xref: /openbmc/linux/net/ipv4/inet_diag.c (revision 2326f3cd) 
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, bool net_admin)
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 	if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
448 				     tw->tw_mark)) {
449 		nlmsg_cancel(skb, nlh);
450 		return -EMSGSIZE;
451 	}
452 
453 	nlmsg_end(skb, nlh);
454 	return 0;
455 }
456 
457 static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb,
458 			      struct netlink_callback *cb,
459 			      u16 nlmsg_flags, bool net_admin)
460 {
461 	struct request_sock *reqsk = inet_reqsk(sk);
462 	struct inet_diag_msg *r;
463 	struct nlmsghdr *nlh;
464 	long tmo;
465 
466 	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
467 			cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
468 	if (!nlh)
469 		return -EMSGSIZE;
470 
471 	r = nlmsg_data(nlh);
472 	inet_diag_msg_common_fill(r, sk);
473 	r->idiag_state = TCP_SYN_RECV;
474 	r->idiag_timer = 1;
475 	r->idiag_retrans = reqsk->num_retrans;
476 
477 	BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) !=
478 		     offsetof(struct sock, sk_cookie));
479 
480 	tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies;
481 	r->idiag_expires = jiffies_delta_to_msecs(tmo);
482 	r->idiag_rqueue	= 0;
483 	r->idiag_wqueue	= 0;
484 	r->idiag_uid	= 0;
485 	r->idiag_inode	= 0;
486 
487 	if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
488 				     inet_rsk(reqsk)->ir_mark)) {
489 		nlmsg_cancel(skb, nlh);
490 		return -EMSGSIZE;
491 	}
492 
493 	nlmsg_end(skb, nlh);
494 	return 0;
495 }
496 
497 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
498 			struct netlink_callback *cb,
499 			const struct inet_diag_req_v2 *r,
500 			u16 nlmsg_flags, bool net_admin)
501 {
502 	if (sk->sk_state == TCP_TIME_WAIT)
503 		return inet_twsk_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
504 
505 	if (sk->sk_state == TCP_NEW_SYN_RECV)
506 		return inet_req_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
507 
508 	return inet_sk_diag_fill(sk, inet_csk(sk), skb, cb, r, nlmsg_flags,
509 				 net_admin);
510 }
511 
512 struct sock *inet_diag_find_one_icsk(struct net *net,
513 				     struct inet_hashinfo *hashinfo,
514 				     const struct inet_diag_req_v2 *req)
515 {
516 	struct sock *sk;
517 
518 	rcu_read_lock();
519 	if (req->sdiag_family == AF_INET)
520 		sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0],
521 				 req->id.idiag_dport, req->id.idiag_src[0],
522 				 req->id.idiag_sport, req->id.idiag_if);
523 #if IS_ENABLED(CONFIG_IPV6)
524 	else if (req->sdiag_family == AF_INET6) {
525 		if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) &&
526 		    ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src))
527 			sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3],
528 					 req->id.idiag_dport, req->id.idiag_src[3],
529 					 req->id.idiag_sport, req->id.idiag_if);
530 		else
531 			sk = inet6_lookup(net, hashinfo, NULL, 0,
532 					  (struct in6_addr *)req->id.idiag_dst,
533 					  req->id.idiag_dport,
534 					  (struct in6_addr *)req->id.idiag_src,
535 					  req->id.idiag_sport,
536 					  req->id.idiag_if);
537 	}
538 #endif
539 	else {
540 		rcu_read_unlock();
541 		return ERR_PTR(-EINVAL);
542 	}
543 	rcu_read_unlock();
544 	if (!sk)
545 		return ERR_PTR(-ENOENT);
546 
547 	if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) {
548 		sock_gen_put(sk);
549 		return ERR_PTR(-ENOENT);
550 	}
551 
552 	return sk;
553 }
554 EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk);
555 
556 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo,
557 			    struct netlink_callback *cb,
558 			    const struct inet_diag_req_v2 *req)
559 {
560 	struct sk_buff *in_skb = cb->skb;
561 	bool net_admin = netlink_net_capable(in_skb, CAP_NET_ADMIN);
562 	struct net *net = sock_net(in_skb->sk);
563 	struct sk_buff *rep;
564 	struct sock *sk;
565 	int err;
566 
567 	sk = inet_diag_find_one_icsk(net, hashinfo, req);
568 	if (IS_ERR(sk))
569 		return PTR_ERR(sk);
570 
571 	rep = nlmsg_new(inet_sk_attr_size(sk, req, net_admin), GFP_KERNEL);
572 	if (!rep) {
573 		err = -ENOMEM;
574 		goto out;
575 	}
576 
577 	err = sk_diag_fill(sk, rep, cb, req, 0, net_admin);
578 	if (err < 0) {
579 		WARN_ON(err == -EMSGSIZE);
580 		nlmsg_free(rep);
581 		goto out;
582 	}
583 	err = nlmsg_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid);
584 
585 out:
586 	if (sk)
587 		sock_gen_put(sk);
588 
589 	return err;
590 }
591 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
592 
593 static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb,
594 			       const struct nlmsghdr *nlh,
595 			       int hdrlen,
596 			       const struct inet_diag_req_v2 *req)
597 {
598 	const struct inet_diag_handler *handler;
599 	struct inet_diag_dump_data dump_data;
600 	int err, protocol;
601 
602 	memset(&dump_data, 0, sizeof(dump_data));
603 	err = inet_diag_parse_attrs(nlh, hdrlen, dump_data.req_nlas);
604 	if (err)
605 		return err;
606 
607 	protocol = inet_diag_get_protocol(req, &dump_data);
608 
609 	handler = inet_diag_lock_handler(protocol);
610 	if (IS_ERR(handler)) {
611 		err = PTR_ERR(handler);
612 	} else if (cmd == SOCK_DIAG_BY_FAMILY) {
613 		struct netlink_callback cb = {
614 			.nlh = nlh,
615 			.skb = in_skb,
616 			.data = &dump_data,
617 		};
618 		err = handler->dump_one(&cb, req);
619 	} else if (cmd == SOCK_DESTROY && handler->destroy) {
620 		err = handler->destroy(in_skb, req);
621 	} else {
622 		err = -EOPNOTSUPP;
623 	}
624 	inet_diag_unlock_handler(handler);
625 
626 	return err;
627 }
628 
629 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
630 {
631 	int words = bits >> 5;
632 
633 	bits &= 0x1f;
634 
635 	if (words) {
636 		if (memcmp(a1, a2, words << 2))
637 			return 0;
638 	}
639 	if (bits) {
640 		__be32 w1, w2;
641 		__be32 mask;
642 
643 		w1 = a1[words];
644 		w2 = a2[words];
645 
646 		mask = htonl((0xffffffff) << (32 - bits));
647 
648 		if ((w1 ^ w2) & mask)
649 			return 0;
650 	}
651 
652 	return 1;
653 }
654 
655 static int inet_diag_bc_run(const struct nlattr *_bc,
656 			    const struct inet_diag_entry *entry)
657 {
658 	const void *bc = nla_data(_bc);
659 	int len = nla_len(_bc);
660 
661 	while (len > 0) {
662 		int yes = 1;
663 		const struct inet_diag_bc_op *op = bc;
664 
665 		switch (op->code) {
666 		case INET_DIAG_BC_NOP:
667 			break;
668 		case INET_DIAG_BC_JMP:
669 			yes = 0;
670 			break;
671 		case INET_DIAG_BC_S_EQ:
672 			yes = entry->sport == op[1].no;
673 			break;
674 		case INET_DIAG_BC_S_GE:
675 			yes = entry->sport >= op[1].no;
676 			break;
677 		case INET_DIAG_BC_S_LE:
678 			yes = entry->sport <= op[1].no;
679 			break;
680 		case INET_DIAG_BC_D_EQ:
681 			yes = entry->dport == op[1].no;
682 			break;
683 		case INET_DIAG_BC_D_GE:
684 			yes = entry->dport >= op[1].no;
685 			break;
686 		case INET_DIAG_BC_D_LE:
687 			yes = entry->dport <= op[1].no;
688 			break;
689 		case INET_DIAG_BC_AUTO:
690 			yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
691 			break;
692 		case INET_DIAG_BC_S_COND:
693 		case INET_DIAG_BC_D_COND: {
694 			const struct inet_diag_hostcond *cond;
695 			const __be32 *addr;
696 
697 			cond = (const struct inet_diag_hostcond *)(op + 1);
698 			if (cond->port != -1 &&
699 			    cond->port != (op->code == INET_DIAG_BC_S_COND ?
700 					     entry->sport : entry->dport)) {
701 				yes = 0;
702 				break;
703 			}
704 
705 			if (op->code == INET_DIAG_BC_S_COND)
706 				addr = entry->saddr;
707 			else
708 				addr = entry->daddr;
709 
710 			if (cond->family != AF_UNSPEC &&
711 			    cond->family != entry->family) {
712 				if (entry->family == AF_INET6 &&
713 				    cond->family == AF_INET) {
714 					if (addr[0] == 0 && addr[1] == 0 &&
715 					    addr[2] == htonl(0xffff) &&
716 					    bitstring_match(addr + 3,
717 							    cond->addr,
718 							    cond->prefix_len))
719 						break;
720 				}
721 				yes = 0;
722 				break;
723 			}
724 
725 			if (cond->prefix_len == 0)
726 				break;
727 			if (bitstring_match(addr, cond->addr,
728 					    cond->prefix_len))
729 				break;
730 			yes = 0;
731 			break;
732 		}
733 		case INET_DIAG_BC_DEV_COND: {
734 			u32 ifindex;
735 
736 			ifindex = *((const u32 *)(op + 1));
737 			if (ifindex != entry->ifindex)
738 				yes = 0;
739 			break;
740 		}
741 		case INET_DIAG_BC_MARK_COND: {
742 			struct inet_diag_markcond *cond;
743 
744 			cond = (struct inet_diag_markcond *)(op + 1);
745 			if ((entry->mark & cond->mask) != cond->mark)
746 				yes = 0;
747 			break;
748 		}
749 #ifdef CONFIG_SOCK_CGROUP_DATA
750 		case INET_DIAG_BC_CGROUP_COND: {
751 			u64 cgroup_id;
752 
753 			cgroup_id = get_unaligned((const u64 *)(op + 1));
754 			if (cgroup_id != entry->cgroup_id)
755 				yes = 0;
756 			break;
757 		}
758 #endif
759 		}
760 
761 		if (yes) {
762 			len -= op->yes;
763 			bc += op->yes;
764 		} else {
765 			len -= op->no;
766 			bc += op->no;
767 		}
768 	}
769 	return len == 0;
770 }
771 
772 /* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV)
773  */
774 static void entry_fill_addrs(struct inet_diag_entry *entry,
775 			     const struct sock *sk)
776 {
777 #if IS_ENABLED(CONFIG_IPV6)
778 	if (sk->sk_family == AF_INET6) {
779 		entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32;
780 		entry->daddr = sk->sk_v6_daddr.s6_addr32;
781 	} else
782 #endif
783 	{
784 		entry->saddr = &sk->sk_rcv_saddr;
785 		entry->daddr = &sk->sk_daddr;
786 	}
787 }
788 
789 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
790 {
791 	struct inet_sock *inet = inet_sk(sk);
792 	struct inet_diag_entry entry;
793 
794 	if (!bc)
795 		return 1;
796 
797 	entry.family = sk->sk_family;
798 	entry_fill_addrs(&entry, sk);
799 	entry.sport = inet->inet_num;
800 	entry.dport = ntohs(inet->inet_dport);
801 	entry.ifindex = sk->sk_bound_dev_if;
802 	entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0;
803 	if (sk_fullsock(sk))
804 		entry.mark = sk->sk_mark;
805 	else if (sk->sk_state == TCP_NEW_SYN_RECV)
806 		entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark;
807 	else if (sk->sk_state == TCP_TIME_WAIT)
808 		entry.mark = inet_twsk(sk)->tw_mark;
809 	else
810 		entry.mark = 0;
811 #ifdef CONFIG_SOCK_CGROUP_DATA
812 	entry.cgroup_id = sk_fullsock(sk) ?
813 		cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)) : 0;
814 #endif
815 
816 	return inet_diag_bc_run(bc, &entry);
817 }
818 EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
819 
820 static int valid_cc(const void *bc, int len, int cc)
821 {
822 	while (len >= 0) {
823 		const struct inet_diag_bc_op *op = bc;
824 
825 		if (cc > len)
826 			return 0;
827 		if (cc == len)
828 			return 1;
829 		if (op->yes < 4 || op->yes & 3)
830 			return 0;
831 		len -= op->yes;
832 		bc  += op->yes;
833 	}
834 	return 0;
835 }
836 
837 /* data is u32 ifindex */
838 static bool valid_devcond(const struct inet_diag_bc_op *op, int len,
839 			  int *min_len)
840 {
841 	/* Check ifindex space. */
842 	*min_len += sizeof(u32);
843 	if (len < *min_len)
844 		return false;
845 
846 	return true;
847 }
848 /* Validate an inet_diag_hostcond. */
849 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
850 			   int *min_len)
851 {
852 	struct inet_diag_hostcond *cond;
853 	int addr_len;
854 
855 	/* Check hostcond space. */
856 	*min_len += sizeof(struct inet_diag_hostcond);
857 	if (len < *min_len)
858 		return false;
859 	cond = (struct inet_diag_hostcond *)(op + 1);
860 
861 	/* Check address family and address length. */
862 	switch (cond->family) {
863 	case AF_UNSPEC:
864 		addr_len = 0;
865 		break;
866 	case AF_INET:
867 		addr_len = sizeof(struct in_addr);
868 		break;
869 	case AF_INET6:
870 		addr_len = sizeof(struct in6_addr);
871 		break;
872 	default:
873 		return false;
874 	}
875 	*min_len += addr_len;
876 	if (len < *min_len)
877 		return false;
878 
879 	/* Check prefix length (in bits) vs address length (in bytes). */
880 	if (cond->prefix_len > 8 * addr_len)
881 		return false;
882 
883 	return true;
884 }
885 
886 /* Validate a port comparison operator. */
887 static bool valid_port_comparison(const struct inet_diag_bc_op *op,
888 				  int len, int *min_len)
889 {
890 	/* Port comparisons put the port in a follow-on inet_diag_bc_op. */
891 	*min_len += sizeof(struct inet_diag_bc_op);
892 	if (len < *min_len)
893 		return false;
894 	return true;
895 }
896 
897 static bool valid_markcond(const struct inet_diag_bc_op *op, int len,
898 			   int *min_len)
899 {
900 	*min_len += sizeof(struct inet_diag_markcond);
901 	return len >= *min_len;
902 }
903 
904 #ifdef CONFIG_SOCK_CGROUP_DATA
905 static bool valid_cgroupcond(const struct inet_diag_bc_op *op, int len,
906 			     int *min_len)
907 {
908 	*min_len += sizeof(u64);
909 	return len >= *min_len;
910 }
911 #endif
912 
913 static int inet_diag_bc_audit(const struct nlattr *attr,
914 			      const struct sk_buff *skb)
915 {
916 	bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN);
917 	const void *bytecode, *bc;
918 	int bytecode_len, len;
919 
920 	if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op))
921 		return -EINVAL;
922 
923 	bytecode = bc = nla_data(attr);
924 	len = bytecode_len = nla_len(attr);
925 
926 	while (len > 0) {
927 		int min_len = sizeof(struct inet_diag_bc_op);
928 		const struct inet_diag_bc_op *op = bc;
929 
930 		switch (op->code) {
931 		case INET_DIAG_BC_S_COND:
932 		case INET_DIAG_BC_D_COND:
933 			if (!valid_hostcond(bc, len, &min_len))
934 				return -EINVAL;
935 			break;
936 		case INET_DIAG_BC_DEV_COND:
937 			if (!valid_devcond(bc, len, &min_len))
938 				return -EINVAL;
939 			break;
940 		case INET_DIAG_BC_S_EQ:
941 		case INET_DIAG_BC_S_GE:
942 		case INET_DIAG_BC_S_LE:
943 		case INET_DIAG_BC_D_EQ:
944 		case INET_DIAG_BC_D_GE:
945 		case INET_DIAG_BC_D_LE:
946 			if (!valid_port_comparison(bc, len, &min_len))
947 				return -EINVAL;
948 			break;
949 		case INET_DIAG_BC_MARK_COND:
950 			if (!net_admin)
951 				return -EPERM;
952 			if (!valid_markcond(bc, len, &min_len))
953 				return -EINVAL;
954 			break;
955 #ifdef CONFIG_SOCK_CGROUP_DATA
956 		case INET_DIAG_BC_CGROUP_COND:
957 			if (!valid_cgroupcond(bc, len, &min_len))
958 				return -EINVAL;
959 			break;
960 #endif
961 		case INET_DIAG_BC_AUTO:
962 		case INET_DIAG_BC_JMP:
963 		case INET_DIAG_BC_NOP:
964 			break;
965 		default:
966 			return -EINVAL;
967 		}
968 
969 		if (op->code != INET_DIAG_BC_NOP) {
970 			if (op->no < min_len || op->no > len + 4 || op->no & 3)
971 				return -EINVAL;
972 			if (op->no < len &&
973 			    !valid_cc(bytecode, bytecode_len, len - op->no))
974 				return -EINVAL;
975 		}
976 
977 		if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
978 			return -EINVAL;
979 		bc  += op->yes;
980 		len -= op->yes;
981 	}
982 	return len == 0 ? 0 : -EINVAL;
983 }
984 
985 static void twsk_build_assert(void)
986 {
987 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) !=
988 		     offsetof(struct sock, sk_family));
989 
990 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) !=
991 		     offsetof(struct inet_sock, inet_num));
992 
993 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) !=
994 		     offsetof(struct inet_sock, inet_dport));
995 
996 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) !=
997 		     offsetof(struct inet_sock, inet_rcv_saddr));
998 
999 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) !=
1000 		     offsetof(struct inet_sock, inet_daddr));
1001 
1002 #if IS_ENABLED(CONFIG_IPV6)
1003 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) !=
1004 		     offsetof(struct sock, sk_v6_rcv_saddr));
1005 
1006 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) !=
1007 		     offsetof(struct sock, sk_v6_daddr));
1008 #endif
1009 }
1010 
1011 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
1012 			 struct netlink_callback *cb,
1013 			 const struct inet_diag_req_v2 *r)
1014 {
1015 	bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN);
1016 	struct inet_diag_dump_data *cb_data = cb->data;
1017 	struct net *net = sock_net(skb->sk);
1018 	u32 idiag_states = r->idiag_states;
1019 	int i, num, s_i, s_num;
1020 	struct nlattr *bc;
1021 	struct sock *sk;
1022 
1023 	bc = cb_data->inet_diag_nla_bc;
1024 	if (idiag_states & TCPF_SYN_RECV)
1025 		idiag_states |= TCPF_NEW_SYN_RECV;
1026 	s_i = cb->args[1];
1027 	s_num = num = cb->args[2];
1028 
1029 	if (cb->args[0] == 0) {
1030 		if (!(idiag_states & TCPF_LISTEN) || r->id.idiag_dport)
1031 			goto skip_listen_ht;
1032 
1033 		for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
1034 			struct inet_listen_hashbucket *ilb;
1035 			struct hlist_nulls_node *node;
1036 
1037 			num = 0;
1038 			ilb = &hashinfo->listening_hash[i];
1039 			spin_lock(&ilb->lock);
1040 			sk_nulls_for_each(sk, node, &ilb->nulls_head) {
1041 				struct inet_sock *inet = inet_sk(sk);
1042 
1043 				if (!net_eq(sock_net(sk), net))
1044 					continue;
1045 
1046 				if (num < s_num) {
1047 					num++;
1048 					continue;
1049 				}
1050 
1051 				if (r->sdiag_family != AF_UNSPEC &&
1052 				    sk->sk_family != r->sdiag_family)
1053 					goto next_listen;
1054 
1055 				if (r->id.idiag_sport != inet->inet_sport &&
1056 				    r->id.idiag_sport)
1057 					goto next_listen;
1058 
1059 				if (!inet_diag_bc_sk(bc, sk))
1060 					goto next_listen;
1061 
1062 				if (inet_sk_diag_fill(sk, inet_csk(sk), skb,
1063 						      cb, r, NLM_F_MULTI,
1064 						      net_admin) < 0) {
1065 					spin_unlock(&ilb->lock);
1066 					goto done;
1067 				}
1068 
1069 next_listen:
1070 				++num;
1071 			}
1072 			spin_unlock(&ilb->lock);
1073 
1074 			s_num = 0;
1075 		}
1076 skip_listen_ht:
1077 		cb->args[0] = 1;
1078 		s_i = num = s_num = 0;
1079 	}
1080 
1081 	if (!(idiag_states & ~TCPF_LISTEN))
1082 		goto out;
1083 
1084 #define SKARR_SZ 16
1085 	for (i = s_i; i <= hashinfo->ehash_mask; i++) {
1086 		struct inet_ehash_bucket *head = &hashinfo->ehash[i];
1087 		spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
1088 		struct hlist_nulls_node *node;
1089 		struct sock *sk_arr[SKARR_SZ];
1090 		int num_arr[SKARR_SZ];
1091 		int idx, accum, res;
1092 
1093 		if (hlist_nulls_empty(&head->chain))
1094 			continue;
1095 
1096 		if (i > s_i)
1097 			s_num = 0;
1098 
1099 next_chunk:
1100 		num = 0;
1101 		accum = 0;
1102 		spin_lock_bh(lock);
1103 		sk_nulls_for_each(sk, node, &head->chain) {
1104 			int state;
1105 
1106 			if (!net_eq(sock_net(sk), net))
1107 				continue;
1108 			if (num < s_num)
1109 				goto next_normal;
1110 			state = (sk->sk_state == TCP_TIME_WAIT) ?
1111 				inet_twsk(sk)->tw_substate : sk->sk_state;
1112 			if (!(idiag_states & (1 << state)))
1113 				goto next_normal;
1114 			if (r->sdiag_family != AF_UNSPEC &&
1115 			    sk->sk_family != r->sdiag_family)
1116 				goto next_normal;
1117 			if (r->id.idiag_sport != htons(sk->sk_num) &&
1118 			    r->id.idiag_sport)
1119 				goto next_normal;
1120 			if (r->id.idiag_dport != sk->sk_dport &&
1121 			    r->id.idiag_dport)
1122 				goto next_normal;
1123 			twsk_build_assert();
1124 
1125 			if (!inet_diag_bc_sk(bc, sk))
1126 				goto next_normal;
1127 
1128 			if (!refcount_inc_not_zero(&sk->sk_refcnt))
1129 				goto next_normal;
1130 
1131 			num_arr[accum] = num;
1132 			sk_arr[accum] = sk;
1133 			if (++accum == SKARR_SZ)
1134 				break;
1135 next_normal:
1136 			++num;
1137 		}
1138 		spin_unlock_bh(lock);
1139 		res = 0;
1140 		for (idx = 0; idx < accum; idx++) {
1141 			if (res >= 0) {
1142 				res = sk_diag_fill(sk_arr[idx], skb, cb, r,
1143 						   NLM_F_MULTI, net_admin);
1144 				if (res < 0)
1145 					num = num_arr[idx];
1146 			}
1147 			sock_gen_put(sk_arr[idx]);
1148 		}
1149 		if (res < 0)
1150 			break;
1151 		cond_resched();
1152 		if (accum == SKARR_SZ) {
1153 			s_num = num + 1;
1154 			goto next_chunk;
1155 		}
1156 	}
1157 
1158 done:
1159 	cb->args[1] = i;
1160 	cb->args[2] = num;
1161 out:
1162 	;
1163 }
1164 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
1165 
1166 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
1167 			    const struct inet_diag_req_v2 *r)
1168 {
1169 	struct inet_diag_dump_data *cb_data = cb->data;
1170 	const struct inet_diag_handler *handler;
1171 	u32 prev_min_dump_alloc;
1172 	int protocol, err = 0;
1173 
1174 	protocol = inet_diag_get_protocol(r, cb_data);
1175 
1176 again:
1177 	prev_min_dump_alloc = cb->min_dump_alloc;
1178 	handler = inet_diag_lock_handler(protocol);
1179 	if (!IS_ERR(handler))
1180 		handler->dump(skb, cb, r);
1181 	else
1182 		err = PTR_ERR(handler);
1183 	inet_diag_unlock_handler(handler);
1184 
1185 	/* The skb is not large enough to fit one sk info and
1186 	 * inet_sk_diag_fill() has requested for a larger skb.
1187 	 */
1188 	if (!skb->len && cb->min_dump_alloc > prev_min_dump_alloc) {
1189 		err = pskb_expand_head(skb, 0, cb->min_dump_alloc, GFP_KERNEL);
1190 		if (!err)
1191 			goto again;
1192 	}
1193 
1194 	return err ? : skb->len;
1195 }
1196 
1197 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
1198 {
1199 	return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh));
1200 }
1201 
1202 static int __inet_diag_dump_start(struct netlink_callback *cb, int hdrlen)
1203 {
1204 	const struct nlmsghdr *nlh = cb->nlh;
1205 	struct inet_diag_dump_data *cb_data;
1206 	struct sk_buff *skb = cb->skb;
1207 	struct nlattr *nla;
1208 	int err;
1209 
1210 	cb_data = kzalloc(sizeof(*cb_data), GFP_KERNEL);
1211 	if (!cb_data)
1212 		return -ENOMEM;
1213 
1214 	err = inet_diag_parse_attrs(nlh, hdrlen, cb_data->req_nlas);
1215 	if (err) {
1216 		kfree(cb_data);
1217 		return err;
1218 	}
1219 	nla = cb_data->inet_diag_nla_bc;
1220 	if (nla) {
1221 		err = inet_diag_bc_audit(nla, skb);
1222 		if (err) {
1223 			kfree(cb_data);
1224 			return err;
1225 		}
1226 	}
1227 
1228 	nla = cb_data->inet_diag_nla_bpf_stgs;
1229 	if (nla) {
1230 		struct bpf_sk_storage_diag *bpf_stg_diag;
1231 
1232 		bpf_stg_diag = bpf_sk_storage_diag_alloc(nla);
1233 		if (IS_ERR(bpf_stg_diag)) {
1234 			kfree(cb_data);
1235 			return PTR_ERR(bpf_stg_diag);
1236 		}
1237 		cb_data->bpf_stg_diag = bpf_stg_diag;
1238 	}
1239 
1240 	cb->data = cb_data;
1241 	return 0;
1242 }
1243 
1244 static int inet_diag_dump_start(struct netlink_callback *cb)
1245 {
1246 	return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req_v2));
1247 }
1248 
1249 static int inet_diag_dump_start_compat(struct netlink_callback *cb)
1250 {
1251 	return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req));
1252 }
1253 
1254 static int inet_diag_dump_done(struct netlink_callback *cb)
1255 {
1256 	struct inet_diag_dump_data *cb_data = cb->data;
1257 
1258 	bpf_sk_storage_diag_free(cb_data->bpf_stg_diag);
1259 	kfree(cb->data);
1260 
1261 	return 0;
1262 }
1263 
1264 static int inet_diag_type2proto(int type)
1265 {
1266 	switch (type) {
1267 	case TCPDIAG_GETSOCK:
1268 		return IPPROTO_TCP;
1269 	case DCCPDIAG_GETSOCK:
1270 		return IPPROTO_DCCP;
1271 	default:
1272 		return 0;
1273 	}
1274 }
1275 
1276 static int inet_diag_dump_compat(struct sk_buff *skb,
1277 				 struct netlink_callback *cb)
1278 {
1279 	struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1280 	struct inet_diag_req_v2 req;
1281 
1282 	req.sdiag_family = AF_UNSPEC; /* compatibility */
1283 	req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1284 	req.idiag_ext = rc->idiag_ext;
1285 	req.idiag_states = rc->idiag_states;
1286 	req.id = rc->id;
1287 
1288 	return __inet_diag_dump(skb, cb, &req);
1289 }
1290 
1291 static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1292 				      const struct nlmsghdr *nlh)
1293 {
1294 	struct inet_diag_req *rc = nlmsg_data(nlh);
1295 	struct inet_diag_req_v2 req;
1296 
1297 	req.sdiag_family = rc->idiag_family;
1298 	req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1299 	req.idiag_ext = rc->idiag_ext;
1300 	req.idiag_states = rc->idiag_states;
1301 	req.id = rc->id;
1302 
1303 	return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh,
1304 				   sizeof(struct inet_diag_req), &req);
1305 }
1306 
1307 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1308 {
1309 	int hdrlen = sizeof(struct inet_diag_req);
1310 	struct net *net = sock_net(skb->sk);
1311 
1312 	if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1313 	    nlmsg_len(nlh) < hdrlen)
1314 		return -EINVAL;
1315 
1316 	if (nlh->nlmsg_flags & NLM_F_DUMP) {
1317 		struct netlink_dump_control c = {
1318 			.start = inet_diag_dump_start_compat,
1319 			.done = inet_diag_dump_done,
1320 			.dump = inet_diag_dump_compat,
1321 		};
1322 		return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1323 	}
1324 
1325 	return inet_diag_get_exact_compat(skb, nlh);
1326 }
1327 
1328 static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h)
1329 {
1330 	int hdrlen = sizeof(struct inet_diag_req_v2);
1331 	struct net *net = sock_net(skb->sk);
1332 
1333 	if (nlmsg_len(h) < hdrlen)
1334 		return -EINVAL;
1335 
1336 	if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY &&
1337 	    h->nlmsg_flags & NLM_F_DUMP) {
1338 		struct netlink_dump_control c = {
1339 			.start = inet_diag_dump_start,
1340 			.done = inet_diag_dump_done,
1341 			.dump = inet_diag_dump,
1342 		};
1343 		return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1344 	}
1345 
1346 	return inet_diag_cmd_exact(h->nlmsg_type, skb, h, hdrlen,
1347 				   nlmsg_data(h));
1348 }
1349 
1350 static
1351 int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk)
1352 {
1353 	const struct inet_diag_handler *handler;
1354 	struct nlmsghdr *nlh;
1355 	struct nlattr *attr;
1356 	struct inet_diag_msg *r;
1357 	void *info = NULL;
1358 	int err = 0;
1359 
1360 	nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0);
1361 	if (!nlh)
1362 		return -ENOMEM;
1363 
1364 	r = nlmsg_data(nlh);
1365 	memset(r, 0, sizeof(*r));
1366 	inet_diag_msg_common_fill(r, sk);
1367 	if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM)
1368 		r->id.idiag_sport = inet_sk(sk)->inet_sport;
1369 	r->idiag_state = sk->sk_state;
1370 
1371 	if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) {
1372 		nlmsg_cancel(skb, nlh);
1373 		return err;
1374 	}
1375 
1376 	handler = inet_diag_lock_handler(sk->sk_protocol);
1377 	if (IS_ERR(handler)) {
1378 		inet_diag_unlock_handler(handler);
1379 		nlmsg_cancel(skb, nlh);
1380 		return PTR_ERR(handler);
1381 	}
1382 
1383 	attr = handler->idiag_info_size
1384 		? nla_reserve_64bit(skb, INET_DIAG_INFO,
1385 				    handler->idiag_info_size,
1386 				    INET_DIAG_PAD)
1387 		: NULL;
1388 	if (attr)
1389 		info = nla_data(attr);
1390 
1391 	handler->idiag_get_info(sk, r, info);
1392 	inet_diag_unlock_handler(handler);
1393 
1394 	nlmsg_end(skb, nlh);
1395 	return 0;
1396 }
1397 
1398 static const struct sock_diag_handler inet_diag_handler = {
1399 	.family = AF_INET,
1400 	.dump = inet_diag_handler_cmd,
1401 	.get_info = inet_diag_handler_get_info,
1402 	.destroy = inet_diag_handler_cmd,
1403 };
1404 
1405 static const struct sock_diag_handler inet6_diag_handler = {
1406 	.family = AF_INET6,
1407 	.dump = inet_diag_handler_cmd,
1408 	.get_info = inet_diag_handler_get_info,
1409 	.destroy = inet_diag_handler_cmd,
1410 };
1411 
1412 int inet_diag_register(const struct inet_diag_handler *h)
1413 {
1414 	const __u16 type = h->idiag_type;
1415 	int err = -EINVAL;
1416 
1417 	if (type >= IPPROTO_MAX)
1418 		goto out;
1419 
1420 	mutex_lock(&inet_diag_table_mutex);
1421 	err = -EEXIST;
1422 	if (!inet_diag_table[type]) {
1423 		inet_diag_table[type] = h;
1424 		err = 0;
1425 	}
1426 	mutex_unlock(&inet_diag_table_mutex);
1427 out:
1428 	return err;
1429 }
1430 EXPORT_SYMBOL_GPL(inet_diag_register);
1431 
1432 void inet_diag_unregister(const struct inet_diag_handler *h)
1433 {
1434 	const __u16 type = h->idiag_type;
1435 
1436 	if (type >= IPPROTO_MAX)
1437 		return;
1438 
1439 	mutex_lock(&inet_diag_table_mutex);
1440 	inet_diag_table[type] = NULL;
1441 	mutex_unlock(&inet_diag_table_mutex);
1442 }
1443 EXPORT_SYMBOL_GPL(inet_diag_unregister);
1444 
1445 static int __init inet_diag_init(void)
1446 {
1447 	const int inet_diag_table_size = (IPPROTO_MAX *
1448 					  sizeof(struct inet_diag_handler *));
1449 	int err = -ENOMEM;
1450 
1451 	inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1452 	if (!inet_diag_table)
1453 		goto out;
1454 
1455 	err = sock_diag_register(&inet_diag_handler);
1456 	if (err)
1457 		goto out_free_nl;
1458 
1459 	err = sock_diag_register(&inet6_diag_handler);
1460 	if (err)
1461 		goto out_free_inet;
1462 
1463 	sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1464 out:
1465 	return err;
1466 
1467 out_free_inet:
1468 	sock_diag_unregister(&inet_diag_handler);
1469 out_free_nl:
1470 	kfree(inet_diag_table);
1471 	goto out;
1472 }
1473 
1474 static void __exit inet_diag_exit(void)
1475 {
1476 	sock_diag_unregister(&inet6_diag_handler);
1477 	sock_diag_unregister(&inet_diag_handler);
1478 	sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1479 	kfree(inet_diag_table);
1480 }
1481 
1482 module_init(inet_diag_init);
1483 module_exit(inet_diag_exit);
1484 MODULE_LICENSE("GPL");
1485 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1486 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);
1487