Linux Audio

Check our new training course

Loading...
v6.13.7
   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 */);
v6.8
   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 **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 DEFINE_MUTEX(inet_diag_table_mutex);
  52
  53static const struct inet_diag_handler *inet_diag_lock_handler(int proto)
  54{
  55	if (proto < 0 || proto >= IPPROTO_MAX) {
  56		mutex_lock(&inet_diag_table_mutex);
  57		return ERR_PTR(-ENOENT);
  58	}
  59
  60	if (!inet_diag_table[proto])
  61		sock_load_diag_module(AF_INET, proto);
  62
  63	mutex_lock(&inet_diag_table_mutex);
  64	if (!inet_diag_table[proto])
  65		return ERR_PTR(-ENOENT);
 
 
  66
  67	return inet_diag_table[proto];
  68}
  69
  70static void inet_diag_unlock_handler(const struct inet_diag_handler *handler)
  71{
  72	mutex_unlock(&inet_diag_table_mutex);
  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	handler = inet_diag_table[req->sdiag_protocol];
 
 
 108	if (handler && handler->idiag_get_aux_size)
 109		aux = handler->idiag_get_aux_size(sk, net_admin);
 
 110
 111	return	  nla_total_size(sizeof(struct tcp_info))
 112		+ nla_total_size(sizeof(struct inet_diag_msg))
 113		+ inet_diag_msg_attrs_size()
 114		+ nla_total_size(sizeof(struct inet_diag_meminfo))
 115		+ nla_total_size(SK_MEMINFO_VARS * sizeof(u32))
 116		+ nla_total_size(TCP_CA_NAME_MAX)
 117		+ nla_total_size(sizeof(struct tcpvegas_info))
 118		+ aux
 119		+ 64;
 120}
 121
 122int inet_diag_msg_attrs_fill(struct sock *sk, struct sk_buff *skb,
 123			     struct inet_diag_msg *r, int ext,
 124			     struct user_namespace *user_ns,
 125			     bool net_admin)
 126{
 127	const struct inet_sock *inet = inet_sk(sk);
 128	struct inet_diag_sockopt inet_sockopt;
 129
 130	if (nla_put_u8(skb, INET_DIAG_SHUTDOWN, sk->sk_shutdown))
 131		goto errout;
 132
 133	/* IPv6 dual-stack sockets use inet->tos for IPv4 connections,
 134	 * hence this needs to be included regardless of socket family.
 135	 */
 136	if (ext & (1 << (INET_DIAG_TOS - 1)))
 137		if (nla_put_u8(skb, INET_DIAG_TOS, READ_ONCE(inet->tos)) < 0)
 138			goto errout;
 139
 140#if IS_ENABLED(CONFIG_IPV6)
 141	if (r->idiag_family == AF_INET6) {
 142		if (ext & (1 << (INET_DIAG_TCLASS - 1)))
 143			if (nla_put_u8(skb, INET_DIAG_TCLASS,
 144				       inet6_sk(sk)->tclass) < 0)
 145				goto errout;
 146
 147		if (((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
 148		    nla_put_u8(skb, INET_DIAG_SKV6ONLY, ipv6_only_sock(sk)))
 149			goto errout;
 150	}
 151#endif
 152
 153	if (net_admin && nla_put_u32(skb, INET_DIAG_MARK, READ_ONCE(sk->sk_mark)))
 154		goto errout;
 155
 156	if (ext & (1 << (INET_DIAG_CLASS_ID - 1)) ||
 157	    ext & (1 << (INET_DIAG_TCLASS - 1))) {
 158		u32 classid = 0;
 159
 160#ifdef CONFIG_SOCK_CGROUP_DATA
 161		classid = sock_cgroup_classid(&sk->sk_cgrp_data);
 162#endif
 163		/* Fallback to socket priority if class id isn't set.
 164		 * Classful qdiscs use it as direct reference to class.
 165		 * For cgroup2 classid is always zero.
 166		 */
 167		if (!classid)
 168			classid = READ_ONCE(sk->sk_priority);
 169
 170		if (nla_put_u32(skb, INET_DIAG_CLASS_ID, classid))
 171			goto errout;
 172	}
 173
 174#ifdef CONFIG_SOCK_CGROUP_DATA
 175	if (nla_put_u64_64bit(skb, INET_DIAG_CGROUP_ID,
 176			      cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)),
 177			      INET_DIAG_PAD))
 178		goto errout;
 179#endif
 180
 181	r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk));
 182	r->idiag_inode = sock_i_ino(sk);
 183
 184	memset(&inet_sockopt, 0, sizeof(inet_sockopt));
 185	inet_sockopt.recverr	= inet_test_bit(RECVERR, sk);
 186	inet_sockopt.is_icsk	= inet_test_bit(IS_ICSK, sk);
 187	inet_sockopt.freebind	= inet_test_bit(FREEBIND, sk);
 188	inet_sockopt.hdrincl	= inet_test_bit(HDRINCL, sk);
 189	inet_sockopt.mc_loop	= inet_test_bit(MC_LOOP, sk);
 190	inet_sockopt.transparent = inet_test_bit(TRANSPARENT, sk);
 191	inet_sockopt.mc_all	= inet_test_bit(MC_ALL, sk);
 192	inet_sockopt.nodefrag	= inet_test_bit(NODEFRAG, sk);
 193	inet_sockopt.bind_address_no_port = inet_test_bit(BIND_ADDRESS_NO_PORT, sk);
 194	inet_sockopt.recverr_rfc4884 = inet_test_bit(RECVERR_RFC4884, sk);
 195	inet_sockopt.defer_connect = inet_test_bit(DEFER_CONNECT, sk);
 196	if (nla_put(skb, INET_DIAG_SOCKOPT, sizeof(inet_sockopt),
 197		    &inet_sockopt))
 198		goto errout;
 199
 200	return 0;
 201errout:
 202	return 1;
 203}
 204EXPORT_SYMBOL_GPL(inet_diag_msg_attrs_fill);
 205
 206static int inet_diag_parse_attrs(const struct nlmsghdr *nlh, int hdrlen,
 207				 struct nlattr **req_nlas)
 208{
 209	struct nlattr *nla;
 210	int remaining;
 211
 212	nlmsg_for_each_attr(nla, nlh, hdrlen, remaining) {
 213		int type = nla_type(nla);
 214
 215		if (type == INET_DIAG_REQ_PROTOCOL && nla_len(nla) != sizeof(u32))
 216			return -EINVAL;
 217
 218		if (type < __INET_DIAG_REQ_MAX)
 219			req_nlas[type] = nla;
 220	}
 221	return 0;
 222}
 223
 224static int inet_diag_get_protocol(const struct inet_diag_req_v2 *req,
 225				  const struct inet_diag_dump_data *data)
 226{
 227	if (data->req_nlas[INET_DIAG_REQ_PROTOCOL])
 228		return nla_get_u32(data->req_nlas[INET_DIAG_REQ_PROTOCOL]);
 229	return req->sdiag_protocol;
 230}
 231
 232#define MAX_DUMP_ALLOC_SIZE (KMALLOC_MAX_SIZE - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
 233
 234int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk,
 235		      struct sk_buff *skb, struct netlink_callback *cb,
 236		      const struct inet_diag_req_v2 *req,
 237		      u16 nlmsg_flags, bool net_admin)
 238{
 239	const struct tcp_congestion_ops *ca_ops;
 240	const struct inet_diag_handler *handler;
 241	struct inet_diag_dump_data *cb_data;
 242	int ext = req->idiag_ext;
 243	struct inet_diag_msg *r;
 244	struct nlmsghdr  *nlh;
 245	struct nlattr *attr;
 246	void *info = NULL;
 
 
 247
 248	cb_data = cb->data;
 249	handler = inet_diag_table[inet_diag_get_protocol(req, cb_data)];
 250	BUG_ON(!handler);
 
 
 
 
 
 251
 252	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
 253			cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
 254	if (!nlh)
 255		return -EMSGSIZE;
 256
 257	r = nlmsg_data(nlh);
 258	BUG_ON(!sk_fullsock(sk));
 259
 260	inet_diag_msg_common_fill(r, sk);
 261	r->idiag_state = sk->sk_state;
 262	r->idiag_timer = 0;
 263	r->idiag_retrans = 0;
 264	r->idiag_expires = 0;
 265
 266	if (inet_diag_msg_attrs_fill(sk, skb, r, ext,
 267				     sk_user_ns(NETLINK_CB(cb->skb).sk),
 268				     net_admin))
 269		goto errout;
 270
 271	if (ext & (1 << (INET_DIAG_MEMINFO - 1))) {
 272		struct inet_diag_meminfo minfo = {
 273			.idiag_rmem = sk_rmem_alloc_get(sk),
 274			.idiag_wmem = READ_ONCE(sk->sk_wmem_queued),
 275			.idiag_fmem = sk_forward_alloc_get(sk),
 276			.idiag_tmem = sk_wmem_alloc_get(sk),
 277		};
 278
 279		if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0)
 280			goto errout;
 281	}
 282
 283	if (ext & (1 << (INET_DIAG_SKMEMINFO - 1)))
 284		if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO))
 285			goto errout;
 286
 287	/*
 288	 * RAW sockets might have user-defined protocols assigned,
 289	 * so report the one supplied on socket creation.
 290	 */
 291	if (sk->sk_type == SOCK_RAW) {
 292		if (nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))
 293			goto errout;
 294	}
 295
 296	if (!icsk) {
 297		handler->idiag_get_info(sk, r, NULL);
 298		goto out;
 299	}
 300
 301	if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
 302	    icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
 303	    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
 
 304		r->idiag_timer = 1;
 305		r->idiag_retrans = icsk->icsk_retransmits;
 306		r->idiag_expires =
 307			jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
 308	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
 309		r->idiag_timer = 4;
 310		r->idiag_retrans = icsk->icsk_probes_out;
 311		r->idiag_expires =
 312			jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
 313	} else if (timer_pending(&sk->sk_timer)) {
 314		r->idiag_timer = 2;
 315		r->idiag_retrans = icsk->icsk_probes_out;
 316		r->idiag_expires =
 317			jiffies_delta_to_msecs(sk->sk_timer.expires - jiffies);
 318	}
 319
 320	if ((ext & (1 << (INET_DIAG_INFO - 1))) && handler->idiag_info_size) {
 321		attr = nla_reserve_64bit(skb, INET_DIAG_INFO,
 322					 handler->idiag_info_size,
 323					 INET_DIAG_PAD);
 324		if (!attr)
 325			goto errout;
 326
 327		info = nla_data(attr);
 328	}
 329
 330	if (ext & (1 << (INET_DIAG_CONG - 1))) {
 331		int err = 0;
 332
 333		rcu_read_lock();
 334		ca_ops = READ_ONCE(icsk->icsk_ca_ops);
 335		if (ca_ops)
 336			err = nla_put_string(skb, INET_DIAG_CONG, ca_ops->name);
 337		rcu_read_unlock();
 338		if (err < 0)
 339			goto errout;
 340	}
 341
 342	handler->idiag_get_info(sk, r, info);
 343
 344	if (ext & (1 << (INET_DIAG_INFO - 1)) && handler->idiag_get_aux)
 345		if (handler->idiag_get_aux(sk, net_admin, skb) < 0)
 346			goto errout;
 347
 348	if (sk->sk_state < TCP_TIME_WAIT) {
 349		union tcp_cc_info info;
 350		size_t sz = 0;
 351		int attr;
 352
 353		rcu_read_lock();
 354		ca_ops = READ_ONCE(icsk->icsk_ca_ops);
 355		if (ca_ops && ca_ops->get_info)
 356			sz = ca_ops->get_info(sk, ext, &attr, &info);
 357		rcu_read_unlock();
 358		if (sz && nla_put(skb, attr, sz, &info) < 0)
 359			goto errout;
 360	}
 361
 362	/* Keep it at the end for potential retry with a larger skb,
 363	 * or else do best-effort fitting, which is only done for the
 364	 * first_nlmsg.
 365	 */
 366	if (cb_data->bpf_stg_diag) {
 367		bool first_nlmsg = ((unsigned char *)nlh == skb->data);
 368		unsigned int prev_min_dump_alloc;
 369		unsigned int total_nla_size = 0;
 370		unsigned int msg_len;
 371		int err;
 372
 373		msg_len = skb_tail_pointer(skb) - (unsigned char *)nlh;
 374		err = bpf_sk_storage_diag_put(cb_data->bpf_stg_diag, sk, skb,
 375					      INET_DIAG_SK_BPF_STORAGES,
 376					      &total_nla_size);
 377
 378		if (!err)
 379			goto out;
 380
 381		total_nla_size += msg_len;
 382		prev_min_dump_alloc = cb->min_dump_alloc;
 383		if (total_nla_size > prev_min_dump_alloc)
 384			cb->min_dump_alloc = min_t(u32, total_nla_size,
 385						   MAX_DUMP_ALLOC_SIZE);
 386
 387		if (!first_nlmsg)
 388			goto errout;
 389
 390		if (cb->min_dump_alloc > prev_min_dump_alloc)
 391			/* Retry with pskb_expand_head() with
 392			 * __GFP_DIRECT_RECLAIM
 393			 */
 394			goto errout;
 395
 396		WARN_ON_ONCE(total_nla_size <= prev_min_dump_alloc);
 397
 398		/* Send what we have for this sk
 399		 * and move on to the next sk in the following
 400		 * dump()
 401		 */
 402	}
 403
 404out:
 405	nlmsg_end(skb, nlh);
 406	return 0;
 407
 408errout:
 409	nlmsg_cancel(skb, nlh);
 410	return -EMSGSIZE;
 411}
 412EXPORT_SYMBOL_GPL(inet_sk_diag_fill);
 413
 414static int inet_twsk_diag_fill(struct sock *sk,
 415			       struct sk_buff *skb,
 416			       struct netlink_callback *cb,
 417			       u16 nlmsg_flags, bool net_admin)
 418{
 419	struct inet_timewait_sock *tw = inet_twsk(sk);
 420	struct inet_diag_msg *r;
 421	struct nlmsghdr *nlh;
 422	long tmo;
 423
 424	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
 425			cb->nlh->nlmsg_seq, cb->nlh->nlmsg_type,
 426			sizeof(*r), nlmsg_flags);
 427	if (!nlh)
 428		return -EMSGSIZE;
 429
 430	r = nlmsg_data(nlh);
 431	BUG_ON(tw->tw_state != TCP_TIME_WAIT);
 432
 433	inet_diag_msg_common_fill(r, sk);
 434	r->idiag_retrans      = 0;
 435
 436	r->idiag_state	      = tw->tw_substate;
 437	r->idiag_timer	      = 3;
 438	tmo = tw->tw_timer.expires - jiffies;
 439	r->idiag_expires      = jiffies_delta_to_msecs(tmo);
 440	r->idiag_rqueue	      = 0;
 441	r->idiag_wqueue	      = 0;
 442	r->idiag_uid	      = 0;
 443	r->idiag_inode	      = 0;
 444
 445	if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
 446				     tw->tw_mark)) {
 447		nlmsg_cancel(skb, nlh);
 448		return -EMSGSIZE;
 449	}
 450
 451	nlmsg_end(skb, nlh);
 452	return 0;
 453}
 454
 455static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb,
 456			      struct netlink_callback *cb,
 457			      u16 nlmsg_flags, bool net_admin)
 458{
 459	struct request_sock *reqsk = inet_reqsk(sk);
 460	struct inet_diag_msg *r;
 461	struct nlmsghdr *nlh;
 462	long tmo;
 463
 464	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
 465			cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
 466	if (!nlh)
 467		return -EMSGSIZE;
 468
 469	r = nlmsg_data(nlh);
 470	inet_diag_msg_common_fill(r, sk);
 471	r->idiag_state = TCP_SYN_RECV;
 472	r->idiag_timer = 1;
 473	r->idiag_retrans = reqsk->num_retrans;
 474
 475	BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) !=
 476		     offsetof(struct sock, sk_cookie));
 477
 478	tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies;
 479	r->idiag_expires = jiffies_delta_to_msecs(tmo);
 480	r->idiag_rqueue	= 0;
 481	r->idiag_wqueue	= 0;
 482	r->idiag_uid	= 0;
 483	r->idiag_inode	= 0;
 484
 485	if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
 486				     inet_rsk(reqsk)->ir_mark)) {
 487		nlmsg_cancel(skb, nlh);
 488		return -EMSGSIZE;
 489	}
 490
 491	nlmsg_end(skb, nlh);
 492	return 0;
 493}
 494
 495static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
 496			struct netlink_callback *cb,
 497			const struct inet_diag_req_v2 *r,
 498			u16 nlmsg_flags, bool net_admin)
 499{
 500	if (sk->sk_state == TCP_TIME_WAIT)
 501		return inet_twsk_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
 502
 503	if (sk->sk_state == TCP_NEW_SYN_RECV)
 504		return inet_req_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
 505
 506	return inet_sk_diag_fill(sk, inet_csk(sk), skb, cb, r, nlmsg_flags,
 507				 net_admin);
 508}
 509
 510struct sock *inet_diag_find_one_icsk(struct net *net,
 511				     struct inet_hashinfo *hashinfo,
 512				     const struct inet_diag_req_v2 *req)
 513{
 514	struct sock *sk;
 515
 516	rcu_read_lock();
 517	if (req->sdiag_family == AF_INET)
 518		sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0],
 519				 req->id.idiag_dport, req->id.idiag_src[0],
 520				 req->id.idiag_sport, req->id.idiag_if);
 521#if IS_ENABLED(CONFIG_IPV6)
 522	else if (req->sdiag_family == AF_INET6) {
 523		if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) &&
 524		    ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src))
 525			sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3],
 526					 req->id.idiag_dport, req->id.idiag_src[3],
 527					 req->id.idiag_sport, req->id.idiag_if);
 528		else
 529			sk = inet6_lookup(net, hashinfo, NULL, 0,
 530					  (struct in6_addr *)req->id.idiag_dst,
 531					  req->id.idiag_dport,
 532					  (struct in6_addr *)req->id.idiag_src,
 533					  req->id.idiag_sport,
 534					  req->id.idiag_if);
 535	}
 536#endif
 537	else {
 538		rcu_read_unlock();
 539		return ERR_PTR(-EINVAL);
 540	}
 541	rcu_read_unlock();
 542	if (!sk)
 543		return ERR_PTR(-ENOENT);
 544
 545	if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) {
 546		sock_gen_put(sk);
 547		return ERR_PTR(-ENOENT);
 548	}
 549
 550	return sk;
 551}
 552EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk);
 553
 554int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo,
 555			    struct netlink_callback *cb,
 556			    const struct inet_diag_req_v2 *req)
 557{
 558	struct sk_buff *in_skb = cb->skb;
 559	bool net_admin = netlink_net_capable(in_skb, CAP_NET_ADMIN);
 560	struct net *net = sock_net(in_skb->sk);
 561	struct sk_buff *rep;
 562	struct sock *sk;
 563	int err;
 564
 565	sk = inet_diag_find_one_icsk(net, hashinfo, req);
 566	if (IS_ERR(sk))
 567		return PTR_ERR(sk);
 568
 569	rep = nlmsg_new(inet_sk_attr_size(sk, req, net_admin), GFP_KERNEL);
 570	if (!rep) {
 571		err = -ENOMEM;
 572		goto out;
 573	}
 574
 575	err = sk_diag_fill(sk, rep, cb, req, 0, net_admin);
 576	if (err < 0) {
 577		WARN_ON(err == -EMSGSIZE);
 578		nlmsg_free(rep);
 579		goto out;
 580	}
 581	err = nlmsg_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid);
 582
 583out:
 584	if (sk)
 585		sock_gen_put(sk);
 586
 587	return err;
 588}
 589EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
 590
 591static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb,
 592			       const struct nlmsghdr *nlh,
 593			       int hdrlen,
 594			       const struct inet_diag_req_v2 *req)
 595{
 596	const struct inet_diag_handler *handler;
 597	struct inet_diag_dump_data dump_data;
 598	int err, protocol;
 599
 600	memset(&dump_data, 0, sizeof(dump_data));
 601	err = inet_diag_parse_attrs(nlh, hdrlen, dump_data.req_nlas);
 602	if (err)
 603		return err;
 604
 605	protocol = inet_diag_get_protocol(req, &dump_data);
 606
 607	handler = inet_diag_lock_handler(protocol);
 608	if (IS_ERR(handler)) {
 609		err = PTR_ERR(handler);
 610	} else if (cmd == SOCK_DIAG_BY_FAMILY) {
 
 611		struct netlink_callback cb = {
 612			.nlh = nlh,
 613			.skb = in_skb,
 614			.data = &dump_data,
 615		};
 616		err = handler->dump_one(&cb, req);
 617	} else if (cmd == SOCK_DESTROY && handler->destroy) {
 618		err = handler->destroy(in_skb, req);
 619	} else {
 620		err = -EOPNOTSUPP;
 621	}
 622	inet_diag_unlock_handler(handler);
 623
 624	return err;
 625}
 626
 627static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
 628{
 629	int words = bits >> 5;
 630
 631	bits &= 0x1f;
 632
 633	if (words) {
 634		if (memcmp(a1, a2, words << 2))
 635			return 0;
 636	}
 637	if (bits) {
 638		__be32 w1, w2;
 639		__be32 mask;
 640
 641		w1 = a1[words];
 642		w2 = a2[words];
 643
 644		mask = htonl((0xffffffff) << (32 - bits));
 645
 646		if ((w1 ^ w2) & mask)
 647			return 0;
 648	}
 649
 650	return 1;
 651}
 652
 653static int inet_diag_bc_run(const struct nlattr *_bc,
 654			    const struct inet_diag_entry *entry)
 655{
 656	const void *bc = nla_data(_bc);
 657	int len = nla_len(_bc);
 658
 659	while (len > 0) {
 660		int yes = 1;
 661		const struct inet_diag_bc_op *op = bc;
 662
 663		switch (op->code) {
 664		case INET_DIAG_BC_NOP:
 665			break;
 666		case INET_DIAG_BC_JMP:
 667			yes = 0;
 668			break;
 669		case INET_DIAG_BC_S_EQ:
 670			yes = entry->sport == op[1].no;
 671			break;
 672		case INET_DIAG_BC_S_GE:
 673			yes = entry->sport >= op[1].no;
 674			break;
 675		case INET_DIAG_BC_S_LE:
 676			yes = entry->sport <= op[1].no;
 677			break;
 678		case INET_DIAG_BC_D_EQ:
 679			yes = entry->dport == op[1].no;
 680			break;
 681		case INET_DIAG_BC_D_GE:
 682			yes = entry->dport >= op[1].no;
 683			break;
 684		case INET_DIAG_BC_D_LE:
 685			yes = entry->dport <= op[1].no;
 686			break;
 687		case INET_DIAG_BC_AUTO:
 688			yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
 689			break;
 690		case INET_DIAG_BC_S_COND:
 691		case INET_DIAG_BC_D_COND: {
 692			const struct inet_diag_hostcond *cond;
 693			const __be32 *addr;
 694
 695			cond = (const struct inet_diag_hostcond *)(op + 1);
 696			if (cond->port != -1 &&
 697			    cond->port != (op->code == INET_DIAG_BC_S_COND ?
 698					     entry->sport : entry->dport)) {
 699				yes = 0;
 700				break;
 701			}
 702
 703			if (op->code == INET_DIAG_BC_S_COND)
 704				addr = entry->saddr;
 705			else
 706				addr = entry->daddr;
 707
 708			if (cond->family != AF_UNSPEC &&
 709			    cond->family != entry->family) {
 710				if (entry->family == AF_INET6 &&
 711				    cond->family == AF_INET) {
 712					if (addr[0] == 0 && addr[1] == 0 &&
 713					    addr[2] == htonl(0xffff) &&
 714					    bitstring_match(addr + 3,
 715							    cond->addr,
 716							    cond->prefix_len))
 717						break;
 718				}
 719				yes = 0;
 720				break;
 721			}
 722
 723			if (cond->prefix_len == 0)
 724				break;
 725			if (bitstring_match(addr, cond->addr,
 726					    cond->prefix_len))
 727				break;
 728			yes = 0;
 729			break;
 730		}
 731		case INET_DIAG_BC_DEV_COND: {
 732			u32 ifindex;
 733
 734			ifindex = *((const u32 *)(op + 1));
 735			if (ifindex != entry->ifindex)
 736				yes = 0;
 737			break;
 738		}
 739		case INET_DIAG_BC_MARK_COND: {
 740			struct inet_diag_markcond *cond;
 741
 742			cond = (struct inet_diag_markcond *)(op + 1);
 743			if ((entry->mark & cond->mask) != cond->mark)
 744				yes = 0;
 745			break;
 746		}
 747#ifdef CONFIG_SOCK_CGROUP_DATA
 748		case INET_DIAG_BC_CGROUP_COND: {
 749			u64 cgroup_id;
 750
 751			cgroup_id = get_unaligned((const u64 *)(op + 1));
 752			if (cgroup_id != entry->cgroup_id)
 753				yes = 0;
 754			break;
 755		}
 756#endif
 757		}
 758
 759		if (yes) {
 760			len -= op->yes;
 761			bc += op->yes;
 762		} else {
 763			len -= op->no;
 764			bc += op->no;
 765		}
 766	}
 767	return len == 0;
 768}
 769
 770/* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV)
 771 */
 772static void entry_fill_addrs(struct inet_diag_entry *entry,
 773			     const struct sock *sk)
 774{
 775#if IS_ENABLED(CONFIG_IPV6)
 776	if (sk->sk_family == AF_INET6) {
 777		entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32;
 778		entry->daddr = sk->sk_v6_daddr.s6_addr32;
 779	} else
 780#endif
 781	{
 782		entry->saddr = &sk->sk_rcv_saddr;
 783		entry->daddr = &sk->sk_daddr;
 784	}
 785}
 786
 787int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
 788{
 789	struct inet_sock *inet = inet_sk(sk);
 790	struct inet_diag_entry entry;
 791
 792	if (!bc)
 793		return 1;
 794
 795	entry.family = sk->sk_family;
 796	entry_fill_addrs(&entry, sk);
 797	entry.sport = inet->inet_num;
 798	entry.dport = ntohs(inet->inet_dport);
 799	entry.ifindex = sk->sk_bound_dev_if;
 800	entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0;
 801	if (sk_fullsock(sk))
 802		entry.mark = READ_ONCE(sk->sk_mark);
 803	else if (sk->sk_state == TCP_NEW_SYN_RECV)
 804		entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark;
 805	else if (sk->sk_state == TCP_TIME_WAIT)
 806		entry.mark = inet_twsk(sk)->tw_mark;
 807	else
 808		entry.mark = 0;
 809#ifdef CONFIG_SOCK_CGROUP_DATA
 810	entry.cgroup_id = sk_fullsock(sk) ?
 811		cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)) : 0;
 812#endif
 813
 814	return inet_diag_bc_run(bc, &entry);
 815}
 816EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
 817
 818static int valid_cc(const void *bc, int len, int cc)
 819{
 820	while (len >= 0) {
 821		const struct inet_diag_bc_op *op = bc;
 822
 823		if (cc > len)
 824			return 0;
 825		if (cc == len)
 826			return 1;
 827		if (op->yes < 4 || op->yes & 3)
 828			return 0;
 829		len -= op->yes;
 830		bc  += op->yes;
 831	}
 832	return 0;
 833}
 834
 835/* data is u32 ifindex */
 836static bool valid_devcond(const struct inet_diag_bc_op *op, int len,
 837			  int *min_len)
 838{
 839	/* Check ifindex space. */
 840	*min_len += sizeof(u32);
 841	if (len < *min_len)
 842		return false;
 843
 844	return true;
 845}
 846/* Validate an inet_diag_hostcond. */
 847static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
 848			   int *min_len)
 849{
 850	struct inet_diag_hostcond *cond;
 851	int addr_len;
 852
 853	/* Check hostcond space. */
 854	*min_len += sizeof(struct inet_diag_hostcond);
 855	if (len < *min_len)
 856		return false;
 857	cond = (struct inet_diag_hostcond *)(op + 1);
 858
 859	/* Check address family and address length. */
 860	switch (cond->family) {
 861	case AF_UNSPEC:
 862		addr_len = 0;
 863		break;
 864	case AF_INET:
 865		addr_len = sizeof(struct in_addr);
 866		break;
 867	case AF_INET6:
 868		addr_len = sizeof(struct in6_addr);
 869		break;
 870	default:
 871		return false;
 872	}
 873	*min_len += addr_len;
 874	if (len < *min_len)
 875		return false;
 876
 877	/* Check prefix length (in bits) vs address length (in bytes). */
 878	if (cond->prefix_len > 8 * addr_len)
 879		return false;
 880
 881	return true;
 882}
 883
 884/* Validate a port comparison operator. */
 885static bool valid_port_comparison(const struct inet_diag_bc_op *op,
 886				  int len, int *min_len)
 887{
 888	/* Port comparisons put the port in a follow-on inet_diag_bc_op. */
 889	*min_len += sizeof(struct inet_diag_bc_op);
 890	if (len < *min_len)
 891		return false;
 892	return true;
 893}
 894
 895static bool valid_markcond(const struct inet_diag_bc_op *op, int len,
 896			   int *min_len)
 897{
 898	*min_len += sizeof(struct inet_diag_markcond);
 899	return len >= *min_len;
 900}
 901
 902#ifdef CONFIG_SOCK_CGROUP_DATA
 903static bool valid_cgroupcond(const struct inet_diag_bc_op *op, int len,
 904			     int *min_len)
 905{
 906	*min_len += sizeof(u64);
 907	return len >= *min_len;
 908}
 909#endif
 910
 911static int inet_diag_bc_audit(const struct nlattr *attr,
 912			      const struct sk_buff *skb)
 913{
 914	bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN);
 915	const void *bytecode, *bc;
 916	int bytecode_len, len;
 917
 918	if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op))
 919		return -EINVAL;
 920
 921	bytecode = bc = nla_data(attr);
 922	len = bytecode_len = nla_len(attr);
 923
 924	while (len > 0) {
 925		int min_len = sizeof(struct inet_diag_bc_op);
 926		const struct inet_diag_bc_op *op = bc;
 927
 928		switch (op->code) {
 929		case INET_DIAG_BC_S_COND:
 930		case INET_DIAG_BC_D_COND:
 931			if (!valid_hostcond(bc, len, &min_len))
 932				return -EINVAL;
 933			break;
 934		case INET_DIAG_BC_DEV_COND:
 935			if (!valid_devcond(bc, len, &min_len))
 936				return -EINVAL;
 937			break;
 938		case INET_DIAG_BC_S_EQ:
 939		case INET_DIAG_BC_S_GE:
 940		case INET_DIAG_BC_S_LE:
 941		case INET_DIAG_BC_D_EQ:
 942		case INET_DIAG_BC_D_GE:
 943		case INET_DIAG_BC_D_LE:
 944			if (!valid_port_comparison(bc, len, &min_len))
 945				return -EINVAL;
 946			break;
 947		case INET_DIAG_BC_MARK_COND:
 948			if (!net_admin)
 949				return -EPERM;
 950			if (!valid_markcond(bc, len, &min_len))
 951				return -EINVAL;
 952			break;
 953#ifdef CONFIG_SOCK_CGROUP_DATA
 954		case INET_DIAG_BC_CGROUP_COND:
 955			if (!valid_cgroupcond(bc, len, &min_len))
 956				return -EINVAL;
 957			break;
 958#endif
 959		case INET_DIAG_BC_AUTO:
 960		case INET_DIAG_BC_JMP:
 961		case INET_DIAG_BC_NOP:
 962			break;
 963		default:
 964			return -EINVAL;
 965		}
 966
 967		if (op->code != INET_DIAG_BC_NOP) {
 968			if (op->no < min_len || op->no > len + 4 || op->no & 3)
 969				return -EINVAL;
 970			if (op->no < len &&
 971			    !valid_cc(bytecode, bytecode_len, len - op->no))
 972				return -EINVAL;
 973		}
 974
 975		if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
 976			return -EINVAL;
 977		bc  += op->yes;
 978		len -= op->yes;
 979	}
 980	return len == 0 ? 0 : -EINVAL;
 981}
 982
 983static void twsk_build_assert(void)
 984{
 985	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) !=
 986		     offsetof(struct sock, sk_family));
 987
 988	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) !=
 989		     offsetof(struct inet_sock, inet_num));
 990
 991	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) !=
 992		     offsetof(struct inet_sock, inet_dport));
 993
 994	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) !=
 995		     offsetof(struct inet_sock, inet_rcv_saddr));
 996
 997	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) !=
 998		     offsetof(struct inet_sock, inet_daddr));
 999
1000#if IS_ENABLED(CONFIG_IPV6)
1001	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) !=
1002		     offsetof(struct sock, sk_v6_rcv_saddr));
1003
1004	BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) !=
1005		     offsetof(struct sock, sk_v6_daddr));
1006#endif
1007}
1008
1009void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
1010			 struct netlink_callback *cb,
1011			 const struct inet_diag_req_v2 *r)
1012{
1013	bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN);
1014	struct inet_diag_dump_data *cb_data = cb->data;
1015	struct net *net = sock_net(skb->sk);
1016	u32 idiag_states = r->idiag_states;
1017	int i, num, s_i, s_num;
1018	struct nlattr *bc;
1019	struct sock *sk;
1020
1021	bc = cb_data->inet_diag_nla_bc;
1022	if (idiag_states & TCPF_SYN_RECV)
1023		idiag_states |= TCPF_NEW_SYN_RECV;
1024	s_i = cb->args[1];
1025	s_num = num = cb->args[2];
1026
1027	if (cb->args[0] == 0) {
1028		if (!(idiag_states & TCPF_LISTEN) || r->id.idiag_dport)
1029			goto skip_listen_ht;
1030
1031		for (i = s_i; i <= hashinfo->lhash2_mask; i++) {
1032			struct inet_listen_hashbucket *ilb;
1033			struct hlist_nulls_node *node;
1034
1035			num = 0;
1036			ilb = &hashinfo->lhash2[i];
1037
 
 
 
 
1038			spin_lock(&ilb->lock);
1039			sk_nulls_for_each(sk, node, &ilb->nulls_head) {
1040				struct inet_sock *inet = inet_sk(sk);
1041
1042				if (!net_eq(sock_net(sk), net))
1043					continue;
1044
1045				if (num < s_num) {
1046					num++;
1047					continue;
1048				}
1049
1050				if (r->sdiag_family != AF_UNSPEC &&
1051				    sk->sk_family != r->sdiag_family)
1052					goto next_listen;
1053
1054				if (r->id.idiag_sport != inet->inet_sport &&
1055				    r->id.idiag_sport)
1056					goto next_listen;
1057
1058				if (!inet_diag_bc_sk(bc, sk))
1059					goto next_listen;
1060
1061				if (inet_sk_diag_fill(sk, inet_csk(sk), skb,
1062						      cb, r, NLM_F_MULTI,
1063						      net_admin) < 0) {
1064					spin_unlock(&ilb->lock);
1065					goto done;
1066				}
1067
1068next_listen:
1069				++num;
1070			}
1071			spin_unlock(&ilb->lock);
1072
1073			s_num = 0;
1074		}
1075skip_listen_ht:
1076		cb->args[0] = 1;
1077		s_i = num = s_num = 0;
1078	}
1079
1080/* Process a maximum of SKARR_SZ sockets at a time when walking hash buckets
1081 * with bh disabled.
1082 */
1083#define SKARR_SZ 16
1084
1085	/* Dump bound but inactive (not listening, connecting, etc.) sockets */
1086	if (cb->args[0] == 1) {
1087		if (!(idiag_states & TCPF_BOUND_INACTIVE))
1088			goto skip_bind_ht;
1089
1090		for (i = s_i; i < hashinfo->bhash_size; i++) {
1091			struct inet_bind_hashbucket *ibb;
1092			struct inet_bind2_bucket *tb2;
1093			struct sock *sk_arr[SKARR_SZ];
1094			int num_arr[SKARR_SZ];
1095			int idx, accum, res;
1096
1097resume_bind_walk:
1098			num = 0;
1099			accum = 0;
1100			ibb = &hashinfo->bhash2[i];
1101
 
 
 
 
1102			spin_lock_bh(&ibb->lock);
1103			inet_bind_bucket_for_each(tb2, &ibb->chain) {
1104				if (!net_eq(ib2_net(tb2), net))
1105					continue;
1106
1107				sk_for_each_bound(sk, &tb2->owners) {
1108					struct inet_sock *inet = inet_sk(sk);
1109
1110					if (num < s_num)
1111						goto next_bind;
1112
1113					if (sk->sk_state != TCP_CLOSE ||
1114					    !inet->inet_num)
1115						goto next_bind;
1116
1117					if (r->sdiag_family != AF_UNSPEC &&
1118					    r->sdiag_family != sk->sk_family)
1119						goto next_bind;
1120
1121					if (!inet_diag_bc_sk(bc, sk))
1122						goto next_bind;
1123
1124					sock_hold(sk);
1125					num_arr[accum] = num;
1126					sk_arr[accum] = sk;
1127					if (++accum == SKARR_SZ)
1128						goto pause_bind_walk;
1129next_bind:
1130					num++;
1131				}
1132			}
1133pause_bind_walk:
1134			spin_unlock_bh(&ibb->lock);
1135
1136			res = 0;
1137			for (idx = 0; idx < accum; idx++) {
1138				if (res >= 0) {
1139					res = inet_sk_diag_fill(sk_arr[idx],
1140								NULL, skb, cb,
1141								r, NLM_F_MULTI,
1142								net_admin);
1143					if (res < 0)
1144						num = num_arr[idx];
1145				}
1146				sock_put(sk_arr[idx]);
1147			}
1148			if (res < 0)
1149				goto done;
1150
1151			cond_resched();
1152
1153			if (accum == SKARR_SZ) {
1154				s_num = num + 1;
1155				goto resume_bind_walk;
1156			}
1157
1158			s_num = 0;
1159		}
1160skip_bind_ht:
1161		cb->args[0] = 2;
1162		s_i = num = s_num = 0;
1163	}
1164
1165	if (!(idiag_states & ~TCPF_LISTEN))
1166		goto out;
1167
1168	for (i = s_i; i <= hashinfo->ehash_mask; i++) {
1169		struct inet_ehash_bucket *head = &hashinfo->ehash[i];
1170		spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
1171		struct hlist_nulls_node *node;
1172		struct sock *sk_arr[SKARR_SZ];
1173		int num_arr[SKARR_SZ];
1174		int idx, accum, res;
1175
1176		if (hlist_nulls_empty(&head->chain))
1177			continue;
1178
1179		if (i > s_i)
1180			s_num = 0;
1181
1182next_chunk:
1183		num = 0;
1184		accum = 0;
1185		spin_lock_bh(lock);
1186		sk_nulls_for_each(sk, node, &head->chain) {
1187			int state;
1188
1189			if (!net_eq(sock_net(sk), net))
1190				continue;
1191			if (num < s_num)
1192				goto next_normal;
1193			state = (sk->sk_state == TCP_TIME_WAIT) ?
1194				inet_twsk(sk)->tw_substate : sk->sk_state;
1195			if (!(idiag_states & (1 << state)))
1196				goto next_normal;
1197			if (r->sdiag_family != AF_UNSPEC &&
1198			    sk->sk_family != r->sdiag_family)
1199				goto next_normal;
1200			if (r->id.idiag_sport != htons(sk->sk_num) &&
1201			    r->id.idiag_sport)
1202				goto next_normal;
1203			if (r->id.idiag_dport != sk->sk_dport &&
1204			    r->id.idiag_dport)
1205				goto next_normal;
1206			twsk_build_assert();
1207
1208			if (!inet_diag_bc_sk(bc, sk))
1209				goto next_normal;
1210
1211			if (!refcount_inc_not_zero(&sk->sk_refcnt))
1212				goto next_normal;
1213
1214			num_arr[accum] = num;
1215			sk_arr[accum] = sk;
1216			if (++accum == SKARR_SZ)
1217				break;
1218next_normal:
1219			++num;
1220		}
1221		spin_unlock_bh(lock);
1222		res = 0;
1223		for (idx = 0; idx < accum; idx++) {
1224			if (res >= 0) {
1225				res = sk_diag_fill(sk_arr[idx], skb, cb, r,
1226						   NLM_F_MULTI, net_admin);
1227				if (res < 0)
1228					num = num_arr[idx];
1229			}
1230			sock_gen_put(sk_arr[idx]);
1231		}
1232		if (res < 0)
1233			break;
1234		cond_resched();
1235		if (accum == SKARR_SZ) {
1236			s_num = num + 1;
1237			goto next_chunk;
1238		}
1239	}
1240
1241done:
1242	cb->args[1] = i;
1243	cb->args[2] = num;
1244out:
1245	;
1246}
1247EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
1248
1249static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
1250			    const struct inet_diag_req_v2 *r)
1251{
1252	struct inet_diag_dump_data *cb_data = cb->data;
1253	const struct inet_diag_handler *handler;
1254	u32 prev_min_dump_alloc;
1255	int protocol, err = 0;
1256
1257	protocol = inet_diag_get_protocol(r, cb_data);
1258
1259again:
1260	prev_min_dump_alloc = cb->min_dump_alloc;
1261	handler = inet_diag_lock_handler(protocol);
1262	if (!IS_ERR(handler))
1263		handler->dump(skb, cb, r);
1264	else
1265		err = PTR_ERR(handler);
1266	inet_diag_unlock_handler(handler);
1267
1268	/* The skb is not large enough to fit one sk info and
1269	 * inet_sk_diag_fill() has requested for a larger skb.
1270	 */
1271	if (!skb->len && cb->min_dump_alloc > prev_min_dump_alloc) {
1272		err = pskb_expand_head(skb, 0, cb->min_dump_alloc, GFP_KERNEL);
1273		if (!err)
1274			goto again;
1275	}
1276
1277	return err ? : skb->len;
1278}
1279
1280static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
1281{
1282	return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh));
1283}
1284
1285static int __inet_diag_dump_start(struct netlink_callback *cb, int hdrlen)
1286{
1287	const struct nlmsghdr *nlh = cb->nlh;
1288	struct inet_diag_dump_data *cb_data;
1289	struct sk_buff *skb = cb->skb;
1290	struct nlattr *nla;
1291	int err;
1292
1293	cb_data = kzalloc(sizeof(*cb_data), GFP_KERNEL);
1294	if (!cb_data)
1295		return -ENOMEM;
1296
1297	err = inet_diag_parse_attrs(nlh, hdrlen, cb_data->req_nlas);
1298	if (err) {
1299		kfree(cb_data);
1300		return err;
1301	}
1302	nla = cb_data->inet_diag_nla_bc;
1303	if (nla) {
1304		err = inet_diag_bc_audit(nla, skb);
1305		if (err) {
1306			kfree(cb_data);
1307			return err;
1308		}
1309	}
1310
1311	nla = cb_data->inet_diag_nla_bpf_stgs;
1312	if (nla) {
1313		struct bpf_sk_storage_diag *bpf_stg_diag;
1314
1315		bpf_stg_diag = bpf_sk_storage_diag_alloc(nla);
1316		if (IS_ERR(bpf_stg_diag)) {
1317			kfree(cb_data);
1318			return PTR_ERR(bpf_stg_diag);
1319		}
1320		cb_data->bpf_stg_diag = bpf_stg_diag;
1321	}
1322
1323	cb->data = cb_data;
1324	return 0;
1325}
1326
1327static int inet_diag_dump_start(struct netlink_callback *cb)
1328{
1329	return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req_v2));
1330}
1331
1332static int inet_diag_dump_start_compat(struct netlink_callback *cb)
1333{
1334	return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req));
1335}
1336
1337static int inet_diag_dump_done(struct netlink_callback *cb)
1338{
1339	struct inet_diag_dump_data *cb_data = cb->data;
1340
1341	bpf_sk_storage_diag_free(cb_data->bpf_stg_diag);
1342	kfree(cb->data);
1343
1344	return 0;
1345}
1346
1347static int inet_diag_type2proto(int type)
1348{
1349	switch (type) {
1350	case TCPDIAG_GETSOCK:
1351		return IPPROTO_TCP;
1352	case DCCPDIAG_GETSOCK:
1353		return IPPROTO_DCCP;
1354	default:
1355		return 0;
1356	}
1357}
1358
1359static int inet_diag_dump_compat(struct sk_buff *skb,
1360				 struct netlink_callback *cb)
1361{
1362	struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1363	struct inet_diag_req_v2 req;
1364
1365	req.sdiag_family = AF_UNSPEC; /* compatibility */
1366	req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1367	req.idiag_ext = rc->idiag_ext;
 
1368	req.idiag_states = rc->idiag_states;
1369	req.id = rc->id;
1370
1371	return __inet_diag_dump(skb, cb, &req);
1372}
1373
1374static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1375				      const struct nlmsghdr *nlh)
1376{
1377	struct inet_diag_req *rc = nlmsg_data(nlh);
1378	struct inet_diag_req_v2 req;
1379
1380	req.sdiag_family = rc->idiag_family;
1381	req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1382	req.idiag_ext = rc->idiag_ext;
 
1383	req.idiag_states = rc->idiag_states;
1384	req.id = rc->id;
1385
1386	return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh,
1387				   sizeof(struct inet_diag_req), &req);
1388}
1389
1390static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1391{
1392	int hdrlen = sizeof(struct inet_diag_req);
1393	struct net *net = sock_net(skb->sk);
1394
1395	if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1396	    nlmsg_len(nlh) < hdrlen)
1397		return -EINVAL;
1398
1399	if (nlh->nlmsg_flags & NLM_F_DUMP) {
1400		struct netlink_dump_control c = {
1401			.start = inet_diag_dump_start_compat,
1402			.done = inet_diag_dump_done,
1403			.dump = inet_diag_dump_compat,
1404		};
1405		return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1406	}
1407
1408	return inet_diag_get_exact_compat(skb, nlh);
1409}
1410
1411static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h)
1412{
1413	int hdrlen = sizeof(struct inet_diag_req_v2);
1414	struct net *net = sock_net(skb->sk);
1415
1416	if (nlmsg_len(h) < hdrlen)
1417		return -EINVAL;
1418
1419	if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY &&
1420	    h->nlmsg_flags & NLM_F_DUMP) {
1421		struct netlink_dump_control c = {
1422			.start = inet_diag_dump_start,
1423			.done = inet_diag_dump_done,
1424			.dump = inet_diag_dump,
1425		};
1426		return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1427	}
1428
1429	return inet_diag_cmd_exact(h->nlmsg_type, skb, h, hdrlen,
1430				   nlmsg_data(h));
1431}
1432
1433static
1434int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk)
1435{
1436	const struct inet_diag_handler *handler;
1437	struct nlmsghdr *nlh;
1438	struct nlattr *attr;
1439	struct inet_diag_msg *r;
1440	void *info = NULL;
1441	int err = 0;
1442
1443	nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0);
1444	if (!nlh)
1445		return -ENOMEM;
1446
1447	r = nlmsg_data(nlh);
1448	memset(r, 0, sizeof(*r));
1449	inet_diag_msg_common_fill(r, sk);
1450	if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM)
1451		r->id.idiag_sport = inet_sk(sk)->inet_sport;
1452	r->idiag_state = sk->sk_state;
1453
1454	if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) {
1455		nlmsg_cancel(skb, nlh);
1456		return err;
1457	}
1458
1459	handler = inet_diag_lock_handler(sk->sk_protocol);
1460	if (IS_ERR(handler)) {
1461		inet_diag_unlock_handler(handler);
1462		nlmsg_cancel(skb, nlh);
1463		return PTR_ERR(handler);
1464	}
1465
1466	attr = handler->idiag_info_size
1467		? nla_reserve_64bit(skb, INET_DIAG_INFO,
1468				    handler->idiag_info_size,
1469				    INET_DIAG_PAD)
1470		: NULL;
1471	if (attr)
1472		info = nla_data(attr);
1473
1474	handler->idiag_get_info(sk, r, info);
1475	inet_diag_unlock_handler(handler);
1476
1477	nlmsg_end(skb, nlh);
1478	return 0;
1479}
1480
1481static const struct sock_diag_handler inet_diag_handler = {
 
1482	.family = AF_INET,
1483	.dump = inet_diag_handler_cmd,
1484	.get_info = inet_diag_handler_get_info,
1485	.destroy = inet_diag_handler_cmd,
1486};
1487
1488static const struct sock_diag_handler inet6_diag_handler = {
 
1489	.family = AF_INET6,
1490	.dump = inet_diag_handler_cmd,
1491	.get_info = inet_diag_handler_get_info,
1492	.destroy = inet_diag_handler_cmd,
1493};
1494
1495int inet_diag_register(const struct inet_diag_handler *h)
1496{
1497	const __u16 type = h->idiag_type;
1498	int err = -EINVAL;
1499
1500	if (type >= IPPROTO_MAX)
1501		goto out;
1502
1503	mutex_lock(&inet_diag_table_mutex);
1504	err = -EEXIST;
1505	if (!inet_diag_table[type]) {
1506		inet_diag_table[type] = h;
1507		err = 0;
1508	}
1509	mutex_unlock(&inet_diag_table_mutex);
1510out:
1511	return err;
1512}
1513EXPORT_SYMBOL_GPL(inet_diag_register);
1514
1515void inet_diag_unregister(const struct inet_diag_handler *h)
1516{
1517	const __u16 type = h->idiag_type;
1518
1519	if (type >= IPPROTO_MAX)
1520		return;
1521
1522	mutex_lock(&inet_diag_table_mutex);
1523	inet_diag_table[type] = NULL;
1524	mutex_unlock(&inet_diag_table_mutex);
1525}
1526EXPORT_SYMBOL_GPL(inet_diag_unregister);
1527
 
 
 
 
 
1528static int __init inet_diag_init(void)
1529{
1530	const int inet_diag_table_size = (IPPROTO_MAX *
1531					  sizeof(struct inet_diag_handler *));
1532	int err = -ENOMEM;
1533
1534	inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1535	if (!inet_diag_table)
1536		goto out;
1537
1538	err = sock_diag_register(&inet_diag_handler);
1539	if (err)
1540		goto out_free_nl;
1541
1542	err = sock_diag_register(&inet6_diag_handler);
1543	if (err)
1544		goto out_free_inet;
1545
1546	sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1547out:
1548	return err;
1549
1550out_free_inet:
1551	sock_diag_unregister(&inet_diag_handler);
1552out_free_nl:
1553	kfree(inet_diag_table);
1554	goto out;
1555}
1556
1557static void __exit inet_diag_exit(void)
1558{
1559	sock_diag_unregister(&inet6_diag_handler);
1560	sock_diag_unregister(&inet_diag_handler);
1561	sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1562	kfree(inet_diag_table);
1563}
1564
1565module_init(inet_diag_init);
1566module_exit(inet_diag_exit);
1567MODULE_LICENSE("GPL");
1568MODULE_DESCRIPTION("INET/INET6: socket monitoring via SOCK_DIAG");
1569MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1570MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);