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 struct hlist_head raw_v4_htable[RAWV4_HTABLE_SIZE];
84 DEFINE_RWLOCK(raw_v4_lock);
86 static void raw_v4_hash(struct sock *sk)
88 struct hlist_head *head = &raw_v4_htable[inet_sk(sk)->num &
89 (RAWV4_HTABLE_SIZE - 1)];
91 write_lock_bh(&raw_v4_lock);
92 sk_add_node(sk, head);
93 sock_prot_inc_use(sk->sk_prot);
94 write_unlock_bh(&raw_v4_lock);
97 static void raw_v4_unhash(struct sock *sk)
99 write_lock_bh(&raw_v4_lock);
100 if (sk_del_node_init(sk))
101 sock_prot_dec_use(sk->sk_prot);
102 write_unlock_bh(&raw_v4_lock);
105 struct sock *__raw_v4_lookup(struct sock *sk, unsigned short num,
106 __be32 raddr, __be32 laddr,
109 struct hlist_node *node;
111 sk_for_each_from(sk, node) {
112 struct inet_sock *inet = inet_sk(sk);
114 if (inet->num == num &&
115 !(inet->daddr && inet->daddr != raddr) &&
116 !(inet->rcv_saddr && inet->rcv_saddr != laddr) &&
117 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
118 goto found; /* gotcha */
129 static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
133 if (!pskb_may_pull(skb, sizeof(struct icmphdr)))
136 type = icmp_hdr(skb)->type;
138 __u32 data = raw_sk(sk)->filter.data;
140 return ((1 << type) & data) != 0;
143 /* Do not block unknown ICMP types */
147 /* IP input processing comes here for RAW socket delivery.
148 * Caller owns SKB, so we must make clones.
150 * RFC 1122: SHOULD pass TOS value up to the transport layer.
151 * -> It does. And not only TOS, but all IP header.
153 int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
156 struct hlist_head *head;
159 read_lock(&raw_v4_lock);
160 head = &raw_v4_htable[hash];
161 if (hlist_empty(head))
163 sk = __raw_v4_lookup(__sk_head(head), iph->protocol,
164 iph->saddr, iph->daddr,
169 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
170 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
172 /* Not releasing hash table! */
176 sk = __raw_v4_lookup(sk_next(sk), iph->protocol,
177 iph->saddr, iph->daddr,
181 read_unlock(&raw_v4_lock);
185 void raw_err (struct sock *sk, struct sk_buff *skb, u32 info)
187 struct inet_sock *inet = inet_sk(sk);
188 const int type = icmp_hdr(skb)->type;
189 const int code = icmp_hdr(skb)->code;
193 /* Report error on raw socket, if:
194 1. User requested ip_recverr.
195 2. Socket is connected (otherwise the error indication
196 is useless without ip_recverr and error is hard.
198 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
203 case ICMP_TIME_EXCEEDED:
206 case ICMP_SOURCE_QUENCH:
208 case ICMP_PARAMETERPROB:
212 case ICMP_DEST_UNREACH:
214 if (code > NR_ICMP_UNREACH)
216 err = icmp_err_convert[code].errno;
217 harderr = icmp_err_convert[code].fatal;
218 if (code == ICMP_FRAG_NEEDED) {
219 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
225 struct iphdr *iph = (struct iphdr*)skb->data;
226 u8 *payload = skb->data + (iph->ihl << 2);
230 ip_icmp_error(sk, skb, err, 0, info, payload);
233 if (inet->recverr || harderr) {
235 sk->sk_error_report(sk);
239 static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
241 /* Charge it to the socket. */
243 if (sock_queue_rcv_skb(sk, skb) < 0) {
244 /* FIXME: increment a raw drops counter here */
249 return NET_RX_SUCCESS;
252 int raw_rcv(struct sock *sk, struct sk_buff *skb)
254 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
260 skb_push(skb, skb->data - skb_network_header(skb));
262 raw_rcv_skb(sk, skb);
266 static int raw_send_hdrinc(struct sock *sk, void *from, size_t length,
270 struct inet_sock *inet = inet_sk(sk);
276 if (length > rt->u.dst.dev->mtu) {
277 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
284 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
286 skb = sock_alloc_send_skb(sk, length+hh_len+15,
287 flags&MSG_DONTWAIT, &err);
290 skb_reserve(skb, hh_len);
292 skb->priority = sk->sk_priority;
293 skb->dst = dst_clone(&rt->u.dst);
295 skb_reset_network_header(skb);
297 skb_put(skb, length);
299 skb->ip_summed = CHECKSUM_NONE;
301 skb->transport_header = skb->network_header;
302 err = memcpy_fromiovecend((void *)iph, from, 0, length);
306 /* We don't modify invalid header */
307 if (length >= sizeof(*iph) && iph->ihl * 4U <= length) {
309 iph->saddr = rt->rt_src;
311 iph->tot_len = htons(length);
313 ip_select_ident(iph, &rt->u.dst, NULL);
315 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
317 if (iph->protocol == IPPROTO_ICMP)
318 icmp_out_count(((struct icmphdr *)
319 skb_transport_header(skb))->type);
321 err = NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
324 err = inet->recverr ? net_xmit_errno(err) : 0;
334 IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
338 static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
341 u8 __user *type = NULL;
342 u8 __user *code = NULL;
349 for (i = 0; i < msg->msg_iovlen; i++) {
350 iov = &msg->msg_iov[i];
356 /* check if one-byte field is readable or not. */
357 if (iov->iov_base && iov->iov_len < 1)
361 type = iov->iov_base;
362 /* check if code field is readable or not. */
363 if (iov->iov_len > 1)
366 code = iov->iov_base;
369 if (get_user(fl->fl_icmp_type, type) ||
370 get_user(fl->fl_icmp_code, code))
385 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
388 struct inet_sock *inet = inet_sk(sk);
389 struct ipcm_cookie ipc;
390 struct rtable *rt = NULL;
406 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
407 goto out; /* compatibility */
410 * Get and verify the address.
413 if (msg->msg_namelen) {
414 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
416 if (msg->msg_namelen < sizeof(*usin))
418 if (usin->sin_family != AF_INET) {
419 static int complained;
421 printk(KERN_INFO "%s forgot to set AF_INET in "
422 "raw sendmsg. Fix it!\n",
425 if (usin->sin_family)
428 daddr = usin->sin_addr.s_addr;
429 /* ANK: I did not forget to get protocol from port field.
430 * I just do not know, who uses this weirdness.
431 * IP_HDRINCL is much more convenient.
435 if (sk->sk_state != TCP_ESTABLISHED)
440 ipc.addr = inet->saddr;
442 ipc.oif = sk->sk_bound_dev_if;
444 if (msg->msg_controllen) {
445 err = ip_cmsg_send(msg, &ipc);
460 /* Linux does not mangle headers on raw sockets,
461 * so that IP options + IP_HDRINCL is non-sense.
468 daddr = ipc.opt->faddr;
471 tos = RT_CONN_FLAGS(sk);
472 if (msg->msg_flags & MSG_DONTROUTE)
475 if (MULTICAST(daddr)) {
477 ipc.oif = inet->mc_index;
479 saddr = inet->mc_addr;
483 struct flowi fl = { .oif = ipc.oif,
488 .proto = inet->hdrincl ? IPPROTO_RAW :
491 if (!inet->hdrincl) {
492 err = raw_probe_proto_opt(&fl, msg);
497 security_sk_classify_flow(sk, &fl);
498 err = ip_route_output_flow(&rt, &fl, sk, 1);
504 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
507 if (msg->msg_flags & MSG_CONFIRM)
512 err = raw_send_hdrinc(sk, msg->msg_iov, len,
517 ipc.addr = rt->rt_dst;
519 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
520 &ipc, rt, msg->msg_flags);
522 ip_flush_pending_frames(sk);
523 else if (!(msg->msg_flags & MSG_MORE))
524 err = ip_push_pending_frames(sk);
538 dst_confirm(&rt->u.dst);
539 if (!(msg->msg_flags & MSG_PROBE) || len)
540 goto back_from_confirm;
545 static void raw_close(struct sock *sk, long timeout)
548 * Raw sockets may have direct kernel refereneces. Kill them.
550 ip_ra_control(sk, 0, NULL);
552 sk_common_release(sk);
555 /* This gets rid of all the nasties in af_inet. -DaveM */
556 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
558 struct inet_sock *inet = inet_sk(sk);
559 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
563 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
565 chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
566 ret = -EADDRNOTAVAIL;
567 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
568 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
570 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
571 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
572 inet->saddr = 0; /* Use device */
579 * This should be easy, if there is something there
580 * we return it, otherwise we block.
583 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
584 size_t len, int noblock, int flags, int *addr_len)
586 struct inet_sock *inet = inet_sk(sk);
588 int err = -EOPNOTSUPP;
589 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
596 *addr_len = sizeof(*sin);
598 if (flags & MSG_ERRQUEUE) {
599 err = ip_recv_error(sk, msg, len);
603 skb = skb_recv_datagram(sk, flags, noblock, &err);
609 msg->msg_flags |= MSG_TRUNC;
613 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
617 sock_recv_timestamp(msg, sk, skb);
619 /* Copy the address. */
621 sin->sin_family = AF_INET;
622 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
624 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
626 if (inet->cmsg_flags)
627 ip_cmsg_recv(msg, skb);
628 if (flags & MSG_TRUNC)
631 skb_free_datagram(sk, skb);
638 static int raw_init(struct sock *sk)
640 struct raw_sock *rp = raw_sk(sk);
642 if (inet_sk(sk)->num == IPPROTO_ICMP)
643 memset(&rp->filter, 0, sizeof(rp->filter));
647 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
649 if (optlen > sizeof(struct icmp_filter))
650 optlen = sizeof(struct icmp_filter);
651 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
656 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
658 int len, ret = -EFAULT;
660 if (get_user(len, optlen))
665 if (len > sizeof(struct icmp_filter))
666 len = sizeof(struct icmp_filter);
668 if (put_user(len, optlen) ||
669 copy_to_user(optval, &raw_sk(sk)->filter, len))
675 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
676 char __user *optval, int optlen)
678 if (optname == ICMP_FILTER) {
679 if (inet_sk(sk)->num != IPPROTO_ICMP)
682 return raw_seticmpfilter(sk, optval, optlen);
687 static int raw_setsockopt(struct sock *sk, int level, int optname,
688 char __user *optval, int optlen)
690 if (level != SOL_RAW)
691 return ip_setsockopt(sk, level, optname, optval, optlen);
692 return do_raw_setsockopt(sk, level, optname, optval, optlen);
696 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
697 char __user *optval, int optlen)
699 if (level != SOL_RAW)
700 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
701 return do_raw_setsockopt(sk, level, optname, optval, optlen);
705 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
706 char __user *optval, int __user *optlen)
708 if (optname == ICMP_FILTER) {
709 if (inet_sk(sk)->num != IPPROTO_ICMP)
712 return raw_geticmpfilter(sk, optval, optlen);
717 static int raw_getsockopt(struct sock *sk, int level, int optname,
718 char __user *optval, int __user *optlen)
720 if (level != SOL_RAW)
721 return ip_getsockopt(sk, level, optname, optval, optlen);
722 return do_raw_getsockopt(sk, level, optname, optval, optlen);
726 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
727 char __user *optval, int __user *optlen)
729 if (level != SOL_RAW)
730 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
731 return do_raw_getsockopt(sk, level, optname, optval, optlen);
735 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
739 int amount = atomic_read(&sk->sk_wmem_alloc);
740 return put_user(amount, (int __user *)arg);
746 spin_lock_bh(&sk->sk_receive_queue.lock);
747 skb = skb_peek(&sk->sk_receive_queue);
750 spin_unlock_bh(&sk->sk_receive_queue.lock);
751 return put_user(amount, (int __user *)arg);
755 #ifdef CONFIG_IP_MROUTE
756 return ipmr_ioctl(sk, cmd, (void __user *)arg);
763 struct proto raw_prot = {
765 .owner = THIS_MODULE,
767 .connect = ip4_datagram_connect,
768 .disconnect = udp_disconnect,
771 .setsockopt = raw_setsockopt,
772 .getsockopt = raw_getsockopt,
773 .sendmsg = raw_sendmsg,
774 .recvmsg = raw_recvmsg,
776 .backlog_rcv = raw_rcv_skb,
778 .unhash = raw_v4_unhash,
779 .obj_size = sizeof(struct raw_sock),
781 .compat_setsockopt = compat_raw_setsockopt,
782 .compat_getsockopt = compat_raw_getsockopt,
786 #ifdef CONFIG_PROC_FS
787 struct raw_iter_state {
791 #define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
793 static struct sock *raw_get_first(struct seq_file *seq)
796 struct raw_iter_state* state = raw_seq_private(seq);
798 for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) {
799 struct hlist_node *node;
801 sk_for_each(sk, node, &raw_v4_htable[state->bucket])
802 if (sk->sk_family == PF_INET)
810 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
812 struct raw_iter_state* state = raw_seq_private(seq);
818 } while (sk && sk->sk_family != PF_INET);
820 if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) {
821 sk = sk_head(&raw_v4_htable[state->bucket]);
827 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
829 struct sock *sk = raw_get_first(seq);
832 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
834 return pos ? NULL : sk;
837 static void *raw_seq_start(struct seq_file *seq, loff_t *pos)
839 read_lock(&raw_v4_lock);
840 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
843 static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
847 if (v == SEQ_START_TOKEN)
848 sk = raw_get_first(seq);
850 sk = raw_get_next(seq, v);
855 static void raw_seq_stop(struct seq_file *seq, void *v)
857 read_unlock(&raw_v4_lock);
860 static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i)
862 struct inet_sock *inet = inet_sk(sp);
863 __be32 dest = inet->daddr,
864 src = inet->rcv_saddr;
868 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
869 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
870 i, src, srcp, dest, destp, sp->sk_state,
871 atomic_read(&sp->sk_wmem_alloc),
872 atomic_read(&sp->sk_rmem_alloc),
873 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
874 atomic_read(&sp->sk_refcnt), sp);
878 static int raw_seq_show(struct seq_file *seq, void *v)
882 if (v == SEQ_START_TOKEN)
883 seq_printf(seq, "%-127s\n",
884 " sl local_address rem_address st tx_queue "
885 "rx_queue tr tm->when retrnsmt uid timeout "
888 struct raw_iter_state *state = raw_seq_private(seq);
890 seq_printf(seq, "%-127s\n",
891 get_raw_sock(v, tmpbuf, state->bucket));
896 static const struct seq_operations raw_seq_ops = {
897 .start = raw_seq_start,
898 .next = raw_seq_next,
899 .stop = raw_seq_stop,
900 .show = raw_seq_show,
903 static int raw_seq_open(struct inode *inode, struct file *file)
905 return seq_open_private(file, &raw_seq_ops,
906 sizeof(struct raw_iter_state));
909 static const struct file_operations raw_seq_fops = {
910 .owner = THIS_MODULE,
911 .open = raw_seq_open,
914 .release = seq_release_private,
917 int __init raw_proc_init(void)
919 if (!proc_net_fops_create(&init_net, "raw", S_IRUGO, &raw_seq_fops))
924 void __init raw_proc_exit(void)
926 proc_net_remove(&init_net, "raw");
928 #endif /* CONFIG_PROC_FS */