if;
105 	sock_diag_save_cookie(sk, r->id.idiag_cookie);
106 
107 	r->id.idiag_sport = inet->inet_sport;
108 	r->id.idiag_dport = inet->inet_dport;
109 
110 	memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src));
111 	memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst));
112 
113 	r->id.idiag_src[0] = inet->inet_rcv_saddr;
114 	r->id.idiag_dst[0] = inet->inet_daddr;
115 
116 	if (nla_put_u8(skb, INET_DIAG_SHUTDOWN, sk->sk_shutdown))
117 		goto errout;
118 
119 	/* IPv6 dual-stack sockets use inet->tos for IPv4 connections,
120 	 * hence this needs to be included regardless of socket family.
121 	 */
122 	if (ext & (1 << (INET_DIAG_TOS - 1)))
123 		if (nla_put_u8(skb, INET_DIAG_TOS, inet->tos) < 0)
124 			goto errout;
125 
126 #if IS_ENABLED(CONFIG_IPV6)
127 	if (r->idiag_family == AF_INET6) {
128 
129 		*(struct in6_addr *)r->id.idiag_src = sk->sk_v6_rcv_saddr;
130 		*(struct in6_addr *)r->id.idiag_dst = sk->sk_v6_daddr;
131 
132 		if (ext & (1 << (INET_DIAG_TCLASS - 1)))
133 			if (nla_put_u8(skb, INET_DIAG_TCLASS,
134 				       inet6_sk(sk)->tclass) < 0)
135 				goto errout;
136 	}
137 #endif
138 
139 	r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk));
140 	r->idiag_inode = sock_i_ino(sk);
141 
142 	if (ext & (1 << (INET_DIAG_MEMINFO - 1))) {
143 		struct inet_diag_meminfo minfo = {
144 			.idiag_rmem = sk_rmem_alloc_get(sk),
145 			.idiag_wmem = sk->sk_wmem_queued,
146 			.idiag_fmem = sk->sk_forward_alloc,
147 			.idiag_tmem = sk_wmem_alloc_get(sk),
148 		};
149 
150 		if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0)
151 			goto errout;
152 	}
153 
154 	if (ext & (1 << (INET_DIAG_SKMEMINFO - 1)))
155 		if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO))
156 			goto errout;
157 
158 	if (icsk == NULL) {
159 		handler->idiag_get_info(sk, r, NULL);
160 		goto out;
161 	}
162 
163 #define EXPIRES_IN_MS(tmo)  DIV_ROUND_UP((tmo - jiffies) * 1000, HZ)
164 
165 	if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
166 	    icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS ||
167 	    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
168 		r->idiag_timer = 1;
169 		r->idiag_retrans = icsk->icsk_retransmits;
170 		r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
171 	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
172 		r->idiag_timer = 4;
173 		r->idiag_retrans = icsk->icsk_probes_out;
174 		r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
175 	} else if (timer_pending(&sk->sk_timer)) {
176 		r->idiag_timer = 2;
177 		r->idiag_retrans = icsk->icsk_probes_out;
178 		r->idiag_expires = EXPIRES_IN_MS(sk->sk_timer.expires);
179 	} else {
180 		r->idiag_timer = 0;
181 		r->idiag_expires = 0;
182 	}
183 #undef EXPIRES_IN_MS
184 
185 	if (ext & (1 << (INET_DIAG_INFO - 1))) {
186 		attr = nla_reserve(skb, INET_DIAG_INFO,
187 				   sizeof(struct tcp_info));
188 		if (!attr)
189 			goto errout;
190 
191 		info = nla_data(attr);
192 	}
193 
194 	if ((ext & (1 << (INET_DIAG_CONG - 1))) && icsk->icsk_ca_ops)
195 		if (nla_put_string(skb, INET_DIAG_CONG,
196 				   icsk->icsk_ca_ops->name) < 0)
197 			goto errout;
198 
199 	handler->idiag_get_info(sk, r, info);
200 
201 	if (sk->sk_state < TCP_TIME_WAIT &&
202 	    icsk->icsk_ca_ops && icsk->icsk_ca_ops->get_info)
203 		icsk->icsk_ca_ops->get_info(sk, ext, skb);
204 
205 out:
206 	nlmsg_end(skb, nlh);
207 	return 0;
208 
209 errout:
210 	nlmsg_cancel(skb, nlh);
211 	return -EMSGSIZE;
212 }
213 EXPORT_SYMBOL_GPL(inet_sk_diag_fill);
214 
215 static int inet_csk_diag_fill(struct sock *sk,
216 			      struct sk_buff *skb, struct inet_diag_req_v2 *req,
217 			      struct user_namespace *user_ns,
218 			      u32 portid, u32 seq, u16 nlmsg_flags,
219 			      const struct nlmsghdr *unlh)
220 {
221 	return inet_sk_diag_fill(sk, inet_csk(sk),
222 			skb, req, user_ns, portid, seq, nlmsg_flags, unlh);
223 }
224 
225 static int inet_twsk_diag_fill(struct inet_timewait_sock *tw,
226 			       struct sk_buff *skb, struct inet_diag_req_v2 *req,
227 			       u32 portid, u32 seq, u16 nlmsg_flags,
228 			       const struct nlmsghdr *unlh)
229 {
230 	s32 tmo;
231 	struct inet_diag_msg *r;
232 	struct nlmsghdr *nlh;
233 
234 	nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
235 			nlmsg_flags);
236 	if (!nlh)
237 		return -EMSGSIZE;
238 
239 	r = nlmsg_data(nlh);
240 	BUG_ON(tw->tw_state != TCP_TIME_WAIT);
241 
242 	tmo = tw->tw_ttd - inet_tw_time_stamp();
243 	if (tmo < 0)
244 		tmo = 0;
245 
246 	r->idiag_family	      = tw->tw_family;
247 	r->idiag_retrans      = 0;
248 
249 	r->id.idiag_if	      = tw->tw_bound_dev_if;
250 	sock_diag_save_cookie(tw, r->id.idiag_cookie);
251 
252 	r->id.idiag_sport     = tw->tw_sport;
253 	r->id.idiag_dport     = tw->tw_dport;
254 
255 	memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src));
256 	memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst));
257 
258 	r->id.idiag_src[0]    = tw->tw_rcv_saddr;
259 	r->id.idiag_dst[0]    = tw->tw_daddr;
260 
261 	r->idiag_state	      = tw->tw_substate;
262 	r->idiag_timer	      = 3;
263 	r->idiag_expires      = jiffies_to_msecs(tmo);
264 	r->idiag_rqueue	      = 0;
265 	r->idiag_wqueue	      = 0;
266 	r->idiag_uid	      = 0;
267 	r->idiag_inode	      = 0;
268 #if IS_ENABLED(CONFIG_IPV6)
269 	if (tw->tw_family == AF_INET6) {
270 		*(struct in6_addr *)r->id.idiag_src = tw->tw_v6_rcv_saddr;
271 		*(struct in6_addr *)r->id.idiag_dst = tw->tw_v6_daddr;
272 	}
273 #endif
274 
275 	nlmsg_end(skb, nlh);
276 	return 0;
277 }
278 
279 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
280 			struct inet_diag_req_v2 *r,
281 			struct user_namespace *user_ns,
282 			u32 portid, u32 seq, u16 nlmsg_flags,
283 			const struct nlmsghdr *unlh)
284 {
285 	if (sk->sk_state == TCP_TIME_WAIT)
286 		return inet_twsk_diag_fill(inet_twsk(sk), skb, r, portid, seq,
287 					   nlmsg_flags, unlh);
288 
289 	return inet_csk_diag_fill(sk, skb, r, user_ns, portid, seq,
290 				  nlmsg_flags, unlh);
291 }
292 
293 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *in_skb,
294 		const struct nlmsghdr *nlh, struct inet_diag_req_v2 *req)
295 {
296 	int err;
297 	struct sock *sk;
298 	struct sk_buff *rep;
299 	struct net *net = sock_net(in_skb->sk);
300 
301 	err = -EINVAL;
302 	if (req->sdiag_family == AF_INET) {
303 		sk = inet_lookup(net, hashinfo, req->id.idiag_dst[0],
304 				 req->id.idiag_dport, req->id.idiag_src[0],
305 				 req->id.idiag_sport, req->id.idiag_if);
306 	}
307 #if IS_ENABLED(CONFIG_IPV6)
308 	else if (req->sdiag_family == AF_INET6) {
309 		sk = inet6_lookup(net, hashinfo,
310 				  (struct in6_addr *)req->id.idiag_dst,
311 				  req->id.idiag_dport,
312 				  (struct in6_addr *)req->id.idiag_src,
313 				  req->id.idiag_sport,
314 				  req->id.idiag_if);
315 	}
316 #endif
317 	else {
318 		goto out_nosk;
319 	}
320 
321 	err = -ENOENT;
322 	if (sk == NULL)
323 		goto out_nosk;
324 
325 	err = sock_diag_check_cookie(sk, req->id.idiag_cookie);
326 	if (err)
327 		goto out;
328 
329 	rep = nlmsg_new(sizeof(struct inet_diag_msg) +
330 			sizeof(struct inet_diag_meminfo) +
331 			sizeof(struct tcp_info) + 64, GFP_KERNEL);
332 	if (!rep) {
333 		err = -ENOMEM;
334 		goto out;
335 	}
336 
337 	err = sk_diag_fill(sk, rep, req,
338 			   sk_user_ns(NETLINK_CB(in_skb).sk),
339 			   NETLINK_CB(in_skb).portid,
340 			   nlh->nlmsg_seq, 0, nlh);
341 	if (err < 0) {
342 		WARN_ON(err == -EMSGSIZE);
343 		nlmsg_free(rep);
344 		goto out;
345 	}
346 	err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid,
347 			      MSG_DONTWAIT);
348 	if (err > 0)
349 		err = 0;
350 
351 out:
352 	if (sk)
353 		sock_gen_put(sk);
354 
355 out_nosk:
356 	return err;
357 }
358 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
359 
360 static int inet_diag_get_exact(struct sk_buff *in_skb,
361 			       const struct nlmsghdr *nlh,
362 			       struct inet_diag_req_v2 *req)
363 {
364 	const struct inet_diag_handler *handler;
365 	int err;
366 
367 	handler = inet_diag_lock_handler(req->sdiag_protocol);
368 	if (IS_ERR(handler))
369 		err = PTR_ERR(handler);
370 	else
371 		err = handler->dump_one(in_skb, nlh, req);
372 	inet_diag_unlock_handler(handler);
373 
374 	return err;
375 }
376 
377 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
378 {
379 	int words = bits >> 5;
380 
381 	bits &= 0x1f;
382 
383 	if (words) {
384 		if (memcmp(a1, a2, words << 2))
385 			return 0;
386 	}
387 	if (bits) {
388 		__be32 w1, w2;
389 		__be32 mask;
390 
391 		w1 = a1[words];
392 		w2 = a2[words];
393 
394 		mask = htonl((0xffffffff) << (32 - bits));
395 
396 		if ((w1 ^ w2) & mask)
397 			return 0;
398 	}
399 
400 	return 1;
401 }
402 
403 
404 static int inet_diag_bc_run(const struct nlattr *_bc,
405 		const struct inet_diag_entry *entry)
406 {
407 	const void *bc = nla_data(_bc);
408 	int len = nla_len(_bc);
409 
410 	while (len > 0) {
411 		int yes = 1;
412 		const struct inet_diag_bc_op *op = bc;
413 
414 		switch (op->code) {
415 		case INET_DIAG_BC_NOP:
416 			break;
417 		case INET_DIAG_BC_JMP:
418 			yes = 0;
419 			break;
420 		case INET_DIAG_BC_S_GE:
421 			yes = entry->sport >= op[1].no;
422 			break;
423 		case INET_DIAG_BC_S_LE:
424 			yes = entry->sport <= op[1].no;
425 			break;
426 		case INET_DIAG_BC_D_GE:
427 			yes = entry->dport >= op[1].no;
428 			break;
429 		case INET_DIAG_BC_D_LE:
430 			yes = entry->dport <= op[1].no;
431 			break;
432 		case INET_DIAG_BC_AUTO:
433 			yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
434 			break;
435 		case INET_DIAG_BC_S_COND:
436 		case INET_DIAG_BC_D_COND: {
437 			struct inet_diag_hostcond *cond;
438 			__be32 *addr;
439 
440 			cond = (struct inet_diag_hostcond *)(op + 1);
441 			if (cond->port != -1 &&
442 			    cond->port != (op->code == INET_DIAG_BC_S_COND ?
443 					     entry->sport : entry->dport)) {
444 				yes = 0;
445 				break;
446 			}
447 
448 			if (op->code == INET_DIAG_BC_S_COND)
449 				addr = entry->saddr;
450 			else
451 				addr = entry->daddr;
452 
453 			if (cond->family != AF_UNSPEC &&
454 			    cond->family != entry->family) {
455 				if (entry->family == AF_INET6 &&
456 				    cond->family == AF_INET) {
457 					if (addr[0] == 0 && addr[1] == 0 &&
458 					    addr[2] == htonl(0xffff) &&
459 					    bitstring_match(addr + 3,
460 							    cond->addr,
461 							    cond->prefix_len))
462 						break;
463 				}
464 				yes = 0;
465 				break;
466 			}
467 
468 			if (cond->prefix_len == 0)
469 				break;
470 			if (bitstring_match(addr, cond->addr,
471 					    cond->prefix_len))
472 				break;
473 			yes = 0;
474 			break;
475 		}
476 		}
477 
478 		if (yes) {
479 			len -= op->yes;
480 			bc += op->yes;
481 		} else {
482 			len -= op->no;
483 			bc += op->no;
484 		}
485 	}
486 	return len == 0;
487 }
488 
489 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
490 {
491 	struct inet_diag_entry entry;
492 	struct inet_sock *inet = inet_sk(sk);
493 
494 	if (bc == NULL)
495 		return 1;
496 
497 	entry.family = sk->sk_family;
498 #if IS_ENABLED(CONFIG_IPV6)
499 	if (entry.family == AF_INET6) {
500 
501 		entry.saddr = sk->sk_v6_rcv_saddr.s6_addr32;
502 		entry.daddr = sk->sk_v6_daddr.s6_addr32;
503 	} else
504 #endif
505 	{
506 		entry.saddr = &inet->inet_rcv_saddr;
507 		entry.daddr = &inet->inet_daddr;
508 	}
509 	entry.sport = inet->inet_num;
510 	entry.dport = ntohs(inet->inet_dport);
511 	entry.userlocks = sk->sk_userlocks;
512 
513 	return inet_diag_bc_run(bc, &entry);
514 }
515 EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
516 
517 static int valid_cc(const void *bc, int len, int cc)
518 {
519 	while (len >= 0) {
520 		const struct inet_diag_bc_op *op = bc;
521 
522 		if (cc > len)
523 			return 0;
524 		if (cc == len)
525 			return 1;
526 		if (op->yes < 4 || op->yes & 3)
527 			return 0;
528 		len -= op->yes;
529 		bc  += op->yes;
530 	}
531 	return 0;
532 }
533 
534 /* Validate an inet_diag_hostcond. */
535 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
536 			   int *min_len)
537 {
538 	int addr_len;
539 	struct inet_diag_hostcond *cond;
540 
541 	/* Check hostcond space. */
542 	*min_len += sizeof(struct inet_diag_hostcond);
543 	if (len < *min_len)
544 		return false;
545 	cond = (struct inet_diag_hostcond *)(op + 1);
546 
547 	/* Check address family and address length. */
548 	switch (cond->family) {
549 	case AF_UNSPEC:
550 		addr_len = 0;
551 		break;
552 	case AF_INET:
553 		addr_len = sizeof(struct in_addr);
554 		break;
555 	case AF_INET6:
556 		addr_len = sizeof(struct in6_addr);
557 		break;
558 	default:
559 		return false;
560 	}
561 	*min_len += addr_len;
562 	if (len < *min_len)
563 		return false;
564 
565 	/* Check prefix length (in bits) vs address length (in bytes). */
566 	if (cond->prefix_len > 8 * addr_len)
567 		return false;
568 
569 	return true;
570 }
571 
572 /* Validate a port comparison operator. */
573 static inline bool valid_port_comparison(const struct inet_diag_bc_op *op,
574 					 int len, int *min_len)
575 {
576 	/* Port comparisons put the port in a follow-on inet_diag_bc_op. */
577 	*min_len += sizeof(struct inet_diag_bc_op);
578 	if (len < *min_len)
579 		return false;
580 	return true;
581 }
582 
583 static int inet_diag_bc_audit(const void *bytecode, int bytecode_len)
584 {
585 	const void *bc = bytecode;
586 	int  len = bytecode_len;
587 
588 	while (len > 0) {
589 		const struct inet_diag_bc_op *op = bc;
590 		int min_len = sizeof(struct inet_diag_bc_op);
591 
592 //printk("BC: %d %d %d {%d} / %d\n", op->code, op->yes, op->no, op[1].no, len);
593 		switch (op->code) {
594 		case INET_DIAG_BC_S_COND:
595 		case INET_DIAG_BC_D_COND:
596 			if (!valid_hostcond(bc, len, &min_len))
597 				return -EINVAL;
598 			break;
599 		case INET_DIAG_BC_S_GE:
600 		case INET_DIAG_BC_S_LE:
601 		case INET_DIAG_BC_D_GE:
602 		case INET_DIAG_BC_D_LE:
603 			if (!valid_port_comparison(bc, len, &min_len))
604 				return -EINVAL;
605 			break;
606 		case INET_DIAG_BC_AUTO:
607 		case INET_DIAG_BC_JMP:
608 		case INET_DIAG_BC_NOP:
609 			break;
610 		default:
611 			return -EINVAL;
612 		}
613 
614 		if (op->code != INET_DIAG_BC_NOP) {
615 			if (op->no < min_len || op->no > len + 4 || op->no & 3)
616 				return -EINVAL;
617 			if (op->no < len &&
618 			    !valid_cc(bytecode, bytecode_len, len - op->no))
619 				return -EINVAL;
620 		}
621 
622 		if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
623 			return -EINVAL;
624 		bc  += op->yes;
625 		len -= op->yes;
626 	}
627 	return len == 0 ? 0 : -EINVAL;
628 }
629 
630 static int inet_csk_diag_dump(struct sock *sk,
631 			      struct sk_buff *skb,
632 			      struct netlink_callback *cb,
633 			      struct inet_diag_req_v2 *r,
634 			      const struct nlattr *bc)
635 {
636 	if (!inet_diag_bc_sk(bc, sk))
637 		return 0;
638 
639 	return inet_csk_diag_fill(sk, skb, r,
640 				  sk_user_ns(NETLINK_CB(cb->skb).sk),
641 				  NETLINK_CB(cb->skb).portid,
642 				  cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
643 }
644 
645 static int inet_twsk_diag_dump(struct sock *sk,
646 			       struct sk_buff *skb,
647 			       struct netlink_callback *cb,
648 			       struct inet_diag_req_v2 *r,
649 			       const struct nlattr *bc)
650 {
651 	struct inet_timewait_sock *tw = inet_twsk(sk);
652 
653 	if (bc != NULL) {
654 		struct inet_diag_entry entry;
655 
656 		entry.family = tw->tw_family;
657 #if IS_ENABLED(CONFIG_IPV6)
658 		if (tw->tw_family == AF_INET6) {
659 			entry.saddr = tw->tw_v6_rcv_saddr.s6_addr32;
660 			entry.daddr = tw->tw_v6_daddr.s6_addr32;
661 		} else
662 #endif
663 		{
664 			entry.saddr = &tw->tw_rcv_saddr;
665 			entry.daddr = &tw->tw_daddr;
666 		}
667 		entry.sport = tw->tw_num;
668 		entry.dport = ntohs(tw->tw_dport);
669 		entry.userlocks = 0;
670 
671 		if (!inet_diag_bc_run(bc, &entry))
672 			return 0;
673 	}
674 
675 	return inet_twsk_diag_fill(tw, skb, r,
676 				   NETLINK_CB(cb->skb).portid,
677 				   cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
678 }
679 
680 /* Get the IPv4, IPv6, or IPv4-mapped-IPv6 local and remote addresses
681  * from a request_sock. For IPv4-mapped-IPv6 we must map IPv4 to IPv6.
682  */
683 static inline void inet_diag_req_addrs(const struct sock *sk,
684 				       const struct request_sock *req,
685 				       struct inet_diag_entry *entry)
686 {
687 	struct inet_request_sock *ireq = inet_rsk(req);
688 
689 #if IS_ENABLED(CONFIG_IPV6)
690 	if (sk->sk_family == AF_INET6) {
691 		if (req->rsk_ops->family == AF_INET6) {
692 			entry->saddr = ireq->ir_v6_loc_addr.s6_addr32;
693 			entry->daddr = ireq->ir_v6_rmt_addr.s6_addr32;
694 		} else if (req->rsk_ops->family == AF_INET) {
695 			ipv6_addr_set_v4mapped(ireq->ir_loc_addr,
696 					       &entry->saddr_storage);
697 			ipv6_addr_set_v4mapped(ireq->ir_rmt_addr,
698 					       &entry->daddr_storage);
699 			entry->saddr = entry->saddr_storage.s6_addr32;
700 			entry->daddr = entry->daddr_storage.s6_addr32;
701 		}
702 	} else
703 #endif
704 	{
705 		entry->saddr = &ireq->ir_loc_addr;
706 		entry->daddr = &ireq->ir_rmt_addr;
707 	}
708 }
709 
710 static int inet_diag_fill_req(struct sk_buff *skb, struct sock *sk,
711 			      struct request_sock *req,
712 			      struct user_namespace *user_ns,
713 			      u32 portid, u32 seq,
714 			      const struct nlmsghdr *unlh)
715 {
716 	const struct inet_request_sock *ireq = inet_rsk(req);
717 	struct inet_sock *inet = inet_sk(sk);
718 	struct inet_diag_msg *r;
719 	struct nlmsghdr *nlh;
720 	long tmo;
721 
722 	nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
723 			NLM_F_MULTI);
724 	if (!nlh)
725 		return -EMSGSIZE;
726 
727 	r = nlmsg_data(nlh);
728 	r->idiag_family = sk->sk_family;
729 	r->idiag_state = TCP_SYN_RECV;
730 	r->idiag_timer = 1;
731 	r->idiag_retrans = req->num_retrans;
732 
733 	r->id.idiag_if = sk->sk_bound_dev_if;
734 	sock_diag_save_cookie(req, r->id.idiag_cookie);
735 
736 	tmo = req->expires - jiffies;
737 	if (tmo < 0)
738 		tmo = 0;
739 
740 	r->id.idiag_sport = inet->inet_sport;
741 	r->id.idiag_dport = ireq->ir_rmt_port;
742 
743 	memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src));
744 	memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst));
745 
746 	r->id.idiag_src[0] = ireq->ir_loc_addr;
747 	r->id.idiag_dst[0] = ireq->ir_rmt_addr;
748 
749 	r->idiag_expires = jiffies_to_msecs(tmo);
750 	r->idiag_rqueue = 0;
751 	r->idiag_wqueue = 0;
752 	r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk));
753 	r->idiag_inode = 0;
754 #if IS_ENABLED(CONFIG_IPV6)
755 	if (r->idiag_family == AF_INET6) {
756 		struct inet_diag_entry entry;
757 		inet_diag_req_addrs(sk, req, &entry);
758 		memcpy(r->id.idiag_src, entry.saddr, sizeof(struct in6_addr));
759 		memcpy(r->id.idiag_dst, entry.daddr, sizeof(struct in6_addr));
760 	}
761 #endif
762 
763 	nlmsg_end(skb, nlh);
764 	return 0;
765 }
766 
767 static int inet_diag_dump_reqs(struct sk_buff *skb, struct sock *sk,
768 			       struct netlink_callback *cb,
769 			       struct inet_diag_req_v2 *r,
770 			       const struct nlattr *bc)
771 {
772 	struct inet_diag_entry entry;
773 	struct inet_connection_sock *icsk = inet_csk(sk);
774 	struct listen_sock *lopt;
775 	struct inet_sock *inet = inet_sk(sk);
776 	int j, s_j;
777 	int reqnum, s_reqnum;
778 	int err = 0;
779 
780 	s_j = cb->args[3];
781 	s_reqnum = cb->args[4];
782 
783 	if (s_j > 0)
784 		s_j--;
785 
786 	entry.family = sk->sk_family;
787 
788 	read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
789 
790 	lopt = icsk->icsk_accept_queue.listen_opt;
791 	if (!lopt || !lopt->qlen)
792 		goto out;
793 
794 	if (bc != NULL) {
795 		entry.sport = inet->inet_num;
796 		entry.userlocks = sk->sk_userlocks;
797 	}
798 
799 	for (j = s_j; j < lopt->nr_table_entries; j++) {
800 		struct request_sock *req, *head = lopt->syn_table[j];
801 
802 		reqnum = 0;
803 		for (req = head; req; reqnum++, req = req->dl_next) {
804 			struct inet_request_sock *ireq = inet_rsk(req);
805 
806 			if (reqnum < s_reqnum)
807 				continue;
808 			if (r->id.idiag_dport != ireq->ir_rmt_port &&
809 			    r->id.idiag_dport)
810 				continue;
811 
812 			if (bc) {
813 				inet_diag_req_addrs(sk, req, &entry);
814 				entry.dport = ntohs(ireq->ir_rmt_port);
815 
816 				if (!inet_diag_bc_run(bc, &entry))
817 					continue;
818 			}
819 
820 			err = inet_diag_fill_req(skb, sk, req,
821 					       sk_user_ns(NETLINK_CB(cb->skb).sk),
822 					       NETLINK_CB(cb->skb).portid,
823 					       cb->nlh->nlmsg_seq, cb->nlh);
824 			if (err < 0) {
825 				cb->args[3] = j + 1;
826 				cb->args[4] = reqnum;
827 				goto out;
828 			}
829 		}
830 
831 		s_reqnum = 0;
832 	}
833 
834 out:
835 	read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
836 
837 	return err;
838 }
839 
840 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
841 		struct netlink_callback *cb, struct inet_diag_req_v2 *r, struct nlattr *bc)
842 {
843 	int i, num;
844 	int s_i, s_num;
845 	struct net *net = sock_net(skb->sk);
846 
847 	s_i = cb->args[1];
848 	s_num = num = cb->args[2];
849 
850 	if (cb->args[0] == 0) {
851 		if (!(r->idiag_states & (TCPF_LISTEN | TCPF_SYN_RECV)))
852 			goto skip_listen_ht;
853 
854 		for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
855 			struct sock *sk;
856 			struct hlist_nulls_node *node;
857 			struct inet_listen_hashbucket *ilb;
858 
859 			num = 0;
860 			ilb = &hashinfo->listening_hash[i];
861 			spin_lock_bh(&ilb->lock);
862 			sk_nulls_for_each(sk, node, &ilb->head) {
863 				struct inet_sock *inet = inet_sk(sk);
864 
865 				if (!net_eq(sock_net(sk), net))
866 					continue;
867 
868 				if (num < s_num) {
869 					num++;
870 					continue;
871 				}
872 
873 				if (r->sdiag_family != AF_UNSPEC &&
874 						sk->sk_family != r->sdiag_family)
875 					goto next_listen;
876 
877 				if (r->id.idiag_sport != inet->inet_sport &&
878 				    r->id.idiag_sport)
879 					goto next_listen;
880 
881 				if (!(r->idiag_states & TCPF_LISTEN) ||
882 				    r->id.idiag_dport ||
883 				    cb->args[3] > 0)
884 					goto syn_recv;
885 
886 				if (inet_csk_diag_dump(sk, skb, cb, r, bc) < 0) {
887 					spin_unlock_bh(&ilb->lock);
888 					goto done;
889 				}
890 
891 syn_recv:
892 				if (!(r->idiag_states & TCPF_SYN_RECV))
893 					goto next_listen;
894 
895 				if (inet_diag_dump_reqs(skb, sk, cb, r, bc) < 0) {
896 					spin_unlock_bh(&ilb->lock);
897 					goto done;
898 				}
899 
900 next_listen:
901 				cb->args[3] = 0;
902 				cb->args[4] = 0;
903 				++num;
904 			}
905 			spin_unlock_bh(&ilb->lock);
906 
907 			s_num = 0;
908 			cb->args[3] = 0;
909 			cb->args[4] = 0;
910 		}
911 skip_listen_ht:
912 		cb->args[0] = 1;
913 		s_i = num = s_num = 0;
914 	}
915 
916 	if (!(r->idiag_states & ~(TCPF_LISTEN | TCPF_SYN_RECV)))
917 		goto out;
918 
919 	for (i = s_i; i <= hashinfo->ehash_mask; i++) {
920 		struct inet_ehash_bucket *head = &hashinfo->ehash[i];
921 		spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
922 		struct sock *sk;
923 		struct hlist_nulls_node *node;
924 
925 		num = 0;
926 
927 		if (hlist_nulls_empty(&head->chain))
928 			continue;
929 
930 		if (i > s_i)
931 			s_num = 0;
932 
933 		spin_lock_bh(lock);
934 		sk_nulls_for_each(sk, node, &head->chain) {
935 			int res;
936 			int state;
937 
938 			if (!net_eq(sock_net(sk), net))
939 				continue;
940 			if (num < s_num)
941 				goto next_normal;
942 			state = (sk->sk_state == TCP_TIME_WAIT) ?
943 				inet_twsk(sk)->tw_substate : sk->sk_state;
944 			if (!(r->idiag_states & (1 << state)))
945 				goto next_normal;
946 			if (r->sdiag_family != AF_UNSPEC &&
947 			    sk->sk_family != r->sdiag_family)
948 				goto next_normal;
949 			if (r->id.idiag_sport != htons(sk->sk_num) &&
950 			    r->id.idiag_sport)
951 				goto next_normal;
952 			if (r->id.idiag_dport != sk->sk_dport &&
953 			    r->id.idiag_dport)
954 				goto next_normal;
955 			if (sk->sk_state == TCP_TIME_WAIT)
956 				res = inet_twsk_diag_dump(sk, skb, cb, r, bc);
957 			else
958 				res = inet_csk_diag_dump(sk, skb, cb, r, bc);
959 			if (res < 0) {
960 				spin_unlock_bh(lock);
961 				goto done;
962 			}
963 next_normal:
964 			++num;
965 		}
966 
967 		spin_unlock_bh(lock);
968 	}
969 
970 done:
971 	cb->args[1] = i;
972 	cb->args[2] = num;
973 out:
974 	;
975 }
976 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
977 
978 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
979 		struct inet_diag_req_v2 *r, struct nlattr *bc)
980 {
981 	const struct inet_diag_handler *handler;
982 	int err = 0;
983 
984 	handler = inet_diag_lock_handler(r->sdiag_protocol);
985 	if (!IS_ERR(handler))
986 		handler->dump(skb, cb, r, bc);
987 	else
988 		err = PTR_ERR(handler);
989 	inet_diag_unlock_handler(handler);
990 
991 	return err ? : skb->len;
992 }
993 
994 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
995 {
996 	struct nlattr *bc = NULL;
997 	int hdrlen = sizeof(struct inet_diag_req_v2);
998 
999 	if (nlmsg_attrlen(cb->nlh, hdrlen))
1000 		bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE);
1001 
1002 	return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh), bc);
1003 }
1004 
1005 static inline int inet_diag_type2proto(int type)
1006 {
1007 	switch (type) {
1008 	case TCPDIAG_GETSOCK:
1009 		return IPPROTO_TCP;
1010 	case DCCPDIAG_GETSOCK:
1011 		return IPPROTO_DCCP;
1012 	default:
1013 		return 0;
1014 	}
1015 }
1016 
1017 static int inet_diag_dump_compat(struct sk_buff *skb, struct netlink_callback *cb)
1018 {
1019 	struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1020 	struct inet_diag_req_v2 req;
1021 	struct nlattr *bc = NULL;
1022 	int hdrlen = sizeof(struct inet_diag_req);
1023 
1024 	req.sdiag_family = AF_UNSPEC; /* compatibility */
1025 	req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1026 	req.idiag_ext = rc->idiag_ext;
1027 	req.idiag_states = rc->idiag_states;
1028 	req.id = rc->id;
1029 
1030 	if (nlmsg_attrlen(cb->nlh, hdrlen))
1031 		bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE);
1032 
1033 	return __inet_diag_dump(skb, cb, &req, bc);
1034 }
1035 
1036 static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1037 			       const struct nlmsghdr *nlh)
1038 {
1039 	struct inet_diag_req *rc = nlmsg_data(nlh);
1040 	struct inet_diag_req_v2 req;
1041 
1042 	req.sdiag_family = rc->idiag_family;
1043 	req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1044 	req.idiag_ext = rc->idiag_ext;
1045 	req.idiag_states = rc->idiag_states;
1046 	req.id = rc->id;
1047 
1048 	return inet_diag_get_exact(in_skb, nlh, &req);
1049 }
1050 
1051 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1052 {
1053 	int hdrlen = sizeof(struct inet_diag_req);
1054 	struct net *net = sock_net(skb->sk);
1055 
1056 	if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1057 	    nlmsg_len(nlh) < hdrlen)
1058 		return -EINVAL;
1059 
1060 	if (nlh->nlmsg_flags & NLM_F_DUMP) {
1061 		if (nlmsg_attrlen(nlh, hdrlen)) {
1062 			struct nlattr *attr;
1063 
1064 			attr = nlmsg_find_attr(nlh, hdrlen,
1065 					       INET_DIAG_REQ_BYTECODE);
1066 			if (attr == NULL ||
1067 			    nla_len(attr) < sizeof(struct inet_diag_bc_op) ||
1068 			    inet_diag_bc_audit(nla_data(attr), nla_len(attr)))
1069 				return -EINVAL;
1070 		}
1071 		{
1072 			struct netlink_dump_control c = {
1073 				.dump = inet_diag_dump_compat,
1074 			};
1075 			return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1076 		}
1077 	}
1078 
1079 	return inet_diag_get_exact_compat(skb, nlh);
1080 }
1081 
1082 static int inet_diag_handler_dump(struct sk_buff *skb, struct nlmsghdr *h)
1083 {
1084 	int hdrlen = sizeof(struct inet_diag_req_v2);
1085 	struct net *net = sock_net(skb->sk);
1086 
1087 	if (nlmsg_len(h) < hdrlen)
1088 		return -EINVAL;
1089 
1090 	if (h->nlmsg_flags & NLM_F_DUMP) {
1091 		if (nlmsg_attrlen(h, hdrlen)) {
1092 			struct nlattr *attr;
1093 			attr = nlmsg_find_attr(h, hdrlen,
1094 					       INET_DIAG_REQ_BYTECODE);
1095 			if (attr == NULL ||
1096 			    nla_len(attr) < sizeof(struct inet_diag_bc_op) ||
1097 			    inet_diag_bc_audit(nla_data(attr), nla_len(attr)))
1098 				return -EINVAL;
1099 		}
1100 		{
1101 			struct netlink_dump_control c = {
1102 				.dump = inet_diag_dump,
1103 			};
1104 			return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1105 		}
1106 	}
1107 
1108 	return inet_diag_get_exact(skb, h, nlmsg_data(h));
1109 }
1110 
1111 static const struct sock_diag_handler inet_diag_handler = {
1112 	.family = AF_INET,
1113 	.dump = inet_diag_handler_dump,
1114 };
1115 
1116 static const struct sock_diag_handler inet6_diag_handler = {
1117 	.family = AF_INET6,
1118 	.dump = inet_diag_handler_dump,
1119 };
1120 
1121 int inet_diag_register(const struct inet_diag_handler *h)
1122 {
1123 	const __u16 type = h->idiag_type;
1124 	int err = -EINVAL;
1125 
1126 	if (type >= IPPROTO_MAX)
1127 		goto out;
1128 
1129 	mutex_lock(&inet_diag_table_mutex);
1130 	err = -EEXIST;
1131 	if (inet_diag_table[type] == NULL) {
1132 		inet_diag_table[type] = h;
1133 		err = 0;
1134 	}
1135 	mutex_unlock(&inet_diag_table_mutex);
1136 out:
1137 	return err;
1138 }
1139 EXPORT_SYMBOL_GPL(inet_diag_register);
1140 
1141 void inet_diag_unregister(const struct inet_diag_handler *h)
1142 {
1143 	const __u16 type = h->idiag_type;
1144 
1145 	if (type >= IPPROTO_MAX)
1146 		return;
1147 
1148 	mutex_lock(&inet_diag_table_mutex);
1149 	inet_diag_table[type] = NULL;
1150 	mutex_unlock(&inet_diag_table_mutex);
1151 }
1152 EXPORT_SYMBOL_GPL(inet_diag_unregister);
1153 
1154 static int __init inet_diag_init(void)
1155 {
1156 	const int inet_diag_table_size = (IPPROTO_MAX *
1157 					  sizeof(struct inet_diag_handler *));
1158 	int err = -ENOMEM;
1159 
1160 	inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1161 	if (!inet_diag_table)
1162 		goto out;
1163 
1164 	err = sock_diag_register(&inet_diag_handler);
1165 	if (err)
1166 		goto out_free_nl;
1167 
1168 	err = sock_diag_register(&inet6_diag_handler);
1169 	if (err)
1170 		goto out_free_inet;
1171 
1172 	sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1173 out:
1174 	return err;
1175 
1176 out_free_inet:
1177 	sock_diag_unregister(&inet_diag_handler);
1178 out_free_nl:
1179 	kfree(inet_diag_table);
1180 	goto out;
1181 }
1182 
1183 static void __exit inet_diag_exit(void)
1184 {
1185 	sock_diag_unregister(&inet6_diag_handler);
1186 	sock_diag_unregister(&inet_diag_handler);
1187 	sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1188 	kfree(inet_diag_table);
1189 }
1190 
1191 module_init(inet_diag_init);
1192 module_exit(inet_diag_exit);
1193 MODULE_LICENSE("GPL");
1194 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1195 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);
1196