xref: /openbmc/linux/net/ipv4/inet_diag.c (revision 7dd65feb)
1 /*
2  * inet_diag.c	Module for monitoring INET transport protocols sockets.
3  *
4  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
5  *
6  *	This program is free software; you can redistribute it and/or
7  *      modify it under the terms of the GNU General Public License
8  *      as published by the Free Software Foundation; either version
9  *      2 of the License, or (at your option) any later version.
10  */
11 
12 #include <linux/kernel.h>
13 #include <linux/module.h>
14 #include <linux/types.h>
15 #include <linux/fcntl.h>
16 #include <linux/random.h>
17 #include <linux/cache.h>
18 #include <linux/init.h>
19 #include <linux/time.h>
20 
21 #include <net/icmp.h>
22 #include <net/tcp.h>
23 #include <net/ipv6.h>
24 #include <net/inet_common.h>
25 #include <net/inet_connection_sock.h>
26 #include <net/inet_hashtables.h>
27 #include <net/inet_timewait_sock.h>
28 #include <net/inet6_hashtables.h>
29 #include <net/netlink.h>
30 
31 #include <linux/inet.h>
32 #include <linux/stddef.h>
33 
34 #include <linux/inet_diag.h>
35 
36 static const struct inet_diag_handler **inet_diag_table;
37 
38 struct inet_diag_entry {
39 	__be32 *saddr;
40 	__be32 *daddr;
41 	u16 sport;
42 	u16 dport;
43 	u16 family;
44 	u16 userlocks;
45 };
46 
47 static struct sock *idiagnl;
48 
49 #define INET_DIAG_PUT(skb, attrtype, attrlen) \
50 	RTA_DATA(__RTA_PUT(skb, attrtype, attrlen))
51 
52 static DEFINE_MUTEX(inet_diag_table_mutex);
53 
54 static const struct inet_diag_handler *inet_diag_lock_handler(int type)
55 {
56 	if (!inet_diag_table[type])
57 		request_module("net-pf-%d-proto-%d-type-%d", PF_NETLINK,
58 			       NETLINK_INET_DIAG, type);
59 
60 	mutex_lock(&inet_diag_table_mutex);
61 	if (!inet_diag_table[type])
62 		return ERR_PTR(-ENOENT);
63 
64 	return inet_diag_table[type];
65 }
66 
67 static inline void inet_diag_unlock_handler(
68 	const struct inet_diag_handler *handler)
69 {
70 	mutex_unlock(&inet_diag_table_mutex);
71 }
72 
73 static int inet_csk_diag_fill(struct sock *sk,
74 			      struct sk_buff *skb,
75 			      int ext, u32 pid, u32 seq, u16 nlmsg_flags,
76 			      const struct nlmsghdr *unlh)
77 {
78 	const struct inet_sock *inet = inet_sk(sk);
79 	const struct inet_connection_sock *icsk = inet_csk(sk);
80 	struct inet_diag_msg *r;
81 	struct nlmsghdr  *nlh;
82 	void *info = NULL;
83 	struct inet_diag_meminfo  *minfo = NULL;
84 	unsigned char	 *b = skb_tail_pointer(skb);
85 	const struct inet_diag_handler *handler;
86 
87 	handler = inet_diag_table[unlh->nlmsg_type];
88 	BUG_ON(handler == NULL);
89 
90 	nlh = NLMSG_PUT(skb, pid, seq, unlh->nlmsg_type, sizeof(*r));
91 	nlh->nlmsg_flags = nlmsg_flags;
92 
93 	r = NLMSG_DATA(nlh);
94 	BUG_ON(sk->sk_state == TCP_TIME_WAIT);
95 
96 	if (ext & (1 << (INET_DIAG_MEMINFO - 1)))
97 		minfo = INET_DIAG_PUT(skb, INET_DIAG_MEMINFO, sizeof(*minfo));
98 
99 	if (ext & (1 << (INET_DIAG_INFO - 1)))
100 		info = INET_DIAG_PUT(skb, INET_DIAG_INFO,
101 				     handler->idiag_info_size);
102 
103 	if ((ext & (1 << (INET_DIAG_CONG - 1))) && icsk->icsk_ca_ops) {
104 		const size_t len = strlen(icsk->icsk_ca_ops->name);
105 
106 		strcpy(INET_DIAG_PUT(skb, INET_DIAG_CONG, len + 1),
107 		       icsk->icsk_ca_ops->name);
108 	}
109 
110 	r->idiag_family = sk->sk_family;
111 	r->idiag_state = sk->sk_state;
112 	r->idiag_timer = 0;
113 	r->idiag_retrans = 0;
114 
115 	r->id.idiag_if = sk->sk_bound_dev_if;
116 	r->id.idiag_cookie[0] = (u32)(unsigned long)sk;
117 	r->id.idiag_cookie[1] = (u32)(((unsigned long)sk >> 31) >> 1);
118 
119 	r->id.idiag_sport = inet->inet_sport;
120 	r->id.idiag_dport = inet->inet_dport;
121 	r->id.idiag_src[0] = inet->inet_rcv_saddr;
122 	r->id.idiag_dst[0] = inet->inet_daddr;
123 
124 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
125 	if (r->idiag_family == AF_INET6) {
126 		struct ipv6_pinfo *np = inet6_sk(sk);
127 
128 		ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
129 			       &np->rcv_saddr);
130 		ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
131 			       &np->daddr);
132 	}
133 #endif
134 
135 #define EXPIRES_IN_MS(tmo)  DIV_ROUND_UP((tmo - jiffies) * 1000, HZ)
136 
137 	if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
138 		r->idiag_timer = 1;
139 		r->idiag_retrans = icsk->icsk_retransmits;
140 		r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
141 	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
142 		r->idiag_timer = 4;
143 		r->idiag_retrans = icsk->icsk_probes_out;
144 		r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
145 	} else if (timer_pending(&sk->sk_timer)) {
146 		r->idiag_timer = 2;
147 		r->idiag_retrans = icsk->icsk_probes_out;
148 		r->idiag_expires = EXPIRES_IN_MS(sk->sk_timer.expires);
149 	} else {
150 		r->idiag_timer = 0;
151 		r->idiag_expires = 0;
152 	}
153 #undef EXPIRES_IN_MS
154 
155 	r->idiag_uid = sock_i_uid(sk);
156 	r->idiag_inode = sock_i_ino(sk);
157 
158 	if (minfo) {
159 		minfo->idiag_rmem = sk_rmem_alloc_get(sk);
160 		minfo->idiag_wmem = sk->sk_wmem_queued;
161 		minfo->idiag_fmem = sk->sk_forward_alloc;
162 		minfo->idiag_tmem = sk_wmem_alloc_get(sk);
163 	}
164 
165 	handler->idiag_get_info(sk, r, info);
166 
167 	if (sk->sk_state < TCP_TIME_WAIT &&
168 	    icsk->icsk_ca_ops && icsk->icsk_ca_ops->get_info)
169 		icsk->icsk_ca_ops->get_info(sk, ext, skb);
170 
171 	nlh->nlmsg_len = skb_tail_pointer(skb) - b;
172 	return skb->len;
173 
174 rtattr_failure:
175 nlmsg_failure:
176 	nlmsg_trim(skb, b);
177 	return -EMSGSIZE;
178 }
179 
180 static int inet_twsk_diag_fill(struct inet_timewait_sock *tw,
181 			       struct sk_buff *skb, int ext, u32 pid,
182 			       u32 seq, u16 nlmsg_flags,
183 			       const struct nlmsghdr *unlh)
184 {
185 	long tmo;
186 	struct inet_diag_msg *r;
187 	const unsigned char *previous_tail = skb_tail_pointer(skb);
188 	struct nlmsghdr *nlh = NLMSG_PUT(skb, pid, seq,
189 					 unlh->nlmsg_type, sizeof(*r));
190 
191 	r = NLMSG_DATA(nlh);
192 	BUG_ON(tw->tw_state != TCP_TIME_WAIT);
193 
194 	nlh->nlmsg_flags = nlmsg_flags;
195 
196 	tmo = tw->tw_ttd - jiffies;
197 	if (tmo < 0)
198 		tmo = 0;
199 
200 	r->idiag_family	      = tw->tw_family;
201 	r->idiag_retrans      = 0;
202 	r->id.idiag_if	      = tw->tw_bound_dev_if;
203 	r->id.idiag_cookie[0] = (u32)(unsigned long)tw;
204 	r->id.idiag_cookie[1] = (u32)(((unsigned long)tw >> 31) >> 1);
205 	r->id.idiag_sport     = tw->tw_sport;
206 	r->id.idiag_dport     = tw->tw_dport;
207 	r->id.idiag_src[0]    = tw->tw_rcv_saddr;
208 	r->id.idiag_dst[0]    = tw->tw_daddr;
209 	r->idiag_state	      = tw->tw_substate;
210 	r->idiag_timer	      = 3;
211 	r->idiag_expires      = DIV_ROUND_UP(tmo * 1000, HZ);
212 	r->idiag_rqueue	      = 0;
213 	r->idiag_wqueue	      = 0;
214 	r->idiag_uid	      = 0;
215 	r->idiag_inode	      = 0;
216 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
217 	if (tw->tw_family == AF_INET6) {
218 		const struct inet6_timewait_sock *tw6 =
219 						inet6_twsk((struct sock *)tw);
220 
221 		ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
222 			       &tw6->tw_v6_rcv_saddr);
223 		ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
224 			       &tw6->tw_v6_daddr);
225 	}
226 #endif
227 	nlh->nlmsg_len = skb_tail_pointer(skb) - previous_tail;
228 	return skb->len;
229 nlmsg_failure:
230 	nlmsg_trim(skb, previous_tail);
231 	return -EMSGSIZE;
232 }
233 
234 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
235 			int ext, u32 pid, u32 seq, u16 nlmsg_flags,
236 			const struct nlmsghdr *unlh)
237 {
238 	if (sk->sk_state == TCP_TIME_WAIT)
239 		return inet_twsk_diag_fill((struct inet_timewait_sock *)sk,
240 					   skb, ext, pid, seq, nlmsg_flags,
241 					   unlh);
242 	return inet_csk_diag_fill(sk, skb, ext, pid, seq, nlmsg_flags, unlh);
243 }
244 
245 static int inet_diag_get_exact(struct sk_buff *in_skb,
246 			       const struct nlmsghdr *nlh)
247 {
248 	int err;
249 	struct sock *sk;
250 	struct inet_diag_req *req = NLMSG_DATA(nlh);
251 	struct sk_buff *rep;
252 	struct inet_hashinfo *hashinfo;
253 	const struct inet_diag_handler *handler;
254 
255 	handler = inet_diag_lock_handler(nlh->nlmsg_type);
256 	if (IS_ERR(handler)) {
257 		err = PTR_ERR(handler);
258 		goto unlock;
259 	}
260 
261 	hashinfo = handler->idiag_hashinfo;
262 	err = -EINVAL;
263 
264 	if (req->idiag_family == AF_INET) {
265 		sk = inet_lookup(&init_net, hashinfo, req->id.idiag_dst[0],
266 				 req->id.idiag_dport, req->id.idiag_src[0],
267 				 req->id.idiag_sport, req->id.idiag_if);
268 	}
269 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
270 	else if (req->idiag_family == AF_INET6) {
271 		sk = inet6_lookup(&init_net, hashinfo,
272 				  (struct in6_addr *)req->id.idiag_dst,
273 				  req->id.idiag_dport,
274 				  (struct in6_addr *)req->id.idiag_src,
275 				  req->id.idiag_sport,
276 				  req->id.idiag_if);
277 	}
278 #endif
279 	else {
280 		goto unlock;
281 	}
282 
283 	err = -ENOENT;
284 	if (sk == NULL)
285 		goto unlock;
286 
287 	err = -ESTALE;
288 	if ((req->id.idiag_cookie[0] != INET_DIAG_NOCOOKIE ||
289 	     req->id.idiag_cookie[1] != INET_DIAG_NOCOOKIE) &&
290 	    ((u32)(unsigned long)sk != req->id.idiag_cookie[0] ||
291 	     (u32)((((unsigned long)sk) >> 31) >> 1) != req->id.idiag_cookie[1]))
292 		goto out;
293 
294 	err = -ENOMEM;
295 	rep = alloc_skb(NLMSG_SPACE((sizeof(struct inet_diag_msg) +
296 				     sizeof(struct inet_diag_meminfo) +
297 				     handler->idiag_info_size + 64)),
298 			GFP_KERNEL);
299 	if (!rep)
300 		goto out;
301 
302 	err = sk_diag_fill(sk, rep, req->idiag_ext,
303 			   NETLINK_CB(in_skb).pid,
304 			   nlh->nlmsg_seq, 0, nlh);
305 	if (err < 0) {
306 		WARN_ON(err == -EMSGSIZE);
307 		kfree_skb(rep);
308 		goto out;
309 	}
310 	err = netlink_unicast(idiagnl, rep, NETLINK_CB(in_skb).pid,
311 			      MSG_DONTWAIT);
312 	if (err > 0)
313 		err = 0;
314 
315 out:
316 	if (sk) {
317 		if (sk->sk_state == TCP_TIME_WAIT)
318 			inet_twsk_put((struct inet_timewait_sock *)sk);
319 		else
320 			sock_put(sk);
321 	}
322 unlock:
323 	inet_diag_unlock_handler(handler);
324 	return err;
325 }
326 
327 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
328 {
329 	int words = bits >> 5;
330 
331 	bits &= 0x1f;
332 
333 	if (words) {
334 		if (memcmp(a1, a2, words << 2))
335 			return 0;
336 	}
337 	if (bits) {
338 		__be32 w1, w2;
339 		__be32 mask;
340 
341 		w1 = a1[words];
342 		w2 = a2[words];
343 
344 		mask = htonl((0xffffffff) << (32 - bits));
345 
346 		if ((w1 ^ w2) & mask)
347 			return 0;
348 	}
349 
350 	return 1;
351 }
352 
353 
354 static int inet_diag_bc_run(const void *bc, int len,
355 			    const struct inet_diag_entry *entry)
356 {
357 	while (len > 0) {
358 		int yes = 1;
359 		const struct inet_diag_bc_op *op = bc;
360 
361 		switch (op->code) {
362 		case INET_DIAG_BC_NOP:
363 			break;
364 		case INET_DIAG_BC_JMP:
365 			yes = 0;
366 			break;
367 		case INET_DIAG_BC_S_GE:
368 			yes = entry->sport >= op[1].no;
369 			break;
370 		case INET_DIAG_BC_S_LE:
371 			yes = entry->dport <= op[1].no;
372 			break;
373 		case INET_DIAG_BC_D_GE:
374 			yes = entry->dport >= op[1].no;
375 			break;
376 		case INET_DIAG_BC_D_LE:
377 			yes = entry->dport <= op[1].no;
378 			break;
379 		case INET_DIAG_BC_AUTO:
380 			yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
381 			break;
382 		case INET_DIAG_BC_S_COND:
383 		case INET_DIAG_BC_D_COND: {
384 			struct inet_diag_hostcond *cond;
385 			__be32 *addr;
386 
387 			cond = (struct inet_diag_hostcond *)(op + 1);
388 			if (cond->port != -1 &&
389 			    cond->port != (op->code == INET_DIAG_BC_S_COND ?
390 					     entry->sport : entry->dport)) {
391 				yes = 0;
392 				break;
393 			}
394 
395 			if (cond->prefix_len == 0)
396 				break;
397 
398 			if (op->code == INET_DIAG_BC_S_COND)
399 				addr = entry->saddr;
400 			else
401 				addr = entry->daddr;
402 
403 			if (bitstring_match(addr, cond->addr,
404 					    cond->prefix_len))
405 				break;
406 			if (entry->family == AF_INET6 &&
407 			    cond->family == AF_INET) {
408 				if (addr[0] == 0 && addr[1] == 0 &&
409 				    addr[2] == htonl(0xffff) &&
410 				    bitstring_match(addr + 3, cond->addr,
411 						    cond->prefix_len))
412 					break;
413 			}
414 			yes = 0;
415 			break;
416 		}
417 		}
418 
419 		if (yes) {
420 			len -= op->yes;
421 			bc += op->yes;
422 		} else {
423 			len -= op->no;
424 			bc += op->no;
425 		}
426 	}
427 	return (len == 0);
428 }
429 
430 static int valid_cc(const void *bc, int len, int cc)
431 {
432 	while (len >= 0) {
433 		const struct inet_diag_bc_op *op = bc;
434 
435 		if (cc > len)
436 			return 0;
437 		if (cc == len)
438 			return 1;
439 		if (op->yes < 4)
440 			return 0;
441 		len -= op->yes;
442 		bc  += op->yes;
443 	}
444 	return 0;
445 }
446 
447 static int inet_diag_bc_audit(const void *bytecode, int bytecode_len)
448 {
449 	const unsigned char *bc = bytecode;
450 	int  len = bytecode_len;
451 
452 	while (len > 0) {
453 		struct inet_diag_bc_op *op = (struct inet_diag_bc_op *)bc;
454 
455 //printk("BC: %d %d %d {%d} / %d\n", op->code, op->yes, op->no, op[1].no, len);
456 		switch (op->code) {
457 		case INET_DIAG_BC_AUTO:
458 		case INET_DIAG_BC_S_COND:
459 		case INET_DIAG_BC_D_COND:
460 		case INET_DIAG_BC_S_GE:
461 		case INET_DIAG_BC_S_LE:
462 		case INET_DIAG_BC_D_GE:
463 		case INET_DIAG_BC_D_LE:
464 			if (op->yes < 4 || op->yes > len + 4)
465 				return -EINVAL;
466 		case INET_DIAG_BC_JMP:
467 			if (op->no < 4 || op->no > len + 4)
468 				return -EINVAL;
469 			if (op->no < len &&
470 			    !valid_cc(bytecode, bytecode_len, len - op->no))
471 				return -EINVAL;
472 			break;
473 		case INET_DIAG_BC_NOP:
474 			if (op->yes < 4 || op->yes > len + 4)
475 				return -EINVAL;
476 			break;
477 		default:
478 			return -EINVAL;
479 		}
480 		bc  += op->yes;
481 		len -= op->yes;
482 	}
483 	return len == 0 ? 0 : -EINVAL;
484 }
485 
486 static int inet_csk_diag_dump(struct sock *sk,
487 			      struct sk_buff *skb,
488 			      struct netlink_callback *cb)
489 {
490 	struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
491 
492 	if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) {
493 		struct inet_diag_entry entry;
494 		struct rtattr *bc = (struct rtattr *)(r + 1);
495 		struct inet_sock *inet = inet_sk(sk);
496 
497 		entry.family = sk->sk_family;
498 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
499 		if (entry.family == AF_INET6) {
500 			struct ipv6_pinfo *np = inet6_sk(sk);
501 
502 			entry.saddr = np->rcv_saddr.s6_addr32;
503 			entry.daddr = np->daddr.s6_addr32;
504 		} else
505 #endif
506 		{
507 			entry.saddr = &inet->inet_rcv_saddr;
508 			entry.daddr = &inet->inet_daddr;
509 		}
510 		entry.sport = inet->inet_num;
511 		entry.dport = ntohs(inet->inet_dport);
512 		entry.userlocks = sk->sk_userlocks;
513 
514 		if (!inet_diag_bc_run(RTA_DATA(bc), RTA_PAYLOAD(bc), &entry))
515 			return 0;
516 	}
517 
518 	return inet_csk_diag_fill(sk, skb, r->idiag_ext,
519 				  NETLINK_CB(cb->skb).pid,
520 				  cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
521 }
522 
523 static int inet_twsk_diag_dump(struct inet_timewait_sock *tw,
524 			       struct sk_buff *skb,
525 			       struct netlink_callback *cb)
526 {
527 	struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
528 
529 	if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) {
530 		struct inet_diag_entry entry;
531 		struct rtattr *bc = (struct rtattr *)(r + 1);
532 
533 		entry.family = tw->tw_family;
534 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
535 		if (tw->tw_family == AF_INET6) {
536 			struct inet6_timewait_sock *tw6 =
537 						inet6_twsk((struct sock *)tw);
538 			entry.saddr = tw6->tw_v6_rcv_saddr.s6_addr32;
539 			entry.daddr = tw6->tw_v6_daddr.s6_addr32;
540 		} else
541 #endif
542 		{
543 			entry.saddr = &tw->tw_rcv_saddr;
544 			entry.daddr = &tw->tw_daddr;
545 		}
546 		entry.sport = tw->tw_num;
547 		entry.dport = ntohs(tw->tw_dport);
548 		entry.userlocks = 0;
549 
550 		if (!inet_diag_bc_run(RTA_DATA(bc), RTA_PAYLOAD(bc), &entry))
551 			return 0;
552 	}
553 
554 	return inet_twsk_diag_fill(tw, skb, r->idiag_ext,
555 				   NETLINK_CB(cb->skb).pid,
556 				   cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
557 }
558 
559 static int inet_diag_fill_req(struct sk_buff *skb, struct sock *sk,
560 			      struct request_sock *req, u32 pid, u32 seq,
561 			      const struct nlmsghdr *unlh)
562 {
563 	const struct inet_request_sock *ireq = inet_rsk(req);
564 	struct inet_sock *inet = inet_sk(sk);
565 	unsigned char *b = skb_tail_pointer(skb);
566 	struct inet_diag_msg *r;
567 	struct nlmsghdr *nlh;
568 	long tmo;
569 
570 	nlh = NLMSG_PUT(skb, pid, seq, unlh->nlmsg_type, sizeof(*r));
571 	nlh->nlmsg_flags = NLM_F_MULTI;
572 	r = NLMSG_DATA(nlh);
573 
574 	r->idiag_family = sk->sk_family;
575 	r->idiag_state = TCP_SYN_RECV;
576 	r->idiag_timer = 1;
577 	r->idiag_retrans = req->retrans;
578 
579 	r->id.idiag_if = sk->sk_bound_dev_if;
580 	r->id.idiag_cookie[0] = (u32)(unsigned long)req;
581 	r->id.idiag_cookie[1] = (u32)(((unsigned long)req >> 31) >> 1);
582 
583 	tmo = req->expires - jiffies;
584 	if (tmo < 0)
585 		tmo = 0;
586 
587 	r->id.idiag_sport = inet->inet_sport;
588 	r->id.idiag_dport = ireq->rmt_port;
589 	r->id.idiag_src[0] = ireq->loc_addr;
590 	r->id.idiag_dst[0] = ireq->rmt_addr;
591 	r->idiag_expires = jiffies_to_msecs(tmo);
592 	r->idiag_rqueue = 0;
593 	r->idiag_wqueue = 0;
594 	r->idiag_uid = sock_i_uid(sk);
595 	r->idiag_inode = 0;
596 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
597 	if (r->idiag_family == AF_INET6) {
598 		ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
599 			       &inet6_rsk(req)->loc_addr);
600 		ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
601 			       &inet6_rsk(req)->rmt_addr);
602 	}
603 #endif
604 	nlh->nlmsg_len = skb_tail_pointer(skb) - b;
605 
606 	return skb->len;
607 
608 nlmsg_failure:
609 	nlmsg_trim(skb, b);
610 	return -1;
611 }
612 
613 static int inet_diag_dump_reqs(struct sk_buff *skb, struct sock *sk,
614 			       struct netlink_callback *cb)
615 {
616 	struct inet_diag_entry entry;
617 	struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
618 	struct inet_connection_sock *icsk = inet_csk(sk);
619 	struct listen_sock *lopt;
620 	struct rtattr *bc = NULL;
621 	struct inet_sock *inet = inet_sk(sk);
622 	int j, s_j;
623 	int reqnum, s_reqnum;
624 	int err = 0;
625 
626 	s_j = cb->args[3];
627 	s_reqnum = cb->args[4];
628 
629 	if (s_j > 0)
630 		s_j--;
631 
632 	entry.family = sk->sk_family;
633 
634 	read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
635 
636 	lopt = icsk->icsk_accept_queue.listen_opt;
637 	if (!lopt || !lopt->qlen)
638 		goto out;
639 
640 	if (cb->nlh->nlmsg_len > 4 + NLMSG_SPACE(sizeof(*r))) {
641 		bc = (struct rtattr *)(r + 1);
642 		entry.sport = inet->inet_num;
643 		entry.userlocks = sk->sk_userlocks;
644 	}
645 
646 	for (j = s_j; j < lopt->nr_table_entries; j++) {
647 		struct request_sock *req, *head = lopt->syn_table[j];
648 
649 		reqnum = 0;
650 		for (req = head; req; reqnum++, req = req->dl_next) {
651 			struct inet_request_sock *ireq = inet_rsk(req);
652 
653 			if (reqnum < s_reqnum)
654 				continue;
655 			if (r->id.idiag_dport != ireq->rmt_port &&
656 			    r->id.idiag_dport)
657 				continue;
658 
659 			if (bc) {
660 				entry.saddr =
661 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
662 					(entry.family == AF_INET6) ?
663 					inet6_rsk(req)->loc_addr.s6_addr32 :
664 #endif
665 					&ireq->loc_addr;
666 				entry.daddr =
667 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
668 					(entry.family == AF_INET6) ?
669 					inet6_rsk(req)->rmt_addr.s6_addr32 :
670 #endif
671 					&ireq->rmt_addr;
672 				entry.dport = ntohs(ireq->rmt_port);
673 
674 				if (!inet_diag_bc_run(RTA_DATA(bc),
675 						    RTA_PAYLOAD(bc), &entry))
676 					continue;
677 			}
678 
679 			err = inet_diag_fill_req(skb, sk, req,
680 					       NETLINK_CB(cb->skb).pid,
681 					       cb->nlh->nlmsg_seq, cb->nlh);
682 			if (err < 0) {
683 				cb->args[3] = j + 1;
684 				cb->args[4] = reqnum;
685 				goto out;
686 			}
687 		}
688 
689 		s_reqnum = 0;
690 	}
691 
692 out:
693 	read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
694 
695 	return err;
696 }
697 
698 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
699 {
700 	int i, num;
701 	int s_i, s_num;
702 	struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
703 	const struct inet_diag_handler *handler;
704 	struct inet_hashinfo *hashinfo;
705 
706 	handler = inet_diag_lock_handler(cb->nlh->nlmsg_type);
707 	if (IS_ERR(handler))
708 		goto unlock;
709 
710 	hashinfo = handler->idiag_hashinfo;
711 
712 	s_i = cb->args[1];
713 	s_num = num = cb->args[2];
714 
715 	if (cb->args[0] == 0) {
716 		if (!(r->idiag_states & (TCPF_LISTEN | TCPF_SYN_RECV)))
717 			goto skip_listen_ht;
718 
719 		for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
720 			struct sock *sk;
721 			struct hlist_nulls_node *node;
722 			struct inet_listen_hashbucket *ilb;
723 
724 			num = 0;
725 			ilb = &hashinfo->listening_hash[i];
726 			spin_lock_bh(&ilb->lock);
727 			sk_nulls_for_each(sk, node, &ilb->head) {
728 				struct inet_sock *inet = inet_sk(sk);
729 
730 				if (num < s_num) {
731 					num++;
732 					continue;
733 				}
734 
735 				if (r->id.idiag_sport != inet->inet_sport &&
736 				    r->id.idiag_sport)
737 					goto next_listen;
738 
739 				if (!(r->idiag_states & TCPF_LISTEN) ||
740 				    r->id.idiag_dport ||
741 				    cb->args[3] > 0)
742 					goto syn_recv;
743 
744 				if (inet_csk_diag_dump(sk, skb, cb) < 0) {
745 					spin_unlock_bh(&ilb->lock);
746 					goto done;
747 				}
748 
749 syn_recv:
750 				if (!(r->idiag_states & TCPF_SYN_RECV))
751 					goto next_listen;
752 
753 				if (inet_diag_dump_reqs(skb, sk, cb) < 0) {
754 					spin_unlock_bh(&ilb->lock);
755 					goto done;
756 				}
757 
758 next_listen:
759 				cb->args[3] = 0;
760 				cb->args[4] = 0;
761 				++num;
762 			}
763 			spin_unlock_bh(&ilb->lock);
764 
765 			s_num = 0;
766 			cb->args[3] = 0;
767 			cb->args[4] = 0;
768 		}
769 skip_listen_ht:
770 		cb->args[0] = 1;
771 		s_i = num = s_num = 0;
772 	}
773 
774 	if (!(r->idiag_states & ~(TCPF_LISTEN | TCPF_SYN_RECV)))
775 		goto unlock;
776 
777 	for (i = s_i; i <= hashinfo->ehash_mask; i++) {
778 		struct inet_ehash_bucket *head = &hashinfo->ehash[i];
779 		spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
780 		struct sock *sk;
781 		struct hlist_nulls_node *node;
782 
783 		num = 0;
784 
785 		if (hlist_nulls_empty(&head->chain) &&
786 			hlist_nulls_empty(&head->twchain))
787 			continue;
788 
789 		if (i > s_i)
790 			s_num = 0;
791 
792 		spin_lock_bh(lock);
793 		sk_nulls_for_each(sk, node, &head->chain) {
794 			struct inet_sock *inet = inet_sk(sk);
795 
796 			if (num < s_num)
797 				goto next_normal;
798 			if (!(r->idiag_states & (1 << sk->sk_state)))
799 				goto next_normal;
800 			if (r->id.idiag_sport != inet->inet_sport &&
801 			    r->id.idiag_sport)
802 				goto next_normal;
803 			if (r->id.idiag_dport != inet->inet_dport &&
804 			    r->id.idiag_dport)
805 				goto next_normal;
806 			if (inet_csk_diag_dump(sk, skb, cb) < 0) {
807 				spin_unlock_bh(lock);
808 				goto done;
809 			}
810 next_normal:
811 			++num;
812 		}
813 
814 		if (r->idiag_states & TCPF_TIME_WAIT) {
815 			struct inet_timewait_sock *tw;
816 
817 			inet_twsk_for_each(tw, node,
818 				    &head->twchain) {
819 
820 				if (num < s_num)
821 					goto next_dying;
822 				if (r->id.idiag_sport != tw->tw_sport &&
823 				    r->id.idiag_sport)
824 					goto next_dying;
825 				if (r->id.idiag_dport != tw->tw_dport &&
826 				    r->id.idiag_dport)
827 					goto next_dying;
828 				if (inet_twsk_diag_dump(tw, skb, cb) < 0) {
829 					spin_unlock_bh(lock);
830 					goto done;
831 				}
832 next_dying:
833 				++num;
834 			}
835 		}
836 		spin_unlock_bh(lock);
837 	}
838 
839 done:
840 	cb->args[1] = i;
841 	cb->args[2] = num;
842 unlock:
843 	inet_diag_unlock_handler(handler);
844 	return skb->len;
845 }
846 
847 static int inet_diag_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
848 {
849 	int hdrlen = sizeof(struct inet_diag_req);
850 
851 	if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
852 	    nlmsg_len(nlh) < hdrlen)
853 		return -EINVAL;
854 
855 	if (nlh->nlmsg_flags & NLM_F_DUMP) {
856 		if (nlmsg_attrlen(nlh, hdrlen)) {
857 			struct nlattr *attr;
858 
859 			attr = nlmsg_find_attr(nlh, hdrlen,
860 					       INET_DIAG_REQ_BYTECODE);
861 			if (attr == NULL ||
862 			    nla_len(attr) < sizeof(struct inet_diag_bc_op) ||
863 			    inet_diag_bc_audit(nla_data(attr), nla_len(attr)))
864 				return -EINVAL;
865 		}
866 
867 		return netlink_dump_start(idiagnl, skb, nlh,
868 					  inet_diag_dump, NULL);
869 	}
870 
871 	return inet_diag_get_exact(skb, nlh);
872 }
873 
874 static DEFINE_MUTEX(inet_diag_mutex);
875 
876 static void inet_diag_rcv(struct sk_buff *skb)
877 {
878 	mutex_lock(&inet_diag_mutex);
879 	netlink_rcv_skb(skb, &inet_diag_rcv_msg);
880 	mutex_unlock(&inet_diag_mutex);
881 }
882 
883 int inet_diag_register(const struct inet_diag_handler *h)
884 {
885 	const __u16 type = h->idiag_type;
886 	int err = -EINVAL;
887 
888 	if (type >= INET_DIAG_GETSOCK_MAX)
889 		goto out;
890 
891 	mutex_lock(&inet_diag_table_mutex);
892 	err = -EEXIST;
893 	if (inet_diag_table[type] == NULL) {
894 		inet_diag_table[type] = h;
895 		err = 0;
896 	}
897 	mutex_unlock(&inet_diag_table_mutex);
898 out:
899 	return err;
900 }
901 EXPORT_SYMBOL_GPL(inet_diag_register);
902 
903 void inet_diag_unregister(const struct inet_diag_handler *h)
904 {
905 	const __u16 type = h->idiag_type;
906 
907 	if (type >= INET_DIAG_GETSOCK_MAX)
908 		return;
909 
910 	mutex_lock(&inet_diag_table_mutex);
911 	inet_diag_table[type] = NULL;
912 	mutex_unlock(&inet_diag_table_mutex);
913 }
914 EXPORT_SYMBOL_GPL(inet_diag_unregister);
915 
916 static int __init inet_diag_init(void)
917 {
918 	const int inet_diag_table_size = (INET_DIAG_GETSOCK_MAX *
919 					  sizeof(struct inet_diag_handler *));
920 	int err = -ENOMEM;
921 
922 	inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
923 	if (!inet_diag_table)
924 		goto out;
925 
926 	idiagnl = netlink_kernel_create(&init_net, NETLINK_INET_DIAG, 0,
927 					inet_diag_rcv, NULL, THIS_MODULE);
928 	if (idiagnl == NULL)
929 		goto out_free_table;
930 	err = 0;
931 out:
932 	return err;
933 out_free_table:
934 	kfree(inet_diag_table);
935 	goto out;
936 }
937 
938 static void __exit inet_diag_exit(void)
939 {
940 	netlink_kernel_release(idiagnl);
941 	kfree(inet_diag_table);
942 }
943 
944 module_init(inet_diag_init);
945 module_exit(inet_diag_exit);
946 MODULE_LICENSE("GPL");
947 MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK, NETLINK_INET_DIAG);
948