Loading...
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
35static const struct inet_diag_handler __rcu **inet_diag_table;
36
37struct 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
51static const struct inet_diag_handler *inet_diag_lock_handler(int proto)
52{
53 const struct inet_diag_handler *handler;
54
55 if (proto < 0 || proto >= IPPROTO_MAX)
56 return NULL;
57
58 if (!READ_ONCE(inet_diag_table[proto]))
59 sock_load_diag_module(AF_INET, proto);
60
61 rcu_read_lock();
62 handler = rcu_dereference(inet_diag_table[proto]);
63 if (handler && !try_module_get(handler->owner))
64 handler = NULL;
65 rcu_read_unlock();
66
67 return handler;
68}
69
70static void inet_diag_unlock_handler(const struct inet_diag_handler *handler)
71{
72 module_put(handler->owner);
73}
74
75void 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}
98EXPORT_SYMBOL_GPL(inet_diag_msg_common_fill);
99
100static 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 rcu_read_lock();
108 handler = rcu_dereference(inet_diag_table[req->sdiag_protocol]);
109 DEBUG_NET_WARN_ON_ONCE(!handler);
110 if (handler && handler->idiag_get_aux_size)
111 aux = handler->idiag_get_aux_size(sk, net_admin);
112 rcu_read_unlock();
113
114 return nla_total_size(sizeof(struct tcp_info))
115 + nla_total_size(sizeof(struct inet_diag_msg))
116 + inet_diag_msg_attrs_size()
117 + nla_total_size(sizeof(struct inet_diag_meminfo))
118 + nla_total_size(SK_MEMINFO_VARS * sizeof(u32))
119 + nla_total_size(TCP_CA_NAME_MAX)
120 + nla_total_size(sizeof(struct tcpvegas_info))
121 + aux
122 + 64;
123}
124
125int inet_diag_msg_attrs_fill(struct sock *sk, struct sk_buff *skb,
126 struct inet_diag_msg *r, int ext,
127 struct user_namespace *user_ns,
128 bool net_admin)
129{
130 const struct inet_sock *inet = inet_sk(sk);
131 struct inet_diag_sockopt inet_sockopt;
132
133 if (nla_put_u8(skb, INET_DIAG_SHUTDOWN, sk->sk_shutdown))
134 goto errout;
135
136 /* IPv6 dual-stack sockets use inet->tos for IPv4 connections,
137 * hence this needs to be included regardless of socket family.
138 */
139 if (ext & (1 << (INET_DIAG_TOS - 1)))
140 if (nla_put_u8(skb, INET_DIAG_TOS, READ_ONCE(inet->tos)) < 0)
141 goto errout;
142
143#if IS_ENABLED(CONFIG_IPV6)
144 if (r->idiag_family == AF_INET6) {
145 if (ext & (1 << (INET_DIAG_TCLASS - 1)))
146 if (nla_put_u8(skb, INET_DIAG_TCLASS,
147 inet6_sk(sk)->tclass) < 0)
148 goto errout;
149
150 if (((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
151 nla_put_u8(skb, INET_DIAG_SKV6ONLY, ipv6_only_sock(sk)))
152 goto errout;
153 }
154#endif
155
156 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK, READ_ONCE(sk->sk_mark)))
157 goto errout;
158
159 if (ext & (1 << (INET_DIAG_CLASS_ID - 1)) ||
160 ext & (1 << (INET_DIAG_TCLASS - 1))) {
161 u32 classid = 0;
162
163#ifdef CONFIG_SOCK_CGROUP_DATA
164 classid = sock_cgroup_classid(&sk->sk_cgrp_data);
165#endif
166 /* Fallback to socket priority if class id isn't set.
167 * Classful qdiscs use it as direct reference to class.
168 * For cgroup2 classid is always zero.
169 */
170 if (!classid)
171 classid = READ_ONCE(sk->sk_priority);
172
173 if (nla_put_u32(skb, INET_DIAG_CLASS_ID, classid))
174 goto errout;
175 }
176
177#ifdef CONFIG_SOCK_CGROUP_DATA
178 if (nla_put_u64_64bit(skb, INET_DIAG_CGROUP_ID,
179 cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)),
180 INET_DIAG_PAD))
181 goto errout;
182#endif
183
184 r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk));
185 r->idiag_inode = sock_i_ino(sk);
186
187 memset(&inet_sockopt, 0, sizeof(inet_sockopt));
188 inet_sockopt.recverr = inet_test_bit(RECVERR, sk);
189 inet_sockopt.is_icsk = inet_test_bit(IS_ICSK, sk);
190 inet_sockopt.freebind = inet_test_bit(FREEBIND, sk);
191 inet_sockopt.hdrincl = inet_test_bit(HDRINCL, sk);
192 inet_sockopt.mc_loop = inet_test_bit(MC_LOOP, sk);
193 inet_sockopt.transparent = inet_test_bit(TRANSPARENT, sk);
194 inet_sockopt.mc_all = inet_test_bit(MC_ALL, sk);
195 inet_sockopt.nodefrag = inet_test_bit(NODEFRAG, sk);
196 inet_sockopt.bind_address_no_port = inet_test_bit(BIND_ADDRESS_NO_PORT, sk);
197 inet_sockopt.recverr_rfc4884 = inet_test_bit(RECVERR_RFC4884, sk);
198 inet_sockopt.defer_connect = inet_test_bit(DEFER_CONNECT, sk);
199 if (nla_put(skb, INET_DIAG_SOCKOPT, sizeof(inet_sockopt),
200 &inet_sockopt))
201 goto errout;
202
203 return 0;
204errout:
205 return 1;
206}
207EXPORT_SYMBOL_GPL(inet_diag_msg_attrs_fill);
208
209static int inet_diag_parse_attrs(const struct nlmsghdr *nlh, int hdrlen,
210 struct nlattr **req_nlas)
211{
212 struct nlattr *nla;
213 int remaining;
214
215 nlmsg_for_each_attr(nla, nlh, hdrlen, remaining) {
216 int type = nla_type(nla);
217
218 if (type == INET_DIAG_REQ_PROTOCOL && nla_len(nla) != sizeof(u32))
219 return -EINVAL;
220
221 if (type < __INET_DIAG_REQ_MAX)
222 req_nlas[type] = nla;
223 }
224 return 0;
225}
226
227static int inet_diag_get_protocol(const struct inet_diag_req_v2 *req,
228 const struct inet_diag_dump_data *data)
229{
230 if (data->req_nlas[INET_DIAG_REQ_PROTOCOL])
231 return nla_get_u32(data->req_nlas[INET_DIAG_REQ_PROTOCOL]);
232 return req->sdiag_protocol;
233}
234
235#define MAX_DUMP_ALLOC_SIZE (KMALLOC_MAX_SIZE - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
236
237int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk,
238 struct sk_buff *skb, struct netlink_callback *cb,
239 const struct inet_diag_req_v2 *req,
240 u16 nlmsg_flags, bool net_admin)
241{
242 const struct tcp_congestion_ops *ca_ops;
243 const struct inet_diag_handler *handler;
244 struct inet_diag_dump_data *cb_data;
245 int ext = req->idiag_ext;
246 struct inet_diag_msg *r;
247 struct nlmsghdr *nlh;
248 struct nlattr *attr;
249 void *info = NULL;
250 int protocol;
251
252 cb_data = cb->data;
253 protocol = inet_diag_get_protocol(req, cb_data);
254
255 /* inet_diag_lock_handler() made sure inet_diag_table[] is stable. */
256 handler = rcu_dereference_protected(inet_diag_table[protocol], 1);
257 DEBUG_NET_WARN_ON_ONCE(!handler);
258 if (!handler)
259 return -ENXIO;
260
261 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
262 cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
263 if (!nlh)
264 return -EMSGSIZE;
265
266 r = nlmsg_data(nlh);
267 BUG_ON(!sk_fullsock(sk));
268
269 inet_diag_msg_common_fill(r, sk);
270 r->idiag_state = sk->sk_state;
271 r->idiag_timer = 0;
272 r->idiag_retrans = 0;
273 r->idiag_expires = 0;
274
275 if (inet_diag_msg_attrs_fill(sk, skb, r, ext,
276 sk_user_ns(NETLINK_CB(cb->skb).sk),
277 net_admin))
278 goto errout;
279
280 if (ext & (1 << (INET_DIAG_MEMINFO - 1))) {
281 struct inet_diag_meminfo minfo = {
282 .idiag_rmem = sk_rmem_alloc_get(sk),
283 .idiag_wmem = READ_ONCE(sk->sk_wmem_queued),
284 .idiag_fmem = sk_forward_alloc_get(sk),
285 .idiag_tmem = sk_wmem_alloc_get(sk),
286 };
287
288 if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0)
289 goto errout;
290 }
291
292 if (ext & (1 << (INET_DIAG_SKMEMINFO - 1)))
293 if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO))
294 goto errout;
295
296 /*
297 * RAW sockets might have user-defined protocols assigned,
298 * so report the one supplied on socket creation.
299 */
300 if (sk->sk_type == SOCK_RAW) {
301 if (nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))
302 goto errout;
303 }
304
305 if (!icsk) {
306 handler->idiag_get_info(sk, r, NULL);
307 goto out;
308 }
309
310 if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
311 icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
312 icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
313 r->idiag_timer = 1;
314 r->idiag_retrans = icsk->icsk_retransmits;
315 r->idiag_expires =
316 jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
317 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
318 r->idiag_timer = 4;
319 r->idiag_retrans = icsk->icsk_probes_out;
320 r->idiag_expires =
321 jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
322 } else if (timer_pending(&sk->sk_timer)) {
323 r->idiag_timer = 2;
324 r->idiag_retrans = icsk->icsk_probes_out;
325 r->idiag_expires =
326 jiffies_delta_to_msecs(sk->sk_timer.expires - jiffies);
327 }
328
329 if ((ext & (1 << (INET_DIAG_INFO - 1))) && handler->idiag_info_size) {
330 attr = nla_reserve_64bit(skb, INET_DIAG_INFO,
331 handler->idiag_info_size,
332 INET_DIAG_PAD);
333 if (!attr)
334 goto errout;
335
336 info = nla_data(attr);
337 }
338
339 if (ext & (1 << (INET_DIAG_CONG - 1))) {
340 int err = 0;
341
342 rcu_read_lock();
343 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
344 if (ca_ops)
345 err = nla_put_string(skb, INET_DIAG_CONG, ca_ops->name);
346 rcu_read_unlock();
347 if (err < 0)
348 goto errout;
349 }
350
351 handler->idiag_get_info(sk, r, info);
352
353 if (ext & (1 << (INET_DIAG_INFO - 1)) && handler->idiag_get_aux)
354 if (handler->idiag_get_aux(sk, net_admin, skb) < 0)
355 goto errout;
356
357 if (sk->sk_state < TCP_TIME_WAIT) {
358 union tcp_cc_info info;
359 size_t sz = 0;
360 int attr;
361
362 rcu_read_lock();
363 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
364 if (ca_ops && ca_ops->get_info)
365 sz = ca_ops->get_info(sk, ext, &attr, &info);
366 rcu_read_unlock();
367 if (sz && nla_put(skb, attr, sz, &info) < 0)
368 goto errout;
369 }
370
371 /* Keep it at the end for potential retry with a larger skb,
372 * or else do best-effort fitting, which is only done for the
373 * first_nlmsg.
374 */
375 if (cb_data->bpf_stg_diag) {
376 bool first_nlmsg = ((unsigned char *)nlh == skb->data);
377 unsigned int prev_min_dump_alloc;
378 unsigned int total_nla_size = 0;
379 unsigned int msg_len;
380 int err;
381
382 msg_len = skb_tail_pointer(skb) - (unsigned char *)nlh;
383 err = bpf_sk_storage_diag_put(cb_data->bpf_stg_diag, sk, skb,
384 INET_DIAG_SK_BPF_STORAGES,
385 &total_nla_size);
386
387 if (!err)
388 goto out;
389
390 total_nla_size += msg_len;
391 prev_min_dump_alloc = cb->min_dump_alloc;
392 if (total_nla_size > prev_min_dump_alloc)
393 cb->min_dump_alloc = min_t(u32, total_nla_size,
394 MAX_DUMP_ALLOC_SIZE);
395
396 if (!first_nlmsg)
397 goto errout;
398
399 if (cb->min_dump_alloc > prev_min_dump_alloc)
400 /* Retry with pskb_expand_head() with
401 * __GFP_DIRECT_RECLAIM
402 */
403 goto errout;
404
405 WARN_ON_ONCE(total_nla_size <= prev_min_dump_alloc);
406
407 /* Send what we have for this sk
408 * and move on to the next sk in the following
409 * dump()
410 */
411 }
412
413out:
414 nlmsg_end(skb, nlh);
415 return 0;
416
417errout:
418 nlmsg_cancel(skb, nlh);
419 return -EMSGSIZE;
420}
421EXPORT_SYMBOL_GPL(inet_sk_diag_fill);
422
423static int inet_twsk_diag_fill(struct sock *sk,
424 struct sk_buff *skb,
425 struct netlink_callback *cb,
426 u16 nlmsg_flags, bool net_admin)
427{
428 struct inet_timewait_sock *tw = inet_twsk(sk);
429 struct inet_diag_msg *r;
430 struct nlmsghdr *nlh;
431 long tmo;
432
433 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
434 cb->nlh->nlmsg_seq, cb->nlh->nlmsg_type,
435 sizeof(*r), nlmsg_flags);
436 if (!nlh)
437 return -EMSGSIZE;
438
439 r = nlmsg_data(nlh);
440 BUG_ON(tw->tw_state != TCP_TIME_WAIT);
441
442 inet_diag_msg_common_fill(r, sk);
443 r->idiag_retrans = 0;
444
445 r->idiag_state = tw->tw_substate;
446 r->idiag_timer = 3;
447 tmo = tw->tw_timer.expires - jiffies;
448 r->idiag_expires = jiffies_delta_to_msecs(tmo);
449 r->idiag_rqueue = 0;
450 r->idiag_wqueue = 0;
451 r->idiag_uid = 0;
452 r->idiag_inode = 0;
453
454 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
455 tw->tw_mark)) {
456 nlmsg_cancel(skb, nlh);
457 return -EMSGSIZE;
458 }
459
460 nlmsg_end(skb, nlh);
461 return 0;
462}
463
464static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb,
465 struct netlink_callback *cb,
466 u16 nlmsg_flags, bool net_admin)
467{
468 struct request_sock *reqsk = inet_reqsk(sk);
469 struct inet_diag_msg *r;
470 struct nlmsghdr *nlh;
471 long tmo;
472
473 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
474 cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
475 if (!nlh)
476 return -EMSGSIZE;
477
478 r = nlmsg_data(nlh);
479 inet_diag_msg_common_fill(r, sk);
480 r->idiag_state = TCP_SYN_RECV;
481 r->idiag_timer = 1;
482 r->idiag_retrans = reqsk->num_retrans;
483
484 BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) !=
485 offsetof(struct sock, sk_cookie));
486
487 tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies;
488 r->idiag_expires = jiffies_delta_to_msecs(tmo);
489 r->idiag_rqueue = 0;
490 r->idiag_wqueue = 0;
491 r->idiag_uid = 0;
492 r->idiag_inode = 0;
493
494 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
495 inet_rsk(reqsk)->ir_mark)) {
496 nlmsg_cancel(skb, nlh);
497 return -EMSGSIZE;
498 }
499
500 nlmsg_end(skb, nlh);
501 return 0;
502}
503
504static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
505 struct netlink_callback *cb,
506 const struct inet_diag_req_v2 *r,
507 u16 nlmsg_flags, bool net_admin)
508{
509 if (sk->sk_state == TCP_TIME_WAIT)
510 return inet_twsk_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
511
512 if (sk->sk_state == TCP_NEW_SYN_RECV)
513 return inet_req_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
514
515 return inet_sk_diag_fill(sk, inet_csk(sk), skb, cb, r, nlmsg_flags,
516 net_admin);
517}
518
519struct sock *inet_diag_find_one_icsk(struct net *net,
520 struct inet_hashinfo *hashinfo,
521 const struct inet_diag_req_v2 *req)
522{
523 struct sock *sk;
524
525 rcu_read_lock();
526 if (req->sdiag_family == AF_INET)
527 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0],
528 req->id.idiag_dport, req->id.idiag_src[0],
529 req->id.idiag_sport, req->id.idiag_if);
530#if IS_ENABLED(CONFIG_IPV6)
531 else if (req->sdiag_family == AF_INET6) {
532 if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) &&
533 ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src))
534 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3],
535 req->id.idiag_dport, req->id.idiag_src[3],
536 req->id.idiag_sport, req->id.idiag_if);
537 else
538 sk = inet6_lookup(net, hashinfo, NULL, 0,
539 (struct in6_addr *)req->id.idiag_dst,
540 req->id.idiag_dport,
541 (struct in6_addr *)req->id.idiag_src,
542 req->id.idiag_sport,
543 req->id.idiag_if);
544 }
545#endif
546 else {
547 rcu_read_unlock();
548 return ERR_PTR(-EINVAL);
549 }
550 rcu_read_unlock();
551 if (!sk)
552 return ERR_PTR(-ENOENT);
553
554 if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) {
555 sock_gen_put(sk);
556 return ERR_PTR(-ENOENT);
557 }
558
559 return sk;
560}
561EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk);
562
563int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo,
564 struct netlink_callback *cb,
565 const struct inet_diag_req_v2 *req)
566{
567 struct sk_buff *in_skb = cb->skb;
568 bool net_admin = netlink_net_capable(in_skb, CAP_NET_ADMIN);
569 struct net *net = sock_net(in_skb->sk);
570 struct sk_buff *rep;
571 struct sock *sk;
572 int err;
573
574 sk = inet_diag_find_one_icsk(net, hashinfo, req);
575 if (IS_ERR(sk))
576 return PTR_ERR(sk);
577
578 rep = nlmsg_new(inet_sk_attr_size(sk, req, net_admin), GFP_KERNEL);
579 if (!rep) {
580 err = -ENOMEM;
581 goto out;
582 }
583
584 err = sk_diag_fill(sk, rep, cb, req, 0, net_admin);
585 if (err < 0) {
586 WARN_ON(err == -EMSGSIZE);
587 nlmsg_free(rep);
588 goto out;
589 }
590 err = nlmsg_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid);
591
592out:
593 if (sk)
594 sock_gen_put(sk);
595
596 return err;
597}
598EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
599
600static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb,
601 const struct nlmsghdr *nlh,
602 int hdrlen,
603 const struct inet_diag_req_v2 *req)
604{
605 const struct inet_diag_handler *handler;
606 struct inet_diag_dump_data dump_data;
607 int err, protocol;
608
609 memset(&dump_data, 0, sizeof(dump_data));
610 err = inet_diag_parse_attrs(nlh, hdrlen, dump_data.req_nlas);
611 if (err)
612 return err;
613
614 protocol = inet_diag_get_protocol(req, &dump_data);
615
616 handler = inet_diag_lock_handler(protocol);
617 if (!handler)
618 return -ENOENT;
619
620 if (cmd == SOCK_DIAG_BY_FAMILY) {
621 struct netlink_callback cb = {
622 .nlh = nlh,
623 .skb = in_skb,
624 .data = &dump_data,
625 };
626 err = handler->dump_one(&cb, req);
627 } else if (cmd == SOCK_DESTROY && handler->destroy) {
628 err = handler->destroy(in_skb, req);
629 } else {
630 err = -EOPNOTSUPP;
631 }
632 inet_diag_unlock_handler(handler);
633
634 return err;
635}
636
637static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
638{
639 int words = bits >> 5;
640
641 bits &= 0x1f;
642
643 if (words) {
644 if (memcmp(a1, a2, words << 2))
645 return 0;
646 }
647 if (bits) {
648 __be32 w1, w2;
649 __be32 mask;
650
651 w1 = a1[words];
652 w2 = a2[words];
653
654 mask = htonl((0xffffffff) << (32 - bits));
655
656 if ((w1 ^ w2) & mask)
657 return 0;
658 }
659
660 return 1;
661}
662
663static int inet_diag_bc_run(const struct nlattr *_bc,
664 const struct inet_diag_entry *entry)
665{
666 const void *bc = nla_data(_bc);
667 int len = nla_len(_bc);
668
669 while (len > 0) {
670 int yes = 1;
671 const struct inet_diag_bc_op *op = bc;
672
673 switch (op->code) {
674 case INET_DIAG_BC_NOP:
675 break;
676 case INET_DIAG_BC_JMP:
677 yes = 0;
678 break;
679 case INET_DIAG_BC_S_EQ:
680 yes = entry->sport == op[1].no;
681 break;
682 case INET_DIAG_BC_S_GE:
683 yes = entry->sport >= op[1].no;
684 break;
685 case INET_DIAG_BC_S_LE:
686 yes = entry->sport <= op[1].no;
687 break;
688 case INET_DIAG_BC_D_EQ:
689 yes = entry->dport == op[1].no;
690 break;
691 case INET_DIAG_BC_D_GE:
692 yes = entry->dport >= op[1].no;
693 break;
694 case INET_DIAG_BC_D_LE:
695 yes = entry->dport <= op[1].no;
696 break;
697 case INET_DIAG_BC_AUTO:
698 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
699 break;
700 case INET_DIAG_BC_S_COND:
701 case INET_DIAG_BC_D_COND: {
702 const struct inet_diag_hostcond *cond;
703 const __be32 *addr;
704
705 cond = (const struct inet_diag_hostcond *)(op + 1);
706 if (cond->port != -1 &&
707 cond->port != (op->code == INET_DIAG_BC_S_COND ?
708 entry->sport : entry->dport)) {
709 yes = 0;
710 break;
711 }
712
713 if (op->code == INET_DIAG_BC_S_COND)
714 addr = entry->saddr;
715 else
716 addr = entry->daddr;
717
718 if (cond->family != AF_UNSPEC &&
719 cond->family != entry->family) {
720 if (entry->family == AF_INET6 &&
721 cond->family == AF_INET) {
722 if (addr[0] == 0 && addr[1] == 0 &&
723 addr[2] == htonl(0xffff) &&
724 bitstring_match(addr + 3,
725 cond->addr,
726 cond->prefix_len))
727 break;
728 }
729 yes = 0;
730 break;
731 }
732
733 if (cond->prefix_len == 0)
734 break;
735 if (bitstring_match(addr, cond->addr,
736 cond->prefix_len))
737 break;
738 yes = 0;
739 break;
740 }
741 case INET_DIAG_BC_DEV_COND: {
742 u32 ifindex;
743
744 ifindex = *((const u32 *)(op + 1));
745 if (ifindex != entry->ifindex)
746 yes = 0;
747 break;
748 }
749 case INET_DIAG_BC_MARK_COND: {
750 struct inet_diag_markcond *cond;
751
752 cond = (struct inet_diag_markcond *)(op + 1);
753 if ((entry->mark & cond->mask) != cond->mark)
754 yes = 0;
755 break;
756 }
757#ifdef CONFIG_SOCK_CGROUP_DATA
758 case INET_DIAG_BC_CGROUP_COND: {
759 u64 cgroup_id;
760
761 cgroup_id = get_unaligned((const u64 *)(op + 1));
762 if (cgroup_id != entry->cgroup_id)
763 yes = 0;
764 break;
765 }
766#endif
767 }
768
769 if (yes) {
770 len -= op->yes;
771 bc += op->yes;
772 } else {
773 len -= op->no;
774 bc += op->no;
775 }
776 }
777 return len == 0;
778}
779
780/* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV)
781 */
782static void entry_fill_addrs(struct inet_diag_entry *entry,
783 const struct sock *sk)
784{
785#if IS_ENABLED(CONFIG_IPV6)
786 if (sk->sk_family == AF_INET6) {
787 entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32;
788 entry->daddr = sk->sk_v6_daddr.s6_addr32;
789 } else
790#endif
791 {
792 entry->saddr = &sk->sk_rcv_saddr;
793 entry->daddr = &sk->sk_daddr;
794 }
795}
796
797int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
798{
799 struct inet_sock *inet = inet_sk(sk);
800 struct inet_diag_entry entry;
801
802 if (!bc)
803 return 1;
804
805 entry.family = sk->sk_family;
806 entry_fill_addrs(&entry, sk);
807 entry.sport = inet->inet_num;
808 entry.dport = ntohs(inet->inet_dport);
809 entry.ifindex = sk->sk_bound_dev_if;
810 entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0;
811 if (sk_fullsock(sk))
812 entry.mark = READ_ONCE(sk->sk_mark);
813 else if (sk->sk_state == TCP_NEW_SYN_RECV)
814 entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark;
815 else if (sk->sk_state == TCP_TIME_WAIT)
816 entry.mark = inet_twsk(sk)->tw_mark;
817 else
818 entry.mark = 0;
819#ifdef CONFIG_SOCK_CGROUP_DATA
820 entry.cgroup_id = sk_fullsock(sk) ?
821 cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)) : 0;
822#endif
823
824 return inet_diag_bc_run(bc, &entry);
825}
826EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
827
828static int valid_cc(const void *bc, int len, int cc)
829{
830 while (len >= 0) {
831 const struct inet_diag_bc_op *op = bc;
832
833 if (cc > len)
834 return 0;
835 if (cc == len)
836 return 1;
837 if (op->yes < 4 || op->yes & 3)
838 return 0;
839 len -= op->yes;
840 bc += op->yes;
841 }
842 return 0;
843}
844
845/* data is u32 ifindex */
846static bool valid_devcond(const struct inet_diag_bc_op *op, int len,
847 int *min_len)
848{
849 /* Check ifindex space. */
850 *min_len += sizeof(u32);
851 if (len < *min_len)
852 return false;
853
854 return true;
855}
856/* Validate an inet_diag_hostcond. */
857static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
858 int *min_len)
859{
860 struct inet_diag_hostcond *cond;
861 int addr_len;
862
863 /* Check hostcond space. */
864 *min_len += sizeof(struct inet_diag_hostcond);
865 if (len < *min_len)
866 return false;
867 cond = (struct inet_diag_hostcond *)(op + 1);
868
869 /* Check address family and address length. */
870 switch (cond->family) {
871 case AF_UNSPEC:
872 addr_len = 0;
873 break;
874 case AF_INET:
875 addr_len = sizeof(struct in_addr);
876 break;
877 case AF_INET6:
878 addr_len = sizeof(struct in6_addr);
879 break;
880 default:
881 return false;
882 }
883 *min_len += addr_len;
884 if (len < *min_len)
885 return false;
886
887 /* Check prefix length (in bits) vs address length (in bytes). */
888 if (cond->prefix_len > 8 * addr_len)
889 return false;
890
891 return true;
892}
893
894/* Validate a port comparison operator. */
895static bool valid_port_comparison(const struct inet_diag_bc_op *op,
896 int len, int *min_len)
897{
898 /* Port comparisons put the port in a follow-on inet_diag_bc_op. */
899 *min_len += sizeof(struct inet_diag_bc_op);
900 if (len < *min_len)
901 return false;
902 return true;
903}
904
905static bool valid_markcond(const struct inet_diag_bc_op *op, int len,
906 int *min_len)
907{
908 *min_len += sizeof(struct inet_diag_markcond);
909 return len >= *min_len;
910}
911
912#ifdef CONFIG_SOCK_CGROUP_DATA
913static bool valid_cgroupcond(const struct inet_diag_bc_op *op, int len,
914 int *min_len)
915{
916 *min_len += sizeof(u64);
917 return len >= *min_len;
918}
919#endif
920
921static int inet_diag_bc_audit(const struct nlattr *attr,
922 const struct sk_buff *skb)
923{
924 bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN);
925 const void *bytecode, *bc;
926 int bytecode_len, len;
927
928 if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op))
929 return -EINVAL;
930
931 bytecode = bc = nla_data(attr);
932 len = bytecode_len = nla_len(attr);
933
934 while (len > 0) {
935 int min_len = sizeof(struct inet_diag_bc_op);
936 const struct inet_diag_bc_op *op = bc;
937
938 switch (op->code) {
939 case INET_DIAG_BC_S_COND:
940 case INET_DIAG_BC_D_COND:
941 if (!valid_hostcond(bc, len, &min_len))
942 return -EINVAL;
943 break;
944 case INET_DIAG_BC_DEV_COND:
945 if (!valid_devcond(bc, len, &min_len))
946 return -EINVAL;
947 break;
948 case INET_DIAG_BC_S_EQ:
949 case INET_DIAG_BC_S_GE:
950 case INET_DIAG_BC_S_LE:
951 case INET_DIAG_BC_D_EQ:
952 case INET_DIAG_BC_D_GE:
953 case INET_DIAG_BC_D_LE:
954 if (!valid_port_comparison(bc, len, &min_len))
955 return -EINVAL;
956 break;
957 case INET_DIAG_BC_MARK_COND:
958 if (!net_admin)
959 return -EPERM;
960 if (!valid_markcond(bc, len, &min_len))
961 return -EINVAL;
962 break;
963#ifdef CONFIG_SOCK_CGROUP_DATA
964 case INET_DIAG_BC_CGROUP_COND:
965 if (!valid_cgroupcond(bc, len, &min_len))
966 return -EINVAL;
967 break;
968#endif
969 case INET_DIAG_BC_AUTO:
970 case INET_DIAG_BC_JMP:
971 case INET_DIAG_BC_NOP:
972 break;
973 default:
974 return -EINVAL;
975 }
976
977 if (op->code != INET_DIAG_BC_NOP) {
978 if (op->no < min_len || op->no > len + 4 || op->no & 3)
979 return -EINVAL;
980 if (op->no < len &&
981 !valid_cc(bytecode, bytecode_len, len - op->no))
982 return -EINVAL;
983 }
984
985 if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
986 return -EINVAL;
987 bc += op->yes;
988 len -= op->yes;
989 }
990 return len == 0 ? 0 : -EINVAL;
991}
992
993static void twsk_build_assert(void)
994{
995 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) !=
996 offsetof(struct sock, sk_family));
997
998 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) !=
999 offsetof(struct inet_sock, inet_num));
1000
1001 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) !=
1002 offsetof(struct inet_sock, inet_dport));
1003
1004 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) !=
1005 offsetof(struct inet_sock, inet_rcv_saddr));
1006
1007 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) !=
1008 offsetof(struct inet_sock, inet_daddr));
1009
1010#if IS_ENABLED(CONFIG_IPV6)
1011 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) !=
1012 offsetof(struct sock, sk_v6_rcv_saddr));
1013
1014 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) !=
1015 offsetof(struct sock, sk_v6_daddr));
1016#endif
1017}
1018
1019void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
1020 struct netlink_callback *cb,
1021 const struct inet_diag_req_v2 *r)
1022{
1023 bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN);
1024 struct inet_diag_dump_data *cb_data = cb->data;
1025 struct net *net = sock_net(skb->sk);
1026 u32 idiag_states = r->idiag_states;
1027 int i, num, s_i, s_num;
1028 struct nlattr *bc;
1029 struct sock *sk;
1030
1031 bc = cb_data->inet_diag_nla_bc;
1032 if (idiag_states & TCPF_SYN_RECV)
1033 idiag_states |= TCPF_NEW_SYN_RECV;
1034 s_i = cb->args[1];
1035 s_num = num = cb->args[2];
1036
1037 if (cb->args[0] == 0) {
1038 if (!(idiag_states & TCPF_LISTEN) || r->id.idiag_dport)
1039 goto skip_listen_ht;
1040
1041 for (i = s_i; i <= hashinfo->lhash2_mask; i++) {
1042 struct inet_listen_hashbucket *ilb;
1043 struct hlist_nulls_node *node;
1044
1045 num = 0;
1046 ilb = &hashinfo->lhash2[i];
1047
1048 if (hlist_nulls_empty(&ilb->nulls_head)) {
1049 s_num = 0;
1050 continue;
1051 }
1052 spin_lock(&ilb->lock);
1053 sk_nulls_for_each(sk, node, &ilb->nulls_head) {
1054 struct inet_sock *inet = inet_sk(sk);
1055
1056 if (!net_eq(sock_net(sk), net))
1057 continue;
1058
1059 if (num < s_num) {
1060 num++;
1061 continue;
1062 }
1063
1064 if (r->sdiag_family != AF_UNSPEC &&
1065 sk->sk_family != r->sdiag_family)
1066 goto next_listen;
1067
1068 if (r->id.idiag_sport != inet->inet_sport &&
1069 r->id.idiag_sport)
1070 goto next_listen;
1071
1072 if (!inet_diag_bc_sk(bc, sk))
1073 goto next_listen;
1074
1075 if (inet_sk_diag_fill(sk, inet_csk(sk), skb,
1076 cb, r, NLM_F_MULTI,
1077 net_admin) < 0) {
1078 spin_unlock(&ilb->lock);
1079 goto done;
1080 }
1081
1082next_listen:
1083 ++num;
1084 }
1085 spin_unlock(&ilb->lock);
1086
1087 s_num = 0;
1088 }
1089skip_listen_ht:
1090 cb->args[0] = 1;
1091 s_i = num = s_num = 0;
1092 }
1093
1094/* Process a maximum of SKARR_SZ sockets at a time when walking hash buckets
1095 * with bh disabled.
1096 */
1097#define SKARR_SZ 16
1098
1099 /* Dump bound but inactive (not listening, connecting, etc.) sockets */
1100 if (cb->args[0] == 1) {
1101 if (!(idiag_states & TCPF_BOUND_INACTIVE))
1102 goto skip_bind_ht;
1103
1104 for (i = s_i; i < hashinfo->bhash_size; i++) {
1105 struct inet_bind_hashbucket *ibb;
1106 struct inet_bind2_bucket *tb2;
1107 struct sock *sk_arr[SKARR_SZ];
1108 int num_arr[SKARR_SZ];
1109 int idx, accum, res;
1110
1111resume_bind_walk:
1112 num = 0;
1113 accum = 0;
1114 ibb = &hashinfo->bhash2[i];
1115
1116 if (hlist_empty(&ibb->chain)) {
1117 s_num = 0;
1118 continue;
1119 }
1120 spin_lock_bh(&ibb->lock);
1121 inet_bind_bucket_for_each(tb2, &ibb->chain) {
1122 if (!net_eq(ib2_net(tb2), net))
1123 continue;
1124
1125 sk_for_each_bound(sk, &tb2->owners) {
1126 struct inet_sock *inet = inet_sk(sk);
1127
1128 if (num < s_num)
1129 goto next_bind;
1130
1131 if (sk->sk_state != TCP_CLOSE ||
1132 !inet->inet_num)
1133 goto next_bind;
1134
1135 if (r->sdiag_family != AF_UNSPEC &&
1136 r->sdiag_family != sk->sk_family)
1137 goto next_bind;
1138
1139 if (!inet_diag_bc_sk(bc, sk))
1140 goto next_bind;
1141
1142 sock_hold(sk);
1143 num_arr[accum] = num;
1144 sk_arr[accum] = sk;
1145 if (++accum == SKARR_SZ)
1146 goto pause_bind_walk;
1147next_bind:
1148 num++;
1149 }
1150 }
1151pause_bind_walk:
1152 spin_unlock_bh(&ibb->lock);
1153
1154 res = 0;
1155 for (idx = 0; idx < accum; idx++) {
1156 if (res >= 0) {
1157 res = inet_sk_diag_fill(sk_arr[idx],
1158 NULL, skb, cb,
1159 r, NLM_F_MULTI,
1160 net_admin);
1161 if (res < 0)
1162 num = num_arr[idx];
1163 }
1164 sock_put(sk_arr[idx]);
1165 }
1166 if (res < 0)
1167 goto done;
1168
1169 cond_resched();
1170
1171 if (accum == SKARR_SZ) {
1172 s_num = num + 1;
1173 goto resume_bind_walk;
1174 }
1175
1176 s_num = 0;
1177 }
1178skip_bind_ht:
1179 cb->args[0] = 2;
1180 s_i = num = s_num = 0;
1181 }
1182
1183 if (!(idiag_states & ~TCPF_LISTEN))
1184 goto out;
1185
1186 for (i = s_i; i <= hashinfo->ehash_mask; i++) {
1187 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
1188 spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
1189 struct hlist_nulls_node *node;
1190 struct sock *sk_arr[SKARR_SZ];
1191 int num_arr[SKARR_SZ];
1192 int idx, accum, res;
1193
1194 if (hlist_nulls_empty(&head->chain))
1195 continue;
1196
1197 if (i > s_i)
1198 s_num = 0;
1199
1200next_chunk:
1201 num = 0;
1202 accum = 0;
1203 spin_lock_bh(lock);
1204 sk_nulls_for_each(sk, node, &head->chain) {
1205 int state;
1206
1207 if (!net_eq(sock_net(sk), net))
1208 continue;
1209 if (num < s_num)
1210 goto next_normal;
1211 state = (sk->sk_state == TCP_TIME_WAIT) ?
1212 inet_twsk(sk)->tw_substate : sk->sk_state;
1213 if (!(idiag_states & (1 << state)))
1214 goto next_normal;
1215 if (r->sdiag_family != AF_UNSPEC &&
1216 sk->sk_family != r->sdiag_family)
1217 goto next_normal;
1218 if (r->id.idiag_sport != htons(sk->sk_num) &&
1219 r->id.idiag_sport)
1220 goto next_normal;
1221 if (r->id.idiag_dport != sk->sk_dport &&
1222 r->id.idiag_dport)
1223 goto next_normal;
1224 twsk_build_assert();
1225
1226 if (!inet_diag_bc_sk(bc, sk))
1227 goto next_normal;
1228
1229 if (!refcount_inc_not_zero(&sk->sk_refcnt))
1230 goto next_normal;
1231
1232 num_arr[accum] = num;
1233 sk_arr[accum] = sk;
1234 if (++accum == SKARR_SZ)
1235 break;
1236next_normal:
1237 ++num;
1238 }
1239 spin_unlock_bh(lock);
1240 res = 0;
1241 for (idx = 0; idx < accum; idx++) {
1242 if (res >= 0) {
1243 res = sk_diag_fill(sk_arr[idx], skb, cb, r,
1244 NLM_F_MULTI, net_admin);
1245 if (res < 0)
1246 num = num_arr[idx];
1247 }
1248 sock_gen_put(sk_arr[idx]);
1249 }
1250 if (res < 0)
1251 break;
1252 cond_resched();
1253 if (accum == SKARR_SZ) {
1254 s_num = num + 1;
1255 goto next_chunk;
1256 }
1257 }
1258
1259done:
1260 cb->args[1] = i;
1261 cb->args[2] = num;
1262out:
1263 ;
1264}
1265EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
1266
1267static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
1268 const struct inet_diag_req_v2 *r)
1269{
1270 struct inet_diag_dump_data *cb_data = cb->data;
1271 const struct inet_diag_handler *handler;
1272 u32 prev_min_dump_alloc;
1273 int protocol, err = 0;
1274
1275 protocol = inet_diag_get_protocol(r, cb_data);
1276
1277again:
1278 prev_min_dump_alloc = cb->min_dump_alloc;
1279 handler = inet_diag_lock_handler(protocol);
1280 if (handler) {
1281 handler->dump(skb, cb, r);
1282 inet_diag_unlock_handler(handler);
1283 } else {
1284 err = -ENOENT;
1285 }
1286 /* The skb is not large enough to fit one sk info and
1287 * inet_sk_diag_fill() has requested for a larger skb.
1288 */
1289 if (!skb->len && cb->min_dump_alloc > prev_min_dump_alloc) {
1290 err = pskb_expand_head(skb, 0, cb->min_dump_alloc, GFP_KERNEL);
1291 if (!err)
1292 goto again;
1293 }
1294
1295 return err ? : skb->len;
1296}
1297
1298static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
1299{
1300 return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh));
1301}
1302
1303static int __inet_diag_dump_start(struct netlink_callback *cb, int hdrlen)
1304{
1305 const struct nlmsghdr *nlh = cb->nlh;
1306 struct inet_diag_dump_data *cb_data;
1307 struct sk_buff *skb = cb->skb;
1308 struct nlattr *nla;
1309 int err;
1310
1311 cb_data = kzalloc(sizeof(*cb_data), GFP_KERNEL);
1312 if (!cb_data)
1313 return -ENOMEM;
1314
1315 err = inet_diag_parse_attrs(nlh, hdrlen, cb_data->req_nlas);
1316 if (err) {
1317 kfree(cb_data);
1318 return err;
1319 }
1320 nla = cb_data->inet_diag_nla_bc;
1321 if (nla) {
1322 err = inet_diag_bc_audit(nla, skb);
1323 if (err) {
1324 kfree(cb_data);
1325 return err;
1326 }
1327 }
1328
1329 nla = cb_data->inet_diag_nla_bpf_stgs;
1330 if (nla) {
1331 struct bpf_sk_storage_diag *bpf_stg_diag;
1332
1333 bpf_stg_diag = bpf_sk_storage_diag_alloc(nla);
1334 if (IS_ERR(bpf_stg_diag)) {
1335 kfree(cb_data);
1336 return PTR_ERR(bpf_stg_diag);
1337 }
1338 cb_data->bpf_stg_diag = bpf_stg_diag;
1339 }
1340
1341 cb->data = cb_data;
1342 return 0;
1343}
1344
1345static int inet_diag_dump_start(struct netlink_callback *cb)
1346{
1347 return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req_v2));
1348}
1349
1350static int inet_diag_dump_start_compat(struct netlink_callback *cb)
1351{
1352 return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req));
1353}
1354
1355static int inet_diag_dump_done(struct netlink_callback *cb)
1356{
1357 struct inet_diag_dump_data *cb_data = cb->data;
1358
1359 bpf_sk_storage_diag_free(cb_data->bpf_stg_diag);
1360 kfree(cb->data);
1361
1362 return 0;
1363}
1364
1365static int inet_diag_type2proto(int type)
1366{
1367 switch (type) {
1368 case TCPDIAG_GETSOCK:
1369 return IPPROTO_TCP;
1370 case DCCPDIAG_GETSOCK:
1371 return IPPROTO_DCCP;
1372 default:
1373 return 0;
1374 }
1375}
1376
1377static int inet_diag_dump_compat(struct sk_buff *skb,
1378 struct netlink_callback *cb)
1379{
1380 struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1381 struct inet_diag_req_v2 req;
1382
1383 req.sdiag_family = AF_UNSPEC; /* compatibility */
1384 req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1385 req.idiag_ext = rc->idiag_ext;
1386 req.idiag_states = rc->idiag_states;
1387 req.id = rc->id;
1388
1389 return __inet_diag_dump(skb, cb, &req);
1390}
1391
1392static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1393 const struct nlmsghdr *nlh)
1394{
1395 struct inet_diag_req *rc = nlmsg_data(nlh);
1396 struct inet_diag_req_v2 req;
1397
1398 req.sdiag_family = rc->idiag_family;
1399 req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1400 req.idiag_ext = rc->idiag_ext;
1401 req.idiag_states = rc->idiag_states;
1402 req.id = rc->id;
1403
1404 return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh,
1405 sizeof(struct inet_diag_req), &req);
1406}
1407
1408static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1409{
1410 int hdrlen = sizeof(struct inet_diag_req);
1411 struct net *net = sock_net(skb->sk);
1412
1413 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1414 nlmsg_len(nlh) < hdrlen)
1415 return -EINVAL;
1416
1417 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1418 struct netlink_dump_control c = {
1419 .start = inet_diag_dump_start_compat,
1420 .done = inet_diag_dump_done,
1421 .dump = inet_diag_dump_compat,
1422 };
1423 return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1424 }
1425
1426 return inet_diag_get_exact_compat(skb, nlh);
1427}
1428
1429static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h)
1430{
1431 int hdrlen = sizeof(struct inet_diag_req_v2);
1432 struct net *net = sock_net(skb->sk);
1433
1434 if (nlmsg_len(h) < hdrlen)
1435 return -EINVAL;
1436
1437 if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY &&
1438 h->nlmsg_flags & NLM_F_DUMP) {
1439 struct netlink_dump_control c = {
1440 .start = inet_diag_dump_start,
1441 .done = inet_diag_dump_done,
1442 .dump = inet_diag_dump,
1443 };
1444 return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1445 }
1446
1447 return inet_diag_cmd_exact(h->nlmsg_type, skb, h, hdrlen,
1448 nlmsg_data(h));
1449}
1450
1451static
1452int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk)
1453{
1454 const struct inet_diag_handler *handler;
1455 struct nlmsghdr *nlh;
1456 struct nlattr *attr;
1457 struct inet_diag_msg *r;
1458 void *info = NULL;
1459 int err = 0;
1460
1461 nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0);
1462 if (!nlh)
1463 return -ENOMEM;
1464
1465 r = nlmsg_data(nlh);
1466 memset(r, 0, sizeof(*r));
1467 inet_diag_msg_common_fill(r, sk);
1468 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM)
1469 r->id.idiag_sport = inet_sk(sk)->inet_sport;
1470 r->idiag_state = sk->sk_state;
1471
1472 if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) {
1473 nlmsg_cancel(skb, nlh);
1474 return err;
1475 }
1476
1477 handler = inet_diag_lock_handler(sk->sk_protocol);
1478 if (!handler) {
1479 nlmsg_cancel(skb, nlh);
1480 return -ENOENT;
1481 }
1482
1483 attr = handler->idiag_info_size
1484 ? nla_reserve_64bit(skb, INET_DIAG_INFO,
1485 handler->idiag_info_size,
1486 INET_DIAG_PAD)
1487 : NULL;
1488 if (attr)
1489 info = nla_data(attr);
1490
1491 handler->idiag_get_info(sk, r, info);
1492 inet_diag_unlock_handler(handler);
1493
1494 nlmsg_end(skb, nlh);
1495 return 0;
1496}
1497
1498static const struct sock_diag_handler inet_diag_handler = {
1499 .owner = THIS_MODULE,
1500 .family = AF_INET,
1501 .dump = inet_diag_handler_cmd,
1502 .get_info = inet_diag_handler_get_info,
1503 .destroy = inet_diag_handler_cmd,
1504};
1505
1506static const struct sock_diag_handler inet6_diag_handler = {
1507 .owner = THIS_MODULE,
1508 .family = AF_INET6,
1509 .dump = inet_diag_handler_cmd,
1510 .get_info = inet_diag_handler_get_info,
1511 .destroy = inet_diag_handler_cmd,
1512};
1513
1514int inet_diag_register(const struct inet_diag_handler *h)
1515{
1516 const __u16 type = h->idiag_type;
1517
1518 if (type >= IPPROTO_MAX)
1519 return -EINVAL;
1520
1521 return !cmpxchg((const struct inet_diag_handler **)&inet_diag_table[type],
1522 NULL, h) ? 0 : -EEXIST;
1523}
1524EXPORT_SYMBOL_GPL(inet_diag_register);
1525
1526void inet_diag_unregister(const struct inet_diag_handler *h)
1527{
1528 const __u16 type = h->idiag_type;
1529
1530 if (type >= IPPROTO_MAX)
1531 return;
1532
1533 xchg((const struct inet_diag_handler **)&inet_diag_table[type],
1534 NULL);
1535}
1536EXPORT_SYMBOL_GPL(inet_diag_unregister);
1537
1538static const struct sock_diag_inet_compat inet_diag_compat = {
1539 .owner = THIS_MODULE,
1540 .fn = inet_diag_rcv_msg_compat,
1541};
1542
1543static int __init inet_diag_init(void)
1544{
1545 const int inet_diag_table_size = (IPPROTO_MAX *
1546 sizeof(struct inet_diag_handler *));
1547 int err = -ENOMEM;
1548
1549 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1550 if (!inet_diag_table)
1551 goto out;
1552
1553 err = sock_diag_register(&inet_diag_handler);
1554 if (err)
1555 goto out_free_nl;
1556
1557 err = sock_diag_register(&inet6_diag_handler);
1558 if (err)
1559 goto out_free_inet;
1560
1561 sock_diag_register_inet_compat(&inet_diag_compat);
1562out:
1563 return err;
1564
1565out_free_inet:
1566 sock_diag_unregister(&inet_diag_handler);
1567out_free_nl:
1568 kfree(inet_diag_table);
1569 goto out;
1570}
1571
1572static void __exit inet_diag_exit(void)
1573{
1574 sock_diag_unregister(&inet6_diag_handler);
1575 sock_diag_unregister(&inet_diag_handler);
1576 sock_diag_unregister_inet_compat(&inet_diag_compat);
1577 kfree(inet_diag_table);
1578}
1579
1580module_init(inet_diag_init);
1581module_exit(inet_diag_exit);
1582MODULE_LICENSE("GPL");
1583MODULE_DESCRIPTION("INET/INET6: socket monitoring via SOCK_DIAG");
1584MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1585MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);
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
37static const struct inet_diag_handler **inet_diag_table;
38
39struct inet_diag_entry {
40 __be32 *saddr;
41 __be32 *daddr;
42 u16 sport;
43 u16 dport;
44 u16 family;
45 u16 userlocks;
46};
47
48static struct sock *idiagnl;
49
50#define INET_DIAG_PUT(skb, attrtype, attrlen) \
51 RTA_DATA(__RTA_PUT(skb, attrtype, attrlen))
52
53static DEFINE_MUTEX(inet_diag_table_mutex);
54
55static const struct inet_diag_handler *inet_diag_lock_handler(int type)
56{
57 if (!inet_diag_table[type])
58 request_module("net-pf-%d-proto-%d-type-%d", PF_NETLINK,
59 NETLINK_INET_DIAG, type);
60
61 mutex_lock(&inet_diag_table_mutex);
62 if (!inet_diag_table[type])
63 return ERR_PTR(-ENOENT);
64
65 return inet_diag_table[type];
66}
67
68static inline void inet_diag_unlock_handler(
69 const struct inet_diag_handler *handler)
70{
71 mutex_unlock(&inet_diag_table_mutex);
72}
73
74static int inet_csk_diag_fill(struct sock *sk,
75 struct sk_buff *skb,
76 int ext, u32 pid, u32 seq, u16 nlmsg_flags,
77 const struct nlmsghdr *unlh)
78{
79 const struct inet_sock *inet = inet_sk(sk);
80 const struct inet_connection_sock *icsk = inet_csk(sk);
81 struct inet_diag_msg *r;
82 struct nlmsghdr *nlh;
83 void *info = NULL;
84 struct inet_diag_meminfo *minfo = NULL;
85 unsigned char *b = skb_tail_pointer(skb);
86 const struct inet_diag_handler *handler;
87
88 handler = inet_diag_table[unlh->nlmsg_type];
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,
102 handler->idiag_info_size);
103
104 if ((ext & (1 << (INET_DIAG_CONG - 1))) && icsk->icsk_ca_ops) {
105 const size_t len = strlen(icsk->icsk_ca_ops->name);
106
107 strcpy(INET_DIAG_PUT(skb, INET_DIAG_CONG, len + 1),
108 icsk->icsk_ca_ops->name);
109 }
110
111 r->idiag_family = sk->sk_family;
112 r->idiag_state = sk->sk_state;
113 r->idiag_timer = 0;
114 r->idiag_retrans = 0;
115
116 r->id.idiag_if = sk->sk_bound_dev_if;
117 r->id.idiag_cookie[0] = (u32)(unsigned long)sk;
118 r->id.idiag_cookie[1] = (u32)(((unsigned long)sk >> 31) >> 1);
119
120 r->id.idiag_sport = inet->inet_sport;
121 r->id.idiag_dport = inet->inet_dport;
122 r->id.idiag_src[0] = inet->inet_rcv_saddr;
123 r->id.idiag_dst[0] = inet->inet_daddr;
124
125#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
126 if (r->idiag_family == AF_INET6) {
127 const struct ipv6_pinfo *np = inet6_sk(sk);
128
129 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
130 &np->rcv_saddr);
131 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
132 &np->daddr);
133 }
134#endif
135
136#define EXPIRES_IN_MS(tmo) DIV_ROUND_UP((tmo - jiffies) * 1000, HZ)
137
138 if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
139 r->idiag_timer = 1;
140 r->idiag_retrans = icsk->icsk_retransmits;
141 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
142 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
143 r->idiag_timer = 4;
144 r->idiag_retrans = icsk->icsk_probes_out;
145 r->idiag_expires = EXPIRES_IN_MS(icsk->icsk_timeout);
146 } else if (timer_pending(&sk->sk_timer)) {
147 r->idiag_timer = 2;
148 r->idiag_retrans = icsk->icsk_probes_out;
149 r->idiag_expires = EXPIRES_IN_MS(sk->sk_timer.expires);
150 } else {
151 r->idiag_timer = 0;
152 r->idiag_expires = 0;
153 }
154#undef EXPIRES_IN_MS
155
156 r->idiag_uid = sock_i_uid(sk);
157 r->idiag_inode = sock_i_ino(sk);
158
159 if (minfo) {
160 minfo->idiag_rmem = sk_rmem_alloc_get(sk);
161 minfo->idiag_wmem = sk->sk_wmem_queued;
162 minfo->idiag_fmem = sk->sk_forward_alloc;
163 minfo->idiag_tmem = sk_wmem_alloc_get(sk);
164 }
165
166 handler->idiag_get_info(sk, r, info);
167
168 if (sk->sk_state < TCP_TIME_WAIT &&
169 icsk->icsk_ca_ops && icsk->icsk_ca_ops->get_info)
170 icsk->icsk_ca_ops->get_info(sk, ext, skb);
171
172 nlh->nlmsg_len = skb_tail_pointer(skb) - b;
173 return skb->len;
174
175rtattr_failure:
176nlmsg_failure:
177 nlmsg_trim(skb, b);
178 return -EMSGSIZE;
179}
180
181static int inet_twsk_diag_fill(struct inet_timewait_sock *tw,
182 struct sk_buff *skb, int ext, u32 pid,
183 u32 seq, u16 nlmsg_flags,
184 const struct nlmsghdr *unlh)
185{
186 long tmo;
187 struct inet_diag_msg *r;
188 const unsigned char *previous_tail = skb_tail_pointer(skb);
189 struct nlmsghdr *nlh = NLMSG_PUT(skb, pid, seq,
190 unlh->nlmsg_type, sizeof(*r));
191
192 r = NLMSG_DATA(nlh);
193 BUG_ON(tw->tw_state != TCP_TIME_WAIT);
194
195 nlh->nlmsg_flags = nlmsg_flags;
196
197 tmo = tw->tw_ttd - jiffies;
198 if (tmo < 0)
199 tmo = 0;
200
201 r->idiag_family = tw->tw_family;
202 r->idiag_retrans = 0;
203 r->id.idiag_if = tw->tw_bound_dev_if;
204 r->id.idiag_cookie[0] = (u32)(unsigned long)tw;
205 r->id.idiag_cookie[1] = (u32)(((unsigned long)tw >> 31) >> 1);
206 r->id.idiag_sport = tw->tw_sport;
207 r->id.idiag_dport = tw->tw_dport;
208 r->id.idiag_src[0] = tw->tw_rcv_saddr;
209 r->id.idiag_dst[0] = tw->tw_daddr;
210 r->idiag_state = tw->tw_substate;
211 r->idiag_timer = 3;
212 r->idiag_expires = DIV_ROUND_UP(tmo * 1000, HZ);
213 r->idiag_rqueue = 0;
214 r->idiag_wqueue = 0;
215 r->idiag_uid = 0;
216 r->idiag_inode = 0;
217#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
218 if (tw->tw_family == AF_INET6) {
219 const struct inet6_timewait_sock *tw6 =
220 inet6_twsk((struct sock *)tw);
221
222 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
223 &tw6->tw_v6_rcv_saddr);
224 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
225 &tw6->tw_v6_daddr);
226 }
227#endif
228 nlh->nlmsg_len = skb_tail_pointer(skb) - previous_tail;
229 return skb->len;
230nlmsg_failure:
231 nlmsg_trim(skb, previous_tail);
232 return -EMSGSIZE;
233}
234
235static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
236 int ext, u32 pid, u32 seq, u16 nlmsg_flags,
237 const struct nlmsghdr *unlh)
238{
239 if (sk->sk_state == TCP_TIME_WAIT)
240 return inet_twsk_diag_fill((struct inet_timewait_sock *)sk,
241 skb, ext, pid, seq, nlmsg_flags,
242 unlh);
243 return inet_csk_diag_fill(sk, skb, ext, pid, seq, nlmsg_flags, unlh);
244}
245
246static int inet_diag_get_exact(struct sk_buff *in_skb,
247 const struct nlmsghdr *nlh)
248{
249 int err;
250 struct sock *sk;
251 struct inet_diag_req *req = NLMSG_DATA(nlh);
252 struct sk_buff *rep;
253 struct inet_hashinfo *hashinfo;
254 const struct inet_diag_handler *handler;
255
256 handler = inet_diag_lock_handler(nlh->nlmsg_type);
257 if (IS_ERR(handler)) {
258 err = PTR_ERR(handler);
259 goto unlock;
260 }
261
262 hashinfo = handler->idiag_hashinfo;
263 err = -EINVAL;
264
265 if (req->idiag_family == AF_INET) {
266 sk = inet_lookup(&init_net, hashinfo, req->id.idiag_dst[0],
267 req->id.idiag_dport, req->id.idiag_src[0],
268 req->id.idiag_sport, req->id.idiag_if);
269 }
270#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
271 else if (req->idiag_family == AF_INET6) {
272 sk = inet6_lookup(&init_net, hashinfo,
273 (struct in6_addr *)req->id.idiag_dst,
274 req->id.idiag_dport,
275 (struct in6_addr *)req->id.idiag_src,
276 req->id.idiag_sport,
277 req->id.idiag_if);
278 }
279#endif
280 else {
281 goto unlock;
282 }
283
284 err = -ENOENT;
285 if (sk == NULL)
286 goto unlock;
287
288 err = -ESTALE;
289 if ((req->id.idiag_cookie[0] != INET_DIAG_NOCOOKIE ||
290 req->id.idiag_cookie[1] != INET_DIAG_NOCOOKIE) &&
291 ((u32)(unsigned long)sk != req->id.idiag_cookie[0] ||
292 (u32)((((unsigned long)sk) >> 31) >> 1) != req->id.idiag_cookie[1]))
293 goto out;
294
295 err = -ENOMEM;
296 rep = alloc_skb(NLMSG_SPACE((sizeof(struct inet_diag_msg) +
297 sizeof(struct inet_diag_meminfo) +
298 handler->idiag_info_size + 64)),
299 GFP_KERNEL);
300 if (!rep)
301 goto out;
302
303 err = sk_diag_fill(sk, rep, req->idiag_ext,
304 NETLINK_CB(in_skb).pid,
305 nlh->nlmsg_seq, 0, nlh);
306 if (err < 0) {
307 WARN_ON(err == -EMSGSIZE);
308 kfree_skb(rep);
309 goto out;
310 }
311 err = netlink_unicast(idiagnl, rep, NETLINK_CB(in_skb).pid,
312 MSG_DONTWAIT);
313 if (err > 0)
314 err = 0;
315
316out:
317 if (sk) {
318 if (sk->sk_state == TCP_TIME_WAIT)
319 inet_twsk_put((struct inet_timewait_sock *)sk);
320 else
321 sock_put(sk);
322 }
323unlock:
324 inet_diag_unlock_handler(handler);
325 return err;
326}
327
328static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
329{
330 int words = bits >> 5;
331
332 bits &= 0x1f;
333
334 if (words) {
335 if (memcmp(a1, a2, words << 2))
336 return 0;
337 }
338 if (bits) {
339 __be32 w1, w2;
340 __be32 mask;
341
342 w1 = a1[words];
343 w2 = a2[words];
344
345 mask = htonl((0xffffffff) << (32 - bits));
346
347 if ((w1 ^ w2) & mask)
348 return 0;
349 }
350
351 return 1;
352}
353
354
355static int inet_diag_bc_run(const void *bc, int len,
356 const struct inet_diag_entry *entry)
357{
358 while (len > 0) {
359 int yes = 1;
360 const struct inet_diag_bc_op *op = bc;
361
362 switch (op->code) {
363 case INET_DIAG_BC_NOP:
364 break;
365 case INET_DIAG_BC_JMP:
366 yes = 0;
367 break;
368 case INET_DIAG_BC_S_GE:
369 yes = entry->sport >= op[1].no;
370 break;
371 case INET_DIAG_BC_S_LE:
372 yes = entry->sport <= op[1].no;
373 break;
374 case INET_DIAG_BC_D_GE:
375 yes = entry->dport >= op[1].no;
376 break;
377 case INET_DIAG_BC_D_LE:
378 yes = entry->dport <= op[1].no;
379 break;
380 case INET_DIAG_BC_AUTO:
381 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
382 break;
383 case INET_DIAG_BC_S_COND:
384 case INET_DIAG_BC_D_COND: {
385 struct inet_diag_hostcond *cond;
386 __be32 *addr;
387
388 cond = (struct inet_diag_hostcond *)(op + 1);
389 if (cond->port != -1 &&
390 cond->port != (op->code == INET_DIAG_BC_S_COND ?
391 entry->sport : entry->dport)) {
392 yes = 0;
393 break;
394 }
395
396 if (cond->prefix_len == 0)
397 break;
398
399 if (op->code == INET_DIAG_BC_S_COND)
400 addr = entry->saddr;
401 else
402 addr = entry->daddr;
403
404 if (bitstring_match(addr, cond->addr,
405 cond->prefix_len))
406 break;
407 if (entry->family == AF_INET6 &&
408 cond->family == AF_INET) {
409 if (addr[0] == 0 && addr[1] == 0 &&
410 addr[2] == htonl(0xffff) &&
411 bitstring_match(addr + 3, cond->addr,
412 cond->prefix_len))
413 break;
414 }
415 yes = 0;
416 break;
417 }
418 }
419
420 if (yes) {
421 len -= op->yes;
422 bc += op->yes;
423 } else {
424 len -= op->no;
425 bc += op->no;
426 }
427 }
428 return len == 0;
429}
430
431static int valid_cc(const void *bc, int len, int cc)
432{
433 while (len >= 0) {
434 const struct inet_diag_bc_op *op = bc;
435
436 if (cc > len)
437 return 0;
438 if (cc == len)
439 return 1;
440 if (op->yes < 4 || op->yes & 3)
441 return 0;
442 len -= op->yes;
443 bc += op->yes;
444 }
445 return 0;
446}
447
448static int inet_diag_bc_audit(const void *bytecode, int bytecode_len)
449{
450 const void *bc = bytecode;
451 int len = bytecode_len;
452
453 while (len > 0) {
454 const struct inet_diag_bc_op *op = bc;
455
456//printk("BC: %d %d %d {%d} / %d\n", op->code, op->yes, op->no, op[1].no, len);
457 switch (op->code) {
458 case INET_DIAG_BC_AUTO:
459 case INET_DIAG_BC_S_COND:
460 case INET_DIAG_BC_D_COND:
461 case INET_DIAG_BC_S_GE:
462 case INET_DIAG_BC_S_LE:
463 case INET_DIAG_BC_D_GE:
464 case INET_DIAG_BC_D_LE:
465 case INET_DIAG_BC_JMP:
466 if (op->no < 4 || op->no > len + 4 || op->no & 3)
467 return -EINVAL;
468 if (op->no < len &&
469 !valid_cc(bytecode, bytecode_len, len - op->no))
470 return -EINVAL;
471 break;
472 case INET_DIAG_BC_NOP:
473 break;
474 default:
475 return -EINVAL;
476 }
477 if (op->yes < 4 || op->yes > len + 4 || op->yes & 3)
478 return -EINVAL;
479 bc += op->yes;
480 len -= op->yes;
481 }
482 return len == 0 ? 0 : -EINVAL;
483}
484
485static int inet_csk_diag_dump(struct sock *sk,
486 struct sk_buff *skb,
487 struct netlink_callback *cb)
488{
489 struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
490
491 if (nlmsg_attrlen(cb->nlh, sizeof(*r))) {
492 struct inet_diag_entry entry;
493 const struct nlattr *bc = nlmsg_find_attr(cb->nlh,
494 sizeof(*r),
495 INET_DIAG_REQ_BYTECODE);
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->idiag_ext,
520 NETLINK_CB(cb->skb).pid,
521 cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
522}
523
524static int inet_twsk_diag_dump(struct inet_timewait_sock *tw,
525 struct sk_buff *skb,
526 struct netlink_callback *cb)
527{
528 struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
529
530 if (nlmsg_attrlen(cb->nlh, sizeof(*r))) {
531 struct inet_diag_entry entry;
532 const struct nlattr *bc = nlmsg_find_attr(cb->nlh,
533 sizeof(*r),
534 INET_DIAG_REQ_BYTECODE);
535
536 entry.family = tw->tw_family;
537#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
538 if (tw->tw_family == AF_INET6) {
539 struct inet6_timewait_sock *tw6 =
540 inet6_twsk((struct sock *)tw);
541 entry.saddr = tw6->tw_v6_rcv_saddr.s6_addr32;
542 entry.daddr = tw6->tw_v6_daddr.s6_addr32;
543 } else
544#endif
545 {
546 entry.saddr = &tw->tw_rcv_saddr;
547 entry.daddr = &tw->tw_daddr;
548 }
549 entry.sport = tw->tw_num;
550 entry.dport = ntohs(tw->tw_dport);
551 entry.userlocks = 0;
552
553 if (!inet_diag_bc_run(nla_data(bc), nla_len(bc), &entry))
554 return 0;
555 }
556
557 return inet_twsk_diag_fill(tw, skb, r->idiag_ext,
558 NETLINK_CB(cb->skb).pid,
559 cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh);
560}
561
562static int inet_diag_fill_req(struct sk_buff *skb, struct sock *sk,
563 struct request_sock *req, u32 pid, u32 seq,
564 const struct nlmsghdr *unlh)
565{
566 const struct inet_request_sock *ireq = inet_rsk(req);
567 struct inet_sock *inet = inet_sk(sk);
568 unsigned char *b = skb_tail_pointer(skb);
569 struct inet_diag_msg *r;
570 struct nlmsghdr *nlh;
571 long tmo;
572
573 nlh = NLMSG_PUT(skb, pid, seq, unlh->nlmsg_type, sizeof(*r));
574 nlh->nlmsg_flags = NLM_F_MULTI;
575 r = NLMSG_DATA(nlh);
576
577 r->idiag_family = sk->sk_family;
578 r->idiag_state = TCP_SYN_RECV;
579 r->idiag_timer = 1;
580 r->idiag_retrans = req->retrans;
581
582 r->id.idiag_if = sk->sk_bound_dev_if;
583 r->id.idiag_cookie[0] = (u32)(unsigned long)req;
584 r->id.idiag_cookie[1] = (u32)(((unsigned long)req >> 31) >> 1);
585
586 tmo = req->expires - jiffies;
587 if (tmo < 0)
588 tmo = 0;
589
590 r->id.idiag_sport = inet->inet_sport;
591 r->id.idiag_dport = ireq->rmt_port;
592 r->id.idiag_src[0] = ireq->loc_addr;
593 r->id.idiag_dst[0] = ireq->rmt_addr;
594 r->idiag_expires = jiffies_to_msecs(tmo);
595 r->idiag_rqueue = 0;
596 r->idiag_wqueue = 0;
597 r->idiag_uid = sock_i_uid(sk);
598 r->idiag_inode = 0;
599#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
600 if (r->idiag_family == AF_INET6) {
601 ipv6_addr_copy((struct in6_addr *)r->id.idiag_src,
602 &inet6_rsk(req)->loc_addr);
603 ipv6_addr_copy((struct in6_addr *)r->id.idiag_dst,
604 &inet6_rsk(req)->rmt_addr);
605 }
606#endif
607 nlh->nlmsg_len = skb_tail_pointer(skb) - b;
608
609 return skb->len;
610
611nlmsg_failure:
612 nlmsg_trim(skb, b);
613 return -1;
614}
615
616static int inet_diag_dump_reqs(struct sk_buff *skb, struct sock *sk,
617 struct netlink_callback *cb)
618{
619 struct inet_diag_entry entry;
620 struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
621 struct inet_connection_sock *icsk = inet_csk(sk);
622 struct listen_sock *lopt;
623 const struct nlattr *bc = NULL;
624 struct inet_sock *inet = inet_sk(sk);
625 int j, s_j;
626 int reqnum, s_reqnum;
627 int err = 0;
628
629 s_j = cb->args[3];
630 s_reqnum = cb->args[4];
631
632 if (s_j > 0)
633 s_j--;
634
635 entry.family = sk->sk_family;
636
637 read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
638
639 lopt = icsk->icsk_accept_queue.listen_opt;
640 if (!lopt || !lopt->qlen)
641 goto out;
642
643 if (nlmsg_attrlen(cb->nlh, sizeof(*r))) {
644 bc = nlmsg_find_attr(cb->nlh, sizeof(*r),
645 INET_DIAG_REQ_BYTECODE);
646 entry.sport = inet->inet_num;
647 entry.userlocks = sk->sk_userlocks;
648 }
649
650 for (j = s_j; j < lopt->nr_table_entries; j++) {
651 struct request_sock *req, *head = lopt->syn_table[j];
652
653 reqnum = 0;
654 for (req = head; req; reqnum++, req = req->dl_next) {
655 struct inet_request_sock *ireq = inet_rsk(req);
656
657 if (reqnum < s_reqnum)
658 continue;
659 if (r->id.idiag_dport != ireq->rmt_port &&
660 r->id.idiag_dport)
661 continue;
662
663 if (bc) {
664 entry.saddr =
665#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
666 (entry.family == AF_INET6) ?
667 inet6_rsk(req)->loc_addr.s6_addr32 :
668#endif
669 &ireq->loc_addr;
670 entry.daddr =
671#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
672 (entry.family == AF_INET6) ?
673 inet6_rsk(req)->rmt_addr.s6_addr32 :
674#endif
675 &ireq->rmt_addr;
676 entry.dport = ntohs(ireq->rmt_port);
677
678 if (!inet_diag_bc_run(nla_data(bc),
679 nla_len(bc), &entry))
680 continue;
681 }
682
683 err = inet_diag_fill_req(skb, sk, req,
684 NETLINK_CB(cb->skb).pid,
685 cb->nlh->nlmsg_seq, cb->nlh);
686 if (err < 0) {
687 cb->args[3] = j + 1;
688 cb->args[4] = reqnum;
689 goto out;
690 }
691 }
692
693 s_reqnum = 0;
694 }
695
696out:
697 read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
698
699 return err;
700}
701
702static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
703{
704 int i, num;
705 int s_i, s_num;
706 struct inet_diag_req *r = NLMSG_DATA(cb->nlh);
707 const struct inet_diag_handler *handler;
708 struct inet_hashinfo *hashinfo;
709
710 handler = inet_diag_lock_handler(cb->nlh->nlmsg_type);
711 if (IS_ERR(handler))
712 goto unlock;
713
714 hashinfo = handler->idiag_hashinfo;
715
716 s_i = cb->args[1];
717 s_num = num = cb->args[2];
718
719 if (cb->args[0] == 0) {
720 if (!(r->idiag_states & (TCPF_LISTEN | TCPF_SYN_RECV)))
721 goto skip_listen_ht;
722
723 for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
724 struct sock *sk;
725 struct hlist_nulls_node *node;
726 struct inet_listen_hashbucket *ilb;
727
728 num = 0;
729 ilb = &hashinfo->listening_hash[i];
730 spin_lock_bh(&ilb->lock);
731 sk_nulls_for_each(sk, node, &ilb->head) {
732 struct inet_sock *inet = inet_sk(sk);
733
734 if (num < s_num) {
735 num++;
736 continue;
737 }
738
739 if (r->id.idiag_sport != inet->inet_sport &&
740 r->id.idiag_sport)
741 goto next_listen;
742
743 if (!(r->idiag_states & TCPF_LISTEN) ||
744 r->id.idiag_dport ||
745 cb->args[3] > 0)
746 goto syn_recv;
747
748 if (inet_csk_diag_dump(sk, skb, cb) < 0) {
749 spin_unlock_bh(&ilb->lock);
750 goto done;
751 }
752
753syn_recv:
754 if (!(r->idiag_states & TCPF_SYN_RECV))
755 goto next_listen;
756
757 if (inet_diag_dump_reqs(skb, sk, cb) < 0) {
758 spin_unlock_bh(&ilb->lock);
759 goto done;
760 }
761
762next_listen:
763 cb->args[3] = 0;
764 cb->args[4] = 0;
765 ++num;
766 }
767 spin_unlock_bh(&ilb->lock);
768
769 s_num = 0;
770 cb->args[3] = 0;
771 cb->args[4] = 0;
772 }
773skip_listen_ht:
774 cb->args[0] = 1;
775 s_i = num = s_num = 0;
776 }
777
778 if (!(r->idiag_states & ~(TCPF_LISTEN | TCPF_SYN_RECV)))
779 goto unlock;
780
781 for (i = s_i; i <= hashinfo->ehash_mask; i++) {
782 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
783 spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
784 struct sock *sk;
785 struct hlist_nulls_node *node;
786
787 num = 0;
788
789 if (hlist_nulls_empty(&head->chain) &&
790 hlist_nulls_empty(&head->twchain))
791 continue;
792
793 if (i > s_i)
794 s_num = 0;
795
796 spin_lock_bh(lock);
797 sk_nulls_for_each(sk, node, &head->chain) {
798 struct inet_sock *inet = inet_sk(sk);
799
800 if (num < s_num)
801 goto next_normal;
802 if (!(r->idiag_states & (1 << sk->sk_state)))
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) < 0) {
811 spin_unlock_bh(lock);
812 goto done;
813 }
814next_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->id.idiag_sport != tw->tw_sport &&
827 r->id.idiag_sport)
828 goto next_dying;
829 if (r->id.idiag_dport != tw->tw_dport &&
830 r->id.idiag_dport)
831 goto next_dying;
832 if (inet_twsk_diag_dump(tw, skb, cb) < 0) {
833 spin_unlock_bh(lock);
834 goto done;
835 }
836next_dying:
837 ++num;
838 }
839 }
840 spin_unlock_bh(lock);
841 }
842
843done:
844 cb->args[1] = i;
845 cb->args[2] = num;
846unlock:
847 inet_diag_unlock_handler(handler);
848 return skb->len;
849}
850
851static int inet_diag_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
852{
853 int hdrlen = sizeof(struct inet_diag_req);
854
855 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
856 nlmsg_len(nlh) < hdrlen)
857 return -EINVAL;
858
859 if (nlh->nlmsg_flags & NLM_F_DUMP) {
860 if (nlmsg_attrlen(nlh, hdrlen)) {
861 struct nlattr *attr;
862
863 attr = nlmsg_find_attr(nlh, hdrlen,
864 INET_DIAG_REQ_BYTECODE);
865 if (attr == NULL ||
866 nla_len(attr) < sizeof(struct inet_diag_bc_op) ||
867 inet_diag_bc_audit(nla_data(attr), nla_len(attr)))
868 return -EINVAL;
869 }
870
871 return netlink_dump_start(idiagnl, skb, nlh,
872 inet_diag_dump, NULL, 0);
873 }
874
875 return inet_diag_get_exact(skb, nlh);
876}
877
878static DEFINE_MUTEX(inet_diag_mutex);
879
880static void inet_diag_rcv(struct sk_buff *skb)
881{
882 mutex_lock(&inet_diag_mutex);
883 netlink_rcv_skb(skb, &inet_diag_rcv_msg);
884 mutex_unlock(&inet_diag_mutex);
885}
886
887int inet_diag_register(const struct inet_diag_handler *h)
888{
889 const __u16 type = h->idiag_type;
890 int err = -EINVAL;
891
892 if (type >= INET_DIAG_GETSOCK_MAX)
893 goto out;
894
895 mutex_lock(&inet_diag_table_mutex);
896 err = -EEXIST;
897 if (inet_diag_table[type] == NULL) {
898 inet_diag_table[type] = h;
899 err = 0;
900 }
901 mutex_unlock(&inet_diag_table_mutex);
902out:
903 return err;
904}
905EXPORT_SYMBOL_GPL(inet_diag_register);
906
907void inet_diag_unregister(const struct inet_diag_handler *h)
908{
909 const __u16 type = h->idiag_type;
910
911 if (type >= INET_DIAG_GETSOCK_MAX)
912 return;
913
914 mutex_lock(&inet_diag_table_mutex);
915 inet_diag_table[type] = NULL;
916 mutex_unlock(&inet_diag_table_mutex);
917}
918EXPORT_SYMBOL_GPL(inet_diag_unregister);
919
920static int __init inet_diag_init(void)
921{
922 const int inet_diag_table_size = (INET_DIAG_GETSOCK_MAX *
923 sizeof(struct inet_diag_handler *));
924 int err = -ENOMEM;
925
926 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
927 if (!inet_diag_table)
928 goto out;
929
930 idiagnl = netlink_kernel_create(&init_net, NETLINK_INET_DIAG, 0,
931 inet_diag_rcv, NULL, THIS_MODULE);
932 if (idiagnl == NULL)
933 goto out_free_table;
934 err = 0;
935out:
936 return err;
937out_free_table:
938 kfree(inet_diag_table);
939 goto out;
940}
941
942static void __exit inet_diag_exit(void)
943{
944 netlink_kernel_release(idiagnl);
945 kfree(inet_diag_table);
946}
947
948module_init(inet_diag_init);
949module_exit(inet_diag_exit);
950MODULE_LICENSE("GPL");
951MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK, NETLINK_INET_DIAG);