xref: /openbmc/linux/net/ipv4/inet_diag.c (revision 3ab0a7a0) 
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * inet_diag.c	Module for monitoring INET transport protocols sockets.
4  *
5  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
6  */
7 
8 #include <linux/kernel.h>
9 #include <linux/module.h>
10 #include <linux/types.h>
11 #include <linux/fcntl.h>
12 #include <linux/random.h>
13 #include <linux/slab.h>
14 #include <linux/cache.h>
15 #include <linux/init.h>
16 #include <linux/time.h>
17 
18 #include <net/icmp.h>
19 #include <net/tcp.h>
20 #include <net/ipv6.h>
21 #include <net/inet_common.h>
22 #include <net/inet_connection_sock.h>
23 #include <net/inet_hashtables.h>
24 #include <net/inet_timewait_sock.h>
25 #include <net/inet6_hashtables.h>
26 #include <net/bpf_sk_storage.h>
27 #include <net/netlink.h>
28 
29 #include <linux/inet.h>
30 #include <linux/stddef.h>
31 
32 #include <linux/inet_diag.h>
33 #include <linux/sock_diag.h>
34 
35 static const struct inet_diag_handler **inet_diag_table;
36 
37 struct inet_diag_entry {
38 	const __be32 *saddr;
39 	const __be32 *daddr;
40 	u16 sport;
41 	u16 dport;
42 	u16 family;
43 	u16 userlocks;
44 	u32 ifindex;
45 	u32 mark;
46 #ifdef CONFIG_SOCK_CGROUP_DATA
47 	u64 cgroup_id;
48 #endif
49 };
50 
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 (proto < 0 || proto >= IPPROTO_MAX) {
56 		mutex_lock(&inet_diag_table_mutex);
57 		return ERR_PTR(-ENOENT);
58 	}
59 
60 	if (!inet_diag_table[proto])
61 		sock_load_diag_module(AF_INET, proto);
62 
63 	mutex_lock(&inet_diag_table_mutex);
64 	if (!inet_diag_table[proto])
65 		return ERR_PTR(-ENOENT);
66 
67 	return inet_diag_table[proto];
68 }
69 
70 static void inet_diag_unlock_handler(const struct inet_diag_handler *handler)
71 {
72 	mutex_unlock(&inet_diag_table_mutex);
73 }
74 
75 void inet_diag_msg_common_fill(struct inet_diag_msg *r, struct sock *sk)
76 {
77 	r->idiag_family = sk->sk_family;
78 
79 	r->id.idiag_sport = htons(sk->sk_num);
80 	r->id.idiag_dport = sk->sk_dport;
81 	r->id.idiag_if = sk->sk_bound_dev_if;
82 	sock_diag_save_cookie(sk, r->id.idiag_cookie);
83 
84 #if IS_ENABLED(CONFIG_IPV6)
85 	if (sk->sk_family == AF_INET6) {
86 		*(struct in6_addr *)r->id.idiag_src = sk->sk_v6_rcv_saddr;
87 		*(struct in6_addr *)r->id.idiag_dst = sk->sk_v6_daddr;
88 	} else
89 #endif
90 	{
91 	memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src));
92 	memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst));
93 
94 	r->id.idiag_src[0] = sk->sk_rcv_saddr;
95 	r->id.idiag_dst[0] = sk->sk_daddr;
96 	}
97 }
98 EXPORT_SYMBOL_GPL(inet_diag_msg_common_fill);
99 
100 static size_t inet_sk_attr_size(struct sock *sk,
101 				const struct inet_diag_req_v2 *req,
102 				bool net_admin)
103 {
104 	const struct inet_diag_handler *handler;
105 	size_t aux = 0;
106 
107 	handler = inet_diag_table[req->sdiag_protocol];
108 	if (handler && handler->idiag_get_aux_size)
109 		aux = handler->idiag_get_aux_size(sk, net_admin);
110 
111 	return	  nla_total_size(sizeof(struct tcp_info))
112 		+ nla_total_size(sizeof(struct inet_diag_msg))
113 		+ inet_diag_msg_attrs_size()
114 		+ nla_total_size(sizeof(struct inet_diag_meminfo))
115 		+ nla_total_size(SK_MEMINFO_VARS * sizeof(u32))
116 		+ nla_total_size(TCP_CA_NAME_MAX)
117 		+ nla_total_size(sizeof(struct tcpvegas_info))
118 		+ aux
119 		+ 64;
120 }
121 
122 int inet_diag_msg_attrs_fill(struct sock *sk, struct sk_buff *skb,
123 			     struct inet_diag_msg *r, int ext,
124 			     struct user_namespace *user_ns,
125 			     bool net_admin)
126 {
127 	const struct inet_sock *inet = inet_sk(sk);
128 	struct inet_diag_sockopt inet_sockopt;
129 
130 	if (nla_put_u8(skb, INET_DIAG_SHUTDOWN, sk->sk_shutdown))
131 		goto errout;
132 
133 	/* IPv6 dual-stack sockets use inet->tos for IPv4 connections,
134 	 * hence this needs to be included regardless of socket family.
135 	 */
136 	if (ext & (1 << (INET_DIAG_TOS - 1)))
137 		if (nla_put_u8(skb, INET_DIAG_TOS, inet->tos) < 0)
138 			goto errout;
139 
140 #if IS_ENABLED(CONFIG_IPV6)
141 	if (r->idiag_family == AF_INET6) {
142 		if (ext & (1 << (INET_DIAG_TCLASS - 1)))
143 			if (nla_put_u8(skb, INET_DIAG_TCLASS,
144 				       inet6_sk(sk)->tclass) < 0)
145 				goto errout;
146 
147 		if (((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
148 		    nla_put_u8(skb, INET_DIAG_SKV6ONLY, ipv6_only_sock(sk)))
149 			goto errout;
150 	}
151 #endif
152 
153 	if (net_admin && nla_put_u32(skb, INET_DIAG_MARK, sk->sk_mark))
154 		goto errout;
155 
156 	if (ext & (1 << (INET_DIAG_CLASS_ID - 1)) ||
157 	    ext & (1 << (INET_DIAG_TCLASS - 1))) {
158 		u32 classid = 0;
159 
160 #ifdef CONFIG_SOCK_CGROUP_DATA
161 		classid = sock_cgroup_classid(&sk->sk_cgrp_data);
162 #endif
163 		/* Fallback to socket priority if class id isn't set.
164 		 * Classful qdiscs use it as direct reference to class.
165 		 * For cgroup2 classid is always zero.
166 		 */
167 		if (!classid)
168 			classid = sk->sk_priority;
169 
170 		if (nla_put_u32(skb, INET_DIAG_CLASS_ID, classid))
171 			goto errout;
172 	}
173 
174 #ifdef CONFIG_SOCK_CGROUP_DATA
175 	if (nla_put_u64_64bit(skb, INET_DIAG_CGROUP_ID,
176 			      cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)),
177 			      INET_DIAG_PAD))
178 		goto errout;
179 #endif
180 
181 	r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk));
182 	r->idiag_inode = sock_i_ino(sk);
183 
184 	memset(&inet_sockopt, 0, sizeof(inet_sockopt));
185 	inet_sockopt.recverr	= inet->recverr;
186 	inet_sockopt.is_icsk	= inet->is_icsk;
187 	inet_sockopt.freebind	= inet->freebind;
188 	inet_sockopt.hdrincl	= inet->hdrincl;
189 	inet_sockopt.mc_loop	= inet->mc_loop;
190 	inet_sockopt.transparent = inet->transparent;
191 	inet_sockopt.mc_all	= inet->mc_all;
192 	inet_sockopt.nodefrag	= inet->nodefrag;
193 	inet_sockopt.bind_address_no_port = inet->bind_address_no_port;
194 	inet_sockopt.recverr_rfc4884 = inet->recverr_rfc4884;
195 	inet_sockopt.defer_connect = inet->defer_connect;
196 	if (nla_put(skb, INET_DIAG_SOCKOPT, sizeof(inet_sockopt),
197 		    &inet_sockopt))
198 		goto errout;
199 
200 	return 0;
201 errout:
202 	return 1;
203 }
204 EXPORT_SYMBOL_GPL(inet_diag_msg_attrs_fill);
205 
206 static int inet_diag_parse_attrs(const struct nlmsghdr *nlh, int hdrlen,
207 				 struct nlattr **req_nlas)
208 {
209 	struct nlattr *nla;
210 	int remaining;
211 
212 	nlmsg_for_each_attr(nla, nlh, hdrlen, remaining) {
213 		int type = nla_type(nla);
214 
215 		if (type == INET_DIAG_REQ_PROTOCOL && nla_len(nla) != sizeof(u32))
216 			return -EINVAL;
217 
218 		if (type < __INET_DIAG_REQ_MAX)
219 			req_nlas[type] = nla;
220 	}
221 	return 0;
222 }
223 
224 static int inet_diag_get_protocol(const struct inet_diag_req_v2 *req,
225 				  const struct inet_diag_dump_data *data)
226 {
227 	if (data->req_nlas[INET_DIAG_REQ_PROTOCOL])
228 		return nla_get_u32(data->req_nlas[INET_DIAG_REQ_PROTOCOL]);
229 	return req->sdiag_protocol;
230 }
231 
232 #define MAX_DUMP_ALLOC_SIZE (KMALLOC_MAX_SIZE - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
233 
234 int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk,
235 		      struct sk_buff *skb, struct netlink_callback *cb,
236 		      const struct inet_diag_req_v2 *req,
237 		      u16 nlmsg_flags, bool net_admin)
238 {
239 	const struct tcp_congestion_ops *ca_ops;
240 	const struct inet_diag_handler *handler;
241 	struct inet_diag_dump_data *cb_data;
242 	int ext = req->idiag_ext;
243 	struct inet_diag_msg *r;
244 	struct nlmsghdr  *nlh;
245 	struct nlattr *attr;
246 	void *info = NULL;
247 
248 	cb_data = cb->data;
249 	handler = inet_diag_table[inet_diag_get_protocol(req, cb_data)];
250 	BUG_ON(!handler);
251 
252 	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
253 			cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
254 	if (!nlh)
255 		return -EMSGSIZE;
256 
257 	r = nlmsg_data(nlh);
258 	BUG_ON(!sk_fullsock(sk));
259 
260 	inet_diag_msg_common_fill(r, sk);
261 	r->idiag_state = sk->sk_state;
262 	r->idiag_timer = 0;
263 	r->idiag_retrans = 0;
264 
265 	if (inet_diag_msg_attrs_fill(sk, skb, r, ext,
266 				     sk_user_ns(NETLINK_CB(cb->skb).sk),
267 				     net_admin))
268 		goto errout;
269 
270 	if (ext & (1 << (INET_DIAG_MEMINFO - 1))) {
271 		struct inet_diag_meminfo minfo = {
272 			.idiag_rmem = sk_rmem_alloc_get(sk),
273 			.idiag_wmem = READ_ONCE(sk->sk_wmem_queued),
274 			.idiag_fmem = sk->sk_forward_alloc,
275 			.idiag_tmem = sk_wmem_alloc_get(sk),
276 		};
277 
278 		if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0)
279 			goto errout;
280 	}
281 
282 	if (ext & (1 << (INET_DIAG_SKMEMINFO - 1)))
283 		if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO))
284 			goto errout;
285 
286 	/*
287 	 * RAW sockets might have user-defined protocols assigned,
288 	 * so report the one supplied on socket creation.
289 	 */
290 	if (sk->sk_type == SOCK_RAW) {
291 		if (nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))
292 			goto errout;
293 	}
294 
295 	if (!icsk) {
296 		handler->idiag_get_info(sk, r, NULL);
297 		goto out;
298 	}
299 
300 	if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
301 	    icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
302 	    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
303 		r->idiag_timer = 1;
304 		r->idiag_retrans = icsk->icsk_retransmits;
305 		r->idiag_expires =
306 			jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
307 	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
308 		r->idiag_timer = 4;
309 		r->idiag_retrans = icsk->icsk_probes_out;
310 		r->idiag_expires =
311 			jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
312 	} else if (timer_pending(&sk->sk_timer)) {
313 		r->idiag_timer = 2;
314 		r->idiag_retrans = icsk->icsk_probes_out;
315 		r->idiag_expires =
316 			jiffies_delta_to_msecs(sk->sk_timer.expires - jiffies);
317 	} else {
318 		r->idiag_timer = 0;
319 		r->idiag_expires = 0;
320 	}
321 
322 	if ((ext & (1 << (INET_DIAG_INFO - 1))) && handler->idiag_info_size) {
323 		attr = nla_reserve_64bit(skb, INET_DIAG_INFO,
324 					 handler->idiag_info_size,
325 					 INET_DIAG_PAD);
326 		if (!attr)
327 			goto errout;
328 
329 		info = nla_data(attr);
330 	}
331 
332 	if (ext & (1 << (INET_DIAG_CONG - 1))) {
333 		int err = 0;
334 
335 		rcu_read_lock();
336 		ca_ops = READ_ONCE(icsk->icsk_ca_ops);
337 		if (ca_ops)
338 			err = nla_put_string(skb, INET_DIAG_CONG, ca_ops->name);
339 		rcu_read_unlock();
340 		if (err < 0)
341 			goto errout;
342 	}
343 
344 	handler->idiag_get_info(sk, r, info);
345 
346 	if (ext & (1 << (INET_DIAG_INFO - 1)) && handler->idiag_get_aux)
347 		if (handler->idiag_get_aux(sk, net_admin, skb) < 0)
348 			goto errout;
349 
350 	if (sk->sk_state < TCP_TIME_WAIT) {
351 		union tcp_cc_info info;
352 		size_t sz = 0;
353 		int attr;
354 
355 		rcu_read_lock();
356 		ca_ops = READ_ONCE(icsk->icsk_ca_ops);
357 		if (ca_ops && ca_ops->get_info)
358 			sz = ca_ops->get_info(sk, ext, &attr, &info);
359 		rcu_read_unlock();
360 		if (sz && nla_put(skb, attr, sz, &info) < 0)
361 			goto errout;
362 	}
363 
364 	/* Keep it at the end for potential retry with a larger skb,
365 	 * or else do best-effort fitting, which is only done for the
366 	 * first_nlmsg.
367 	 */
368 	if (cb_data->bpf_stg_diag) {
369 		bool first_nlmsg = ((unsigned char *)nlh == skb->data);
370 		unsigned int prev_min_dump_alloc;
371 		unsigned int total_nla_size = 0;
372 		unsigned int msg_len;
373 		int err;
374 
375 		msg_len = skb_tail_pointer(skb) - (unsigned char *)nlh;
376 		err = bpf_sk_storage_diag_put(cb_data->bpf_stg_diag, sk, skb,
377 					      INET_DIAG_SK_BPF_STORAGES,
378 					      &total_nla_size);
379 
380 		if (!err)
381 			goto out;
382 
383 		total_nla_size += msg_len;
384 		prev_min_dump_alloc = cb->min_dump_alloc;
385 		if (total_nla_size > prev_min_dump_alloc)
386 			cb->min_dump_alloc = min_t(u32, total_nla_size,
387 						   MAX_DUMP_ALLOC_SIZE);
388 
389 		if (!first_nlmsg)
390 			goto errout;
391 
392 		if (cb->min_dump_alloc > prev_min_dump_alloc)
393 			/* Retry with pskb_expand_head() with
394 			 * __GFP_DIRECT_RECLAIM
395 			 */
396 			goto errout;
397 
398 		WARN_ON_ONCE(total_nla_size <= prev_min_dump_alloc);
399 
400 		/* Send what we have for this sk
401 		 * and move on to the next sk in the following
402 		 * dump()
403 		 */
404 	}
405 
406 out:
407 	nlmsg_end(skb, nlh);
408 	return 0;
409 
410 errout:
411 	nlmsg_cancel(skb, nlh);
412 	return -EMSGSIZE;
413 }
414 EXPORT_SYMBOL_GPL(inet_sk_diag_fill);
415 
416 static int inet_twsk_diag_fill(struct sock *sk,
417 			       struct sk_buff *skb,
418 			       struct netlink_callback *cb,
419 			       u16 nlmsg_flags)
420 {
421 	struct inet_timewait_sock *tw = inet_twsk(sk);
422 	struct inet_diag_msg *r;
423 	struct nlmsghdr *nlh;
424 	long tmo;
425 
426 	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
427 			cb->nlh->nlmsg_seq, cb->nlh->nlmsg_type,
428 			sizeof(*r), nlmsg_flags);
429 	if (!nlh)
430 		return -EMSGSIZE;
431 
432 	r = nlmsg_data(nlh);
433 	BUG_ON(tw->tw_state != TCP_TIME_WAIT);
434 
435 	inet_diag_msg_common_fill(r, sk);
436 	r->idiag_retrans      = 0;
437 
438 	r->idiag_state	      = tw->tw_substate;
439 	r->idiag_timer	      = 3;
440 	tmo = tw->tw_timer.expires - jiffies;
441 	r->idiag_expires      = jiffies_delta_to_msecs(tmo);
442 	r->idiag_rqueue	      = 0;
443 	r->idiag_wqueue	      = 0;
444 	r->idiag_uid	      = 0;
445 	r->idiag_inode	      = 0;
446 
447 	nlmsg_end(skb, nlh);
448 	return 0;
449 }
450 
451 static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb,
452 			      struct netlink_callback *cb,
453 			      u16 nlmsg_flags, bool net_admin)
454 {
455 	struct request_sock *reqsk = inet_reqsk(sk);
456 	struct inet_diag_msg *r;
457 	struct nlmsghdr *nlh;
458 	long tmo;
459 
460 	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
461 			cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
462 	if (!nlh)
463 		return -EMSGSIZE;
464 
465 	r = nlmsg_data(nlh);
466 	inet_diag_msg_common_fill(r, sk);
467 	r->idiag_state = TCP_SYN_RECV;
468 	r->idiag_timer = 1;
469 	r->idiag_retrans = reqsk->num_retrans;
470 
471 	BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) !=
472 		     offsetof(struct sock, sk_cookie));
473 
474 	tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies;
475 	r->idiag_expires = jiffies_delta_to_msecs(tmo);
476 	r->idiag_rqueue	= 0;
477 	r->idiag_wqueue	= 0;
478 	r->idiag_uid	= 0;
479 	r->idiag_inode	= 0;
480 
481 	if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
482 				     inet_rsk(reqsk)->ir_mark))
483 		return -EMSGSIZE;
484 
485 	nlmsg_end(skb, nlh);
486 	return 0;
487 }
488 
489 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
490 			struct netlink_callback *cb,
491 			const struct inet_diag_req_v2 *r,
492 			u16 nlmsg_flags, bool net_admin)
493 {
494 	if (sk->sk_state == TCP_TIME_WAIT)
495 		return inet_twsk_diag_fill(sk, skb, cb, nlmsg_flags);
496 
497 	if (sk->sk_state == TCP_NEW_SYN_RECV)
498 		return inet_req_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
499 
500 	return inet_sk_diag_fill(sk, inet_csk(sk), skb, cb, r, nlmsg_flags,
501 				 net_admin);
502 }
503 
504 struct sock *inet_diag_find_one_icsk(struct net *net,
505 				     struct inet_hashinfo *hashinfo,
506 				     const struct inet_diag_req_v2 *req)
507 {
508 	struct sock *sk;
509 
510 	rcu_read_lock();
511 	if (req->sdiag_family == AF_INET)
512 		sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0],
513 				 req->id.idiag_dport, req->id.idiag_src[0],
514 				 req->id.idiag_sport, req->id.idiag_if);
515 #if IS_ENABLED(CONFIG_IPV6)
516 	else if (req->sdiag_family == AF_INET6) {
517 		if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) &&
518 		    ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src))
519 			sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3],
520 					 req->id.idiag_dport, req->id.idiag_src[3],
521 					 req->id.idiag_sport, req->id.idiag_if);
522 		else
523 			sk = inet6_lookup(net, hashinfo, NULL, 0,
524 					  (struct in6_addr *)req->id.idiag_dst,
525 					  req->id.idiag_dport,
526 					  (struct in6_addr *)req->id.idiag_src,
527 					  req->id.idiag_sport,
528 					  req->id.idiag_if);
529 	}
530 #endif
531 	else {
532 		rcu_read_unlock();
533 		return ERR_PTR(-EINVAL);
534 	}
535 	rcu_read_unlock();
536 	if (!sk)
537 		return ERR_PTR(-ENOENT);
538 
539 	if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) {
540 		sock_gen_put(sk);
541 		return ERR_PTR(-ENOENT);
542 	}
543 
544 	return sk;
545 }
546 EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk);
547 
548 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo,
549 			    struct netlink_callback *cb,
550 			    const struct inet_diag_req_v2 *req)
551 {
552 	struct sk_buff *in_skb = cb->skb;
553 	bool net_admin = netlink_net_capable(in_skb, CAP_NET_ADMIN);
554 	struct net *net = sock_net(in_skb->sk);
555 	struct sk_buff *rep;
556 	struct sock *sk;
557 	int err;
558 
559 	sk = inet_diag_find_one_icsk(net, hashinfo, req);
560 	if (IS_ERR(sk))
561 		return PTR_ERR(sk);
562 
563 	rep = nlmsg_new(inet_sk_attr_size(sk, req, net_admin), GFP_KERNEL);
564 	if (!rep) {
565 		err = -ENOMEM;
566 		goto out;
567 	}
568 
569 	err = sk_diag_fill(sk, rep, cb, req, 0, net_admin);
570 	if (err < 0) {
571 		WARN_ON(err == -EMSGSIZE);
572 		nlmsg_free(rep);
573 		goto out;
574 	}
575 	err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid,
576 			      MSG_DONTWAIT);
577 	if (err > 0)
578 		err = 0;
579 
580 out:
581 	if (sk)
582 		sock_gen_put(sk);
583 
584 	return err;
585 }
586 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
587 
588 static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb,
589 			       const struct nlmsghdr *nlh,
590 			       int hdrlen,
591 			       const struct inet_diag_req_v2 *req)
592 {
593 	const struct inet_diag_handler *handler;
594 	struct inet_diag_dump_data dump_data;
595 	int err, protocol;
596 
597 	memset(&dump_data, 0, sizeof(dump_data));
598 	err = inet_diag_parse_attrs(nlh, hdrlen, dump_data.req_nlas);
599 	if (err)
600 		return err;
601 
602 	protocol = inet_diag_get_protocol(req, &dump_data);
603 
604 	handler = inet_diag_lock_handler(protocol);
605 	if (IS_ERR(handler)) {
606 		err = PTR_ERR(handler);
607 	} else if (cmd == SOCK_DIAG_BY_FAMILY) {
608 		struct netlink_callback cb = {
609 			.nlh = nlh,
610 			.skb = in_skb,
611 			.data = &dump_data,
612 		};
613 		err = handler->dump_one(&cb, req);
614 	} else if (cmd == SOCK_DESTROY && handler->destroy) {
615 		err = handler->destroy(in_skb, req);
616 	} else {
617 		err = -EOPNOTSUPP;
618 	}
619 	inet_diag_unlock_handler(handler);
620 
621 	return err;
622 }
623 
624 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
625 {
626 	int words = bits >> 5;
627 
628 	bits &= 0x1f;
629 
630 	if (words) {
631 		if (memcmp(a1, a2, words << 2))
632 			return 0;
633 	}
634 	if (bits) {
635 		__be32 w1, w2;
636 		__be32 mask;
637 
638 		w1 = a1[words];
639 		w2 = a2[words];
640 
641 		mask = htonl((0xffffffff) << (32 - bits));
642 
643 		if ((w1 ^ w2) & mask)
644 			return 0;
645 	}
646 
647 	return 1;
648 }
649 
650 static int inet_diag_bc_run(const struct nlattr *_bc,
651 			    const struct inet_diag_entry *entry)
652 {
653 	const void *bc = nla_data(_bc);
654 	int len = nla_len(_bc);
655 
656 	while (len > 0) {
657 		int yes = 1;
658 		const struct inet_diag_bc_op *op = bc;
659 
660 		switch (op->code) {
661 		case INET_DIAG_BC_NOP:
662 			break;
663 		case INET_DIAG_BC_JMP:
664 			yes = 0;
665 			break;
666 		case INET_DIAG_BC_S_EQ:
667 			yes = entry->sport == op[1].no;
668 			break;
669 		case INET_DIAG_BC_S_GE:
670 			yes = entry->sport >= op[1].no;
671 			break;
672 		case INET_DIAG_BC_S_LE:
673 			yes = entry->sport <= op[1].no;
674 			break;
675 		case INET_DIAG_BC_D_EQ:
676 			yes = entry->dport == op[1].no;
677 			break;
678 		case INET_DIAG_BC_D_GE:
679 			yes = entry->dport >= op[1].no;
680 			break;
681 		case INET_DIAG_BC_D_LE:
682 			yes = entry->dport <= op[1].no;
683 			break;
684 		case INET_DIAG_BC_AUTO:
685 			yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
686 			break;
687 		case INET_DIAG_BC_S_COND:
688 		case INET_DIAG_BC_D_COND: {
689 			const struct inet_diag_hostcond *cond;
690 			const __be32 *addr;
691 
692 			cond = (const struct inet_diag_hostcond *)(op + 1);
693 			if (cond->port != -1 &&
694 			    cond->port != (op->code == INET_DIAG_BC_S_COND ?
695 					     entry->sport : entry->dport)) {
696 				yes = 0;
697 				break;
698 			}
699 
700 			if (op->code == INET_DIAG_BC_S_COND)
701 				addr = entry->saddr;
702 			else
703 				addr = entry->daddr;
704 
705 			if (cond->family != AF_UNSPEC &&
706 			    cond->family != entry->family) {
707 				if (entry->family == AF_INET6 &&
708 				    cond->family == AF_INET) {
709 					if (addr[0] == 0 && addr[1] == 0 &&
710 					    addr[2] == htonl(0xffff) &&
711 					    bitstring_match(addr + 3,
712 							    cond->addr,
713 							    cond->prefix_len))
714 						break;
715 				}
716 				yes = 0;
717 				break;
718 			}
719 
720 			if (cond->prefix_len == 0)
721 				break;
722 			if (bitstring_match(addr, cond->addr,
723 					    cond->prefix_len))
724 				break;
725 			yes = 0;
726 			break;
727 		}
728 		case INET_DIAG_BC_DEV_COND: {
729 			u32 ifindex;
730 
731 			ifindex = *((const u32 *)(op + 1));
732 			if (ifindex != entry->ifindex)
733 				yes = 0;
734 			break;
735 		}
736 		case INET_DIAG_BC_MARK_COND: {
737 			struct inet_diag_markcond *cond;
738 
739 			cond = (struct inet_diag_markcond *)(op + 1);
740 			if ((entry->mark & cond->mask) != cond->mark)
741 				yes = 0;
742 			break;
743 		}
744 #ifdef CONFIG_SOCK_CGROUP_DATA
745 		case INET_DIAG_BC_CGROUP_COND: {
746 			u64 cgroup_id;
747 
748 			cgroup_id = get_unaligned((const u64 *)(op + 1));
749 			if (cgroup_id != entry->cgroup_id)
750 				yes = 0;
751 			break;
752 		}
753 #endif
754 		}
755 
756 		if (yes) {
757 			len -= op->yes;
758 			bc += op->yes;
759 		} else {
760 			len -= op->no;
761 			bc += op->no;
762 		}
763 	}
764 	return len == 0;
765 }
766 
767 /* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV)
768  */
769 static void entry_fill_addrs(struct inet_diag_entry *entry,
770 			     const struct sock *sk)
771 {
772 #if IS_ENABLED(CONFIG_IPV6)
773 	if (sk->sk_family == AF_INET6) {
774 		entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32;
775 		entry->daddr = sk->sk_v6_daddr.s6_addr32;
776 	} else
777 #endif
778 	{
779 		entry->saddr = &sk->sk_rcv_saddr;
780 		entry->daddr = &sk->sk_daddr;
781 	}
782 }
783 
784 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
785 {
786 	struct inet_sock *inet = inet_sk(sk);
787 	struct inet_diag_entry entry;
788 
789 	if (!bc)
790 		return 1;
791 
792 	entry.family = sk->sk_family;
793 	entry_fill_addrs(&entry, sk);
794 	entry.sport = inet->inet_num;
795 	entry.dport = ntohs(inet->inet_dport);
796 	entry.ifindex = sk->sk_bound_dev_if;
797 	entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0;
798 	if (sk_fullsock(sk))
799 		entry.mark = sk->sk_mark;
800 	else if (sk->sk_state == TCP_NEW_SYN_RECV)
801 		entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark;
802 	else
803 		entry.mark = 0;
804 #ifdef CONFIG_SOCK_CGROUP_DATA
805 	entry.cgroup_id = sk_fullsock(sk) ?
806 		cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)) : 0;
807 #endif
808 
809 	return inet_diag_bc_run(bc, &entry);
810 }
811 EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
812 
813 static int valid_cc(const void *bc, int len, int cc)
814 {
815 	while (len >= 0) {
816 		const struct inet_diag_bc_op *op = bc;
817 
818 		if (cc > len)
819 			return 0;
820 		if (cc == len)
821 			return 1;
822 		if (op->yes < 4 || op->yes & 3)
823 			return 0;
824 		len -= op->yes;
825 		bc  += op->yes;
826 	}
827 	return 0;
828 }
829 
830 /* data is u32 ifindex */
831 static bool valid_devcond(const struct inet_diag_bc_op *op, int len,
832 			  int *min_len)
833 {
834 	/* Check ifindex space. */
835 	*min_len += sizeof(u32);
836 	if (len < *min_len)
837 		return false;
838 
839 	return true;
840 }
841 /* Validate an inet_diag_hostcond. */
842 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
843 			   int *min_len)
844 {
845 	struct inet_diag_hostcond *cond;
846 	int addr_len;
847 
848 	/* Check hostcond space. */
849 	*min_len += sizeof(struct inet_diag_hostcond);
850 	if (len < *min_len)
851 		return false;
852 	cond = (struct inet_diag_hostcond *)(op + 1);
853 
854 	/* Check address family and address length. */
855 	switch (cond->family) {
856 	case AF_UNSPEC:
857 		addr_len = 0;
858 		break;
859 	case AF_INET:
860 		addr_len = sizeof(struct in_addr);
861 		break;
862 	case AF_INET6:
863 		addr_len = sizeof(struct in6_addr);
864 		break;
865 	default:
866 		return false;
867 	}
868 	*min_len += addr_len;
869 	if (len < *min_len)
870 		return false;
871 
872 	/* Check prefix length (in bits) vs address length (in bytes). */
873 	if (cond->prefix_len > 8 * addr_len)
874 		return false;
875 
876 	return true;
877 }
878 
879 /* Validate a port comparison operator. */
880 static bool valid_port_comparison(const struct inet_diag_bc_op *op,
881 				  int len, int *min_len)
882 {
883 	/* Port comparisons put the port in a follow-on inet_diag_bc_op. */
884 	*min_len += sizeof(struct inet_diag_bc_op);
885 	if (len < *min_len)
886 		return false;
887 	return true;
888 }
889 
890 static bool valid_markcond(const struct inet_diag_bc_op *op, int len,
891 			   int *min_len)
892 {
893 	*min_len += sizeof(struct inet_diag_markcond);
894 	return len >= *min_len;
895 }
896 
897 #ifdef CONFIG_SOCK_CGROUP_DATA
898 static bool valid_cgroupcond(const struct inet_diag_bc_op *op, int len,
899 			     int *min_len)
900 {
901 	*min_len += sizeof(u64);
902 	return len >= *min_len;
903 }
904 #endif
905 
906 static int inet_diag_bc_audit(const struct nlattr *attr,
907 			      const struct sk_buff *skb)
908 {
909 	bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN);
910 	const void *bytecode, *bc;
911 	int bytecode_len, len;
912 
913 	if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op))
914 		return -EINVAL;
915 
916 	bytecode = bc = nla_data(attr);
917 	len = bytecode_len = nla_len(attr);
918 
919 	while (len > 0) {
920 		int min_len = sizeof(struct inet_diag_bc_op);
921 		const struct inet_diag_bc_op *op = bc;
922 
923 		switch (op->code) {
924 		case INET_DIAG_BC_S_COND:
925 		case INET_DIAG_BC_D_COND:
926 			if (!valid_hostcond(bc, len, &min_len))
927 				return -EINVAL;
928 			break;
929 		case INET_DIAG_BC_DEV_COND:
930 			if (!valid_devcond(bc, len, &min_len))
931 				return -EINVAL;
932 			break;
933 		case INET_DIAG_BC_S_EQ:
934 		case INET_DIAG_BC_S_GE:
935 		case INET_DIAG_BC_S_LE:
936 		case INET_DIAG_BC_D_EQ:
937 		case INET_DIAG_BC_D_GE:
938 		case INET_DIAG_BC_D_LE:
939 			if (!valid_port_comparison(bc, len, &min_len))
940 				return -EINVAL;
941 			break;
942 		case INET_DIAG_BC_MARK_COND:
943 			if (!net_admin)
944 				return -EPERM;
945 			if (!valid_markcond(bc, len, &min_len))
946 				return -EINVAL;
947 			break;
948 #ifdef CONFIG_SOCK_CGROUP_DATA
949 		case INET_DIAG_BC_CGROUP_COND:
950 			if (!valid_cgroupcond(bc, len, &min_len))
951 				return -EINVAL;
952 			break;
953 #endif
954 		case INET_DIAG_BC_AUTO:
955 		case INET_DIAG_BC_JMP:
956 		case INET_DIAG_BC_NOP:
957 			break;
958 		default:
959 			return -EINVAL;
960 		}
961 
962 		if (op->code != INET_DIAG_BC_NOP) {
963 			if (op->no < min_len || op->no > len + 4 || op->no & 3)
964 				return -EINVAL;
965 			if (op->no < len &&
966 			    !valid_cc(bytecode, bytecode_len, len - op->no))
967 				return -EINVAL;
968 		}
969 
970 		if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
971 			return -EINVAL;
972 		bc  += op->yes;
973 		len -= op->yes;
974 	}
975 	return len == 0 ? 0 : -EINVAL;
976 }
977 
978 static void twsk_build_assert(void)
979 {
980 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) !=
981 		     offsetof(struct sock, sk_family));
982 
983 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) !=
984 		     offsetof(struct inet_sock, inet_num));
985 
986 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) !=
987 		     offsetof(struct inet_sock, inet_dport));
988 
989 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) !=
990 		     offsetof(struct inet_sock, inet_rcv_saddr));
991 
992 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) !=
993 		     offsetof(struct inet_sock, inet_daddr));
994 
995 #if IS_ENABLED(CONFIG_IPV6)
996 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) !=
997 		     offsetof(struct sock, sk_v6_rcv_saddr));
998 
999 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) !=
1000 		     offsetof(struct sock, sk_v6_daddr));
1001 #endif
1002 }
1003 
1004 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
1005 			 struct netlink_callback *cb,
1006 			 const struct inet_diag_req_v2 *r)
1007 {
1008 	bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN);
1009 	struct inet_diag_dump_data *cb_data = cb->data;
1010 	struct net *net = sock_net(skb->sk);
1011 	u32 idiag_states = r->idiag_states;
1012 	int i, num, s_i, s_num;
1013 	struct nlattr *bc;
1014 	struct sock *sk;
1015 
1016 	bc = cb_data->inet_diag_nla_bc;
1017 	if (idiag_states & TCPF_SYN_RECV)
1018 		idiag_states |= TCPF_NEW_SYN_RECV;
1019 	s_i = cb->args[1];
1020 	s_num = num = cb->args[2];
1021 
1022 	if (cb->args[0] == 0) {
1023 		if (!(idiag_states & TCPF_LISTEN) || r->id.idiag_dport)
1024 			goto skip_listen_ht;
1025 
1026 		for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
1027 			struct inet_listen_hashbucket *ilb;
1028 			struct hlist_nulls_node *node;
1029 
1030 			num = 0;
1031 			ilb = &hashinfo->listening_hash[i];
1032 			spin_lock(&ilb->lock);
1033 			sk_nulls_for_each(sk, node, &ilb->nulls_head) {
1034 				struct inet_sock *inet = inet_sk(sk);
1035 
1036 				if (!net_eq(sock_net(sk), net))
1037 					continue;
1038 
1039 				if (num < s_num) {
1040 					num++;
1041 					continue;
1042 				}
1043 
1044 				if (r->sdiag_family != AF_UNSPEC &&
1045 				    sk->sk_family != r->sdiag_family)
1046 					goto next_listen;
1047 
1048 				if (r->id.idiag_sport != inet->inet_sport &&
1049 				    r->id.idiag_sport)
1050 					goto next_listen;
1051 
1052 				if (!inet_diag_bc_sk(bc, sk))
1053 					goto next_listen;
1054 
1055 				if (inet_sk_diag_fill(sk, inet_csk(sk), skb,
1056 						      cb, r, NLM_F_MULTI,
1057 						      net_admin) < 0) {
1058 					spin_unlock(&ilb->lock);
1059 					goto done;
1060 				}
1061 
1062 next_listen:
1063 				++num;
1064 			}
1065 			spin_unlock(&ilb->lock);
1066 
1067 			s_num = 0;
1068 		}
1069 skip_listen_ht:
1070 		cb->args[0] = 1;
1071 		s_i = num = s_num = 0;
1072 	}
1073 
1074 	if (!(idiag_states & ~TCPF_LISTEN))
1075 		goto out;
1076 
1077 #define SKARR_SZ 16
1078 	for (i = s_i; i <= hashinfo->ehash_mask; i++) {
1079 		struct inet_ehash_bucket *head = &hashinfo->ehash[i];
1080 		spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
1081 		struct hlist_nulls_node *node;
1082 		struct sock *sk_arr[SKARR_SZ];
1083 		int num_arr[SKARR_SZ];
1084 		int idx, accum, res;
1085 
1086 		if (hlist_nulls_empty(&head->chain))
1087 			continue;
1088 
1089 		if (i > s_i)
1090 			s_num = 0;
1091 
1092 next_chunk:
1093 		num = 0;
1094 		accum = 0;
1095 		spin_lock_bh(lock);
1096 		sk_nulls_for_each(sk, node, &head->chain) {
1097 			int state;
1098 
1099 			if (!net_eq(sock_net(sk), net))
1100 				continue;
1101 			if (num < s_num)
1102 				goto next_normal;
1103 			state = (sk->sk_state == TCP_TIME_WAIT) ?
1104 				inet_twsk(sk)->tw_substate : sk->sk_state;
1105 			if (!(idiag_states & (1 << state)))
1106 				goto next_normal;
1107 			if (r->sdiag_family != AF_UNSPEC &&
1108 			    sk->sk_family != r->sdiag_family)
1109 				goto next_normal;
1110 			if (r->id.idiag_sport != htons(sk->sk_num) &&
1111 			    r->id.idiag_sport)
1112 				goto next_normal;
1113 			if (r->id.idiag_dport != sk->sk_dport &&
1114 			    r->id.idiag_dport)
1115 				goto next_normal;
1116 			twsk_build_assert();
1117 
1118 			if (!inet_diag_bc_sk(bc, sk))
1119 				goto next_normal;
1120 
1121 			if (!refcount_inc_not_zero(&sk->sk_refcnt))
1122 				goto next_normal;
1123 
1124 			num_arr[accum] = num;
1125 			sk_arr[accum] = sk;
1126 			if (++accum == SKARR_SZ)
1127 				break;
1128 next_normal:
1129 			++num;
1130 		}
1131 		spin_unlock_bh(lock);
1132 		res = 0;
1133 		for (idx = 0; idx < accum; idx++) {
1134 			if (res >= 0) {
1135 				res = sk_diag_fill(sk_arr[idx], skb, cb, r,
1136 						   NLM_F_MULTI, net_admin);
1137 				if (res < 0)
1138 					num = num_arr[idx];
1139 			}
1140 			sock_gen_put(sk_arr[idx]);
1141 		}
1142 		if (res < 0)
1143 			break;
1144 		cond_resched();
1145 		if (accum == SKARR_SZ) {
1146 			s_num = num + 1;
1147 			goto next_chunk;
1148 		}
1149 	}
1150 
1151 done:
1152 	cb->args[1] = i;
1153 	cb->args[2] = num;
1154 out:
1155 	;
1156 }
1157 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
1158 
1159 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
1160 			    const struct inet_diag_req_v2 *r)
1161 {
1162 	struct inet_diag_dump_data *cb_data = cb->data;
1163 	const struct inet_diag_handler *handler;
1164 	u32 prev_min_dump_alloc;
1165 	int protocol, err = 0;
1166 
1167 	protocol = inet_diag_get_protocol(r, cb_data);
1168 
1169 again:
1170 	prev_min_dump_alloc = cb->min_dump_alloc;
1171 	handler = inet_diag_lock_handler(protocol);
1172 	if (!IS_ERR(handler))
1173 		handler->dump(skb, cb, r);
1174 	else
1175 		err = PTR_ERR(handler);
1176 	inet_diag_unlock_handler(handler);
1177 
1178 	/* The skb is not large enough to fit one sk info and
1179 	 * inet_sk_diag_fill() has requested for a larger skb.
1180 	 */
1181 	if (!skb->len && cb->min_dump_alloc > prev_min_dump_alloc) {
1182 		err = pskb_expand_head(skb, 0, cb->min_dump_alloc, GFP_KERNEL);
1183 		if (!err)
1184 			goto again;
1185 	}
1186 
1187 	return err ? : skb->len;
1188 }
1189 
1190 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
1191 {
1192 	return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh));
1193 }
1194 
1195 static int __inet_diag_dump_start(struct netlink_callback *cb, int hdrlen)
1196 {
1197 	const struct nlmsghdr *nlh = cb->nlh;
1198 	struct inet_diag_dump_data *cb_data;
1199 	struct sk_buff *skb = cb->skb;
1200 	struct nlattr *nla;
1201 	int err;
1202 
1203 	cb_data = kzalloc(sizeof(*cb_data), GFP_KERNEL);
1204 	if (!cb_data)
1205 		return -ENOMEM;
1206 
1207 	err = inet_diag_parse_attrs(nlh, hdrlen, cb_data->req_nlas);
1208 	if (err) {
1209 		kfree(cb_data);
1210 		return err;
1211 	}
1212 	nla = cb_data->inet_diag_nla_bc;
1213 	if (nla) {
1214 		err = inet_diag_bc_audit(nla, skb);
1215 		if (err) {
1216 			kfree(cb_data);
1217 			return err;
1218 		}
1219 	}
1220 
1221 	nla = cb_data->inet_diag_nla_bpf_stgs;
1222 	if (nla) {
1223 		struct bpf_sk_storage_diag *bpf_stg_diag;
1224 
1225 		bpf_stg_diag = bpf_sk_storage_diag_alloc(nla);
1226 		if (IS_ERR(bpf_stg_diag)) {
1227 			kfree(cb_data);
1228 			return PTR_ERR(bpf_stg_diag);
1229 		}
1230 		cb_data->bpf_stg_diag = bpf_stg_diag;
1231 	}
1232 
1233 	cb->data = cb_data;
1234 	return 0;
1235 }
1236 
1237 static int inet_diag_dump_start(struct netlink_callback *cb)
1238 {
1239 	return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req_v2));
1240 }
1241 
1242 static int inet_diag_dump_start_compat(struct netlink_callback *cb)
1243 {
1244 	return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req));
1245 }
1246 
1247 static int inet_diag_dump_done(struct netlink_callback *cb)
1248 {
1249 	struct inet_diag_dump_data *cb_data = cb->data;
1250 
1251 	bpf_sk_storage_diag_free(cb_data->bpf_stg_diag);
1252 	kfree(cb->data);
1253 
1254 	return 0;
1255 }
1256 
1257 static int inet_diag_type2proto(int type)
1258 {
1259 	switch (type) {
1260 	case TCPDIAG_GETSOCK:
1261 		return IPPROTO_TCP;
1262 	case DCCPDIAG_GETSOCK:
1263 		return IPPROTO_DCCP;
1264 	default:
1265 		return 0;
1266 	}
1267 }
1268 
1269 static int inet_diag_dump_compat(struct sk_buff *skb,
1270 				 struct netlink_callback *cb)
1271 {
1272 	struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1273 	struct inet_diag_req_v2 req;
1274 
1275 	req.sdiag_family = AF_UNSPEC; /* compatibility */
1276 	req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1277 	req.idiag_ext = rc->idiag_ext;
1278 	req.idiag_states = rc->idiag_states;
1279 	req.id = rc->id;
1280 
1281 	return __inet_diag_dump(skb, cb, &req);
1282 }
1283 
1284 static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1285 				      const struct nlmsghdr *nlh)
1286 {
1287 	struct inet_diag_req *rc = nlmsg_data(nlh);
1288 	struct inet_diag_req_v2 req;
1289 
1290 	req.sdiag_family = rc->idiag_family;
1291 	req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1292 	req.idiag_ext = rc->idiag_ext;
1293 	req.idiag_states = rc->idiag_states;
1294 	req.id = rc->id;
1295 
1296 	return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh,
1297 				   sizeof(struct inet_diag_req), &req);
1298 }
1299 
1300 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1301 {
1302 	int hdrlen = sizeof(struct inet_diag_req);
1303 	struct net *net = sock_net(skb->sk);
1304 
1305 	if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1306 	    nlmsg_len(nlh) < hdrlen)
1307 		return -EINVAL;
1308 
1309 	if (nlh->nlmsg_flags & NLM_F_DUMP) {
1310 		struct netlink_dump_control c = {
1311 			.start = inet_diag_dump_start_compat,
1312 			.done = inet_diag_dump_done,
1313 			.dump = inet_diag_dump_compat,
1314 		};
1315 		return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1316 	}
1317 
1318 	return inet_diag_get_exact_compat(skb, nlh);
1319 }
1320 
1321 static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h)
1322 {
1323 	int hdrlen = sizeof(struct inet_diag_req_v2);
1324 	struct net *net = sock_net(skb->sk);
1325 
1326 	if (nlmsg_len(h) < hdrlen)
1327 		return -EINVAL;
1328 
1329 	if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY &&
1330 	    h->nlmsg_flags & NLM_F_DUMP) {
1331 		struct netlink_dump_control c = {
1332 			.start = inet_diag_dump_start,
1333 			.done = inet_diag_dump_done,
1334 			.dump = inet_diag_dump,
1335 		};
1336 		return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1337 	}
1338 
1339 	return inet_diag_cmd_exact(h->nlmsg_type, skb, h, hdrlen,
1340 				   nlmsg_data(h));
1341 }
1342 
1343 static
1344 int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk)
1345 {
1346 	const struct inet_diag_handler *handler;
1347 	struct nlmsghdr *nlh;
1348 	struct nlattr *attr;
1349 	struct inet_diag_msg *r;
1350 	void *info = NULL;
1351 	int err = 0;
1352 
1353 	nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0);
1354 	if (!nlh)
1355 		return -ENOMEM;
1356 
1357 	r = nlmsg_data(nlh);
1358 	memset(r, 0, sizeof(*r));
1359 	inet_diag_msg_common_fill(r, sk);
1360 	if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM)
1361 		r->id.idiag_sport = inet_sk(sk)->inet_sport;
1362 	r->idiag_state = sk->sk_state;
1363 
1364 	if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) {
1365 		nlmsg_cancel(skb, nlh);
1366 		return err;
1367 	}
1368 
1369 	handler = inet_diag_lock_handler(sk->sk_protocol);
1370 	if (IS_ERR(handler)) {
1371 		inet_diag_unlock_handler(handler);
1372 		nlmsg_cancel(skb, nlh);
1373 		return PTR_ERR(handler);
1374 	}
1375 
1376 	attr = handler->idiag_info_size
1377 		? nla_reserve_64bit(skb, INET_DIAG_INFO,
1378 				    handler->idiag_info_size,
1379 				    INET_DIAG_PAD)
1380 		: NULL;
1381 	if (attr)
1382 		info = nla_data(attr);
1383 
1384 	handler->idiag_get_info(sk, r, info);
1385 	inet_diag_unlock_handler(handler);
1386 
1387 	nlmsg_end(skb, nlh);
1388 	return 0;
1389 }
1390 
1391 static const struct sock_diag_handler inet_diag_handler = {
1392 	.family = AF_INET,
1393 	.dump = inet_diag_handler_cmd,
1394 	.get_info = inet_diag_handler_get_info,
1395 	.destroy = inet_diag_handler_cmd,
1396 };
1397 
1398 static const struct sock_diag_handler inet6_diag_handler = {
1399 	.family = AF_INET6,
1400 	.dump = inet_diag_handler_cmd,
1401 	.get_info = inet_diag_handler_get_info,
1402 	.destroy = inet_diag_handler_cmd,
1403 };
1404 
1405 int inet_diag_register(const struct inet_diag_handler *h)
1406 {
1407 	const __u16 type = h->idiag_type;
1408 	int err = -EINVAL;
1409 
1410 	if (type >= IPPROTO_MAX)
1411 		goto out;
1412 
1413 	mutex_lock(&inet_diag_table_mutex);
1414 	err = -EEXIST;
1415 	if (!inet_diag_table[type]) {
1416 		inet_diag_table[type] = h;
1417 		err = 0;
1418 	}
1419 	mutex_unlock(&inet_diag_table_mutex);
1420 out:
1421 	return err;
1422 }
1423 EXPORT_SYMBOL_GPL(inet_diag_register);
1424 
1425 void inet_diag_unregister(const struct inet_diag_handler *h)
1426 {
1427 	const __u16 type = h->idiag_type;
1428 
1429 	if (type >= IPPROTO_MAX)
1430 		return;
1431 
1432 	mutex_lock(&inet_diag_table_mutex);
1433 	inet_diag_table[type] = NULL;
1434 	mutex_unlock(&inet_diag_table_mutex);
1435 }
1436 EXPORT_SYMBOL_GPL(inet_diag_unregister);
1437 
1438 static int __init inet_diag_init(void)
1439 {
1440 	const int inet_diag_table_size = (IPPROTO_MAX *
1441 					  sizeof(struct inet_diag_handler *));
1442 	int err = -ENOMEM;
1443 
1444 	inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1445 	if (!inet_diag_table)
1446 		goto out;
1447 
1448 	err = sock_diag_register(&inet_diag_handler);
1449 	if (err)
1450 		goto out_free_nl;
1451 
1452 	err = sock_diag_register(&inet6_diag_handler);
1453 	if (err)
1454 		goto out_free_inet;
1455 
1456 	sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1457 out:
1458 	return err;
1459 
1460 out_free_inet:
1461 	sock_diag_unregister(&inet_diag_handler);
1462 out_free_nl:
1463 	kfree(inet_diag_table);
1464 	goto out;
1465 }
1466 
1467 static void __exit inet_diag_exit(void)
1468 {
1469 	sock_diag_unregister(&inet6_diag_handler);
1470 	sock_diag_unregister(&inet_diag_handler);
1471 	sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1472 	kfree(inet_diag_table);
1473 }
1474 
1475 module_init(inet_diag_init);
1476 module_exit(inet_diag_exit);
1477 MODULE_LICENSE("GPL");
1478 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1479 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);
1480