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/config.h>
43 #include <linux/types.h>
44 #include <asm/atomic.h>
45 #include <asm/byteorder.h>
46 #include <asm/current.h>
47 #include <asm/uaccess.h>
48 #include <asm/ioctls.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/skbuff.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 = skb->h.icmph->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 int type = skb->h.icmph->type;
189 int code = skb->h.icmph->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->nh.raw);
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->nh.iph = iph = (struct iphdr *)skb_put(skb, length);
297 skb->ip_summed = CHECKSUM_NONE;
299 skb->h.raw = skb->nh.raw;
300 err = memcpy_fromiovecend((void *)iph, from, 0, length);
304 /* We don't modify invalid header */
305 if (length >= sizeof(*iph) && iph->ihl * 4U <= length) {
307 iph->saddr = rt->rt_src;
309 iph->tot_len = htons(length);
311 ip_select_ident(iph, &rt->u.dst, NULL);
313 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
316 err = NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
319 err = inet->recverr ? net_xmit_errno(err) : 0;
329 IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
333 static void raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
336 u8 __user *type = NULL;
337 u8 __user *code = NULL;
344 for (i = 0; i < msg->msg_iovlen; i++) {
345 iov = &msg->msg_iov[i];
351 /* check if one-byte field is readable or not. */
352 if (iov->iov_base && iov->iov_len < 1)
356 type = iov->iov_base;
357 /* check if code field is readable or not. */
358 if (iov->iov_len > 1)
361 code = iov->iov_base;
364 get_user(fl->fl_icmp_type, type);
365 get_user(fl->fl_icmp_code, code);
378 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
381 struct inet_sock *inet = inet_sk(sk);
382 struct ipcm_cookie ipc;
383 struct rtable *rt = NULL;
399 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
400 goto out; /* compatibility */
403 * Get and verify the address.
406 if (msg->msg_namelen) {
407 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
409 if (msg->msg_namelen < sizeof(*usin))
411 if (usin->sin_family != AF_INET) {
412 static int complained;
414 printk(KERN_INFO "%s forgot to set AF_INET in "
415 "raw sendmsg. Fix it!\n",
418 if (usin->sin_family)
421 daddr = usin->sin_addr.s_addr;
422 /* ANK: I did not forget to get protocol from port field.
423 * I just do not know, who uses this weirdness.
424 * IP_HDRINCL is much more convenient.
428 if (sk->sk_state != TCP_ESTABLISHED)
433 ipc.addr = inet->saddr;
435 ipc.oif = sk->sk_bound_dev_if;
437 if (msg->msg_controllen) {
438 err = ip_cmsg_send(msg, &ipc);
453 /* Linux does not mangle headers on raw sockets,
454 * so that IP options + IP_HDRINCL is non-sense.
461 daddr = ipc.opt->faddr;
464 tos = RT_CONN_FLAGS(sk);
465 if (msg->msg_flags & MSG_DONTROUTE)
468 if (MULTICAST(daddr)) {
470 ipc.oif = inet->mc_index;
472 saddr = inet->mc_addr;
476 struct flowi fl = { .oif = ipc.oif,
481 .proto = inet->hdrincl ? IPPROTO_RAW :
485 raw_probe_proto_opt(&fl, msg);
487 err = ip_route_output_flow(&rt, &fl, sk, !(msg->msg_flags&MSG_DONTWAIT));
493 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
496 if (msg->msg_flags & MSG_CONFIRM)
501 err = raw_send_hdrinc(sk, msg->msg_iov, len,
506 ipc.addr = rt->rt_dst;
508 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
509 &ipc, rt, msg->msg_flags);
511 ip_flush_pending_frames(sk);
512 else if (!(msg->msg_flags & MSG_MORE))
513 err = ip_push_pending_frames(sk);
527 dst_confirm(&rt->u.dst);
528 if (!(msg->msg_flags & MSG_PROBE) || len)
529 goto back_from_confirm;
534 static void raw_close(struct sock *sk, long timeout)
537 * Raw sockets may have direct kernel refereneces. Kill them.
539 ip_ra_control(sk, 0, NULL);
541 sk_common_release(sk);
544 /* This gets rid of all the nasties in af_inet. -DaveM */
545 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
547 struct inet_sock *inet = inet_sk(sk);
548 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
552 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
554 chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
555 ret = -EADDRNOTAVAIL;
556 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
557 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
559 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
560 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
561 inet->saddr = 0; /* Use device */
568 * This should be easy, if there is something there
569 * we return it, otherwise we block.
572 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
573 size_t len, int noblock, int flags, int *addr_len)
575 struct inet_sock *inet = inet_sk(sk);
577 int err = -EOPNOTSUPP;
578 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
585 *addr_len = sizeof(*sin);
587 if (flags & MSG_ERRQUEUE) {
588 err = ip_recv_error(sk, msg, len);
592 skb = skb_recv_datagram(sk, flags, noblock, &err);
598 msg->msg_flags |= MSG_TRUNC;
602 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
606 sock_recv_timestamp(msg, sk, skb);
608 /* Copy the address. */
610 sin->sin_family = AF_INET;
611 sin->sin_addr.s_addr = skb->nh.iph->saddr;
612 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
614 if (inet->cmsg_flags)
615 ip_cmsg_recv(msg, skb);
616 if (flags & MSG_TRUNC)
619 skb_free_datagram(sk, skb);
626 static int raw_init(struct sock *sk)
628 struct raw_sock *rp = raw_sk(sk);
630 if (inet_sk(sk)->num == IPPROTO_ICMP)
631 memset(&rp->filter, 0, sizeof(rp->filter));
635 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
637 if (optlen > sizeof(struct icmp_filter))
638 optlen = sizeof(struct icmp_filter);
639 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
644 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
646 int len, ret = -EFAULT;
648 if (get_user(len, optlen))
653 if (len > sizeof(struct icmp_filter))
654 len = sizeof(struct icmp_filter);
656 if (put_user(len, optlen) ||
657 copy_to_user(optval, &raw_sk(sk)->filter, len))
663 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
664 char __user *optval, int optlen)
666 if (optname == ICMP_FILTER) {
667 if (inet_sk(sk)->num != IPPROTO_ICMP)
670 return raw_seticmpfilter(sk, optval, optlen);
675 static int raw_setsockopt(struct sock *sk, int level, int optname,
676 char __user *optval, int optlen)
678 if (level != SOL_RAW)
679 return ip_setsockopt(sk, level, optname, optval, optlen);
680 return do_raw_setsockopt(sk, level, optname, optval, optlen);
684 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
685 char __user *optval, int optlen)
687 if (level != SOL_RAW)
688 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
689 return do_raw_setsockopt(sk, level, optname, optval, optlen);
693 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
694 char __user *optval, int __user *optlen)
696 if (optname == ICMP_FILTER) {
697 if (inet_sk(sk)->num != IPPROTO_ICMP)
700 return raw_geticmpfilter(sk, optval, optlen);
705 static int raw_getsockopt(struct sock *sk, int level, int optname,
706 char __user *optval, int __user *optlen)
708 if (level != SOL_RAW)
709 return ip_getsockopt(sk, level, optname, optval, optlen);
710 return do_raw_getsockopt(sk, level, optname, optval, optlen);
714 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
715 char __user *optval, int __user *optlen)
717 if (level != SOL_RAW)
718 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
719 return do_raw_getsockopt(sk, level, optname, optval, optlen);
723 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
727 int amount = atomic_read(&sk->sk_wmem_alloc);
728 return put_user(amount, (int __user *)arg);
734 spin_lock_bh(&sk->sk_receive_queue.lock);
735 skb = skb_peek(&sk->sk_receive_queue);
738 spin_unlock_bh(&sk->sk_receive_queue.lock);
739 return put_user(amount, (int __user *)arg);
743 #ifdef CONFIG_IP_MROUTE
744 return ipmr_ioctl(sk, cmd, (void __user *)arg);
751 struct proto raw_prot = {
753 .owner = THIS_MODULE,
755 .connect = ip4_datagram_connect,
756 .disconnect = udp_disconnect,
759 .setsockopt = raw_setsockopt,
760 .getsockopt = raw_getsockopt,
761 .sendmsg = raw_sendmsg,
762 .recvmsg = raw_recvmsg,
764 .backlog_rcv = raw_rcv_skb,
766 .unhash = raw_v4_unhash,
767 .obj_size = sizeof(struct raw_sock),
769 .compat_setsockopt = compat_raw_setsockopt,
770 .compat_getsockopt = compat_raw_getsockopt,
774 #ifdef CONFIG_PROC_FS
775 struct raw_iter_state {
779 #define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
781 static struct sock *raw_get_first(struct seq_file *seq)
784 struct raw_iter_state* state = raw_seq_private(seq);
786 for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) {
787 struct hlist_node *node;
789 sk_for_each(sk, node, &raw_v4_htable[state->bucket])
790 if (sk->sk_family == PF_INET)
798 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
800 struct raw_iter_state* state = raw_seq_private(seq);
806 } while (sk && sk->sk_family != PF_INET);
808 if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) {
809 sk = sk_head(&raw_v4_htable[state->bucket]);
815 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
817 struct sock *sk = raw_get_first(seq);
820 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
822 return pos ? NULL : sk;
825 static void *raw_seq_start(struct seq_file *seq, loff_t *pos)
827 read_lock(&raw_v4_lock);
828 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
831 static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
835 if (v == SEQ_START_TOKEN)
836 sk = raw_get_first(seq);
838 sk = raw_get_next(seq, v);
843 static void raw_seq_stop(struct seq_file *seq, void *v)
845 read_unlock(&raw_v4_lock);
848 static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i)
850 struct inet_sock *inet = inet_sk(sp);
851 unsigned int dest = inet->daddr,
852 src = inet->rcv_saddr;
856 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
857 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
858 i, src, srcp, dest, destp, sp->sk_state,
859 atomic_read(&sp->sk_wmem_alloc),
860 atomic_read(&sp->sk_rmem_alloc),
861 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
862 atomic_read(&sp->sk_refcnt), sp);
866 static int raw_seq_show(struct seq_file *seq, void *v)
870 if (v == SEQ_START_TOKEN)
871 seq_printf(seq, "%-127s\n",
872 " sl local_address rem_address st tx_queue "
873 "rx_queue tr tm->when retrnsmt uid timeout "
876 struct raw_iter_state *state = raw_seq_private(seq);
878 seq_printf(seq, "%-127s\n",
879 get_raw_sock(v, tmpbuf, state->bucket));
884 static struct seq_operations raw_seq_ops = {
885 .start = raw_seq_start,
886 .next = raw_seq_next,
887 .stop = raw_seq_stop,
888 .show = raw_seq_show,
891 static int raw_seq_open(struct inode *inode, struct file *file)
893 struct seq_file *seq;
895 struct raw_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
899 rc = seq_open(file, &raw_seq_ops);
903 seq = file->private_data;
905 memset(s, 0, sizeof(*s));
913 static struct file_operations raw_seq_fops = {
914 .owner = THIS_MODULE,
915 .open = raw_seq_open,
918 .release = seq_release_private,
921 int __init raw_proc_init(void)
923 if (!proc_net_fops_create("raw", S_IRUGO, &raw_seq_fops))
928 void __init raw_proc_exit(void)
930 proc_net_remove("raw");
932 #endif /* CONFIG_PROC_FS */