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);
105 struct sock *__raw_v4_lookup(struct sock *sk, unsigned short num,
106 unsigned long raddr, unsigned long laddr,
109 struct hlist_node *node;
111 sk_for_each_from(sk, node) {
112 struct inet_opt *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 type = skb->h.icmph->type;
135 __u32 data = raw4_sk(sk)->filter.data;
137 return ((1 << type) & data) != 0;
140 /* Do not block unknown ICMP types */
144 /* IP input processing comes here for RAW socket delivery.
145 * Caller owns SKB, so we must make clones.
147 * RFC 1122: SHOULD pass TOS value up to the transport layer.
148 * -> It does. And not only TOS, but all IP header.
150 void raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
153 struct hlist_head *head;
155 read_lock(&raw_v4_lock);
156 head = &raw_v4_htable[hash];
157 if (hlist_empty(head))
159 sk = __raw_v4_lookup(__sk_head(head), iph->protocol,
160 iph->saddr, iph->daddr,
164 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
165 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
167 /* Not releasing hash table! */
171 sk = __raw_v4_lookup(sk_next(sk), iph->protocol,
172 iph->saddr, iph->daddr,
176 read_unlock(&raw_v4_lock);
179 void raw_err (struct sock *sk, struct sk_buff *skb, u32 info)
181 struct inet_opt *inet = inet_sk(sk);
182 int type = skb->h.icmph->type;
183 int code = skb->h.icmph->code;
187 /* Report error on raw socket, if:
188 1. User requested ip_recverr.
189 2. Socket is connected (otherwise the error indication
190 is useless without ip_recverr and error is hard.
192 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
197 case ICMP_TIME_EXCEEDED:
200 case ICMP_SOURCE_QUENCH:
202 case ICMP_PARAMETERPROB:
206 case ICMP_DEST_UNREACH:
208 if (code > NR_ICMP_UNREACH)
210 err = icmp_err_convert[code].errno;
211 harderr = icmp_err_convert[code].fatal;
212 if (code == ICMP_FRAG_NEEDED) {
213 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
219 struct iphdr *iph = (struct iphdr*)skb->data;
220 u8 *payload = skb->data + (iph->ihl << 2);
224 ip_icmp_error(sk, skb, err, 0, info, payload);
227 if (inet->recverr || harderr) {
229 sk->sk_error_report(sk);
233 static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
235 /* Charge it to the socket. */
237 if (sock_queue_rcv_skb(sk, skb) < 0) {
238 /* FIXME: increment a raw drops counter here */
243 return NET_RX_SUCCESS;
246 int raw_rcv(struct sock *sk, struct sk_buff *skb)
248 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
253 skb_push(skb, skb->data - skb->nh.raw);
255 raw_rcv_skb(sk, skb);
259 static int raw_send_hdrinc(struct sock *sk, void *from, int length,
263 struct inet_opt *inet = inet_sk(sk);
269 if (length > rt->u.dst.dev->mtu) {
270 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
277 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
279 skb = sock_alloc_send_skb(sk, length+hh_len+15,
280 flags&MSG_DONTWAIT, &err);
283 skb_reserve(skb, hh_len);
285 skb->priority = sk->sk_priority;
286 skb->dst = dst_clone(&rt->u.dst);
288 skb->nh.iph = iph = (struct iphdr *)skb_put(skb, length);
290 skb->ip_summed = CHECKSUM_NONE;
292 skb->h.raw = skb->nh.raw;
293 err = memcpy_fromiovecend((void *)iph, from, 0, length);
297 /* We don't modify invalid header */
298 if (length >= sizeof(*iph) && iph->ihl * 4 <= length) {
300 iph->saddr = rt->rt_src;
302 iph->tot_len = htons(length);
304 ip_select_ident(iph, &rt->u.dst, NULL);
306 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
309 err = NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
312 err = inet->recverr ? net_xmit_errno(err) : 0;
322 IP_INC_STATS(IpOutDiscards);
326 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
329 struct inet_opt *inet = inet_sk(sk);
330 struct ipcm_cookie ipc;
331 struct rtable *rt = NULL;
339 if (len < 0 || len > 0xFFFF)
347 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
348 goto out; /* compatibility */
351 * Get and verify the address.
354 if (msg->msg_namelen) {
355 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
357 if (msg->msg_namelen < sizeof(*usin))
359 if (usin->sin_family != AF_INET) {
360 static int complained;
362 printk(KERN_INFO "%s forgot to set AF_INET in "
363 "raw sendmsg. Fix it!\n",
366 if (usin->sin_family)
369 daddr = usin->sin_addr.s_addr;
370 /* ANK: I did not forget to get protocol from port field.
371 * I just do not know, who uses this weirdness.
372 * IP_HDRINCL is much more convenient.
376 if (sk->sk_state != TCP_ESTABLISHED)
381 ipc.addr = inet->saddr;
383 ipc.oif = sk->sk_bound_dev_if;
385 if (msg->msg_controllen) {
386 err = ip_cmsg_send(msg, &ipc);
401 /* Linux does not mangle headers on raw sockets,
402 * so that IP options + IP_HDRINCL is non-sense.
409 daddr = ipc.opt->faddr;
412 tos = RT_TOS(inet->tos) | sk->sk_localroute;
413 if (msg->msg_flags & MSG_DONTROUTE)
416 if (MULTICAST(daddr)) {
418 ipc.oif = inet->mc_index;
420 saddr = inet->mc_addr;
424 struct flowi fl = { .oif = ipc.oif,
429 .proto = inet->hdrincl ? IPPROTO_RAW :
432 err = ip_route_output_flow(&rt, &fl, sk, !(msg->msg_flags&MSG_DONTWAIT));
438 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
441 if (msg->msg_flags & MSG_CONFIRM)
446 err = raw_send_hdrinc(sk, msg->msg_iov, len,
451 ipc.addr = rt->rt_dst;
453 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
454 &ipc, rt, msg->msg_flags);
456 ip_flush_pending_frames(sk);
457 else if (!(msg->msg_flags & MSG_MORE))
458 err = ip_push_pending_frames(sk);
466 out: return err < 0 ? err : len;
469 dst_confirm(&rt->u.dst);
470 if (!(msg->msg_flags & MSG_PROBE) || len)
471 goto back_from_confirm;
476 static void raw_close(struct sock *sk, long timeout)
479 * Raw sockets may have direct kernel refereneces. Kill them.
481 ip_ra_control(sk, 0, NULL);
483 inet_sock_release(sk);
486 /* This gets rid of all the nasties in af_inet. -DaveM */
487 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
489 struct inet_opt *inet = inet_sk(sk);
490 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
494 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
496 chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
497 ret = -EADDRNOTAVAIL;
498 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
499 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
501 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
502 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
503 inet->saddr = 0; /* Use device */
510 * This should be easy, if there is something there
511 * we return it, otherwise we block.
514 int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
515 size_t len, int noblock, int flags, int *addr_len)
517 struct inet_opt *inet = inet_sk(sk);
519 int err = -EOPNOTSUPP;
520 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
527 *addr_len = sizeof(*sin);
529 if (flags & MSG_ERRQUEUE) {
530 err = ip_recv_error(sk, msg, len);
534 skb = skb_recv_datagram(sk, flags, noblock, &err);
540 msg->msg_flags |= MSG_TRUNC;
544 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
548 sock_recv_timestamp(msg, sk, skb);
550 /* Copy the address. */
552 sin->sin_family = AF_INET;
553 sin->sin_addr.s_addr = skb->nh.iph->saddr;
554 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
556 if (inet->cmsg_flags)
557 ip_cmsg_recv(msg, skb);
559 skb_free_datagram(sk, skb);
560 out: return err ? : copied;
563 static int raw_init(struct sock *sk)
565 struct raw_opt *tp = raw4_sk(sk);
566 if (inet_sk(sk)->num == IPPROTO_ICMP)
567 memset(&tp->filter, 0, sizeof(tp->filter));
571 static int raw_seticmpfilter(struct sock *sk, char *optval, int optlen)
573 if (optlen > sizeof(struct icmp_filter))
574 optlen = sizeof(struct icmp_filter);
575 if (copy_from_user(&raw4_sk(sk)->filter, optval, optlen))
580 static int raw_geticmpfilter(struct sock *sk, char *optval, int *optlen)
582 int len, ret = -EFAULT;
584 if (get_user(len, optlen))
589 if (len > sizeof(struct icmp_filter))
590 len = sizeof(struct icmp_filter);
592 if (put_user(len, optlen) ||
593 copy_to_user(optval, &raw4_sk(sk)->filter, len))
599 static int raw_setsockopt(struct sock *sk, int level, int optname,
600 char *optval, int optlen)
602 if (level != SOL_RAW)
603 return ip_setsockopt(sk, level, optname, optval, optlen);
605 if (optname == ICMP_FILTER) {
606 if (inet_sk(sk)->num != IPPROTO_ICMP)
609 return raw_seticmpfilter(sk, optval, optlen);
614 static int raw_getsockopt(struct sock *sk, int level, int optname,
615 char *optval, int *optlen)
617 if (level != SOL_RAW)
618 return ip_getsockopt(sk, level, optname, optval, optlen);
620 if (optname == ICMP_FILTER) {
621 if (inet_sk(sk)->num != IPPROTO_ICMP)
624 return raw_geticmpfilter(sk, optval, optlen);
629 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
633 int amount = atomic_read(&sk->sk_wmem_alloc);
634 return put_user(amount, (int *)arg);
640 spin_lock_irq(&sk->sk_receive_queue.lock);
641 skb = skb_peek(&sk->sk_receive_queue);
644 spin_unlock_irq(&sk->sk_receive_queue.lock);
645 return put_user(amount, (int *)arg);
649 #ifdef CONFIG_IP_MROUTE
650 return ipmr_ioctl(sk, cmd, arg);
657 struct proto raw_prot = {
660 .connect = udp_connect,
661 .disconnect = udp_disconnect,
664 .setsockopt = raw_setsockopt,
665 .getsockopt = raw_getsockopt,
666 .sendmsg = raw_sendmsg,
667 .recvmsg = raw_recvmsg,
669 .backlog_rcv = raw_rcv_skb,
671 .unhash = raw_v4_unhash,
674 #ifdef CONFIG_PROC_FS
675 struct raw_iter_state {
679 #define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
681 static struct sock *raw_get_first(struct seq_file *seq)
684 struct raw_iter_state* state = raw_seq_private(seq);
686 for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) {
687 struct hlist_node *node;
689 sk_for_each(sk, node, &raw_v4_htable[state->bucket])
690 if (sk->sk_family == PF_INET)
698 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
700 struct raw_iter_state* state = raw_seq_private(seq);
706 } while (sk && sk->sk_family != PF_INET);
708 if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) {
709 sk = sk_head(&raw_v4_htable[state->bucket]);
715 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
717 struct sock *sk = raw_get_first(seq);
720 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
722 return pos ? NULL : sk;
725 static void *raw_seq_start(struct seq_file *seq, loff_t *pos)
727 read_lock(&raw_v4_lock);
728 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
731 static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
735 if (v == SEQ_START_TOKEN)
736 sk = raw_get_first(seq);
738 sk = raw_get_next(seq, v);
743 static void raw_seq_stop(struct seq_file *seq, void *v)
745 read_unlock(&raw_v4_lock);
748 static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i)
750 struct inet_opt *inet = inet_sk(sp);
751 unsigned int dest = inet->daddr,
752 src = inet->rcv_saddr;
756 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
757 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
758 i, src, srcp, dest, destp, sp->sk_state,
759 atomic_read(&sp->sk_wmem_alloc),
760 atomic_read(&sp->sk_rmem_alloc),
761 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
762 atomic_read(&sp->sk_refcnt), sp);
766 static int raw_seq_show(struct seq_file *seq, void *v)
770 if (v == SEQ_START_TOKEN)
771 seq_printf(seq, "%-127s\n",
772 " sl local_address rem_address st tx_queue "
773 "rx_queue tr tm->when retrnsmt uid timeout "
776 struct raw_iter_state *state = raw_seq_private(seq);
778 seq_printf(seq, "%-127s\n",
779 get_raw_sock(v, tmpbuf, state->bucket));
784 static struct seq_operations raw_seq_ops = {
785 .start = raw_seq_start,
786 .next = raw_seq_next,
787 .stop = raw_seq_stop,
788 .show = raw_seq_show,
791 static int raw_seq_open(struct inode *inode, struct file *file)
793 struct seq_file *seq;
795 struct raw_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
799 rc = seq_open(file, &raw_seq_ops);
803 seq = file->private_data;
805 memset(s, 0, sizeof(*s));
813 static struct file_operations raw_seq_fops = {
814 .owner = THIS_MODULE,
815 .open = raw_seq_open,
818 .release = seq_release_private,
821 int __init raw_proc_init(void)
823 if (!proc_net_fops_create("raw", S_IRUGO, &raw_seq_fops))
828 void __init raw_proc_exit(void)
830 proc_net_remove("raw");
832 #endif /* CONFIG_PROC_FS */