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.
9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Alan Cox : verify_area() fixed up
13 * Alan Cox : ICMP error handling
14 * Alan Cox : EMSGSIZE if you send too big a packet
15 * Alan Cox : Now uses generic datagrams and shared
16 * skbuff library. No more peek crashes,
18 * Alan Cox : Checks sk->broadcast.
19 * Alan Cox : Uses skb_free_datagram/skb_copy_datagram
20 * Alan Cox : Raw passes ip options too
21 * Alan Cox : Setsocketopt added
22 * Alan Cox : Fixed error return for broadcasts
23 * Alan Cox : Removed wake_up calls
24 * Alan Cox : Use ttl/tos
25 * Alan Cox : Cleaned up old debugging
26 * Alan Cox : Use new kernel side addresses
27 * Arnt Gulbrandsen : Fixed MSG_DONTROUTE in raw sockets.
28 * Alan Cox : BSD style RAW socket demultiplexing.
29 * Alan Cox : Beginnings of mrouted support.
30 * Alan Cox : Added IP_HDRINCL option.
31 * Alan Cox : Skip broadcast check if BSDism set.
32 * David S. Miller : New socket lookup architecture.
34 * This program is free software; you can redistribute it and/or
35 * modify it under the terms of the GNU General Public License
36 * as published by the Free Software Foundation; either version
37 * 2 of the License, or (at your option) any later version.
40 #include <linux/types.h>
41 #include <linux/atomic.h>
42 #include <asm/byteorder.h>
43 #include <asm/current.h>
44 #include <linux/uaccess.h>
45 #include <asm/ioctls.h>
46 #include <linux/stddef.h>
47 #include <linux/slab.h>
48 #include <linux/errno.h>
49 #include <linux/kernel.h>
50 #include <linux/export.h>
51 #include <linux/spinlock.h>
52 #include <linux/sockios.h>
53 #include <linux/socket.h>
55 #include <linux/mroute.h>
56 #include <linux/netdevice.h>
57 #include <linux/in_route.h>
58 #include <linux/route.h>
59 #include <linux/skbuff.h>
60 #include <linux/igmp.h>
61 #include <net/net_namespace.h>
65 #include <linux/net.h>
71 #include <net/tcp_states.h>
72 #include <net/inet_common.h>
73 #include <net/checksum.h>
75 #include <linux/rtnetlink.h>
76 #include <linux/proc_fs.h>
77 #include <linux/seq_file.h>
78 #include <linux/netfilter.h>
79 #include <linux/netfilter_ipv4.h>
80 #include <linux/compat.h>
81 #include <linux/uio.h>
92 struct raw_hashinfo raw_v4_hashinfo
= {
93 .lock
= __RW_LOCK_UNLOCKED(raw_v4_hashinfo
.lock
),
95 EXPORT_SYMBOL_GPL(raw_v4_hashinfo
);
97 int raw_hash_sk(struct sock
*sk
)
99 struct raw_hashinfo
*h
= sk
->sk_prot
->h
.raw_hash
;
100 struct hlist_head
*head
;
102 head
= &h
->ht
[inet_sk(sk
)->inet_num
& (RAW_HTABLE_SIZE
- 1)];
104 write_lock_bh(&h
->lock
);
105 sk_add_node(sk
, head
);
106 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, 1);
107 write_unlock_bh(&h
->lock
);
111 EXPORT_SYMBOL_GPL(raw_hash_sk
);
113 void raw_unhash_sk(struct sock
*sk
)
115 struct raw_hashinfo
*h
= sk
->sk_prot
->h
.raw_hash
;
117 write_lock_bh(&h
->lock
);
118 if (sk_del_node_init(sk
))
119 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, -1);
120 write_unlock_bh(&h
->lock
);
122 EXPORT_SYMBOL_GPL(raw_unhash_sk
);
124 struct sock
*__raw_v4_lookup(struct net
*net
, struct sock
*sk
,
125 unsigned short num
, __be32 raddr
, __be32 laddr
,
128 sk_for_each_from(sk
) {
129 struct inet_sock
*inet
= inet_sk(sk
);
131 if (net_eq(sock_net(sk
), net
) && inet
->inet_num
== num
&&
132 !(inet
->inet_daddr
&& inet
->inet_daddr
!= raddr
) &&
133 !(inet
->inet_rcv_saddr
&& inet
->inet_rcv_saddr
!= laddr
) &&
134 raw_sk_bound_dev_eq(net
, sk
->sk_bound_dev_if
, dif
, sdif
))
135 goto found
; /* gotcha */
141 EXPORT_SYMBOL_GPL(__raw_v4_lookup
);
147 static int icmp_filter(const struct sock
*sk
, const struct sk_buff
*skb
)
150 const struct icmphdr
*hdr
;
152 hdr
= skb_header_pointer(skb
, skb_transport_offset(skb
),
153 sizeof(_hdr
), &_hdr
);
157 if (hdr
->type
< 32) {
158 __u32 data
= raw_sk(sk
)->filter
.data
;
160 return ((1U << hdr
->type
) & data
) != 0;
163 /* Do not block unknown ICMP types */
167 /* IP input processing comes here for RAW socket delivery.
168 * Caller owns SKB, so we must make clones.
170 * RFC 1122: SHOULD pass TOS value up to the transport layer.
171 * -> It does. And not only TOS, but all IP header.
173 static int raw_v4_input(struct sk_buff
*skb
, const struct iphdr
*iph
, int hash
)
175 int sdif
= inet_sdif(skb
);
176 int dif
= inet_iif(skb
);
178 struct hlist_head
*head
;
182 read_lock(&raw_v4_hashinfo
.lock
);
183 head
= &raw_v4_hashinfo
.ht
[hash
];
184 if (hlist_empty(head
))
187 net
= dev_net(skb
->dev
);
188 sk
= __raw_v4_lookup(net
, __sk_head(head
), iph
->protocol
,
189 iph
->saddr
, iph
->daddr
, dif
, sdif
);
193 if ((iph
->protocol
!= IPPROTO_ICMP
|| !icmp_filter(sk
, skb
)) &&
194 ip_mc_sf_allow(sk
, iph
->daddr
, iph
->saddr
,
195 skb
->dev
->ifindex
, sdif
)) {
196 struct sk_buff
*clone
= skb_clone(skb
, GFP_ATOMIC
);
198 /* Not releasing hash table! */
202 sk
= __raw_v4_lookup(net
, sk_next(sk
), iph
->protocol
,
203 iph
->saddr
, iph
->daddr
,
204 skb
->dev
->ifindex
, sdif
);
207 read_unlock(&raw_v4_hashinfo
.lock
);
211 int raw_local_deliver(struct sk_buff
*skb
, int protocol
)
216 hash
= protocol
& (RAW_HTABLE_SIZE
- 1);
217 raw_sk
= sk_head(&raw_v4_hashinfo
.ht
[hash
]);
219 /* If there maybe a raw socket we must check - if not we
222 if (raw_sk
&& !raw_v4_input(skb
, ip_hdr(skb
), hash
))
225 return raw_sk
!= NULL
;
229 static void raw_err(struct sock
*sk
, struct sk_buff
*skb
, u32 info
)
231 struct inet_sock
*inet
= inet_sk(sk
);
232 const int type
= icmp_hdr(skb
)->type
;
233 const int code
= icmp_hdr(skb
)->code
;
237 if (type
== ICMP_DEST_UNREACH
&& code
== ICMP_FRAG_NEEDED
)
238 ipv4_sk_update_pmtu(skb
, sk
, info
);
239 else if (type
== ICMP_REDIRECT
) {
240 ipv4_sk_redirect(skb
, sk
);
244 /* Report error on raw socket, if:
245 1. User requested ip_recverr.
246 2. Socket is connected (otherwise the error indication
247 is useless without ip_recverr and error is hard.
249 if (!inet
->recverr
&& sk
->sk_state
!= TCP_ESTABLISHED
)
254 case ICMP_TIME_EXCEEDED
:
257 case ICMP_SOURCE_QUENCH
:
259 case ICMP_PARAMETERPROB
:
263 case ICMP_DEST_UNREACH
:
265 if (code
> NR_ICMP_UNREACH
)
267 err
= icmp_err_convert
[code
].errno
;
268 harderr
= icmp_err_convert
[code
].fatal
;
269 if (code
== ICMP_FRAG_NEEDED
) {
270 harderr
= inet
->pmtudisc
!= IP_PMTUDISC_DONT
;
276 const struct iphdr
*iph
= (const struct iphdr
*)skb
->data
;
277 u8
*payload
= skb
->data
+ (iph
->ihl
<< 2);
281 ip_icmp_error(sk
, skb
, err
, 0, info
, payload
);
284 if (inet
->recverr
|| harderr
) {
286 sk
->sk_error_report(sk
);
290 void raw_icmp_error(struct sk_buff
*skb
, int protocol
, u32 info
)
294 const struct iphdr
*iph
;
297 hash
= protocol
& (RAW_HTABLE_SIZE
- 1);
299 read_lock(&raw_v4_hashinfo
.lock
);
300 raw_sk
= sk_head(&raw_v4_hashinfo
.ht
[hash
]);
302 int dif
= skb
->dev
->ifindex
;
303 int sdif
= inet_sdif(skb
);
305 iph
= (const struct iphdr
*)skb
->data
;
306 net
= dev_net(skb
->dev
);
308 while ((raw_sk
= __raw_v4_lookup(net
, raw_sk
, protocol
,
309 iph
->daddr
, iph
->saddr
,
310 dif
, sdif
)) != NULL
) {
311 raw_err(raw_sk
, skb
, info
);
312 raw_sk
= sk_next(raw_sk
);
313 iph
= (const struct iphdr
*)skb
->data
;
316 read_unlock(&raw_v4_hashinfo
.lock
);
319 static int raw_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
)
321 /* Charge it to the socket. */
323 ipv4_pktinfo_prepare(sk
, skb
);
324 if (sock_queue_rcv_skb(sk
, skb
) < 0) {
329 return NET_RX_SUCCESS
;
332 int raw_rcv(struct sock
*sk
, struct sk_buff
*skb
)
334 if (!xfrm4_policy_check(sk
, XFRM_POLICY_IN
, skb
)) {
335 atomic_inc(&sk
->sk_drops
);
341 skb_push(skb
, skb
->data
- skb_network_header(skb
));
343 raw_rcv_skb(sk
, skb
);
347 static int raw_send_hdrinc(struct sock
*sk
, struct flowi4
*fl4
,
348 struct msghdr
*msg
, size_t length
,
349 struct rtable
**rtp
, unsigned int flags
,
350 const struct sockcm_cookie
*sockc
)
352 struct inet_sock
*inet
= inet_sk(sk
);
353 struct net
*net
= sock_net(sk
);
358 struct rtable
*rt
= *rtp
;
361 if (length
> rt
->dst
.dev
->mtu
) {
362 ip_local_error(sk
, EMSGSIZE
, fl4
->daddr
, inet
->inet_dport
,
366 if (length
< sizeof(struct iphdr
))
372 hlen
= LL_RESERVED_SPACE(rt
->dst
.dev
);
373 tlen
= rt
->dst
.dev
->needed_tailroom
;
374 skb
= sock_alloc_send_skb(sk
,
375 length
+ hlen
+ tlen
+ 15,
376 flags
& MSG_DONTWAIT
, &err
);
379 skb_reserve(skb
, hlen
);
381 skb
->priority
= sk
->sk_priority
;
382 skb
->mark
= sk
->sk_mark
;
383 skb
->tstamp
= sockc
->transmit_time
;
384 skb_dst_set(skb
, &rt
->dst
);
387 skb_reset_network_header(skb
);
389 skb_put(skb
, length
);
391 skb
->ip_summed
= CHECKSUM_NONE
;
393 skb_setup_tx_timestamp(skb
, sockc
->tsflags
);
395 if (flags
& MSG_CONFIRM
)
396 skb_set_dst_pending_confirm(skb
, 1);
398 skb
->transport_header
= skb
->network_header
;
400 if (memcpy_from_msg(iph
, msg
, length
))
403 iphlen
= iph
->ihl
* 4;
406 * We don't want to modify the ip header, but we do need to
407 * be sure that it won't cause problems later along the network
408 * stack. Specifically we want to make sure that iph->ihl is a
409 * sane value. If ihl points beyond the length of the buffer passed
410 * in, reject the frame as invalid
416 if (iphlen
>= sizeof(*iph
)) {
418 iph
->saddr
= fl4
->saddr
;
420 iph
->tot_len
= htons(length
);
422 ip_select_ident(net
, skb
, NULL
);
424 iph
->check
= ip_fast_csum((unsigned char *)iph
, iph
->ihl
);
425 skb
->transport_header
+= iphlen
;
426 if (iph
->protocol
== IPPROTO_ICMP
&&
427 length
>= iphlen
+ sizeof(struct icmphdr
))
428 icmp_out_count(net
, ((struct icmphdr
*)
429 skb_transport_header(skb
))->type
);
432 err
= NF_HOOK(NFPROTO_IPV4
, NF_INET_LOCAL_OUT
,
433 net
, sk
, skb
, NULL
, rt
->dst
.dev
,
436 err
= net_xmit_errno(err
);
445 IP_INC_STATS(net
, IPSTATS_MIB_OUTDISCARDS
);
446 if (err
== -ENOBUFS
&& !inet
->recverr
)
451 static int raw_probe_proto_opt(struct raw_frag_vec
*rfv
, struct flowi4
*fl4
)
455 if (fl4
->flowi4_proto
!= IPPROTO_ICMP
)
458 /* We only need the first two bytes. */
461 err
= memcpy_from_msg(rfv
->hdr
.c
, rfv
->msg
, rfv
->hlen
);
465 fl4
->fl4_icmp_type
= rfv
->hdr
.icmph
.type
;
466 fl4
->fl4_icmp_code
= rfv
->hdr
.icmph
.code
;
471 static int raw_getfrag(void *from
, char *to
, int offset
, int len
, int odd
,
474 struct raw_frag_vec
*rfv
= from
;
476 if (offset
< rfv
->hlen
) {
477 int copy
= min(rfv
->hlen
- offset
, len
);
479 if (skb
->ip_summed
== CHECKSUM_PARTIAL
)
480 memcpy(to
, rfv
->hdr
.c
+ offset
, copy
);
482 skb
->csum
= csum_block_add(
484 csum_partial_copy_nocheck(rfv
->hdr
.c
+ offset
,
499 return ip_generic_getfrag(rfv
->msg
, to
, offset
, len
, odd
, skb
);
502 static int raw_sendmsg(struct sock
*sk
, struct msghdr
*msg
, size_t len
)
504 struct inet_sock
*inet
= inet_sk(sk
);
505 struct net
*net
= sock_net(sk
);
506 struct ipcm_cookie ipc
;
507 struct rtable
*rt
= NULL
;
514 struct ip_options_data opt_copy
;
515 struct raw_frag_vec rfv
;
522 /* hdrincl should be READ_ONCE(inet->hdrincl)
523 * but READ_ONCE() doesn't work with bit fields.
524 * Doing this indirectly yields the same result.
526 hdrincl
= inet
->hdrincl
;
527 hdrincl
= READ_ONCE(hdrincl
);
533 if (msg
->msg_flags
& MSG_OOB
) /* Mirror BSD error message */
534 goto out
; /* compatibility */
537 * Get and verify the address.
540 if (msg
->msg_namelen
) {
541 DECLARE_SOCKADDR(struct sockaddr_in
*, usin
, msg
->msg_name
);
543 if (msg
->msg_namelen
< sizeof(*usin
))
545 if (usin
->sin_family
!= AF_INET
) {
546 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
547 __func__
, current
->comm
);
549 if (usin
->sin_family
)
552 daddr
= usin
->sin_addr
.s_addr
;
553 /* ANK: I did not forget to get protocol from port field.
554 * I just do not know, who uses this weirdness.
555 * IP_HDRINCL is much more convenient.
559 if (sk
->sk_state
!= TCP_ESTABLISHED
)
561 daddr
= inet
->inet_daddr
;
564 ipcm_init_sk(&ipc
, inet
);
566 if (msg
->msg_controllen
) {
567 err
= ip_cmsg_send(sk
, msg
, &ipc
, false);
580 struct ip_options_rcu
*inet_opt
;
583 inet_opt
= rcu_dereference(inet
->inet_opt
);
585 memcpy(&opt_copy
, inet_opt
,
586 sizeof(*inet_opt
) + inet_opt
->opt
.optlen
);
587 ipc
.opt
= &opt_copy
.opt
;
594 /* Linux does not mangle headers on raw sockets,
595 * so that IP options + IP_HDRINCL is non-sense.
599 if (ipc
.opt
->opt
.srr
) {
602 daddr
= ipc
.opt
->opt
.faddr
;
605 tos
= get_rtconn_flags(&ipc
, sk
);
606 if (msg
->msg_flags
& MSG_DONTROUTE
)
609 if (ipv4_is_multicast(daddr
)) {
610 if (!ipc
.oif
|| netif_index_is_l3_master(sock_net(sk
), ipc
.oif
))
611 ipc
.oif
= inet
->mc_index
;
613 saddr
= inet
->mc_addr
;
614 } else if (!ipc
.oif
) {
615 ipc
.oif
= inet
->uc_index
;
616 } else if (ipv4_is_lbcast(daddr
) && inet
->uc_index
) {
617 /* oif is set, packet is to local broadcast and
618 * and uc_index is set. oif is most likely set
619 * by sk_bound_dev_if. If uc_index != oif check if the
620 * oif is an L3 master and uc_index is an L3 slave.
621 * If so, we want to allow the send using the uc_index.
623 if (ipc
.oif
!= inet
->uc_index
&&
624 ipc
.oif
== l3mdev_master_ifindex_by_index(sock_net(sk
),
626 ipc
.oif
= inet
->uc_index
;
630 flowi4_init_output(&fl4
, ipc
.oif
, sk
->sk_mark
, tos
,
632 hdrincl
? IPPROTO_RAW
: sk
->sk_protocol
,
633 inet_sk_flowi_flags(sk
) |
634 (hdrincl
? FLOWI_FLAG_KNOWN_NH
: 0),
635 daddr
, saddr
, 0, 0, sk
->sk_uid
);
641 err
= raw_probe_proto_opt(&rfv
, &fl4
);
646 security_sk_classify_flow(sk
, flowi4_to_flowi(&fl4
));
647 rt
= ip_route_output_flow(net
, &fl4
, sk
);
655 if (rt
->rt_flags
& RTCF_BROADCAST
&& !sock_flag(sk
, SOCK_BROADCAST
))
658 if (msg
->msg_flags
& MSG_CONFIRM
)
663 err
= raw_send_hdrinc(sk
, &fl4
, msg
, len
,
664 &rt
, msg
->msg_flags
, &ipc
.sockc
);
668 ipc
.addr
= fl4
.daddr
;
670 err
= ip_append_data(sk
, &fl4
, raw_getfrag
,
672 &ipc
, &rt
, msg
->msg_flags
);
674 ip_flush_pending_frames(sk
);
675 else if (!(msg
->msg_flags
& MSG_MORE
)) {
676 err
= ip_push_pending_frames(sk
, &fl4
);
677 if (err
== -ENOBUFS
&& !inet
->recverr
)
693 if (msg
->msg_flags
& MSG_PROBE
)
694 dst_confirm_neigh(&rt
->dst
, &fl4
.daddr
);
695 if (!(msg
->msg_flags
& MSG_PROBE
) || len
)
696 goto back_from_confirm
;
701 static void raw_close(struct sock
*sk
, long timeout
)
704 * Raw sockets may have direct kernel references. Kill them.
706 ip_ra_control(sk
, 0, NULL
);
708 sk_common_release(sk
);
711 static void raw_destroy(struct sock
*sk
)
714 ip_flush_pending_frames(sk
);
718 /* This gets rid of all the nasties in af_inet. -DaveM */
719 static int raw_bind(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
721 struct inet_sock
*inet
= inet_sk(sk
);
722 struct sockaddr_in
*addr
= (struct sockaddr_in
*) uaddr
;
723 u32 tb_id
= RT_TABLE_LOCAL
;
727 if (sk
->sk_state
!= TCP_CLOSE
|| addr_len
< sizeof(struct sockaddr_in
))
730 if (sk
->sk_bound_dev_if
)
731 tb_id
= l3mdev_fib_table_by_index(sock_net(sk
),
732 sk
->sk_bound_dev_if
) ? : tb_id
;
734 chk_addr_ret
= inet_addr_type_table(sock_net(sk
), addr
->sin_addr
.s_addr
,
737 ret
= -EADDRNOTAVAIL
;
738 if (addr
->sin_addr
.s_addr
&& chk_addr_ret
!= RTN_LOCAL
&&
739 chk_addr_ret
!= RTN_MULTICAST
&& chk_addr_ret
!= RTN_BROADCAST
)
741 inet
->inet_rcv_saddr
= inet
->inet_saddr
= addr
->sin_addr
.s_addr
;
742 if (chk_addr_ret
== RTN_MULTICAST
|| chk_addr_ret
== RTN_BROADCAST
)
743 inet
->inet_saddr
= 0; /* Use device */
750 * This should be easy, if there is something there
751 * we return it, otherwise we block.
754 static int raw_recvmsg(struct sock
*sk
, struct msghdr
*msg
, size_t len
,
755 int noblock
, int flags
, int *addr_len
)
757 struct inet_sock
*inet
= inet_sk(sk
);
759 int err
= -EOPNOTSUPP
;
760 DECLARE_SOCKADDR(struct sockaddr_in
*, sin
, msg
->msg_name
);
766 if (flags
& MSG_ERRQUEUE
) {
767 err
= ip_recv_error(sk
, msg
, len
, addr_len
);
771 skb
= skb_recv_datagram(sk
, flags
, noblock
, &err
);
777 msg
->msg_flags
|= MSG_TRUNC
;
781 err
= skb_copy_datagram_msg(skb
, 0, msg
, copied
);
785 sock_recv_ts_and_drops(msg
, sk
, skb
);
787 /* Copy the address. */
789 sin
->sin_family
= AF_INET
;
790 sin
->sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
792 memset(&sin
->sin_zero
, 0, sizeof(sin
->sin_zero
));
793 *addr_len
= sizeof(*sin
);
795 if (inet
->cmsg_flags
)
796 ip_cmsg_recv(msg
, skb
);
797 if (flags
& MSG_TRUNC
)
800 skb_free_datagram(sk
, skb
);
807 static int raw_sk_init(struct sock
*sk
)
809 struct raw_sock
*rp
= raw_sk(sk
);
811 if (inet_sk(sk
)->inet_num
== IPPROTO_ICMP
)
812 memset(&rp
->filter
, 0, sizeof(rp
->filter
));
816 static int raw_seticmpfilter(struct sock
*sk
, char __user
*optval
, int optlen
)
818 if (optlen
> sizeof(struct icmp_filter
))
819 optlen
= sizeof(struct icmp_filter
);
820 if (copy_from_user(&raw_sk(sk
)->filter
, optval
, optlen
))
825 static int raw_geticmpfilter(struct sock
*sk
, char __user
*optval
, int __user
*optlen
)
827 int len
, ret
= -EFAULT
;
829 if (get_user(len
, optlen
))
834 if (len
> sizeof(struct icmp_filter
))
835 len
= sizeof(struct icmp_filter
);
837 if (put_user(len
, optlen
) ||
838 copy_to_user(optval
, &raw_sk(sk
)->filter
, len
))
844 static int do_raw_setsockopt(struct sock
*sk
, int level
, int optname
,
845 char __user
*optval
, unsigned int optlen
)
847 if (optname
== ICMP_FILTER
) {
848 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMP
)
851 return raw_seticmpfilter(sk
, optval
, optlen
);
856 static int raw_setsockopt(struct sock
*sk
, int level
, int optname
,
857 char __user
*optval
, unsigned int optlen
)
859 if (level
!= SOL_RAW
)
860 return ip_setsockopt(sk
, level
, optname
, optval
, optlen
);
861 return do_raw_setsockopt(sk
, level
, optname
, optval
, optlen
);
865 static int compat_raw_setsockopt(struct sock
*sk
, int level
, int optname
,
866 char __user
*optval
, unsigned int optlen
)
868 if (level
!= SOL_RAW
)
869 return compat_ip_setsockopt(sk
, level
, optname
, optval
, optlen
);
870 return do_raw_setsockopt(sk
, level
, optname
, optval
, optlen
);
874 static int do_raw_getsockopt(struct sock
*sk
, int level
, int optname
,
875 char __user
*optval
, int __user
*optlen
)
877 if (optname
== ICMP_FILTER
) {
878 if (inet_sk(sk
)->inet_num
!= IPPROTO_ICMP
)
881 return raw_geticmpfilter(sk
, optval
, optlen
);
886 static int raw_getsockopt(struct sock
*sk
, int level
, int optname
,
887 char __user
*optval
, int __user
*optlen
)
889 if (level
!= SOL_RAW
)
890 return ip_getsockopt(sk
, level
, optname
, optval
, optlen
);
891 return do_raw_getsockopt(sk
, level
, optname
, optval
, optlen
);
895 static int compat_raw_getsockopt(struct sock
*sk
, int level
, int optname
,
896 char __user
*optval
, int __user
*optlen
)
898 if (level
!= SOL_RAW
)
899 return compat_ip_getsockopt(sk
, level
, optname
, optval
, optlen
);
900 return do_raw_getsockopt(sk
, level
, optname
, optval
, optlen
);
904 static int raw_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
908 int amount
= sk_wmem_alloc_get(sk
);
910 return put_user(amount
, (int __user
*)arg
);
916 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
917 skb
= skb_peek(&sk
->sk_receive_queue
);
920 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
921 return put_user(amount
, (int __user
*)arg
);
925 #ifdef CONFIG_IP_MROUTE
926 return ipmr_ioctl(sk
, cmd
, (void __user
*)arg
);
934 static int compat_raw_ioctl(struct sock
*sk
, unsigned int cmd
, unsigned long arg
)
941 #ifdef CONFIG_IP_MROUTE
942 return ipmr_compat_ioctl(sk
, cmd
, compat_ptr(arg
));
950 int raw_abort(struct sock
*sk
, int err
)
955 sk
->sk_error_report(sk
);
956 __udp_disconnect(sk
, 0);
962 EXPORT_SYMBOL_GPL(raw_abort
);
964 struct proto raw_prot
= {
966 .owner
= THIS_MODULE
,
968 .destroy
= raw_destroy
,
969 .connect
= ip4_datagram_connect
,
970 .disconnect
= __udp_disconnect
,
973 .setsockopt
= raw_setsockopt
,
974 .getsockopt
= raw_getsockopt
,
975 .sendmsg
= raw_sendmsg
,
976 .recvmsg
= raw_recvmsg
,
978 .backlog_rcv
= raw_rcv_skb
,
979 .release_cb
= ip4_datagram_release_cb
,
981 .unhash
= raw_unhash_sk
,
982 .obj_size
= sizeof(struct raw_sock
),
983 .useroffset
= offsetof(struct raw_sock
, filter
),
984 .usersize
= sizeof_field(struct raw_sock
, filter
),
985 .h
.raw_hash
= &raw_v4_hashinfo
,
987 .compat_setsockopt
= compat_raw_setsockopt
,
988 .compat_getsockopt
= compat_raw_getsockopt
,
989 .compat_ioctl
= compat_raw_ioctl
,
991 .diag_destroy
= raw_abort
,
994 #ifdef CONFIG_PROC_FS
995 static struct sock
*raw_get_first(struct seq_file
*seq
)
998 struct raw_hashinfo
*h
= PDE_DATA(file_inode(seq
->file
));
999 struct raw_iter_state
*state
= raw_seq_private(seq
);
1001 for (state
->bucket
= 0; state
->bucket
< RAW_HTABLE_SIZE
;
1003 sk_for_each(sk
, &h
->ht
[state
->bucket
])
1004 if (sock_net(sk
) == seq_file_net(seq
))
1012 static struct sock
*raw_get_next(struct seq_file
*seq
, struct sock
*sk
)
1014 struct raw_hashinfo
*h
= PDE_DATA(file_inode(seq
->file
));
1015 struct raw_iter_state
*state
= raw_seq_private(seq
);
1021 } while (sk
&& sock_net(sk
) != seq_file_net(seq
));
1023 if (!sk
&& ++state
->bucket
< RAW_HTABLE_SIZE
) {
1024 sk
= sk_head(&h
->ht
[state
->bucket
]);
1030 static struct sock
*raw_get_idx(struct seq_file
*seq
, loff_t pos
)
1032 struct sock
*sk
= raw_get_first(seq
);
1035 while (pos
&& (sk
= raw_get_next(seq
, sk
)) != NULL
)
1037 return pos
? NULL
: sk
;
1040 void *raw_seq_start(struct seq_file
*seq
, loff_t
*pos
)
1042 struct raw_hashinfo
*h
= PDE_DATA(file_inode(seq
->file
));
1044 read_lock(&h
->lock
);
1045 return *pos
? raw_get_idx(seq
, *pos
- 1) : SEQ_START_TOKEN
;
1047 EXPORT_SYMBOL_GPL(raw_seq_start
);
1049 void *raw_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
1053 if (v
== SEQ_START_TOKEN
)
1054 sk
= raw_get_first(seq
);
1056 sk
= raw_get_next(seq
, v
);
1060 EXPORT_SYMBOL_GPL(raw_seq_next
);
1062 void raw_seq_stop(struct seq_file
*seq
, void *v
)
1064 struct raw_hashinfo
*h
= PDE_DATA(file_inode(seq
->file
));
1066 read_unlock(&h
->lock
);
1068 EXPORT_SYMBOL_GPL(raw_seq_stop
);
1070 static void raw_sock_seq_show(struct seq_file
*seq
, struct sock
*sp
, int i
)
1072 struct inet_sock
*inet
= inet_sk(sp
);
1073 __be32 dest
= inet
->inet_daddr
,
1074 src
= inet
->inet_rcv_saddr
;
1076 srcp
= inet
->inet_num
;
1078 seq_printf(seq
, "%4d: %08X:%04X %08X:%04X"
1079 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u\n",
1080 i
, src
, srcp
, dest
, destp
, sp
->sk_state
,
1081 sk_wmem_alloc_get(sp
),
1082 sk_rmem_alloc_get(sp
),
1084 from_kuid_munged(seq_user_ns(seq
), sock_i_uid(sp
)),
1086 refcount_read(&sp
->sk_refcnt
), sp
, atomic_read(&sp
->sk_drops
));
1089 static int raw_seq_show(struct seq_file
*seq
, void *v
)
1091 if (v
== SEQ_START_TOKEN
)
1092 seq_printf(seq
, " sl local_address rem_address st tx_queue "
1093 "rx_queue tr tm->when retrnsmt uid timeout "
1094 "inode ref pointer drops\n");
1096 raw_sock_seq_show(seq
, v
, raw_seq_private(seq
)->bucket
);
1100 static const struct seq_operations raw_seq_ops
= {
1101 .start
= raw_seq_start
,
1102 .next
= raw_seq_next
,
1103 .stop
= raw_seq_stop
,
1104 .show
= raw_seq_show
,
1107 static __net_init
int raw_init_net(struct net
*net
)
1109 if (!proc_create_net_data("raw", 0444, net
->proc_net
, &raw_seq_ops
,
1110 sizeof(struct raw_iter_state
), &raw_v4_hashinfo
))
1116 static __net_exit
void raw_exit_net(struct net
*net
)
1118 remove_proc_entry("raw", net
->proc_net
);
1121 static __net_initdata
struct pernet_operations raw_net_ops
= {
1122 .init
= raw_init_net
,
1123 .exit
= raw_exit_net
,
1126 int __init
raw_proc_init(void)
1128 return register_pernet_subsys(&raw_net_ops
);
1131 void __init
raw_proc_exit(void)
1133 unregister_pernet_subsys(&raw_net_ops
);
1135 #endif /* CONFIG_PROC_FS */
1137 static void raw_sysctl_init_net(struct net
*net
)
1139 #ifdef CONFIG_NET_L3_MASTER_DEV
1140 net
->ipv4
.sysctl_raw_l3mdev_accept
= 1;
1144 static int __net_init
raw_sysctl_init(struct net
*net
)
1146 raw_sysctl_init_net(net
);
1150 static struct pernet_operations __net_initdata raw_sysctl_ops
= {
1151 .init
= raw_sysctl_init
,
1154 void __init
raw_init(void)
1156 raw_sysctl_init_net(&init_net
);
1157 if (register_pernet_subsys(&raw_sysctl_ops
))
1158 panic("RAW: failed to init sysctl parameters.\n");