xref: /openbmc/linux/net/ipv4/inet_diag.c (revision 7b35eadd)
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/slab.h>
18 #include <linux/cache.h>
19 #include <linux/init.h>
20 #include <linux/time.h>
21 
22 #include <net/icmp.h>
23 #include <net/tcp.h>
24 #include <net/ipv6.h>
25 #include <net/inet_common.h>
26 #include <net/inet_connection_sock.h>
27 #include <net/inet_hashtables.h>
28 #include <net/inet_timewait_sock.h>
29 #include <net/inet6_hashtables.h>
30 #include <net/netlink.h>
31 
32 #include <linux/inet.h>
33 #include <linux/stddef.h>
34 
35 #include <linux/inet_diag.h>
36 #include <linux/sock_diag.h>
37 
38 static const struct inet_diag_handler **inet_diag_table;
39 
40 struct inet_diag_entry {
41 	__be32 *saddr;
42 	__be32 *daddr;
43 	u16 sport;
44 	u16 dport;
45 	u16 family;
46 	u16 userlocks;
47 };
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 proto)
55 {
56 	if (!inet_diag_table[proto])
57 		request_module("net-pf-%d-proto-%d-type-%d", PF_NETLINK,
58 			       NETLINK_SOCK_DIAG, proto);
59 
60 	mutex_lock(&inet_diag_table_mutex);
61 	if (!inet_diag_table[proto])
62 		return ERR_PTR(-ENOENT);
63 
64 	return inet_diag_table[proto];
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, struct inet_diag_req *req,
75 			      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 	int ext = req->idiag_ext;
87 
88 	handler = inet_diag_table[req->sdiag_protocol];
89 	BUG_ON(handler == NULL);
90 
91 	nlh = NLMSG_PUT(skb, pid, seq, unlh->nlmsg_type, sizeof(*r));
92 	nlh->nlmsg_flags = nlmsg_flags;
93 
94 	r = NLMSG_DATA(nlh);
95 	BUG_ON(sk->sk_state == TCP_TIME_WAIT);
96 
97 	if (ext & (1 << (INET_DIAG_MEMINFO - 1)))
98 		minfo = INET_DIAG_PUT(skb, INET_DIAG_MEMINFO, sizeof(*minfo));
99 
100 	if (ext & (1 << (INET_DIAG_INFO - 1)))
101 		info = INET_DIAG_PUT(skb, INET_DIAG_INFO, sizeof(struct tcp_info));
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 	/* IPv6 dual-stack sockets use inet->tos for IPv4 connections,
125 	 * hence this needs to be included regardless of socket family.
126 	 */
127 	if (ext & (1 << (INET_DIAG_TOS - 1)))
128 		RTA_PUT_U8(skb, INET_DIAG_TOS, inet->tos);
129 
130 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
131 	if (r->idiag_family == AF_INET6) {
132 		const struct ipv6_pinfo *np = inet6_sk(sk);
133 
134 		*(struct in6_addr *)r->id.idiag_src = np->rcv_saddr;
135 		*(struct in6_addr *)r->id.idiag_dst = np->daddr;
136 		if (ext & (1 << (INET_DIAG_TCLASS - 1)))
137 			RTA_PUT_U8(skb, INET_DIAG_TCLASS, np->tclass);
138 	}
139 #endif
140 
141 #define EXPIRES_IN_MS(tmo)  DIV_ROUND_UP((tmo - jiffies) * 1000, HZ)
142 
143 	if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
144 		r->idiag_timer = 1;
145 		r->idiag_retrans = icsk->icsk_retransmits;
146 		r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
147 	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
148 		r->idiag_timer = 4;
149 		r->idiag_retrans = icsk->icsk_probes_out;
150 		r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
151 	} else if (timer_pending(&sk->sk_timer)) {
152 		r->idiag_timer = 2;
153 		r->idiag_retrans = icsk->icsk_probes_out;
154 		r->idiag_expires = EXPIRES_IN_MS(sk->sk_timer.expires);
155 	} else {
156 		r->idiag_timer = 0;
157 		r->idiag_expires = 0;
158 	}
159 #undef EXPIRES_IN_MS
160 
161 	r->idiag_uid = sock_i_uid(sk);
162 	r->idiag_inode = sock_i_ino(sk);
163 
164 	if (minfo) {
165 		minfo->idiag_rmem = sk_rmem_alloc_get(sk);
166 		minfo->idiag_wmem = sk->sk_wmem_queued;
167 		minfo->idiag_fmem = sk->sk_forward_alloc;
168 		minfo->idiag_tmem = sk_wmem_alloc_get(sk);
169 	}
170 
171 	handler->idiag_get_info(sk, r, info);
172 
173 	if (sk->sk_state < TCP_TIME_WAIT &&
174 	    icsk->icsk_ca_ops && icsk->icsk_ca_ops->get_info)
175 		icsk->icsk_ca_ops->get_info(sk, ext, skb);
176 
177 	nlh->nlmsg_len = skb_tail_pointer(skb) - b;
178 	return skb->len;
179 
180 rtattr_failure:
181 nlmsg_failure:
182 	nlmsg_trim(skb, b);
183 	return -EMSGSIZE;
184 }
185 
186 static int inet_twsk_diag_fill(struct inet_timewait_sock *tw,
187 			       struct sk_buff *skb, struct inet_diag_req *req,
188 			       u32 pid, u32 seq, u16 nlmsg_flags,
189 			       const struct nlmsghdr *unlh)
190 {
191 	long tmo;
192 	struct inet_diag_msg *r;
193 	const unsigned char *previous_tail = skb_tail_pointer(skb);
194 	struct nlmsghdr *nlh = NLMSG_PUT(skb, pid, seq,
195 					 unlh->nlmsg_type, sizeof(*r));
196 
197 	r = NLMSG_DATA(nlh);
198 	BUG_ON(tw->tw_state != TCP_TIME_WAIT);
199 
200 	nlh->nlmsg_flags = nlmsg_flags;
201 
202 	tmo = tw->tw_ttd - jiffies;
203 	if (tmo < 0)
204 		tmo = 0;
205 
206 	r->idiag_family	      = tw->tw_family;
207 	r->idiag_retrans      = 0;
208 	r->id.idiag_if	      = tw->tw_bound_dev_if;
209 	r->id.idiag_cookie[0] = (u32)(unsigned long)tw;
210 	r->id.idiag_cookie[1] = (u32)(((unsigned long)tw >> 31) >> 1);
211 	r->id.idiag_sport     = tw->tw_sport;
212 	r->id.idiag_dport     = tw->tw_dport;
213 	r->id.idiag_src[0]    = tw->tw_rcv_saddr;
214 	r->id.idiag_dst[0]    = tw->tw_daddr;
215 	r->idiag_state	      = tw->tw_substate;
216 	r->idiag_timer	      = 3;
217 	r->idiag_expires      = DIV_ROUND_UP(tmo * 1000, HZ);
218 	r->idiag_rqueue	      = 0;
219 	r->idiag_wqueue	      = 0;
220 	r->idiag_uid	      = 0;
221 	r->idiag_inode	      = 0;
222 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
223 	if (tw->tw_family == AF_INET6) {
224 		const struct inet6_timewait_sock *tw6 =
225 						inet6_twsk((struct sock *)tw);
226 
227 		*(struct in6_addr *)r->id.idiag_src = tw6->tw_v6_rcv_saddr;
228 		*(struct in6_addr *)r->id.idiag_dst = tw6->tw_v6_daddr;
229 	}
230 #endif
231 	nlh->nlmsg_len = skb_tail_pointer(skb) - previous_tail;
232 	return skb->len;
233 nlmsg_failure:
234 	nlmsg_trim(skb, previous_tail);
235 	return -EMSGSIZE;
236 }
237 
238 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
239 			struct inet_diag_req *r, u32 pid, u32 seq, u16 nlmsg_flags,
240 			const struct nlmsghdr *unlh)
241 {
242 	if (sk->sk_state == TCP_TIME_WAIT)
243 		return inet_twsk_diag_fill((struct inet_timewait_sock *)sk,
244 					   skb, r, pid, seq, nlmsg_flags,
245 					   unlh);
246 	return inet_csk_diag_fill(sk, skb, r, pid, seq, nlmsg_flags, unlh);
247 }
248 
249 static int inet_diag_get_exact(struct sk_buff *in_skb,
250 			       const struct nlmsghdr *nlh,
251 			       struct inet_diag_req *req)
252 {
253 	int err;
254 	struct sock *sk;
255 	struct sk_buff *rep;
256 	struct inet_hashinfo *hashinfo;
257 	const struct inet_diag_handler *handler;
258 
259 	handler = inet_diag_lock_handler(req->sdiag_protocol);
260 	if (IS_ERR(handler)) {
261 		err = PTR_ERR(handler);
262 		goto unlock;
263 	}
264 
265 	hashinfo = handler->idiag_hashinfo;
266 	err = -EINVAL;
267 
268 	if (req->sdiag_family == AF_INET) {
269 		sk = inet_lookup(&init_net, hashinfo, req->id.idiag_dst[0],
270 				 req->id.idiag_dport, req->id.idiag_src[0],
271 				 req->id.idiag_sport, req->id.idiag_if);
272 	}
273 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
274 	else if (req->sdiag_family == AF_INET6) {
275 		sk = inet6_lookup(&init_net, hashinfo,
276 				  (struct in6_addr *)req->id.idiag_dst,
277 				  req->id.idiag_dport,
278 				  (struct in6_addr *)req->id.idiag_src,
279 				  req->id.idiag_sport,
280 				  req->id.idiag_if);
281 	}
282 #endif
283 	else {
284 		goto unlock;
285 	}
286 
287 	err = -ENOENT;
288 	if (sk == NULL)
289 		goto unlock;
290 
291 	err = -ESTALE;
292 	if ((req->id.idiag_cookie[0] != INET_DIAG_NOCOOKIE ||
293 	     req->id.idiag_cookie[1] != INET_DIAG_NOCOOKIE) &&
294 	    ((u32)(unsigned long)sk != req->id.idiag_cookie[0] ||
295 	     (u32)((((unsigned long)sk) >> 31) >> 1) != req->id.idiag_cookie[1]))
296 		goto out;
297 
298 	err = -ENOMEM;
299 	rep = alloc_skb(NLMSG_SPACE((sizeof(struct inet_diag_msg) +
300 				     sizeof(struct inet_diag_meminfo) +
301 				     sizeof(struct tcp_info) + 64)),
302 			GFP_KERNEL);
303 	if (!rep)
304 		goto out;
305 
306 	err = sk_diag_fill(sk, rep, req,
307 			   NETLINK_CB(in_skb).pid,
308 			   nlh->nlmsg_seq, 0, nlh);
309 	if (err < 0) {
310 		WARN_ON(err == -EMSGSIZE);
311 		kfree_skb(rep);
312 		goto out;
313 	}
314 	err = netlink_unicast(sock_diag_nlsk, rep, NETLINK_CB(in_skb).pid,
315 			      MSG_DONTWAIT);
316 	if (err > 0)
317 		err = 0;
318 
319 out:
320 	if (sk) {
321 		if (sk->sk_state == TCP_TIME_WAIT)
322 			inet_twsk_put((struct inet_timewait_sock *)sk);
323 		else
324 			sock_put(sk);
325 	}
326 unlock:
327 	inet_diag_unlock_handler(handler);
328 	return err;
329 }
330 
331 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
332 {
333 	int words = bits >> 5;
334 
335 	bits &= 0x1f;
336 
337 	if (words) {
338 		if (memcmp(a1, a2, words << 2))
339 			return 0;
340 	}
341 	if (bits) {
342 		__be32 w1, w2;
343 		__be32 mask;
344 
345 		w1 = a1[words];
346 		w2 = a2[words];
347 
348 		mask = htonl((0xffffffff) << (32 - bits));
349 
350 		if ((w1 ^ w2) & mask)
351 			return 0;
352 	}
353 
354 	return 1;
355 }
356 
357 
358 static int inet_diag_bc_run(const void *bc, int len,
359 			    const struct inet_diag_entry *entry)
360 {
361 	while (len > 0) {
362 		int yes = 1;
363 		const struct inet_diag_bc_op *op = bc;
364 
365 		switch (op->code) {
366 		case INET_DIAG_BC_NOP:
367 			break;
368 		case INET_DIAG_BC_JMP:
369 			yes = 0;
370 			break;
371 		case INET_DIAG_BC_S_GE:
372 			yes = entry->sport >= op[1].no;
373 			break;
374 		case INET_DIAG_BC_S_LE:
375 			yes = entry->sport <= op[1].no;
376 			break;
377 		case INET_DIAG_BC_D_GE:
378 			yes = entry->dport >= op[1].no;
379 			break;
380 		case INET_DIAG_BC_D_LE:
381 			yes = entry->dport <= op[1].no;
382 			break;
383 		case INET_DIAG_BC_AUTO:
384 			yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
385 			break;
386 		case INET_DIAG_BC_S_COND:
387 		case INET_DIAG_BC_D_COND: {
388 			struct inet_diag_hostcond *cond;
389 			__be32 *addr;
390 
391 			cond = (struct inet_diag_hostcond *)(op + 1);
392 			if (cond->port != -1 &&
393 			    cond->port != (op->code == INET_DIAG_BC_S_COND ?
394 					     entry->sport : entry->dport)) {
395 				yes = 0;
396 				break;
397 			}
398 
399 			if (cond->prefix_len == 0)
400 				break;
401 
402 			if (op->code == INET_DIAG_BC_S_COND)
403 				addr = entry->saddr;
404 			else
405 				addr = entry->daddr;
406 
407 			if (bitstring_match(addr, cond->addr,
408 					    cond->prefix_len))
409 				break;
410 			if (entry->family == AF_INET6 &&
411 			    cond->family == AF_INET) {
412 				if (addr[0] == 0 && addr[1] == 0 &&
413 				    addr[2] == htonl(0xffff) &&
414 				    bitstring_match(addr + 3, cond->addr,
415 						    cond->prefix_len))
416 					break;
417 			}
418 			yes = 0;
419 			break;
420 		}
421 		}
422 
423 		if (yes) {
424 			len -= op->yes;
425 			bc += op->yes;
426 		} else {
427 			len -= op->no;
428 			bc += op->no;
429 		}
430 	}
431 	return len == 0;
432 }
433 
434 static int valid_cc(const void *bc, int len, int cc)
435 {
436 	while (len >= 0) {
437 		const struct inet_diag_bc_op *op = bc;
438 
439 		if (cc > len)
440 			return 0;
441 		if (cc == len)
442 			return 1;
443 		if (op->yes < 4 || op->yes & 3)
444 			return 0;
445 		len -= op->yes;
446 		bc  += op->yes;
447 	}
448 	return 0;
449 }
450 
451 static int inet_diag_bc_audit(const void *bytecode, int bytecode_len)
452 {
453 	const void *bc = bytecode;
454 	int  len = bytecode_len;
455 
456 	while (len > 0) {
457 		const struct inet_diag_bc_op *op = bc;
458 
459 //printk("BC: %d %d %d {%d} / %d\n", op->code, op->yes, op->no, op[1].no, len);
460 		switch (op->code) {
461 		case INET_DIAG_BC_AUTO:
462 		case INET_DIAG_BC_S_COND:
463 		case INET_DIAG_BC_D_COND:
464 		case INET_DIAG_BC_S_GE:
465 		case INET_DIAG_BC_S_LE:
466 		case INET_DIAG_BC_D_GE:
467 		case INET_DIAG_BC_D_LE:
468 		case INET_DIAG_BC_JMP:
469 			if (op->no < 4 || op->no > len + 4 || op->no & 3)
470 				return -EINVAL;
471 			if (op->no < len &&
472 			    !valid_cc(bytecode, bytecode_len, len - op->no))
473 				return -EINVAL;
474 			break;
475 		case INET_DIAG_BC_NOP:
476 			break;
477 		default:
478 			return -EINVAL;
479 		}
480 		if (op->yes < 4 || op->yes > len + 4 || op->yes & 3)
481 			return -EINVAL;
482 		bc  += op->yes;
483 		len -= op->yes;
484 	}
485 	return len == 0 ? 0 : -EINVAL;
486 }
487 
488 static int inet_csk_diag_dump(struct sock *sk,
489 			      struct sk_buff *skb,
490 			      struct netlink_callback *cb,
491 			      struct inet_diag_req *r,
492 			      const struct nlattr *bc)
493 {
494 	if (bc != NULL) {
495 		struct inet_diag_entry entry;
496 		struct inet_sock *inet = inet_sk(sk);
497 
498 		entry.family = sk->sk_family;
499 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
500 		if (entry.family == AF_INET6) {
501 			struct ipv6_pinfo *np = inet6_sk(sk);
502 
503 			entry.saddr = np->rcv_saddr.s6_addr32;
504 			entry.daddr = np->daddr.s6_addr32;
505 		} else
506 #endif
507 		{
508 			entry.saddr = &inet->inet_rcv_saddr;
509 			entry.daddr = &inet->inet_daddr;
510 		}
511 		entry.sport = inet->inet_num;
512 		entry.dport = ntohs(inet->inet_dport);
513 		entry.userlocks = sk->sk_userlocks;
514 
515 		if (!inet_diag_bc_run(nla_data(bc), nla_len(bc), &entry))
516 			return 0;
517 	}
518 
519 	return inet_csk_diag_fill(sk, skb, r,
520 				  NETLINK_CB(cb->skb).pid,
521 				  cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
522 }
523 
524 static int inet_twsk_diag_dump(struct inet_timewait_sock *tw,
525 			       struct sk_buff *skb,
526 			       struct netlink_callback *cb,
527 			       struct inet_diag_req *r,
528 			       const struct nlattr *bc)
529 {
530 	if (bc != NULL) {
531 		struct inet_diag_entry entry;
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(nla_data(bc), nla_len(bc), &entry))
551 			return 0;
552 	}
553 
554 	return inet_twsk_diag_fill(tw, skb, r,
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 		*(struct in6_addr *)r->id.idiag_src = inet6_rsk(req)->loc_addr;
599 		*(struct in6_addr *)r->id.idiag_dst = inet6_rsk(req)->rmt_addr;
600 	}
601 #endif
602 	nlh->nlmsg_len = skb_tail_pointer(skb) - b;
603 
604 	return skb->len;
605 
606 nlmsg_failure:
607 	nlmsg_trim(skb, b);
608 	return -1;
609 }
610 
611 static int inet_diag_dump_reqs(struct sk_buff *skb, struct sock *sk,
612 			       struct netlink_callback *cb,
613 			       struct inet_diag_req *r,
614 			       const struct nlattr *bc)
615 {
616 	struct inet_diag_entry entry;
617 	struct inet_connection_sock *icsk = inet_csk(sk);
618 	struct listen_sock *lopt;
619 	struct inet_sock *inet = inet_sk(sk);
620 	int j, s_j;
621 	int reqnum, s_reqnum;
622 	int err = 0;
623 
624 	s_j = cb->args[3];
625 	s_reqnum = cb->args[4];
626 
627 	if (s_j > 0)
628 		s_j--;
629 
630 	entry.family = sk->sk_family;
631 
632 	read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
633 
634 	lopt = icsk->icsk_accept_queue.listen_opt;
635 	if (!lopt || !lopt->qlen)
636 		goto out;
637 
638 	if (bc != NULL) {
639 		entry.sport = inet->inet_num;
640 		entry.userlocks = sk->sk_userlocks;
641 	}
642 
643 	for (j = s_j; j < lopt->nr_table_entries; j++) {
644 		struct request_sock *req, *head = lopt->syn_table[j];
645 
646 		reqnum = 0;
647 		for (req = head; req; reqnum++, req = req->dl_next) {
648 			struct inet_request_sock *ireq = inet_rsk(req);
649 
650 			if (reqnum < s_reqnum)
651 				continue;
652 			if (r->id.idiag_dport != ireq->rmt_port &&
653 			    r->id.idiag_dport)
654 				continue;
655 
656 			if (bc) {
657 				entry.saddr =
658 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
659 					(entry.family == AF_INET6) ?
660 					inet6_rsk(req)->loc_addr.s6_addr32 :
661 #endif
662 					&ireq->loc_addr;
663 				entry.daddr =
664 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
665 					(entry.family == AF_INET6) ?
666 					inet6_rsk(req)->rmt_addr.s6_addr32 :
667 #endif
668 					&ireq->rmt_addr;
669 				entry.dport = ntohs(ireq->rmt_port);
670 
671 				if (!inet_diag_bc_run(nla_data(bc),
672 						      nla_len(bc), &entry))
673 					continue;
674 			}
675 
676 			err = inet_diag_fill_req(skb, sk, req,
677 					       NETLINK_CB(cb->skb).pid,
678 					       cb->nlh->nlmsg_seq, cb->nlh);
679 			if (err < 0) {
680 				cb->args[3] = j + 1;
681 				cb->args[4] = reqnum;
682 				goto out;
683 			}
684 		}
685 
686 		s_reqnum = 0;
687 	}
688 
689 out:
690 	read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
691 
692 	return err;
693 }
694 
695 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
696 		struct inet_diag_req *r, struct nlattr *bc)
697 {
698 	int i, num;
699 	int s_i, s_num;
700 	const struct inet_diag_handler *handler;
701 	struct inet_hashinfo *hashinfo;
702 
703 	handler = inet_diag_lock_handler(r->sdiag_protocol);
704 	if (IS_ERR(handler))
705 		goto unlock;
706 
707 	hashinfo = handler->idiag_hashinfo;
708 
709 	s_i = cb->args[1];
710 	s_num = num = cb->args[2];
711 
712 	if (cb->args[0] == 0) {
713 		if (!(r->idiag_states & (TCPF_LISTEN | TCPF_SYN_RECV)))
714 			goto skip_listen_ht;
715 
716 		for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
717 			struct sock *sk;
718 			struct hlist_nulls_node *node;
719 			struct inet_listen_hashbucket *ilb;
720 
721 			num = 0;
722 			ilb = &hashinfo->listening_hash[i];
723 			spin_lock_bh(&ilb->lock);
724 			sk_nulls_for_each(sk, node, &ilb->head) {
725 				struct inet_sock *inet = inet_sk(sk);
726 
727 				if (num < s_num) {
728 					num++;
729 					continue;
730 				}
731 
732 				if (r->sdiag_family != AF_UNSPEC &&
733 						sk->sk_family != r->sdiag_family)
734 					goto next_listen;
735 
736 				if (r->id.idiag_sport != inet->inet_sport &&
737 				    r->id.idiag_sport)
738 					goto next_listen;
739 
740 				if (!(r->idiag_states & TCPF_LISTEN) ||
741 				    r->id.idiag_dport ||
742 				    cb->args[3] > 0)
743 					goto syn_recv;
744 
745 				if (inet_csk_diag_dump(sk, skb, cb, r, bc) < 0) {
746 					spin_unlock_bh(&ilb->lock);
747 					goto done;
748 				}
749 
750 syn_recv:
751 				if (!(r->idiag_states & TCPF_SYN_RECV))
752 					goto next_listen;
753 
754 				if (inet_diag_dump_reqs(skb, sk, cb, r, bc) < 0) {
755 					spin_unlock_bh(&ilb->lock);
756 					goto done;
757 				}
758 
759 next_listen:
760 				cb->args[3] = 0;
761 				cb->args[4] = 0;
762 				++num;
763 			}
764 			spin_unlock_bh(&ilb->lock);
765 
766 			s_num = 0;
767 			cb->args[3] = 0;
768 			cb->args[4] = 0;
769 		}
770 skip_listen_ht:
771 		cb->args[0] = 1;
772 		s_i = num = s_num = 0;
773 	}
774 
775 	if (!(r->idiag_states & ~(TCPF_LISTEN | TCPF_SYN_RECV)))
776 		goto unlock;
777 
778 	for (i = s_i; i <= hashinfo->ehash_mask; i++) {
779 		struct inet_ehash_bucket *head = &hashinfo->ehash[i];
780 		spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
781 		struct sock *sk;
782 		struct hlist_nulls_node *node;
783 
784 		num = 0;
785 
786 		if (hlist_nulls_empty(&head->chain) &&
787 			hlist_nulls_empty(&head->twchain))
788 			continue;
789 
790 		if (i > s_i)
791 			s_num = 0;
792 
793 		spin_lock_bh(lock);
794 		sk_nulls_for_each(sk, node, &head->chain) {
795 			struct inet_sock *inet = inet_sk(sk);
796 
797 			if (num < s_num)
798 				goto next_normal;
799 			if (!(r->idiag_states & (1 << sk->sk_state)))
800 				goto next_normal;
801 			if (r->sdiag_family != AF_UNSPEC &&
802 					sk->sk_family != r->sdiag_family)
803 				goto next_normal;
804 			if (r->id.idiag_sport != inet->inet_sport &&
805 			    r->id.idiag_sport)
806 				goto next_normal;
807 			if (r->id.idiag_dport != inet->inet_dport &&
808 			    r->id.idiag_dport)
809 				goto next_normal;
810 			if (inet_csk_diag_dump(sk, skb, cb, r, bc) < 0) {
811 				spin_unlock_bh(lock);
812 				goto done;
813 			}
814 next_normal:
815 			++num;
816 		}
817 
818 		if (r->idiag_states & TCPF_TIME_WAIT) {
819 			struct inet_timewait_sock *tw;
820 
821 			inet_twsk_for_each(tw, node,
822 				    &head->twchain) {
823 
824 				if (num < s_num)
825 					goto next_dying;
826 				if (r->sdiag_family != AF_UNSPEC &&
827 						tw->tw_family != r->sdiag_family)
828 					goto next_dying;
829 				if (r->id.idiag_sport != tw->tw_sport &&
830 				    r->id.idiag_sport)
831 					goto next_dying;
832 				if (r->id.idiag_dport != tw->tw_dport &&
833 				    r->id.idiag_dport)
834 					goto next_dying;
835 				if (inet_twsk_diag_dump(tw, skb, cb, r, bc) < 0) {
836 					spin_unlock_bh(lock);
837 					goto done;
838 				}
839 next_dying:
840 				++num;
841 			}
842 		}
843 		spin_unlock_bh(lock);
844 	}
845 
846 done:
847 	cb->args[1] = i;
848 	cb->args[2] = num;
849 unlock:
850 	inet_diag_unlock_handler(handler);
851 	return skb->len;
852 }
853 
854 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
855 {
856 	struct nlattr *bc = NULL;
857 	int hdrlen = sizeof(struct inet_diag_req);
858 
859 	if (nlmsg_attrlen(cb->nlh, hdrlen))
860 		bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE);
861 
862 	return __inet_diag_dump(skb, cb, (struct inet_diag_req *)NLMSG_DATA(cb->nlh), bc);
863 }
864 
865 static inline int inet_diag_type2proto(int type)
866 {
867 	switch (type) {
868 	case TCPDIAG_GETSOCK:
869 		return IPPROTO_TCP;
870 	case DCCPDIAG_GETSOCK:
871 		return IPPROTO_DCCP;
872 	default:
873 		return 0;
874 	}
875 }
876 
877 static int inet_diag_dump_compat(struct sk_buff *skb, struct netlink_callback *cb)
878 {
879 	struct inet_diag_req_compat *rc = NLMSG_DATA(cb->nlh);
880 	struct inet_diag_req req;
881 	struct nlattr *bc = NULL;
882 	int hdrlen = sizeof(struct inet_diag_req_compat);
883 
884 	req.sdiag_family = AF_UNSPEC; /* compatibility */
885 	req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
886 	req.idiag_ext = rc->idiag_ext;
887 	req.idiag_states = rc->idiag_states;
888 	req.id = rc->id;
889 
890 	if (nlmsg_attrlen(cb->nlh, hdrlen))
891 		bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE);
892 
893 	return __inet_diag_dump(skb, cb, &req, bc);
894 }
895 
896 static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
897 			       const struct nlmsghdr *nlh)
898 {
899 	struct inet_diag_req_compat *rc = NLMSG_DATA(nlh);
900 	struct inet_diag_req req;
901 
902 	req.sdiag_family = rc->idiag_family;
903 	req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
904 	req.idiag_ext = rc->idiag_ext;
905 	req.idiag_states = rc->idiag_states;
906 	req.id = rc->id;
907 
908 	return inet_diag_get_exact(in_skb, nlh, &req);
909 }
910 
911 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
912 {
913 	int hdrlen = sizeof(struct inet_diag_req_compat);
914 
915 	if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
916 	    nlmsg_len(nlh) < hdrlen)
917 		return -EINVAL;
918 
919 	if (nlh->nlmsg_flags & NLM_F_DUMP) {
920 		if (nlmsg_attrlen(nlh, hdrlen)) {
921 			struct nlattr *attr;
922 
923 			attr = nlmsg_find_attr(nlh, hdrlen,
924 					       INET_DIAG_REQ_BYTECODE);
925 			if (attr == NULL ||
926 			    nla_len(attr) < sizeof(struct inet_diag_bc_op) ||
927 			    inet_diag_bc_audit(nla_data(attr), nla_len(attr)))
928 				return -EINVAL;
929 		}
930 
931 		return netlink_dump_start(sock_diag_nlsk, skb, nlh,
932 					  inet_diag_dump_compat, NULL, 0);
933 	}
934 
935 	return inet_diag_get_exact_compat(skb, nlh);
936 }
937 
938 static int inet_diag_handler_dump(struct sk_buff *skb, struct nlmsghdr *h)
939 {
940 	int hdrlen = sizeof(struct inet_diag_req);
941 
942 	if (nlmsg_len(h) < hdrlen)
943 		return -EINVAL;
944 
945 	if (h->nlmsg_flags & NLM_F_DUMP) {
946 		if (nlmsg_attrlen(h, hdrlen)) {
947 			struct nlattr *attr;
948 			attr = nlmsg_find_attr(h, hdrlen,
949 					       INET_DIAG_REQ_BYTECODE);
950 			if (attr == NULL ||
951 			    nla_len(attr) < sizeof(struct inet_diag_bc_op) ||
952 			    inet_diag_bc_audit(nla_data(attr), nla_len(attr)))
953 				return -EINVAL;
954 		}
955 
956 		return netlink_dump_start(sock_diag_nlsk, skb, h,
957 					  inet_diag_dump, NULL, 0);
958 	}
959 
960 	return inet_diag_get_exact(skb, h, (struct inet_diag_req *)NLMSG_DATA(h));
961 }
962 
963 static struct sock_diag_handler inet_diag_handler = {
964 	.family = AF_INET,
965 	.dump = inet_diag_handler_dump,
966 };
967 
968 static struct sock_diag_handler inet6_diag_handler = {
969 	.family = AF_INET6,
970 	.dump = inet_diag_handler_dump,
971 };
972 
973 int inet_diag_register(const struct inet_diag_handler *h)
974 {
975 	const __u16 type = h->idiag_type;
976 	int err = -EINVAL;
977 
978 	if (type >= IPPROTO_MAX)
979 		goto out;
980 
981 	mutex_lock(&inet_diag_table_mutex);
982 	err = -EEXIST;
983 	if (inet_diag_table[type] == NULL) {
984 		inet_diag_table[type] = h;
985 		err = 0;
986 	}
987 	mutex_unlock(&inet_diag_table_mutex);
988 out:
989 	return err;
990 }
991 EXPORT_SYMBOL_GPL(inet_diag_register);
992 
993 void inet_diag_unregister(const struct inet_diag_handler *h)
994 {
995 	const __u16 type = h->idiag_type;
996 
997 	if (type >= IPPROTO_MAX)
998 		return;
999 
1000 	mutex_lock(&inet_diag_table_mutex);
1001 	inet_diag_table[type] = NULL;
1002 	mutex_unlock(&inet_diag_table_mutex);
1003 }
1004 EXPORT_SYMBOL_GPL(inet_diag_unregister);
1005 
1006 static int __init inet_diag_init(void)
1007 {
1008 	const int inet_diag_table_size = (IPPROTO_MAX *
1009 					  sizeof(struct inet_diag_handler *));
1010 	int err = -ENOMEM;
1011 
1012 	inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1013 	if (!inet_diag_table)
1014 		goto out;
1015 
1016 	err = sock_diag_register(&inet_diag_handler);
1017 	if (err)
1018 		goto out_free_nl;
1019 
1020 	err = sock_diag_register(&inet6_diag_handler);
1021 	if (err)
1022 		goto out_free_inet;
1023 
1024 	sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1025 out:
1026 	return err;
1027 
1028 out_free_inet:
1029 	sock_diag_unregister(&inet_diag_handler);
1030 out_free_nl:
1031 	kfree(inet_diag_table);
1032 	goto out;
1033 }
1034 
1035 static void __exit inet_diag_exit(void)
1036 {
1037 	sock_diag_unregister(&inet6_diag_handler);
1038 	sock_diag_unregister(&inet_diag_handler);
1039 	sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1040 	kfree(inet_diag_table);
1041 }
1042 
1043 module_init(inet_diag_init);
1044 module_exit(inet_diag_exit);
1045 MODULE_LICENSE("GPL");
1046 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 0);
1047