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 num, unsigned mask)
56 int res = (num + net_hash_mix(net)) & mask;
57 pr_debug("hash(%d) = %d\n", num, res);
61 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
62 struct net *net, unsigned num)
64 return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
67 static int ping_v4_get_port(struct sock *sk, unsigned short ident)
69 struct hlist_nulls_node *node;
70 struct hlist_nulls_head *hlist;
71 struct inet_sock *isk, *isk2;
72 struct sock *sk2 = NULL;
75 write_lock_bh(&ping_table.lock);
78 u16 result = ping_port_rover + 1;
80 for (i = 0; i < (1L << 16); i++, result++) {
82 result++; /* avoid zero */
83 hlist = ping_hashslot(&ping_table, sock_net(sk),
85 ping_portaddr_for_each_entry(sk2, node, hlist) {
88 if (isk2->inet_num == result)
93 ping_port_rover = ident = result;
101 hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
102 ping_portaddr_for_each_entry(sk2, node, hlist) {
105 if ((isk2->inet_num == ident) &&
107 (!sk2->sk_reuse || !sk->sk_reuse))
112 pr_debug("found port/ident = %d\n", ident);
113 isk->inet_num = ident;
114 if (sk_unhashed(sk)) {
115 pr_debug("was not hashed\n");
117 hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
118 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
120 write_unlock_bh(&ping_table.lock);
124 write_unlock_bh(&ping_table.lock);
128 static void ping_v4_hash(struct sock *sk)
130 pr_debug("ping_v4_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
131 BUG(); /* "Please do not press this button again." */
134 static void ping_v4_unhash(struct sock *sk)
136 struct inet_sock *isk = inet_sk(sk);
137 pr_debug("ping_v4_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
139 write_lock_bh(&ping_table.lock);
140 hlist_nulls_del(&sk->sk_nulls_node);
144 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
145 write_unlock_bh(&ping_table.lock);
149 static struct sock *ping_v4_lookup(struct net *net, __be32 saddr, __be32 daddr,
152 struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
153 struct sock *sk = NULL;
154 struct inet_sock *isk;
155 struct hlist_nulls_node *hnode;
157 pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
158 (int)ident, &daddr, dif);
159 read_lock_bh(&ping_table.lock);
161 ping_portaddr_for_each_entry(sk, hnode, hslot) {
164 pr_debug("found: %p: num = %d, daddr = %pI4, dif = %d\n", sk,
165 (int)isk->inet_num, &isk->inet_rcv_saddr,
166 sk->sk_bound_dev_if);
168 pr_debug("iterate\n");
169 if (isk->inet_num != ident)
171 if (isk->inet_rcv_saddr && isk->inet_rcv_saddr != daddr)
173 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
182 read_unlock_bh(&ping_table.lock);
187 static void inet_get_ping_group_range_net(struct net *net, gid_t *low,
190 gid_t *data = net->ipv4.sysctl_ping_group_range;
193 seq = read_seqbegin(&sysctl_local_ports.lock);
197 } while (read_seqretry(&sysctl_local_ports.lock, seq));
201 static int ping_init_sock(struct sock *sk)
203 struct net *net = sock_net(sk);
204 gid_t group = current_egid();
206 struct group_info *group_info = get_current_groups();
207 int i, j, count = group_info->ngroups;
209 inet_get_ping_group_range_net(net, range, range+1);
210 if (range[0] <= group && group <= range[1])
213 for (i = 0; i < group_info->nblocks; i++) {
214 int cp_count = min_t(int, NGROUPS_PER_BLOCK, count);
216 for (j = 0; j < cp_count; j++) {
217 group = group_info->blocks[i][j];
218 if (range[0] <= group && group <= range[1])
228 static void ping_close(struct sock *sk, long timeout)
230 pr_debug("ping_close(sk=%p,sk->num=%u)\n",
231 inet_sk(sk), inet_sk(sk)->inet_num);
232 pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
234 sk_common_release(sk);
238 * We need our own bind because there are no privileged id's == local ports.
239 * Moreover, we don't allow binding to multi- and broadcast addresses.
242 static int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
244 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
245 struct inet_sock *isk = inet_sk(sk);
250 if (addr_len < sizeof(struct sockaddr_in))
253 pr_debug("ping_v4_bind(sk=%p,sa_addr=%08x,sa_port=%d)\n",
254 sk, addr->sin_addr.s_addr, ntohs(addr->sin_port));
256 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
257 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
258 chk_addr_ret = RTN_LOCAL;
260 if ((sysctl_ip_nonlocal_bind == 0 &&
261 isk->freebind == 0 && isk->transparent == 0 &&
262 chk_addr_ret != RTN_LOCAL) ||
263 chk_addr_ret == RTN_MULTICAST ||
264 chk_addr_ret == RTN_BROADCAST)
265 return -EADDRNOTAVAIL;
270 if (isk->inet_num != 0)
274 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
275 snum = ntohs(addr->sin_port);
276 if (ping_v4_get_port(sk, snum) != 0) {
277 isk->inet_saddr = isk->inet_rcv_saddr = 0;
281 pr_debug("after bind(): num = %d, daddr = %pI4, dif = %d\n",
283 &isk->inet_rcv_saddr,
284 (int)sk->sk_bound_dev_if);
287 if (isk->inet_rcv_saddr)
288 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
290 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
291 isk->inet_sport = htons(isk->inet_num);
297 pr_debug("ping_v4_bind -> %d\n", err);
302 * Is this a supported type of ICMP message?
305 static inline int ping_supported(int type, int code)
307 if (type == ICMP_ECHO && code == 0)
313 * This routine is called by the ICMP module when it gets some
314 * sort of error condition.
317 static int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
319 void ping_err(struct sk_buff *skb, u32 info)
321 struct iphdr *iph = (struct iphdr *)skb->data;
322 struct icmphdr *icmph = (struct icmphdr *)(skb->data+(iph->ihl<<2));
323 struct inet_sock *inet_sock;
324 int type = icmph->type;
325 int code = icmph->code;
326 struct net *net = dev_net(skb->dev);
331 /* We assume the packet has already been checked by icmp_unreach */
333 if (!ping_supported(icmph->type, icmph->code))
336 pr_debug("ping_err(type=%04x,code=%04x,id=%04x,seq=%04x)\n", type,
337 code, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
339 sk = ping_v4_lookup(net, iph->daddr, iph->saddr,
340 ntohs(icmph->un.echo.id), skb->dev->ifindex);
342 pr_debug("no socket, dropping\n");
343 return; /* No socket for error */
345 pr_debug("err on socket %p\n", sk);
349 inet_sock = inet_sk(sk);
353 case ICMP_TIME_EXCEEDED:
356 case ICMP_SOURCE_QUENCH:
357 /* This is not a real error but ping wants to see it.
358 * Report it with some fake errno. */
361 case ICMP_PARAMETERPROB:
365 case ICMP_DEST_UNREACH:
366 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
367 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
375 if (code <= NR_ICMP_UNREACH) {
376 harderr = icmp_err_convert[code].fatal;
377 err = icmp_err_convert[code].errno;
381 /* See ICMP_SOURCE_QUENCH */
387 * RFC1122: OK. Passes ICMP errors back to application, as per
390 if (!inet_sock->recverr) {
391 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
394 ip_icmp_error(sk, skb, err, 0 /* no remote port */,
398 sk->sk_error_report(sk);
404 * Copy and checksum an ICMP Echo packet from user space into a buffer.
408 struct icmphdr icmph;
413 static int ping_getfrag(void *from, char * to,
414 int offset, int fraglen, int odd, struct sk_buff *skb)
416 struct pingfakehdr *pfh = (struct pingfakehdr *)from;
419 if (fraglen < sizeof(struct icmphdr))
421 if (csum_partial_copy_fromiovecend(to + sizeof(struct icmphdr),
422 pfh->iov, 0, fraglen - sizeof(struct icmphdr),
428 if (offset < sizeof(struct icmphdr))
430 if (csum_partial_copy_fromiovecend
431 (to, pfh->iov, offset - sizeof(struct icmphdr),
432 fraglen, &pfh->wcheck))
437 static int ping_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
440 struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
442 pfh->wcheck = csum_partial((char *)&pfh->icmph,
443 sizeof(struct icmphdr), pfh->wcheck);
444 pfh->icmph.checksum = csum_fold(pfh->wcheck);
445 memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
446 skb->ip_summed = CHECKSUM_NONE;
447 return ip_push_pending_frames(sk, fl4);
450 static int ping_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
453 struct net *net = sock_net(sk);
455 struct inet_sock *inet = inet_sk(sk);
456 struct ipcm_cookie ipc;
457 struct icmphdr user_icmph;
458 struct pingfakehdr pfh;
459 struct rtable *rt = NULL;
460 struct ip_options_data opt_copy;
462 __be32 saddr, daddr, faddr;
466 pr_debug("ping_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
476 /* Mirror BSD error message compatibility */
477 if (msg->msg_flags & MSG_OOB)
481 * Fetch the ICMP header provided by the userland.
485 if (memcpy_fromiovec((u8 *)&user_icmph, msg->msg_iov,
486 sizeof(struct icmphdr)))
488 if (!ping_supported(user_icmph.type, user_icmph.code))
492 * Get and verify the address.
496 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
497 if (msg->msg_namelen < sizeof(*usin))
499 if (usin->sin_family != AF_INET)
501 daddr = usin->sin_addr.s_addr;
504 if (sk->sk_state != TCP_ESTABLISHED)
505 return -EDESTADDRREQ;
506 daddr = inet->inet_daddr;
510 ipc.addr = inet->inet_saddr;
512 ipc.oif = sk->sk_bound_dev_if;
514 err = sock_tx_timestamp(sk, &ipc.tx_flags);
518 if (msg->msg_controllen) {
519 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
526 struct ip_options_rcu *inet_opt;
529 inet_opt = rcu_dereference(inet->inet_opt);
531 memcpy(&opt_copy, inet_opt,
532 sizeof(*inet_opt) + inet_opt->opt.optlen);
533 ipc.opt = &opt_copy.opt;
539 ipc.addr = faddr = daddr;
541 if (ipc.opt && ipc.opt->opt.srr) {
544 faddr = ipc.opt->opt.faddr;
546 tos = RT_TOS(inet->tos);
547 if (sock_flag(sk, SOCK_LOCALROUTE) ||
548 (msg->msg_flags & MSG_DONTROUTE) ||
549 (ipc.opt && ipc.opt->opt.is_strictroute)) {
553 if (ipv4_is_multicast(daddr)) {
555 ipc.oif = inet->mc_index;
557 saddr = inet->mc_addr;
559 ipc.oif = inet->uc_index;
561 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
562 RT_SCOPE_UNIVERSE, sk->sk_protocol,
563 inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
565 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
566 rt = ip_route_output_flow(net, &fl4, sk);
570 if (err == -ENETUNREACH)
571 IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
576 if ((rt->rt_flags & RTCF_BROADCAST) &&
577 !sock_flag(sk, SOCK_BROADCAST))
580 if (msg->msg_flags & MSG_CONFIRM)
585 ipc.addr = fl4.daddr;
589 pfh.icmph.type = user_icmph.type; /* already checked */
590 pfh.icmph.code = user_icmph.code; /* ditto */
591 pfh.icmph.checksum = 0;
592 pfh.icmph.un.echo.id = inet->inet_sport;
593 pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
594 pfh.iov = msg->msg_iov;
597 err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
598 0, &ipc, &rt, msg->msg_flags);
600 ip_flush_pending_frames(sk);
602 err = ping_push_pending_frames(sk, &pfh, &fl4);
610 icmp_out_count(sock_net(sk), user_icmph.type);
616 dst_confirm(&rt->dst);
617 if (!(msg->msg_flags & MSG_PROBE) || len)
618 goto back_from_confirm;
623 static int ping_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
624 size_t len, int noblock, int flags, int *addr_len)
626 struct inet_sock *isk = inet_sk(sk);
627 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
631 pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
638 *addr_len = sizeof(*sin);
640 if (flags & MSG_ERRQUEUE)
641 return ip_recv_error(sk, msg, len);
643 skb = skb_recv_datagram(sk, flags, noblock, &err);
649 msg->msg_flags |= MSG_TRUNC;
653 /* Don't bother checking the checksum */
654 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
658 sock_recv_timestamp(msg, sk, skb);
660 /* Copy the address. */
662 sin->sin_family = AF_INET;
663 sin->sin_port = 0 /* skb->h.uh->source */;
664 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
665 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
668 ip_cmsg_recv(msg, skb);
672 skb_free_datagram(sk, skb);
674 pr_debug("ping_recvmsg -> %d\n", err);
678 static int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
680 pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
681 inet_sk(sk), inet_sk(sk)->inet_num, skb);
682 if (sock_queue_rcv_skb(sk, skb) < 0) {
684 pr_debug("ping_queue_rcv_skb -> failed\n");
692 * All we need to do is get the socket.
695 void ping_rcv(struct sk_buff *skb)
698 struct net *net = dev_net(skb->dev);
699 struct iphdr *iph = ip_hdr(skb);
700 struct icmphdr *icmph = icmp_hdr(skb);
701 __be32 saddr = iph->saddr;
702 __be32 daddr = iph->daddr;
704 /* We assume the packet has already been checked by icmp_rcv */
706 pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
707 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
709 /* Push ICMP header back */
710 skb_push(skb, skb->data - (u8 *)icmph);
712 sk = ping_v4_lookup(net, saddr, daddr, ntohs(icmph->un.echo.id),
715 pr_debug("rcv on socket %p\n", sk);
716 ping_queue_rcv_skb(sk, skb_get(skb));
720 pr_debug("no socket, dropping\n");
722 /* We're called from icmp_rcv(). kfree_skb() is done there. */
725 struct proto ping_prot = {
727 .owner = THIS_MODULE,
728 .init = ping_init_sock,
730 .connect = ip4_datagram_connect,
731 .disconnect = udp_disconnect,
732 .setsockopt = ip_setsockopt,
733 .getsockopt = ip_getsockopt,
734 .sendmsg = ping_sendmsg,
735 .recvmsg = ping_recvmsg,
737 .backlog_rcv = ping_queue_rcv_skb,
738 .hash = ping_v4_hash,
739 .unhash = ping_v4_unhash,
740 .get_port = ping_v4_get_port,
741 .obj_size = sizeof(struct inet_sock),
743 EXPORT_SYMBOL(ping_prot);
745 #ifdef CONFIG_PROC_FS
747 static struct sock *ping_get_first(struct seq_file *seq, int start)
750 struct ping_iter_state *state = seq->private;
751 struct net *net = seq_file_net(seq);
753 for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
755 struct hlist_nulls_node *node;
756 struct hlist_nulls_head *hslot;
758 hslot = &ping_table.hash[state->bucket];
760 if (hlist_nulls_empty(hslot))
763 sk_nulls_for_each(sk, node, hslot) {
764 if (net_eq(sock_net(sk), net))
773 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
775 struct ping_iter_state *state = seq->private;
776 struct net *net = seq_file_net(seq);
779 sk = sk_nulls_next(sk);
780 } while (sk && (!net_eq(sock_net(sk), net)));
783 return ping_get_first(seq, state->bucket + 1);
787 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
789 struct sock *sk = ping_get_first(seq, 0);
792 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
794 return pos ? NULL : sk;
797 static void *ping_seq_start(struct seq_file *seq, loff_t *pos)
799 struct ping_iter_state *state = seq->private;
802 read_lock_bh(&ping_table.lock);
804 return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
807 static void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
811 if (v == SEQ_START_TOKEN)
812 sk = ping_get_idx(seq, 0);
814 sk = ping_get_next(seq, v);
820 static void ping_seq_stop(struct seq_file *seq, void *v)
822 read_unlock_bh(&ping_table.lock);
825 static void ping_format_sock(struct sock *sp, struct seq_file *f,
826 int bucket, int *len)
828 struct inet_sock *inet = inet_sk(sp);
829 __be32 dest = inet->inet_daddr;
830 __be32 src = inet->inet_rcv_saddr;
831 __u16 destp = ntohs(inet->inet_dport);
832 __u16 srcp = ntohs(inet->inet_sport);
834 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
835 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d%n",
836 bucket, src, srcp, dest, destp, sp->sk_state,
837 sk_wmem_alloc_get(sp),
838 sk_rmem_alloc_get(sp),
839 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
840 atomic_read(&sp->sk_refcnt), sp,
841 atomic_read(&sp->sk_drops), len);
844 static int ping_seq_show(struct seq_file *seq, void *v)
846 if (v == SEQ_START_TOKEN)
847 seq_printf(seq, "%-127s\n",
848 " sl local_address rem_address st tx_queue "
849 "rx_queue tr tm->when retrnsmt uid timeout "
850 "inode ref pointer drops");
852 struct ping_iter_state *state = seq->private;
855 ping_format_sock(v, seq, state->bucket, &len);
856 seq_printf(seq, "%*s\n", 127 - len, "");
861 static const struct seq_operations ping_seq_ops = {
862 .show = ping_seq_show,
863 .start = ping_seq_start,
864 .next = ping_seq_next,
865 .stop = ping_seq_stop,
868 static int ping_seq_open(struct inode *inode, struct file *file)
870 return seq_open_net(inode, file, &ping_seq_ops,
871 sizeof(struct ping_iter_state));
874 static const struct file_operations ping_seq_fops = {
875 .open = ping_seq_open,
878 .release = seq_release_net,
881 static int ping_proc_register(struct net *net)
883 struct proc_dir_entry *p;
886 p = proc_net_fops_create(net, "icmp", S_IRUGO, &ping_seq_fops);
892 static void ping_proc_unregister(struct net *net)
894 proc_net_remove(net, "icmp");
898 static int __net_init ping_proc_init_net(struct net *net)
900 return ping_proc_register(net);
903 static void __net_exit ping_proc_exit_net(struct net *net)
905 ping_proc_unregister(net);
908 static struct pernet_operations ping_net_ops = {
909 .init = ping_proc_init_net,
910 .exit = ping_proc_exit_net,
913 int __init ping_proc_init(void)
915 return register_pernet_subsys(&ping_net_ops);
918 void ping_proc_exit(void)
920 unregister_pernet_subsys(&ping_net_ops);
925 void __init ping_init(void)
929 for (i = 0; i < PING_HTABLE_SIZE; i++)
930 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
931 rwlock_init(&ping_table.lock);