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.
6 * RAW - implementation of IP "raw" sockets.
8 * Version: $Id: raw.c,v 1.64 2002/02/01 22:01:04 davem Exp $
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
14 * Alan Cox : verify_area() fixed up
15 * Alan Cox : ICMP error handling
16 * Alan Cox : EMSGSIZE if you send too big a packet
17 * Alan Cox : Now uses generic datagrams and shared
18 * skbuff library. No more peek crashes,
20 * Alan Cox : Checks sk->broadcast.
21 * Alan Cox : Uses skb_free_datagram/skb_copy_datagram
22 * Alan Cox : Raw passes ip options too
23 * Alan Cox : Setsocketopt added
24 * Alan Cox : Fixed error return for broadcasts
25 * Alan Cox : Removed wake_up calls
26 * Alan Cox : Use ttl/tos
27 * Alan Cox : Cleaned up old debugging
28 * Alan Cox : Use new kernel side addresses
29 * Arnt Gulbrandsen : Fixed MSG_DONTROUTE in raw sockets.
30 * Alan Cox : BSD style RAW socket demultiplexing.
31 * Alan Cox : Beginnings of mrouted support.
32 * Alan Cox : Added IP_HDRINCL option.
33 * Alan Cox : Skip broadcast check if BSDism set.
34 * David S. Miller : New socket lookup architecture.
36 * This program is free software; you can redistribute it and/or
37 * modify it under the terms of the GNU General Public License
38 * as published by the Free Software Foundation; either version
39 * 2 of the License, or (at your option) any later version.
42 #include <linux/types.h>
43 #include <asm/atomic.h>
44 #include <asm/byteorder.h>
45 #include <asm/current.h>
46 #include <asm/uaccess.h>
47 #include <asm/ioctls.h>
48 #include <linux/stddef.h>
49 #include <linux/slab.h>
50 #include <linux/errno.h>
51 #include <linux/aio.h>
52 #include <linux/kernel.h>
53 #include <linux/spinlock.h>
54 #include <linux/sockios.h>
55 #include <linux/socket.h>
57 #include <linux/mroute.h>
58 #include <linux/netdevice.h>
59 #include <linux/in_route.h>
60 #include <linux/route.h>
61 #include <linux/skbuff.h>
62 #include <net/net_namespace.h>
65 #include <linux/gfp.h>
67 #include <linux/net.h>
73 #include <net/tcp_states.h>
74 #include <net/inet_common.h>
75 #include <net/checksum.h>
77 #include <linux/rtnetlink.h>
78 #include <linux/proc_fs.h>
79 #include <linux/seq_file.h>
80 #include <linux/netfilter.h>
81 #include <linux/netfilter_ipv4.h>
83 static struct raw_hashinfo raw_v4_hashinfo = {
84 .lock = __RW_LOCK_UNLOCKED(),
87 void raw_hash_sk(struct sock *sk, struct raw_hashinfo *h)
89 struct hlist_head *head;
91 head = &h->ht[inet_sk(sk)->num & (RAW_HTABLE_SIZE - 1)];
93 write_lock_bh(&h->lock);
94 sk_add_node(sk, head);
95 sock_prot_inc_use(sk->sk_prot);
96 write_unlock_bh(&h->lock);
98 EXPORT_SYMBOL_GPL(raw_hash_sk);
100 void raw_unhash_sk(struct sock *sk, struct raw_hashinfo *h)
102 write_lock_bh(&h->lock);
103 if (sk_del_node_init(sk))
104 sock_prot_dec_use(sk->sk_prot);
105 write_unlock_bh(&h->lock);
107 EXPORT_SYMBOL_GPL(raw_unhash_sk);
109 static void raw_v4_hash(struct sock *sk)
111 raw_hash_sk(sk, &raw_v4_hashinfo);
114 static void raw_v4_unhash(struct sock *sk)
116 raw_unhash_sk(sk, &raw_v4_hashinfo);
119 static struct sock *__raw_v4_lookup(struct sock *sk, unsigned short num,
120 __be32 raddr, __be32 laddr,
123 struct hlist_node *node;
125 sk_for_each_from(sk, node) {
126 struct inet_sock *inet = inet_sk(sk);
128 if (inet->num == num &&
129 !(inet->daddr && inet->daddr != raddr) &&
130 !(inet->rcv_saddr && inet->rcv_saddr != laddr) &&
131 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
132 goto found; /* gotcha */
143 static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
147 if (!pskb_may_pull(skb, sizeof(struct icmphdr)))
150 type = icmp_hdr(skb)->type;
152 __u32 data = raw_sk(sk)->filter.data;
154 return ((1 << type) & data) != 0;
157 /* Do not block unknown ICMP types */
161 /* IP input processing comes here for RAW socket delivery.
162 * Caller owns SKB, so we must make clones.
164 * RFC 1122: SHOULD pass TOS value up to the transport layer.
165 * -> It does. And not only TOS, but all IP header.
167 static int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
170 struct hlist_head *head;
173 read_lock(&raw_v4_hashinfo.lock);
174 head = &raw_v4_hashinfo.ht[hash];
175 if (hlist_empty(head))
177 sk = __raw_v4_lookup(__sk_head(head), iph->protocol,
178 iph->saddr, iph->daddr,
183 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
184 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
186 /* Not releasing hash table! */
190 sk = __raw_v4_lookup(sk_next(sk), iph->protocol,
191 iph->saddr, iph->daddr,
195 read_unlock(&raw_v4_hashinfo.lock);
199 int raw_local_deliver(struct sk_buff *skb, int protocol)
204 hash = protocol & (RAW_HTABLE_SIZE - 1);
205 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
207 /* If there maybe a raw socket we must check - if not we
210 if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
213 return raw_sk != NULL;
217 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
219 struct inet_sock *inet = inet_sk(sk);
220 const int type = icmp_hdr(skb)->type;
221 const int code = icmp_hdr(skb)->code;
225 /* Report error on raw socket, if:
226 1. User requested ip_recverr.
227 2. Socket is connected (otherwise the error indication
228 is useless without ip_recverr and error is hard.
230 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
235 case ICMP_TIME_EXCEEDED:
238 case ICMP_SOURCE_QUENCH:
240 case ICMP_PARAMETERPROB:
244 case ICMP_DEST_UNREACH:
246 if (code > NR_ICMP_UNREACH)
248 err = icmp_err_convert[code].errno;
249 harderr = icmp_err_convert[code].fatal;
250 if (code == ICMP_FRAG_NEEDED) {
251 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
257 struct iphdr *iph = (struct iphdr*)skb->data;
258 u8 *payload = skb->data + (iph->ihl << 2);
262 ip_icmp_error(sk, skb, err, 0, info, payload);
265 if (inet->recverr || harderr) {
267 sk->sk_error_report(sk);
271 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
277 hash = protocol & (RAW_HTABLE_SIZE - 1);
279 read_lock(&raw_v4_hashinfo.lock);
280 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
281 if (raw_sk != NULL) {
282 iph = (struct iphdr *)skb->data;
283 while ((raw_sk = __raw_v4_lookup(raw_sk, protocol, iph->daddr,
285 skb->dev->ifindex)) != NULL) {
286 raw_err(raw_sk, skb, info);
287 raw_sk = sk_next(raw_sk);
288 iph = (struct iphdr *)skb->data;
291 read_unlock(&raw_v4_hashinfo.lock);
294 static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
296 /* Charge it to the socket. */
298 if (sock_queue_rcv_skb(sk, skb) < 0) {
299 atomic_inc(&sk->sk_drops);
304 return NET_RX_SUCCESS;
307 int raw_rcv(struct sock *sk, struct sk_buff *skb)
309 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
310 atomic_inc(&sk->sk_drops);
316 skb_push(skb, skb->data - skb_network_header(skb));
318 raw_rcv_skb(sk, skb);
322 static int raw_send_hdrinc(struct sock *sk, void *from, size_t length,
326 struct inet_sock *inet = inet_sk(sk);
333 if (length > rt->u.dst.dev->mtu) {
334 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
341 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
343 skb = sock_alloc_send_skb(sk, length+hh_len+15,
344 flags&MSG_DONTWAIT, &err);
347 skb_reserve(skb, hh_len);
349 skb->priority = sk->sk_priority;
350 skb->dst = dst_clone(&rt->u.dst);
352 skb_reset_network_header(skb);
354 skb_put(skb, length);
356 skb->ip_summed = CHECKSUM_NONE;
358 skb->transport_header = skb->network_header;
359 err = memcpy_fromiovecend((void *)iph, from, 0, length);
363 /* We don't modify invalid header */
364 iphlen = iph->ihl * 4;
365 if (iphlen >= sizeof(*iph) && iphlen <= length) {
367 iph->saddr = rt->rt_src;
369 iph->tot_len = htons(length);
371 ip_select_ident(iph, &rt->u.dst, NULL);
373 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
375 if (iph->protocol == IPPROTO_ICMP)
376 icmp_out_count(((struct icmphdr *)
377 skb_transport_header(skb))->type);
379 err = NF_HOOK(PF_INET, NF_INET_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
382 err = inet->recverr ? net_xmit_errno(err) : 0;
392 IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
396 static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
399 u8 __user *type = NULL;
400 u8 __user *code = NULL;
407 for (i = 0; i < msg->msg_iovlen; i++) {
408 iov = &msg->msg_iov[i];
414 /* check if one-byte field is readable or not. */
415 if (iov->iov_base && iov->iov_len < 1)
419 type = iov->iov_base;
420 /* check if code field is readable or not. */
421 if (iov->iov_len > 1)
424 code = iov->iov_base;
427 if (get_user(fl->fl_icmp_type, type) ||
428 get_user(fl->fl_icmp_code, code))
443 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
446 struct inet_sock *inet = inet_sk(sk);
447 struct ipcm_cookie ipc;
448 struct rtable *rt = NULL;
464 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
465 goto out; /* compatibility */
468 * Get and verify the address.
471 if (msg->msg_namelen) {
472 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
474 if (msg->msg_namelen < sizeof(*usin))
476 if (usin->sin_family != AF_INET) {
477 static int complained;
479 printk(KERN_INFO "%s forgot to set AF_INET in "
480 "raw sendmsg. Fix it!\n",
483 if (usin->sin_family)
486 daddr = usin->sin_addr.s_addr;
487 /* ANK: I did not forget to get protocol from port field.
488 * I just do not know, who uses this weirdness.
489 * IP_HDRINCL is much more convenient.
493 if (sk->sk_state != TCP_ESTABLISHED)
498 ipc.addr = inet->saddr;
500 ipc.oif = sk->sk_bound_dev_if;
502 if (msg->msg_controllen) {
503 err = ip_cmsg_send(msg, &ipc);
518 /* Linux does not mangle headers on raw sockets,
519 * so that IP options + IP_HDRINCL is non-sense.
526 daddr = ipc.opt->faddr;
529 tos = RT_CONN_FLAGS(sk);
530 if (msg->msg_flags & MSG_DONTROUTE)
533 if (MULTICAST(daddr)) {
535 ipc.oif = inet->mc_index;
537 saddr = inet->mc_addr;
541 struct flowi fl = { .oif = ipc.oif,
546 .proto = inet->hdrincl ? IPPROTO_RAW :
549 if (!inet->hdrincl) {
550 err = raw_probe_proto_opt(&fl, msg);
555 security_sk_classify_flow(sk, &fl);
556 err = ip_route_output_flow(&rt, &fl, sk, 1);
562 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
565 if (msg->msg_flags & MSG_CONFIRM)
570 err = raw_send_hdrinc(sk, msg->msg_iov, len,
575 ipc.addr = rt->rt_dst;
577 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
578 &ipc, rt, msg->msg_flags);
580 ip_flush_pending_frames(sk);
581 else if (!(msg->msg_flags & MSG_MORE))
582 err = ip_push_pending_frames(sk);
596 dst_confirm(&rt->u.dst);
597 if (!(msg->msg_flags & MSG_PROBE) || len)
598 goto back_from_confirm;
603 static void raw_close(struct sock *sk, long timeout)
606 * Raw sockets may have direct kernel refereneces. Kill them.
608 ip_ra_control(sk, 0, NULL);
610 sk_common_release(sk);
613 /* This gets rid of all the nasties in af_inet. -DaveM */
614 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
616 struct inet_sock *inet = inet_sk(sk);
617 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
621 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
623 chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
624 ret = -EADDRNOTAVAIL;
625 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
626 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
628 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
629 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
630 inet->saddr = 0; /* Use device */
637 * This should be easy, if there is something there
638 * we return it, otherwise we block.
641 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
642 size_t len, int noblock, int flags, int *addr_len)
644 struct inet_sock *inet = inet_sk(sk);
646 int err = -EOPNOTSUPP;
647 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
654 *addr_len = sizeof(*sin);
656 if (flags & MSG_ERRQUEUE) {
657 err = ip_recv_error(sk, msg, len);
661 skb = skb_recv_datagram(sk, flags, noblock, &err);
667 msg->msg_flags |= MSG_TRUNC;
671 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
675 sock_recv_timestamp(msg, sk, skb);
677 /* Copy the address. */
679 sin->sin_family = AF_INET;
680 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
682 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
684 if (inet->cmsg_flags)
685 ip_cmsg_recv(msg, skb);
686 if (flags & MSG_TRUNC)
689 skb_free_datagram(sk, skb);
696 static int raw_init(struct sock *sk)
698 struct raw_sock *rp = raw_sk(sk);
700 if (inet_sk(sk)->num == IPPROTO_ICMP)
701 memset(&rp->filter, 0, sizeof(rp->filter));
705 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
707 if (optlen > sizeof(struct icmp_filter))
708 optlen = sizeof(struct icmp_filter);
709 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
714 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
716 int len, ret = -EFAULT;
718 if (get_user(len, optlen))
723 if (len > sizeof(struct icmp_filter))
724 len = sizeof(struct icmp_filter);
726 if (put_user(len, optlen) ||
727 copy_to_user(optval, &raw_sk(sk)->filter, len))
733 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
734 char __user *optval, int optlen)
736 if (optname == ICMP_FILTER) {
737 if (inet_sk(sk)->num != IPPROTO_ICMP)
740 return raw_seticmpfilter(sk, optval, optlen);
745 static int raw_setsockopt(struct sock *sk, int level, int optname,
746 char __user *optval, int optlen)
748 if (level != SOL_RAW)
749 return ip_setsockopt(sk, level, optname, optval, optlen);
750 return do_raw_setsockopt(sk, level, optname, optval, optlen);
754 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
755 char __user *optval, int optlen)
757 if (level != SOL_RAW)
758 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
759 return do_raw_setsockopt(sk, level, optname, optval, optlen);
763 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
764 char __user *optval, int __user *optlen)
766 if (optname == ICMP_FILTER) {
767 if (inet_sk(sk)->num != IPPROTO_ICMP)
770 return raw_geticmpfilter(sk, optval, optlen);
775 static int raw_getsockopt(struct sock *sk, int level, int optname,
776 char __user *optval, int __user *optlen)
778 if (level != SOL_RAW)
779 return ip_getsockopt(sk, level, optname, optval, optlen);
780 return do_raw_getsockopt(sk, level, optname, optval, optlen);
784 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
785 char __user *optval, int __user *optlen)
787 if (level != SOL_RAW)
788 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
789 return do_raw_getsockopt(sk, level, optname, optval, optlen);
793 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
797 int amount = atomic_read(&sk->sk_wmem_alloc);
798 return put_user(amount, (int __user *)arg);
804 spin_lock_bh(&sk->sk_receive_queue.lock);
805 skb = skb_peek(&sk->sk_receive_queue);
808 spin_unlock_bh(&sk->sk_receive_queue.lock);
809 return put_user(amount, (int __user *)arg);
813 #ifdef CONFIG_IP_MROUTE
814 return ipmr_ioctl(sk, cmd, (void __user *)arg);
821 DEFINE_PROTO_INUSE(raw)
823 struct proto raw_prot = {
825 .owner = THIS_MODULE,
827 .connect = ip4_datagram_connect,
828 .disconnect = udp_disconnect,
831 .setsockopt = raw_setsockopt,
832 .getsockopt = raw_getsockopt,
833 .sendmsg = raw_sendmsg,
834 .recvmsg = raw_recvmsg,
836 .backlog_rcv = raw_rcv_skb,
838 .unhash = raw_v4_unhash,
839 .obj_size = sizeof(struct raw_sock),
841 .compat_setsockopt = compat_raw_setsockopt,
842 .compat_getsockopt = compat_raw_getsockopt,
847 #ifdef CONFIG_PROC_FS
848 static struct sock *raw_get_first(struct seq_file *seq)
851 struct raw_iter_state* state = raw_seq_private(seq);
853 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
855 struct hlist_node *node;
857 sk_for_each(sk, node, &state->h->ht[state->bucket])
858 if (sk->sk_family == state->family)
866 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
868 struct raw_iter_state* state = raw_seq_private(seq);
874 } while (sk && sk->sk_family != state->family);
876 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
877 sk = sk_head(&state->h->ht[state->bucket]);
883 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
885 struct sock *sk = raw_get_first(seq);
888 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
890 return pos ? NULL : sk;
893 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
895 struct raw_iter_state *state = raw_seq_private(seq);
897 read_lock(&state->h->lock);
898 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
900 EXPORT_SYMBOL_GPL(raw_seq_start);
902 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
906 if (v == SEQ_START_TOKEN)
907 sk = raw_get_first(seq);
909 sk = raw_get_next(seq, v);
913 EXPORT_SYMBOL_GPL(raw_seq_next);
915 void raw_seq_stop(struct seq_file *seq, void *v)
917 struct raw_iter_state *state = raw_seq_private(seq);
919 read_unlock(&state->h->lock);
921 EXPORT_SYMBOL_GPL(raw_seq_stop);
923 static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i)
925 struct inet_sock *inet = inet_sk(sp);
926 __be32 dest = inet->daddr,
927 src = inet->rcv_saddr;
931 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
932 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d",
933 i, src, srcp, dest, destp, sp->sk_state,
934 atomic_read(&sp->sk_wmem_alloc),
935 atomic_read(&sp->sk_rmem_alloc),
936 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
937 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
943 static int raw_seq_show(struct seq_file *seq, void *v)
945 char tmpbuf[TMPSZ+1];
947 if (v == SEQ_START_TOKEN)
948 seq_printf(seq, "%-*s\n", TMPSZ-1,
949 " sl local_address rem_address st tx_queue "
950 "rx_queue tr tm->when retrnsmt uid timeout "
953 struct raw_iter_state *state = raw_seq_private(seq);
955 seq_printf(seq, "%-*s\n", TMPSZ-1,
956 get_raw_sock(v, tmpbuf, state->bucket));
961 static const struct seq_operations raw_seq_ops = {
962 .start = raw_seq_start,
963 .next = raw_seq_next,
964 .stop = raw_seq_stop,
965 .show = raw_seq_show,
968 int raw_seq_open(struct file *file, struct raw_hashinfo *h,
969 unsigned short family)
971 struct raw_iter_state *i;
973 i = __seq_open_private(file, &raw_seq_ops,
974 sizeof(struct raw_iter_state));
982 EXPORT_SYMBOL_GPL(raw_seq_open);
984 static int raw_v4_seq_open(struct inode *inode, struct file *file)
986 return raw_seq_open(file, &raw_v4_hashinfo, PF_INET);
989 static const struct file_operations raw_seq_fops = {
990 .owner = THIS_MODULE,
991 .open = raw_v4_seq_open,
994 .release = seq_release_private,
997 int __init raw_proc_init(void)
999 if (!proc_net_fops_create(&init_net, "raw", S_IRUGO, &raw_seq_fops))
1004 void __init raw_proc_exit(void)
1006 proc_net_remove(&init_net, "raw");
1008 #endif /* CONFIG_PROC_FS */