2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
13 * Based on ipv4/udp.c code.
15 * Authors: Vasiliy Kulikov / Openwall (for Linux 2.6),
16 * Pavel Kankovsky (for Linux 2.4.32)
18 * Pavel gave all rights to bugs to Vasiliy,
19 * none of the bugs are Pavel's now.
23 #include <linux/uaccess.h>
24 #include <linux/types.h>
25 #include <linux/fcntl.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
29 #include <linux/errno.h>
30 #include <linux/timer.h>
32 #include <linux/inet.h>
33 #include <linux/netdevice.h>
38 #include <net/protocol.h>
39 #include <linux/skbuff.h>
40 #include <linux/proc_fs.h>
41 #include <linux/export.h>
45 #include <net/route.h>
46 #include <net/inet_common.h>
47 #include <net/checksum.h>
50 static struct ping_table ping_table;
52 static u16 ping_port_rover;
54 static inline int ping_hashfn(struct net *net, unsigned int num, unsigned int mask)
56 int res = (num + net_hash_mix(net)) & mask;
58 pr_debug("hash(%d) = %d\n", num, res);
62 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
63 struct net *net, unsigned int num)
65 return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
68 static int ping_v4_get_port(struct sock *sk, unsigned short ident)
70 struct hlist_nulls_node *node;
71 struct hlist_nulls_head *hlist;
72 struct inet_sock *isk, *isk2;
73 struct sock *sk2 = NULL;
76 write_lock_bh(&ping_table.lock);
79 u16 result = ping_port_rover + 1;
81 for (i = 0; i < (1L << 16); i++, result++) {
83 result++; /* avoid zero */
84 hlist = ping_hashslot(&ping_table, sock_net(sk),
86 ping_portaddr_for_each_entry(sk2, node, hlist) {
89 if (isk2->inet_num == result)
94 ping_port_rover = ident = result;
102 hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
103 ping_portaddr_for_each_entry(sk2, node, hlist) {
106 if ((isk2->inet_num == ident) &&
108 (!sk2->sk_reuse || !sk->sk_reuse))
113 pr_debug("found port/ident = %d\n", ident);
114 isk->inet_num = ident;
115 if (sk_unhashed(sk)) {
116 pr_debug("was not hashed\n");
118 hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
119 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
121 write_unlock_bh(&ping_table.lock);
125 write_unlock_bh(&ping_table.lock);
129 static void ping_v4_hash(struct sock *sk)
131 pr_debug("ping_v4_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
132 BUG(); /* "Please do not press this button again." */
135 static void ping_v4_unhash(struct sock *sk)
137 struct inet_sock *isk = inet_sk(sk);
138 pr_debug("ping_v4_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
140 write_lock_bh(&ping_table.lock);
141 hlist_nulls_del(&sk->sk_nulls_node);
145 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
146 write_unlock_bh(&ping_table.lock);
150 static struct sock *ping_v4_lookup(struct net *net, __be32 saddr, __be32 daddr,
153 struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
154 struct sock *sk = NULL;
155 struct inet_sock *isk;
156 struct hlist_nulls_node *hnode;
158 pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
159 (int)ident, &daddr, dif);
160 read_lock_bh(&ping_table.lock);
162 ping_portaddr_for_each_entry(sk, hnode, hslot) {
165 pr_debug("found: %p: num = %d, daddr = %pI4, dif = %d\n", sk,
166 (int)isk->inet_num, &isk->inet_rcv_saddr,
167 sk->sk_bound_dev_if);
169 pr_debug("iterate\n");
170 if (isk->inet_num != ident)
172 if (isk->inet_rcv_saddr && isk->inet_rcv_saddr != daddr)
174 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
183 read_unlock_bh(&ping_table.lock);
188 static void inet_get_ping_group_range_net(struct net *net, gid_t *low,
191 gid_t *data = net->ipv4.sysctl_ping_group_range;
195 seq = read_seqbegin(&sysctl_local_ports.lock);
199 } while (read_seqretry(&sysctl_local_ports.lock, seq));
203 static int ping_init_sock(struct sock *sk)
205 struct net *net = sock_net(sk);
206 gid_t group = current_egid();
208 struct group_info *group_info = get_current_groups();
209 int i, j, count = group_info->ngroups;
212 inet_get_ping_group_range_net(net, range, range+1);
213 low = make_kgid(&init_user_ns, range[0]);
214 high = make_kgid(&init_user_ns, range[1]);
215 if (!gid_valid(low) || !gid_valid(high) || gid_lt(high, low))
218 if (range[0] <= group && group <= range[1])
221 for (i = 0; i < group_info->nblocks; i++) {
222 int cp_count = min_t(int, NGROUPS_PER_BLOCK, count);
223 for (j = 0; j < cp_count; j++) {
224 kgid_t gid = group_info->blocks[i][j];
225 if (gid_lte(low, gid) && gid_lte(gid, high))
235 static void ping_close(struct sock *sk, long timeout)
237 pr_debug("ping_close(sk=%p,sk->num=%u)\n",
238 inet_sk(sk), inet_sk(sk)->inet_num);
239 pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
241 sk_common_release(sk);
245 * We need our own bind because there are no privileged id's == local ports.
246 * Moreover, we don't allow binding to multi- and broadcast addresses.
249 static int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
251 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
252 struct inet_sock *isk = inet_sk(sk);
257 if (addr_len < sizeof(struct sockaddr_in))
260 pr_debug("ping_v4_bind(sk=%p,sa_addr=%08x,sa_port=%d)\n",
261 sk, addr->sin_addr.s_addr, ntohs(addr->sin_port));
263 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
264 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
265 chk_addr_ret = RTN_LOCAL;
267 if ((sysctl_ip_nonlocal_bind == 0 &&
268 isk->freebind == 0 && isk->transparent == 0 &&
269 chk_addr_ret != RTN_LOCAL) ||
270 chk_addr_ret == RTN_MULTICAST ||
271 chk_addr_ret == RTN_BROADCAST)
272 return -EADDRNOTAVAIL;
277 if (isk->inet_num != 0)
281 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
282 snum = ntohs(addr->sin_port);
283 if (ping_v4_get_port(sk, snum) != 0) {
284 isk->inet_saddr = isk->inet_rcv_saddr = 0;
288 pr_debug("after bind(): num = %d, daddr = %pI4, dif = %d\n",
290 &isk->inet_rcv_saddr,
291 (int)sk->sk_bound_dev_if);
294 if (isk->inet_rcv_saddr)
295 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
297 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
298 isk->inet_sport = htons(isk->inet_num);
304 pr_debug("ping_v4_bind -> %d\n", err);
309 * Is this a supported type of ICMP message?
312 static inline int ping_supported(int type, int code)
314 if (type == ICMP_ECHO && code == 0)
320 * This routine is called by the ICMP module when it gets some
321 * sort of error condition.
324 static int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
326 void ping_err(struct sk_buff *skb, u32 info)
328 struct iphdr *iph = (struct iphdr *)skb->data;
329 struct icmphdr *icmph = (struct icmphdr *)(skb->data+(iph->ihl<<2));
330 struct inet_sock *inet_sock;
331 int type = icmph->type;
332 int code = icmph->code;
333 struct net *net = dev_net(skb->dev);
338 /* We assume the packet has already been checked by icmp_unreach */
340 if (!ping_supported(icmph->type, icmph->code))
343 pr_debug("ping_err(type=%04x,code=%04x,id=%04x,seq=%04x)\n", type,
344 code, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
346 sk = ping_v4_lookup(net, iph->daddr, iph->saddr,
347 ntohs(icmph->un.echo.id), skb->dev->ifindex);
349 pr_debug("no socket, dropping\n");
350 return; /* No socket for error */
352 pr_debug("err on socket %p\n", sk);
356 inet_sock = inet_sk(sk);
360 case ICMP_TIME_EXCEEDED:
363 case ICMP_SOURCE_QUENCH:
364 /* This is not a real error but ping wants to see it.
365 * Report it with some fake errno. */
368 case ICMP_PARAMETERPROB:
372 case ICMP_DEST_UNREACH:
373 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
374 ipv4_sk_update_pmtu(skb, sk, info);
375 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
383 if (code <= NR_ICMP_UNREACH) {
384 harderr = icmp_err_convert[code].fatal;
385 err = icmp_err_convert[code].errno;
389 /* See ICMP_SOURCE_QUENCH */
390 ipv4_sk_redirect(skb, sk);
396 * RFC1122: OK. Passes ICMP errors back to application, as per
399 if (!inet_sock->recverr) {
400 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
403 ip_icmp_error(sk, skb, err, 0 /* no remote port */,
407 sk->sk_error_report(sk);
413 * Copy and checksum an ICMP Echo packet from user space into a buffer.
417 struct icmphdr icmph;
422 static int ping_getfrag(void *from, char *to,
423 int offset, int fraglen, int odd, struct sk_buff *skb)
425 struct pingfakehdr *pfh = (struct pingfakehdr *)from;
428 if (fraglen < sizeof(struct icmphdr))
430 if (csum_partial_copy_fromiovecend(to + sizeof(struct icmphdr),
431 pfh->iov, 0, fraglen - sizeof(struct icmphdr),
437 if (offset < sizeof(struct icmphdr))
439 if (csum_partial_copy_fromiovecend
440 (to, pfh->iov, offset - sizeof(struct icmphdr),
441 fraglen, &pfh->wcheck))
446 static int ping_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
449 struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
451 pfh->wcheck = csum_partial((char *)&pfh->icmph,
452 sizeof(struct icmphdr), pfh->wcheck);
453 pfh->icmph.checksum = csum_fold(pfh->wcheck);
454 memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
455 skb->ip_summed = CHECKSUM_NONE;
456 return ip_push_pending_frames(sk, fl4);
459 static int ping_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
462 struct net *net = sock_net(sk);
464 struct inet_sock *inet = inet_sk(sk);
465 struct ipcm_cookie ipc;
466 struct icmphdr user_icmph;
467 struct pingfakehdr pfh;
468 struct rtable *rt = NULL;
469 struct ip_options_data opt_copy;
471 __be32 saddr, daddr, faddr;
475 pr_debug("ping_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
485 /* Mirror BSD error message compatibility */
486 if (msg->msg_flags & MSG_OOB)
490 * Fetch the ICMP header provided by the userland.
494 if (memcpy_fromiovec((u8 *)&user_icmph, msg->msg_iov,
495 sizeof(struct icmphdr)))
497 if (!ping_supported(user_icmph.type, user_icmph.code))
501 * Get and verify the address.
505 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
506 if (msg->msg_namelen < sizeof(*usin))
508 if (usin->sin_family != AF_INET)
510 daddr = usin->sin_addr.s_addr;
513 if (sk->sk_state != TCP_ESTABLISHED)
514 return -EDESTADDRREQ;
515 daddr = inet->inet_daddr;
519 ipc.addr = inet->inet_saddr;
521 ipc.oif = sk->sk_bound_dev_if;
523 err = sock_tx_timestamp(sk, &ipc.tx_flags);
527 if (msg->msg_controllen) {
528 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
535 struct ip_options_rcu *inet_opt;
538 inet_opt = rcu_dereference(inet->inet_opt);
540 memcpy(&opt_copy, inet_opt,
541 sizeof(*inet_opt) + inet_opt->opt.optlen);
542 ipc.opt = &opt_copy.opt;
548 ipc.addr = faddr = daddr;
550 if (ipc.opt && ipc.opt->opt.srr) {
553 faddr = ipc.opt->opt.faddr;
555 tos = RT_TOS(inet->tos);
556 if (sock_flag(sk, SOCK_LOCALROUTE) ||
557 (msg->msg_flags & MSG_DONTROUTE) ||
558 (ipc.opt && ipc.opt->opt.is_strictroute)) {
562 if (ipv4_is_multicast(daddr)) {
564 ipc.oif = inet->mc_index;
566 saddr = inet->mc_addr;
568 ipc.oif = inet->uc_index;
570 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
571 RT_SCOPE_UNIVERSE, sk->sk_protocol,
572 inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
574 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
575 rt = ip_route_output_flow(net, &fl4, sk);
579 if (err == -ENETUNREACH)
580 IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
585 if ((rt->rt_flags & RTCF_BROADCAST) &&
586 !sock_flag(sk, SOCK_BROADCAST))
589 if (msg->msg_flags & MSG_CONFIRM)
594 ipc.addr = fl4.daddr;
598 pfh.icmph.type = user_icmph.type; /* already checked */
599 pfh.icmph.code = user_icmph.code; /* ditto */
600 pfh.icmph.checksum = 0;
601 pfh.icmph.un.echo.id = inet->inet_sport;
602 pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
603 pfh.iov = msg->msg_iov;
606 err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
607 0, &ipc, &rt, msg->msg_flags);
609 ip_flush_pending_frames(sk);
611 err = ping_push_pending_frames(sk, &pfh, &fl4);
619 icmp_out_count(sock_net(sk), user_icmph.type);
625 dst_confirm(&rt->dst);
626 if (!(msg->msg_flags & MSG_PROBE) || len)
627 goto back_from_confirm;
632 static int ping_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
633 size_t len, int noblock, int flags, int *addr_len)
635 struct inet_sock *isk = inet_sk(sk);
636 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
640 pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
647 *addr_len = sizeof(*sin);
649 if (flags & MSG_ERRQUEUE)
650 return ip_recv_error(sk, msg, len);
652 skb = skb_recv_datagram(sk, flags, noblock, &err);
658 msg->msg_flags |= MSG_TRUNC;
662 /* Don't bother checking the checksum */
663 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
667 sock_recv_timestamp(msg, sk, skb);
669 /* Copy the address. */
671 sin->sin_family = AF_INET;
672 sin->sin_port = 0 /* skb->h.uh->source */;
673 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
674 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
677 ip_cmsg_recv(msg, skb);
681 skb_free_datagram(sk, skb);
683 pr_debug("ping_recvmsg -> %d\n", err);
687 static int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
689 pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
690 inet_sk(sk), inet_sk(sk)->inet_num, skb);
691 if (sock_queue_rcv_skb(sk, skb) < 0) {
693 pr_debug("ping_queue_rcv_skb -> failed\n");
701 * All we need to do is get the socket.
704 void ping_rcv(struct sk_buff *skb)
707 struct net *net = dev_net(skb->dev);
708 struct iphdr *iph = ip_hdr(skb);
709 struct icmphdr *icmph = icmp_hdr(skb);
710 __be32 saddr = iph->saddr;
711 __be32 daddr = iph->daddr;
713 /* We assume the packet has already been checked by icmp_rcv */
715 pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
716 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
718 /* Push ICMP header back */
719 skb_push(skb, skb->data - (u8 *)icmph);
721 sk = ping_v4_lookup(net, saddr, daddr, ntohs(icmph->un.echo.id),
724 pr_debug("rcv on socket %p\n", sk);
725 ping_queue_rcv_skb(sk, skb_get(skb));
729 pr_debug("no socket, dropping\n");
731 /* We're called from icmp_rcv(). kfree_skb() is done there. */
734 struct proto ping_prot = {
736 .owner = THIS_MODULE,
737 .init = ping_init_sock,
739 .connect = ip4_datagram_connect,
740 .disconnect = udp_disconnect,
741 .setsockopt = ip_setsockopt,
742 .getsockopt = ip_getsockopt,
743 .sendmsg = ping_sendmsg,
744 .recvmsg = ping_recvmsg,
746 .backlog_rcv = ping_queue_rcv_skb,
747 .hash = ping_v4_hash,
748 .unhash = ping_v4_unhash,
749 .get_port = ping_v4_get_port,
750 .obj_size = sizeof(struct inet_sock),
752 EXPORT_SYMBOL(ping_prot);
754 #ifdef CONFIG_PROC_FS
756 static struct sock *ping_get_first(struct seq_file *seq, int start)
759 struct ping_iter_state *state = seq->private;
760 struct net *net = seq_file_net(seq);
762 for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
764 struct hlist_nulls_node *node;
765 struct hlist_nulls_head *hslot;
767 hslot = &ping_table.hash[state->bucket];
769 if (hlist_nulls_empty(hslot))
772 sk_nulls_for_each(sk, node, hslot) {
773 if (net_eq(sock_net(sk), net))
782 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
784 struct ping_iter_state *state = seq->private;
785 struct net *net = seq_file_net(seq);
788 sk = sk_nulls_next(sk);
789 } while (sk && (!net_eq(sock_net(sk), net)));
792 return ping_get_first(seq, state->bucket + 1);
796 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
798 struct sock *sk = ping_get_first(seq, 0);
801 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
803 return pos ? NULL : sk;
806 static void *ping_seq_start(struct seq_file *seq, loff_t *pos)
808 struct ping_iter_state *state = seq->private;
811 read_lock_bh(&ping_table.lock);
813 return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
816 static void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
820 if (v == SEQ_START_TOKEN)
821 sk = ping_get_idx(seq, 0);
823 sk = ping_get_next(seq, v);
829 static void ping_seq_stop(struct seq_file *seq, void *v)
831 read_unlock_bh(&ping_table.lock);
834 static void ping_format_sock(struct sock *sp, struct seq_file *f,
835 int bucket, int *len)
837 struct inet_sock *inet = inet_sk(sp);
838 __be32 dest = inet->inet_daddr;
839 __be32 src = inet->inet_rcv_saddr;
840 __u16 destp = ntohs(inet->inet_dport);
841 __u16 srcp = ntohs(inet->inet_sport);
843 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
844 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d%n",
845 bucket, src, srcp, dest, destp, sp->sk_state,
846 sk_wmem_alloc_get(sp),
847 sk_rmem_alloc_get(sp),
848 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
849 atomic_read(&sp->sk_refcnt), sp,
850 atomic_read(&sp->sk_drops), len);
853 static int ping_seq_show(struct seq_file *seq, void *v)
855 if (v == SEQ_START_TOKEN)
856 seq_printf(seq, "%-127s\n",
857 " sl local_address rem_address st tx_queue "
858 "rx_queue tr tm->when retrnsmt uid timeout "
859 "inode ref pointer drops");
861 struct ping_iter_state *state = seq->private;
864 ping_format_sock(v, seq, state->bucket, &len);
865 seq_printf(seq, "%*s\n", 127 - len, "");
870 static const struct seq_operations ping_seq_ops = {
871 .show = ping_seq_show,
872 .start = ping_seq_start,
873 .next = ping_seq_next,
874 .stop = ping_seq_stop,
877 static int ping_seq_open(struct inode *inode, struct file *file)
879 return seq_open_net(inode, file, &ping_seq_ops,
880 sizeof(struct ping_iter_state));
883 static const struct file_operations ping_seq_fops = {
884 .open = ping_seq_open,
887 .release = seq_release_net,
890 static int ping_proc_register(struct net *net)
892 struct proc_dir_entry *p;
895 p = proc_net_fops_create(net, "icmp", S_IRUGO, &ping_seq_fops);
901 static void ping_proc_unregister(struct net *net)
903 proc_net_remove(net, "icmp");
907 static int __net_init ping_proc_init_net(struct net *net)
909 return ping_proc_register(net);
912 static void __net_exit ping_proc_exit_net(struct net *net)
914 ping_proc_unregister(net);
917 static struct pernet_operations ping_net_ops = {
918 .init = ping_proc_init_net,
919 .exit = ping_proc_exit_net,
922 int __init ping_proc_init(void)
924 return register_pernet_subsys(&ping_net_ops);
927 void ping_proc_exit(void)
929 unregister_pernet_subsys(&ping_net_ops);
934 void __init ping_init(void)
938 for (i = 0; i < PING_HTABLE_SIZE; i++)
939 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
940 rwlock_init(&ping_table.lock);