xref: /openbmc/linux/net/ipv4/inet_diag.c (revision 0760aad038b5a032c31ea124feed63d88627d2f1)
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 void 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_MAX)
216 			req_nlas[type] = nla;
217 	}
218 }
219 
220 static int inet_diag_get_protocol(const struct inet_diag_req_v2 *req,
221 				  const struct inet_diag_dump_data *data)
222 {
223 	if (data->req_nlas[INET_DIAG_REQ_PROTOCOL])
224 		return nla_get_u32(data->req_nlas[INET_DIAG_REQ_PROTOCOL]);
225 	return req->sdiag_protocol;
226 }
227 
228 #define MAX_DUMP_ALLOC_SIZE (KMALLOC_MAX_SIZE - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
229 
230 int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk,
231 		      struct sk_buff *skb, struct netlink_callback *cb,
232 		      const struct inet_diag_req_v2 *req,
233 		      u16 nlmsg_flags, bool net_admin)
234 {
235 	const struct tcp_congestion_ops *ca_ops;
236 	const struct inet_diag_handler *handler;
237 	struct inet_diag_dump_data *cb_data;
238 	int ext = req->idiag_ext;
239 	struct inet_diag_msg *r;
240 	struct nlmsghdr  *nlh;
241 	struct nlattr *attr;
242 	void *info = NULL;
243 
244 	cb_data = cb->data;
245 	handler = inet_diag_table[inet_diag_get_protocol(req, cb_data)];
246 	BUG_ON(!handler);
247 
248 	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
249 			cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
250 	if (!nlh)
251 		return -EMSGSIZE;
252 
253 	r = nlmsg_data(nlh);
254 	BUG_ON(!sk_fullsock(sk));
255 
256 	inet_diag_msg_common_fill(r, sk);
257 	r->idiag_state = sk->sk_state;
258 	r->idiag_timer = 0;
259 	r->idiag_retrans = 0;
260 
261 	if (inet_diag_msg_attrs_fill(sk, skb, r, ext,
262 				     sk_user_ns(NETLINK_CB(cb->skb).sk),
263 				     net_admin))
264 		goto errout;
265 
266 	if (ext & (1 << (INET_DIAG_MEMINFO - 1))) {
267 		struct inet_diag_meminfo minfo = {
268 			.idiag_rmem = sk_rmem_alloc_get(sk),
269 			.idiag_wmem = READ_ONCE(sk->sk_wmem_queued),
270 			.idiag_fmem = sk->sk_forward_alloc,
271 			.idiag_tmem = sk_wmem_alloc_get(sk),
272 		};
273 
274 		if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0)
275 			goto errout;
276 	}
277 
278 	if (ext & (1 << (INET_DIAG_SKMEMINFO - 1)))
279 		if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO))
280 			goto errout;
281 
282 	/*
283 	 * RAW sockets might have user-defined protocols assigned,
284 	 * so report the one supplied on socket creation.
285 	 */
286 	if (sk->sk_type == SOCK_RAW) {
287 		if (nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))
288 			goto errout;
289 	}
290 
291 	if (!icsk) {
292 		handler->idiag_get_info(sk, r, NULL);
293 		goto out;
294 	}
295 
296 	if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
297 	    icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
298 	    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
299 		r->idiag_timer = 1;
300 		r->idiag_retrans = icsk->icsk_retransmits;
301 		r->idiag_expires =
302 			jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
303 	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
304 		r->idiag_timer = 4;
305 		r->idiag_retrans = icsk->icsk_probes_out;
306 		r->idiag_expires =
307 			jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
308 	} else if (timer_pending(&sk->sk_timer)) {
309 		r->idiag_timer = 2;
310 		r->idiag_retrans = icsk->icsk_probes_out;
311 		r->idiag_expires =
312 			jiffies_delta_to_msecs(sk->sk_timer.expires - jiffies);
313 	} else {
314 		r->idiag_timer = 0;
315 		r->idiag_expires = 0;
316 	}
317 
318 	if ((ext & (1 << (INET_DIAG_INFO - 1))) && handler->idiag_info_size) {
319 		attr = nla_reserve_64bit(skb, INET_DIAG_INFO,
320 					 handler->idiag_info_size,
321 					 INET_DIAG_PAD);
322 		if (!attr)
323 			goto errout;
324 
325 		info = nla_data(attr);
326 	}
327 
328 	if (ext & (1 << (INET_DIAG_CONG - 1))) {
329 		int err = 0;
330 
331 		rcu_read_lock();
332 		ca_ops = READ_ONCE(icsk->icsk_ca_ops);
333 		if (ca_ops)
334 			err = nla_put_string(skb, INET_DIAG_CONG, ca_ops->name);
335 		rcu_read_unlock();
336 		if (err < 0)
337 			goto errout;
338 	}
339 
340 	handler->idiag_get_info(sk, r, info);
341 
342 	if (ext & (1 << (INET_DIAG_INFO - 1)) && handler->idiag_get_aux)
343 		if (handler->idiag_get_aux(sk, net_admin, skb) < 0)
344 			goto errout;
345 
346 	if (sk->sk_state < TCP_TIME_WAIT) {
347 		union tcp_cc_info info;
348 		size_t sz = 0;
349 		int attr;
350 
351 		rcu_read_lock();
352 		ca_ops = READ_ONCE(icsk->icsk_ca_ops);
353 		if (ca_ops && ca_ops->get_info)
354 			sz = ca_ops->get_info(sk, ext, &attr, &info);
355 		rcu_read_unlock();
356 		if (sz && nla_put(skb, attr, sz, &info) < 0)
357 			goto errout;
358 	}
359 
360 	/* Keep it at the end for potential retry with a larger skb,
361 	 * or else do best-effort fitting, which is only done for the
362 	 * first_nlmsg.
363 	 */
364 	if (cb_data->bpf_stg_diag) {
365 		bool first_nlmsg = ((unsigned char *)nlh == skb->data);
366 		unsigned int prev_min_dump_alloc;
367 		unsigned int total_nla_size = 0;
368 		unsigned int msg_len;
369 		int err;
370 
371 		msg_len = skb_tail_pointer(skb) - (unsigned char *)nlh;
372 		err = bpf_sk_storage_diag_put(cb_data->bpf_stg_diag, sk, skb,
373 					      INET_DIAG_SK_BPF_STORAGES,
374 					      &total_nla_size);
375 
376 		if (!err)
377 			goto out;
378 
379 		total_nla_size += msg_len;
380 		prev_min_dump_alloc = cb->min_dump_alloc;
381 		if (total_nla_size > prev_min_dump_alloc)
382 			cb->min_dump_alloc = min_t(u32, total_nla_size,
383 						   MAX_DUMP_ALLOC_SIZE);
384 
385 		if (!first_nlmsg)
386 			goto errout;
387 
388 		if (cb->min_dump_alloc > prev_min_dump_alloc)
389 			/* Retry with pskb_expand_head() with
390 			 * __GFP_DIRECT_RECLAIM
391 			 */
392 			goto errout;
393 
394 		WARN_ON_ONCE(total_nla_size <= prev_min_dump_alloc);
395 
396 		/* Send what we have for this sk
397 		 * and move on to the next sk in the following
398 		 * dump()
399 		 */
400 	}
401 
402 out:
403 	nlmsg_end(skb, nlh);
404 	return 0;
405 
406 errout:
407 	nlmsg_cancel(skb, nlh);
408 	return -EMSGSIZE;
409 }
410 EXPORT_SYMBOL_GPL(inet_sk_diag_fill);
411 
412 static int inet_twsk_diag_fill(struct sock *sk,
413 			       struct sk_buff *skb,
414 			       struct netlink_callback *cb,
415 			       u16 nlmsg_flags)
416 {
417 	struct inet_timewait_sock *tw = inet_twsk(sk);
418 	struct inet_diag_msg *r;
419 	struct nlmsghdr *nlh;
420 	long tmo;
421 
422 	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
423 			cb->nlh->nlmsg_seq, cb->nlh->nlmsg_type,
424 			sizeof(*r), nlmsg_flags);
425 	if (!nlh)
426 		return -EMSGSIZE;
427 
428 	r = nlmsg_data(nlh);
429 	BUG_ON(tw->tw_state != TCP_TIME_WAIT);
430 
431 	inet_diag_msg_common_fill(r, sk);
432 	r->idiag_retrans      = 0;
433 
434 	r->idiag_state	      = tw->tw_substate;
435 	r->idiag_timer	      = 3;
436 	tmo = tw->tw_timer.expires - jiffies;
437 	r->idiag_expires      = jiffies_delta_to_msecs(tmo);
438 	r->idiag_rqueue	      = 0;
439 	r->idiag_wqueue	      = 0;
440 	r->idiag_uid	      = 0;
441 	r->idiag_inode	      = 0;
442 
443 	nlmsg_end(skb, nlh);
444 	return 0;
445 }
446 
447 static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb,
448 			      struct netlink_callback *cb,
449 			      u16 nlmsg_flags, bool net_admin)
450 {
451 	struct request_sock *reqsk = inet_reqsk(sk);
452 	struct inet_diag_msg *r;
453 	struct nlmsghdr *nlh;
454 	long tmo;
455 
456 	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
457 			cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
458 	if (!nlh)
459 		return -EMSGSIZE;
460 
461 	r = nlmsg_data(nlh);
462 	inet_diag_msg_common_fill(r, sk);
463 	r->idiag_state = TCP_SYN_RECV;
464 	r->idiag_timer = 1;
465 	r->idiag_retrans = reqsk->num_retrans;
466 
467 	BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) !=
468 		     offsetof(struct sock, sk_cookie));
469 
470 	tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies;
471 	r->idiag_expires = jiffies_delta_to_msecs(tmo);
472 	r->idiag_rqueue	= 0;
473 	r->idiag_wqueue	= 0;
474 	r->idiag_uid	= 0;
475 	r->idiag_inode	= 0;
476 
477 	if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
478 				     inet_rsk(reqsk)->ir_mark))
479 		return -EMSGSIZE;
480 
481 	nlmsg_end(skb, nlh);
482 	return 0;
483 }
484 
485 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
486 			struct netlink_callback *cb,
487 			const struct inet_diag_req_v2 *r,
488 			u16 nlmsg_flags, bool net_admin)
489 {
490 	if (sk->sk_state == TCP_TIME_WAIT)
491 		return inet_twsk_diag_fill(sk, skb, cb, nlmsg_flags);
492 
493 	if (sk->sk_state == TCP_NEW_SYN_RECV)
494 		return inet_req_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
495 
496 	return inet_sk_diag_fill(sk, inet_csk(sk), skb, cb, r, nlmsg_flags,
497 				 net_admin);
498 }
499 
500 struct sock *inet_diag_find_one_icsk(struct net *net,
501 				     struct inet_hashinfo *hashinfo,
502 				     const struct inet_diag_req_v2 *req)
503 {
504 	struct sock *sk;
505 
506 	rcu_read_lock();
507 	if (req->sdiag_family == AF_INET)
508 		sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0],
509 				 req->id.idiag_dport, req->id.idiag_src[0],
510 				 req->id.idiag_sport, req->id.idiag_if);
511 #if IS_ENABLED(CONFIG_IPV6)
512 	else if (req->sdiag_family == AF_INET6) {
513 		if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) &&
514 		    ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src))
515 			sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3],
516 					 req->id.idiag_dport, req->id.idiag_src[3],
517 					 req->id.idiag_sport, req->id.idiag_if);
518 		else
519 			sk = inet6_lookup(net, hashinfo, NULL, 0,
520 					  (struct in6_addr *)req->id.idiag_dst,
521 					  req->id.idiag_dport,
522 					  (struct in6_addr *)req->id.idiag_src,
523 					  req->id.idiag_sport,
524 					  req->id.idiag_if);
525 	}
526 #endif
527 	else {
528 		rcu_read_unlock();
529 		return ERR_PTR(-EINVAL);
530 	}
531 	rcu_read_unlock();
532 	if (!sk)
533 		return ERR_PTR(-ENOENT);
534 
535 	if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) {
536 		sock_gen_put(sk);
537 		return ERR_PTR(-ENOENT);
538 	}
539 
540 	return sk;
541 }
542 EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk);
543 
544 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo,
545 			    struct netlink_callback *cb,
546 			    const struct inet_diag_req_v2 *req)
547 {
548 	struct sk_buff *in_skb = cb->skb;
549 	bool net_admin = netlink_net_capable(in_skb, CAP_NET_ADMIN);
550 	struct net *net = sock_net(in_skb->sk);
551 	struct sk_buff *rep;
552 	struct sock *sk;
553 	int err;
554 
555 	sk = inet_diag_find_one_icsk(net, hashinfo, req);
556 	if (IS_ERR(sk))
557 		return PTR_ERR(sk);
558 
559 	rep = nlmsg_new(inet_sk_attr_size(sk, req, net_admin), GFP_KERNEL);
560 	if (!rep) {
561 		err = -ENOMEM;
562 		goto out;
563 	}
564 
565 	err = sk_diag_fill(sk, rep, cb, req, 0, net_admin);
566 	if (err < 0) {
567 		WARN_ON(err == -EMSGSIZE);
568 		nlmsg_free(rep);
569 		goto out;
570 	}
571 	err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid,
572 			      MSG_DONTWAIT);
573 	if (err > 0)
574 		err = 0;
575 
576 out:
577 	if (sk)
578 		sock_gen_put(sk);
579 
580 	return err;
581 }
582 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
583 
584 static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb,
585 			       const struct nlmsghdr *nlh,
586 			       int hdrlen,
587 			       const struct inet_diag_req_v2 *req)
588 {
589 	const struct inet_diag_handler *handler;
590 	struct inet_diag_dump_data dump_data;
591 	int err, protocol;
592 
593 	memset(&dump_data, 0, sizeof(dump_data));
594 	inet_diag_parse_attrs(nlh, hdrlen, dump_data.req_nlas);
595 	protocol = inet_diag_get_protocol(req, &dump_data);
596 
597 	handler = inet_diag_lock_handler(protocol);
598 	if (IS_ERR(handler)) {
599 		err = PTR_ERR(handler);
600 	} else if (cmd == SOCK_DIAG_BY_FAMILY) {
601 		struct netlink_callback cb = {
602 			.nlh = nlh,
603 			.skb = in_skb,
604 			.data = &dump_data,
605 		};
606 		err = handler->dump_one(&cb, req);
607 	} else if (cmd == SOCK_DESTROY && handler->destroy) {
608 		err = handler->destroy(in_skb, req);
609 	} else {
610 		err = -EOPNOTSUPP;
611 	}
612 	inet_diag_unlock_handler(handler);
613 
614 	return err;
615 }
616 
617 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
618 {
619 	int words = bits >> 5;
620 
621 	bits &= 0x1f;
622 
623 	if (words) {
624 		if (memcmp(a1, a2, words << 2))
625 			return 0;
626 	}
627 	if (bits) {
628 		__be32 w1, w2;
629 		__be32 mask;
630 
631 		w1 = a1[words];
632 		w2 = a2[words];
633 
634 		mask = htonl((0xffffffff) << (32 - bits));
635 
636 		if ((w1 ^ w2) & mask)
637 			return 0;
638 	}
639 
640 	return 1;
641 }
642 
643 static int inet_diag_bc_run(const struct nlattr *_bc,
644 			    const struct inet_diag_entry *entry)
645 {
646 	const void *bc = nla_data(_bc);
647 	int len = nla_len(_bc);
648 
649 	while (len > 0) {
650 		int yes = 1;
651 		const struct inet_diag_bc_op *op = bc;
652 
653 		switch (op->code) {
654 		case INET_DIAG_BC_NOP:
655 			break;
656 		case INET_DIAG_BC_JMP:
657 			yes = 0;
658 			break;
659 		case INET_DIAG_BC_S_EQ:
660 			yes = entry->sport == op[1].no;
661 			break;
662 		case INET_DIAG_BC_S_GE:
663 			yes = entry->sport >= op[1].no;
664 			break;
665 		case INET_DIAG_BC_S_LE:
666 			yes = entry->sport <= op[1].no;
667 			break;
668 		case INET_DIAG_BC_D_EQ:
669 			yes = entry->dport == op[1].no;
670 			break;
671 		case INET_DIAG_BC_D_GE:
672 			yes = entry->dport >= op[1].no;
673 			break;
674 		case INET_DIAG_BC_D_LE:
675 			yes = entry->dport <= op[1].no;
676 			break;
677 		case INET_DIAG_BC_AUTO:
678 			yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
679 			break;
680 		case INET_DIAG_BC_S_COND:
681 		case INET_DIAG_BC_D_COND: {
682 			const struct inet_diag_hostcond *cond;
683 			const __be32 *addr;
684 
685 			cond = (const struct inet_diag_hostcond *)(op + 1);
686 			if (cond->port != -1 &&
687 			    cond->port != (op->code == INET_DIAG_BC_S_COND ?
688 					     entry->sport : entry->dport)) {
689 				yes = 0;
690 				break;
691 			}
692 
693 			if (op->code == INET_DIAG_BC_S_COND)
694 				addr = entry->saddr;
695 			else
696 				addr = entry->daddr;
697 
698 			if (cond->family != AF_UNSPEC &&
699 			    cond->family != entry->family) {
700 				if (entry->family == AF_INET6 &&
701 				    cond->family == AF_INET) {
702 					if (addr[0] == 0 && addr[1] == 0 &&
703 					    addr[2] == htonl(0xffff) &&
704 					    bitstring_match(addr + 3,
705 							    cond->addr,
706 							    cond->prefix_len))
707 						break;
708 				}
709 				yes = 0;
710 				break;
711 			}
712 
713 			if (cond->prefix_len == 0)
714 				break;
715 			if (bitstring_match(addr, cond->addr,
716 					    cond->prefix_len))
717 				break;
718 			yes = 0;
719 			break;
720 		}
721 		case INET_DIAG_BC_DEV_COND: {
722 			u32 ifindex;
723 
724 			ifindex = *((const u32 *)(op + 1));
725 			if (ifindex != entry->ifindex)
726 				yes = 0;
727 			break;
728 		}
729 		case INET_DIAG_BC_MARK_COND: {
730 			struct inet_diag_markcond *cond;
731 
732 			cond = (struct inet_diag_markcond *)(op + 1);
733 			if ((entry->mark & cond->mask) != cond->mark)
734 				yes = 0;
735 			break;
736 		}
737 #ifdef CONFIG_SOCK_CGROUP_DATA
738 		case INET_DIAG_BC_CGROUP_COND: {
739 			u64 cgroup_id;
740 
741 			cgroup_id = get_unaligned((const u64 *)(op + 1));
742 			if (cgroup_id != entry->cgroup_id)
743 				yes = 0;
744 			break;
745 		}
746 #endif
747 		}
748 
749 		if (yes) {
750 			len -= op->yes;
751 			bc += op->yes;
752 		} else {
753 			len -= op->no;
754 			bc += op->no;
755 		}
756 	}
757 	return len == 0;
758 }
759 
760 /* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV)
761  */
762 static void entry_fill_addrs(struct inet_diag_entry *entry,
763 			     const struct sock *sk)
764 {
765 #if IS_ENABLED(CONFIG_IPV6)
766 	if (sk->sk_family == AF_INET6) {
767 		entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32;
768 		entry->daddr = sk->sk_v6_daddr.s6_addr32;
769 	} else
770 #endif
771 	{
772 		entry->saddr = &sk->sk_rcv_saddr;
773 		entry->daddr = &sk->sk_daddr;
774 	}
775 }
776 
777 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
778 {
779 	struct inet_sock *inet = inet_sk(sk);
780 	struct inet_diag_entry entry;
781 
782 	if (!bc)
783 		return 1;
784 
785 	entry.family = sk->sk_family;
786 	entry_fill_addrs(&entry, sk);
787 	entry.sport = inet->inet_num;
788 	entry.dport = ntohs(inet->inet_dport);
789 	entry.ifindex = sk->sk_bound_dev_if;
790 	entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0;
791 	if (sk_fullsock(sk))
792 		entry.mark = sk->sk_mark;
793 	else if (sk->sk_state == TCP_NEW_SYN_RECV)
794 		entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark;
795 	else
796 		entry.mark = 0;
797 #ifdef CONFIG_SOCK_CGROUP_DATA
798 	entry.cgroup_id = sk_fullsock(sk) ?
799 		cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)) : 0;
800 #endif
801 
802 	return inet_diag_bc_run(bc, &entry);
803 }
804 EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
805 
806 static int valid_cc(const void *bc, int len, int cc)
807 {
808 	while (len >= 0) {
809 		const struct inet_diag_bc_op *op = bc;
810 
811 		if (cc > len)
812 			return 0;
813 		if (cc == len)
814 			return 1;
815 		if (op->yes < 4 || op->yes & 3)
816 			return 0;
817 		len -= op->yes;
818 		bc  += op->yes;
819 	}
820 	return 0;
821 }
822 
823 /* data is u32 ifindex */
824 static bool valid_devcond(const struct inet_diag_bc_op *op, int len,
825 			  int *min_len)
826 {
827 	/* Check ifindex space. */
828 	*min_len += sizeof(u32);
829 	if (len < *min_len)
830 		return false;
831 
832 	return true;
833 }
834 /* Validate an inet_diag_hostcond. */
835 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
836 			   int *min_len)
837 {
838 	struct inet_diag_hostcond *cond;
839 	int addr_len;
840 
841 	/* Check hostcond space. */
842 	*min_len += sizeof(struct inet_diag_hostcond);
843 	if (len < *min_len)
844 		return false;
845 	cond = (struct inet_diag_hostcond *)(op + 1);
846 
847 	/* Check address family and address length. */
848 	switch (cond->family) {
849 	case AF_UNSPEC:
850 		addr_len = 0;
851 		break;
852 	case AF_INET:
853 		addr_len = sizeof(struct in_addr);
854 		break;
855 	case AF_INET6:
856 		addr_len = sizeof(struct in6_addr);
857 		break;
858 	default:
859 		return false;
860 	}
861 	*min_len += addr_len;
862 	if (len < *min_len)
863 		return false;
864 
865 	/* Check prefix length (in bits) vs address length (in bytes). */
866 	if (cond->prefix_len > 8 * addr_len)
867 		return false;
868 
869 	return true;
870 }
871 
872 /* Validate a port comparison operator. */
873 static bool valid_port_comparison(const struct inet_diag_bc_op *op,
874 				  int len, int *min_len)
875 {
876 	/* Port comparisons put the port in a follow-on inet_diag_bc_op. */
877 	*min_len += sizeof(struct inet_diag_bc_op);
878 	if (len < *min_len)
879 		return false;
880 	return true;
881 }
882 
883 static bool valid_markcond(const struct inet_diag_bc_op *op, int len,
884 			   int *min_len)
885 {
886 	*min_len += sizeof(struct inet_diag_markcond);
887 	return len >= *min_len;
888 }
889 
890 #ifdef CONFIG_SOCK_CGROUP_DATA
891 static bool valid_cgroupcond(const struct inet_diag_bc_op *op, int len,
892 			     int *min_len)
893 {
894 	*min_len += sizeof(u64);
895 	return len >= *min_len;
896 }
897 #endif
898 
899 static int inet_diag_bc_audit(const struct nlattr *attr,
900 			      const struct sk_buff *skb)
901 {
902 	bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN);
903 	const void *bytecode, *bc;
904 	int bytecode_len, len;
905 
906 	if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op))
907 		return -EINVAL;
908 
909 	bytecode = bc = nla_data(attr);
910 	len = bytecode_len = nla_len(attr);
911 
912 	while (len > 0) {
913 		int min_len = sizeof(struct inet_diag_bc_op);
914 		const struct inet_diag_bc_op *op = bc;
915 
916 		switch (op->code) {
917 		case INET_DIAG_BC_S_COND:
918 		case INET_DIAG_BC_D_COND:
919 			if (!valid_hostcond(bc, len, &min_len))
920 				return -EINVAL;
921 			break;
922 		case INET_DIAG_BC_DEV_COND:
923 			if (!valid_devcond(bc, len, &min_len))
924 				return -EINVAL;
925 			break;
926 		case INET_DIAG_BC_S_EQ:
927 		case INET_DIAG_BC_S_GE:
928 		case INET_DIAG_BC_S_LE:
929 		case INET_DIAG_BC_D_EQ:
930 		case INET_DIAG_BC_D_GE:
931 		case INET_DIAG_BC_D_LE:
932 			if (!valid_port_comparison(bc, len, &min_len))
933 				return -EINVAL;
934 			break;
935 		case INET_DIAG_BC_MARK_COND:
936 			if (!net_admin)
937 				return -EPERM;
938 			if (!valid_markcond(bc, len, &min_len))
939 				return -EINVAL;
940 			break;
941 #ifdef CONFIG_SOCK_CGROUP_DATA
942 		case INET_DIAG_BC_CGROUP_COND:
943 			if (!valid_cgroupcond(bc, len, &min_len))
944 				return -EINVAL;
945 			break;
946 #endif
947 		case INET_DIAG_BC_AUTO:
948 		case INET_DIAG_BC_JMP:
949 		case INET_DIAG_BC_NOP:
950 			break;
951 		default:
952 			return -EINVAL;
953 		}
954 
955 		if (op->code != INET_DIAG_BC_NOP) {
956 			if (op->no < min_len || op->no > len + 4 || op->no & 3)
957 				return -EINVAL;
958 			if (op->no < len &&
959 			    !valid_cc(bytecode, bytecode_len, len - op->no))
960 				return -EINVAL;
961 		}
962 
963 		if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
964 			return -EINVAL;
965 		bc  += op->yes;
966 		len -= op->yes;
967 	}
968 	return len == 0 ? 0 : -EINVAL;
969 }
970 
971 static void twsk_build_assert(void)
972 {
973 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) !=
974 		     offsetof(struct sock, sk_family));
975 
976 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) !=
977 		     offsetof(struct inet_sock, inet_num));
978 
979 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) !=
980 		     offsetof(struct inet_sock, inet_dport));
981 
982 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) !=
983 		     offsetof(struct inet_sock, inet_rcv_saddr));
984 
985 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) !=
986 		     offsetof(struct inet_sock, inet_daddr));
987 
988 #if IS_ENABLED(CONFIG_IPV6)
989 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) !=
990 		     offsetof(struct sock, sk_v6_rcv_saddr));
991 
992 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) !=
993 		     offsetof(struct sock, sk_v6_daddr));
994 #endif
995 }
996 
997 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
998 			 struct netlink_callback *cb,
999 			 const struct inet_diag_req_v2 *r)
1000 {
1001 	bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN);
1002 	struct inet_diag_dump_data *cb_data = cb->data;
1003 	struct net *net = sock_net(skb->sk);
1004 	u32 idiag_states = r->idiag_states;
1005 	int i, num, s_i, s_num;
1006 	struct nlattr *bc;
1007 	struct sock *sk;
1008 
1009 	bc = cb_data->inet_diag_nla_bc;
1010 	if (idiag_states & TCPF_SYN_RECV)
1011 		idiag_states |= TCPF_NEW_SYN_RECV;
1012 	s_i = cb->args[1];
1013 	s_num = num = cb->args[2];
1014 
1015 	if (cb->args[0] == 0) {
1016 		if (!(idiag_states & TCPF_LISTEN) || r->id.idiag_dport)
1017 			goto skip_listen_ht;
1018 
1019 		for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
1020 			struct inet_listen_hashbucket *ilb;
1021 			struct hlist_nulls_node *node;
1022 
1023 			num = 0;
1024 			ilb = &hashinfo->listening_hash[i];
1025 			spin_lock(&ilb->lock);
1026 			sk_nulls_for_each(sk, node, &ilb->nulls_head) {
1027 				struct inet_sock *inet = inet_sk(sk);
1028 
1029 				if (!net_eq(sock_net(sk), net))
1030 					continue;
1031 
1032 				if (num < s_num) {
1033 					num++;
1034 					continue;
1035 				}
1036 
1037 				if (r->sdiag_family != AF_UNSPEC &&
1038 				    sk->sk_family != r->sdiag_family)
1039 					goto next_listen;
1040 
1041 				if (r->id.idiag_sport != inet->inet_sport &&
1042 				    r->id.idiag_sport)
1043 					goto next_listen;
1044 
1045 				if (!inet_diag_bc_sk(bc, sk))
1046 					goto next_listen;
1047 
1048 				if (inet_sk_diag_fill(sk, inet_csk(sk), skb,
1049 						      cb, r, NLM_F_MULTI,
1050 						      net_admin) < 0) {
1051 					spin_unlock(&ilb->lock);
1052 					goto done;
1053 				}
1054 
1055 next_listen:
1056 				++num;
1057 			}
1058 			spin_unlock(&ilb->lock);
1059 
1060 			s_num = 0;
1061 		}
1062 skip_listen_ht:
1063 		cb->args[0] = 1;
1064 		s_i = num = s_num = 0;
1065 	}
1066 
1067 	if (!(idiag_states & ~TCPF_LISTEN))
1068 		goto out;
1069 
1070 #define SKARR_SZ 16
1071 	for (i = s_i; i <= hashinfo->ehash_mask; i++) {
1072 		struct inet_ehash_bucket *head = &hashinfo->ehash[i];
1073 		spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
1074 		struct hlist_nulls_node *node;
1075 		struct sock *sk_arr[SKARR_SZ];
1076 		int num_arr[SKARR_SZ];
1077 		int idx, accum, res;
1078 
1079 		if (hlist_nulls_empty(&head->chain))
1080 			continue;
1081 
1082 		if (i > s_i)
1083 			s_num = 0;
1084 
1085 next_chunk:
1086 		num = 0;
1087 		accum = 0;
1088 		spin_lock_bh(lock);
1089 		sk_nulls_for_each(sk, node, &head->chain) {
1090 			int state;
1091 
1092 			if (!net_eq(sock_net(sk), net))
1093 				continue;
1094 			if (num < s_num)
1095 				goto next_normal;
1096 			state = (sk->sk_state == TCP_TIME_WAIT) ?
1097 				inet_twsk(sk)->tw_substate : sk->sk_state;
1098 			if (!(idiag_states & (1 << state)))
1099 				goto next_normal;
1100 			if (r->sdiag_family != AF_UNSPEC &&
1101 			    sk->sk_family != r->sdiag_family)
1102 				goto next_normal;
1103 			if (r->id.idiag_sport != htons(sk->sk_num) &&
1104 			    r->id.idiag_sport)
1105 				goto next_normal;
1106 			if (r->id.idiag_dport != sk->sk_dport &&
1107 			    r->id.idiag_dport)
1108 				goto next_normal;
1109 			twsk_build_assert();
1110 
1111 			if (!inet_diag_bc_sk(bc, sk))
1112 				goto next_normal;
1113 
1114 			if (!refcount_inc_not_zero(&sk->sk_refcnt))
1115 				goto next_normal;
1116 
1117 			num_arr[accum] = num;
1118 			sk_arr[accum] = sk;
1119 			if (++accum == SKARR_SZ)
1120 				break;
1121 next_normal:
1122 			++num;
1123 		}
1124 		spin_unlock_bh(lock);
1125 		res = 0;
1126 		for (idx = 0; idx < accum; idx++) {
1127 			if (res >= 0) {
1128 				res = sk_diag_fill(sk_arr[idx], skb, cb, r,
1129 						   NLM_F_MULTI, net_admin);
1130 				if (res < 0)
1131 					num = num_arr[idx];
1132 			}
1133 			sock_gen_put(sk_arr[idx]);
1134 		}
1135 		if (res < 0)
1136 			break;
1137 		cond_resched();
1138 		if (accum == SKARR_SZ) {
1139 			s_num = num + 1;
1140 			goto next_chunk;
1141 		}
1142 	}
1143 
1144 done:
1145 	cb->args[1] = i;
1146 	cb->args[2] = num;
1147 out:
1148 	;
1149 }
1150 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
1151 
1152 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
1153 			    const struct inet_diag_req_v2 *r)
1154 {
1155 	struct inet_diag_dump_data *cb_data = cb->data;
1156 	const struct inet_diag_handler *handler;
1157 	u32 prev_min_dump_alloc;
1158 	int protocol, err = 0;
1159 
1160 	protocol = inet_diag_get_protocol(r, cb_data);
1161 
1162 again:
1163 	prev_min_dump_alloc = cb->min_dump_alloc;
1164 	handler = inet_diag_lock_handler(protocol);
1165 	if (!IS_ERR(handler))
1166 		handler->dump(skb, cb, r);
1167 	else
1168 		err = PTR_ERR(handler);
1169 	inet_diag_unlock_handler(handler);
1170 
1171 	/* The skb is not large enough to fit one sk info and
1172 	 * inet_sk_diag_fill() has requested for a larger skb.
1173 	 */
1174 	if (!skb->len && cb->min_dump_alloc > prev_min_dump_alloc) {
1175 		err = pskb_expand_head(skb, 0, cb->min_dump_alloc, GFP_KERNEL);
1176 		if (!err)
1177 			goto again;
1178 	}
1179 
1180 	return err ? : skb->len;
1181 }
1182 
1183 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
1184 {
1185 	return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh));
1186 }
1187 
1188 static int __inet_diag_dump_start(struct netlink_callback *cb, int hdrlen)
1189 {
1190 	const struct nlmsghdr *nlh = cb->nlh;
1191 	struct inet_diag_dump_data *cb_data;
1192 	struct sk_buff *skb = cb->skb;
1193 	struct nlattr *nla;
1194 	int err;
1195 
1196 	cb_data = kzalloc(sizeof(*cb_data), GFP_KERNEL);
1197 	if (!cb_data)
1198 		return -ENOMEM;
1199 
1200 	inet_diag_parse_attrs(nlh, hdrlen, cb_data->req_nlas);
1201 
1202 	nla = cb_data->inet_diag_nla_bc;
1203 	if (nla) {
1204 		err = inet_diag_bc_audit(nla, skb);
1205 		if (err) {
1206 			kfree(cb_data);
1207 			return err;
1208 		}
1209 	}
1210 
1211 	nla = cb_data->inet_diag_nla_bpf_stgs;
1212 	if (nla) {
1213 		struct bpf_sk_storage_diag *bpf_stg_diag;
1214 
1215 		bpf_stg_diag = bpf_sk_storage_diag_alloc(nla);
1216 		if (IS_ERR(bpf_stg_diag)) {
1217 			kfree(cb_data);
1218 			return PTR_ERR(bpf_stg_diag);
1219 		}
1220 		cb_data->bpf_stg_diag = bpf_stg_diag;
1221 	}
1222 
1223 	cb->data = cb_data;
1224 	return 0;
1225 }
1226 
1227 static int inet_diag_dump_start(struct netlink_callback *cb)
1228 {
1229 	return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req_v2));
1230 }
1231 
1232 static int inet_diag_dump_start_compat(struct netlink_callback *cb)
1233 {
1234 	return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req));
1235 }
1236 
1237 static int inet_diag_dump_done(struct netlink_callback *cb)
1238 {
1239 	struct inet_diag_dump_data *cb_data = cb->data;
1240 
1241 	bpf_sk_storage_diag_free(cb_data->bpf_stg_diag);
1242 	kfree(cb->data);
1243 
1244 	return 0;
1245 }
1246 
1247 static int inet_diag_type2proto(int type)
1248 {
1249 	switch (type) {
1250 	case TCPDIAG_GETSOCK:
1251 		return IPPROTO_TCP;
1252 	case DCCPDIAG_GETSOCK:
1253 		return IPPROTO_DCCP;
1254 	default:
1255 		return 0;
1256 	}
1257 }
1258 
1259 static int inet_diag_dump_compat(struct sk_buff *skb,
1260 				 struct netlink_callback *cb)
1261 {
1262 	struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1263 	struct inet_diag_req_v2 req;
1264 
1265 	req.sdiag_family = AF_UNSPEC; /* compatibility */
1266 	req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1267 	req.idiag_ext = rc->idiag_ext;
1268 	req.idiag_states = rc->idiag_states;
1269 	req.id = rc->id;
1270 
1271 	return __inet_diag_dump(skb, cb, &req);
1272 }
1273 
1274 static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1275 				      const struct nlmsghdr *nlh)
1276 {
1277 	struct inet_diag_req *rc = nlmsg_data(nlh);
1278 	struct inet_diag_req_v2 req;
1279 
1280 	req.sdiag_family = rc->idiag_family;
1281 	req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1282 	req.idiag_ext = rc->idiag_ext;
1283 	req.idiag_states = rc->idiag_states;
1284 	req.id = rc->id;
1285 
1286 	return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh,
1287 				   sizeof(struct inet_diag_req), &req);
1288 }
1289 
1290 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1291 {
1292 	int hdrlen = sizeof(struct inet_diag_req);
1293 	struct net *net = sock_net(skb->sk);
1294 
1295 	if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1296 	    nlmsg_len(nlh) < hdrlen)
1297 		return -EINVAL;
1298 
1299 	if (nlh->nlmsg_flags & NLM_F_DUMP) {
1300 		struct netlink_dump_control c = {
1301 			.start = inet_diag_dump_start_compat,
1302 			.done = inet_diag_dump_done,
1303 			.dump = inet_diag_dump_compat,
1304 		};
1305 		return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1306 	}
1307 
1308 	return inet_diag_get_exact_compat(skb, nlh);
1309 }
1310 
1311 static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h)
1312 {
1313 	int hdrlen = sizeof(struct inet_diag_req_v2);
1314 	struct net *net = sock_net(skb->sk);
1315 
1316 	if (nlmsg_len(h) < hdrlen)
1317 		return -EINVAL;
1318 
1319 	if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY &&
1320 	    h->nlmsg_flags & NLM_F_DUMP) {
1321 		struct netlink_dump_control c = {
1322 			.start = inet_diag_dump_start,
1323 			.done = inet_diag_dump_done,
1324 			.dump = inet_diag_dump,
1325 		};
1326 		return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1327 	}
1328 
1329 	return inet_diag_cmd_exact(h->nlmsg_type, skb, h, hdrlen,
1330 				   nlmsg_data(h));
1331 }
1332 
1333 static
1334 int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk)
1335 {
1336 	const struct inet_diag_handler *handler;
1337 	struct nlmsghdr *nlh;
1338 	struct nlattr *attr;
1339 	struct inet_diag_msg *r;
1340 	void *info = NULL;
1341 	int err = 0;
1342 
1343 	nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0);
1344 	if (!nlh)
1345 		return -ENOMEM;
1346 
1347 	r = nlmsg_data(nlh);
1348 	memset(r, 0, sizeof(*r));
1349 	inet_diag_msg_common_fill(r, sk);
1350 	if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM)
1351 		r->id.idiag_sport = inet_sk(sk)->inet_sport;
1352 	r->idiag_state = sk->sk_state;
1353 
1354 	if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) {
1355 		nlmsg_cancel(skb, nlh);
1356 		return err;
1357 	}
1358 
1359 	handler = inet_diag_lock_handler(sk->sk_protocol);
1360 	if (IS_ERR(handler)) {
1361 		inet_diag_unlock_handler(handler);
1362 		nlmsg_cancel(skb, nlh);
1363 		return PTR_ERR(handler);
1364 	}
1365 
1366 	attr = handler->idiag_info_size
1367 		? nla_reserve_64bit(skb, INET_DIAG_INFO,
1368 				    handler->idiag_info_size,
1369 				    INET_DIAG_PAD)
1370 		: NULL;
1371 	if (attr)
1372 		info = nla_data(attr);
1373 
1374 	handler->idiag_get_info(sk, r, info);
1375 	inet_diag_unlock_handler(handler);
1376 
1377 	nlmsg_end(skb, nlh);
1378 	return 0;
1379 }
1380 
1381 static const struct sock_diag_handler inet_diag_handler = {
1382 	.family = AF_INET,
1383 	.dump = inet_diag_handler_cmd,
1384 	.get_info = inet_diag_handler_get_info,
1385 	.destroy = inet_diag_handler_cmd,
1386 };
1387 
1388 static const struct sock_diag_handler inet6_diag_handler = {
1389 	.family = AF_INET6,
1390 	.dump = inet_diag_handler_cmd,
1391 	.get_info = inet_diag_handler_get_info,
1392 	.destroy = inet_diag_handler_cmd,
1393 };
1394 
1395 int inet_diag_register(const struct inet_diag_handler *h)
1396 {
1397 	const __u16 type = h->idiag_type;
1398 	int err = -EINVAL;
1399 
1400 	if (type >= IPPROTO_MAX)
1401 		goto out;
1402 
1403 	mutex_lock(&inet_diag_table_mutex);
1404 	err = -EEXIST;
1405 	if (!inet_diag_table[type]) {
1406 		inet_diag_table[type] = h;
1407 		err = 0;
1408 	}
1409 	mutex_unlock(&inet_diag_table_mutex);
1410 out:
1411 	return err;
1412 }
1413 EXPORT_SYMBOL_GPL(inet_diag_register);
1414 
1415 void inet_diag_unregister(const struct inet_diag_handler *h)
1416 {
1417 	const __u16 type = h->idiag_type;
1418 
1419 	if (type >= IPPROTO_MAX)
1420 		return;
1421 
1422 	mutex_lock(&inet_diag_table_mutex);
1423 	inet_diag_table[type] = NULL;
1424 	mutex_unlock(&inet_diag_table_mutex);
1425 }
1426 EXPORT_SYMBOL_GPL(inet_diag_unregister);
1427 
1428 static int __init inet_diag_init(void)
1429 {
1430 	const int inet_diag_table_size = (IPPROTO_MAX *
1431 					  sizeof(struct inet_diag_handler *));
1432 	int err = -ENOMEM;
1433 
1434 	inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1435 	if (!inet_diag_table)
1436 		goto out;
1437 
1438 	err = sock_diag_register(&inet_diag_handler);
1439 	if (err)
1440 		goto out_free_nl;
1441 
1442 	err = sock_diag_register(&inet6_diag_handler);
1443 	if (err)
1444 		goto out_free_inet;
1445 
1446 	sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1447 out:
1448 	return err;
1449 
1450 out_free_inet:
1451 	sock_diag_unregister(&inet_diag_handler);
1452 out_free_nl:
1453 	kfree(inet_diag_table);
1454 	goto out;
1455 }
1456 
1457 static void __exit inet_diag_exit(void)
1458 {
1459 	sock_diag_unregister(&inet6_diag_handler);
1460 	sock_diag_unregister(&inet_diag_handler);
1461 	sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1462 	kfree(inet_diag_table);
1463 }
1464 
1465 module_init(inet_diag_init);
1466 module_exit(inet_diag_exit);
1467 MODULE_LICENSE("GPL");
1468 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1469 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);
1470