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 a given address matches for a socket
109 * nxi: the socket's nx_info if any
110 * addr: to be verified address
111 * saddr/baddr: socket addresses
113 static inline int raw_addr_match (
119 if (addr && (saddr == addr || baddr == addr))
122 return addr_in_nx_info(nxi, addr);
126 struct sock *__raw_v4_lookup(struct sock *sk, unsigned short num,
127 unsigned long raddr, unsigned long laddr,
130 struct hlist_node *node;
132 sk_for_each_from(sk, node) {
133 struct inet_opt *inet = inet_sk(sk);
135 if (inet->num == num &&
136 !(inet->daddr && inet->daddr != raddr) &&
137 raw_addr_match(sk->sk_nx_info, laddr,
138 inet->rcv_saddr, inet->rcv_saddr2) &&
139 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
140 goto found; /* gotcha */
151 static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
155 type = skb->h.icmph->type;
157 __u32 data = raw4_sk(sk)->filter.data;
159 return ((1 << type) & data) != 0;
162 /* Do not block unknown ICMP types */
166 /* IP input processing comes here for RAW socket delivery.
167 * Caller owns SKB, so we must make clones.
169 * RFC 1122: SHOULD pass TOS value up to the transport layer.
170 * -> It does. And not only TOS, but all IP header.
172 void raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
175 struct hlist_head *head;
177 read_lock(&raw_v4_lock);
178 head = &raw_v4_htable[hash];
179 if (hlist_empty(head))
181 sk = __raw_v4_lookup(__sk_head(head), iph->protocol,
182 iph->saddr, iph->daddr,
186 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
187 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
189 /* Not releasing hash table! */
193 sk = __raw_v4_lookup(sk_next(sk), iph->protocol,
194 iph->saddr, iph->daddr,
198 read_unlock(&raw_v4_lock);
201 void raw_err (struct sock *sk, struct sk_buff *skb, u32 info)
203 struct inet_opt *inet = inet_sk(sk);
204 int type = skb->h.icmph->type;
205 int code = skb->h.icmph->code;
209 /* Report error on raw socket, if:
210 1. User requested ip_recverr.
211 2. Socket is connected (otherwise the error indication
212 is useless without ip_recverr and error is hard.
214 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
219 case ICMP_TIME_EXCEEDED:
222 case ICMP_SOURCE_QUENCH:
224 case ICMP_PARAMETERPROB:
228 case ICMP_DEST_UNREACH:
230 if (code > NR_ICMP_UNREACH)
232 err = icmp_err_convert[code].errno;
233 harderr = icmp_err_convert[code].fatal;
234 if (code == ICMP_FRAG_NEEDED) {
235 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
241 struct iphdr *iph = (struct iphdr*)skb->data;
242 u8 *payload = skb->data + (iph->ihl << 2);
246 ip_icmp_error(sk, skb, err, 0, info, payload);
249 if (inet->recverr || harderr) {
251 sk->sk_error_report(sk);
255 static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
257 /* Charge it to the socket. */
259 if (sock_queue_rcv_skb(sk, skb) < 0) {
260 /* FIXME: increment a raw drops counter here */
265 return NET_RX_SUCCESS;
268 int raw_rcv(struct sock *sk, struct sk_buff *skb)
270 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
275 skb_push(skb, skb->data - skb->nh.raw);
277 raw_rcv_skb(sk, skb);
281 static int raw_send_hdrinc(struct sock *sk, void *from, int length,
285 struct inet_opt *inet = inet_sk(sk);
291 if (length > rt->u.dst.dev->mtu) {
292 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
299 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
301 skb = sock_alloc_send_skb(sk, length+hh_len+15,
302 flags&MSG_DONTWAIT, &err);
305 skb_reserve(skb, hh_len);
307 skb->priority = sk->sk_priority;
308 skb->dst = dst_clone(&rt->u.dst);
310 skb->nh.iph = iph = (struct iphdr *)skb_put(skb, length);
312 skb->ip_summed = CHECKSUM_NONE;
314 skb->h.raw = skb->nh.raw;
315 err = memcpy_fromiovecend((void *)iph, from, 0, length);
319 /* We don't modify invalid header */
320 if (length >= sizeof(*iph) && iph->ihl * 4 <= length) {
322 iph->saddr = rt->rt_src;
324 iph->tot_len = htons(length);
326 ip_select_ident(iph, &rt->u.dst, NULL);
328 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
331 if (!vx_check(0, VX_ADMIN) && !capable(CAP_NET_RAW)
332 && (!addr_in_nx_info(sk->sk_nx_info, iph->saddr)))
335 err = NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
338 err = inet->recverr ? net_xmit_errno(err) : 0;
348 IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
352 static void raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
355 u8 __user *type = NULL;
356 u8 __user *code = NULL;
363 for (i = 0; i < msg->msg_iovlen; i++) {
364 iov = &msg->msg_iov[i];
370 /* check if one-byte field is readable or not. */
371 if (iov->iov_base && iov->iov_len < 1)
375 type = iov->iov_base;
376 /* check if code field is readable or not. */
377 if (iov->iov_len > 1)
380 code = iov->iov_base;
383 get_user(fl->fl_icmp_type, type);
384 __get_user(fl->fl_icmp_code, code);
397 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
400 struct inet_opt *inet = inet_sk(sk);
401 struct ipcm_cookie ipc;
402 struct rtable *rt = NULL;
410 if (len < 0 || len > 0xFFFF)
418 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
419 goto out; /* compatibility */
422 * Get and verify the address.
425 if (msg->msg_namelen) {
426 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
428 if (msg->msg_namelen < sizeof(*usin))
430 if (usin->sin_family != AF_INET) {
431 static int complained;
433 printk(KERN_INFO "%s forgot to set AF_INET in "
434 "raw sendmsg. Fix it!\n",
437 if (usin->sin_family)
440 daddr = usin->sin_addr.s_addr;
441 /* ANK: I did not forget to get protocol from port field.
442 * I just do not know, who uses this weirdness.
443 * IP_HDRINCL is much more convenient.
447 if (sk->sk_state != TCP_ESTABLISHED)
452 ipc.addr = inet->saddr;
454 ipc.oif = sk->sk_bound_dev_if;
456 if (msg->msg_controllen) {
457 err = ip_cmsg_send(msg, &ipc);
472 /* Linux does not mangle headers on raw sockets,
473 * so that IP options + IP_HDRINCL is non-sense.
480 daddr = ipc.opt->faddr;
483 tos = RT_TOS(inet->tos) | sk->sk_localroute;
484 if (msg->msg_flags & MSG_DONTROUTE)
487 if (MULTICAST(daddr)) {
489 ipc.oif = inet->mc_index;
491 saddr = inet->mc_addr;
495 struct flowi fl = { .oif = ipc.oif,
500 .proto = inet->hdrincl ? IPPROTO_RAW :
504 raw_probe_proto_opt(&fl, msg);
506 if (sk->sk_nx_info) {
507 err = ip_find_src(sk->sk_nx_info, &rt, &fl);
512 err = ip_route_output_flow(&rt, &fl, sk, !(msg->msg_flags&MSG_DONTWAIT));
518 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
521 if (msg->msg_flags & MSG_CONFIRM)
526 err = raw_send_hdrinc(sk, msg->msg_iov, len,
531 ipc.addr = rt->rt_dst;
533 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
534 &ipc, rt, msg->msg_flags);
536 ip_flush_pending_frames(sk);
537 else if (!(msg->msg_flags & MSG_MORE))
538 err = ip_push_pending_frames(sk);
546 out: return err < 0 ? err : len;
549 dst_confirm(&rt->u.dst);
550 if (!(msg->msg_flags & MSG_PROBE) || len)
551 goto back_from_confirm;
556 static void raw_close(struct sock *sk, long timeout)
559 * Raw sockets may have direct kernel refereneces. Kill them.
561 ip_ra_control(sk, 0, NULL);
563 sk_common_release(sk);
566 /* This gets rid of all the nasties in af_inet. -DaveM */
567 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
569 struct inet_opt *inet = inet_sk(sk);
570 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
574 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
576 chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
577 ret = -EADDRNOTAVAIL;
578 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
579 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
581 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
582 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
583 inet->saddr = 0; /* Use device */
590 * This should be easy, if there is something there
591 * we return it, otherwise we block.
594 int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
595 size_t len, int noblock, int flags, int *addr_len)
597 struct inet_opt *inet = inet_sk(sk);
599 int err = -EOPNOTSUPP;
600 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
607 *addr_len = sizeof(*sin);
609 if (flags & MSG_ERRQUEUE) {
610 err = ip_recv_error(sk, msg, len);
614 skb = skb_recv_datagram(sk, flags, noblock, &err);
620 msg->msg_flags |= MSG_TRUNC;
624 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
628 sock_recv_timestamp(msg, sk, skb);
630 /* Copy the address. */
632 sin->sin_family = AF_INET;
633 sin->sin_addr.s_addr = skb->nh.iph->saddr;
634 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
636 if (inet->cmsg_flags)
637 ip_cmsg_recv(msg, skb);
638 if (flags & MSG_TRUNC)
641 skb_free_datagram(sk, skb);
642 out: return err ? err : copied;
645 static int raw_init(struct sock *sk)
647 struct raw_opt *tp = raw4_sk(sk);
648 if (inet_sk(sk)->num == IPPROTO_ICMP)
649 memset(&tp->filter, 0, sizeof(tp->filter));
653 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
655 if (optlen > sizeof(struct icmp_filter))
656 optlen = sizeof(struct icmp_filter);
657 if (copy_from_user(&raw4_sk(sk)->filter, optval, optlen))
662 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
664 int len, ret = -EFAULT;
666 if (get_user(len, optlen))
671 if (len > sizeof(struct icmp_filter))
672 len = sizeof(struct icmp_filter);
674 if (put_user(len, optlen) ||
675 copy_to_user(optval, &raw4_sk(sk)->filter, len))
681 static int raw_setsockopt(struct sock *sk, int level, int optname,
682 char __user *optval, int optlen)
684 if (level != SOL_RAW)
685 return ip_setsockopt(sk, level, optname, optval, optlen);
687 if (optname == ICMP_FILTER) {
688 if (inet_sk(sk)->num != IPPROTO_ICMP)
691 return raw_seticmpfilter(sk, optval, optlen);
696 static int raw_getsockopt(struct sock *sk, int level, int optname,
697 char __user *optval, int __user *optlen)
699 if (level != SOL_RAW)
700 return ip_getsockopt(sk, level, optname, optval, optlen);
702 if (optname == ICMP_FILTER) {
703 if (inet_sk(sk)->num != IPPROTO_ICMP)
706 return raw_geticmpfilter(sk, optval, optlen);
711 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
715 int amount = atomic_read(&sk->sk_wmem_alloc);
716 return put_user(amount, (int __user *)arg);
722 spin_lock_irq(&sk->sk_receive_queue.lock);
723 skb = skb_peek(&sk->sk_receive_queue);
726 spin_unlock_irq(&sk->sk_receive_queue.lock);
727 return put_user(amount, (int __user *)arg);
731 #ifdef CONFIG_IP_MROUTE
732 return ipmr_ioctl(sk, cmd, (void __user *)arg);
739 struct proto raw_prot = {
742 .connect = ip4_datagram_connect,
743 .disconnect = udp_disconnect,
746 .setsockopt = raw_setsockopt,
747 .getsockopt = raw_getsockopt,
748 .sendmsg = raw_sendmsg,
749 .recvmsg = raw_recvmsg,
751 .backlog_rcv = raw_rcv_skb,
753 .unhash = raw_v4_unhash,
754 .slab_obj_size = sizeof(struct raw_sock),
757 #ifdef CONFIG_PROC_FS
758 struct raw_iter_state {
762 #define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
764 static struct sock *raw_get_first(struct seq_file *seq)
767 struct raw_iter_state* state = raw_seq_private(seq);
769 for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) {
770 struct hlist_node *node;
772 sk_for_each(sk, node, &raw_v4_htable[state->bucket])
773 if (sk->sk_family == PF_INET &&
774 vx_check(sk->sk_xid, VX_WATCH|VX_IDENT))
782 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
784 struct raw_iter_state* state = raw_seq_private(seq);
790 } while (sk && (sk->sk_family != PF_INET ||
791 !vx_check(sk->sk_xid, VX_WATCH|VX_IDENT)));
793 if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) {
794 sk = sk_head(&raw_v4_htable[state->bucket]);
800 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
802 struct sock *sk = raw_get_first(seq);
805 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
807 return pos ? NULL : sk;
810 static void *raw_seq_start(struct seq_file *seq, loff_t *pos)
812 read_lock(&raw_v4_lock);
813 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
816 static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
820 if (v == SEQ_START_TOKEN)
821 sk = raw_get_first(seq);
823 sk = raw_get_next(seq, v);
828 static void raw_seq_stop(struct seq_file *seq, void *v)
830 read_unlock(&raw_v4_lock);
833 static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i)
835 struct inet_opt *inet = inet_sk(sp);
836 unsigned int dest = inet->daddr,
837 src = inet->rcv_saddr;
841 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
842 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
843 i, src, srcp, dest, destp, sp->sk_state,
844 atomic_read(&sp->sk_wmem_alloc),
845 atomic_read(&sp->sk_rmem_alloc),
846 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
847 atomic_read(&sp->sk_refcnt), sp);
851 static int raw_seq_show(struct seq_file *seq, void *v)
855 if (v == SEQ_START_TOKEN)
856 seq_printf(seq, "%-127s\n",
857 " sl local_address rem_address st tx_queue "
858 "rx_queue tr tm->when retrnsmt uid timeout "
861 struct raw_iter_state *state = raw_seq_private(seq);
863 seq_printf(seq, "%-127s\n",
864 get_raw_sock(v, tmpbuf, state->bucket));
869 static struct seq_operations raw_seq_ops = {
870 .start = raw_seq_start,
871 .next = raw_seq_next,
872 .stop = raw_seq_stop,
873 .show = raw_seq_show,
876 static int raw_seq_open(struct inode *inode, struct file *file)
878 struct seq_file *seq;
880 struct raw_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
884 rc = seq_open(file, &raw_seq_ops);
888 seq = file->private_data;
890 memset(s, 0, sizeof(*s));
898 static struct file_operations raw_seq_fops = {
899 .owner = THIS_MODULE,
900 .open = raw_seq_open,
903 .release = seq_release_private,
906 int __init raw_proc_init(void)
908 if (!proc_net_fops_create("raw", S_IRUGO, &raw_seq_fops))
913 void __init raw_proc_exit(void)
915 proc_net_remove("raw");
917 #endif /* CONFIG_PROC_FS */