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
;
338 /* This check is ONLY to check for arithmetic overflow
339 on integer(!) len. Not more! Real check will be made
340 in ip_build_xmit --ANK
342 BTW socket.c -> af_*.c -> ... make multiple
343 invalid conversions size_t -> int. We MUST repair it f.e.
344 by replacing all of them with size_t and revise all
345 the places sort of len += sizeof(struct iphdr)
346 If len was ULONG_MAX-10 it would be cathastrophe --ANK
350 if (len
< 0 || len
> 0xFFFF)
358 if (msg
->msg_flags
& MSG_OOB
) /* Mirror BSD error message */
359 goto out
; /* compatibility */
362 * Get and verify the address.
365 if (msg
->msg_namelen
) {
366 struct sockaddr_in
*usin
= (struct sockaddr_in
*)msg
->msg_name
;
368 if (msg
->msg_namelen
< sizeof(*usin
))
370 if (usin
->sin_family
!= AF_INET
) {
371 static int complained
;
373 printk(KERN_INFO
"%s forgot to set AF_INET in "
374 "raw sendmsg. Fix it!\n",
377 if (usin
->sin_family
)
380 daddr
= usin
->sin_addr
.s_addr
;
381 /* ANK: I did not forget to get protocol from port field.
382 * I just do not know, who uses this weirdness.
383 * IP_HDRINCL is much more convenient.
387 if (sk
->sk_state
!= TCP_ESTABLISHED
)
392 ipc
.addr
= inet
->saddr
;
394 ipc
.oif
= sk
->sk_bound_dev_if
;
396 if (msg
->msg_controllen
) {
397 err
= ip_cmsg_send(msg
, &ipc
);
412 /* Linux does not mangle headers on raw sockets,
413 * so that IP options + IP_HDRINCL is non-sense.
420 daddr
= ipc
.opt
->faddr
;
423 tos
= RT_TOS(inet
->tos
) | sk
->sk_localroute
;
424 if (msg
->msg_flags
& MSG_DONTROUTE
)
427 if (MULTICAST(daddr
)) {
429 ipc
.oif
= inet
->mc_index
;
431 saddr
= inet
->mc_addr
;
435 struct flowi fl
= { .oif
= ipc
.oif
,
440 .proto
= inet
->hdrincl
? IPPROTO_RAW
:
443 err
= ip_route_output_flow(&rt
, &fl
, sk
, !(msg
->msg_flags
&MSG_DONTWAIT
));
449 if (rt
->rt_flags
& RTCF_BROADCAST
&& !sock_flag(sk
, SOCK_BROADCAST
))
452 if (msg
->msg_flags
& MSG_CONFIRM
)
457 err
= raw_send_hdrinc(sk
, msg
->msg_iov
, len
,
462 ipc
.addr
= rt
->rt_dst
;
464 err
= ip_append_data(sk
, ip_generic_getfrag
, msg
->msg_iov
, len
, 0,
465 &ipc
, rt
, msg
->msg_flags
);
467 ip_flush_pending_frames(sk
);
468 else if (!(msg
->msg_flags
& MSG_MORE
))
469 err
= ip_push_pending_frames(sk
);
477 out
: return err
< 0 ? err
: len
;
480 dst_confirm(&rt
->u
.dst
);
481 if (!(msg
->msg_flags
& MSG_PROBE
) || len
)
482 goto back_from_confirm
;
487 static void raw_close(struct sock
*sk
, long timeout
)
490 * Raw sockets may have direct kernel refereneces. Kill them.
492 ip_ra_control(sk
, 0, NULL
);
494 inet_sock_release(sk
);
497 /* This gets rid of all the nasties in af_inet. -DaveM */
498 static int raw_bind(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
500 struct inet_opt
*inet
= inet_sk(sk
);
501 struct sockaddr_in
*addr
= (struct sockaddr_in
*) uaddr
;
505 if (sk
->sk_state
!= TCP_CLOSE
|| addr_len
< sizeof(struct sockaddr_in
))
507 chk_addr_ret
= inet_addr_type(addr
->sin_addr
.s_addr
);
508 ret
= -EADDRNOTAVAIL
;
509 if (addr
->sin_addr
.s_addr
&& chk_addr_ret
!= RTN_LOCAL
&&
510 chk_addr_ret
!= RTN_MULTICAST
&& chk_addr_ret
!= RTN_BROADCAST
)
512 inet
->rcv_saddr
= inet
->saddr
= addr
->sin_addr
.s_addr
;
513 if (chk_addr_ret
== RTN_MULTICAST
|| chk_addr_ret
== RTN_BROADCAST
)
514 inet
->saddr
= 0; /* Use device */
521 * This should be easy, if there is something there
522 * we return it, otherwise we block.
525 int raw_recvmsg(struct kiocb
*iocb
, struct sock
*sk
, struct msghdr
*msg
,
526 int len
, int noblock
, int flags
, int *addr_len
)
528 struct inet_opt
*inet
= inet_sk(sk
);
530 int err
= -EOPNOTSUPP
;
531 struct sockaddr_in
*sin
= (struct sockaddr_in
*)msg
->msg_name
;
538 *addr_len
= sizeof(*sin
);
540 if (flags
& MSG_ERRQUEUE
) {
541 err
= ip_recv_error(sk
, msg
, len
);
545 skb
= skb_recv_datagram(sk
, flags
, noblock
, &err
);
551 msg
->msg_flags
|= MSG_TRUNC
;
555 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
559 sock_recv_timestamp(msg
, sk
, skb
);
561 /* Copy the address. */
563 sin
->sin_family
= AF_INET
;
564 sin
->sin_addr
.s_addr
= skb
->nh
.iph
->saddr
;
565 memset(&sin
->sin_zero
, 0, sizeof(sin
->sin_zero
));
567 if (inet
->cmsg_flags
)
568 ip_cmsg_recv(msg
, skb
);
570 skb_free_datagram(sk
, skb
);
571 out
: return err
? : copied
;
574 static int raw_init(struct sock
*sk
)
576 struct raw_opt
*tp
= raw4_sk(sk
);
577 if (inet_sk(sk
)->num
== IPPROTO_ICMP
)
578 memset(&tp
->filter
, 0, sizeof(tp
->filter
));
582 static int raw_seticmpfilter(struct sock
*sk
, char *optval
, int optlen
)
584 if (optlen
> sizeof(struct icmp_filter
))
585 optlen
= sizeof(struct icmp_filter
);
586 if (copy_from_user(&raw4_sk(sk
)->filter
, optval
, optlen
))
591 static int raw_geticmpfilter(struct sock
*sk
, char *optval
, int *optlen
)
593 int len
, ret
= -EFAULT
;
595 if (get_user(len
, optlen
))
600 if (len
> sizeof(struct icmp_filter
))
601 len
= sizeof(struct icmp_filter
);
603 if (put_user(len
, optlen
) ||
604 copy_to_user(optval
, &raw4_sk(sk
)->filter
, len
))
610 static int raw_setsockopt(struct sock
*sk
, int level
, int optname
,
611 char *optval
, int optlen
)
613 if (level
!= SOL_RAW
)
614 return ip_setsockopt(sk
, level
, optname
, optval
, optlen
);
616 if (optname
== ICMP_FILTER
) {
617 if (inet_sk(sk
)->num
!= IPPROTO_ICMP
)
620 return raw_seticmpfilter(sk
, optval
, optlen
);
625 static int raw_getsockopt(struct sock
*sk
, int level
, int optname
,
626 char *optval
, int *optlen
)
628 if (level
!= SOL_RAW
)
629 return ip_getsockopt(sk
, level
, optname
, optval
, optlen
);
631 if (optname
== ICMP_FILTER
) {
632 if (inet_sk(sk
)->num
!= IPPROTO_ICMP
)
635 return raw_geticmpfilter(sk
, optval
, optlen
);
640 static int raw_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
644 int amount
= atomic_read(&sk
->sk_wmem_alloc
);
645 return put_user(amount
, (int *)arg
);
651 spin_lock_irq(&sk
->sk_receive_queue
.lock
);
652 skb
= skb_peek(&sk
->sk_receive_queue
);
655 spin_unlock_irq(&sk
->sk_receive_queue
.lock
);
656 return put_user(amount
, (int *)arg
);
660 #ifdef CONFIG_IP_MROUTE
661 return ipmr_ioctl(sk
, cmd
, arg
);
668 struct proto raw_prot
= {
671 .connect
= udp_connect
,
672 .disconnect
= udp_disconnect
,
675 .setsockopt
= raw_setsockopt
,
676 .getsockopt
= raw_getsockopt
,
677 .sendmsg
= raw_sendmsg
,
678 .recvmsg
= raw_recvmsg
,
680 .backlog_rcv
= raw_rcv_skb
,
682 .unhash
= raw_v4_unhash
,
685 #ifdef CONFIG_PROC_FS
686 struct raw_iter_state
{
690 #define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
692 static struct sock
*raw_get_first(struct seq_file
*seq
)
695 struct raw_iter_state
* state
= raw_seq_private(seq
);
697 for (state
->bucket
= 0; state
->bucket
< RAWV4_HTABLE_SIZE
; ++state
->bucket
) {
698 struct hlist_node
*node
;
700 sk_for_each(sk
, node
, &raw_v4_htable
[state
->bucket
])
701 if (sk
->sk_family
== PF_INET
)
709 static struct sock
*raw_get_next(struct seq_file
*seq
, struct sock
*sk
)
711 struct raw_iter_state
* state
= raw_seq_private(seq
);
717 } while (sk
&& sk
->sk_family
!= PF_INET
);
719 if (!sk
&& ++state
->bucket
< RAWV4_HTABLE_SIZE
) {
720 sk
= sk_head(&raw_v4_htable
[state
->bucket
]);
726 static struct sock
*raw_get_idx(struct seq_file
*seq
, loff_t pos
)
728 struct sock
*sk
= raw_get_first(seq
);
731 while (pos
&& (sk
= raw_get_next(seq
, sk
)) != NULL
)
733 return pos
? NULL
: sk
;
736 static void *raw_seq_start(struct seq_file
*seq
, loff_t
*pos
)
738 read_lock(&raw_v4_lock
);
739 return *pos
? raw_get_idx(seq
, *pos
) : (void *)1;
742 static void *raw_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
747 sk
= raw_get_first(seq
);
749 sk
= raw_get_next(seq
, v
);
754 static void raw_seq_stop(struct seq_file
*seq
, void *v
)
756 read_unlock(&raw_v4_lock
);
759 static __inline__
char *get_raw_sock(struct sock
*sp
, char *tmpbuf
, int i
)
761 struct inet_opt
*inet
= inet_sk(sp
);
762 unsigned int dest
= inet
->daddr
,
763 src
= inet
->rcv_saddr
;
767 sprintf(tmpbuf
, "%4d: %08X:%04X %08X:%04X"
768 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
769 i
, src
, srcp
, dest
, destp
, sp
->sk_state
,
770 atomic_read(&sp
->sk_wmem_alloc
),
771 atomic_read(&sp
->sk_rmem_alloc
),
772 0, 0L, 0, sock_i_uid(sp
), 0, sock_i_ino(sp
),
773 atomic_read(&sp
->sk_refcnt
), sp
);
777 static int raw_seq_show(struct seq_file
*seq
, void *v
)
782 seq_printf(seq
, "%-127s\n",
783 " sl local_address rem_address st tx_queue "
784 "rx_queue tr tm->when retrnsmt uid timeout "
787 struct raw_iter_state
*state
= raw_seq_private(seq
);
789 seq_printf(seq
, "%-127s\n",
790 get_raw_sock(v
, tmpbuf
, state
->bucket
));
795 static struct seq_operations raw_seq_ops
= {
796 .start
= raw_seq_start
,
797 .next
= raw_seq_next
,
798 .stop
= raw_seq_stop
,
799 .show
= raw_seq_show
,
802 static int raw_seq_open(struct inode
*inode
, struct file
*file
)
804 struct seq_file
*seq
;
806 struct raw_iter_state
*s
= kmalloc(sizeof(*s
), GFP_KERNEL
);
810 rc
= seq_open(file
, &raw_seq_ops
);
814 seq
= file
->private_data
;
816 memset(s
, 0, sizeof(*s
));
824 static struct file_operations raw_seq_fops
= {
825 .owner
= THIS_MODULE
,
826 .open
= raw_seq_open
,
829 .release
= seq_release_private
,
832 int __init
raw_proc_init(void)
834 struct proc_dir_entry
*p
;
837 p
= create_proc_entry("raw", S_IRUGO
, proc_net
);
839 p
->proc_fops
= &raw_seq_fops
;
845 void __init
raw_proc_exit(void)
847 remove_proc_entry("raw", proc_net
);
849 #endif /* CONFIG_PROC_FS */