1 // SPDX-License-Identifier: GPL-2.0-or-later
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.
9 * Based on ipv4/udp.c code.
11 * Authors: Vasiliy Kulikov / Openwall (for Linux 2.6),
12 * Pavel Kankovsky (for Linux 2.4.32)
14 * Pavel gave all rights to bugs to Vasiliy,
15 * none of the bugs are Pavel's now.
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>
24 #include <linux/errno.h>
25 #include <linux/timer.h>
27 #include <linux/inet.h>
28 #include <linux/netdevice.h>
32 #include <net/protocol.h>
33 #include <linux/skbuff.h>
34 #include <linux/proc_fs.h>
35 #include <linux/export.h>
36 #include <linux/bpf-cgroup.h>
40 #include <net/route.h>
41 #include <net/inet_common.h>
42 #include <net/checksum.h>
44 #if IS_ENABLED(CONFIG_IPV6)
45 #include <linux/in6.h>
46 #include <linux/icmpv6.h>
47 #include <net/addrconf.h>
49 #include <net/transp_v6.h>
53 struct hlist_nulls_head hash[PING_HTABLE_SIZE];
57 static struct ping_table ping_table;
58 struct pingv6_ops pingv6_ops;
59 EXPORT_SYMBOL_GPL(pingv6_ops);
61 static u16 ping_port_rover;
63 static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
65 u32 res = (num + net_hash_mix(net)) & mask;
67 pr_debug("hash(%u) = %u\n", num, res);
70 EXPORT_SYMBOL_GPL(ping_hash);
72 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
73 struct net *net, unsigned int num)
75 return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
78 int ping_get_port(struct sock *sk, unsigned short ident)
80 struct hlist_nulls_node *node;
81 struct hlist_nulls_head *hlist;
82 struct inet_sock *isk, *isk2;
83 struct sock *sk2 = NULL;
86 spin_lock(&ping_table.lock);
89 u16 result = ping_port_rover + 1;
91 for (i = 0; i < (1L << 16); i++, result++) {
93 result++; /* avoid zero */
94 hlist = ping_hashslot(&ping_table, sock_net(sk),
96 ping_portaddr_for_each_entry(sk2, node, hlist) {
99 if (isk2->inet_num == result)
104 ping_port_rover = ident = result;
112 hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
113 ping_portaddr_for_each_entry(sk2, node, hlist) {
116 /* BUG? Why is this reuse and not reuseaddr? ping.c
117 * doesn't turn off SO_REUSEADDR, and it doesn't expect
118 * that other ping processes can steal its packets.
120 if ((isk2->inet_num == ident) &&
122 (!sk2->sk_reuse || !sk->sk_reuse))
127 pr_debug("found port/ident = %d\n", ident);
128 isk->inet_num = ident;
129 if (sk_unhashed(sk)) {
130 pr_debug("was not hashed\n");
132 sock_set_flag(sk, SOCK_RCU_FREE);
133 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, hlist);
134 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
136 spin_unlock(&ping_table.lock);
140 spin_unlock(&ping_table.lock);
143 EXPORT_SYMBOL_GPL(ping_get_port);
145 int ping_hash(struct sock *sk)
147 pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
148 BUG(); /* "Please do not press this button again." */
153 void ping_unhash(struct sock *sk)
155 struct inet_sock *isk = inet_sk(sk);
157 pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
158 spin_lock(&ping_table.lock);
160 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
164 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
166 spin_unlock(&ping_table.lock);
168 EXPORT_SYMBOL_GPL(ping_unhash);
170 /* Called under rcu_read_lock() */
171 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
173 struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
174 struct sock *sk = NULL;
175 struct inet_sock *isk;
176 struct hlist_nulls_node *hnode;
179 if (skb->protocol == htons(ETH_P_IP)) {
181 sdif = inet_sdif(skb);
182 pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
183 (int)ident, &ip_hdr(skb)->daddr, dif);
184 #if IS_ENABLED(CONFIG_IPV6)
185 } else if (skb->protocol == htons(ETH_P_IPV6)) {
186 dif = inet6_iif(skb);
187 sdif = inet6_sdif(skb);
188 pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
189 (int)ident, &ipv6_hdr(skb)->daddr, dif);
195 ping_portaddr_for_each_entry(sk, hnode, hslot) {
198 pr_debug("iterate\n");
199 if (isk->inet_num != ident)
202 if (skb->protocol == htons(ETH_P_IP) &&
203 sk->sk_family == AF_INET) {
204 pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
205 (int) isk->inet_num, &isk->inet_rcv_saddr,
206 sk->sk_bound_dev_if);
208 if (isk->inet_rcv_saddr &&
209 isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
211 #if IS_ENABLED(CONFIG_IPV6)
212 } else if (skb->protocol == htons(ETH_P_IPV6) &&
213 sk->sk_family == AF_INET6) {
215 pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
217 &sk->sk_v6_rcv_saddr,
218 sk->sk_bound_dev_if);
220 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
221 !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
222 &ipv6_hdr(skb)->daddr))
229 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif &&
230 sk->sk_bound_dev_if != sdif)
242 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
245 kgid_t *data = net->ipv4.ping_group_range.range;
249 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
253 } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
257 int ping_init_sock(struct sock *sk)
259 struct net *net = sock_net(sk);
260 kgid_t group = current_egid();
261 struct group_info *group_info;
266 if (sk->sk_family == AF_INET6)
269 inet_get_ping_group_range_net(net, &low, &high);
270 if (gid_lte(low, group) && gid_lte(group, high))
273 group_info = get_current_groups();
274 for (i = 0; i < group_info->ngroups; i++) {
275 kgid_t gid = group_info->gid[i];
277 if (gid_lte(low, gid) && gid_lte(gid, high))
278 goto out_release_group;
284 put_group_info(group_info);
287 EXPORT_SYMBOL_GPL(ping_init_sock);
289 void ping_close(struct sock *sk, long timeout)
291 pr_debug("ping_close(sk=%p,sk->num=%u)\n",
292 inet_sk(sk), inet_sk(sk)->inet_num);
293 pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
295 sk_common_release(sk);
297 EXPORT_SYMBOL_GPL(ping_close);
299 static int ping_pre_connect(struct sock *sk, struct sockaddr *uaddr,
302 /* This check is replicated from __ip4_datagram_connect() and
303 * intended to prevent BPF program called below from accessing bytes
304 * that are out of the bound specified by user in addr_len.
306 if (addr_len < sizeof(struct sockaddr_in))
309 return BPF_CGROUP_RUN_PROG_INET4_CONNECT_LOCK(sk, uaddr);
312 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
313 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
314 struct sockaddr *uaddr, int addr_len)
316 struct net *net = sock_net(sk);
317 if (sk->sk_family == AF_INET) {
318 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
319 u32 tb_id = RT_TABLE_LOCAL;
322 if (addr_len < sizeof(*addr))
325 if (addr->sin_family != AF_INET &&
326 !(addr->sin_family == AF_UNSPEC &&
327 addr->sin_addr.s_addr == htonl(INADDR_ANY)))
328 return -EAFNOSUPPORT;
330 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
331 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
333 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
336 tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id;
337 chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
339 if (chk_addr_ret == RTN_MULTICAST ||
340 chk_addr_ret == RTN_BROADCAST ||
341 (chk_addr_ret != RTN_LOCAL &&
342 !inet_can_nonlocal_bind(net, isk)))
343 return -EADDRNOTAVAIL;
345 #if IS_ENABLED(CONFIG_IPV6)
346 } else if (sk->sk_family == AF_INET6) {
347 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
348 int addr_type, scoped, has_addr;
349 struct net_device *dev = NULL;
351 if (addr_len < sizeof(*addr))
354 if (addr->sin6_family != AF_INET6)
355 return -EAFNOSUPPORT;
357 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
358 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
360 addr_type = ipv6_addr_type(&addr->sin6_addr);
361 scoped = __ipv6_addr_needs_scope_id(addr_type);
362 if ((addr_type != IPV6_ADDR_ANY &&
363 !(addr_type & IPV6_ADDR_UNICAST)) ||
364 (scoped && !addr->sin6_scope_id))
368 if (addr->sin6_scope_id) {
369 dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
376 if (!dev && sk->sk_bound_dev_if) {
377 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
383 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
387 if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
388 addr_type == IPV6_ADDR_ANY))
389 return -EADDRNOTAVAIL;
392 sk->sk_bound_dev_if = addr->sin6_scope_id;
395 return -EAFNOSUPPORT;
400 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
402 if (saddr->sa_family == AF_INET) {
403 struct inet_sock *isk = inet_sk(sk);
404 struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
405 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
406 #if IS_ENABLED(CONFIG_IPV6)
407 } else if (saddr->sa_family == AF_INET6) {
408 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
409 struct ipv6_pinfo *np = inet6_sk(sk);
410 sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
416 * We need our own bind because there are no privileged id's == local ports.
417 * Moreover, we don't allow binding to multi- and broadcast addresses.
420 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
422 struct inet_sock *isk = inet_sk(sk);
425 int dif = sk->sk_bound_dev_if;
427 err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
434 if (isk->inet_num != 0)
438 snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
439 if (ping_get_port(sk, snum) != 0) {
440 /* Restore possibly modified sk->sk_bound_dev_if by ping_check_bind_addr(). */
441 sk->sk_bound_dev_if = dif;
444 ping_set_saddr(sk, uaddr);
446 pr_debug("after bind(): num = %hu, dif = %d\n",
448 sk->sk_bound_dev_if);
451 if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
452 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
453 #if IS_ENABLED(CONFIG_IPV6)
454 if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
455 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
459 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
460 isk->inet_sport = htons(isk->inet_num);
464 #if IS_ENABLED(CONFIG_IPV6)
465 if (sk->sk_family == AF_INET6)
466 memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
472 pr_debug("ping_v4_bind -> %d\n", err);
475 EXPORT_SYMBOL_GPL(ping_bind);
478 * Is this a supported type of ICMP message?
481 static inline int ping_supported(int family, int type, int code)
483 return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
484 (family == AF_INET && type == ICMP_EXT_ECHO && code == 0) ||
485 (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0) ||
486 (family == AF_INET6 && type == ICMPV6_EXT_ECHO_REQUEST && code == 0);
490 * This routine is called by the ICMP module when it gets some
491 * sort of error condition.
494 void ping_err(struct sk_buff *skb, int offset, u32 info)
497 struct icmphdr *icmph;
498 struct inet_sock *inet_sock;
501 struct net *net = dev_net(skb->dev);
506 if (skb->protocol == htons(ETH_P_IP)) {
508 type = icmp_hdr(skb)->type;
509 code = icmp_hdr(skb)->code;
510 icmph = (struct icmphdr *)(skb->data + offset);
511 } else if (skb->protocol == htons(ETH_P_IPV6)) {
513 type = icmp6_hdr(skb)->icmp6_type;
514 code = icmp6_hdr(skb)->icmp6_code;
515 icmph = (struct icmphdr *) (skb->data + offset);
520 /* We assume the packet has already been checked by icmp_unreach */
522 if (!ping_supported(family, icmph->type, icmph->code))
525 pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
526 skb->protocol, type, code, ntohs(icmph->un.echo.id),
527 ntohs(icmph->un.echo.sequence));
529 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
531 pr_debug("no socket, dropping\n");
532 return; /* No socket for error */
534 pr_debug("err on socket %p\n", sk);
538 inet_sock = inet_sk(sk);
540 if (skb->protocol == htons(ETH_P_IP)) {
543 case ICMP_TIME_EXCEEDED:
546 case ICMP_SOURCE_QUENCH:
547 /* This is not a real error but ping wants to see it.
548 * Report it with some fake errno.
552 case ICMP_PARAMETERPROB:
556 case ICMP_DEST_UNREACH:
557 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
558 ipv4_sk_update_pmtu(skb, sk, info);
559 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
567 if (code <= NR_ICMP_UNREACH) {
568 harderr = icmp_err_convert[code].fatal;
569 err = icmp_err_convert[code].errno;
573 /* See ICMP_SOURCE_QUENCH */
574 ipv4_sk_redirect(skb, sk);
578 #if IS_ENABLED(CONFIG_IPV6)
579 } else if (skb->protocol == htons(ETH_P_IPV6)) {
580 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
585 * RFC1122: OK. Passes ICMP errors back to application, as per
588 if ((family == AF_INET && !inet_sock->recverr) ||
589 (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
590 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
593 if (family == AF_INET) {
594 ip_icmp_error(sk, skb, err, 0 /* no remote port */,
596 #if IS_ENABLED(CONFIG_IPV6)
597 } else if (family == AF_INET6) {
598 pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
608 EXPORT_SYMBOL_GPL(ping_err);
611 * Copy and checksum an ICMP Echo packet from user space into a buffer
612 * starting from the payload.
615 int ping_getfrag(void *from, char *to,
616 int offset, int fraglen, int odd, struct sk_buff *skb)
618 struct pingfakehdr *pfh = from;
621 fraglen -= sizeof(struct icmphdr);
624 if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
625 fraglen, &pfh->wcheck,
626 &pfh->msg->msg_iter))
628 } else if (offset < sizeof(struct icmphdr)) {
631 if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
632 &pfh->msg->msg_iter))
636 #if IS_ENABLED(CONFIG_IPV6)
637 /* For IPv6, checksum each skb as we go along, as expected by
638 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
639 * wcheck, it will be finalized in ping_v4_push_pending_frames.
641 if (pfh->family == AF_INET6) {
642 skb->csum = pfh->wcheck;
643 skb->ip_summed = CHECKSUM_NONE;
650 EXPORT_SYMBOL_GPL(ping_getfrag);
652 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
655 struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
659 pfh->wcheck = csum_partial((char *)&pfh->icmph,
660 sizeof(struct icmphdr), pfh->wcheck);
661 pfh->icmph.checksum = csum_fold(pfh->wcheck);
662 memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
663 skb->ip_summed = CHECKSUM_NONE;
664 return ip_push_pending_frames(sk, fl4);
667 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
668 void *user_icmph, size_t icmph_len)
675 /* Must have at least a full ICMP header. */
683 /* Mirror BSD error message compatibility */
684 if (msg->msg_flags & MSG_OOB)
688 * Fetch the ICMP header provided by the userland.
689 * iovec is modified! The ICMP header is consumed.
691 if (memcpy_from_msg(user_icmph, msg, icmph_len))
694 if (family == AF_INET) {
695 type = ((struct icmphdr *) user_icmph)->type;
696 code = ((struct icmphdr *) user_icmph)->code;
697 #if IS_ENABLED(CONFIG_IPV6)
698 } else if (family == AF_INET6) {
699 type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
700 code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
706 if (!ping_supported(family, type, code))
711 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
713 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
715 struct net *net = sock_net(sk);
717 struct inet_sock *inet = inet_sk(sk);
718 struct ipcm_cookie ipc;
719 struct icmphdr user_icmph;
720 struct pingfakehdr pfh;
721 struct rtable *rt = NULL;
722 struct ip_options_data opt_copy;
724 __be32 saddr, daddr, faddr;
728 pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
730 err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
736 * Get and verify the address.
740 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
741 if (msg->msg_namelen < sizeof(*usin))
743 if (usin->sin_family != AF_INET)
744 return -EAFNOSUPPORT;
745 daddr = usin->sin_addr.s_addr;
748 if (sk->sk_state != TCP_ESTABLISHED)
749 return -EDESTADDRREQ;
750 daddr = inet->inet_daddr;
754 ipcm_init_sk(&ipc, inet);
756 if (msg->msg_controllen) {
757 err = ip_cmsg_send(sk, msg, &ipc, false);
766 struct ip_options_rcu *inet_opt;
769 inet_opt = rcu_dereference(inet->inet_opt);
771 memcpy(&opt_copy, inet_opt,
772 sizeof(*inet_opt) + inet_opt->opt.optlen);
773 ipc.opt = &opt_copy.opt;
779 ipc.addr = faddr = daddr;
781 if (ipc.opt && ipc.opt->opt.srr) {
786 faddr = ipc.opt->opt.faddr;
788 tos = get_rttos(&ipc, inet);
789 if (sock_flag(sk, SOCK_LOCALROUTE) ||
790 (msg->msg_flags & MSG_DONTROUTE) ||
791 (ipc.opt && ipc.opt->opt.is_strictroute)) {
795 if (ipv4_is_multicast(daddr)) {
796 if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
797 ipc.oif = inet->mc_index;
799 saddr = inet->mc_addr;
801 ipc.oif = inet->uc_index;
803 flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
804 RT_SCOPE_UNIVERSE, sk->sk_protocol,
805 inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
808 fl4.fl4_icmp_type = user_icmph.type;
809 fl4.fl4_icmp_code = user_icmph.code;
811 security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4));
812 rt = ip_route_output_flow(net, &fl4, sk);
816 if (err == -ENETUNREACH)
817 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
822 if ((rt->rt_flags & RTCF_BROADCAST) &&
823 !sock_flag(sk, SOCK_BROADCAST))
826 if (msg->msg_flags & MSG_CONFIRM)
831 ipc.addr = fl4.daddr;
835 pfh.icmph.type = user_icmph.type; /* already checked */
836 pfh.icmph.code = user_icmph.code; /* ditto */
837 pfh.icmph.checksum = 0;
838 pfh.icmph.un.echo.id = inet->inet_sport;
839 pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
842 pfh.family = AF_INET;
844 err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
845 0, &ipc, &rt, msg->msg_flags);
847 ip_flush_pending_frames(sk);
849 err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
858 icmp_out_count(sock_net(sk), user_icmph.type);
864 if (msg->msg_flags & MSG_PROBE)
865 dst_confirm_neigh(&rt->dst, &fl4.daddr);
866 if (!(msg->msg_flags & MSG_PROBE) || len)
867 goto back_from_confirm;
872 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int flags,
875 struct inet_sock *isk = inet_sk(sk);
876 int family = sk->sk_family;
880 pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
886 if (flags & MSG_ERRQUEUE)
887 return inet_recv_error(sk, msg, len, addr_len);
889 skb = skb_recv_datagram(sk, flags, &err);
895 msg->msg_flags |= MSG_TRUNC;
899 /* Don't bother checking the checksum */
900 err = skb_copy_datagram_msg(skb, 0, msg, copied);
904 sock_recv_timestamp(msg, sk, skb);
906 /* Copy the address and add cmsg data. */
907 if (family == AF_INET) {
908 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
911 sin->sin_family = AF_INET;
912 sin->sin_port = 0 /* skb->h.uh->source */;
913 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
914 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
915 *addr_len = sizeof(*sin);
919 ip_cmsg_recv(msg, skb);
921 #if IS_ENABLED(CONFIG_IPV6)
922 } else if (family == AF_INET6) {
923 struct ipv6_pinfo *np = inet6_sk(sk);
924 struct ipv6hdr *ip6 = ipv6_hdr(skb);
925 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
928 sin6->sin6_family = AF_INET6;
930 sin6->sin6_addr = ip6->saddr;
931 sin6->sin6_flowinfo = 0;
933 sin6->sin6_flowinfo = ip6_flowinfo(ip6);
934 sin6->sin6_scope_id =
935 ipv6_iface_scope_id(&sin6->sin6_addr,
937 *addr_len = sizeof(*sin6);
940 if (inet6_sk(sk)->rxopt.all)
941 pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
942 if (skb->protocol == htons(ETH_P_IPV6) &&
943 inet6_sk(sk)->rxopt.all)
944 pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
945 else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
946 ip_cmsg_recv(msg, skb);
955 skb_free_datagram(sk, skb);
957 pr_debug("ping_recvmsg -> %d\n", err);
960 EXPORT_SYMBOL_GPL(ping_recvmsg);
962 static enum skb_drop_reason __ping_queue_rcv_skb(struct sock *sk,
965 enum skb_drop_reason reason;
967 pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
968 inet_sk(sk), inet_sk(sk)->inet_num, skb);
969 if (sock_queue_rcv_skb_reason(sk, skb, &reason) < 0) {
970 kfree_skb_reason(skb, reason);
971 pr_debug("ping_queue_rcv_skb -> failed\n");
974 return SKB_NOT_DROPPED_YET;
977 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
979 return __ping_queue_rcv_skb(sk, skb) ? -1 : 0;
981 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
985 * All we need to do is get the socket.
988 enum skb_drop_reason ping_rcv(struct sk_buff *skb)
990 enum skb_drop_reason reason = SKB_DROP_REASON_NO_SOCKET;
992 struct net *net = dev_net(skb->dev);
993 struct icmphdr *icmph = icmp_hdr(skb);
995 /* We assume the packet has already been checked by icmp_rcv */
997 pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
998 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
1000 /* Push ICMP header back */
1001 skb_push(skb, skb->data - (u8 *)icmph);
1003 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
1005 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1007 pr_debug("rcv on socket %p\n", sk);
1009 reason = __ping_queue_rcv_skb(sk, skb2);
1011 reason = SKB_DROP_REASON_NOMEM;
1015 pr_debug("no socket, dropping\n");
1019 EXPORT_SYMBOL_GPL(ping_rcv);
1021 struct proto ping_prot = {
1023 .owner = THIS_MODULE,
1024 .init = ping_init_sock,
1025 .close = ping_close,
1026 .pre_connect = ping_pre_connect,
1027 .connect = ip4_datagram_connect,
1028 .disconnect = __udp_disconnect,
1029 .setsockopt = ip_setsockopt,
1030 .getsockopt = ip_getsockopt,
1031 .sendmsg = ping_v4_sendmsg,
1032 .recvmsg = ping_recvmsg,
1034 .backlog_rcv = ping_queue_rcv_skb,
1035 .release_cb = ip4_datagram_release_cb,
1037 .unhash = ping_unhash,
1038 .get_port = ping_get_port,
1039 .put_port = ping_unhash,
1040 .obj_size = sizeof(struct inet_sock),
1042 EXPORT_SYMBOL(ping_prot);
1044 #ifdef CONFIG_PROC_FS
1046 static struct sock *ping_get_first(struct seq_file *seq, int start)
1049 struct ping_iter_state *state = seq->private;
1050 struct net *net = seq_file_net(seq);
1052 for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1054 struct hlist_nulls_node *node;
1055 struct hlist_nulls_head *hslot;
1057 hslot = &ping_table.hash[state->bucket];
1059 if (hlist_nulls_empty(hslot))
1062 sk_nulls_for_each(sk, node, hslot) {
1063 if (net_eq(sock_net(sk), net) &&
1064 sk->sk_family == state->family)
1073 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1075 struct ping_iter_state *state = seq->private;
1076 struct net *net = seq_file_net(seq);
1079 sk = sk_nulls_next(sk);
1080 } while (sk && (!net_eq(sock_net(sk), net)));
1083 return ping_get_first(seq, state->bucket + 1);
1087 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1089 struct sock *sk = ping_get_first(seq, 0);
1092 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1094 return pos ? NULL : sk;
1097 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1100 struct ping_iter_state *state = seq->private;
1102 state->family = family;
1106 return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1108 EXPORT_SYMBOL_GPL(ping_seq_start);
1110 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1112 return ping_seq_start(seq, pos, AF_INET);
1115 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1119 if (v == SEQ_START_TOKEN)
1120 sk = ping_get_idx(seq, 0);
1122 sk = ping_get_next(seq, v);
1127 EXPORT_SYMBOL_GPL(ping_seq_next);
1129 void ping_seq_stop(struct seq_file *seq, void *v)
1134 EXPORT_SYMBOL_GPL(ping_seq_stop);
1136 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1139 struct inet_sock *inet = inet_sk(sp);
1140 __be32 dest = inet->inet_daddr;
1141 __be32 src = inet->inet_rcv_saddr;
1142 __u16 destp = ntohs(inet->inet_dport);
1143 __u16 srcp = ntohs(inet->inet_sport);
1145 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1146 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u",
1147 bucket, src, srcp, dest, destp, sp->sk_state,
1148 sk_wmem_alloc_get(sp),
1149 sk_rmem_alloc_get(sp),
1151 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1153 refcount_read(&sp->sk_refcnt), sp,
1154 atomic_read(&sp->sk_drops));
1157 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1159 seq_setwidth(seq, 127);
1160 if (v == SEQ_START_TOKEN)
1161 seq_puts(seq, " sl local_address rem_address st tx_queue "
1162 "rx_queue tr tm->when retrnsmt uid timeout "
1163 "inode ref pointer drops");
1165 struct ping_iter_state *state = seq->private;
1167 ping_v4_format_sock(v, seq, state->bucket);
1173 static const struct seq_operations ping_v4_seq_ops = {
1174 .start = ping_v4_seq_start,
1175 .show = ping_v4_seq_show,
1176 .next = ping_seq_next,
1177 .stop = ping_seq_stop,
1180 static int __net_init ping_v4_proc_init_net(struct net *net)
1182 if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1183 sizeof(struct ping_iter_state)))
1188 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1190 remove_proc_entry("icmp", net->proc_net);
1193 static struct pernet_operations ping_v4_net_ops = {
1194 .init = ping_v4_proc_init_net,
1195 .exit = ping_v4_proc_exit_net,
1198 int __init ping_proc_init(void)
1200 return register_pernet_subsys(&ping_v4_net_ops);
1203 void ping_proc_exit(void)
1205 unregister_pernet_subsys(&ping_v4_net_ops);
1210 void __init ping_init(void)
1214 for (i = 0; i < PING_HTABLE_SIZE; i++)
1215 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1216 spin_lock_init(&ping_table.lock);