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 u8 icsk_pending;
251 int protocol;
252
253 cb_data = cb->data;
254 protocol = inet_diag_get_protocol(req, cb_data);
255
256 /* inet_diag_lock_handler() made sure inet_diag_table[] is stable. */
257 handler = rcu_dereference_protected(inet_diag_table[protocol], 1);
258 DEBUG_NET_WARN_ON_ONCE(!handler);
259 if (!handler)
260 return -ENXIO;
261
262 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
263 cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
264 if (!nlh)
265 return -EMSGSIZE;
266
267 r = nlmsg_data(nlh);
268 BUG_ON(!sk_fullsock(sk));
269
270 inet_diag_msg_common_fill(r, sk);
271 r->idiag_state = sk->sk_state;
272 r->idiag_timer = 0;
273 r->idiag_retrans = 0;
274 r->idiag_expires = 0;
275
276 if (inet_diag_msg_attrs_fill(sk, skb, r, ext,
277 sk_user_ns(NETLINK_CB(cb->skb).sk),
278 net_admin))
279 goto errout;
280
281 if (ext & (1 << (INET_DIAG_MEMINFO - 1))) {
282 struct inet_diag_meminfo minfo = {
283 .idiag_rmem = sk_rmem_alloc_get(sk),
284 .idiag_wmem = READ_ONCE(sk->sk_wmem_queued),
285 .idiag_fmem = sk_forward_alloc_get(sk),
286 .idiag_tmem = sk_wmem_alloc_get(sk),
287 };
288
289 if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0)
290 goto errout;
291 }
292
293 if (ext & (1 << (INET_DIAG_SKMEMINFO - 1)))
294 if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO))
295 goto errout;
296
297 /*
298 * RAW sockets might have user-defined protocols assigned,
299 * so report the one supplied on socket creation.
300 */
301 if (sk->sk_type == SOCK_RAW) {
302 if (nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))
303 goto errout;
304 }
305
306 if (!icsk) {
307 handler->idiag_get_info(sk, r, NULL);
308 goto out;
309 }
310
311 icsk_pending = smp_load_acquire(&icsk->icsk_pending);
312 if (icsk_pending == ICSK_TIME_RETRANS ||
313 icsk_pending == ICSK_TIME_REO_TIMEOUT ||
314 icsk_pending == ICSK_TIME_LOSS_PROBE) {
315 r->idiag_timer = 1;
316 r->idiag_retrans = icsk->icsk_retransmits;
317 r->idiag_expires =
318 jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
319 } else if (icsk_pending == ICSK_TIME_PROBE0) {
320 r->idiag_timer = 4;
321 r->idiag_retrans = icsk->icsk_probes_out;
322 r->idiag_expires =
323 jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
324 } else if (timer_pending(&sk->sk_timer)) {
325 r->idiag_timer = 2;
326 r->idiag_retrans = icsk->icsk_probes_out;
327 r->idiag_expires =
328 jiffies_delta_to_msecs(sk->sk_timer.expires - jiffies);
329 }
330
331 if ((ext & (1 << (INET_DIAG_INFO - 1))) && handler->idiag_info_size) {
332 attr = nla_reserve_64bit(skb, INET_DIAG_INFO,
333 handler->idiag_info_size,
334 INET_DIAG_PAD);
335 if (!attr)
336 goto errout;
337
338 info = nla_data(attr);
339 }
340
341 if (ext & (1 << (INET_DIAG_CONG - 1))) {
342 int err = 0;
343
344 rcu_read_lock();
345 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
346 if (ca_ops)
347 err = nla_put_string(skb, INET_DIAG_CONG, ca_ops->name);
348 rcu_read_unlock();
349 if (err < 0)
350 goto errout;
351 }
352
353 handler->idiag_get_info(sk, r, info);
354
355 if (ext & (1 << (INET_DIAG_INFO - 1)) && handler->idiag_get_aux)
356 if (handler->idiag_get_aux(sk, net_admin, skb) < 0)
357 goto errout;
358
359 if (sk->sk_state < TCP_TIME_WAIT) {
360 union tcp_cc_info info;
361 size_t sz = 0;
362 int attr;
363
364 rcu_read_lock();
365 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
366 if (ca_ops && ca_ops->get_info)
367 sz = ca_ops->get_info(sk, ext, &attr, &info);
368 rcu_read_unlock();
369 if (sz && nla_put(skb, attr, sz, &info) < 0)
370 goto errout;
371 }
372
373 /* Keep it at the end for potential retry with a larger skb,
374 * or else do best-effort fitting, which is only done for the
375 * first_nlmsg.
376 */
377 if (cb_data->bpf_stg_diag) {
378 bool first_nlmsg = ((unsigned char *)nlh == skb->data);
379 unsigned int prev_min_dump_alloc;
380 unsigned int total_nla_size = 0;
381 unsigned int msg_len;
382 int err;
383
384 msg_len = skb_tail_pointer(skb) - (unsigned char *)nlh;
385 err = bpf_sk_storage_diag_put(cb_data->bpf_stg_diag, sk, skb,
386 INET_DIAG_SK_BPF_STORAGES,
387 &total_nla_size);
388
389 if (!err)
390 goto out;
391
392 total_nla_size += msg_len;
393 prev_min_dump_alloc = cb->min_dump_alloc;
394 if (total_nla_size > prev_min_dump_alloc)
395 cb->min_dump_alloc = min_t(u32, total_nla_size,
396 MAX_DUMP_ALLOC_SIZE);
397
398 if (!first_nlmsg)
399 goto errout;
400
401 if (cb->min_dump_alloc > prev_min_dump_alloc)
402 /* Retry with pskb_expand_head() with
403 * __GFP_DIRECT_RECLAIM
404 */
405 goto errout;
406
407 WARN_ON_ONCE(total_nla_size <= prev_min_dump_alloc);
408
409 /* Send what we have for this sk
410 * and move on to the next sk in the following
411 * dump()
412 */
413 }
414
415out:
416 nlmsg_end(skb, nlh);
417 return 0;
418
419errout:
420 nlmsg_cancel(skb, nlh);
421 return -EMSGSIZE;
422}
423EXPORT_SYMBOL_GPL(inet_sk_diag_fill);
424
425static int inet_twsk_diag_fill(struct sock *sk,
426 struct sk_buff *skb,
427 struct netlink_callback *cb,
428 u16 nlmsg_flags, bool net_admin)
429{
430 struct inet_timewait_sock *tw = inet_twsk(sk);
431 struct inet_diag_msg *r;
432 struct nlmsghdr *nlh;
433 long tmo;
434
435 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
436 cb->nlh->nlmsg_seq, cb->nlh->nlmsg_type,
437 sizeof(*r), nlmsg_flags);
438 if (!nlh)
439 return -EMSGSIZE;
440
441 r = nlmsg_data(nlh);
442 BUG_ON(tw->tw_state != TCP_TIME_WAIT);
443
444 inet_diag_msg_common_fill(r, sk);
445 r->idiag_retrans = 0;
446
447 r->idiag_state = READ_ONCE(tw->tw_substate);
448 r->idiag_timer = 3;
449 tmo = tw->tw_timer.expires - jiffies;
450 r->idiag_expires = jiffies_delta_to_msecs(tmo);
451 r->idiag_rqueue = 0;
452 r->idiag_wqueue = 0;
453 r->idiag_uid = 0;
454 r->idiag_inode = 0;
455
456 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
457 tw->tw_mark)) {
458 nlmsg_cancel(skb, nlh);
459 return -EMSGSIZE;
460 }
461
462 nlmsg_end(skb, nlh);
463 return 0;
464}
465
466static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb,
467 struct netlink_callback *cb,
468 u16 nlmsg_flags, bool net_admin)
469{
470 struct request_sock *reqsk = inet_reqsk(sk);
471 struct inet_diag_msg *r;
472 struct nlmsghdr *nlh;
473 long tmo;
474
475 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
476 cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
477 if (!nlh)
478 return -EMSGSIZE;
479
480 r = nlmsg_data(nlh);
481 inet_diag_msg_common_fill(r, sk);
482 r->idiag_state = TCP_SYN_RECV;
483 r->idiag_timer = 1;
484 r->idiag_retrans = reqsk->num_retrans;
485
486 BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) !=
487 offsetof(struct sock, sk_cookie));
488
489 tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies;
490 r->idiag_expires = jiffies_delta_to_msecs(tmo);
491 r->idiag_rqueue = 0;
492 r->idiag_wqueue = 0;
493 r->idiag_uid = 0;
494 r->idiag_inode = 0;
495
496 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
497 inet_rsk(reqsk)->ir_mark)) {
498 nlmsg_cancel(skb, nlh);
499 return -EMSGSIZE;
500 }
501
502 nlmsg_end(skb, nlh);
503 return 0;
504}
505
506static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
507 struct netlink_callback *cb,
508 const struct inet_diag_req_v2 *r,
509 u16 nlmsg_flags, bool net_admin)
510{
511 if (sk->sk_state == TCP_TIME_WAIT)
512 return inet_twsk_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
513
514 if (sk->sk_state == TCP_NEW_SYN_RECV)
515 return inet_req_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
516
517 return inet_sk_diag_fill(sk, inet_csk(sk), skb, cb, r, nlmsg_flags,
518 net_admin);
519}
520
521struct sock *inet_diag_find_one_icsk(struct net *net,
522 struct inet_hashinfo *hashinfo,
523 const struct inet_diag_req_v2 *req)
524{
525 struct sock *sk;
526
527 rcu_read_lock();
528 if (req->sdiag_family == AF_INET)
529 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0],
530 req->id.idiag_dport, req->id.idiag_src[0],
531 req->id.idiag_sport, req->id.idiag_if);
532#if IS_ENABLED(CONFIG_IPV6)
533 else if (req->sdiag_family == AF_INET6) {
534 if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) &&
535 ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src))
536 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3],
537 req->id.idiag_dport, req->id.idiag_src[3],
538 req->id.idiag_sport, req->id.idiag_if);
539 else
540 sk = inet6_lookup(net, hashinfo, NULL, 0,
541 (struct in6_addr *)req->id.idiag_dst,
542 req->id.idiag_dport,
543 (struct in6_addr *)req->id.idiag_src,
544 req->id.idiag_sport,
545 req->id.idiag_if);
546 }
547#endif
548 else {
549 rcu_read_unlock();
550 return ERR_PTR(-EINVAL);
551 }
552 rcu_read_unlock();
553 if (!sk)
554 return ERR_PTR(-ENOENT);
555
556 if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) {
557 sock_gen_put(sk);
558 return ERR_PTR(-ENOENT);
559 }
560
561 return sk;
562}
563EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk);
564
565int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo,
566 struct netlink_callback *cb,
567 const struct inet_diag_req_v2 *req)
568{
569 struct sk_buff *in_skb = cb->skb;
570 bool net_admin = netlink_net_capable(in_skb, CAP_NET_ADMIN);
571 struct net *net = sock_net(in_skb->sk);
572 struct sk_buff *rep;
573 struct sock *sk;
574 int err;
575
576 sk = inet_diag_find_one_icsk(net, hashinfo, req);
577 if (IS_ERR(sk))
578 return PTR_ERR(sk);
579
580 rep = nlmsg_new(inet_sk_attr_size(sk, req, net_admin), GFP_KERNEL);
581 if (!rep) {
582 err = -ENOMEM;
583 goto out;
584 }
585
586 err = sk_diag_fill(sk, rep, cb, req, 0, net_admin);
587 if (err < 0) {
588 WARN_ON(err == -EMSGSIZE);
589 nlmsg_free(rep);
590 goto out;
591 }
592 err = nlmsg_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid);
593
594out:
595 if (sk)
596 sock_gen_put(sk);
597
598 return err;
599}
600EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
601
602static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb,
603 const struct nlmsghdr *nlh,
604 int hdrlen,
605 const struct inet_diag_req_v2 *req)
606{
607 const struct inet_diag_handler *handler;
608 struct inet_diag_dump_data dump_data;
609 int err, protocol;
610
611 memset(&dump_data, 0, sizeof(dump_data));
612 err = inet_diag_parse_attrs(nlh, hdrlen, dump_data.req_nlas);
613 if (err)
614 return err;
615
616 protocol = inet_diag_get_protocol(req, &dump_data);
617
618 handler = inet_diag_lock_handler(protocol);
619 if (!handler)
620 return -ENOENT;
621
622 if (cmd == SOCK_DIAG_BY_FAMILY) {
623 struct netlink_callback cb = {
624 .nlh = nlh,
625 .skb = in_skb,
626 .data = &dump_data,
627 };
628 err = handler->dump_one(&cb, req);
629 } else if (cmd == SOCK_DESTROY && handler->destroy) {
630 err = handler->destroy(in_skb, req);
631 } else {
632 err = -EOPNOTSUPP;
633 }
634 inet_diag_unlock_handler(handler);
635
636 return err;
637}
638
639static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
640{
641 int words = bits >> 5;
642
643 bits &= 0x1f;
644
645 if (words) {
646 if (memcmp(a1, a2, words << 2))
647 return 0;
648 }
649 if (bits) {
650 __be32 w1, w2;
651 __be32 mask;
652
653 w1 = a1[words];
654 w2 = a2[words];
655
656 mask = htonl((0xffffffff) << (32 - bits));
657
658 if ((w1 ^ w2) & mask)
659 return 0;
660 }
661
662 return 1;
663}
664
665static int inet_diag_bc_run(const struct nlattr *_bc,
666 const struct inet_diag_entry *entry)
667{
668 const void *bc = nla_data(_bc);
669 int len = nla_len(_bc);
670
671 while (len > 0) {
672 int yes = 1;
673 const struct inet_diag_bc_op *op = bc;
674
675 switch (op->code) {
676 case INET_DIAG_BC_NOP:
677 break;
678 case INET_DIAG_BC_JMP:
679 yes = 0;
680 break;
681 case INET_DIAG_BC_S_EQ:
682 yes = entry->sport == op[1].no;
683 break;
684 case INET_DIAG_BC_S_GE:
685 yes = entry->sport >= op[1].no;
686 break;
687 case INET_DIAG_BC_S_LE:
688 yes = entry->sport <= op[1].no;
689 break;
690 case INET_DIAG_BC_D_EQ:
691 yes = entry->dport == op[1].no;
692 break;
693 case INET_DIAG_BC_D_GE:
694 yes = entry->dport >= op[1].no;
695 break;
696 case INET_DIAG_BC_D_LE:
697 yes = entry->dport <= op[1].no;
698 break;
699 case INET_DIAG_BC_AUTO:
700 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
701 break;
702 case INET_DIAG_BC_S_COND:
703 case INET_DIAG_BC_D_COND: {
704 const struct inet_diag_hostcond *cond;
705 const __be32 *addr;
706
707 cond = (const struct inet_diag_hostcond *)(op + 1);
708 if (cond->port != -1 &&
709 cond->port != (op->code == INET_DIAG_BC_S_COND ?
710 entry->sport : entry->dport)) {
711 yes = 0;
712 break;
713 }
714
715 if (op->code == INET_DIAG_BC_S_COND)
716 addr = entry->saddr;
717 else
718 addr = entry->daddr;
719
720 if (cond->family != AF_UNSPEC &&
721 cond->family != entry->family) {
722 if (entry->family == AF_INET6 &&
723 cond->family == AF_INET) {
724 if (addr[0] == 0 && addr[1] == 0 &&
725 addr[2] == htonl(0xffff) &&
726 bitstring_match(addr + 3,
727 cond->addr,
728 cond->prefix_len))
729 break;
730 }
731 yes = 0;
732 break;
733 }
734
735 if (cond->prefix_len == 0)
736 break;
737 if (bitstring_match(addr, cond->addr,
738 cond->prefix_len))
739 break;
740 yes = 0;
741 break;
742 }
743 case INET_DIAG_BC_DEV_COND: {
744 u32 ifindex;
745
746 ifindex = *((const u32 *)(op + 1));
747 if (ifindex != entry->ifindex)
748 yes = 0;
749 break;
750 }
751 case INET_DIAG_BC_MARK_COND: {
752 struct inet_diag_markcond *cond;
753
754 cond = (struct inet_diag_markcond *)(op + 1);
755 if ((entry->mark & cond->mask) != cond->mark)
756 yes = 0;
757 break;
758 }
759#ifdef CONFIG_SOCK_CGROUP_DATA
760 case INET_DIAG_BC_CGROUP_COND: {
761 u64 cgroup_id;
762
763 cgroup_id = get_unaligned((const u64 *)(op + 1));
764 if (cgroup_id != entry->cgroup_id)
765 yes = 0;
766 break;
767 }
768#endif
769 }
770
771 if (yes) {
772 len -= op->yes;
773 bc += op->yes;
774 } else {
775 len -= op->no;
776 bc += op->no;
777 }
778 }
779 return len == 0;
780}
781
782/* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV)
783 */
784static void entry_fill_addrs(struct inet_diag_entry *entry,
785 const struct sock *sk)
786{
787#if IS_ENABLED(CONFIG_IPV6)
788 if (sk->sk_family == AF_INET6) {
789 entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32;
790 entry->daddr = sk->sk_v6_daddr.s6_addr32;
791 } else
792#endif
793 {
794 entry->saddr = &sk->sk_rcv_saddr;
795 entry->daddr = &sk->sk_daddr;
796 }
797}
798
799int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
800{
801 struct inet_sock *inet = inet_sk(sk);
802 struct inet_diag_entry entry;
803
804 if (!bc)
805 return 1;
806
807 entry.family = sk->sk_family;
808 entry_fill_addrs(&entry, sk);
809 entry.sport = inet->inet_num;
810 entry.dport = ntohs(inet->inet_dport);
811 entry.ifindex = sk->sk_bound_dev_if;
812 entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0;
813 if (sk_fullsock(sk))
814 entry.mark = READ_ONCE(sk->sk_mark);
815 else if (sk->sk_state == TCP_NEW_SYN_RECV)
816 entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark;
817 else if (sk->sk_state == TCP_TIME_WAIT)
818 entry.mark = inet_twsk(sk)->tw_mark;
819 else
820 entry.mark = 0;
821#ifdef CONFIG_SOCK_CGROUP_DATA
822 entry.cgroup_id = sk_fullsock(sk) ?
823 cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)) : 0;
824#endif
825
826 return inet_diag_bc_run(bc, &entry);
827}
828EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
829
830static int valid_cc(const void *bc, int len, int cc)
831{
832 while (len >= 0) {
833 const struct inet_diag_bc_op *op = bc;
834
835 if (cc > len)
836 return 0;
837 if (cc == len)
838 return 1;
839 if (op->yes < 4 || op->yes & 3)
840 return 0;
841 len -= op->yes;
842 bc += op->yes;
843 }
844 return 0;
845}
846
847/* data is u32 ifindex */
848static bool valid_devcond(const struct inet_diag_bc_op *op, int len,
849 int *min_len)
850{
851 /* Check ifindex space. */
852 *min_len += sizeof(u32);
853 if (len < *min_len)
854 return false;
855
856 return true;
857}
858/* Validate an inet_diag_hostcond. */
859static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
860 int *min_len)
861{
862 struct inet_diag_hostcond *cond;
863 int addr_len;
864
865 /* Check hostcond space. */
866 *min_len += sizeof(struct inet_diag_hostcond);
867 if (len < *min_len)
868 return false;
869 cond = (struct inet_diag_hostcond *)(op + 1);
870
871 /* Check address family and address length. */
872 switch (cond->family) {
873 case AF_UNSPEC:
874 addr_len = 0;
875 break;
876 case AF_INET:
877 addr_len = sizeof(struct in_addr);
878 break;
879 case AF_INET6:
880 addr_len = sizeof(struct in6_addr);
881 break;
882 default:
883 return false;
884 }
885 *min_len += addr_len;
886 if (len < *min_len)
887 return false;
888
889 /* Check prefix length (in bits) vs address length (in bytes). */
890 if (cond->prefix_len > 8 * addr_len)
891 return false;
892
893 return true;
894}
895
896/* Validate a port comparison operator. */
897static bool valid_port_comparison(const struct inet_diag_bc_op *op,
898 int len, int *min_len)
899{
900 /* Port comparisons put the port in a follow-on inet_diag_bc_op. */
901 *min_len += sizeof(struct inet_diag_bc_op);
902 if (len < *min_len)
903 return false;
904 return true;
905}
906
907static bool valid_markcond(const struct inet_diag_bc_op *op, int len,
908 int *min_len)
909{
910 *min_len += sizeof(struct inet_diag_markcond);
911 return len >= *min_len;
912}
913
914#ifdef CONFIG_SOCK_CGROUP_DATA
915static bool valid_cgroupcond(const struct inet_diag_bc_op *op, int len,
916 int *min_len)
917{
918 *min_len += sizeof(u64);
919 return len >= *min_len;
920}
921#endif
922
923static int inet_diag_bc_audit(const struct nlattr *attr,
924 const struct sk_buff *skb)
925{
926 bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN);
927 const void *bytecode, *bc;
928 int bytecode_len, len;
929
930 if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op))
931 return -EINVAL;
932
933 bytecode = bc = nla_data(attr);
934 len = bytecode_len = nla_len(attr);
935
936 while (len > 0) {
937 int min_len = sizeof(struct inet_diag_bc_op);
938 const struct inet_diag_bc_op *op = bc;
939
940 switch (op->code) {
941 case INET_DIAG_BC_S_COND:
942 case INET_DIAG_BC_D_COND:
943 if (!valid_hostcond(bc, len, &min_len))
944 return -EINVAL;
945 break;
946 case INET_DIAG_BC_DEV_COND:
947 if (!valid_devcond(bc, len, &min_len))
948 return -EINVAL;
949 break;
950 case INET_DIAG_BC_S_EQ:
951 case INET_DIAG_BC_S_GE:
952 case INET_DIAG_BC_S_LE:
953 case INET_DIAG_BC_D_EQ:
954 case INET_DIAG_BC_D_GE:
955 case INET_DIAG_BC_D_LE:
956 if (!valid_port_comparison(bc, len, &min_len))
957 return -EINVAL;
958 break;
959 case INET_DIAG_BC_MARK_COND:
960 if (!net_admin)
961 return -EPERM;
962 if (!valid_markcond(bc, len, &min_len))
963 return -EINVAL;
964 break;
965#ifdef CONFIG_SOCK_CGROUP_DATA
966 case INET_DIAG_BC_CGROUP_COND:
967 if (!valid_cgroupcond(bc, len, &min_len))
968 return -EINVAL;
969 break;
970#endif
971 case INET_DIAG_BC_AUTO:
972 case INET_DIAG_BC_JMP:
973 case INET_DIAG_BC_NOP:
974 break;
975 default:
976 return -EINVAL;
977 }
978
979 if (op->code != INET_DIAG_BC_NOP) {
980 if (op->no < min_len || op->no > len + 4 || op->no & 3)
981 return -EINVAL;
982 if (op->no < len &&
983 !valid_cc(bytecode, bytecode_len, len - op->no))
984 return -EINVAL;
985 }
986
987 if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
988 return -EINVAL;
989 bc += op->yes;
990 len -= op->yes;
991 }
992 return len == 0 ? 0 : -EINVAL;
993}
994
995static void twsk_build_assert(void)
996{
997 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) !=
998 offsetof(struct sock, sk_family));
999
1000 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) !=
1001 offsetof(struct inet_sock, inet_num));
1002
1003 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) !=
1004 offsetof(struct inet_sock, inet_dport));
1005
1006 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) !=
1007 offsetof(struct inet_sock, inet_rcv_saddr));
1008
1009 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) !=
1010 offsetof(struct inet_sock, inet_daddr));
1011
1012#if IS_ENABLED(CONFIG_IPV6)
1013 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) !=
1014 offsetof(struct sock, sk_v6_rcv_saddr));
1015
1016 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) !=
1017 offsetof(struct sock, sk_v6_daddr));
1018#endif
1019}
1020
1021void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
1022 struct netlink_callback *cb,
1023 const struct inet_diag_req_v2 *r)
1024{
1025 bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN);
1026 struct inet_diag_dump_data *cb_data = cb->data;
1027 struct net *net = sock_net(skb->sk);
1028 u32 idiag_states = r->idiag_states;
1029 int i, num, s_i, s_num;
1030 struct nlattr *bc;
1031 struct sock *sk;
1032
1033 bc = cb_data->inet_diag_nla_bc;
1034 if (idiag_states & TCPF_SYN_RECV)
1035 idiag_states |= TCPF_NEW_SYN_RECV;
1036 s_i = cb->args[1];
1037 s_num = num = cb->args[2];
1038
1039 if (cb->args[0] == 0) {
1040 if (!(idiag_states & TCPF_LISTEN) || r->id.idiag_dport)
1041 goto skip_listen_ht;
1042
1043 for (i = s_i; i <= hashinfo->lhash2_mask; i++) {
1044 struct inet_listen_hashbucket *ilb;
1045 struct hlist_nulls_node *node;
1046
1047 num = 0;
1048 ilb = &hashinfo->lhash2[i];
1049
1050 if (hlist_nulls_empty(&ilb->nulls_head)) {
1051 s_num = 0;
1052 continue;
1053 }
1054 spin_lock(&ilb->lock);
1055 sk_nulls_for_each(sk, node, &ilb->nulls_head) {
1056 struct inet_sock *inet = inet_sk(sk);
1057
1058 if (!net_eq(sock_net(sk), net))
1059 continue;
1060
1061 if (num < s_num) {
1062 num++;
1063 continue;
1064 }
1065
1066 if (r->sdiag_family != AF_UNSPEC &&
1067 sk->sk_family != r->sdiag_family)
1068 goto next_listen;
1069
1070 if (r->id.idiag_sport != inet->inet_sport &&
1071 r->id.idiag_sport)
1072 goto next_listen;
1073
1074 if (!inet_diag_bc_sk(bc, sk))
1075 goto next_listen;
1076
1077 if (inet_sk_diag_fill(sk, inet_csk(sk), skb,
1078 cb, r, NLM_F_MULTI,
1079 net_admin) < 0) {
1080 spin_unlock(&ilb->lock);
1081 goto done;
1082 }
1083
1084next_listen:
1085 ++num;
1086 }
1087 spin_unlock(&ilb->lock);
1088
1089 s_num = 0;
1090 }
1091skip_listen_ht:
1092 cb->args[0] = 1;
1093 s_i = num = s_num = 0;
1094 }
1095
1096/* Process a maximum of SKARR_SZ sockets at a time when walking hash buckets
1097 * with bh disabled.
1098 */
1099#define SKARR_SZ 16
1100
1101 /* Dump bound but inactive (not listening, connecting, etc.) sockets */
1102 if (cb->args[0] == 1) {
1103 if (!(idiag_states & TCPF_BOUND_INACTIVE))
1104 goto skip_bind_ht;
1105
1106 for (i = s_i; i < hashinfo->bhash_size; i++) {
1107 struct inet_bind_hashbucket *ibb;
1108 struct inet_bind2_bucket *tb2;
1109 struct sock *sk_arr[SKARR_SZ];
1110 int num_arr[SKARR_SZ];
1111 int idx, accum, res;
1112
1113resume_bind_walk:
1114 num = 0;
1115 accum = 0;
1116 ibb = &hashinfo->bhash2[i];
1117
1118 if (hlist_empty(&ibb->chain)) {
1119 s_num = 0;
1120 continue;
1121 }
1122 spin_lock_bh(&ibb->lock);
1123 inet_bind_bucket_for_each(tb2, &ibb->chain) {
1124 if (!net_eq(ib2_net(tb2), net))
1125 continue;
1126
1127 sk_for_each_bound(sk, &tb2->owners) {
1128 struct inet_sock *inet = inet_sk(sk);
1129
1130 if (num < s_num)
1131 goto next_bind;
1132
1133 if (sk->sk_state != TCP_CLOSE ||
1134 !inet->inet_num)
1135 goto next_bind;
1136
1137 if (r->sdiag_family != AF_UNSPEC &&
1138 r->sdiag_family != sk->sk_family)
1139 goto next_bind;
1140
1141 if (!inet_diag_bc_sk(bc, sk))
1142 goto next_bind;
1143
1144 sock_hold(sk);
1145 num_arr[accum] = num;
1146 sk_arr[accum] = sk;
1147 if (++accum == SKARR_SZ)
1148 goto pause_bind_walk;
1149next_bind:
1150 num++;
1151 }
1152 }
1153pause_bind_walk:
1154 spin_unlock_bh(&ibb->lock);
1155
1156 res = 0;
1157 for (idx = 0; idx < accum; idx++) {
1158 if (res >= 0) {
1159 res = inet_sk_diag_fill(sk_arr[idx],
1160 NULL, skb, cb,
1161 r, NLM_F_MULTI,
1162 net_admin);
1163 if (res < 0)
1164 num = num_arr[idx];
1165 }
1166 sock_put(sk_arr[idx]);
1167 }
1168 if (res < 0)
1169 goto done;
1170
1171 cond_resched();
1172
1173 if (accum == SKARR_SZ) {
1174 s_num = num + 1;
1175 goto resume_bind_walk;
1176 }
1177
1178 s_num = 0;
1179 }
1180skip_bind_ht:
1181 cb->args[0] = 2;
1182 s_i = num = s_num = 0;
1183 }
1184
1185 if (!(idiag_states & ~TCPF_LISTEN))
1186 goto out;
1187
1188 for (i = s_i; i <= hashinfo->ehash_mask; i++) {
1189 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
1190 spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
1191 struct hlist_nulls_node *node;
1192 struct sock *sk_arr[SKARR_SZ];
1193 int num_arr[SKARR_SZ];
1194 int idx, accum, res;
1195
1196 if (hlist_nulls_empty(&head->chain))
1197 continue;
1198
1199 if (i > s_i)
1200 s_num = 0;
1201
1202next_chunk:
1203 num = 0;
1204 accum = 0;
1205 spin_lock_bh(lock);
1206 sk_nulls_for_each(sk, node, &head->chain) {
1207 int state;
1208
1209 if (!net_eq(sock_net(sk), net))
1210 continue;
1211 if (num < s_num)
1212 goto next_normal;
1213 state = (sk->sk_state == TCP_TIME_WAIT) ?
1214 READ_ONCE(inet_twsk(sk)->tw_substate) : sk->sk_state;
1215 if (!(idiag_states & (1 << state)))
1216 goto next_normal;
1217 if (r->sdiag_family != AF_UNSPEC &&
1218 sk->sk_family != r->sdiag_family)
1219 goto next_normal;
1220 if (r->id.idiag_sport != htons(sk->sk_num) &&
1221 r->id.idiag_sport)
1222 goto next_normal;
1223 if (r->id.idiag_dport != sk->sk_dport &&
1224 r->id.idiag_dport)
1225 goto next_normal;
1226 twsk_build_assert();
1227
1228 if (!inet_diag_bc_sk(bc, sk))
1229 goto next_normal;
1230
1231 if (!refcount_inc_not_zero(&sk->sk_refcnt))
1232 goto next_normal;
1233
1234 num_arr[accum] = num;
1235 sk_arr[accum] = sk;
1236 if (++accum == SKARR_SZ)
1237 break;
1238next_normal:
1239 ++num;
1240 }
1241 spin_unlock_bh(lock);
1242 res = 0;
1243 for (idx = 0; idx < accum; idx++) {
1244 if (res >= 0) {
1245 res = sk_diag_fill(sk_arr[idx], skb, cb, r,
1246 NLM_F_MULTI, net_admin);
1247 if (res < 0)
1248 num = num_arr[idx];
1249 }
1250 sock_gen_put(sk_arr[idx]);
1251 }
1252 if (res < 0)
1253 break;
1254 cond_resched();
1255 if (accum == SKARR_SZ) {
1256 s_num = num + 1;
1257 goto next_chunk;
1258 }
1259 }
1260
1261done:
1262 cb->args[1] = i;
1263 cb->args[2] = num;
1264out:
1265 ;
1266}
1267EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
1268
1269static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
1270 const struct inet_diag_req_v2 *r)
1271{
1272 struct inet_diag_dump_data *cb_data = cb->data;
1273 const struct inet_diag_handler *handler;
1274 u32 prev_min_dump_alloc;
1275 int protocol, err = 0;
1276
1277 protocol = inet_diag_get_protocol(r, cb_data);
1278
1279again:
1280 prev_min_dump_alloc = cb->min_dump_alloc;
1281 handler = inet_diag_lock_handler(protocol);
1282 if (handler) {
1283 handler->dump(skb, cb, r);
1284 inet_diag_unlock_handler(handler);
1285 } else {
1286 err = -ENOENT;
1287 }
1288 /* The skb is not large enough to fit one sk info and
1289 * inet_sk_diag_fill() has requested for a larger skb.
1290 */
1291 if (!skb->len && cb->min_dump_alloc > prev_min_dump_alloc) {
1292 err = pskb_expand_head(skb, 0, cb->min_dump_alloc, GFP_KERNEL);
1293 if (!err)
1294 goto again;
1295 }
1296
1297 return err ? : skb->len;
1298}
1299
1300static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
1301{
1302 return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh));
1303}
1304
1305static int __inet_diag_dump_start(struct netlink_callback *cb, int hdrlen)
1306{
1307 const struct nlmsghdr *nlh = cb->nlh;
1308 struct inet_diag_dump_data *cb_data;
1309 struct sk_buff *skb = cb->skb;
1310 struct nlattr *nla;
1311 int err;
1312
1313 cb_data = kzalloc(sizeof(*cb_data), GFP_KERNEL);
1314 if (!cb_data)
1315 return -ENOMEM;
1316
1317 err = inet_diag_parse_attrs(nlh, hdrlen, cb_data->req_nlas);
1318 if (err) {
1319 kfree(cb_data);
1320 return err;
1321 }
1322 nla = cb_data->inet_diag_nla_bc;
1323 if (nla) {
1324 err = inet_diag_bc_audit(nla, skb);
1325 if (err) {
1326 kfree(cb_data);
1327 return err;
1328 }
1329 }
1330
1331 nla = cb_data->inet_diag_nla_bpf_stgs;
1332 if (nla) {
1333 struct bpf_sk_storage_diag *bpf_stg_diag;
1334
1335 bpf_stg_diag = bpf_sk_storage_diag_alloc(nla);
1336 if (IS_ERR(bpf_stg_diag)) {
1337 kfree(cb_data);
1338 return PTR_ERR(bpf_stg_diag);
1339 }
1340 cb_data->bpf_stg_diag = bpf_stg_diag;
1341 }
1342
1343 cb->data = cb_data;
1344 return 0;
1345}
1346
1347static int inet_diag_dump_start(struct netlink_callback *cb)
1348{
1349 return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req_v2));
1350}
1351
1352static int inet_diag_dump_start_compat(struct netlink_callback *cb)
1353{
1354 return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req));
1355}
1356
1357static int inet_diag_dump_done(struct netlink_callback *cb)
1358{
1359 struct inet_diag_dump_data *cb_data = cb->data;
1360
1361 bpf_sk_storage_diag_free(cb_data->bpf_stg_diag);
1362 kfree(cb->data);
1363
1364 return 0;
1365}
1366
1367static int inet_diag_type2proto(int type)
1368{
1369 switch (type) {
1370 case TCPDIAG_GETSOCK:
1371 return IPPROTO_TCP;
1372 case DCCPDIAG_GETSOCK:
1373 return IPPROTO_DCCP;
1374 default:
1375 return 0;
1376 }
1377}
1378
1379static int inet_diag_dump_compat(struct sk_buff *skb,
1380 struct netlink_callback *cb)
1381{
1382 struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1383 struct inet_diag_req_v2 req;
1384
1385 req.sdiag_family = AF_UNSPEC; /* compatibility */
1386 req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1387 req.idiag_ext = rc->idiag_ext;
1388 req.pad = 0;
1389 req.idiag_states = rc->idiag_states;
1390 req.id = rc->id;
1391
1392 return __inet_diag_dump(skb, cb, &req);
1393}
1394
1395static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1396 const struct nlmsghdr *nlh)
1397{
1398 struct inet_diag_req *rc = nlmsg_data(nlh);
1399 struct inet_diag_req_v2 req;
1400
1401 req.sdiag_family = rc->idiag_family;
1402 req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1403 req.idiag_ext = rc->idiag_ext;
1404 req.pad = 0;
1405 req.idiag_states = rc->idiag_states;
1406 req.id = rc->id;
1407
1408 return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh,
1409 sizeof(struct inet_diag_req), &req);
1410}
1411
1412static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1413{
1414 int hdrlen = sizeof(struct inet_diag_req);
1415 struct net *net = sock_net(skb->sk);
1416
1417 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1418 nlmsg_len(nlh) < hdrlen)
1419 return -EINVAL;
1420
1421 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1422 struct netlink_dump_control c = {
1423 .start = inet_diag_dump_start_compat,
1424 .done = inet_diag_dump_done,
1425 .dump = inet_diag_dump_compat,
1426 };
1427 return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1428 }
1429
1430 return inet_diag_get_exact_compat(skb, nlh);
1431}
1432
1433static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h)
1434{
1435 int hdrlen = sizeof(struct inet_diag_req_v2);
1436 struct net *net = sock_net(skb->sk);
1437
1438 if (nlmsg_len(h) < hdrlen)
1439 return -EINVAL;
1440
1441 if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY &&
1442 h->nlmsg_flags & NLM_F_DUMP) {
1443 struct netlink_dump_control c = {
1444 .start = inet_diag_dump_start,
1445 .done = inet_diag_dump_done,
1446 .dump = inet_diag_dump,
1447 };
1448 return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1449 }
1450
1451 return inet_diag_cmd_exact(h->nlmsg_type, skb, h, hdrlen,
1452 nlmsg_data(h));
1453}
1454
1455static
1456int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk)
1457{
1458 const struct inet_diag_handler *handler;
1459 struct nlmsghdr *nlh;
1460 struct nlattr *attr;
1461 struct inet_diag_msg *r;
1462 void *info = NULL;
1463 int err = 0;
1464
1465 nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0);
1466 if (!nlh)
1467 return -ENOMEM;
1468
1469 r = nlmsg_data(nlh);
1470 memset(r, 0, sizeof(*r));
1471 inet_diag_msg_common_fill(r, sk);
1472 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM)
1473 r->id.idiag_sport = inet_sk(sk)->inet_sport;
1474 r->idiag_state = sk->sk_state;
1475
1476 if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) {
1477 nlmsg_cancel(skb, nlh);
1478 return err;
1479 }
1480
1481 handler = inet_diag_lock_handler(sk->sk_protocol);
1482 if (!handler) {
1483 nlmsg_cancel(skb, nlh);
1484 return -ENOENT;
1485 }
1486
1487 attr = handler->idiag_info_size
1488 ? nla_reserve_64bit(skb, INET_DIAG_INFO,
1489 handler->idiag_info_size,
1490 INET_DIAG_PAD)
1491 : NULL;
1492 if (attr)
1493 info = nla_data(attr);
1494
1495 handler->idiag_get_info(sk, r, info);
1496 inet_diag_unlock_handler(handler);
1497
1498 nlmsg_end(skb, nlh);
1499 return 0;
1500}
1501
1502static const struct sock_diag_handler inet_diag_handler = {
1503 .owner = THIS_MODULE,
1504 .family = AF_INET,
1505 .dump = inet_diag_handler_cmd,
1506 .get_info = inet_diag_handler_get_info,
1507 .destroy = inet_diag_handler_cmd,
1508};
1509
1510static const struct sock_diag_handler inet6_diag_handler = {
1511 .owner = THIS_MODULE,
1512 .family = AF_INET6,
1513 .dump = inet_diag_handler_cmd,
1514 .get_info = inet_diag_handler_get_info,
1515 .destroy = inet_diag_handler_cmd,
1516};
1517
1518int inet_diag_register(const struct inet_diag_handler *h)
1519{
1520 const __u16 type = h->idiag_type;
1521
1522 if (type >= IPPROTO_MAX)
1523 return -EINVAL;
1524
1525 return !cmpxchg((const struct inet_diag_handler **)&inet_diag_table[type],
1526 NULL, h) ? 0 : -EEXIST;
1527}
1528EXPORT_SYMBOL_GPL(inet_diag_register);
1529
1530void inet_diag_unregister(const struct inet_diag_handler *h)
1531{
1532 const __u16 type = h->idiag_type;
1533
1534 if (type >= IPPROTO_MAX)
1535 return;
1536
1537 xchg((const struct inet_diag_handler **)&inet_diag_table[type],
1538 NULL);
1539}
1540EXPORT_SYMBOL_GPL(inet_diag_unregister);
1541
1542static const struct sock_diag_inet_compat inet_diag_compat = {
1543 .owner = THIS_MODULE,
1544 .fn = inet_diag_rcv_msg_compat,
1545};
1546
1547static int __init inet_diag_init(void)
1548{
1549 const int inet_diag_table_size = (IPPROTO_MAX *
1550 sizeof(struct inet_diag_handler *));
1551 int err = -ENOMEM;
1552
1553 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1554 if (!inet_diag_table)
1555 goto out;
1556
1557 err = sock_diag_register(&inet_diag_handler);
1558 if (err)
1559 goto out_free_nl;
1560
1561 err = sock_diag_register(&inet6_diag_handler);
1562 if (err)
1563 goto out_free_inet;
1564
1565 sock_diag_register_inet_compat(&inet_diag_compat);
1566out:
1567 return err;
1568
1569out_free_inet:
1570 sock_diag_unregister(&inet_diag_handler);
1571out_free_nl:
1572 kfree(inet_diag_table);
1573 goto out;
1574}
1575
1576static void __exit inet_diag_exit(void)
1577{
1578 sock_diag_unregister(&inet6_diag_handler);
1579 sock_diag_unregister(&inet_diag_handler);
1580 sock_diag_unregister_inet_compat(&inet_diag_compat);
1581 kfree(inet_diag_table);
1582}
1583
1584module_init(inet_diag_init);
1585module_exit(inet_diag_exit);
1586MODULE_LICENSE("GPL");
1587MODULE_DESCRIPTION("INET/INET6: socket monitoring via SOCK_DIAG");
1588MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1589MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);
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 */);