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/types.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/stddef.h>
49 #include <linux/slab.h>
50 #include <linux/errno.h>
51 #include <linux/aio.h>
52 #include <linux/kernel.h>
53 #include <linux/spinlock.h>
54 #include <linux/sockios.h>
55 #include <linux/socket.h>
57 #include <linux/mroute.h>
58 #include <linux/netdevice.h>
59 #include <linux/in_route.h>
60 #include <linux/route.h>
61 #include <linux/skbuff.h>
64 #include <linux/gfp.h>
66 #include <linux/net.h>
72 #include <net/tcp_states.h>
73 #include <net/inet_common.h>
74 #include <net/checksum.h>
76 #include <linux/rtnetlink.h>
77 #include <linux/proc_fs.h>
78 #include <linux/seq_file.h>
79 #include <linux/netfilter.h>
80 #include <linux/netfilter_ipv4.h>
82 struct hlist_head raw_v4_htable
[RAWV4_HTABLE_SIZE
];
83 DEFINE_RWLOCK(raw_v4_lock
);
85 static void raw_v4_hash(struct sock
*sk
)
87 struct hlist_head
*head
= &raw_v4_htable
[inet_sk(sk
)->num
&
88 (RAWV4_HTABLE_SIZE
- 1)];
90 write_lock_bh(&raw_v4_lock
);
91 sk_add_node(sk
, head
);
92 sock_prot_inc_use(sk
->sk_prot
);
93 write_unlock_bh(&raw_v4_lock
);
96 static void raw_v4_unhash(struct sock
*sk
)
98 write_lock_bh(&raw_v4_lock
);
99 if (sk_del_node_init(sk
))
100 sock_prot_dec_use(sk
->sk_prot
);
101 write_unlock_bh(&raw_v4_lock
);
104 struct sock
*__raw_v4_lookup(struct sock
*sk
, unsigned short num
,
105 __be32 raddr
, __be32 laddr
,
108 struct hlist_node
*node
;
110 sk_for_each_from(sk
, node
) {
111 struct inet_sock
*inet
= inet_sk(sk
);
113 if (inet
->num
== num
&&
114 !(inet
->daddr
&& inet
->daddr
!= raddr
) &&
115 !(inet
->rcv_saddr
&& inet
->rcv_saddr
!= laddr
) &&
116 !(sk
->sk_bound_dev_if
&& sk
->sk_bound_dev_if
!= dif
))
117 goto found
; /* gotcha */
128 static __inline__
int icmp_filter(struct sock
*sk
, struct sk_buff
*skb
)
132 if (!pskb_may_pull(skb
, sizeof(struct icmphdr
)))
135 type
= skb
->h
.icmph
->type
;
137 __u32 data
= raw_sk(sk
)->filter
.data
;
139 return ((1 << type
) & data
) != 0;
142 /* Do not block unknown ICMP types */
146 /* IP input processing comes here for RAW socket delivery.
147 * Caller owns SKB, so we must make clones.
149 * RFC 1122: SHOULD pass TOS value up to the transport layer.
150 * -> It does. And not only TOS, but all IP header.
152 int raw_v4_input(struct sk_buff
*skb
, struct iphdr
*iph
, int hash
)
155 struct hlist_head
*head
;
158 read_lock(&raw_v4_lock
);
159 head
= &raw_v4_htable
[hash
];
160 if (hlist_empty(head
))
162 sk
= __raw_v4_lookup(__sk_head(head
), iph
->protocol
,
163 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
);
184 void raw_err (struct sock
*sk
, struct sk_buff
*skb
, u32 info
)
186 struct inet_sock
*inet
= inet_sk(sk
);
187 int type
= skb
->h
.icmph
->type
;
188 int code
= skb
->h
.icmph
->code
;
192 /* Report error on raw socket, if:
193 1. User requested ip_recverr.
194 2. Socket is connected (otherwise the error indication
195 is useless without ip_recverr and error is hard.
197 if (!inet
->recverr
&& sk
->sk_state
!= TCP_ESTABLISHED
)
202 case ICMP_TIME_EXCEEDED
:
205 case ICMP_SOURCE_QUENCH
:
207 case ICMP_PARAMETERPROB
:
211 case ICMP_DEST_UNREACH
:
213 if (code
> NR_ICMP_UNREACH
)
215 err
= icmp_err_convert
[code
].errno
;
216 harderr
= icmp_err_convert
[code
].fatal
;
217 if (code
== ICMP_FRAG_NEEDED
) {
218 harderr
= inet
->pmtudisc
!= IP_PMTUDISC_DONT
;
224 struct iphdr
*iph
= (struct iphdr
*)skb
->data
;
225 u8
*payload
= skb
->data
+ (iph
->ihl
<< 2);
229 ip_icmp_error(sk
, skb
, err
, 0, info
, payload
);
232 if (inet
->recverr
|| harderr
) {
234 sk
->sk_error_report(sk
);
238 static int raw_rcv_skb(struct sock
* sk
, struct sk_buff
* skb
)
240 /* Charge it to the socket. */
242 if (sock_queue_rcv_skb(sk
, skb
) < 0) {
243 /* FIXME: increment a raw drops counter here */
248 return NET_RX_SUCCESS
;
251 int raw_rcv(struct sock
*sk
, struct sk_buff
*skb
)
253 if (!xfrm4_policy_check(sk
, XFRM_POLICY_IN
, skb
)) {
259 skb_push(skb
, skb
->data
- skb_network_header(skb
));
261 raw_rcv_skb(sk
, skb
);
265 static int raw_send_hdrinc(struct sock
*sk
, void *from
, size_t length
,
269 struct inet_sock
*inet
= inet_sk(sk
);
275 if (length
> rt
->u
.dst
.dev
->mtu
) {
276 ip_local_error(sk
, EMSGSIZE
, rt
->rt_dst
, inet
->dport
,
283 hh_len
= LL_RESERVED_SPACE(rt
->u
.dst
.dev
);
285 skb
= sock_alloc_send_skb(sk
, length
+hh_len
+15,
286 flags
&MSG_DONTWAIT
, &err
);
289 skb_reserve(skb
, hh_len
);
291 skb
->priority
= sk
->sk_priority
;
292 skb
->dst
= dst_clone(&rt
->u
.dst
);
294 skb_reset_network_header(skb
);
296 skb_put(skb
, length
);
298 skb
->ip_summed
= CHECKSUM_NONE
;
300 skb
->h
.raw
= skb
->nh
.raw
;
301 err
= memcpy_fromiovecend((void *)iph
, from
, 0, length
);
305 /* We don't modify invalid header */
306 if (length
>= sizeof(*iph
) && iph
->ihl
* 4U <= length
) {
308 iph
->saddr
= rt
->rt_src
;
310 iph
->tot_len
= htons(length
);
312 ip_select_ident(iph
, &rt
->u
.dst
, NULL
);
314 iph
->check
= ip_fast_csum((unsigned char *)iph
, iph
->ihl
);
317 err
= NF_HOOK(PF_INET
, NF_IP_LOCAL_OUT
, skb
, NULL
, rt
->u
.dst
.dev
,
320 err
= inet
->recverr
? net_xmit_errno(err
) : 0;
330 IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS
);
334 static int raw_probe_proto_opt(struct flowi
*fl
, struct msghdr
*msg
)
337 u8 __user
*type
= NULL
;
338 u8 __user
*code
= NULL
;
345 for (i
= 0; i
< msg
->msg_iovlen
; i
++) {
346 iov
= &msg
->msg_iov
[i
];
352 /* check if one-byte field is readable or not. */
353 if (iov
->iov_base
&& iov
->iov_len
< 1)
357 type
= iov
->iov_base
;
358 /* check if code field is readable or not. */
359 if (iov
->iov_len
> 1)
362 code
= iov
->iov_base
;
365 if (get_user(fl
->fl_icmp_type
, type
) ||
366 get_user(fl
->fl_icmp_code
, code
))
381 static int raw_sendmsg(struct kiocb
*iocb
, struct sock
*sk
, struct msghdr
*msg
,
384 struct inet_sock
*inet
= inet_sk(sk
);
385 struct ipcm_cookie ipc
;
386 struct rtable
*rt
= NULL
;
402 if (msg
->msg_flags
& MSG_OOB
) /* Mirror BSD error message */
403 goto out
; /* compatibility */
406 * Get and verify the address.
409 if (msg
->msg_namelen
) {
410 struct sockaddr_in
*usin
= (struct sockaddr_in
*)msg
->msg_name
;
412 if (msg
->msg_namelen
< sizeof(*usin
))
414 if (usin
->sin_family
!= AF_INET
) {
415 static int complained
;
417 printk(KERN_INFO
"%s forgot to set AF_INET in "
418 "raw sendmsg. Fix it!\n",
421 if (usin
->sin_family
)
424 daddr
= usin
->sin_addr
.s_addr
;
425 /* ANK: I did not forget to get protocol from port field.
426 * I just do not know, who uses this weirdness.
427 * IP_HDRINCL is much more convenient.
431 if (sk
->sk_state
!= TCP_ESTABLISHED
)
436 ipc
.addr
= inet
->saddr
;
438 ipc
.oif
= sk
->sk_bound_dev_if
;
440 if (msg
->msg_controllen
) {
441 err
= ip_cmsg_send(msg
, &ipc
);
456 /* Linux does not mangle headers on raw sockets,
457 * so that IP options + IP_HDRINCL is non-sense.
464 daddr
= ipc
.opt
->faddr
;
467 tos
= RT_CONN_FLAGS(sk
);
468 if (msg
->msg_flags
& MSG_DONTROUTE
)
471 if (MULTICAST(daddr
)) {
473 ipc
.oif
= inet
->mc_index
;
475 saddr
= inet
->mc_addr
;
479 struct flowi fl
= { .oif
= ipc
.oif
,
484 .proto
= inet
->hdrincl
? IPPROTO_RAW
:
487 if (!inet
->hdrincl
) {
488 err
= raw_probe_proto_opt(&fl
, msg
);
493 security_sk_classify_flow(sk
, &fl
);
494 err
= ip_route_output_flow(&rt
, &fl
, sk
, 1);
500 if (rt
->rt_flags
& RTCF_BROADCAST
&& !sock_flag(sk
, SOCK_BROADCAST
))
503 if (msg
->msg_flags
& MSG_CONFIRM
)
508 err
= raw_send_hdrinc(sk
, msg
->msg_iov
, len
,
513 ipc
.addr
= rt
->rt_dst
;
515 err
= ip_append_data(sk
, ip_generic_getfrag
, msg
->msg_iov
, len
, 0,
516 &ipc
, rt
, msg
->msg_flags
);
518 ip_flush_pending_frames(sk
);
519 else if (!(msg
->msg_flags
& MSG_MORE
))
520 err
= ip_push_pending_frames(sk
);
534 dst_confirm(&rt
->u
.dst
);
535 if (!(msg
->msg_flags
& MSG_PROBE
) || len
)
536 goto back_from_confirm
;
541 static void raw_close(struct sock
*sk
, long timeout
)
544 * Raw sockets may have direct kernel refereneces. Kill them.
546 ip_ra_control(sk
, 0, NULL
);
548 sk_common_release(sk
);
551 /* This gets rid of all the nasties in af_inet. -DaveM */
552 static int raw_bind(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
554 struct inet_sock
*inet
= inet_sk(sk
);
555 struct sockaddr_in
*addr
= (struct sockaddr_in
*) uaddr
;
559 if (sk
->sk_state
!= TCP_CLOSE
|| addr_len
< sizeof(struct sockaddr_in
))
561 chk_addr_ret
= inet_addr_type(addr
->sin_addr
.s_addr
);
562 ret
= -EADDRNOTAVAIL
;
563 if (addr
->sin_addr
.s_addr
&& chk_addr_ret
!= RTN_LOCAL
&&
564 chk_addr_ret
!= RTN_MULTICAST
&& chk_addr_ret
!= RTN_BROADCAST
)
566 inet
->rcv_saddr
= inet
->saddr
= addr
->sin_addr
.s_addr
;
567 if (chk_addr_ret
== RTN_MULTICAST
|| chk_addr_ret
== RTN_BROADCAST
)
568 inet
->saddr
= 0; /* Use device */
575 * This should be easy, if there is something there
576 * we return it, otherwise we block.
579 static int raw_recvmsg(struct kiocb
*iocb
, struct sock
*sk
, struct msghdr
*msg
,
580 size_t len
, int noblock
, int flags
, int *addr_len
)
582 struct inet_sock
*inet
= inet_sk(sk
);
584 int err
= -EOPNOTSUPP
;
585 struct sockaddr_in
*sin
= (struct sockaddr_in
*)msg
->msg_name
;
592 *addr_len
= sizeof(*sin
);
594 if (flags
& MSG_ERRQUEUE
) {
595 err
= ip_recv_error(sk
, msg
, len
);
599 skb
= skb_recv_datagram(sk
, flags
, noblock
, &err
);
605 msg
->msg_flags
|= MSG_TRUNC
;
609 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
613 sock_recv_timestamp(msg
, sk
, skb
);
615 /* Copy the address. */
617 sin
->sin_family
= AF_INET
;
618 sin
->sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
620 memset(&sin
->sin_zero
, 0, sizeof(sin
->sin_zero
));
622 if (inet
->cmsg_flags
)
623 ip_cmsg_recv(msg
, skb
);
624 if (flags
& MSG_TRUNC
)
627 skb_free_datagram(sk
, skb
);
634 static int raw_init(struct sock
*sk
)
636 struct raw_sock
*rp
= raw_sk(sk
);
638 if (inet_sk(sk
)->num
== IPPROTO_ICMP
)
639 memset(&rp
->filter
, 0, sizeof(rp
->filter
));
643 static int raw_seticmpfilter(struct sock
*sk
, char __user
*optval
, int optlen
)
645 if (optlen
> sizeof(struct icmp_filter
))
646 optlen
= sizeof(struct icmp_filter
);
647 if (copy_from_user(&raw_sk(sk
)->filter
, optval
, optlen
))
652 static int raw_geticmpfilter(struct sock
*sk
, char __user
*optval
, int __user
*optlen
)
654 int len
, ret
= -EFAULT
;
656 if (get_user(len
, optlen
))
661 if (len
> sizeof(struct icmp_filter
))
662 len
= sizeof(struct icmp_filter
);
664 if (put_user(len
, optlen
) ||
665 copy_to_user(optval
, &raw_sk(sk
)->filter
, len
))
671 static int do_raw_setsockopt(struct sock
*sk
, int level
, int optname
,
672 char __user
*optval
, int optlen
)
674 if (optname
== ICMP_FILTER
) {
675 if (inet_sk(sk
)->num
!= IPPROTO_ICMP
)
678 return raw_seticmpfilter(sk
, optval
, optlen
);
683 static int raw_setsockopt(struct sock
*sk
, int level
, int optname
,
684 char __user
*optval
, int optlen
)
686 if (level
!= SOL_RAW
)
687 return ip_setsockopt(sk
, level
, optname
, optval
, optlen
);
688 return do_raw_setsockopt(sk
, level
, optname
, optval
, optlen
);
692 static int compat_raw_setsockopt(struct sock
*sk
, int level
, int optname
,
693 char __user
*optval
, int optlen
)
695 if (level
!= SOL_RAW
)
696 return compat_ip_setsockopt(sk
, level
, optname
, optval
, optlen
);
697 return do_raw_setsockopt(sk
, level
, optname
, optval
, optlen
);
701 static int do_raw_getsockopt(struct sock
*sk
, int level
, int optname
,
702 char __user
*optval
, int __user
*optlen
)
704 if (optname
== ICMP_FILTER
) {
705 if (inet_sk(sk
)->num
!= IPPROTO_ICMP
)
708 return raw_geticmpfilter(sk
, optval
, optlen
);
713 static int raw_getsockopt(struct sock
*sk
, int level
, int optname
,
714 char __user
*optval
, int __user
*optlen
)
716 if (level
!= SOL_RAW
)
717 return ip_getsockopt(sk
, level
, optname
, optval
, optlen
);
718 return do_raw_getsockopt(sk
, level
, optname
, optval
, optlen
);
722 static int compat_raw_getsockopt(struct sock
*sk
, int level
, int optname
,
723 char __user
*optval
, int __user
*optlen
)
725 if (level
!= SOL_RAW
)
726 return compat_ip_getsockopt(sk
, level
, optname
, optval
, optlen
);
727 return do_raw_getsockopt(sk
, level
, optname
, optval
, optlen
);
731 static int raw_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
735 int amount
= atomic_read(&sk
->sk_wmem_alloc
);
736 return put_user(amount
, (int __user
*)arg
);
742 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
743 skb
= skb_peek(&sk
->sk_receive_queue
);
746 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
747 return put_user(amount
, (int __user
*)arg
);
751 #ifdef CONFIG_IP_MROUTE
752 return ipmr_ioctl(sk
, cmd
, (void __user
*)arg
);
759 struct proto raw_prot
= {
761 .owner
= THIS_MODULE
,
763 .connect
= ip4_datagram_connect
,
764 .disconnect
= udp_disconnect
,
767 .setsockopt
= raw_setsockopt
,
768 .getsockopt
= raw_getsockopt
,
769 .sendmsg
= raw_sendmsg
,
770 .recvmsg
= raw_recvmsg
,
772 .backlog_rcv
= raw_rcv_skb
,
774 .unhash
= raw_v4_unhash
,
775 .obj_size
= sizeof(struct raw_sock
),
777 .compat_setsockopt
= compat_raw_setsockopt
,
778 .compat_getsockopt
= compat_raw_getsockopt
,
782 #ifdef CONFIG_PROC_FS
783 struct raw_iter_state
{
787 #define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
789 static struct sock
*raw_get_first(struct seq_file
*seq
)
792 struct raw_iter_state
* state
= raw_seq_private(seq
);
794 for (state
->bucket
= 0; state
->bucket
< RAWV4_HTABLE_SIZE
; ++state
->bucket
) {
795 struct hlist_node
*node
;
797 sk_for_each(sk
, node
, &raw_v4_htable
[state
->bucket
])
798 if (sk
->sk_family
== PF_INET
)
806 static struct sock
*raw_get_next(struct seq_file
*seq
, struct sock
*sk
)
808 struct raw_iter_state
* state
= raw_seq_private(seq
);
814 } while (sk
&& sk
->sk_family
!= PF_INET
);
816 if (!sk
&& ++state
->bucket
< RAWV4_HTABLE_SIZE
) {
817 sk
= sk_head(&raw_v4_htable
[state
->bucket
]);
823 static struct sock
*raw_get_idx(struct seq_file
*seq
, loff_t pos
)
825 struct sock
*sk
= raw_get_first(seq
);
828 while (pos
&& (sk
= raw_get_next(seq
, sk
)) != NULL
)
830 return pos
? NULL
: sk
;
833 static void *raw_seq_start(struct seq_file
*seq
, loff_t
*pos
)
835 read_lock(&raw_v4_lock
);
836 return *pos
? raw_get_idx(seq
, *pos
- 1) : SEQ_START_TOKEN
;
839 static void *raw_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
843 if (v
== SEQ_START_TOKEN
)
844 sk
= raw_get_first(seq
);
846 sk
= raw_get_next(seq
, v
);
851 static void raw_seq_stop(struct seq_file
*seq
, void *v
)
853 read_unlock(&raw_v4_lock
);
856 static __inline__
char *get_raw_sock(struct sock
*sp
, char *tmpbuf
, int i
)
858 struct inet_sock
*inet
= inet_sk(sp
);
859 __be32 dest
= inet
->daddr
,
860 src
= inet
->rcv_saddr
;
864 sprintf(tmpbuf
, "%4d: %08X:%04X %08X:%04X"
865 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
866 i
, src
, srcp
, dest
, destp
, sp
->sk_state
,
867 atomic_read(&sp
->sk_wmem_alloc
),
868 atomic_read(&sp
->sk_rmem_alloc
),
869 0, 0L, 0, sock_i_uid(sp
), 0, sock_i_ino(sp
),
870 atomic_read(&sp
->sk_refcnt
), sp
);
874 static int raw_seq_show(struct seq_file
*seq
, void *v
)
878 if (v
== SEQ_START_TOKEN
)
879 seq_printf(seq
, "%-127s\n",
880 " sl local_address rem_address st tx_queue "
881 "rx_queue tr tm->when retrnsmt uid timeout "
884 struct raw_iter_state
*state
= raw_seq_private(seq
);
886 seq_printf(seq
, "%-127s\n",
887 get_raw_sock(v
, tmpbuf
, state
->bucket
));
892 static const struct seq_operations raw_seq_ops
= {
893 .start
= raw_seq_start
,
894 .next
= raw_seq_next
,
895 .stop
= raw_seq_stop
,
896 .show
= raw_seq_show
,
899 static int raw_seq_open(struct inode
*inode
, struct file
*file
)
901 struct seq_file
*seq
;
903 struct raw_iter_state
*s
= kmalloc(sizeof(*s
), GFP_KERNEL
);
907 rc
= seq_open(file
, &raw_seq_ops
);
911 seq
= file
->private_data
;
913 memset(s
, 0, sizeof(*s
));
921 static const struct file_operations raw_seq_fops
= {
922 .owner
= THIS_MODULE
,
923 .open
= raw_seq_open
,
926 .release
= seq_release_private
,
929 int __init
raw_proc_init(void)
931 if (!proc_net_fops_create("raw", S_IRUGO
, &raw_seq_fops
))
936 void __init
raw_proc_exit(void)
938 proc_net_remove("raw");
940 #endif /* CONFIG_PROC_FS */