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>
82 #include <linux/vs_base.h>
84 struct hlist_head raw_v4_htable[RAWV4_HTABLE_SIZE];
85 rwlock_t raw_v4_lock = RW_LOCK_UNLOCKED;
87 static void raw_v4_hash(struct sock *sk)
89 struct hlist_head *head = &raw_v4_htable[inet_sk(sk)->num &
90 (RAWV4_HTABLE_SIZE - 1)];
92 write_lock_bh(&raw_v4_lock);
93 sk_add_node(sk, head);
94 sock_prot_inc_use(sk->sk_prot);
95 write_unlock_bh(&raw_v4_lock);
98 static void raw_v4_unhash(struct sock *sk)
100 write_lock_bh(&raw_v4_lock);
101 if (sk_del_node_init(sk))
102 sock_prot_dec_use(sk->sk_prot);
103 write_unlock_bh(&raw_v4_lock);
106 struct sock *__raw_v4_lookup(struct sock *sk, unsigned short num,
107 unsigned long raddr, unsigned long laddr,
110 struct hlist_node *node;
112 sk_for_each_from(sk, node) {
113 struct inet_opt *inet = inet_sk(sk);
115 if (inet->num == num &&
116 !(inet->daddr && inet->daddr != raddr) &&
117 !(inet->rcv_saddr && inet->rcv_saddr != laddr) &&
118 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
119 goto found; /* gotcha */
130 static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
134 if (!pskb_may_pull(skb, sizeof(struct icmphdr)))
137 type = skb->h.icmph->type;
139 __u32 data = raw4_sk(sk)->filter.data;
141 return ((1 << type) & data) != 0;
144 /* Do not block unknown ICMP types */
148 /* IP input processing comes here for RAW socket delivery.
149 * Caller owns SKB, so we must make clones.
151 * RFC 1122: SHOULD pass TOS value up to the transport layer.
152 * -> It does. And not only TOS, but all IP header.
154 void raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
157 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,
168 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
169 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
171 /* Not releasing hash table! */
175 sk = __raw_v4_lookup(sk_next(sk), iph->protocol,
176 iph->saddr, iph->daddr,
180 read_unlock(&raw_v4_lock);
183 void raw_err (struct sock *sk, struct sk_buff *skb, u32 info)
185 struct inet_opt *inet = inet_sk(sk);
186 int type = skb->h.icmph->type;
187 int code = skb->h.icmph->code;
191 /* Report error on raw socket, if:
192 1. User requested ip_recverr.
193 2. Socket is connected (otherwise the error indication
194 is useless without ip_recverr and error is hard.
196 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
201 case ICMP_TIME_EXCEEDED:
204 case ICMP_SOURCE_QUENCH:
206 case ICMP_PARAMETERPROB:
210 case ICMP_DEST_UNREACH:
212 if (code > NR_ICMP_UNREACH)
214 err = icmp_err_convert[code].errno;
215 harderr = icmp_err_convert[code].fatal;
216 if (code == ICMP_FRAG_NEEDED) {
217 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
223 struct iphdr *iph = (struct iphdr*)skb->data;
224 u8 *payload = skb->data + (iph->ihl << 2);
228 ip_icmp_error(sk, skb, err, 0, info, payload);
231 if (inet->recverr || harderr) {
233 sk->sk_error_report(sk);
237 static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
239 /* Charge it to the socket. */
241 if (sock_queue_rcv_skb(sk, skb) < 0) {
242 /* FIXME: increment a raw drops counter here */
247 return NET_RX_SUCCESS;
250 int raw_rcv(struct sock *sk, struct sk_buff *skb)
252 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
257 skb_push(skb, skb->data - skb->nh.raw);
259 raw_rcv_skb(sk, skb);
263 static int raw_send_hdrinc(struct sock *sk, void *from, int length,
267 struct inet_opt *inet = inet_sk(sk);
273 if (length > rt->u.dst.dev->mtu) {
274 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
281 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
283 skb = sock_alloc_send_skb(sk, length+hh_len+15,
284 flags&MSG_DONTWAIT, &err);
287 skb_reserve(skb, hh_len);
289 skb->priority = sk->sk_priority;
290 skb->dst = dst_clone(&rt->u.dst);
292 skb->nh.iph = iph = (struct iphdr *)skb_put(skb, length);
294 skb->ip_summed = CHECKSUM_NONE;
296 skb->h.raw = skb->nh.raw;
297 err = memcpy_fromiovecend((void *)iph, from, 0, length);
301 /* We don't modify invalid header */
302 if (length >= sizeof(*iph) && iph->ihl * 4 <= length) {
304 iph->saddr = rt->rt_src;
306 iph->tot_len = htons(length);
308 ip_select_ident(iph, &rt->u.dst, NULL);
310 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
313 err = NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
316 err = inet->recverr ? net_xmit_errno(err) : 0;
326 IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
330 static void raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
333 u8 __user *type = NULL;
334 u8 __user *code = NULL;
341 for (i = 0; i < msg->msg_iovlen; i++) {
342 iov = &msg->msg_iov[i];
348 /* check if one-byte field is readable or not. */
349 if (iov->iov_base && iov->iov_len < 1)
353 type = iov->iov_base;
354 /* check if code field is readable or not. */
355 if (iov->iov_len > 1)
358 code = iov->iov_base;
361 get_user(fl->fl_icmp_type, type);
362 __get_user(fl->fl_icmp_code, code);
375 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
378 struct inet_opt *inet = inet_sk(sk);
379 struct ipcm_cookie ipc;
380 struct rtable *rt = NULL;
388 if (len < 0 || len > 0xFFFF)
396 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
397 goto out; /* compatibility */
400 * Get and verify the address.
403 if (msg->msg_namelen) {
404 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
406 if (msg->msg_namelen < sizeof(*usin))
408 if (usin->sin_family != AF_INET) {
409 static int complained;
411 printk(KERN_INFO "%s forgot to set AF_INET in "
412 "raw sendmsg. Fix it!\n",
415 if (usin->sin_family)
418 daddr = usin->sin_addr.s_addr;
419 /* ANK: I did not forget to get protocol from port field.
420 * I just do not know, who uses this weirdness.
421 * IP_HDRINCL is much more convenient.
425 if (sk->sk_state != TCP_ESTABLISHED)
430 ipc.addr = inet->saddr;
432 ipc.oif = sk->sk_bound_dev_if;
434 if (msg->msg_controllen) {
435 err = ip_cmsg_send(msg, &ipc);
450 /* Linux does not mangle headers on raw sockets,
451 * so that IP options + IP_HDRINCL is non-sense.
458 daddr = ipc.opt->faddr;
461 tos = RT_TOS(inet->tos) | sk->sk_localroute;
462 if (msg->msg_flags & MSG_DONTROUTE)
465 if (MULTICAST(daddr)) {
467 ipc.oif = inet->mc_index;
469 saddr = inet->mc_addr;
473 struct flowi fl = { .oif = ipc.oif,
478 .proto = inet->hdrincl ? IPPROTO_RAW :
482 raw_probe_proto_opt(&fl, msg);
484 err = ip_route_output_flow(&rt, &fl, sk, !(msg->msg_flags&MSG_DONTWAIT));
490 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
493 if (msg->msg_flags & MSG_CONFIRM)
498 err = raw_send_hdrinc(sk, msg->msg_iov, len,
503 ipc.addr = rt->rt_dst;
505 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
506 &ipc, rt, msg->msg_flags);
508 ip_flush_pending_frames(sk);
509 else if (!(msg->msg_flags & MSG_MORE))
510 err = ip_push_pending_frames(sk);
518 out: return err < 0 ? err : len;
521 dst_confirm(&rt->u.dst);
522 if (!(msg->msg_flags & MSG_PROBE) || len)
523 goto back_from_confirm;
528 static void raw_close(struct sock *sk, long timeout)
531 * Raw sockets may have direct kernel refereneces. Kill them.
533 ip_ra_control(sk, 0, NULL);
535 sk_common_release(sk);
538 /* This gets rid of all the nasties in af_inet. -DaveM */
539 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
541 struct inet_opt *inet = inet_sk(sk);
542 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
546 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
548 chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
549 ret = -EADDRNOTAVAIL;
550 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
551 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
553 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
554 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
555 inet->saddr = 0; /* Use device */
562 * This should be easy, if there is something there
563 * we return it, otherwise we block.
566 int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
567 size_t len, int noblock, int flags, int *addr_len)
569 struct inet_opt *inet = inet_sk(sk);
571 int err = -EOPNOTSUPP;
572 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
579 *addr_len = sizeof(*sin);
581 if (flags & MSG_ERRQUEUE) {
582 err = ip_recv_error(sk, msg, len);
586 skb = skb_recv_datagram(sk, flags, noblock, &err);
592 msg->msg_flags |= MSG_TRUNC;
596 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
600 sock_recv_timestamp(msg, sk, skb);
602 /* Copy the address. */
604 sin->sin_family = AF_INET;
605 sin->sin_addr.s_addr = skb->nh.iph->saddr;
606 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
608 if (inet->cmsg_flags)
609 ip_cmsg_recv(msg, skb);
610 if (flags & MSG_TRUNC)
613 skb_free_datagram(sk, skb);
614 out: return err ? err : copied;
617 static int raw_init(struct sock *sk)
619 struct raw_opt *tp = raw4_sk(sk);
620 if (inet_sk(sk)->num == IPPROTO_ICMP)
621 memset(&tp->filter, 0, sizeof(tp->filter));
625 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
627 if (optlen > sizeof(struct icmp_filter))
628 optlen = sizeof(struct icmp_filter);
629 if (copy_from_user(&raw4_sk(sk)->filter, optval, optlen))
634 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
636 int len, ret = -EFAULT;
638 if (get_user(len, optlen))
643 if (len > sizeof(struct icmp_filter))
644 len = sizeof(struct icmp_filter);
646 if (put_user(len, optlen) ||
647 copy_to_user(optval, &raw4_sk(sk)->filter, len))
653 static int raw_setsockopt(struct sock *sk, int level, int optname,
654 char __user *optval, int optlen)
656 if (level != SOL_RAW)
657 return ip_setsockopt(sk, level, optname, optval, optlen);
659 if (optname == ICMP_FILTER) {
660 if (inet_sk(sk)->num != IPPROTO_ICMP)
663 return raw_seticmpfilter(sk, optval, optlen);
668 static int raw_getsockopt(struct sock *sk, int level, int optname,
669 char __user *optval, int __user *optlen)
671 if (level != SOL_RAW)
672 return ip_getsockopt(sk, level, optname, optval, optlen);
674 if (optname == ICMP_FILTER) {
675 if (inet_sk(sk)->num != IPPROTO_ICMP)
678 return raw_geticmpfilter(sk, optval, optlen);
683 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
687 int amount = atomic_read(&sk->sk_wmem_alloc);
688 return put_user(amount, (int __user *)arg);
694 spin_lock_irq(&sk->sk_receive_queue.lock);
695 skb = skb_peek(&sk->sk_receive_queue);
698 spin_unlock_irq(&sk->sk_receive_queue.lock);
699 return put_user(amount, (int __user *)arg);
703 #ifdef CONFIG_IP_MROUTE
704 return ipmr_ioctl(sk, cmd, (void __user *)arg);
711 struct proto raw_prot = {
713 .owner = THIS_MODULE,
715 .connect = ip4_datagram_connect,
716 .disconnect = udp_disconnect,
719 .setsockopt = raw_setsockopt,
720 .getsockopt = raw_getsockopt,
721 .sendmsg = raw_sendmsg,
722 .recvmsg = raw_recvmsg,
724 .backlog_rcv = raw_rcv_skb,
726 .unhash = raw_v4_unhash,
727 .slab_obj_size = sizeof(struct raw_sock),
730 #ifdef CONFIG_PROC_FS
731 struct raw_iter_state {
735 #define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
737 static struct sock *raw_get_first(struct seq_file *seq)
740 struct raw_iter_state* state = raw_seq_private(seq);
742 for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) {
743 struct hlist_node *node;
745 sk_for_each(sk, node, &raw_v4_htable[state->bucket])
746 if (sk->sk_family == PF_INET &&
747 vx_check(sk->sk_xid, VX_WATCH|VX_IDENT))
755 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
757 struct raw_iter_state* state = raw_seq_private(seq);
763 } while (sk && (sk->sk_family != PF_INET ||
764 !vx_check(sk->sk_xid, VX_WATCH|VX_IDENT)));
766 if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) {
767 sk = sk_head(&raw_v4_htable[state->bucket]);
773 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
775 struct sock *sk = raw_get_first(seq);
778 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
780 return pos ? NULL : sk;
783 static void *raw_seq_start(struct seq_file *seq, loff_t *pos)
785 read_lock(&raw_v4_lock);
786 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
789 static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
793 if (v == SEQ_START_TOKEN)
794 sk = raw_get_first(seq);
796 sk = raw_get_next(seq, v);
801 static void raw_seq_stop(struct seq_file *seq, void *v)
803 read_unlock(&raw_v4_lock);
806 static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i)
808 struct inet_opt *inet = inet_sk(sp);
809 unsigned int dest = inet->daddr,
810 src = inet->rcv_saddr;
814 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
815 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
816 i, src, srcp, dest, destp, sp->sk_state,
817 atomic_read(&sp->sk_wmem_alloc),
818 atomic_read(&sp->sk_rmem_alloc),
819 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
820 atomic_read(&sp->sk_refcnt), sp);
824 static int raw_seq_show(struct seq_file *seq, void *v)
828 if (v == SEQ_START_TOKEN)
829 seq_printf(seq, "%-127s\n",
830 " sl local_address rem_address st tx_queue "
831 "rx_queue tr tm->when retrnsmt uid timeout "
834 struct raw_iter_state *state = raw_seq_private(seq);
836 seq_printf(seq, "%-127s\n",
837 get_raw_sock(v, tmpbuf, state->bucket));
842 static struct seq_operations raw_seq_ops = {
843 .start = raw_seq_start,
844 .next = raw_seq_next,
845 .stop = raw_seq_stop,
846 .show = raw_seq_show,
849 static int raw_seq_open(struct inode *inode, struct file *file)
851 struct seq_file *seq;
853 struct raw_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
857 rc = seq_open(file, &raw_seq_ops);
861 seq = file->private_data;
863 memset(s, 0, sizeof(*s));
871 static struct file_operations raw_seq_fops = {
872 .owner = THIS_MODULE,
873 .open = raw_seq_open,
876 .release = seq_release_private,
879 int __init raw_proc_init(void)
881 if (!proc_net_fops_create("raw", S_IRUGO, &raw_seq_fops))
886 void __init raw_proc_exit(void)
888 proc_net_remove("raw");
890 #endif /* CONFIG_PROC_FS */