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 <asm/system.h>
24 #include <linux/uaccess.h>
25 #include <linux/types.h>
26 #include <linux/fcntl.h>
27 #include <linux/socket.h>
28 #include <linux/sockios.h>
30 #include <linux/errno.h>
31 #include <linux/timer.h>
33 #include <linux/inet.h>
34 #include <linux/netdevice.h>
39 #include <net/protocol.h>
40 #include <linux/skbuff.h>
41 #include <linux/proc_fs.h>
46 #include <net/route.h>
47 #include <net/inet_common.h>
48 #include <net/checksum.h>
51 static struct ping_table ping_table;
53 static u16 ping_port_rover;
55 static inline int ping_hashfn(struct net *net, unsigned num, unsigned mask)
57 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 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 struct hlist_nulls_head *hslot;
142 hslot = ping_hashslot(&ping_table, sock_net(sk), isk->inet_num);
143 write_lock_bh(&ping_table.lock);
144 hlist_nulls_del(&sk->sk_nulls_node);
146 isk->inet_num = isk->inet_sport = 0;
147 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
148 write_unlock_bh(&ping_table.lock);
152 static struct sock *ping_v4_lookup(struct net *net, u32 saddr, u32 daddr,
155 struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
156 struct sock *sk = NULL;
157 struct inet_sock *isk;
158 struct hlist_nulls_node *hnode;
160 pr_debug("try to find: num = %d, daddr = %ld, dif = %d\n",
161 (int)ident, (unsigned long)daddr, dif);
162 read_lock_bh(&ping_table.lock);
164 ping_portaddr_for_each_entry(sk, hnode, hslot) {
167 pr_debug("found: %p: num = %d, daddr = %ld, dif = %d\n", sk,
168 (int)isk->inet_num, (unsigned long)isk->inet_rcv_saddr,
169 sk->sk_bound_dev_if);
171 pr_debug("iterate\n");
172 if (isk->inet_num != ident)
174 if (isk->inet_rcv_saddr && isk->inet_rcv_saddr != daddr)
176 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
185 read_unlock_bh(&ping_table.lock);
190 static void inet_get_ping_group_range_net(struct net *net, gid_t *low,
193 gid_t *data = net->ipv4.sysctl_ping_group_range;
196 seq = read_seqbegin(&sysctl_local_ports.lock);
200 } while (read_seqretry(&sysctl_local_ports.lock, seq));
204 static int ping_init_sock(struct sock *sk)
206 struct net *net = sock_net(sk);
207 gid_t group = current_egid();
209 struct group_info *group_info = get_current_groups();
210 int i, j, count = group_info->ngroups;
212 inet_get_ping_group_range_net(net, range, range+1);
213 if (range[0] <= group && group <= range[1])
216 for (i = 0; i < group_info->nblocks; i++) {
217 int cp_count = min_t(int, NGROUPS_PER_BLOCK, count);
219 for (j = 0; j < cp_count; j++) {
220 group = group_info->blocks[i][j];
221 if (range[0] <= group && group <= range[1])
231 static void ping_close(struct sock *sk, long timeout)
233 pr_debug("ping_close(sk=%p,sk->num=%u)\n",
234 inet_sk(sk), inet_sk(sk)->inet_num);
235 pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
237 sk_common_release(sk);
241 * We need our own bind because there are no privileged id's == local ports.
242 * Moreover, we don't allow binding to multi- and broadcast addresses.
245 static int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
247 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
248 struct inet_sock *isk = inet_sk(sk);
253 if (addr_len < sizeof(struct sockaddr_in))
256 pr_debug("ping_v4_bind(sk=%p,sa_addr=%08x,sa_port=%d)\n",
257 sk, addr->sin_addr.s_addr, ntohs(addr->sin_port));
259 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
260 if (addr->sin_addr.s_addr == INADDR_ANY)
261 chk_addr_ret = RTN_LOCAL;
263 if ((sysctl_ip_nonlocal_bind == 0 &&
264 isk->freebind == 0 && isk->transparent == 0 &&
265 chk_addr_ret != RTN_LOCAL) ||
266 chk_addr_ret == RTN_MULTICAST ||
267 chk_addr_ret == RTN_BROADCAST)
268 return -EADDRNOTAVAIL;
273 if (isk->inet_num != 0)
277 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
278 snum = ntohs(addr->sin_port);
279 if (ping_v4_get_port(sk, snum) != 0) {
280 isk->inet_saddr = isk->inet_rcv_saddr = 0;
284 pr_debug("after bind(): num = %d, daddr = %ld, dif = %d\n",
286 (unsigned long) isk->inet_rcv_saddr,
287 (int)sk->sk_bound_dev_if);
290 if (isk->inet_rcv_saddr)
291 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
293 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
294 isk->inet_sport = htons(isk->inet_num);
300 pr_debug("ping_v4_bind -> %d\n", err);
305 * Is this a supported type of ICMP message?
308 static inline int ping_supported(int type, int code)
310 if (type == ICMP_ECHO && code == 0)
316 * This routine is called by the ICMP module when it gets some
317 * sort of error condition.
320 static int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
322 void ping_err(struct sk_buff *skb, u32 info)
324 struct iphdr *iph = (struct iphdr *)skb->data;
325 struct icmphdr *icmph = (struct icmphdr *)(skb->data+(iph->ihl<<2));
326 struct inet_sock *inet_sock;
327 int type = icmph->type;
328 int code = icmph->code;
329 struct net *net = dev_net(skb->dev);
334 /* We assume the packet has already been checked by icmp_unreach */
336 if (!ping_supported(icmph->type, icmph->code))
339 pr_debug("ping_err(type=%04x,code=%04x,id=%04x,seq=%04x)\n", type,
340 code, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
342 sk = ping_v4_lookup(net, iph->daddr, iph->saddr,
343 ntohs(icmph->un.echo.id), skb->dev->ifindex);
345 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
346 pr_debug("no socket, dropping\n");
347 return; /* No socket for error */
349 pr_debug("err on socket %p\n", sk);
353 inet_sock = inet_sk(sk);
357 case ICMP_TIME_EXCEEDED:
360 case ICMP_SOURCE_QUENCH:
361 /* This is not a real error but ping wants to see it.
362 * Report it with some fake errno. */
365 case ICMP_PARAMETERPROB:
369 case ICMP_DEST_UNREACH:
370 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
371 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
379 if (code <= NR_ICMP_UNREACH) {
380 harderr = icmp_err_convert[code].fatal;
381 err = icmp_err_convert[code].errno;
385 /* See ICMP_SOURCE_QUENCH */
391 * RFC1122: OK. Passes ICMP errors back to application, as per
394 if (!inet_sock->recverr) {
395 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
398 ip_icmp_error(sk, skb, err, 0 /* no remote port */,
402 sk->sk_error_report(sk);
408 * Copy and checksum an ICMP Echo packet from user space into a buffer.
412 struct icmphdr icmph;
417 static int ping_getfrag(void *from, char * to,
418 int offset, int fraglen, int odd, struct sk_buff *skb)
420 struct pingfakehdr *pfh = (struct pingfakehdr *)from;
423 if (fraglen < sizeof(struct icmphdr))
425 if (csum_partial_copy_fromiovecend(to + sizeof(struct icmphdr),
426 pfh->iov, 0, fraglen - sizeof(struct icmphdr),
432 if (offset < sizeof(struct icmphdr))
434 if (csum_partial_copy_fromiovecend
435 (to, pfh->iov, offset - sizeof(struct icmphdr),
436 fraglen, &pfh->wcheck))
441 static int ping_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
444 struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
446 pfh->wcheck = csum_partial((char *)&pfh->icmph,
447 sizeof(struct icmphdr), pfh->wcheck);
448 pfh->icmph.checksum = csum_fold(pfh->wcheck);
449 memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
450 skb->ip_summed = CHECKSUM_NONE;
451 return ip_push_pending_frames(sk, fl4);
454 static int ping_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
457 struct net *net = sock_net(sk);
459 struct inet_sock *inet = inet_sk(sk);
460 struct ipcm_cookie ipc;
461 struct icmphdr user_icmph;
462 struct pingfakehdr pfh;
463 struct rtable *rt = NULL;
464 struct ip_options_data opt_copy;
466 u32 saddr, daddr, faddr;
470 pr_debug("ping_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
480 /* Mirror BSD error message compatibility */
481 if (msg->msg_flags & MSG_OOB)
485 * Fetch the ICMP header provided by the userland.
489 if (memcpy_fromiovec((u8 *)&user_icmph, msg->msg_iov,
490 sizeof(struct icmphdr)))
492 if (!ping_supported(user_icmph.type, user_icmph.code))
496 * Get and verify the address.
500 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
501 if (msg->msg_namelen < sizeof(*usin))
503 if (usin->sin_family != AF_INET)
505 daddr = usin->sin_addr.s_addr;
508 if (sk->sk_state != TCP_ESTABLISHED)
509 return -EDESTADDRREQ;
510 daddr = inet->inet_daddr;
514 ipc.addr = inet->inet_saddr;
516 ipc.oif = sk->sk_bound_dev_if;
518 err = sock_tx_timestamp(sk, &ipc.tx_flags);
522 if (msg->msg_controllen) {
523 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
530 struct ip_options_rcu *inet_opt;
533 inet_opt = rcu_dereference(inet->inet_opt);
535 memcpy(&opt_copy, inet_opt,
536 sizeof(*inet_opt) + inet_opt->opt.optlen);
537 ipc.opt = &opt_copy.opt;
543 ipc.addr = faddr = daddr;
545 if (ipc.opt && ipc.opt->opt.srr) {
548 faddr = ipc.opt->opt.faddr;
550 tos = RT_TOS(inet->tos);
551 if (sock_flag(sk, SOCK_LOCALROUTE) ||
552 (msg->msg_flags & MSG_DONTROUTE) ||
553 (ipc.opt && ipc.opt->opt.is_strictroute)) {
557 if (ipv4_is_multicast(daddr)) {
559 ipc.oif = inet->mc_index;
561 saddr = inet->mc_addr;
564 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
565 RT_SCOPE_UNIVERSE, sk->sk_protocol,
566 inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
568 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
569 rt = ip_route_output_flow(net, &fl4, sk);
573 if (err == -ENETUNREACH)
574 IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
579 if ((rt->rt_flags & RTCF_BROADCAST) &&
580 !sock_flag(sk, SOCK_BROADCAST))
583 if (msg->msg_flags & MSG_CONFIRM)
588 ipc.addr = fl4.daddr;
592 pfh.icmph.type = user_icmph.type; /* already checked */
593 pfh.icmph.code = user_icmph.code; /* ditto */
594 pfh.icmph.checksum = 0;
595 pfh.icmph.un.echo.id = inet->inet_sport;
596 pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
597 pfh.iov = msg->msg_iov;
600 err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
601 0, &ipc, &rt, msg->msg_flags);
603 ip_flush_pending_frames(sk);
605 err = ping_push_pending_frames(sk, &pfh, &fl4);
613 icmp_out_count(sock_net(sk), user_icmph.type);
619 dst_confirm(&rt->dst);
620 if (!(msg->msg_flags & MSG_PROBE) || len)
621 goto back_from_confirm;
626 static int ping_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
627 size_t len, int noblock, int flags, int *addr_len)
629 struct inet_sock *isk = inet_sk(sk);
630 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
634 pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
640 *addr_len = sizeof(*sin);
642 if (flags & MSG_ERRQUEUE)
643 return ip_recv_error(sk, msg, len);
645 skb = skb_recv_datagram(sk, flags, noblock, &err);
651 msg->msg_flags |= MSG_TRUNC;
655 /* Don't bother checking the checksum */
656 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
660 sock_recv_timestamp(msg, sk, skb);
662 /* Copy the address. */
664 sin->sin_family = AF_INET;
665 sin->sin_port = 0 /* skb->h.uh->source */;
666 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
667 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
670 ip_cmsg_recv(msg, skb);
674 skb_free_datagram(sk, skb);
676 pr_debug("ping_recvmsg -> %d\n", err);
680 static int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
682 pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
683 inet_sk(sk), inet_sk(sk)->inet_num, skb);
684 if (sock_queue_rcv_skb(sk, skb) < 0) {
685 ICMP_INC_STATS_BH(sock_net(sk), ICMP_MIB_INERRORS);
687 pr_debug("ping_queue_rcv_skb -> failed\n");
695 * All we need to do is get the socket.
698 void ping_rcv(struct sk_buff *skb)
701 struct net *net = dev_net(skb->dev);
702 struct iphdr *iph = ip_hdr(skb);
703 struct icmphdr *icmph = icmp_hdr(skb);
704 u32 saddr = iph->saddr;
705 u32 daddr = iph->daddr;
707 /* We assume the packet has already been checked by icmp_rcv */
709 pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
710 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
712 /* Push ICMP header back */
713 skb_push(skb, skb->data - (u8 *)icmph);
715 sk = ping_v4_lookup(net, saddr, daddr, ntohs(icmph->un.echo.id),
718 pr_debug("rcv on socket %p\n", sk);
719 ping_queue_rcv_skb(sk, skb_get(skb));
723 pr_debug("no socket, dropping\n");
725 /* We're called from icmp_rcv(). kfree_skb() is done there. */
728 struct proto ping_prot = {
730 .owner = THIS_MODULE,
731 .init = ping_init_sock,
733 .connect = ip4_datagram_connect,
734 .disconnect = udp_disconnect,
735 .setsockopt = ip_setsockopt,
736 .getsockopt = ip_getsockopt,
737 .sendmsg = ping_sendmsg,
738 .recvmsg = ping_recvmsg,
740 .backlog_rcv = ping_queue_rcv_skb,
741 .hash = ping_v4_hash,
742 .unhash = ping_v4_unhash,
743 .get_port = ping_v4_get_port,
744 .obj_size = sizeof(struct inet_sock),
746 EXPORT_SYMBOL(ping_prot);
748 #ifdef CONFIG_PROC_FS
750 static struct sock *ping_get_first(struct seq_file *seq, int start)
753 struct ping_iter_state *state = seq->private;
754 struct net *net = seq_file_net(seq);
756 for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
758 struct hlist_nulls_node *node;
759 struct hlist_nulls_head *hslot;
761 hslot = &ping_table.hash[state->bucket];
763 if (hlist_nulls_empty(hslot))
766 sk_nulls_for_each(sk, node, hslot) {
767 if (net_eq(sock_net(sk), net))
776 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
778 struct ping_iter_state *state = seq->private;
779 struct net *net = seq_file_net(seq);
782 sk = sk_nulls_next(sk);
783 } while (sk && (!net_eq(sock_net(sk), net)));
786 return ping_get_first(seq, state->bucket + 1);
790 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
792 struct sock *sk = ping_get_first(seq, 0);
795 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
797 return pos ? NULL : sk;
800 static void *ping_seq_start(struct seq_file *seq, loff_t *pos)
802 struct ping_iter_state *state = seq->private;
805 read_lock_bh(&ping_table.lock);
807 return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
810 static void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
814 if (v == SEQ_START_TOKEN)
815 sk = ping_get_idx(seq, 0);
817 sk = ping_get_next(seq, v);
823 static void ping_seq_stop(struct seq_file *seq, void *v)
825 read_unlock_bh(&ping_table.lock);
828 static void ping_format_sock(struct sock *sp, struct seq_file *f,
829 int bucket, int *len)
831 struct inet_sock *inet = inet_sk(sp);
832 __be32 dest = inet->inet_daddr;
833 __be32 src = inet->inet_rcv_saddr;
834 __u16 destp = ntohs(inet->inet_dport);
835 __u16 srcp = ntohs(inet->inet_sport);
837 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
838 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d%n",
839 bucket, src, srcp, dest, destp, sp->sk_state,
840 sk_wmem_alloc_get(sp),
841 sk_rmem_alloc_get(sp),
842 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
843 atomic_read(&sp->sk_refcnt), sp,
844 atomic_read(&sp->sk_drops), len);
847 static int ping_seq_show(struct seq_file *seq, void *v)
849 if (v == SEQ_START_TOKEN)
850 seq_printf(seq, "%-127s\n",
851 " sl local_address rem_address st tx_queue "
852 "rx_queue tr tm->when retrnsmt uid timeout "
853 "inode ref pointer drops");
855 struct ping_iter_state *state = seq->private;
858 ping_format_sock(v, seq, state->bucket, &len);
859 seq_printf(seq, "%*s\n", 127 - len, "");
864 static const struct seq_operations ping_seq_ops = {
865 .show = ping_seq_show,
866 .start = ping_seq_start,
867 .next = ping_seq_next,
868 .stop = ping_seq_stop,
871 static int ping_seq_open(struct inode *inode, struct file *file)
873 return seq_open_net(inode, file, &ping_seq_ops,
874 sizeof(struct ping_iter_state));
877 static const struct file_operations ping_seq_fops = {
878 .open = ping_seq_open,
881 .release = seq_release_net,
884 static int ping_proc_register(struct net *net)
886 struct proc_dir_entry *p;
889 p = proc_net_fops_create(net, "icmp", S_IRUGO, &ping_seq_fops);
895 static void ping_proc_unregister(struct net *net)
897 proc_net_remove(net, "icmp");
901 static int __net_init ping_proc_init_net(struct net *net)
903 return ping_proc_register(net);
906 static void __net_exit ping_proc_exit_net(struct net *net)
908 ping_proc_unregister(net);
911 static struct pernet_operations ping_net_ops = {
912 .init = ping_proc_init_net,
913 .exit = ping_proc_exit_net,
916 int __init ping_proc_init(void)
918 return register_pernet_subsys(&ping_net_ops);
921 void ping_proc_exit(void)
923 unregister_pernet_subsys(&ping_net_ops);
928 void __init ping_init(void)
932 for (i = 0; i < PING_HTABLE_SIZE; i++)
933 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
934 rwlock_init(&ping_table.lock);