/net/ipv4/ping.c

   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 * INET		An implementation of the TCP/IP protocol suite for the LINUX
   4 *		operating system.  INET is implemented using the  BSD Socket
   5 *		interface as the means of communication with the user level.
   6 *
   7 *		"Ping" sockets
   8 *
   9 * Based on ipv4/udp.c code.
  10 *
  11 * Authors:	Vasiliy Kulikov / Openwall (for Linux 2.6),
  12 *		Pavel Kankovsky (for Linux 2.4.32)
  13 *
  14 * Pavel gave all rights to bugs to Vasiliy,
  15 * none of the bugs are Pavel's now.
  16 */
  17
  18#include <linux/uaccess.h>
  19#include <linux/types.h>
  20#include <linux/fcntl.h>
  21#include <linux/socket.h>
  22#include <linux/sockios.h>
  23#include <linux/in.h>
  24#include <linux/errno.h>
  25#include <linux/timer.h>
  26#include <linux/mm.h>
  27#include <linux/inet.h>
  28#include <linux/netdevice.h>
  29#include <net/snmp.h>
  30#include <net/ip.h>
  31#include <net/icmp.h>
  32#include <net/protocol.h>
  33#include <linux/skbuff.h>
  34#include <linux/proc_fs.h>
  35#include <linux/export.h>
  36#include <net/sock.h>
  37#include <net/ping.h>
  38#include <net/udp.h>
  39#include <net/route.h>
  40#include <net/inet_common.h>
  41#include <net/checksum.h>
  42
  43#if IS_ENABLED(CONFIG_IPV6)
  44#include <linux/in6.h>
  45#include <linux/icmpv6.h>
  46#include <net/addrconf.h>
  47#include <net/ipv6.h>
  48#include <net/transp_v6.h>
  49#endif
  50
  51struct ping_table {
  52	struct hlist_nulls_head	hash[PING_HTABLE_SIZE];
  53	rwlock_t		lock;
  54};
  55
  56static struct ping_table ping_table;
  57struct pingv6_ops pingv6_ops;
  58EXPORT_SYMBOL_GPL(pingv6_ops);
  59
  60static u16 ping_port_rover;
  61
  62static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
  63{
  64	u32 res = (num + net_hash_mix(net)) & mask;
  65
  66	pr_debug("hash(%u) = %u\n", num, res);
  67	return res;
  68}
  69EXPORT_SYMBOL_GPL(ping_hash);
  70
  71static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
  72					     struct net *net, unsigned int num)
  73{
  74	return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
  75}
  76
  77int ping_get_port(struct sock *sk, unsigned short ident)
  78{
  79	struct hlist_nulls_node *node;
  80	struct hlist_nulls_head *hlist;
  81	struct inet_sock *isk, *isk2;
  82	struct sock *sk2 = NULL;
  83
  84	isk = inet_sk(sk);
  85	write_lock_bh(&ping_table.lock);
  86	if (ident == 0) {
  87		u32 i;
  88		u16 result = ping_port_rover + 1;
  89
  90		for (i = 0; i < (1L << 16); i++, result++) {
  91			if (!result)
  92				result++; /* avoid zero */
  93			hlist = ping_hashslot(&ping_table, sock_net(sk),
  94					    result);
  95			ping_portaddr_for_each_entry(sk2, node, hlist) {
  96				isk2 = inet_sk(sk2);
  97
  98				if (isk2->inet_num == result)
  99					goto next_port;
 100			}
 101
 102			/* found */
 103			ping_port_rover = ident = result;
 104			break;
 105next_port:
 106			;
 107		}
 108		if (i >= (1L << 16))
 109			goto fail;
 110	} else {
 111		hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
 112		ping_portaddr_for_each_entry(sk2, node, hlist) {
 113			isk2 = inet_sk(sk2);
 114
 115			/* BUG? Why is this reuse and not reuseaddr? ping.c
 116			 * doesn't turn off SO_REUSEADDR, and it doesn't expect
 117			 * that other ping processes can steal its packets.
 118			 */
 119			if ((isk2->inet_num == ident) &&
 120			    (sk2 != sk) &&
 121			    (!sk2->sk_reuse || !sk->sk_reuse))
 122				goto fail;
 123		}
 124	}
 125
 126	pr_debug("found port/ident = %d\n", ident);
 127	isk->inet_num = ident;
 128	if (sk_unhashed(sk)) {
 129		pr_debug("was not hashed\n");
 130		sock_hold(sk);
 131		hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
 132		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
 133	}
 134	write_unlock_bh(&ping_table.lock);
 135	return 0;
 136
 137fail:
 138	write_unlock_bh(&ping_table.lock);
 139	return 1;
 140}
 141EXPORT_SYMBOL_GPL(ping_get_port);
 142
 143int ping_hash(struct sock *sk)
 144{
 145	pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
 146	BUG(); /* "Please do not press this button again." */
 147
 148	return 0;
 149}
 150
 151void ping_unhash(struct sock *sk)
 152{
 153	struct inet_sock *isk = inet_sk(sk);
 154
 155	pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
 156	write_lock_bh(&ping_table.lock);
 157	if (sk_hashed(sk)) {
 158		hlist_nulls_del(&sk->sk_nulls_node);
 159		sk_nulls_node_init(&sk->sk_nulls_node);
 160		sock_put(sk);
 161		isk->inet_num = 0;
 162		isk->inet_sport = 0;
 163		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
 164	}
 165	write_unlock_bh(&ping_table.lock);
 166}
 167EXPORT_SYMBOL_GPL(ping_unhash);
 168
 169static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
 170{
 171	struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
 172	struct sock *sk = NULL;
 173	struct inet_sock *isk;
 174	struct hlist_nulls_node *hnode;
 175	int dif = skb->dev->ifindex;
 176
 177	if (skb->protocol == htons(ETH_P_IP)) {
 178		pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
 179			 (int)ident, &ip_hdr(skb)->daddr, dif);
 180#if IS_ENABLED(CONFIG_IPV6)
 181	} else if (skb->protocol == htons(ETH_P_IPV6)) {
 182		pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
 183			 (int)ident, &ipv6_hdr(skb)->daddr, dif);
 184#endif
 185	}
 186
 187	read_lock_bh(&ping_table.lock);
 188
 189	ping_portaddr_for_each_entry(sk, hnode, hslot) {
 190		isk = inet_sk(sk);
 191
 192		pr_debug("iterate\n");
 193		if (isk->inet_num != ident)
 194			continue;
 195
 196		if (skb->protocol == htons(ETH_P_IP) &&
 197		    sk->sk_family == AF_INET) {
 198			pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
 199				 (int) isk->inet_num, &isk->inet_rcv_saddr,
 200				 sk->sk_bound_dev_if);
 201
 202			if (isk->inet_rcv_saddr &&
 203			    isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
 204				continue;
 205#if IS_ENABLED(CONFIG_IPV6)
 206		} else if (skb->protocol == htons(ETH_P_IPV6) &&
 207			   sk->sk_family == AF_INET6) {
 208
 209			pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
 210				 (int) isk->inet_num,
 211				 &sk->sk_v6_rcv_saddr,
 212				 sk->sk_bound_dev_if);
 213
 214			if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
 215			    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
 216					     &ipv6_hdr(skb)->daddr))
 217				continue;
 218#endif
 219		} else {
 220			continue;
 221		}
 222
 223		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
 224			continue;
 225
 226		sock_hold(sk);
 227		goto exit;
 228	}
 229
 230	sk = NULL;
 231exit:
 232	read_unlock_bh(&ping_table.lock);
 233
 234	return sk;
 235}
 236
 237static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
 238					  kgid_t *high)
 239{
 240	kgid_t *data = net->ipv4.ping_group_range.range;
 241	unsigned int seq;
 242
 243	do {
 244		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
 245
 246		*low = data[0];
 247		*high = data[1];
 248	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
 249}
 250
 251
 252int ping_init_sock(struct sock *sk)
 253{
 254	struct net *net = sock_net(sk);
 255	kgid_t group = current_egid();
 256	struct group_info *group_info;
 257	int i;
 258	kgid_t low, high;
 259	int ret = 0;
 260
 261	if (sk->sk_family == AF_INET6)
 262		sk->sk_ipv6only = 1;
 263
 264	inet_get_ping_group_range_net(net, &low, &high);
 265	if (gid_lte(low, group) && gid_lte(group, high))
 266		return 0;
 267
 268	group_info = get_current_groups();
 269	for (i = 0; i < group_info->ngroups; i++) {
 270		kgid_t gid = group_info->gid[i];
 271
 272		if (gid_lte(low, gid) && gid_lte(gid, high))
 273			goto out_release_group;
 274	}
 275
 276	ret = -EACCES;
 277
 278out_release_group:
 279	put_group_info(group_info);
 280	return ret;
 281}
 282EXPORT_SYMBOL_GPL(ping_init_sock);
 283
 284void ping_close(struct sock *sk, long timeout)
 285{
 286	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
 287		 inet_sk(sk), inet_sk(sk)->inet_num);
 288	pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
 289
 290	sk_common_release(sk);
 291}
 292EXPORT_SYMBOL_GPL(ping_close);
 293
 294/* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
 295static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
 296				struct sockaddr *uaddr, int addr_len) {
 297	struct net *net = sock_net(sk);
 298	if (sk->sk_family == AF_INET) {
 299		struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 300		int chk_addr_ret;
 301
 302		if (addr_len < sizeof(*addr))
 303			return -EINVAL;
 304
 305		if (addr->sin_family != AF_INET &&
 306		    !(addr->sin_family == AF_UNSPEC &&
 307		      addr->sin_addr.s_addr == htonl(INADDR_ANY)))
 308			return -EAFNOSUPPORT;
 309
 310		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
 311			 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
 312
 313		chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
 314
 315		if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
 316			chk_addr_ret = RTN_LOCAL;
 317
 318		if ((!inet_can_nonlocal_bind(net, isk) &&
 319		     chk_addr_ret != RTN_LOCAL) ||
 320		    chk_addr_ret == RTN_MULTICAST ||
 321		    chk_addr_ret == RTN_BROADCAST)
 322			return -EADDRNOTAVAIL;
 323
 324#if IS_ENABLED(CONFIG_IPV6)
 325	} else if (sk->sk_family == AF_INET6) {
 326		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
 327		int addr_type, scoped, has_addr;
 328		struct net_device *dev = NULL;
 329
 330		if (addr_len < sizeof(*addr))
 331			return -EINVAL;
 332
 333		if (addr->sin6_family != AF_INET6)
 334			return -EAFNOSUPPORT;
 335
 336		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
 337			 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
 338
 339		addr_type = ipv6_addr_type(&addr->sin6_addr);
 340		scoped = __ipv6_addr_needs_scope_id(addr_type);
 341		if ((addr_type != IPV6_ADDR_ANY &&
 342		     !(addr_type & IPV6_ADDR_UNICAST)) ||
 343		    (scoped && !addr->sin6_scope_id))
 344			return -EINVAL;
 345
 346		rcu_read_lock();
 347		if (addr->sin6_scope_id) {
 348			dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
 349			if (!dev) {
 350				rcu_read_unlock();
 351				return -ENODEV;
 352			}
 353		}
 354		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
 355						    scoped);
 356		rcu_read_unlock();
 357
 358		if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
 359		      addr_type == IPV6_ADDR_ANY))
 360			return -EADDRNOTAVAIL;
 361
 362		if (scoped)
 363			sk->sk_bound_dev_if = addr->sin6_scope_id;
 364#endif
 365	} else {
 366		return -EAFNOSUPPORT;
 367	}
 368	return 0;
 369}
 370
 371static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
 372{
 373	if (saddr->sa_family == AF_INET) {
 374		struct inet_sock *isk = inet_sk(sk);
 375		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
 376		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
 377#if IS_ENABLED(CONFIG_IPV6)
 378	} else if (saddr->sa_family == AF_INET6) {
 379		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
 380		struct ipv6_pinfo *np = inet6_sk(sk);
 381		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
 382#endif
 383	}
 384}
 385
 386static void ping_clear_saddr(struct sock *sk, int dif)
 387{
 388	sk->sk_bound_dev_if = dif;
 389	if (sk->sk_family == AF_INET) {
 390		struct inet_sock *isk = inet_sk(sk);
 391		isk->inet_rcv_saddr = isk->inet_saddr = 0;
 392#if IS_ENABLED(CONFIG_IPV6)
 393	} else if (sk->sk_family == AF_INET6) {
 394		struct ipv6_pinfo *np = inet6_sk(sk);
 395		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
 396		memset(&np->saddr, 0, sizeof(np->saddr));
 397#endif
 398	}
 399}
 400/*
 401 * We need our own bind because there are no privileged id's == local ports.
 402 * Moreover, we don't allow binding to multi- and broadcast addresses.
 403 */
 404
 405int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 406{
 407	struct inet_sock *isk = inet_sk(sk);
 408	unsigned short snum;
 409	int err;
 410	int dif = sk->sk_bound_dev_if;
 411
 412	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
 413	if (err)
 414		return err;
 415
 416	lock_sock(sk);
 417
 418	err = -EINVAL;
 419	if (isk->inet_num != 0)
 420		goto out;
 421
 422	err = -EADDRINUSE;
 423	ping_set_saddr(sk, uaddr);
 424	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
 425	if (ping_get_port(sk, snum) != 0) {
 426		ping_clear_saddr(sk, dif);
 427		goto out;
 428	}
 429
 430	pr_debug("after bind(): num = %hu, dif = %d\n",
 431		 isk->inet_num,
 432		 sk->sk_bound_dev_if);
 433
 434	err = 0;
 435	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
 436		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 437#if IS_ENABLED(CONFIG_IPV6)
 438	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 439		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 440#endif
 441
 442	if (snum)
 443		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 444	isk->inet_sport = htons(isk->inet_num);
 445	isk->inet_daddr = 0;
 446	isk->inet_dport = 0;
 447
 448#if IS_ENABLED(CONFIG_IPV6)
 449	if (sk->sk_family == AF_INET6)
 450		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
 451#endif
 452
 453	sk_dst_reset(sk);
 454out:
 455	release_sock(sk);
 456	pr_debug("ping_v4_bind -> %d\n", err);
 457	return err;
 458}
 459EXPORT_SYMBOL_GPL(ping_bind);
 460
 461/*
 462 * Is this a supported type of ICMP message?
 463 */
 464
 465static inline int ping_supported(int family, int type, int code)
 466{
 467	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
 468	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
 469}
 470
 471/*
 472 * This routine is called by the ICMP module when it gets some
 473 * sort of error condition.
 474 */
 475
 476void ping_err(struct sk_buff *skb, int offset, u32 info)
 477{
 478	int family;
 479	struct icmphdr *icmph;
 480	struct inet_sock *inet_sock;
 481	int type;
 482	int code;
 483	struct net *net = dev_net(skb->dev);
 484	struct sock *sk;
 485	int harderr;
 486	int err;
 487
 488	if (skb->protocol == htons(ETH_P_IP)) {
 489		family = AF_INET;
 490		type = icmp_hdr(skb)->type;
 491		code = icmp_hdr(skb)->code;
 492		icmph = (struct icmphdr *)(skb->data + offset);
 493	} else if (skb->protocol == htons(ETH_P_IPV6)) {
 494		family = AF_INET6;
 495		type = icmp6_hdr(skb)->icmp6_type;
 496		code = icmp6_hdr(skb)->icmp6_code;
 497		icmph = (struct icmphdr *) (skb->data + offset);
 498	} else {
 499		BUG();
 500	}
 501
 502	/* We assume the packet has already been checked by icmp_unreach */
 503
 504	if (!ping_supported(family, icmph->type, icmph->code))
 505		return;
 506
 507	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
 508		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
 509		 ntohs(icmph->un.echo.sequence));
 510
 511	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 512	if (!sk) {
 513		pr_debug("no socket, dropping\n");
 514		return;	/* No socket for error */
 515	}
 516	pr_debug("err on socket %p\n", sk);
 517
 518	err = 0;
 519	harderr = 0;
 520	inet_sock = inet_sk(sk);
 521
 522	if (skb->protocol == htons(ETH_P_IP)) {
 523		switch (type) {
 524		default:
 525		case ICMP_TIME_EXCEEDED:
 526			err = EHOSTUNREACH;
 527			break;
 528		case ICMP_SOURCE_QUENCH:
 529			/* This is not a real error but ping wants to see it.
 530			 * Report it with some fake errno.
 531			 */
 532			err = EREMOTEIO;
 533			break;
 534		case ICMP_PARAMETERPROB:
 535			err = EPROTO;
 536			harderr = 1;
 537			break;
 538		case ICMP_DEST_UNREACH:
 539			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
 540				ipv4_sk_update_pmtu(skb, sk, info);
 541				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
 542					err = EMSGSIZE;
 543					harderr = 1;
 544					break;
 545				}
 546				goto out;
 547			}
 548			err = EHOSTUNREACH;
 549			if (code <= NR_ICMP_UNREACH) {
 550				harderr = icmp_err_convert[code].fatal;
 551				err = icmp_err_convert[code].errno;
 552			}
 553			break;
 554		case ICMP_REDIRECT:
 555			/* See ICMP_SOURCE_QUENCH */
 556			ipv4_sk_redirect(skb, sk);
 557			err = EREMOTEIO;
 558			break;
 559		}
 560#if IS_ENABLED(CONFIG_IPV6)
 561	} else if (skb->protocol == htons(ETH_P_IPV6)) {
 562		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
 563#endif
 564	}
 565
 566	/*
 567	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
 568	 *	4.1.3.3.
 569	 */
 570	if ((family == AF_INET && !inet_sock->recverr) ||
 571	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
 572		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
 573			goto out;
 574	} else {
 575		if (family == AF_INET) {
 576			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
 577				      info, (u8 *)icmph);
 578#if IS_ENABLED(CONFIG_IPV6)
 579		} else if (family == AF_INET6) {
 580			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
 581						   info, (u8 *)icmph);
 582#endif
 583		}
 584	}
 585	sk->sk_err = err;
 586	sk->sk_error_report(sk);
 587out:
 588	sock_put(sk);
 589}
 590EXPORT_SYMBOL_GPL(ping_err);
 591
 592/*
 593 *	Copy and checksum an ICMP Echo packet from user space into a buffer
 594 *	starting from the payload.
 595 */
 596
 597int ping_getfrag(void *from, char *to,
 598		 int offset, int fraglen, int odd, struct sk_buff *skb)
 599{
 600	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
 601
 602	if (offset == 0) {
 603		fraglen -= sizeof(struct icmphdr);
 604		if (fraglen < 0)
 605			BUG();
 606		if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
 607			    fraglen, &pfh->wcheck,
 608			    &pfh->msg->msg_iter))
 609			return -EFAULT;
 610	} else if (offset < sizeof(struct icmphdr)) {
 611			BUG();
 612	} else {
 613		if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
 614					    &pfh->msg->msg_iter))
 615			return -EFAULT;
 616	}
 617
 618#if IS_ENABLED(CONFIG_IPV6)
 619	/* For IPv6, checksum each skb as we go along, as expected by
 620	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
 621	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
 622	 */
 623	if (pfh->family == AF_INET6) {
 624		skb->csum = pfh->wcheck;
 625		skb->ip_summed = CHECKSUM_NONE;
 626		pfh->wcheck = 0;
 627	}
 628#endif
 629
 630	return 0;
 631}
 632EXPORT_SYMBOL_GPL(ping_getfrag);
 633
 634static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
 635				       struct flowi4 *fl4)
 636{
 637	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
 638
 639	if (!skb)
 640		return 0;
 641	pfh->wcheck = csum_partial((char *)&pfh->icmph,
 642		sizeof(struct icmphdr), pfh->wcheck);
 643	pfh->icmph.checksum = csum_fold(pfh->wcheck);
 644	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
 645	skb->ip_summed = CHECKSUM_NONE;
 646	return ip_push_pending_frames(sk, fl4);
 647}
 648
 649int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
 650			void *user_icmph, size_t icmph_len) {
 651	u8 type, code;
 652
 653	if (len > 0xFFFF)
 654		return -EMSGSIZE;
 655
 656	/* Must have at least a full ICMP header. */
 657	if (len < icmph_len)
 658		return -EINVAL;
 659
 660	/*
 661	 *	Check the flags.
 662	 */
 663
 664	/* Mirror BSD error message compatibility */
 665	if (msg->msg_flags & MSG_OOB)
 666		return -EOPNOTSUPP;
 667
 668	/*
 669	 *	Fetch the ICMP header provided by the userland.
 670	 *	iovec is modified! The ICMP header is consumed.
 671	 */
 672	if (memcpy_from_msg(user_icmph, msg, icmph_len))
 673		return -EFAULT;
 674
 675	if (family == AF_INET) {
 676		type = ((struct icmphdr *) user_icmph)->type;
 677		code = ((struct icmphdr *) user_icmph)->code;
 678#if IS_ENABLED(CONFIG_IPV6)
 679	} else if (family == AF_INET6) {
 680		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
 681		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
 682#endif
 683	} else {
 684		BUG();
 685	}
 686
 687	if (!ping_supported(family, type, code))
 688		return -EINVAL;
 689
 690	return 0;
 691}
 692EXPORT_SYMBOL_GPL(ping_common_sendmsg);
 693
 694static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 695{
 696	struct net *net = sock_net(sk);
 697	struct flowi4 fl4;
 698	struct inet_sock *inet = inet_sk(sk);
 699	struct ipcm_cookie ipc;
 700	struct icmphdr user_icmph;
 701	struct pingfakehdr pfh;
 702	struct rtable *rt = NULL;
 703	struct ip_options_data opt_copy;
 704	int free = 0;
 705	__be32 saddr, daddr, faddr;
 706	u8  tos;
 707	int err;
 708
 709	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
 710
 711	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
 712				  sizeof(user_icmph));
 713	if (err)
 714		return err;
 715
 716	/*
 717	 *	Get and verify the address.
 718	 */
 719
 720	if (msg->msg_name) {
 721		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
 722		if (msg->msg_namelen < sizeof(*usin))
 723			return -EINVAL;
 724		if (usin->sin_family != AF_INET)
 725			return -EAFNOSUPPORT;
 726		daddr = usin->sin_addr.s_addr;
 727		/* no remote port */
 728	} else {
 729		if (sk->sk_state != TCP_ESTABLISHED)
 730			return -EDESTADDRREQ;
 731		daddr = inet->inet_daddr;
 732		/* no remote port */
 733	}
 734
 735	ipcm_init_sk(&ipc, inet);
 736
 737	if (msg->msg_controllen) {
 738		err = ip_cmsg_send(sk, msg, &ipc, false);
 739		if (unlikely(err)) {
 740			kfree(ipc.opt);
 741			return err;
 742		}
 743		if (ipc.opt)
 744			free = 1;
 745	}
 746	if (!ipc.opt) {
 747		struct ip_options_rcu *inet_opt;
 748
 749		rcu_read_lock();
 750		inet_opt = rcu_dereference(inet->inet_opt);
 751		if (inet_opt) {
 752			memcpy(&opt_copy, inet_opt,
 753			       sizeof(*inet_opt) + inet_opt->opt.optlen);
 754			ipc.opt = &opt_copy.opt;
 755		}
 756		rcu_read_unlock();
 757	}
 758
 759	saddr = ipc.addr;
 760	ipc.addr = faddr = daddr;
 761
 762	if (ipc.opt && ipc.opt->opt.srr) {
 763		if (!daddr) {
 764			err = -EINVAL;
 765			goto out_free;
 766		}
 767		faddr = ipc.opt->opt.faddr;
 768	}
 769	tos = get_rttos(&ipc, inet);
 770	if (sock_flag(sk, SOCK_LOCALROUTE) ||
 771	    (msg->msg_flags & MSG_DONTROUTE) ||
 772	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
 773		tos |= RTO_ONLINK;
 774	}
 775
 776	if (ipv4_is_multicast(daddr)) {
 777		if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
 778			ipc.oif = inet->mc_index;
 779		if (!saddr)
 780			saddr = inet->mc_addr;
 781	} else if (!ipc.oif)
 782		ipc.oif = inet->uc_index;
 783
 784	flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
 785			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
 786			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
 787			   sk->sk_uid);
 788
 789	security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
 790	rt = ip_route_output_flow(net, &fl4, sk);
 791	if (IS_ERR(rt)) {
 792		err = PTR_ERR(rt);
 793		rt = NULL;
 794		if (err == -ENETUNREACH)
 795			IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
 796		goto out;
 797	}
 798
 799	err = -EACCES;
 800	if ((rt->rt_flags & RTCF_BROADCAST) &&
 801	    !sock_flag(sk, SOCK_BROADCAST))
 802		goto out;
 803
 804	if (msg->msg_flags & MSG_CONFIRM)
 805		goto do_confirm;
 806back_from_confirm:
 807
 808	if (!ipc.addr)
 809		ipc.addr = fl4.daddr;
 810
 811	lock_sock(sk);
 812
 813	pfh.icmph.type = user_icmph.type; /* already checked */
 814	pfh.icmph.code = user_icmph.code; /* ditto */
 815	pfh.icmph.checksum = 0;
 816	pfh.icmph.un.echo.id = inet->inet_sport;
 817	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
 818	pfh.msg = msg;
 819	pfh.wcheck = 0;
 820	pfh.family = AF_INET;
 821
 822	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
 823			0, &ipc, &rt, msg->msg_flags);
 824	if (err)
 825		ip_flush_pending_frames(sk);
 826	else
 827		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
 828	release_sock(sk);
 829
 830out:
 831	ip_rt_put(rt);
 832out_free:
 833	if (free)
 834		kfree(ipc.opt);
 835	if (!err) {
 836		icmp_out_count(sock_net(sk), user_icmph.type);
 837		return len;
 838	}
 839	return err;
 840
 841do_confirm:
 842	if (msg->msg_flags & MSG_PROBE)
 843		dst_confirm_neigh(&rt->dst, &fl4.daddr);
 844	if (!(msg->msg_flags & MSG_PROBE) || len)
 845		goto back_from_confirm;
 846	err = 0;
 847	goto out;
 848}
 849
 850int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
 851		 int flags, int *addr_len)
 852{
 853	struct inet_sock *isk = inet_sk(sk);
 854	int family = sk->sk_family;
 855	struct sk_buff *skb;
 856	int copied, err;
 857
 858	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
 859
 860	err = -EOPNOTSUPP;
 861	if (flags & MSG_OOB)
 862		goto out;
 863
 864	if (flags & MSG_ERRQUEUE)
 865		return inet_recv_error(sk, msg, len, addr_len);
 866
 867	skb = skb_recv_datagram(sk, flags, noblock, &err);
 868	if (!skb)
 869		goto out;
 870
 871	copied = skb->len;
 872	if (copied > len) {
 873		msg->msg_flags |= MSG_TRUNC;
 874		copied = len;
 875	}
 876
 877	/* Don't bother checking the checksum */
 878	err = skb_copy_datagram_msg(skb, 0, msg, copied);
 879	if (err)
 880		goto done;
 881
 882	sock_recv_timestamp(msg, sk, skb);
 883
 884	/* Copy the address and add cmsg data. */
 885	if (family == AF_INET) {
 886		DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
 887
 888		if (sin) {
 889			sin->sin_family = AF_INET;
 890			sin->sin_port = 0 /* skb->h.uh->source */;
 891			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 892			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 893			*addr_len = sizeof(*sin);
 894		}
 895
 896		if (isk->cmsg_flags)
 897			ip_cmsg_recv(msg, skb);
 898
 899#if IS_ENABLED(CONFIG_IPV6)
 900	} else if (family == AF_INET6) {
 901		struct ipv6_pinfo *np = inet6_sk(sk);
 902		struct ipv6hdr *ip6 = ipv6_hdr(skb);
 903		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
 904
 905		if (sin6) {
 906			sin6->sin6_family = AF_INET6;
 907			sin6->sin6_port = 0;
 908			sin6->sin6_addr = ip6->saddr;
 909			sin6->sin6_flowinfo = 0;
 910			if (np->sndflow)
 911				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
 912			sin6->sin6_scope_id =
 913				ipv6_iface_scope_id(&sin6->sin6_addr,
 914						    inet6_iif(skb));
 915			*addr_len = sizeof(*sin6);
 916		}
 917
 918		if (inet6_sk(sk)->rxopt.all)
 919			pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
 920		if (skb->protocol == htons(ETH_P_IPV6) &&
 921		    inet6_sk(sk)->rxopt.all)
 922			pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
 923		else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
 924			ip_cmsg_recv(msg, skb);
 925#endif
 926	} else {
 927		BUG();
 928	}
 929
 930	err = copied;
 931
 932done:
 933	skb_free_datagram(sk, skb);
 934out:
 935	pr_debug("ping_recvmsg -> %d\n", err);
 936	return err;
 937}
 938EXPORT_SYMBOL_GPL(ping_recvmsg);
 939
 940int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 941{
 942	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
 943		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
 944	if (sock_queue_rcv_skb(sk, skb) < 0) {
 945		kfree_skb(skb);
 946		pr_debug("ping_queue_rcv_skb -> failed\n");
 947		return -1;
 948	}
 949	return 0;
 950}
 951EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
 952
 953
 954/*
 955 *	All we need to do is get the socket.
 956 */
 957
 958bool ping_rcv(struct sk_buff *skb)
 959{
 960	struct sock *sk;
 961	struct net *net = dev_net(skb->dev);
 962	struct icmphdr *icmph = icmp_hdr(skb);
 963
 964	/* We assume the packet has already been checked by icmp_rcv */
 965
 966	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
 967		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
 968
 969	/* Push ICMP header back */
 970	skb_push(skb, skb->data - (u8 *)icmph);
 971
 972	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 973	if (sk) {
 974		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 975
 976		pr_debug("rcv on socket %p\n", sk);
 977		if (skb2)
 978			ping_queue_rcv_skb(sk, skb2);
 979		sock_put(sk);
 980		return true;
 981	}
 982	pr_debug("no socket, dropping\n");
 983
 984	return false;
 985}
 986EXPORT_SYMBOL_GPL(ping_rcv);
 987
 988struct proto ping_prot = {
 989	.name =		"PING",
 990	.owner =	THIS_MODULE,
 991	.init =		ping_init_sock,
 992	.close =	ping_close,
 993	.connect =	ip4_datagram_connect,
 994	.disconnect =	__udp_disconnect,
 995	.setsockopt =	ip_setsockopt,
 996	.getsockopt =	ip_getsockopt,
 997	.sendmsg =	ping_v4_sendmsg,
 998	.recvmsg =	ping_recvmsg,
 999	.bind =		ping_bind,
1000	.backlog_rcv =	ping_queue_rcv_skb,
1001	.release_cb =	ip4_datagram_release_cb,
1002	.hash =		ping_hash,
1003	.unhash =	ping_unhash,
1004	.get_port =	ping_get_port,
1005	.obj_size =	sizeof(struct inet_sock),
1006};
1007EXPORT_SYMBOL(ping_prot);
1008
1009#ifdef CONFIG_PROC_FS
1010
1011static struct sock *ping_get_first(struct seq_file *seq, int start)
1012{
1013	struct sock *sk;
1014	struct ping_iter_state *state = seq->private;
1015	struct net *net = seq_file_net(seq);
1016
1017	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1018	     ++state->bucket) {
1019		struct hlist_nulls_node *node;
1020		struct hlist_nulls_head *hslot;
1021
1022		hslot = &ping_table.hash[state->bucket];
1023
1024		if (hlist_nulls_empty(hslot))
1025			continue;
1026
1027		sk_nulls_for_each(sk, node, hslot) {
1028			if (net_eq(sock_net(sk), net) &&
1029			    sk->sk_family == state->family)
1030				goto found;
1031		}
1032	}
1033	sk = NULL;
1034found:
1035	return sk;
1036}
1037
1038static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1039{
1040	struct ping_iter_state *state = seq->private;
1041	struct net *net = seq_file_net(seq);
1042
1043	do {
1044		sk = sk_nulls_next(sk);
1045	} while (sk && (!net_eq(sock_net(sk), net)));
1046
1047	if (!sk)
1048		return ping_get_first(seq, state->bucket + 1);
1049	return sk;
1050}
1051
1052static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1053{
1054	struct sock *sk = ping_get_first(seq, 0);
1055
1056	if (sk)
1057		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1058			--pos;
1059	return pos ? NULL : sk;
1060}
1061
1062void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1063	__acquires(ping_table.lock)
1064{
1065	struct ping_iter_state *state = seq->private;
1066	state->bucket = 0;
1067	state->family = family;
1068
1069	read_lock_bh(&ping_table.lock);
1070
1071	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1072}
1073EXPORT_SYMBOL_GPL(ping_seq_start);
1074
1075static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1076{
1077	return ping_seq_start(seq, pos, AF_INET);
1078}
1079
1080void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1081{
1082	struct sock *sk;
1083
1084	if (v == SEQ_START_TOKEN)
1085		sk = ping_get_idx(seq, 0);
1086	else
1087		sk = ping_get_next(seq, v);
1088
1089	++*pos;
1090	return sk;
1091}
1092EXPORT_SYMBOL_GPL(ping_seq_next);
1093
1094void ping_seq_stop(struct seq_file *seq, void *v)
1095	__releases(ping_table.lock)
1096{
1097	read_unlock_bh(&ping_table.lock);
1098}
1099EXPORT_SYMBOL_GPL(ping_seq_stop);
1100
1101static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1102		int bucket)
1103{
1104	struct inet_sock *inet = inet_sk(sp);
1105	__be32 dest = inet->inet_daddr;
1106	__be32 src = inet->inet_rcv_saddr;
1107	__u16 destp = ntohs(inet->inet_dport);
1108	__u16 srcp = ntohs(inet->inet_sport);
1109
1110	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1111		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u",
1112		bucket, src, srcp, dest, destp, sp->sk_state,
1113		sk_wmem_alloc_get(sp),
1114		sk_rmem_alloc_get(sp),
1115		0, 0L, 0,
1116		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1117		0, sock_i_ino(sp),
1118		refcount_read(&sp->sk_refcnt), sp,
1119		atomic_read(&sp->sk_drops));
1120}
1121
1122static int ping_v4_seq_show(struct seq_file *seq, void *v)
1123{
1124	seq_setwidth(seq, 127);
1125	if (v == SEQ_START_TOKEN)
1126		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1127			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1128			   "inode ref pointer drops");
1129	else {
1130		struct ping_iter_state *state = seq->private;
1131
1132		ping_v4_format_sock(v, seq, state->bucket);
1133	}
1134	seq_pad(seq, '\n');
1135	return 0;
1136}
1137
1138static const struct seq_operations ping_v4_seq_ops = {
1139	.start		= ping_v4_seq_start,
1140	.show		= ping_v4_seq_show,
1141	.next		= ping_seq_next,
1142	.stop		= ping_seq_stop,
1143};
1144
1145static int __net_init ping_v4_proc_init_net(struct net *net)
1146{
1147	if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1148			sizeof(struct ping_iter_state)))
1149		return -ENOMEM;
1150	return 0;
1151}
1152
1153static void __net_exit ping_v4_proc_exit_net(struct net *net)
1154{
1155	remove_proc_entry("icmp", net->proc_net);
1156}
1157
1158static struct pernet_operations ping_v4_net_ops = {
1159	.init = ping_v4_proc_init_net,
1160	.exit = ping_v4_proc_exit_net,
1161};
1162
1163int __init ping_proc_init(void)
1164{
1165	return register_pernet_subsys(&ping_v4_net_ops);
1166}
1167
1168void ping_proc_exit(void)
1169{
1170	unregister_pernet_subsys(&ping_v4_net_ops);
1171}
1172
1173#endif
1174
1175void __init ping_init(void)
1176{
1177	int i;
1178
1179	for (i = 0; i < PING_HTABLE_SIZE; i++)
1180		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1181	rwlock_init(&ping_table.lock);
1182}