xref: /openbmc/linux/net/ipv4/inet_diag.c (revision 7f1005dd) 
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 (!READ_ONCE(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, READ_ONCE(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_test_bit(RECVERR, sk);
186 	inet_sockopt.is_icsk	= inet_test_bit(IS_ICSK, sk);
187 	inet_sockopt.freebind	= inet_test_bit(FREEBIND, sk);
188 	inet_sockopt.hdrincl	= inet_test_bit(HDRINCL, sk);
189 	inet_sockopt.mc_loop	= inet_test_bit(MC_LOOP, sk);
190 	inet_sockopt.transparent = inet_test_bit(TRANSPARENT, sk);
191 	inet_sockopt.mc_all	= inet_test_bit(MC_ALL, sk);
192 	inet_sockopt.nodefrag	= inet_test_bit(NODEFRAG, sk);
193 	inet_sockopt.bind_address_no_port = inet_test_bit(BIND_ADDRESS_NO_PORT, sk);
194 	inet_sockopt.recverr_rfc4884 = inet_test_bit(RECVERR_RFC4884, sk);
195 	inet_sockopt.defer_connect = inet_test_bit(DEFER_CONNECT, sk);
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 	r->idiag_expires = 0;
265 
266 	if (inet_diag_msg_attrs_fill(sk, skb, r, ext,
267 				     sk_user_ns(NETLINK_CB(cb->skb).sk),
268 				     net_admin))
269 		goto errout;
270 
271 	if (ext & (1 << (INET_DIAG_MEMINFO - 1))) {
272 		struct inet_diag_meminfo minfo = {
273 			.idiag_rmem = sk_rmem_alloc_get(sk),
274 			.idiag_wmem = READ_ONCE(sk->sk_wmem_queued),
275 			.idiag_fmem = sk_forward_alloc_get(sk),
276 			.idiag_tmem = sk_wmem_alloc_get(sk),
277 		};
278 
279 		if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0)
280 			goto errout;
281 	}
282 
283 	if (ext & (1 << (INET_DIAG_SKMEMINFO - 1)))
284 		if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO))
285 			goto errout;
286 
287 	/*
288 	 * RAW sockets might have user-defined protocols assigned,
289 	 * so report the one supplied on socket creation.
290 	 */
291 	if (sk->sk_type == SOCK_RAW) {
292 		if (nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))
293 			goto errout;
294 	}
295 
296 	if (!icsk) {
297 		handler->idiag_get_info(sk, r, NULL);
298 		goto out;
299 	}
300 
301 	if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
302 	    icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
303 	    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
304 		r->idiag_timer = 1;
305 		r->idiag_retrans = icsk->icsk_retransmits;
306 		r->idiag_expires =
307 			jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
308 	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
309 		r->idiag_timer = 4;
310 		r->idiag_retrans = icsk->icsk_probes_out;
311 		r->idiag_expires =
312 			jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
313 	} else if (timer_pending(&sk->sk_timer)) {
314 		r->idiag_timer = 2;
315 		r->idiag_retrans = icsk->icsk_probes_out;
316 		r->idiag_expires =
317 			jiffies_delta_to_msecs(sk->sk_timer.expires - jiffies);
318 	}
319 
320 	if ((ext & (1 << (INET_DIAG_INFO - 1))) && handler->idiag_info_size) {
321 		attr = nla_reserve_64bit(skb, INET_DIAG_INFO,
322 					 handler->idiag_info_size,
323 					 INET_DIAG_PAD);
324 		if (!attr)
325 			goto errout;
326 
327 		info = nla_data(attr);
328 	}
329 
330 	if (ext & (1 << (INET_DIAG_CONG - 1))) {
331 		int err = 0;
332 
333 		rcu_read_lock();
334 		ca_ops = READ_ONCE(icsk->icsk_ca_ops);
335 		if (ca_ops)
336 			err = nla_put_string(skb, INET_DIAG_CONG, ca_ops->name);
337 		rcu_read_unlock();
338 		if (err < 0)
339 			goto errout;
340 	}
341 
342 	handler->idiag_get_info(sk, r, info);
343 
344 	if (ext & (1 << (INET_DIAG_INFO - 1)) && handler->idiag_get_aux)
345 		if (handler->idiag_get_aux(sk, net_admin, skb) < 0)
346 			goto errout;
347 
348 	if (sk->sk_state < TCP_TIME_WAIT) {
349 		union tcp_cc_info info;
350 		size_t sz = 0;
351 		int attr;
352 
353 		rcu_read_lock();
354 		ca_ops = READ_ONCE(icsk->icsk_ca_ops);
355 		if (ca_ops && ca_ops->get_info)
356 			sz = ca_ops->get_info(sk, ext, &attr, &info);
357 		rcu_read_unlock();
358 		if (sz && nla_put(skb, attr, sz, &info) < 0)
359 			goto errout;
360 	}
361 
362 	/* Keep it at the end for potential retry with a larger skb,
363 	 * or else do best-effort fitting, which is only done for the
364 	 * first_nlmsg.
365 	 */
366 	if (cb_data->bpf_stg_diag) {
367 		bool first_nlmsg = ((unsigned char *)nlh == skb->data);
368 		unsigned int prev_min_dump_alloc;
369 		unsigned int total_nla_size = 0;
370 		unsigned int msg_len;
371 		int err;
372 
373 		msg_len = skb_tail_pointer(skb) - (unsigned char *)nlh;
374 		err = bpf_sk_storage_diag_put(cb_data->bpf_stg_diag, sk, skb,
375 					      INET_DIAG_SK_BPF_STORAGES,
376 					      &total_nla_size);
377 
378 		if (!err)
379 			goto out;
380 
381 		total_nla_size += msg_len;
382 		prev_min_dump_alloc = cb->min_dump_alloc;
383 		if (total_nla_size > prev_min_dump_alloc)
384 			cb->min_dump_alloc = min_t(u32, total_nla_size,
385 						   MAX_DUMP_ALLOC_SIZE);
386 
387 		if (!first_nlmsg)
388 			goto errout;
389 
390 		if (cb->min_dump_alloc > prev_min_dump_alloc)
391 			/* Retry with pskb_expand_head() with
392 			 * __GFP_DIRECT_RECLAIM
393 			 */
394 			goto errout;
395 
396 		WARN_ON_ONCE(total_nla_size <= prev_min_dump_alloc);
397 
398 		/* Send what we have for this sk
399 		 * and move on to the next sk in the following
400 		 * dump()
401 		 */
402 	}
403 
404 out:
405 	nlmsg_end(skb, nlh);
406 	return 0;
407 
408 errout:
409 	nlmsg_cancel(skb, nlh);
410 	return -EMSGSIZE;
411 }
412 EXPORT_SYMBOL_GPL(inet_sk_diag_fill);
413 
414 static int inet_twsk_diag_fill(struct sock *sk,
415 			       struct sk_buff *skb,
416 			       struct netlink_callback *cb,
417 			       u16 nlmsg_flags, bool net_admin)
418 {
419 	struct inet_timewait_sock *tw = inet_twsk(sk);
420 	struct inet_diag_msg *r;
421 	struct nlmsghdr *nlh;
422 	long tmo;
423 
424 	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
425 			cb->nlh->nlmsg_seq, cb->nlh->nlmsg_type,
426 			sizeof(*r), nlmsg_flags);
427 	if (!nlh)
428 		return -EMSGSIZE;
429 
430 	r = nlmsg_data(nlh);
431 	BUG_ON(tw->tw_state != TCP_TIME_WAIT);
432 
433 	inet_diag_msg_common_fill(r, sk);
434 	r->idiag_retrans      = 0;
435 
436 	r->idiag_state	      = tw->tw_substate;
437 	r->idiag_timer	      = 3;
438 	tmo = tw->tw_timer.expires - jiffies;
439 	r->idiag_expires      = jiffies_delta_to_msecs(tmo);
440 	r->idiag_rqueue	      = 0;
441 	r->idiag_wqueue	      = 0;
442 	r->idiag_uid	      = 0;
443 	r->idiag_inode	      = 0;
444 
445 	if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
446 				     tw->tw_mark)) {
447 		nlmsg_cancel(skb, nlh);
448 		return -EMSGSIZE;
449 	}
450 
451 	nlmsg_end(skb, nlh);
452 	return 0;
453 }
454 
455 static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb,
456 			      struct netlink_callback *cb,
457 			      u16 nlmsg_flags, bool net_admin)
458 {
459 	struct request_sock *reqsk = inet_reqsk(sk);
460 	struct inet_diag_msg *r;
461 	struct nlmsghdr *nlh;
462 	long tmo;
463 
464 	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
465 			cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
466 	if (!nlh)
467 		return -EMSGSIZE;
468 
469 	r = nlmsg_data(nlh);
470 	inet_diag_msg_common_fill(r, sk);
471 	r->idiag_state = TCP_SYN_RECV;
472 	r->idiag_timer = 1;
473 	r->idiag_retrans = reqsk->num_retrans;
474 
475 	BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) !=
476 		     offsetof(struct sock, sk_cookie));
477 
478 	tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies;
479 	r->idiag_expires = jiffies_delta_to_msecs(tmo);
480 	r->idiag_rqueue	= 0;
481 	r->idiag_wqueue	= 0;
482 	r->idiag_uid	= 0;
483 	r->idiag_inode	= 0;
484 
485 	if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
486 				     inet_rsk(reqsk)->ir_mark)) {
487 		nlmsg_cancel(skb, nlh);
488 		return -EMSGSIZE;
489 	}
490 
491 	nlmsg_end(skb, nlh);
492 	return 0;
493 }
494 
495 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
496 			struct netlink_callback *cb,
497 			const struct inet_diag_req_v2 *r,
498 			u16 nlmsg_flags, bool net_admin)
499 {
500 	if (sk->sk_state == TCP_TIME_WAIT)
501 		return inet_twsk_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
502 
503 	if (sk->sk_state == TCP_NEW_SYN_RECV)
504 		return inet_req_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
505 
506 	return inet_sk_diag_fill(sk, inet_csk(sk), skb, cb, r, nlmsg_flags,
507 				 net_admin);
508 }
509 
510 struct sock *inet_diag_find_one_icsk(struct net *net,
511 				     struct inet_hashinfo *hashinfo,
512 				     const struct inet_diag_req_v2 *req)
513 {
514 	struct sock *sk;
515 
516 	rcu_read_lock();
517 	if (req->sdiag_family == AF_INET)
518 		sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0],
519 				 req->id.idiag_dport, req->id.idiag_src[0],
520 				 req->id.idiag_sport, req->id.idiag_if);
521 #if IS_ENABLED(CONFIG_IPV6)
522 	else if (req->sdiag_family == AF_INET6) {
523 		if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) &&
524 		    ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src))
525 			sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3],
526 					 req->id.idiag_dport, req->id.idiag_src[3],
527 					 req->id.idiag_sport, req->id.idiag_if);
528 		else
529 			sk = inet6_lookup(net, hashinfo, NULL, 0,
530 					  (struct in6_addr *)req->id.idiag_dst,
531 					  req->id.idiag_dport,
532 					  (struct in6_addr *)req->id.idiag_src,
533 					  req->id.idiag_sport,
534 					  req->id.idiag_if);
535 	}
536 #endif
537 	else {
538 		rcu_read_unlock();
539 		return ERR_PTR(-EINVAL);
540 	}
541 	rcu_read_unlock();
542 	if (!sk)
543 		return ERR_PTR(-ENOENT);
544 
545 	if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) {
546 		sock_gen_put(sk);
547 		return ERR_PTR(-ENOENT);
548 	}
549 
550 	return sk;
551 }
552 EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk);
553 
554 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo,
555 			    struct netlink_callback *cb,
556 			    const struct inet_diag_req_v2 *req)
557 {
558 	struct sk_buff *in_skb = cb->skb;
559 	bool net_admin = netlink_net_capable(in_skb, CAP_NET_ADMIN);
560 	struct net *net = sock_net(in_skb->sk);
561 	struct sk_buff *rep;
562 	struct sock *sk;
563 	int err;
564 
565 	sk = inet_diag_find_one_icsk(net, hashinfo, req);
566 	if (IS_ERR(sk))
567 		return PTR_ERR(sk);
568 
569 	rep = nlmsg_new(inet_sk_attr_size(sk, req, net_admin), GFP_KERNEL);
570 	if (!rep) {
571 		err = -ENOMEM;
572 		goto out;
573 	}
574 
575 	err = sk_diag_fill(sk, rep, cb, req, 0, net_admin);
576 	if (err < 0) {
577 		WARN_ON(err == -EMSGSIZE);
578 		nlmsg_free(rep);
579 		goto out;
580 	}
581 	err = nlmsg_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid);
582 
583 out:
584 	if (sk)
585 		sock_gen_put(sk);
586 
587 	return err;
588 }
589 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
590 
591 static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb,
592 			       const struct nlmsghdr *nlh,
593 			       int hdrlen,
594 			       const struct inet_diag_req_v2 *req)
595 {
596 	const struct inet_diag_handler *handler;
597 	struct inet_diag_dump_data dump_data;
598 	int err, protocol;
599 
600 	memset(&dump_data, 0, sizeof(dump_data));
601 	err = inet_diag_parse_attrs(nlh, hdrlen, dump_data.req_nlas);
602 	if (err)
603 		return err;
604 
605 	protocol = inet_diag_get_protocol(req, &dump_data);
606 
607 	handler = inet_diag_lock_handler(protocol);
608 	if (IS_ERR(handler)) {
609 		err = PTR_ERR(handler);
610 	} else if (cmd == SOCK_DIAG_BY_FAMILY) {
611 		struct netlink_callback cb = {
612 			.nlh = nlh,
613 			.skb = in_skb,
614 			.data = &dump_data,
615 		};
616 		err = handler->dump_one(&cb, req);
617 	} else if (cmd == SOCK_DESTROY && handler->destroy) {
618 		err = handler->destroy(in_skb, req);
619 	} else {
620 		err = -EOPNOTSUPP;
621 	}
622 	inet_diag_unlock_handler(handler);
623 
624 	return err;
625 }
626 
627 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
628 {
629 	int words = bits >> 5;
630 
631 	bits &= 0x1f;
632 
633 	if (words) {
634 		if (memcmp(a1, a2, words << 2))
635 			return 0;
636 	}
637 	if (bits) {
638 		__be32 w1, w2;
639 		__be32 mask;
640 
641 		w1 = a1[words];
642 		w2 = a2[words];
643 
644 		mask = htonl((0xffffffff) << (32 - bits));
645 
646 		if ((w1 ^ w2) & mask)
647 			return 0;
648 	}
649 
650 	return 1;
651 }
652 
653 static int inet_diag_bc_run(const struct nlattr *_bc,
654 			    const struct inet_diag_entry *entry)
655 {
656 	const void *bc = nla_data(_bc);
657 	int len = nla_len(_bc);
658 
659 	while (len > 0) {
660 		int yes = 1;
661 		const struct inet_diag_bc_op *op = bc;
662 
663 		switch (op->code) {
664 		case INET_DIAG_BC_NOP:
665 			break;
666 		case INET_DIAG_BC_JMP:
667 			yes = 0;
668 			break;
669 		case INET_DIAG_BC_S_EQ:
670 			yes = entry->sport == op[1].no;
671 			break;
672 		case INET_DIAG_BC_S_GE:
673 			yes = entry->sport >= op[1].no;
674 			break;
675 		case INET_DIAG_BC_S_LE:
676 			yes = entry->sport <= op[1].no;
677 			break;
678 		case INET_DIAG_BC_D_EQ:
679 			yes = entry->dport == op[1].no;
680 			break;
681 		case INET_DIAG_BC_D_GE:
682 			yes = entry->dport >= op[1].no;
683 			break;
684 		case INET_DIAG_BC_D_LE:
685 			yes = entry->dport <= op[1].no;
686 			break;
687 		case INET_DIAG_BC_AUTO:
688 			yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
689 			break;
690 		case INET_DIAG_BC_S_COND:
691 		case INET_DIAG_BC_D_COND: {
692 			const struct inet_diag_hostcond *cond;
693 			const __be32 *addr;
694 
695 			cond = (const struct inet_diag_hostcond *)(op + 1);
696 			if (cond->port != -1 &&
697 			    cond->port != (op->code == INET_DIAG_BC_S_COND ?
698 					     entry->sport : entry->dport)) {
699 				yes = 0;
700 				break;
701 			}
702 
703 			if (op->code == INET_DIAG_BC_S_COND)
704 				addr = entry->saddr;
705 			else
706 				addr = entry->daddr;
707 
708 			if (cond->family != AF_UNSPEC &&
709 			    cond->family != entry->family) {
710 				if (entry->family == AF_INET6 &&
711 				    cond->family == AF_INET) {
712 					if (addr[0] == 0 && addr[1] == 0 &&
713 					    addr[2] == htonl(0xffff) &&
714 					    bitstring_match(addr + 3,
715 							    cond->addr,
716 							    cond->prefix_len))
717 						break;
718 				}
719 				yes = 0;
720 				break;
721 			}
722 
723 			if (cond->prefix_len == 0)
724 				break;
725 			if (bitstring_match(addr, cond->addr,
726 					    cond->prefix_len))
727 				break;
728 			yes = 0;
729 			break;
730 		}
731 		case INET_DIAG_BC_DEV_COND: {
732 			u32 ifindex;
733 
734 			ifindex = *((const u32 *)(op + 1));
735 			if (ifindex != entry->ifindex)
736 				yes = 0;
737 			break;
738 		}
739 		case INET_DIAG_BC_MARK_COND: {
740 			struct inet_diag_markcond *cond;
741 
742 			cond = (struct inet_diag_markcond *)(op + 1);
743 			if ((entry->mark & cond->mask) != cond->mark)
744 				yes = 0;
745 			break;
746 		}
747 #ifdef CONFIG_SOCK_CGROUP_DATA
748 		case INET_DIAG_BC_CGROUP_COND: {
749 			u64 cgroup_id;
750 
751 			cgroup_id = get_unaligned((const u64 *)(op + 1));
752 			if (cgroup_id != entry->cgroup_id)
753 				yes = 0;
754 			break;
755 		}
756 #endif
757 		}
758 
759 		if (yes) {
760 			len -= op->yes;
761 			bc += op->yes;
762 		} else {
763 			len -= op->no;
764 			bc += op->no;
765 		}
766 	}
767 	return len == 0;
768 }
769 
770 /* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV)
771  */
772 static void entry_fill_addrs(struct inet_diag_entry *entry,
773 			     const struct sock *sk)
774 {
775 #if IS_ENABLED(CONFIG_IPV6)
776 	if (sk->sk_family == AF_INET6) {
777 		entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32;
778 		entry->daddr = sk->sk_v6_daddr.s6_addr32;
779 	} else
780 #endif
781 	{
782 		entry->saddr = &sk->sk_rcv_saddr;
783 		entry->daddr = &sk->sk_daddr;
784 	}
785 }
786 
787 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
788 {
789 	struct inet_sock *inet = inet_sk(sk);
790 	struct inet_diag_entry entry;
791 
792 	if (!bc)
793 		return 1;
794 
795 	entry.family = sk->sk_family;
796 	entry_fill_addrs(&entry, sk);
797 	entry.sport = inet->inet_num;
798 	entry.dport = ntohs(inet->inet_dport);
799 	entry.ifindex = sk->sk_bound_dev_if;
800 	entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0;
801 	if (sk_fullsock(sk))
802 		entry.mark = READ_ONCE(sk->sk_mark);
803 	else if (sk->sk_state == TCP_NEW_SYN_RECV)
804 		entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark;
805 	else if (sk->sk_state == TCP_TIME_WAIT)
806 		entry.mark = inet_twsk(sk)->tw_mark;
807 	else
808 		entry.mark = 0;
809 #ifdef CONFIG_SOCK_CGROUP_DATA
810 	entry.cgroup_id = sk_fullsock(sk) ?
811 		cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)) : 0;
812 #endif
813 
814 	return inet_diag_bc_run(bc, &entry);
815 }
816 EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
817 
818 static int valid_cc(const void *bc, int len, int cc)
819 {
820 	while (len >= 0) {
821 		const struct inet_diag_bc_op *op = bc;
822 
823 		if (cc > len)
824 			return 0;
825 		if (cc == len)
826 			return 1;
827 		if (op->yes < 4 || op->yes & 3)
828 			return 0;
829 		len -= op->yes;
830 		bc  += op->yes;
831 	}
832 	return 0;
833 }
834 
835 /* data is u32 ifindex */
836 static bool valid_devcond(const struct inet_diag_bc_op *op, int len,
837 			  int *min_len)
838 {
839 	/* Check ifindex space. */
840 	*min_len += sizeof(u32);
841 	if (len < *min_len)
842 		return false;
843 
844 	return true;
845 }
846 /* Validate an inet_diag_hostcond. */
847 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
848 			   int *min_len)
849 {
850 	struct inet_diag_hostcond *cond;
851 	int addr_len;
852 
853 	/* Check hostcond space. */
854 	*min_len += sizeof(struct inet_diag_hostcond);
855 	if (len < *min_len)
856 		return false;
857 	cond = (struct inet_diag_hostcond *)(op + 1);
858 
859 	/* Check address family and address length. */
860 	switch (cond->family) {
861 	case AF_UNSPEC:
862 		addr_len = 0;
863 		break;
864 	case AF_INET:
865 		addr_len = sizeof(struct in_addr);
866 		break;
867 	case AF_INET6:
868 		addr_len = sizeof(struct in6_addr);
869 		break;
870 	default:
871 		return false;
872 	}
873 	*min_len += addr_len;
874 	if (len < *min_len)
875 		return false;
876 
877 	/* Check prefix length (in bits) vs address length (in bytes). */
878 	if (cond->prefix_len > 8 * addr_len)
879 		return false;
880 
881 	return true;
882 }
883 
884 /* Validate a port comparison operator. */
885 static bool valid_port_comparison(const struct inet_diag_bc_op *op,
886 				  int len, int *min_len)
887 {
888 	/* Port comparisons put the port in a follow-on inet_diag_bc_op. */
889 	*min_len += sizeof(struct inet_diag_bc_op);
890 	if (len < *min_len)
891 		return false;
892 	return true;
893 }
894 
895 static bool valid_markcond(const struct inet_diag_bc_op *op, int len,
896 			   int *min_len)
897 {
898 	*min_len += sizeof(struct inet_diag_markcond);
899 	return len >= *min_len;
900 }
901 
902 #ifdef CONFIG_SOCK_CGROUP_DATA
903 static bool valid_cgroupcond(const struct inet_diag_bc_op *op, int len,
904 			     int *min_len)
905 {
906 	*min_len += sizeof(u64);
907 	return len >= *min_len;
908 }
909 #endif
910 
911 static int inet_diag_bc_audit(const struct nlattr *attr,
912 			      const struct sk_buff *skb)
913 {
914 	bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN);
915 	const void *bytecode, *bc;
916 	int bytecode_len, len;
917 
918 	if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op))
919 		return -EINVAL;
920 
921 	bytecode = bc = nla_data(attr);
922 	len = bytecode_len = nla_len(attr);
923 
924 	while (len > 0) {
925 		int min_len = sizeof(struct inet_diag_bc_op);
926 		const struct inet_diag_bc_op *op = bc;
927 
928 		switch (op->code) {
929 		case INET_DIAG_BC_S_COND:
930 		case INET_DIAG_BC_D_COND:
931 			if (!valid_hostcond(bc, len, &min_len))
932 				return -EINVAL;
933 			break;
934 		case INET_DIAG_BC_DEV_COND:
935 			if (!valid_devcond(bc, len, &min_len))
936 				return -EINVAL;
937 			break;
938 		case INET_DIAG_BC_S_EQ:
939 		case INET_DIAG_BC_S_GE:
940 		case INET_DIAG_BC_S_LE:
941 		case INET_DIAG_BC_D_EQ:
942 		case INET_DIAG_BC_D_GE:
943 		case INET_DIAG_BC_D_LE:
944 			if (!valid_port_comparison(bc, len, &min_len))
945 				return -EINVAL;
946 			break;
947 		case INET_DIAG_BC_MARK_COND:
948 			if (!net_admin)
949 				return -EPERM;
950 			if (!valid_markcond(bc, len, &min_len))
951 				return -EINVAL;
952 			break;
953 #ifdef CONFIG_SOCK_CGROUP_DATA
954 		case INET_DIAG_BC_CGROUP_COND:
955 			if (!valid_cgroupcond(bc, len, &min_len))
956 				return -EINVAL;
957 			break;
958 #endif
959 		case INET_DIAG_BC_AUTO:
960 		case INET_DIAG_BC_JMP:
961 		case INET_DIAG_BC_NOP:
962 			break;
963 		default:
964 			return -EINVAL;
965 		}
966 
967 		if (op->code != INET_DIAG_BC_NOP) {
968 			if (op->no < min_len || op->no > len + 4 || op->no & 3)
969 				return -EINVAL;
970 			if (op->no < len &&
971 			    !valid_cc(bytecode, bytecode_len, len - op->no))
972 				return -EINVAL;
973 		}
974 
975 		if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
976 			return -EINVAL;
977 		bc  += op->yes;
978 		len -= op->yes;
979 	}
980 	return len == 0 ? 0 : -EINVAL;
981 }
982 
983 static void twsk_build_assert(void)
984 {
985 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) !=
986 		     offsetof(struct sock, sk_family));
987 
988 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) !=
989 		     offsetof(struct inet_sock, inet_num));
990 
991 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) !=
992 		     offsetof(struct inet_sock, inet_dport));
993 
994 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) !=
995 		     offsetof(struct inet_sock, inet_rcv_saddr));
996 
997 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) !=
998 		     offsetof(struct inet_sock, inet_daddr));
999 
1000 #if IS_ENABLED(CONFIG_IPV6)
1001 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) !=
1002 		     offsetof(struct sock, sk_v6_rcv_saddr));
1003 
1004 	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) !=
1005 		     offsetof(struct sock, sk_v6_daddr));
1006 #endif
1007 }
1008 
1009 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
1010 			 struct netlink_callback *cb,
1011 			 const struct inet_diag_req_v2 *r)
1012 {
1013 	bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN);
1014 	struct inet_diag_dump_data *cb_data = cb->data;
1015 	struct net *net = sock_net(skb->sk);
1016 	u32 idiag_states = r->idiag_states;
1017 	int i, num, s_i, s_num;
1018 	struct nlattr *bc;
1019 	struct sock *sk;
1020 
1021 	bc = cb_data->inet_diag_nla_bc;
1022 	if (idiag_states & TCPF_SYN_RECV)
1023 		idiag_states |= TCPF_NEW_SYN_RECV;
1024 	s_i = cb->args[1];
1025 	s_num = num = cb->args[2];
1026 
1027 	if (cb->args[0] == 0) {
1028 		if (!(idiag_states & TCPF_LISTEN) || r->id.idiag_dport)
1029 			goto skip_listen_ht;
1030 
1031 		for (i = s_i; i <= hashinfo->lhash2_mask; i++) {
1032 			struct inet_listen_hashbucket *ilb;
1033 			struct hlist_nulls_node *node;
1034 
1035 			num = 0;
1036 			ilb = &hashinfo->lhash2[i];
1037 
1038 			spin_lock(&ilb->lock);
1039 			sk_nulls_for_each(sk, node, &ilb->nulls_head) {
1040 				struct inet_sock *inet = inet_sk(sk);
1041 
1042 				if (!net_eq(sock_net(sk), net))
1043 					continue;
1044 
1045 				if (num < s_num) {
1046 					num++;
1047 					continue;
1048 				}
1049 
1050 				if (r->sdiag_family != AF_UNSPEC &&
1051 				    sk->sk_family != r->sdiag_family)
1052 					goto next_listen;
1053 
1054 				if (r->id.idiag_sport != inet->inet_sport &&
1055 				    r->id.idiag_sport)
1056 					goto next_listen;
1057 
1058 				if (!inet_diag_bc_sk(bc, sk))
1059 					goto next_listen;
1060 
1061 				if (inet_sk_diag_fill(sk, inet_csk(sk), skb,
1062 						      cb, r, NLM_F_MULTI,
1063 						      net_admin) < 0) {
1064 					spin_unlock(&ilb->lock);
1065 					goto done;
1066 				}
1067 
1068 next_listen:
1069 				++num;
1070 			}
1071 			spin_unlock(&ilb->lock);
1072 
1073 			s_num = 0;
1074 		}
1075 skip_listen_ht:
1076 		cb->args[0] = 1;
1077 		s_i = num = s_num = 0;
1078 	}
1079 
1080 	if (!(idiag_states & ~TCPF_LISTEN))
1081 		goto out;
1082 
1083 #define SKARR_SZ 16
1084 	for (i = s_i; i <= hashinfo->ehash_mask; i++) {
1085 		struct inet_ehash_bucket *head = &hashinfo->ehash[i];
1086 		spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
1087 		struct hlist_nulls_node *node;
1088 		struct sock *sk_arr[SKARR_SZ];
1089 		int num_arr[SKARR_SZ];
1090 		int idx, accum, res;
1091 
1092 		if (hlist_nulls_empty(&head->chain))
1093 			continue;
1094 
1095 		if (i > s_i)
1096 			s_num = 0;
1097 
1098 next_chunk:
1099 		num = 0;
1100 		accum = 0;
1101 		spin_lock_bh(lock);
1102 		sk_nulls_for_each(sk, node, &head->chain) {
1103 			int state;
1104 
1105 			if (!net_eq(sock_net(sk), net))
1106 				continue;
1107 			if (num < s_num)
1108 				goto next_normal;
1109 			state = (sk->sk_state == TCP_TIME_WAIT) ?
1110 				inet_twsk(sk)->tw_substate : sk->sk_state;
1111 			if (!(idiag_states & (1 << state)))
1112 				goto next_normal;
1113 			if (r->sdiag_family != AF_UNSPEC &&
1114 			    sk->sk_family != r->sdiag_family)
1115 				goto next_normal;
1116 			if (r->id.idiag_sport != htons(sk->sk_num) &&
1117 			    r->id.idiag_sport)
1118 				goto next_normal;
1119 			if (r->id.idiag_dport != sk->sk_dport &&
1120 			    r->id.idiag_dport)
1121 				goto next_normal;
1122 			twsk_build_assert();
1123 
1124 			if (!inet_diag_bc_sk(bc, sk))
1125 				goto next_normal;
1126 
1127 			if (!refcount_inc_not_zero(&sk->sk_refcnt))
1128 				goto next_normal;
1129 
1130 			num_arr[accum] = num;
1131 			sk_arr[accum] = sk;
1132 			if (++accum == SKARR_SZ)
1133 				break;
1134 next_normal:
1135 			++num;
1136 		}
1137 		spin_unlock_bh(lock);
1138 		res = 0;
1139 		for (idx = 0; idx < accum; idx++) {
1140 			if (res >= 0) {
1141 				res = sk_diag_fill(sk_arr[idx], skb, cb, r,
1142 						   NLM_F_MULTI, net_admin);
1143 				if (res < 0)
1144 					num = num_arr[idx];
1145 			}
1146 			sock_gen_put(sk_arr[idx]);
1147 		}
1148 		if (res < 0)
1149 			break;
1150 		cond_resched();
1151 		if (accum == SKARR_SZ) {
1152 			s_num = num + 1;
1153 			goto next_chunk;
1154 		}
1155 	}
1156 
1157 done:
1158 	cb->args[1] = i;
1159 	cb->args[2] = num;
1160 out:
1161 	;
1162 }
1163 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
1164 
1165 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
1166 			    const struct inet_diag_req_v2 *r)
1167 {
1168 	struct inet_diag_dump_data *cb_data = cb->data;
1169 	const struct inet_diag_handler *handler;
1170 	u32 prev_min_dump_alloc;
1171 	int protocol, err = 0;
1172 
1173 	protocol = inet_diag_get_protocol(r, cb_data);
1174 
1175 again:
1176 	prev_min_dump_alloc = cb->min_dump_alloc;
1177 	handler = inet_diag_lock_handler(protocol);
1178 	if (!IS_ERR(handler))
1179 		handler->dump(skb, cb, r);
1180 	else
1181 		err = PTR_ERR(handler);
1182 	inet_diag_unlock_handler(handler);
1183 
1184 	/* The skb is not large enough to fit one sk info and
1185 	 * inet_sk_diag_fill() has requested for a larger skb.
1186 	 */
1187 	if (!skb->len && cb->min_dump_alloc > prev_min_dump_alloc) {
1188 		err = pskb_expand_head(skb, 0, cb->min_dump_alloc, GFP_KERNEL);
1189 		if (!err)
1190 			goto again;
1191 	}
1192 
1193 	return err ? : skb->len;
1194 }
1195 
1196 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
1197 {
1198 	return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh));
1199 }
1200 
1201 static int __inet_diag_dump_start(struct netlink_callback *cb, int hdrlen)
1202 {
1203 	const struct nlmsghdr *nlh = cb->nlh;
1204 	struct inet_diag_dump_data *cb_data;
1205 	struct sk_buff *skb = cb->skb;
1206 	struct nlattr *nla;
1207 	int err;
1208 
1209 	cb_data = kzalloc(sizeof(*cb_data), GFP_KERNEL);
1210 	if (!cb_data)
1211 		return -ENOMEM;
1212 
1213 	err = inet_diag_parse_attrs(nlh, hdrlen, cb_data->req_nlas);
1214 	if (err) {
1215 		kfree(cb_data);
1216 		return err;
1217 	}
1218 	nla = cb_data->inet_diag_nla_bc;
1219 	if (nla) {
1220 		err = inet_diag_bc_audit(nla, skb);
1221 		if (err) {
1222 			kfree(cb_data);
1223 			return err;
1224 		}
1225 	}
1226 
1227 	nla = cb_data->inet_diag_nla_bpf_stgs;
1228 	if (nla) {
1229 		struct bpf_sk_storage_diag *bpf_stg_diag;
1230 
1231 		bpf_stg_diag = bpf_sk_storage_diag_alloc(nla);
1232 		if (IS_ERR(bpf_stg_diag)) {
1233 			kfree(cb_data);
1234 			return PTR_ERR(bpf_stg_diag);
1235 		}
1236 		cb_data->bpf_stg_diag = bpf_stg_diag;
1237 	}
1238 
1239 	cb->data = cb_data;
1240 	return 0;
1241 }
1242 
1243 static int inet_diag_dump_start(struct netlink_callback *cb)
1244 {
1245 	return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req_v2));
1246 }
1247 
1248 static int inet_diag_dump_start_compat(struct netlink_callback *cb)
1249 {
1250 	return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req));
1251 }
1252 
1253 static int inet_diag_dump_done(struct netlink_callback *cb)
1254 {
1255 	struct inet_diag_dump_data *cb_data = cb->data;
1256 
1257 	bpf_sk_storage_diag_free(cb_data->bpf_stg_diag);
1258 	kfree(cb->data);
1259 
1260 	return 0;
1261 }
1262 
1263 static int inet_diag_type2proto(int type)
1264 {
1265 	switch (type) {
1266 	case TCPDIAG_GETSOCK:
1267 		return IPPROTO_TCP;
1268 	case DCCPDIAG_GETSOCK:
1269 		return IPPROTO_DCCP;
1270 	default:
1271 		return 0;
1272 	}
1273 }
1274 
1275 static int inet_diag_dump_compat(struct sk_buff *skb,
1276 				 struct netlink_callback *cb)
1277 {
1278 	struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1279 	struct inet_diag_req_v2 req;
1280 
1281 	req.sdiag_family = AF_UNSPEC; /* compatibility */
1282 	req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1283 	req.idiag_ext = rc->idiag_ext;
1284 	req.idiag_states = rc->idiag_states;
1285 	req.id = rc->id;
1286 
1287 	return __inet_diag_dump(skb, cb, &req);
1288 }
1289 
1290 static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1291 				      const struct nlmsghdr *nlh)
1292 {
1293 	struct inet_diag_req *rc = nlmsg_data(nlh);
1294 	struct inet_diag_req_v2 req;
1295 
1296 	req.sdiag_family = rc->idiag_family;
1297 	req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1298 	req.idiag_ext = rc->idiag_ext;
1299 	req.idiag_states = rc->idiag_states;
1300 	req.id = rc->id;
1301 
1302 	return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh,
1303 				   sizeof(struct inet_diag_req), &req);
1304 }
1305 
1306 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1307 {
1308 	int hdrlen = sizeof(struct inet_diag_req);
1309 	struct net *net = sock_net(skb->sk);
1310 
1311 	if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1312 	    nlmsg_len(nlh) < hdrlen)
1313 		return -EINVAL;
1314 
1315 	if (nlh->nlmsg_flags & NLM_F_DUMP) {
1316 		struct netlink_dump_control c = {
1317 			.start = inet_diag_dump_start_compat,
1318 			.done = inet_diag_dump_done,
1319 			.dump = inet_diag_dump_compat,
1320 		};
1321 		return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1322 	}
1323 
1324 	return inet_diag_get_exact_compat(skb, nlh);
1325 }
1326 
1327 static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h)
1328 {
1329 	int hdrlen = sizeof(struct inet_diag_req_v2);
1330 	struct net *net = sock_net(skb->sk);
1331 
1332 	if (nlmsg_len(h) < hdrlen)
1333 		return -EINVAL;
1334 
1335 	if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY &&
1336 	    h->nlmsg_flags & NLM_F_DUMP) {
1337 		struct netlink_dump_control c = {
1338 			.start = inet_diag_dump_start,
1339 			.done = inet_diag_dump_done,
1340 			.dump = inet_diag_dump,
1341 		};
1342 		return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1343 	}
1344 
1345 	return inet_diag_cmd_exact(h->nlmsg_type, skb, h, hdrlen,
1346 				   nlmsg_data(h));
1347 }
1348 
1349 static
1350 int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk)
1351 {
1352 	const struct inet_diag_handler *handler;
1353 	struct nlmsghdr *nlh;
1354 	struct nlattr *attr;
1355 	struct inet_diag_msg *r;
1356 	void *info = NULL;
1357 	int err = 0;
1358 
1359 	nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0);
1360 	if (!nlh)
1361 		return -ENOMEM;
1362 
1363 	r = nlmsg_data(nlh);
1364 	memset(r, 0, sizeof(*r));
1365 	inet_diag_msg_common_fill(r, sk);
1366 	if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM)
1367 		r->id.idiag_sport = inet_sk(sk)->inet_sport;
1368 	r->idiag_state = sk->sk_state;
1369 
1370 	if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) {
1371 		nlmsg_cancel(skb, nlh);
1372 		return err;
1373 	}
1374 
1375 	handler = inet_diag_lock_handler(sk->sk_protocol);
1376 	if (IS_ERR(handler)) {
1377 		inet_diag_unlock_handler(handler);
1378 		nlmsg_cancel(skb, nlh);
1379 		return PTR_ERR(handler);
1380 	}
1381 
1382 	attr = handler->idiag_info_size
1383 		? nla_reserve_64bit(skb, INET_DIAG_INFO,
1384 				    handler->idiag_info_size,
1385 				    INET_DIAG_PAD)
1386 		: NULL;
1387 	if (attr)
1388 		info = nla_data(attr);
1389 
1390 	handler->idiag_get_info(sk, r, info);
1391 	inet_diag_unlock_handler(handler);
1392 
1393 	nlmsg_end(skb, nlh);
1394 	return 0;
1395 }
1396 
1397 static const struct sock_diag_handler inet_diag_handler = {
1398 	.family = AF_INET,
1399 	.dump = inet_diag_handler_cmd,
1400 	.get_info = inet_diag_handler_get_info,
1401 	.destroy = inet_diag_handler_cmd,
1402 };
1403 
1404 static const struct sock_diag_handler inet6_diag_handler = {
1405 	.family = AF_INET6,
1406 	.dump = inet_diag_handler_cmd,
1407 	.get_info = inet_diag_handler_get_info,
1408 	.destroy = inet_diag_handler_cmd,
1409 };
1410 
1411 int inet_diag_register(const struct inet_diag_handler *h)
1412 {
1413 	const __u16 type = h->idiag_type;
1414 	int err = -EINVAL;
1415 
1416 	if (type >= IPPROTO_MAX)
1417 		goto out;
1418 
1419 	mutex_lock(&inet_diag_table_mutex);
1420 	err = -EEXIST;
1421 	if (!inet_diag_table[type]) {
1422 		WRITE_ONCE(inet_diag_table[type], h);
1423 		err = 0;
1424 	}
1425 	mutex_unlock(&inet_diag_table_mutex);
1426 out:
1427 	return err;
1428 }
1429 EXPORT_SYMBOL_GPL(inet_diag_register);
1430 
1431 void inet_diag_unregister(const struct inet_diag_handler *h)
1432 {
1433 	const __u16 type = h->idiag_type;
1434 
1435 	if (type >= IPPROTO_MAX)
1436 		return;
1437 
1438 	mutex_lock(&inet_diag_table_mutex);
1439 	WRITE_ONCE(inet_diag_table[type], NULL);
1440 	mutex_unlock(&inet_diag_table_mutex);
1441 }
1442 EXPORT_SYMBOL_GPL(inet_diag_unregister);
1443 
1444 static int __init inet_diag_init(void)
1445 {
1446 	const int inet_diag_table_size = (IPPROTO_MAX *
1447 					  sizeof(struct inet_diag_handler *));
1448 	int err = -ENOMEM;
1449 
1450 	inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1451 	if (!inet_diag_table)
1452 		goto out;
1453 
1454 	err = sock_diag_register(&inet_diag_handler);
1455 	if (err)
1456 		goto out_free_nl;
1457 
1458 	err = sock_diag_register(&inet6_diag_handler);
1459 	if (err)
1460 		goto out_free_inet;
1461 
1462 	sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1463 out:
1464 	return err;
1465 
1466 out_free_inet:
1467 	sock_diag_unregister(&inet_diag_handler);
1468 out_free_nl:
1469 	kfree(inet_diag_table);
1470 	goto out;
1471 }
1472 
1473 static void __exit inet_diag_exit(void)
1474 {
1475 	sock_diag_unregister(&inet6_diag_handler);
1476 	sock_diag_unregister(&inet_diag_handler);
1477 	sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1478 	kfree(inet_diag_table);
1479 }
1480 
1481 module_init(inet_diag_init);
1482 module_exit(inet_diag_exit);
1483 MODULE_LICENSE("GPL");
1484 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1485 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);
1486