xref: /openbmc/linux/net/ipv4/inet_diag.c (revision 8ffdff6a) 
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 		nlmsg_cancel(skb, nlh);
484 		return -EMSGSIZE;
485 	}
486 
487 	nlmsg_end(skb, nlh);
488 	return 0;
489 }
490 
491 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
492 			struct netlink_callback *cb,
493 			const struct inet_diag_req_v2 *r,
494 			u16 nlmsg_flags, bool net_admin)
495 {
496 	if (sk->sk_state == TCP_TIME_WAIT)
497 		return inet_twsk_diag_fill(sk, skb, cb, nlmsg_flags);
498 
499 	if (sk->sk_state == TCP_NEW_SYN_RECV)
500 		return inet_req_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
501 
502 	return inet_sk_diag_fill(sk, inet_csk(sk), skb, cb, r, nlmsg_flags,
503 				 net_admin);
504 }
505 
506 struct sock *inet_diag_find_one_icsk(struct net *net,
507 				     struct inet_hashinfo *hashinfo,
508 				     const struct inet_diag_req_v2 *req)
509 {
510 	struct sock *sk;
511 
512 	rcu_read_lock();
513 	if (req->sdiag_family == AF_INET)
514 		sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0],
515 				 req->id.idiag_dport, req->id.idiag_src[0],
516 				 req->id.idiag_sport, req->id.idiag_if);
517 #if IS_ENABLED(CONFIG_IPV6)
518 	else if (req->sdiag_family == AF_INET6) {
519 		if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) &&
520 		    ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src))
521 			sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3],
522 					 req->id.idiag_dport, req->id.idiag_src[3],
523 					 req->id.idiag_sport, req->id.idiag_if);
524 		else
525 			sk = inet6_lookup(net, hashinfo, NULL, 0,
526 					  (struct in6_addr *)req->id.idiag_dst,
527 					  req->id.idiag_dport,
528 					  (struct in6_addr *)req->id.idiag_src,
529 					  req->id.idiag_sport,
530 					  req->id.idiag_if);
531 	}
532 #endif
533 	else {
534 		rcu_read_unlock();
535 		return ERR_PTR(-EINVAL);
536 	}
537 	rcu_read_unlock();
538 	if (!sk)
539 		return ERR_PTR(-ENOENT);
540 
541 	if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) {
542 		sock_gen_put(sk);
543 		return ERR_PTR(-ENOENT);
544 	}
545 
546 	return sk;
547 }
548 EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk);
549 
550 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo,
551 			    struct netlink_callback *cb,
552 			    const struct inet_diag_req_v2 *req)
553 {
554 	struct sk_buff *in_skb = cb->skb;
555 	bool net_admin = netlink_net_capable(in_skb, CAP_NET_ADMIN);
556 	struct net *net = sock_net(in_skb->sk);
557 	struct sk_buff *rep;
558 	struct sock *sk;
559 	int err;
560 
561 	sk = inet_diag_find_one_icsk(net, hashinfo, req);
562 	if (IS_ERR(sk))
563 		return PTR_ERR(sk);
564 
565 	rep = nlmsg_new(inet_sk_attr_size(sk, req, net_admin), GFP_KERNEL);
566 	if (!rep) {
567 		err = -ENOMEM;
568 		goto out;
569 	}
570 
571 	err = sk_diag_fill(sk, rep, cb, req, 0, net_admin);
572 	if (err < 0) {
573 		WARN_ON(err == -EMSGSIZE);
574 		nlmsg_free(rep);
575 		goto out;
576 	}
577 	err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid,
578 			      MSG_DONTWAIT);
579 	if (err > 0)
580 		err = 0;
581 
582 out:
583 	if (sk)
584 		sock_gen_put(sk);
585 
586 	return err;
587 }
588 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
589 
590 static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb,
591 			       const struct nlmsghdr *nlh,
592 			       int hdrlen,
593 			       const struct inet_diag_req_v2 *req)
594 {
595 	const struct inet_diag_handler *handler;
596 	struct inet_diag_dump_data dump_data;
597 	int err, protocol;
598 
599 	memset(&dump_data, 0, sizeof(dump_data));
600 	err = inet_diag_parse_attrs(nlh, hdrlen, dump_data.req_nlas);
601 	if (err)
602 		return err;
603 
604 	protocol = inet_diag_get_protocol(req, &dump_data);
605 
606 	handler = inet_diag_lock_handler(protocol);
607 	if (IS_ERR(handler)) {
608 		err = PTR_ERR(handler);
609 	} else if (cmd == SOCK_DIAG_BY_FAMILY) {
610 		struct netlink_callback cb = {
611 			.nlh = nlh,
612 			.skb = in_skb,
613 			.data = &dump_data,
614 		};
615 		err = handler->dump_one(&cb, req);
616 	} else if (cmd == SOCK_DESTROY && handler->destroy) {
617 		err = handler->destroy(in_skb, req);
618 	} else {
619 		err = -EOPNOTSUPP;
620 	}
621 	inet_diag_unlock_handler(handler);
622 
623 	return err;
624 }
625 
626 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
627 {
628 	int words = bits >> 5;
629 
630 	bits &= 0x1f;
631 
632 	if (words) {
633 		if (memcmp(a1, a2, words << 2))
634 			return 0;
635 	}
636 	if (bits) {
637 		__be32 w1, w2;
638 		__be32 mask;
639 
640 		w1 = a1[words];
641 		w2 = a2[words];
642 
643 		mask = htonl((0xffffffff) << (32 - bits));
644 
645 		if ((w1 ^ w2) & mask)
646 			return 0;
647 	}
648 
649 	return 1;
650 }
651 
652 static int inet_diag_bc_run(const struct nlattr *_bc,
653 			    const struct inet_diag_entry *entry)
654 {
655 	const void *bc = nla_data(_bc);
656 	int len = nla_len(_bc);
657 
658 	while (len > 0) {
659 		int yes = 1;
660 		const struct inet_diag_bc_op *op = bc;
661 
662 		switch (op->code) {
663 		case INET_DIAG_BC_NOP:
664 			break;
665 		case INET_DIAG_BC_JMP:
666 			yes = 0;
667 			break;
668 		case INET_DIAG_BC_S_EQ:
669 			yes = entry->sport == op[1].no;
670 			break;
671 		case INET_DIAG_BC_S_GE:
672 			yes = entry->sport >= op[1].no;
673 			break;
674 		case INET_DIAG_BC_S_LE:
675 			yes = entry->sport <= op[1].no;
676 			break;
677 		case INET_DIAG_BC_D_EQ:
678 			yes = entry->dport == op[1].no;
679 			break;
680 		case INET_DIAG_BC_D_GE:
681 			yes = entry->dport >= op[1].no;
682 			break;
683 		case INET_DIAG_BC_D_LE:
684 			yes = entry->dport <= op[1].no;
685 			break;
686 		case INET_DIAG_BC_AUTO:
687 			yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
688 			break;
689 		case INET_DIAG_BC_S_COND:
690 		case INET_DIAG_BC_D_COND: {
691 			const struct inet_diag_hostcond *cond;
692 			const __be32 *addr;
693 
694 			cond = (const struct inet_diag_hostcond *)(op + 1);
695 			if (cond->port != -1 &&
696 			    cond->port != (op->code == INET_DIAG_BC_S_COND ?
697 					     entry->sport : entry->dport)) {
698 				yes = 0;
699 				break;
700 			}
701 
702 			if (op->code == INET_DIAG_BC_S_COND)
703 				addr = entry->saddr;
704 			else
705 				addr = entry->daddr;
706 
707 			if (cond->family != AF_UNSPEC &&
708 			    cond->family != entry->family) {
709 				if (entry->family == AF_INET6 &&
710 				    cond->family == AF_INET) {
711 					if (addr[0] == 0 && addr[1] == 0 &&
712 					    addr[2] == htonl(0xffff) &&
713 					    bitstring_match(addr + 3,
714 							    cond->addr,
715 							    cond->prefix_len))
716 						break;
717 				}
718 				yes = 0;
719 				break;
720 			}
721 
722 			if (cond->prefix_len == 0)
723 				break;
724 			if (bitstring_match(addr, cond->addr,
725 					    cond->prefix_len))
726 				break;
727 			yes = 0;
728 			break;
729 		}
730 		case INET_DIAG_BC_DEV_COND: {
731 			u32 ifindex;
732 
733 			ifindex = *((const u32 *)(op + 1));
734 			if (ifindex != entry->ifindex)
735 				yes = 0;
736 			break;
737 		}
738 		case INET_DIAG_BC_MARK_COND: {
739 			struct inet_diag_markcond *cond;
740 
741 			cond = (struct inet_diag_markcond *)(op + 1);
742 			if ((entry->mark & cond->mask) != cond->mark)
743 				yes = 0;
744 			break;
745 		}
746 #ifdef CONFIG_SOCK_CGROUP_DATA
747 		case INET_DIAG_BC_CGROUP_COND: {
748 			u64 cgroup_id;
749 
750 			cgroup_id = get_unaligned((const u64 *)(op + 1));
751 			if (cgroup_id != entry->cgroup_id)
752 				yes = 0;
753 			break;
754 		}
755 #endif
756 		}
757 
758 		if (yes) {
759 			len -= op->yes;
760 			bc += op->yes;
761 		} else {
762 			len -= op->no;
763 			bc += op->no;
764 		}
765 	}
766 	return len == 0;
767 }
768 
769 /* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV)
770  */
771 static void entry_fill_addrs(struct inet_diag_entry *entry,
772 			     const struct sock *sk)
773 {
774 #if IS_ENABLED(CONFIG_IPV6)
775 	if (sk->sk_family == AF_INET6) {
776 		entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32;
777 		entry->daddr = sk->sk_v6_daddr.s6_addr32;
778 	} else
779 #endif
780 	{
781 		entry->saddr = &sk->sk_rcv_saddr;
782 		entry->daddr = &sk->sk_daddr;
783 	}
784 }
785 
786 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
787 {
788 	struct inet_sock *inet = inet_sk(sk);
789 	struct inet_diag_entry entry;
790 
791 	if (!bc)
792 		return 1;
793 
794 	entry.family = sk->sk_family;
795 	entry_fill_addrs(&entry, sk);
796 	entry.sport = inet->inet_num;
797 	entry.dport = ntohs(inet->inet_dport);
798 	entry.ifindex = sk->sk_bound_dev_if;
799 	entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0;
800 	if (sk_fullsock(sk))
801 		entry.mark = sk->sk_mark;
802 	else if (sk->sk_state == TCP_NEW_SYN_RECV)
803 		entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark;
804 	else
805 		entry.mark = 0;
806 #ifdef CONFIG_SOCK_CGROUP_DATA
807 	entry.cgroup_id = sk_fullsock(sk) ?
808 		cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)) : 0;
809 #endif
810 
811 	return inet_diag_bc_run(bc, &entry);
812 }
813 EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
814 
815 static int valid_cc(const void *bc, int len, int cc)
816 {
817 	while (len >= 0) {
818 		const struct inet_diag_bc_op *op = bc;
819 
820 		if (cc > len)
821 			return 0;
822 		if (cc == len)
823 			return 1;
824 		if (op->yes < 4 || op->yes & 3)
825 			return 0;
826 		len -= op->yes;
827 		bc  += op->yes;
828 	}
829 	return 0;
830 }
831 
832 /* data is u32 ifindex */
833 static bool valid_devcond(const struct inet_diag_bc_op *op, int len,
834 			  int *min_len)
835 {
836 	/* Check ifindex space. */
837 	*min_len += sizeof(u32);
838 	if (len < *min_len)
839 		return false;
840 
841 	return true;
842 }
843 /* Validate an inet_diag_hostcond. */
844 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
845 			   int *min_len)
846 {
847 	struct inet_diag_hostcond *cond;
848 	int addr_len;
849 
850 	/* Check hostcond space. */
851 	*min_len += sizeof(struct inet_diag_hostcond);
852 	if (len < *min_len)
853 		return false;
854 	cond = (struct inet_diag_hostcond *)(op + 1);
855 
856 	/* Check address family and address length. */
857 	switch (cond->family) {
858 	case AF_UNSPEC:
859 		addr_len = 0;
860 		break;
861 	case AF_INET:
862 		addr_len = sizeof(struct in_addr);
863 		break;
864 	case AF_INET6:
865 		addr_len = sizeof(struct in6_addr);
866 		break;
867 	default:
868 		return false;
869 	}
870 	*min_len += addr_len;
871 	if (len < *min_len)
872 		return false;
873 
874 	/* Check prefix length (in bits) vs address length (in bytes). */
875 	if (cond->prefix_len > 8 * addr_len)
876 		return false;
877 
878 	return true;
879 }
880 
881 /* Validate a port comparison operator. */
882 static bool valid_port_comparison(const struct inet_diag_bc_op *op,
883 				  int len, int *min_len)
884 {
885 	/* Port comparisons put the port in a follow-on inet_diag_bc_op. */
886 	*min_len += sizeof(struct inet_diag_bc_op);
887 	if (len < *min_len)
888 		return false;
889 	return true;
890 }
891 
892 static bool valid_markcond(const struct inet_diag_bc_op *op, int len,
893 			   int *min_len)
894 {
895 	*min_len += sizeof(struct inet_diag_markcond);
896 	return len >= *min_len;
897 }
898 
899 #ifdef CONFIG_SOCK_CGROUP_DATA
900 static bool valid_cgroupcond(const struct inet_diag_bc_op *op, int len,
901 			     int *min_len)
902 {
903 	*min_len += sizeof(u64);
904 	return len >= *min_len;
905 }
906 #endif
907 
908 static int inet_diag_bc_audit(const struct nlattr *attr,
909 			      const struct sk_buff *skb)
910 {
911 	bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN);
912 	const void *bytecode, *bc;
913 	int bytecode_len, len;
914 
915 	if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op))
916 		return -EINVAL;
917 
918 	bytecode = bc = nla_data(attr);
919 	len = bytecode_len = nla_len(attr);
920 
921 	while (len > 0) {
922 		int min_len = sizeof(struct inet_diag_bc_op);
923 		const struct inet_diag_bc_op *op = bc;
924 
925 		switch (op->code) {
926 		case INET_DIAG_BC_S_COND:
927 		case INET_DIAG_BC_D_COND:
928 			if (!valid_hostcond(bc, len, &min_len))
929 				return -EINVAL;
930 			break;
931 		case INET_DIAG_BC_DEV_COND:
932 			if (!valid_devcond(bc, len, &min_len))
933 				return -EINVAL;
934 			break;
935 		case INET_DIAG_BC_S_EQ:
936 		case INET_DIAG_BC_S_GE:
937 		case INET_DIAG_BC_S_LE:
938 		case INET_DIAG_BC_D_EQ:
939 		case INET_DIAG_BC_D_GE:
940 		case INET_DIAG_BC_D_LE:
941 			if (!valid_port_comparison(bc, len, &min_len))
942 				return -EINVAL;
943 			break;
944 		case INET_DIAG_BC_MARK_COND:
945 			if (!net_admin)
946 				return -EPERM;
947 			if (!valid_markcond(bc, len, &min_len))
948 				return -EINVAL;
949 			break;
950 #ifdef CONFIG_SOCK_CGROUP_DATA
951 		case INET_DIAG_BC_CGROUP_COND:
952 			if (!valid_cgroupcond(bc, len, &min_len))
953 				return -EINVAL;
954 			break;
955 #endif
956 		case INET_DIAG_BC_AUTO:
957 		case INET_DIAG_BC_JMP:
958 		case INET_DIAG_BC_NOP:
959 			break;
960 		default:
961 			return -EINVAL;
962 		}
963 
964 		if (op->code != INET_DIAG_BC_NOP) {
965 			if (op->no < min_len || op->no > len + 4 || op->no & 3)
966 				return -EINVAL;
967 			if (op->no < len &&
968 			    !valid_cc(bytecode, bytecode_len, len - op->no))
969 				return -EINVAL;
970 		}
971 
972 		if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
973 			return -EINVAL;
974 		bc  += op->yes;
975 		len -= op->yes;
976 	}
977 	return len == 0 ? 0 : -EINVAL;
978 }
979 
980 static void twsk_build_assert(void)
981 {
982 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) !=
983 		     offsetof(struct sock, sk_family));
984 
985 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) !=
986 		     offsetof(struct inet_sock, inet_num));
987 
988 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) !=
989 		     offsetof(struct inet_sock, inet_dport));
990 
991 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) !=
992 		     offsetof(struct inet_sock, inet_rcv_saddr));
993 
994 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) !=
995 		     offsetof(struct inet_sock, inet_daddr));
996 
997 #if IS_ENABLED(CONFIG_IPV6)
998 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) !=
999 		     offsetof(struct sock, sk_v6_rcv_saddr));
1000 
1001 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) !=
1002 		     offsetof(struct sock, sk_v6_daddr));
1003 #endif
1004 }
1005 
1006 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
1007 			 struct netlink_callback *cb,
1008 			 const struct inet_diag_req_v2 *r)
1009 {
1010 	bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN);
1011 	struct inet_diag_dump_data *cb_data = cb->data;
1012 	struct net *net = sock_net(skb->sk);
1013 	u32 idiag_states = r->idiag_states;
1014 	int i, num, s_i, s_num;
1015 	struct nlattr *bc;
1016 	struct sock *sk;
1017 
1018 	bc = cb_data->inet_diag_nla_bc;
1019 	if (idiag_states & TCPF_SYN_RECV)
1020 		idiag_states |= TCPF_NEW_SYN_RECV;
1021 	s_i = cb->args[1];
1022 	s_num = num = cb->args[2];
1023 
1024 	if (cb->args[0] == 0) {
1025 		if (!(idiag_states & TCPF_LISTEN) || r->id.idiag_dport)
1026 			goto skip_listen_ht;
1027 
1028 		for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
1029 			struct inet_listen_hashbucket *ilb;
1030 			struct hlist_nulls_node *node;
1031 
1032 			num = 0;
1033 			ilb = &hashinfo->listening_hash[i];
1034 			spin_lock(&ilb->lock);
1035 			sk_nulls_for_each(sk, node, &ilb->nulls_head) {
1036 				struct inet_sock *inet = inet_sk(sk);
1037 
1038 				if (!net_eq(sock_net(sk), net))
1039 					continue;
1040 
1041 				if (num < s_num) {
1042 					num++;
1043 					continue;
1044 				}
1045 
1046 				if (r->sdiag_family != AF_UNSPEC &&
1047 				    sk->sk_family != r->sdiag_family)
1048 					goto next_listen;
1049 
1050 				if (r->id.idiag_sport != inet->inet_sport &&
1051 				    r->id.idiag_sport)
1052 					goto next_listen;
1053 
1054 				if (!inet_diag_bc_sk(bc, sk))
1055 					goto next_listen;
1056 
1057 				if (inet_sk_diag_fill(sk, inet_csk(sk), skb,
1058 						      cb, r, NLM_F_MULTI,
1059 						      net_admin) < 0) {
1060 					spin_unlock(&ilb->lock);
1061 					goto done;
1062 				}
1063 
1064 next_listen:
1065 				++num;
1066 			}
1067 			spin_unlock(&ilb->lock);
1068 
1069 			s_num = 0;
1070 		}
1071 skip_listen_ht:
1072 		cb->args[0] = 1;
1073 		s_i = num = s_num = 0;
1074 	}
1075 
1076 	if (!(idiag_states & ~TCPF_LISTEN))
1077 		goto out;
1078 
1079 #define SKARR_SZ 16
1080 	for (i = s_i; i <= hashinfo->ehash_mask; i++) {
1081 		struct inet_ehash_bucket *head = &hashinfo->ehash[i];
1082 		spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
1083 		struct hlist_nulls_node *node;
1084 		struct sock *sk_arr[SKARR_SZ];
1085 		int num_arr[SKARR_SZ];
1086 		int idx, accum, res;
1087 
1088 		if (hlist_nulls_empty(&head->chain))
1089 			continue;
1090 
1091 		if (i > s_i)
1092 			s_num = 0;
1093 
1094 next_chunk:
1095 		num = 0;
1096 		accum = 0;
1097 		spin_lock_bh(lock);
1098 		sk_nulls_for_each(sk, node, &head->chain) {
1099 			int state;
1100 
1101 			if (!net_eq(sock_net(sk), net))
1102 				continue;
1103 			if (num < s_num)
1104 				goto next_normal;
1105 			state = (sk->sk_state == TCP_TIME_WAIT) ?
1106 				inet_twsk(sk)->tw_substate : sk->sk_state;
1107 			if (!(idiag_states & (1 << state)))
1108 				goto next_normal;
1109 			if (r->sdiag_family != AF_UNSPEC &&
1110 			    sk->sk_family != r->sdiag_family)
1111 				goto next_normal;
1112 			if (r->id.idiag_sport != htons(sk->sk_num) &&
1113 			    r->id.idiag_sport)
1114 				goto next_normal;
1115 			if (r->id.idiag_dport != sk->sk_dport &&
1116 			    r->id.idiag_dport)
1117 				goto next_normal;
1118 			twsk_build_assert();
1119 
1120 			if (!inet_diag_bc_sk(bc, sk))
1121 				goto next_normal;
1122 
1123 			if (!refcount_inc_not_zero(&sk->sk_refcnt))
1124 				goto next_normal;
1125 
1126 			num_arr[accum] = num;
1127 			sk_arr[accum] = sk;
1128 			if (++accum == SKARR_SZ)
1129 				break;
1130 next_normal:
1131 			++num;
1132 		}
1133 		spin_unlock_bh(lock);
1134 		res = 0;
1135 		for (idx = 0; idx < accum; idx++) {
1136 			if (res >= 0) {
1137 				res = sk_diag_fill(sk_arr[idx], skb, cb, r,
1138 						   NLM_F_MULTI, net_admin);
1139 				if (res < 0)
1140 					num = num_arr[idx];
1141 			}
1142 			sock_gen_put(sk_arr[idx]);
1143 		}
1144 		if (res < 0)
1145 			break;
1146 		cond_resched();
1147 		if (accum == SKARR_SZ) {
1148 			s_num = num + 1;
1149 			goto next_chunk;
1150 		}
1151 	}
1152 
1153 done:
1154 	cb->args[1] = i;
1155 	cb->args[2] = num;
1156 out:
1157 	;
1158 }
1159 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
1160 
1161 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
1162 			    const struct inet_diag_req_v2 *r)
1163 {
1164 	struct inet_diag_dump_data *cb_data = cb->data;
1165 	const struct inet_diag_handler *handler;
1166 	u32 prev_min_dump_alloc;
1167 	int protocol, err = 0;
1168 
1169 	protocol = inet_diag_get_protocol(r, cb_data);
1170 
1171 again:
1172 	prev_min_dump_alloc = cb->min_dump_alloc;
1173 	handler = inet_diag_lock_handler(protocol);
1174 	if (!IS_ERR(handler))
1175 		handler->dump(skb, cb, r);
1176 	else
1177 		err = PTR_ERR(handler);
1178 	inet_diag_unlock_handler(handler);
1179 
1180 	/* The skb is not large enough to fit one sk info and
1181 	 * inet_sk_diag_fill() has requested for a larger skb.
1182 	 */
1183 	if (!skb->len && cb->min_dump_alloc > prev_min_dump_alloc) {
1184 		err = pskb_expand_head(skb, 0, cb->min_dump_alloc, GFP_KERNEL);
1185 		if (!err)
1186 			goto again;
1187 	}
1188 
1189 	return err ? : skb->len;
1190 }
1191 
1192 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
1193 {
1194 	return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh));
1195 }
1196 
1197 static int __inet_diag_dump_start(struct netlink_callback *cb, int hdrlen)
1198 {
1199 	const struct nlmsghdr *nlh = cb->nlh;
1200 	struct inet_diag_dump_data *cb_data;
1201 	struct sk_buff *skb = cb->skb;
1202 	struct nlattr *nla;
1203 	int err;
1204 
1205 	cb_data = kzalloc(sizeof(*cb_data), GFP_KERNEL);
1206 	if (!cb_data)
1207 		return -ENOMEM;
1208 
1209 	err = inet_diag_parse_attrs(nlh, hdrlen, cb_data->req_nlas);
1210 	if (err) {
1211 		kfree(cb_data);
1212 		return err;
1213 	}
1214 	nla = cb_data->inet_diag_nla_bc;
1215 	if (nla) {
1216 		err = inet_diag_bc_audit(nla, skb);
1217 		if (err) {
1218 			kfree(cb_data);
1219 			return err;
1220 		}
1221 	}
1222 
1223 	nla = cb_data->inet_diag_nla_bpf_stgs;
1224 	if (nla) {
1225 		struct bpf_sk_storage_diag *bpf_stg_diag;
1226 
1227 		bpf_stg_diag = bpf_sk_storage_diag_alloc(nla);
1228 		if (IS_ERR(bpf_stg_diag)) {
1229 			kfree(cb_data);
1230 			return PTR_ERR(bpf_stg_diag);
1231 		}
1232 		cb_data->bpf_stg_diag = bpf_stg_diag;
1233 	}
1234 
1235 	cb->data = cb_data;
1236 	return 0;
1237 }
1238 
1239 static int inet_diag_dump_start(struct netlink_callback *cb)
1240 {
1241 	return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req_v2));
1242 }
1243 
1244 static int inet_diag_dump_start_compat(struct netlink_callback *cb)
1245 {
1246 	return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req));
1247 }
1248 
1249 static int inet_diag_dump_done(struct netlink_callback *cb)
1250 {
1251 	struct inet_diag_dump_data *cb_data = cb->data;
1252 
1253 	bpf_sk_storage_diag_free(cb_data->bpf_stg_diag);
1254 	kfree(cb->data);
1255 
1256 	return 0;
1257 }
1258 
1259 static int inet_diag_type2proto(int type)
1260 {
1261 	switch (type) {
1262 	case TCPDIAG_GETSOCK:
1263 		return IPPROTO_TCP;
1264 	case DCCPDIAG_GETSOCK:
1265 		return IPPROTO_DCCP;
1266 	default:
1267 		return 0;
1268 	}
1269 }
1270 
1271 static int inet_diag_dump_compat(struct sk_buff *skb,
1272 				 struct netlink_callback *cb)
1273 {
1274 	struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1275 	struct inet_diag_req_v2 req;
1276 
1277 	req.sdiag_family = AF_UNSPEC; /* compatibility */
1278 	req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1279 	req.idiag_ext = rc->idiag_ext;
1280 	req.idiag_states = rc->idiag_states;
1281 	req.id = rc->id;
1282 
1283 	return __inet_diag_dump(skb, cb, &req);
1284 }
1285 
1286 static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1287 				      const struct nlmsghdr *nlh)
1288 {
1289 	struct inet_diag_req *rc = nlmsg_data(nlh);
1290 	struct inet_diag_req_v2 req;
1291 
1292 	req.sdiag_family = rc->idiag_family;
1293 	req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1294 	req.idiag_ext = rc->idiag_ext;
1295 	req.idiag_states = rc->idiag_states;
1296 	req.id = rc->id;
1297 
1298 	return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh,
1299 				   sizeof(struct inet_diag_req), &req);
1300 }
1301 
1302 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1303 {
1304 	int hdrlen = sizeof(struct inet_diag_req);
1305 	struct net *net = sock_net(skb->sk);
1306 
1307 	if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1308 	    nlmsg_len(nlh) < hdrlen)
1309 		return -EINVAL;
1310 
1311 	if (nlh->nlmsg_flags & NLM_F_DUMP) {
1312 		struct netlink_dump_control c = {
1313 			.start = inet_diag_dump_start_compat,
1314 			.done = inet_diag_dump_done,
1315 			.dump = inet_diag_dump_compat,
1316 		};
1317 		return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1318 	}
1319 
1320 	return inet_diag_get_exact_compat(skb, nlh);
1321 }
1322 
1323 static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h)
1324 {
1325 	int hdrlen = sizeof(struct inet_diag_req_v2);
1326 	struct net *net = sock_net(skb->sk);
1327 
1328 	if (nlmsg_len(h) < hdrlen)
1329 		return -EINVAL;
1330 
1331 	if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY &&
1332 	    h->nlmsg_flags & NLM_F_DUMP) {
1333 		struct netlink_dump_control c = {
1334 			.start = inet_diag_dump_start,
1335 			.done = inet_diag_dump_done,
1336 			.dump = inet_diag_dump,
1337 		};
1338 		return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1339 	}
1340 
1341 	return inet_diag_cmd_exact(h->nlmsg_type, skb, h, hdrlen,
1342 				   nlmsg_data(h));
1343 }
1344 
1345 static
1346 int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk)
1347 {
1348 	const struct inet_diag_handler *handler;
1349 	struct nlmsghdr *nlh;
1350 	struct nlattr *attr;
1351 	struct inet_diag_msg *r;
1352 	void *info = NULL;
1353 	int err = 0;
1354 
1355 	nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0);
1356 	if (!nlh)
1357 		return -ENOMEM;
1358 
1359 	r = nlmsg_data(nlh);
1360 	memset(r, 0, sizeof(*r));
1361 	inet_diag_msg_common_fill(r, sk);
1362 	if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM)
1363 		r->id.idiag_sport = inet_sk(sk)->inet_sport;
1364 	r->idiag_state = sk->sk_state;
1365 
1366 	if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) {
1367 		nlmsg_cancel(skb, nlh);
1368 		return err;
1369 	}
1370 
1371 	handler = inet_diag_lock_handler(sk->sk_protocol);
1372 	if (IS_ERR(handler)) {
1373 		inet_diag_unlock_handler(handler);
1374 		nlmsg_cancel(skb, nlh);
1375 		return PTR_ERR(handler);
1376 	}
1377 
1378 	attr = handler->idiag_info_size
1379 		? nla_reserve_64bit(skb, INET_DIAG_INFO,
1380 				    handler->idiag_info_size,
1381 				    INET_DIAG_PAD)
1382 		: NULL;
1383 	if (attr)
1384 		info = nla_data(attr);
1385 
1386 	handler->idiag_get_info(sk, r, info);
1387 	inet_diag_unlock_handler(handler);
1388 
1389 	nlmsg_end(skb, nlh);
1390 	return 0;
1391 }
1392 
1393 static const struct sock_diag_handler inet_diag_handler = {
1394 	.family = AF_INET,
1395 	.dump = inet_diag_handler_cmd,
1396 	.get_info = inet_diag_handler_get_info,
1397 	.destroy = inet_diag_handler_cmd,
1398 };
1399 
1400 static const struct sock_diag_handler inet6_diag_handler = {
1401 	.family = AF_INET6,
1402 	.dump = inet_diag_handler_cmd,
1403 	.get_info = inet_diag_handler_get_info,
1404 	.destroy = inet_diag_handler_cmd,
1405 };
1406 
1407 int inet_diag_register(const struct inet_diag_handler *h)
1408 {
1409 	const __u16 type = h->idiag_type;
1410 	int err = -EINVAL;
1411 
1412 	if (type >= IPPROTO_MAX)
1413 		goto out;
1414 
1415 	mutex_lock(&inet_diag_table_mutex);
1416 	err = -EEXIST;
1417 	if (!inet_diag_table[type]) {
1418 		inet_diag_table[type] = h;
1419 		err = 0;
1420 	}
1421 	mutex_unlock(&inet_diag_table_mutex);
1422 out:
1423 	return err;
1424 }
1425 EXPORT_SYMBOL_GPL(inet_diag_register);
1426 
1427 void inet_diag_unregister(const struct inet_diag_handler *h)
1428 {
1429 	const __u16 type = h->idiag_type;
1430 
1431 	if (type >= IPPROTO_MAX)
1432 		return;
1433 
1434 	mutex_lock(&inet_diag_table_mutex);
1435 	inet_diag_table[type] = NULL;
1436 	mutex_unlock(&inet_diag_table_mutex);
1437 }
1438 EXPORT_SYMBOL_GPL(inet_diag_unregister);
1439 
1440 static int __init inet_diag_init(void)
1441 {
1442 	const int inet_diag_table_size = (IPPROTO_MAX *
1443 					  sizeof(struct inet_diag_handler *));
1444 	int err = -ENOMEM;
1445 
1446 	inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1447 	if (!inet_diag_table)
1448 		goto out;
1449 
1450 	err = sock_diag_register(&inet_diag_handler);
1451 	if (err)
1452 		goto out_free_nl;
1453 
1454 	err = sock_diag_register(&inet6_diag_handler);
1455 	if (err)
1456 		goto out_free_inet;
1457 
1458 	sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1459 out:
1460 	return err;
1461 
1462 out_free_inet:
1463 	sock_diag_unregister(&inet_diag_handler);
1464 out_free_nl:
1465 	kfree(inet_diag_table);
1466 	goto out;
1467 }
1468 
1469 static void __exit inet_diag_exit(void)
1470 {
1471 	sock_diag_unregister(&inet6_diag_handler);
1472 	sock_diag_unregister(&inet_diag_handler);
1473 	sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1474 	kfree(inet_diag_table);
1475 }
1476 
1477 module_init(inet_diag_init);
1478 module_exit(inet_diag_exit);
1479 MODULE_LICENSE("GPL");
1480 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1481 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);
1482