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 $
10 * Authors: Ross Biro, <bir7@leland.Stanford.Edu>
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/config.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/types.h>
49 #include <linux/stddef.h>
50 #include <linux/slab.h>
51 #include <linux/errno.h>
52 #include <linux/aio.h>
53 #include <linux/kernel.h>
54 #include <linux/spinlock.h>
55 #include <linux/sockios.h>
56 #include <linux/socket.h>
58 #include <linux/mroute.h>
59 #include <linux/netdevice.h>
60 #include <linux/in_route.h>
61 #include <linux/route.h>
62 #include <linux/tcp.h>
63 #include <linux/skbuff.h>
66 #include <linux/gfp.h>
68 #include <linux/net.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 rwlock_t raw_v4_lock = RW_LOCK_UNLOCKED;
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);
107 Check if an address is in the list
109 static inline int raw_addr_in_list (
113 struct nx_info *nx_info)
117 (rcv_saddr1 == loc_addr || rcv_saddr2 == loc_addr))
119 else if (rcv_saddr1 == 0) {
120 /* Accept any address or only the one in the list */
124 int n = nx_info->nbipv4;
126 for (i=0; i<n; i++) {
127 if (nx_info->ipv4[i] == loc_addr) {
137 struct sock *__raw_v4_lookup(struct sock *sk, unsigned short num,
138 unsigned long raddr, unsigned long laddr,
141 struct hlist_node *node;
143 sk_for_each_from(sk, node) {
144 struct inet_opt *inet = inet_sk(sk);
146 if (inet->num == num &&
147 !(inet->daddr && inet->daddr != raddr) &&
148 raw_addr_in_list(inet->rcv_saddr, inet->rcv_saddr2,
149 laddr, sk->sk_nx_info) &&
150 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
151 goto found; /* gotcha */
162 static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
166 type = skb->h.icmph->type;
168 __u32 data = raw4_sk(sk)->filter.data;
170 return ((1 << type) & data) != 0;
173 /* Do not block unknown ICMP types */
177 /* IP input processing comes here for RAW socket delivery.
178 * Caller owns SKB, so we must make clones.
180 * RFC 1122: SHOULD pass TOS value up to the transport layer.
181 * -> It does. And not only TOS, but all IP header.
183 void raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
186 struct hlist_head *head;
188 read_lock(&raw_v4_lock);
189 head = &raw_v4_htable[hash];
190 if (hlist_empty(head))
192 sk = __raw_v4_lookup(__sk_head(head), iph->protocol,
193 iph->saddr, iph->daddr,
197 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
198 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
200 /* Not releasing hash table! */
204 sk = __raw_v4_lookup(sk_next(sk), iph->protocol,
205 iph->saddr, iph->daddr,
209 read_unlock(&raw_v4_lock);
212 void raw_err (struct sock *sk, struct sk_buff *skb, u32 info)
214 struct inet_opt *inet = inet_sk(sk);
215 int type = skb->h.icmph->type;
216 int code = skb->h.icmph->code;
220 /* Report error on raw socket, if:
221 1. User requested ip_recverr.
222 2. Socket is connected (otherwise the error indication
223 is useless without ip_recverr and error is hard.
225 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
230 case ICMP_TIME_EXCEEDED:
233 case ICMP_SOURCE_QUENCH:
235 case ICMP_PARAMETERPROB:
239 case ICMP_DEST_UNREACH:
241 if (code > NR_ICMP_UNREACH)
243 err = icmp_err_convert[code].errno;
244 harderr = icmp_err_convert[code].fatal;
245 if (code == ICMP_FRAG_NEEDED) {
246 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
252 struct iphdr *iph = (struct iphdr*)skb->data;
253 u8 *payload = skb->data + (iph->ihl << 2);
257 ip_icmp_error(sk, skb, err, 0, info, payload);
260 if (inet->recverr || harderr) {
262 sk->sk_error_report(sk);
266 static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
268 /* Charge it to the socket. */
270 if (sock_queue_rcv_skb(sk, skb) < 0) {
271 /* FIXME: increment a raw drops counter here */
276 return NET_RX_SUCCESS;
279 int raw_rcv(struct sock *sk, struct sk_buff *skb)
281 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
286 skb_push(skb, skb->data - skb->nh.raw);
288 raw_rcv_skb(sk, skb);
292 static int raw_send_hdrinc(struct sock *sk, void *from, int length,
296 struct inet_opt *inet = inet_sk(sk);
302 if (length > rt->u.dst.dev->mtu) {
303 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
310 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
312 skb = sock_alloc_send_skb(sk, length+hh_len+15,
313 flags&MSG_DONTWAIT, &err);
316 skb_reserve(skb, hh_len);
318 skb->priority = sk->sk_priority;
319 skb->dst = dst_clone(&rt->u.dst);
321 skb->nh.iph = iph = (struct iphdr *)skb_put(skb, length);
323 skb->ip_summed = CHECKSUM_NONE;
325 skb->h.raw = skb->nh.raw;
326 err = memcpy_fromiovecend((void *)iph, from, 0, length);
330 /* We don't modify invalid header */
331 if (length >= sizeof(*iph) && iph->ihl * 4 <= length) {
333 iph->saddr = rt->rt_src;
335 iph->tot_len = htons(length);
337 ip_select_ident(iph, &rt->u.dst, NULL);
339 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
342 err = NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
345 err = inet->recverr ? net_xmit_errno(err) : 0;
355 IP_INC_STATS(IpOutDiscards);
359 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
362 struct inet_opt *inet = inet_sk(sk);
363 struct ipcm_cookie ipc;
364 struct rtable *rt = NULL;
372 if (len < 0 || len > 0xFFFF)
380 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
381 goto out; /* compatibility */
384 * Get and verify the address.
387 if (msg->msg_namelen) {
388 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
390 if (msg->msg_namelen < sizeof(*usin))
392 if (usin->sin_family != AF_INET) {
393 static int complained;
395 printk(KERN_INFO "%s forgot to set AF_INET in "
396 "raw sendmsg. Fix it!\n",
399 if (usin->sin_family)
402 daddr = usin->sin_addr.s_addr;
403 /* ANK: I did not forget to get protocol from port field.
404 * I just do not know, who uses this weirdness.
405 * IP_HDRINCL is much more convenient.
409 if (sk->sk_state != TCP_ESTABLISHED)
414 ipc.addr = inet->saddr;
416 ipc.oif = sk->sk_bound_dev_if;
418 if (msg->msg_controllen) {
419 err = ip_cmsg_send(msg, &ipc);
434 /* Linux does not mangle headers on raw sockets,
435 * so that IP options + IP_HDRINCL is non-sense.
442 daddr = ipc.opt->faddr;
445 tos = RT_TOS(inet->tos) | sk->sk_localroute;
446 if (msg->msg_flags & MSG_DONTROUTE)
449 if (MULTICAST(daddr)) {
451 ipc.oif = inet->mc_index;
453 saddr = inet->mc_addr;
457 struct flowi fl = { .oif = ipc.oif,
462 .proto = inet->hdrincl ? IPPROTO_RAW :
465 err = ip_route_output_flow(&rt, &fl, sk, !(msg->msg_flags&MSG_DONTWAIT));
471 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
474 if (msg->msg_flags & MSG_CONFIRM)
479 err = raw_send_hdrinc(sk, msg->msg_iov, len,
484 ipc.addr = rt->rt_dst;
486 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
487 &ipc, rt, msg->msg_flags);
489 ip_flush_pending_frames(sk);
490 else if (!(msg->msg_flags & MSG_MORE))
491 err = ip_push_pending_frames(sk);
499 out: return err < 0 ? err : len;
502 dst_confirm(&rt->u.dst);
503 if (!(msg->msg_flags & MSG_PROBE) || len)
504 goto back_from_confirm;
509 static void raw_close(struct sock *sk, long timeout)
512 * Raw sockets may have direct kernel refereneces. Kill them.
514 ip_ra_control(sk, 0, NULL);
516 inet_sock_release(sk);
519 /* This gets rid of all the nasties in af_inet. -DaveM */
520 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
522 struct inet_opt *inet = inet_sk(sk);
523 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
527 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
529 chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
530 ret = -EADDRNOTAVAIL;
531 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
532 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
534 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
535 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
536 inet->saddr = 0; /* Use device */
543 * This should be easy, if there is something there
544 * we return it, otherwise we block.
547 int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
548 size_t len, int noblock, int flags, int *addr_len)
550 struct inet_opt *inet = inet_sk(sk);
552 int err = -EOPNOTSUPP;
553 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
560 *addr_len = sizeof(*sin);
562 if (flags & MSG_ERRQUEUE) {
563 err = ip_recv_error(sk, msg, len);
567 skb = skb_recv_datagram(sk, flags, noblock, &err);
573 msg->msg_flags |= MSG_TRUNC;
577 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
581 sock_recv_timestamp(msg, sk, skb);
583 /* Copy the address. */
585 sin->sin_family = AF_INET;
586 sin->sin_addr.s_addr = skb->nh.iph->saddr;
587 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
589 if (inet->cmsg_flags)
590 ip_cmsg_recv(msg, skb);
592 skb_free_datagram(sk, skb);
593 out: return err ? : copied;
596 static int raw_init(struct sock *sk)
598 struct raw_opt *tp = raw4_sk(sk);
599 if (inet_sk(sk)->num == IPPROTO_ICMP)
600 memset(&tp->filter, 0, sizeof(tp->filter));
604 static int raw_seticmpfilter(struct sock *sk, char *optval, int optlen)
606 if (optlen > sizeof(struct icmp_filter))
607 optlen = sizeof(struct icmp_filter);
608 if (copy_from_user(&raw4_sk(sk)->filter, optval, optlen))
613 static int raw_geticmpfilter(struct sock *sk, char *optval, int *optlen)
615 int len, ret = -EFAULT;
617 if (get_user(len, optlen))
622 if (len > sizeof(struct icmp_filter))
623 len = sizeof(struct icmp_filter);
625 if (put_user(len, optlen) ||
626 copy_to_user(optval, &raw4_sk(sk)->filter, len))
632 static int raw_setsockopt(struct sock *sk, int level, int optname,
633 char *optval, int optlen)
635 if (level != SOL_RAW)
636 return ip_setsockopt(sk, level, optname, optval, optlen);
638 if (optname == ICMP_FILTER) {
639 if (inet_sk(sk)->num != IPPROTO_ICMP)
642 return raw_seticmpfilter(sk, optval, optlen);
647 static int raw_getsockopt(struct sock *sk, int level, int optname,
648 char *optval, int *optlen)
650 if (level != SOL_RAW)
651 return ip_getsockopt(sk, level, optname, optval, optlen);
653 if (optname == ICMP_FILTER) {
654 if (inet_sk(sk)->num != IPPROTO_ICMP)
657 return raw_geticmpfilter(sk, optval, optlen);
662 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
666 int amount = atomic_read(&sk->sk_wmem_alloc);
667 return put_user(amount, (int *)arg);
673 spin_lock_irq(&sk->sk_receive_queue.lock);
674 skb = skb_peek(&sk->sk_receive_queue);
677 spin_unlock_irq(&sk->sk_receive_queue.lock);
678 return put_user(amount, (int *)arg);
682 #ifdef CONFIG_IP_MROUTE
683 return ipmr_ioctl(sk, cmd, arg);
690 struct proto raw_prot = {
693 .connect = udp_connect,
694 .disconnect = udp_disconnect,
697 .setsockopt = raw_setsockopt,
698 .getsockopt = raw_getsockopt,
699 .sendmsg = raw_sendmsg,
700 .recvmsg = raw_recvmsg,
702 .backlog_rcv = raw_rcv_skb,
704 .unhash = raw_v4_unhash,
707 #ifdef CONFIG_PROC_FS
708 struct raw_iter_state {
712 #define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
714 static struct sock *raw_get_first(struct seq_file *seq)
717 struct raw_iter_state* state = raw_seq_private(seq);
719 for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) {
720 struct hlist_node *node;
722 sk_for_each(sk, node, &raw_v4_htable[state->bucket])
723 if (sk->sk_family == PF_INET &&
724 vx_check(sk->sk_xid, VX_WATCH|VX_IDENT))
732 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
734 struct raw_iter_state* state = raw_seq_private(seq);
740 } while (sk && (sk->sk_family != PF_INET ||
741 !vx_check(sk->sk_xid, VX_WATCH|VX_IDENT)));
743 if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) {
744 sk = sk_head(&raw_v4_htable[state->bucket]);
750 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
752 struct sock *sk = raw_get_first(seq);
755 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
757 return pos ? NULL : sk;
760 static void *raw_seq_start(struct seq_file *seq, loff_t *pos)
762 read_lock(&raw_v4_lock);
763 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
766 static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
770 if (v == SEQ_START_TOKEN)
771 sk = raw_get_first(seq);
773 sk = raw_get_next(seq, v);
778 static void raw_seq_stop(struct seq_file *seq, void *v)
780 read_unlock(&raw_v4_lock);
783 static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i)
785 struct inet_opt *inet = inet_sk(sp);
786 unsigned int dest = inet->daddr,
787 src = inet->rcv_saddr;
791 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
792 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
793 i, src, srcp, dest, destp, sp->sk_state,
794 atomic_read(&sp->sk_wmem_alloc),
795 atomic_read(&sp->sk_rmem_alloc),
796 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
797 atomic_read(&sp->sk_refcnt), sp);
801 static int raw_seq_show(struct seq_file *seq, void *v)
805 if (v == SEQ_START_TOKEN)
806 seq_printf(seq, "%-127s\n",
807 " sl local_address rem_address st tx_queue "
808 "rx_queue tr tm->when retrnsmt uid timeout "
811 struct raw_iter_state *state = raw_seq_private(seq);
813 seq_printf(seq, "%-127s\n",
814 get_raw_sock(v, tmpbuf, state->bucket));
819 static struct seq_operations raw_seq_ops = {
820 .start = raw_seq_start,
821 .next = raw_seq_next,
822 .stop = raw_seq_stop,
823 .show = raw_seq_show,
826 static int raw_seq_open(struct inode *inode, struct file *file)
828 struct seq_file *seq;
830 struct raw_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
834 rc = seq_open(file, &raw_seq_ops);
838 seq = file->private_data;
840 memset(s, 0, sizeof(*s));
848 static struct file_operations raw_seq_fops = {
849 .owner = THIS_MODULE,
850 .open = raw_seq_open,
853 .release = seq_release_private,
856 int __init raw_proc_init(void)
858 if (!proc_net_fops_create("raw", S_IRUGO, &raw_seq_fops))
863 void __init raw_proc_exit(void)
865 proc_net_remove("raw");
867 #endif /* CONFIG_PROC_FS */