xref: /openbmc/linux/net/ipv4/inet_diag.c (revision d0b73b48)
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 	r->id.idiag_src[0] = inet->inet_rcv_saddr;
110 	r->id.idiag_dst[0] = inet->inet_daddr;
111 
112 	if (nla_put_u8(skb, INET_DIAG_SHUTDOWN, sk->sk_shutdown))
113 		goto errout;
114 
115 	/* IPv6 dual-stack sockets use inet->tos for IPv4 connections,
116 	 * hence this needs to be included regardless of socket family.
117 	 */
118 	if (ext & (1 << (INET_DIAG_TOS - 1)))
119 		if (nla_put_u8(skb, INET_DIAG_TOS, inet->tos) < 0)
120 			goto errout;
121 
122 #if IS_ENABLED(CONFIG_IPV6)
123 	if (r->idiag_family == AF_INET6) {
124 		const struct ipv6_pinfo *np = inet6_sk(sk);
125 
126 		*(struct in6_addr *)r->id.idiag_src = np->rcv_saddr;
127 		*(struct in6_addr *)r->id.idiag_dst = np->daddr;
128 
129 		if (ext & (1 << (INET_DIAG_TCLASS - 1)))
130 			if (nla_put_u8(skb, INET_DIAG_TCLASS, np->tclass) < 0)
131 				goto errout;
132 	}
133 #endif
134 
135 	r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk));
136 	r->idiag_inode = sock_i_ino(sk);
137 
138 	if (ext & (1 << (INET_DIAG_MEMINFO - 1))) {
139 		struct inet_diag_meminfo minfo = {
140 			.idiag_rmem = sk_rmem_alloc_get(sk),
141 			.idiag_wmem = sk->sk_wmem_queued,
142 			.idiag_fmem = sk->sk_forward_alloc,
143 			.idiag_tmem = sk_wmem_alloc_get(sk),
144 		};
145 
146 		if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0)
147 			goto errout;
148 	}
149 
150 	if (ext & (1 << (INET_DIAG_SKMEMINFO - 1)))
151 		if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO))
152 			goto errout;
153 
154 	if (icsk == NULL) {
155 		handler->idiag_get_info(sk, r, NULL);
156 		goto out;
157 	}
158 
159 #define EXPIRES_IN_MS(tmo)  DIV_ROUND_UP((tmo - jiffies) * 1000, HZ)
160 
161 	if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
162 		r->idiag_timer = 1;
163 		r->idiag_retrans = icsk->icsk_retransmits;
164 		r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
165 	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
166 		r->idiag_timer = 4;
167 		r->idiag_retrans = icsk->icsk_probes_out;
168 		r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
169 	} else if (timer_pending(&sk->sk_timer)) {
170 		r->idiag_timer = 2;
171 		r->idiag_retrans = icsk->icsk_probes_out;
172 		r->idiag_expires = EXPIRES_IN_MS(sk->sk_timer.expires);
173 	} else {
174 		r->idiag_timer = 0;
175 		r->idiag_expires = 0;
176 	}
177 #undef EXPIRES_IN_MS
178 
179 	if (ext & (1 << (INET_DIAG_INFO - 1))) {
180 		attr = nla_reserve(skb, INET_DIAG_INFO,
181 				   sizeof(struct tcp_info));
182 		if (!attr)
183 			goto errout;
184 
185 		info = nla_data(attr);
186 	}
187 
188 	if ((ext & (1 << (INET_DIAG_CONG - 1))) && icsk->icsk_ca_ops)
189 		if (nla_put_string(skb, INET_DIAG_CONG,
190 				   icsk->icsk_ca_ops->name) < 0)
191 			goto errout;
192 
193 	handler->idiag_get_info(sk, r, info);
194 
195 	if (sk->sk_state < TCP_TIME_WAIT &&
196 	    icsk->icsk_ca_ops && icsk->icsk_ca_ops->get_info)
197 		icsk->icsk_ca_ops->get_info(sk, ext, skb);
198 
199 out:
200 	return nlmsg_end(skb, nlh);
201 
202 errout:
203 	nlmsg_cancel(skb, nlh);
204 	return -EMSGSIZE;
205 }
206 EXPORT_SYMBOL_GPL(inet_sk_diag_fill);
207 
208 static int inet_csk_diag_fill(struct sock *sk,
209 			      struct sk_buff *skb, struct inet_diag_req_v2 *req,
210 			      struct user_namespace *user_ns,
211 			      u32 portid, u32 seq, u16 nlmsg_flags,
212 			      const struct nlmsghdr *unlh)
213 {
214 	return inet_sk_diag_fill(sk, inet_csk(sk),
215 			skb, req, user_ns, portid, seq, nlmsg_flags, unlh);
216 }
217 
218 static int inet_twsk_diag_fill(struct inet_timewait_sock *tw,
219 			       struct sk_buff *skb, struct inet_diag_req_v2 *req,
220 			       u32 portid, u32 seq, u16 nlmsg_flags,
221 			       const struct nlmsghdr *unlh)
222 {
223 	long tmo;
224 	struct inet_diag_msg *r;
225 	struct nlmsghdr *nlh;
226 
227 	nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
228 			nlmsg_flags);
229 	if (!nlh)
230 		return -EMSGSIZE;
231 
232 	r = nlmsg_data(nlh);
233 	BUG_ON(tw->tw_state != TCP_TIME_WAIT);
234 
235 	tmo = tw->tw_ttd - jiffies;
236 	if (tmo < 0)
237 		tmo = 0;
238 
239 	r->idiag_family	      = tw->tw_family;
240 	r->idiag_retrans      = 0;
241 	r->id.idiag_if	      = tw->tw_bound_dev_if;
242 	sock_diag_save_cookie(tw, r->id.idiag_cookie);
243 	r->id.idiag_sport     = tw->tw_sport;
244 	r->id.idiag_dport     = tw->tw_dport;
245 	r->id.idiag_src[0]    = tw->tw_rcv_saddr;
246 	r->id.idiag_dst[0]    = tw->tw_daddr;
247 	r->idiag_state	      = tw->tw_substate;
248 	r->idiag_timer	      = 3;
249 	r->idiag_expires      = DIV_ROUND_UP(tmo * 1000, HZ);
250 	r->idiag_rqueue	      = 0;
251 	r->idiag_wqueue	      = 0;
252 	r->idiag_uid	      = 0;
253 	r->idiag_inode	      = 0;
254 #if IS_ENABLED(CONFIG_IPV6)
255 	if (tw->tw_family == AF_INET6) {
256 		const struct inet6_timewait_sock *tw6 =
257 						inet6_twsk((struct sock *)tw);
258 
259 		*(struct in6_addr *)r->id.idiag_src = tw6->tw_v6_rcv_saddr;
260 		*(struct in6_addr *)r->id.idiag_dst = tw6->tw_v6_daddr;
261 	}
262 #endif
263 
264 	return nlmsg_end(skb, nlh);
265 }
266 
267 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
268 			struct inet_diag_req_v2 *r,
269 			struct user_namespace *user_ns,
270 			u32 portid, u32 seq, u16 nlmsg_flags,
271 			const struct nlmsghdr *unlh)
272 {
273 	if (sk->sk_state == TCP_TIME_WAIT)
274 		return inet_twsk_diag_fill((struct inet_timewait_sock *)sk,
275 					   skb, r, portid, seq, nlmsg_flags,
276 					   unlh);
277 	return inet_csk_diag_fill(sk, skb, r, user_ns, portid, seq, nlmsg_flags, unlh);
278 }
279 
280 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *in_skb,
281 		const struct nlmsghdr *nlh, struct inet_diag_req_v2 *req)
282 {
283 	int err;
284 	struct sock *sk;
285 	struct sk_buff *rep;
286 	struct net *net = sock_net(in_skb->sk);
287 
288 	err = -EINVAL;
289 	if (req->sdiag_family == AF_INET) {
290 		sk = inet_lookup(net, hashinfo, req->id.idiag_dst[0],
291 				 req->id.idiag_dport, req->id.idiag_src[0],
292 				 req->id.idiag_sport, req->id.idiag_if);
293 	}
294 #if IS_ENABLED(CONFIG_IPV6)
295 	else if (req->sdiag_family == AF_INET6) {
296 		sk = inet6_lookup(net, hashinfo,
297 				  (struct in6_addr *)req->id.idiag_dst,
298 				  req->id.idiag_dport,
299 				  (struct in6_addr *)req->id.idiag_src,
300 				  req->id.idiag_sport,
301 				  req->id.idiag_if);
302 	}
303 #endif
304 	else {
305 		goto out_nosk;
306 	}
307 
308 	err = -ENOENT;
309 	if (sk == NULL)
310 		goto out_nosk;
311 
312 	err = sock_diag_check_cookie(sk, req->id.idiag_cookie);
313 	if (err)
314 		goto out;
315 
316 	rep = nlmsg_new(sizeof(struct inet_diag_msg) +
317 			sizeof(struct inet_diag_meminfo) +
318 			sizeof(struct tcp_info) + 64, GFP_KERNEL);
319 	if (!rep) {
320 		err = -ENOMEM;
321 		goto out;
322 	}
323 
324 	err = sk_diag_fill(sk, rep, req,
325 			   sk_user_ns(NETLINK_CB(in_skb).ssk),
326 			   NETLINK_CB(in_skb).portid,
327 			   nlh->nlmsg_seq, 0, nlh);
328 	if (err < 0) {
329 		WARN_ON(err == -EMSGSIZE);
330 		nlmsg_free(rep);
331 		goto out;
332 	}
333 	err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid,
334 			      MSG_DONTWAIT);
335 	if (err > 0)
336 		err = 0;
337 
338 out:
339 	if (sk) {
340 		if (sk->sk_state == TCP_TIME_WAIT)
341 			inet_twsk_put((struct inet_timewait_sock *)sk);
342 		else
343 			sock_put(sk);
344 	}
345 out_nosk:
346 	return err;
347 }
348 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
349 
350 static int inet_diag_get_exact(struct sk_buff *in_skb,
351 			       const struct nlmsghdr *nlh,
352 			       struct inet_diag_req_v2 *req)
353 {
354 	const struct inet_diag_handler *handler;
355 	int err;
356 
357 	handler = inet_diag_lock_handler(req->sdiag_protocol);
358 	if (IS_ERR(handler))
359 		err = PTR_ERR(handler);
360 	else
361 		err = handler->dump_one(in_skb, nlh, req);
362 	inet_diag_unlock_handler(handler);
363 
364 	return err;
365 }
366 
367 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
368 {
369 	int words = bits >> 5;
370 
371 	bits &= 0x1f;
372 
373 	if (words) {
374 		if (memcmp(a1, a2, words << 2))
375 			return 0;
376 	}
377 	if (bits) {
378 		__be32 w1, w2;
379 		__be32 mask;
380 
381 		w1 = a1[words];
382 		w2 = a2[words];
383 
384 		mask = htonl((0xffffffff) << (32 - bits));
385 
386 		if ((w1 ^ w2) & mask)
387 			return 0;
388 	}
389 
390 	return 1;
391 }
392 
393 
394 static int inet_diag_bc_run(const struct nlattr *_bc,
395 		const struct inet_diag_entry *entry)
396 {
397 	const void *bc = nla_data(_bc);
398 	int len = nla_len(_bc);
399 
400 	while (len > 0) {
401 		int yes = 1;
402 		const struct inet_diag_bc_op *op = bc;
403 
404 		switch (op->code) {
405 		case INET_DIAG_BC_NOP:
406 			break;
407 		case INET_DIAG_BC_JMP:
408 			yes = 0;
409 			break;
410 		case INET_DIAG_BC_S_GE:
411 			yes = entry->sport >= op[1].no;
412 			break;
413 		case INET_DIAG_BC_S_LE:
414 			yes = entry->sport <= op[1].no;
415 			break;
416 		case INET_DIAG_BC_D_GE:
417 			yes = entry->dport >= op[1].no;
418 			break;
419 		case INET_DIAG_BC_D_LE:
420 			yes = entry->dport <= op[1].no;
421 			break;
422 		case INET_DIAG_BC_AUTO:
423 			yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
424 			break;
425 		case INET_DIAG_BC_S_COND:
426 		case INET_DIAG_BC_D_COND: {
427 			struct inet_diag_hostcond *cond;
428 			__be32 *addr;
429 
430 			cond = (struct inet_diag_hostcond *)(op + 1);
431 			if (cond->port != -1 &&
432 			    cond->port != (op->code == INET_DIAG_BC_S_COND ?
433 					     entry->sport : entry->dport)) {
434 				yes = 0;
435 				break;
436 			}
437 
438 			if (op->code == INET_DIAG_BC_S_COND)
439 				addr = entry->saddr;
440 			else
441 				addr = entry->daddr;
442 
443 			if (cond->family != AF_UNSPEC &&
444 			    cond->family != entry->family) {
445 				if (entry->family == AF_INET6 &&
446 				    cond->family == AF_INET) {
447 					if (addr[0] == 0 && addr[1] == 0 &&
448 					    addr[2] == htonl(0xffff) &&
449 					    bitstring_match(addr + 3,
450 							    cond->addr,
451 							    cond->prefix_len))
452 						break;
453 				}
454 				yes = 0;
455 				break;
456 			}
457 
458 			if (cond->prefix_len == 0)
459 				break;
460 			if (bitstring_match(addr, cond->addr,
461 					    cond->prefix_len))
462 				break;
463 			yes = 0;
464 			break;
465 		}
466 		}
467 
468 		if (yes) {
469 			len -= op->yes;
470 			bc += op->yes;
471 		} else {
472 			len -= op->no;
473 			bc += op->no;
474 		}
475 	}
476 	return len == 0;
477 }
478 
479 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
480 {
481 	struct inet_diag_entry entry;
482 	struct inet_sock *inet = inet_sk(sk);
483 
484 	if (bc == NULL)
485 		return 1;
486 
487 	entry.family = sk->sk_family;
488 #if IS_ENABLED(CONFIG_IPV6)
489 	if (entry.family == AF_INET6) {
490 		struct ipv6_pinfo *np = inet6_sk(sk);
491 
492 		entry.saddr = np->rcv_saddr.s6_addr32;
493 		entry.daddr = np->daddr.s6_addr32;
494 	} else
495 #endif
496 	{
497 		entry.saddr = &inet->inet_rcv_saddr;
498 		entry.daddr = &inet->inet_daddr;
499 	}
500 	entry.sport = inet->inet_num;
501 	entry.dport = ntohs(inet->inet_dport);
502 	entry.userlocks = sk->sk_userlocks;
503 
504 	return inet_diag_bc_run(bc, &entry);
505 }
506 EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
507 
508 static int valid_cc(const void *bc, int len, int cc)
509 {
510 	while (len >= 0) {
511 		const struct inet_diag_bc_op *op = bc;
512 
513 		if (cc > len)
514 			return 0;
515 		if (cc == len)
516 			return 1;
517 		if (op->yes < 4 || op->yes & 3)
518 			return 0;
519 		len -= op->yes;
520 		bc  += op->yes;
521 	}
522 	return 0;
523 }
524 
525 /* Validate an inet_diag_hostcond. */
526 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
527 			   int *min_len)
528 {
529 	int addr_len;
530 	struct inet_diag_hostcond *cond;
531 
532 	/* Check hostcond space. */
533 	*min_len += sizeof(struct inet_diag_hostcond);
534 	if (len < *min_len)
535 		return false;
536 	cond = (struct inet_diag_hostcond *)(op + 1);
537 
538 	/* Check address family and address length. */
539 	switch (cond->family) {
540 	case AF_UNSPEC:
541 		addr_len = 0;
542 		break;
543 	case AF_INET:
544 		addr_len = sizeof(struct in_addr);
545 		break;
546 	case AF_INET6:
547 		addr_len = sizeof(struct in6_addr);
548 		break;
549 	default:
550 		return false;
551 	}
552 	*min_len += addr_len;
553 	if (len < *min_len)
554 		return false;
555 
556 	/* Check prefix length (in bits) vs address length (in bytes). */
557 	if (cond->prefix_len > 8 * addr_len)
558 		return false;
559 
560 	return true;
561 }
562 
563 /* Validate a port comparison operator. */
564 static inline bool valid_port_comparison(const struct inet_diag_bc_op *op,
565 					 int len, int *min_len)
566 {
567 	/* Port comparisons put the port in a follow-on inet_diag_bc_op. */
568 	*min_len += sizeof(struct inet_diag_bc_op);
569 	if (len < *min_len)
570 		return false;
571 	return true;
572 }
573 
574 static int inet_diag_bc_audit(const void *bytecode, int bytecode_len)
575 {
576 	const void *bc = bytecode;
577 	int  len = bytecode_len;
578 
579 	while (len > 0) {
580 		const struct inet_diag_bc_op *op = bc;
581 		int min_len = sizeof(struct inet_diag_bc_op);
582 
583 //printk("BC: %d %d %d {%d} / %d\n", op->code, op->yes, op->no, op[1].no, len);
584 		switch (op->code) {
585 		case INET_DIAG_BC_S_COND:
586 		case INET_DIAG_BC_D_COND:
587 			if (!valid_hostcond(bc, len, &min_len))
588 				return -EINVAL;
589 			break;
590 		case INET_DIAG_BC_S_GE:
591 		case INET_DIAG_BC_S_LE:
592 		case INET_DIAG_BC_D_GE:
593 		case INET_DIAG_BC_D_LE:
594 			if (!valid_port_comparison(bc, len, &min_len))
595 				return -EINVAL;
596 			break;
597 		case INET_DIAG_BC_AUTO:
598 		case INET_DIAG_BC_JMP:
599 		case INET_DIAG_BC_NOP:
600 			break;
601 		default:
602 			return -EINVAL;
603 		}
604 
605 		if (op->code != INET_DIAG_BC_NOP) {
606 			if (op->no < min_len || op->no > len + 4 || op->no & 3)
607 				return -EINVAL;
608 			if (op->no < len &&
609 			    !valid_cc(bytecode, bytecode_len, len - op->no))
610 				return -EINVAL;
611 		}
612 
613 		if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
614 			return -EINVAL;
615 		bc  += op->yes;
616 		len -= op->yes;
617 	}
618 	return len == 0 ? 0 : -EINVAL;
619 }
620 
621 static int inet_csk_diag_dump(struct sock *sk,
622 			      struct sk_buff *skb,
623 			      struct netlink_callback *cb,
624 			      struct inet_diag_req_v2 *r,
625 			      const struct nlattr *bc)
626 {
627 	if (!inet_diag_bc_sk(bc, sk))
628 		return 0;
629 
630 	return inet_csk_diag_fill(sk, skb, r,
631 				  sk_user_ns(NETLINK_CB(cb->skb).ssk),
632 				  NETLINK_CB(cb->skb).portid,
633 				  cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
634 }
635 
636 static int inet_twsk_diag_dump(struct inet_timewait_sock *tw,
637 			       struct sk_buff *skb,
638 			       struct netlink_callback *cb,
639 			       struct inet_diag_req_v2 *r,
640 			       const struct nlattr *bc)
641 {
642 	if (bc != NULL) {
643 		struct inet_diag_entry entry;
644 
645 		entry.family = tw->tw_family;
646 #if IS_ENABLED(CONFIG_IPV6)
647 		if (tw->tw_family == AF_INET6) {
648 			struct inet6_timewait_sock *tw6 =
649 						inet6_twsk((struct sock *)tw);
650 			entry.saddr = tw6->tw_v6_rcv_saddr.s6_addr32;
651 			entry.daddr = tw6->tw_v6_daddr.s6_addr32;
652 		} else
653 #endif
654 		{
655 			entry.saddr = &tw->tw_rcv_saddr;
656 			entry.daddr = &tw->tw_daddr;
657 		}
658 		entry.sport = tw->tw_num;
659 		entry.dport = ntohs(tw->tw_dport);
660 		entry.userlocks = 0;
661 
662 		if (!inet_diag_bc_run(bc, &entry))
663 			return 0;
664 	}
665 
666 	return inet_twsk_diag_fill(tw, skb, r,
667 				   NETLINK_CB(cb->skb).portid,
668 				   cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
669 }
670 
671 /* Get the IPv4, IPv6, or IPv4-mapped-IPv6 local and remote addresses
672  * from a request_sock. For IPv4-mapped-IPv6 we must map IPv4 to IPv6.
673  */
674 static inline void inet_diag_req_addrs(const struct sock *sk,
675 				       const struct request_sock *req,
676 				       struct inet_diag_entry *entry)
677 {
678 	struct inet_request_sock *ireq = inet_rsk(req);
679 
680 #if IS_ENABLED(CONFIG_IPV6)
681 	if (sk->sk_family == AF_INET6) {
682 		if (req->rsk_ops->family == AF_INET6) {
683 			entry->saddr = inet6_rsk(req)->loc_addr.s6_addr32;
684 			entry->daddr = inet6_rsk(req)->rmt_addr.s6_addr32;
685 		} else if (req->rsk_ops->family == AF_INET) {
686 			ipv6_addr_set_v4mapped(ireq->loc_addr,
687 					       &entry->saddr_storage);
688 			ipv6_addr_set_v4mapped(ireq->rmt_addr,
689 					       &entry->daddr_storage);
690 			entry->saddr = entry->saddr_storage.s6_addr32;
691 			entry->daddr = entry->daddr_storage.s6_addr32;
692 		}
693 	} else
694 #endif
695 	{
696 		entry->saddr = &ireq->loc_addr;
697 		entry->daddr = &ireq->rmt_addr;
698 	}
699 }
700 
701 static int inet_diag_fill_req(struct sk_buff *skb, struct sock *sk,
702 			      struct request_sock *req,
703 			      struct user_namespace *user_ns,
704 			      u32 portid, u32 seq,
705 			      const struct nlmsghdr *unlh)
706 {
707 	const struct inet_request_sock *ireq = inet_rsk(req);
708 	struct inet_sock *inet = inet_sk(sk);
709 	struct inet_diag_msg *r;
710 	struct nlmsghdr *nlh;
711 	long tmo;
712 
713 	nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
714 			NLM_F_MULTI);
715 	if (!nlh)
716 		return -EMSGSIZE;
717 
718 	r = nlmsg_data(nlh);
719 	r->idiag_family = sk->sk_family;
720 	r->idiag_state = TCP_SYN_RECV;
721 	r->idiag_timer = 1;
722 	r->idiag_retrans = req->num_retrans;
723 
724 	r->id.idiag_if = sk->sk_bound_dev_if;
725 	sock_diag_save_cookie(req, r->id.idiag_cookie);
726 
727 	tmo = req->expires - jiffies;
728 	if (tmo < 0)
729 		tmo = 0;
730 
731 	r->id.idiag_sport = inet->inet_sport;
732 	r->id.idiag_dport = ireq->rmt_port;
733 	r->id.idiag_src[0] = ireq->loc_addr;
734 	r->id.idiag_dst[0] = ireq->rmt_addr;
735 	r->idiag_expires = jiffies_to_msecs(tmo);
736 	r->idiag_rqueue = 0;
737 	r->idiag_wqueue = 0;
738 	r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk));
739 	r->idiag_inode = 0;
740 #if IS_ENABLED(CONFIG_IPV6)
741 	if (r->idiag_family == AF_INET6) {
742 		struct inet_diag_entry entry;
743 		inet_diag_req_addrs(sk, req, &entry);
744 		memcpy(r->id.idiag_src, entry.saddr, sizeof(struct in6_addr));
745 		memcpy(r->id.idiag_dst, entry.daddr, sizeof(struct in6_addr));
746 	}
747 #endif
748 
749 	return nlmsg_end(skb, nlh);
750 }
751 
752 static int inet_diag_dump_reqs(struct sk_buff *skb, struct sock *sk,
753 			       struct netlink_callback *cb,
754 			       struct inet_diag_req_v2 *r,
755 			       const struct nlattr *bc)
756 {
757 	struct inet_diag_entry entry;
758 	struct inet_connection_sock *icsk = inet_csk(sk);
759 	struct listen_sock *lopt;
760 	struct inet_sock *inet = inet_sk(sk);
761 	int j, s_j;
762 	int reqnum, s_reqnum;
763 	int err = 0;
764 
765 	s_j = cb->args[3];
766 	s_reqnum = cb->args[4];
767 
768 	if (s_j > 0)
769 		s_j--;
770 
771 	entry.family = sk->sk_family;
772 
773 	read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
774 
775 	lopt = icsk->icsk_accept_queue.listen_opt;
776 	if (!lopt || !lopt->qlen)
777 		goto out;
778 
779 	if (bc != NULL) {
780 		entry.sport = inet->inet_num;
781 		entry.userlocks = sk->sk_userlocks;
782 	}
783 
784 	for (j = s_j; j < lopt->nr_table_entries; j++) {
785 		struct request_sock *req, *head = lopt->syn_table[j];
786 
787 		reqnum = 0;
788 		for (req = head; req; reqnum++, req = req->dl_next) {
789 			struct inet_request_sock *ireq = inet_rsk(req);
790 
791 			if (reqnum < s_reqnum)
792 				continue;
793 			if (r->id.idiag_dport != ireq->rmt_port &&
794 			    r->id.idiag_dport)
795 				continue;
796 
797 			if (bc) {
798 				inet_diag_req_addrs(sk, req, &entry);
799 				entry.dport = ntohs(ireq->rmt_port);
800 
801 				if (!inet_diag_bc_run(bc, &entry))
802 					continue;
803 			}
804 
805 			err = inet_diag_fill_req(skb, sk, req,
806 					       sk_user_ns(NETLINK_CB(cb->skb).ssk),
807 					       NETLINK_CB(cb->skb).portid,
808 					       cb->nlh->nlmsg_seq, cb->nlh);
809 			if (err < 0) {
810 				cb->args[3] = j + 1;
811 				cb->args[4] = reqnum;
812 				goto out;
813 			}
814 		}
815 
816 		s_reqnum = 0;
817 	}
818 
819 out:
820 	read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
821 
822 	return err;
823 }
824 
825 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
826 		struct netlink_callback *cb, struct inet_diag_req_v2 *r, struct nlattr *bc)
827 {
828 	int i, num;
829 	int s_i, s_num;
830 	struct net *net = sock_net(skb->sk);
831 
832 	s_i = cb->args[1];
833 	s_num = num = cb->args[2];
834 
835 	if (cb->args[0] == 0) {
836 		if (!(r->idiag_states & (TCPF_LISTEN | TCPF_SYN_RECV)))
837 			goto skip_listen_ht;
838 
839 		for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
840 			struct sock *sk;
841 			struct hlist_nulls_node *node;
842 			struct inet_listen_hashbucket *ilb;
843 
844 			num = 0;
845 			ilb = &hashinfo->listening_hash[i];
846 			spin_lock_bh(&ilb->lock);
847 			sk_nulls_for_each(sk, node, &ilb->head) {
848 				struct inet_sock *inet = inet_sk(sk);
849 
850 				if (!net_eq(sock_net(sk), net))
851 					continue;
852 
853 				if (num < s_num) {
854 					num++;
855 					continue;
856 				}
857 
858 				if (r->sdiag_family != AF_UNSPEC &&
859 						sk->sk_family != r->sdiag_family)
860 					goto next_listen;
861 
862 				if (r->id.idiag_sport != inet->inet_sport &&
863 				    r->id.idiag_sport)
864 					goto next_listen;
865 
866 				if (!(r->idiag_states & TCPF_LISTEN) ||
867 				    r->id.idiag_dport ||
868 				    cb->args[3] > 0)
869 					goto syn_recv;
870 
871 				if (inet_csk_diag_dump(sk, skb, cb, r, bc) < 0) {
872 					spin_unlock_bh(&ilb->lock);
873 					goto done;
874 				}
875 
876 syn_recv:
877 				if (!(r->idiag_states & TCPF_SYN_RECV))
878 					goto next_listen;
879 
880 				if (inet_diag_dump_reqs(skb, sk, cb, r, bc) < 0) {
881 					spin_unlock_bh(&ilb->lock);
882 					goto done;
883 				}
884 
885 next_listen:
886 				cb->args[3] = 0;
887 				cb->args[4] = 0;
888 				++num;
889 			}
890 			spin_unlock_bh(&ilb->lock);
891 
892 			s_num = 0;
893 			cb->args[3] = 0;
894 			cb->args[4] = 0;
895 		}
896 skip_listen_ht:
897 		cb->args[0] = 1;
898 		s_i = num = s_num = 0;
899 	}
900 
901 	if (!(r->idiag_states & ~(TCPF_LISTEN | TCPF_SYN_RECV)))
902 		goto out;
903 
904 	for (i = s_i; i <= hashinfo->ehash_mask; i++) {
905 		struct inet_ehash_bucket *head = &hashinfo->ehash[i];
906 		spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
907 		struct sock *sk;
908 		struct hlist_nulls_node *node;
909 
910 		num = 0;
911 
912 		if (hlist_nulls_empty(&head->chain) &&
913 			hlist_nulls_empty(&head->twchain))
914 			continue;
915 
916 		if (i > s_i)
917 			s_num = 0;
918 
919 		spin_lock_bh(lock);
920 		sk_nulls_for_each(sk, node, &head->chain) {
921 			struct inet_sock *inet = inet_sk(sk);
922 
923 			if (!net_eq(sock_net(sk), net))
924 				continue;
925 			if (num < s_num)
926 				goto next_normal;
927 			if (!(r->idiag_states & (1 << sk->sk_state)))
928 				goto next_normal;
929 			if (r->sdiag_family != AF_UNSPEC &&
930 					sk->sk_family != r->sdiag_family)
931 				goto next_normal;
932 			if (r->id.idiag_sport != inet->inet_sport &&
933 			    r->id.idiag_sport)
934 				goto next_normal;
935 			if (r->id.idiag_dport != inet->inet_dport &&
936 			    r->id.idiag_dport)
937 				goto next_normal;
938 			if (inet_csk_diag_dump(sk, skb, cb, r, bc) < 0) {
939 				spin_unlock_bh(lock);
940 				goto done;
941 			}
942 next_normal:
943 			++num;
944 		}
945 
946 		if (r->idiag_states & TCPF_TIME_WAIT) {
947 			struct inet_timewait_sock *tw;
948 
949 			inet_twsk_for_each(tw, node,
950 				    &head->twchain) {
951 				if (!net_eq(twsk_net(tw), net))
952 					continue;
953 
954 				if (num < s_num)
955 					goto next_dying;
956 				if (r->sdiag_family != AF_UNSPEC &&
957 						tw->tw_family != r->sdiag_family)
958 					goto next_dying;
959 				if (r->id.idiag_sport != tw->tw_sport &&
960 				    r->id.idiag_sport)
961 					goto next_dying;
962 				if (r->id.idiag_dport != tw->tw_dport &&
963 				    r->id.idiag_dport)
964 					goto next_dying;
965 				if (inet_twsk_diag_dump(tw, skb, cb, r, bc) < 0) {
966 					spin_unlock_bh(lock);
967 					goto done;
968 				}
969 next_dying:
970 				++num;
971 			}
972 		}
973 		spin_unlock_bh(lock);
974 	}
975 
976 done:
977 	cb->args[1] = i;
978 	cb->args[2] = num;
979 out:
980 	;
981 }
982 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
983 
984 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
985 		struct inet_diag_req_v2 *r, struct nlattr *bc)
986 {
987 	const struct inet_diag_handler *handler;
988 	int err = 0;
989 
990 	handler = inet_diag_lock_handler(r->sdiag_protocol);
991 	if (!IS_ERR(handler))
992 		handler->dump(skb, cb, r, bc);
993 	else
994 		err = PTR_ERR(handler);
995 	inet_diag_unlock_handler(handler);
996 
997 	return err ? : skb->len;
998 }
999 
1000 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
1001 {
1002 	struct nlattr *bc = NULL;
1003 	int hdrlen = sizeof(struct inet_diag_req_v2);
1004 
1005 	if (nlmsg_attrlen(cb->nlh, hdrlen))
1006 		bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE);
1007 
1008 	return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh), bc);
1009 }
1010 
1011 static inline int inet_diag_type2proto(int type)
1012 {
1013 	switch (type) {
1014 	case TCPDIAG_GETSOCK:
1015 		return IPPROTO_TCP;
1016 	case DCCPDIAG_GETSOCK:
1017 		return IPPROTO_DCCP;
1018 	default:
1019 		return 0;
1020 	}
1021 }
1022 
1023 static int inet_diag_dump_compat(struct sk_buff *skb, struct netlink_callback *cb)
1024 {
1025 	struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1026 	struct inet_diag_req_v2 req;
1027 	struct nlattr *bc = NULL;
1028 	int hdrlen = sizeof(struct inet_diag_req);
1029 
1030 	req.sdiag_family = AF_UNSPEC; /* compatibility */
1031 	req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1032 	req.idiag_ext = rc->idiag_ext;
1033 	req.idiag_states = rc->idiag_states;
1034 	req.id = rc->id;
1035 
1036 	if (nlmsg_attrlen(cb->nlh, hdrlen))
1037 		bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE);
1038 
1039 	return __inet_diag_dump(skb, cb, &req, bc);
1040 }
1041 
1042 static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1043 			       const struct nlmsghdr *nlh)
1044 {
1045 	struct inet_diag_req *rc = nlmsg_data(nlh);
1046 	struct inet_diag_req_v2 req;
1047 
1048 	req.sdiag_family = rc->idiag_family;
1049 	req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1050 	req.idiag_ext = rc->idiag_ext;
1051 	req.idiag_states = rc->idiag_states;
1052 	req.id = rc->id;
1053 
1054 	return inet_diag_get_exact(in_skb, nlh, &req);
1055 }
1056 
1057 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1058 {
1059 	int hdrlen = sizeof(struct inet_diag_req);
1060 	struct net *net = sock_net(skb->sk);
1061 
1062 	if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1063 	    nlmsg_len(nlh) < hdrlen)
1064 		return -EINVAL;
1065 
1066 	if (nlh->nlmsg_flags & NLM_F_DUMP) {
1067 		if (nlmsg_attrlen(nlh, hdrlen)) {
1068 			struct nlattr *attr;
1069 
1070 			attr = nlmsg_find_attr(nlh, hdrlen,
1071 					       INET_DIAG_REQ_BYTECODE);
1072 			if (attr == NULL ||
1073 			    nla_len(attr) < sizeof(struct inet_diag_bc_op) ||
1074 			    inet_diag_bc_audit(nla_data(attr), nla_len(attr)))
1075 				return -EINVAL;
1076 		}
1077 		{
1078 			struct netlink_dump_control c = {
1079 				.dump = inet_diag_dump_compat,
1080 			};
1081 			return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1082 		}
1083 	}
1084 
1085 	return inet_diag_get_exact_compat(skb, nlh);
1086 }
1087 
1088 static int inet_diag_handler_dump(struct sk_buff *skb, struct nlmsghdr *h)
1089 {
1090 	int hdrlen = sizeof(struct inet_diag_req_v2);
1091 	struct net *net = sock_net(skb->sk);
1092 
1093 	if (nlmsg_len(h) < hdrlen)
1094 		return -EINVAL;
1095 
1096 	if (h->nlmsg_flags & NLM_F_DUMP) {
1097 		if (nlmsg_attrlen(h, hdrlen)) {
1098 			struct nlattr *attr;
1099 			attr = nlmsg_find_attr(h, hdrlen,
1100 					       INET_DIAG_REQ_BYTECODE);
1101 			if (attr == NULL ||
1102 			    nla_len(attr) < sizeof(struct inet_diag_bc_op) ||
1103 			    inet_diag_bc_audit(nla_data(attr), nla_len(attr)))
1104 				return -EINVAL;
1105 		}
1106 		{
1107 			struct netlink_dump_control c = {
1108 				.dump = inet_diag_dump,
1109 			};
1110 			return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1111 		}
1112 	}
1113 
1114 	return inet_diag_get_exact(skb, h, nlmsg_data(h));
1115 }
1116 
1117 static const struct sock_diag_handler inet_diag_handler = {
1118 	.family = AF_INET,
1119 	.dump = inet_diag_handler_dump,
1120 };
1121 
1122 static const struct sock_diag_handler inet6_diag_handler = {
1123 	.family = AF_INET6,
1124 	.dump = inet_diag_handler_dump,
1125 };
1126 
1127 int inet_diag_register(const struct inet_diag_handler *h)
1128 {
1129 	const __u16 type = h->idiag_type;
1130 	int err = -EINVAL;
1131 
1132 	if (type >= IPPROTO_MAX)
1133 		goto out;
1134 
1135 	mutex_lock(&inet_diag_table_mutex);
1136 	err = -EEXIST;
1137 	if (inet_diag_table[type] == NULL) {
1138 		inet_diag_table[type] = h;
1139 		err = 0;
1140 	}
1141 	mutex_unlock(&inet_diag_table_mutex);
1142 out:
1143 	return err;
1144 }
1145 EXPORT_SYMBOL_GPL(inet_diag_register);
1146 
1147 void inet_diag_unregister(const struct inet_diag_handler *h)
1148 {
1149 	const __u16 type = h->idiag_type;
1150 
1151 	if (type >= IPPROTO_MAX)
1152 		return;
1153 
1154 	mutex_lock(&inet_diag_table_mutex);
1155 	inet_diag_table[type] = NULL;
1156 	mutex_unlock(&inet_diag_table_mutex);
1157 }
1158 EXPORT_SYMBOL_GPL(inet_diag_unregister);
1159 
1160 static int __init inet_diag_init(void)
1161 {
1162 	const int inet_diag_table_size = (IPPROTO_MAX *
1163 					  sizeof(struct inet_diag_handler *));
1164 	int err = -ENOMEM;
1165 
1166 	inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1167 	if (!inet_diag_table)
1168 		goto out;
1169 
1170 	err = sock_diag_register(&inet_diag_handler);
1171 	if (err)
1172 		goto out_free_nl;
1173 
1174 	err = sock_diag_register(&inet6_diag_handler);
1175 	if (err)
1176 		goto out_free_inet;
1177 
1178 	sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1179 out:
1180 	return err;
1181 
1182 out_free_inet:
1183 	sock_diag_unregister(&inet_diag_handler);
1184 out_free_nl:
1185 	kfree(inet_diag_table);
1186 	goto out;
1187 }
1188 
1189 static void __exit inet_diag_exit(void)
1190 {
1191 	sock_diag_unregister(&inet6_diag_handler);
1192 	sock_diag_unregister(&inet_diag_handler);
1193 	sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1194 	kfree(inet_diag_table);
1195 }
1196 
1197 module_init(inet_diag_init);
1198 module_exit(inet_diag_exit);
1199 MODULE_LICENSE("GPL");
1200 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1201 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);
1202