1 // SPDX-License-Identifier: GPL-2.0
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
12 * Many : Split from ip.c , see ip.c for history.
13 * Martin Mares : TOS setting fixed.
14 * Alan Cox : Fixed a couple of oopses in Martin's
16 * Mike McLagan : Routing by source
19 #include <linux/module.h>
20 #include <linux/types.h>
22 #include <linux/skbuff.h>
24 #include <linux/icmp.h>
25 #include <linux/inetdevice.h>
26 #include <linux/netdevice.h>
27 #include <linux/slab.h>
31 #include <net/tcp_states.h>
32 #include <linux/udp.h>
33 #include <linux/igmp.h>
34 #include <linux/netfilter.h>
35 #include <linux/route.h>
36 #include <linux/mroute.h>
37 #include <net/inet_ecn.h>
38 #include <net/route.h>
40 #include <net/compat.h>
41 #include <net/checksum.h>
42 #if IS_ENABLED(CONFIG_IPV6)
43 #include <net/transp_v6.h>
45 #include <net/ip_fib.h>
47 #include <linux/errqueue.h>
48 #include <linux/uaccess.h>
51 * SOL_IP control messages.
54 static void ip_cmsg_recv_pktinfo(struct msghdr
*msg
, struct sk_buff
*skb
)
56 struct in_pktinfo info
= *PKTINFO_SKB_CB(skb
);
58 info
.ipi_addr
.s_addr
= ip_hdr(skb
)->daddr
;
60 put_cmsg(msg
, SOL_IP
, IP_PKTINFO
, sizeof(info
), &info
);
63 static void ip_cmsg_recv_ttl(struct msghdr
*msg
, struct sk_buff
*skb
)
65 int ttl
= ip_hdr(skb
)->ttl
;
66 put_cmsg(msg
, SOL_IP
, IP_TTL
, sizeof(int), &ttl
);
69 static void ip_cmsg_recv_tos(struct msghdr
*msg
, struct sk_buff
*skb
)
71 put_cmsg(msg
, SOL_IP
, IP_TOS
, 1, &ip_hdr(skb
)->tos
);
74 static void ip_cmsg_recv_opts(struct msghdr
*msg
, struct sk_buff
*skb
)
76 if (IPCB(skb
)->opt
.optlen
== 0)
79 put_cmsg(msg
, SOL_IP
, IP_RECVOPTS
, IPCB(skb
)->opt
.optlen
,
84 static void ip_cmsg_recv_retopts(struct net
*net
, struct msghdr
*msg
,
87 unsigned char optbuf
[sizeof(struct ip_options
) + 40];
88 struct ip_options
*opt
= (struct ip_options
*)optbuf
;
90 if (IPCB(skb
)->opt
.optlen
== 0)
93 if (ip_options_echo(net
, opt
, skb
)) {
94 msg
->msg_flags
|= MSG_CTRUNC
;
99 put_cmsg(msg
, SOL_IP
, IP_RETOPTS
, opt
->optlen
, opt
->__data
);
102 static void ip_cmsg_recv_fragsize(struct msghdr
*msg
, struct sk_buff
*skb
)
106 if (IPCB(skb
)->frag_max_size
== 0)
109 val
= IPCB(skb
)->frag_max_size
;
110 put_cmsg(msg
, SOL_IP
, IP_RECVFRAGSIZE
, sizeof(val
), &val
);
113 static void ip_cmsg_recv_checksum(struct msghdr
*msg
, struct sk_buff
*skb
,
114 int tlen
, int offset
)
116 __wsum csum
= skb
->csum
;
118 if (skb
->ip_summed
!= CHECKSUM_COMPLETE
)
122 int tend_off
= skb_transport_offset(skb
) + tlen
;
123 csum
= csum_sub(csum
, skb_checksum(skb
, tend_off
, offset
, 0));
126 put_cmsg(msg
, SOL_IP
, IP_CHECKSUM
, sizeof(__wsum
), &csum
);
129 static void ip_cmsg_recv_security(struct msghdr
*msg
, struct sk_buff
*skb
)
135 err
= security_socket_getpeersec_dgram(NULL
, skb
, &secid
);
139 err
= security_secid_to_secctx(secid
, &secdata
, &seclen
);
143 put_cmsg(msg
, SOL_IP
, SCM_SECURITY
, seclen
, secdata
);
144 security_release_secctx(secdata
, seclen
);
147 static void ip_cmsg_recv_dstaddr(struct msghdr
*msg
, struct sk_buff
*skb
)
149 __be16 _ports
[2], *ports
;
150 struct sockaddr_in sin
;
152 /* All current transport protocols have the port numbers in the
153 * first four bytes of the transport header and this function is
154 * written with this assumption in mind.
156 ports
= skb_header_pointer(skb
, skb_transport_offset(skb
),
157 sizeof(_ports
), &_ports
);
161 sin
.sin_family
= AF_INET
;
162 sin
.sin_addr
.s_addr
= ip_hdr(skb
)->daddr
;
163 sin
.sin_port
= ports
[1];
164 memset(sin
.sin_zero
, 0, sizeof(sin
.sin_zero
));
166 put_cmsg(msg
, SOL_IP
, IP_ORIGDSTADDR
, sizeof(sin
), &sin
);
169 void ip_cmsg_recv_offset(struct msghdr
*msg
, struct sock
*sk
,
170 struct sk_buff
*skb
, int tlen
, int offset
)
172 struct inet_sock
*inet
= inet_sk(sk
);
173 unsigned int flags
= inet
->cmsg_flags
;
175 /* Ordered by supposed usage frequency */
176 if (flags
& IP_CMSG_PKTINFO
) {
177 ip_cmsg_recv_pktinfo(msg
, skb
);
179 flags
&= ~IP_CMSG_PKTINFO
;
184 if (flags
& IP_CMSG_TTL
) {
185 ip_cmsg_recv_ttl(msg
, skb
);
187 flags
&= ~IP_CMSG_TTL
;
192 if (flags
& IP_CMSG_TOS
) {
193 ip_cmsg_recv_tos(msg
, skb
);
195 flags
&= ~IP_CMSG_TOS
;
200 if (flags
& IP_CMSG_RECVOPTS
) {
201 ip_cmsg_recv_opts(msg
, skb
);
203 flags
&= ~IP_CMSG_RECVOPTS
;
208 if (flags
& IP_CMSG_RETOPTS
) {
209 ip_cmsg_recv_retopts(sock_net(sk
), msg
, skb
);
211 flags
&= ~IP_CMSG_RETOPTS
;
216 if (flags
& IP_CMSG_PASSSEC
) {
217 ip_cmsg_recv_security(msg
, skb
);
219 flags
&= ~IP_CMSG_PASSSEC
;
224 if (flags
& IP_CMSG_ORIGDSTADDR
) {
225 ip_cmsg_recv_dstaddr(msg
, skb
);
227 flags
&= ~IP_CMSG_ORIGDSTADDR
;
232 if (flags
& IP_CMSG_CHECKSUM
)
233 ip_cmsg_recv_checksum(msg
, skb
, tlen
, offset
);
235 if (flags
& IP_CMSG_RECVFRAGSIZE
)
236 ip_cmsg_recv_fragsize(msg
, skb
);
238 EXPORT_SYMBOL(ip_cmsg_recv_offset
);
240 int ip_cmsg_send(struct sock
*sk
, struct msghdr
*msg
, struct ipcm_cookie
*ipc
,
244 struct cmsghdr
*cmsg
;
245 struct net
*net
= sock_net(sk
);
247 for_each_cmsghdr(cmsg
, msg
) {
248 if (!CMSG_OK(msg
, cmsg
))
250 #if IS_ENABLED(CONFIG_IPV6)
252 cmsg
->cmsg_level
== SOL_IPV6
&&
253 cmsg
->cmsg_type
== IPV6_PKTINFO
) {
254 struct in6_pktinfo
*src_info
;
256 if (cmsg
->cmsg_len
< CMSG_LEN(sizeof(*src_info
)))
258 src_info
= (struct in6_pktinfo
*)CMSG_DATA(cmsg
);
259 if (!ipv6_addr_v4mapped(&src_info
->ipi6_addr
))
261 if (src_info
->ipi6_ifindex
)
262 ipc
->oif
= src_info
->ipi6_ifindex
;
263 ipc
->addr
= src_info
->ipi6_addr
.s6_addr32
[3];
267 if (cmsg
->cmsg_level
== SOL_SOCKET
) {
268 err
= __sock_cmsg_send(sk
, msg
, cmsg
, &ipc
->sockc
);
274 if (cmsg
->cmsg_level
!= SOL_IP
)
276 switch (cmsg
->cmsg_type
) {
278 err
= cmsg
->cmsg_len
- sizeof(struct cmsghdr
);
280 /* Our caller is responsible for freeing ipc->opt */
281 err
= ip_options_get(net
, &ipc
->opt
, CMSG_DATA(cmsg
),
282 err
< 40 ? err
: 40);
288 struct in_pktinfo
*info
;
289 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct in_pktinfo
)))
291 info
= (struct in_pktinfo
*)CMSG_DATA(cmsg
);
292 if (info
->ipi_ifindex
)
293 ipc
->oif
= info
->ipi_ifindex
;
294 ipc
->addr
= info
->ipi_spec_dst
.s_addr
;
298 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(int)))
300 val
= *(int *)CMSG_DATA(cmsg
);
301 if (val
< 1 || val
> 255)
306 if (cmsg
->cmsg_len
== CMSG_LEN(sizeof(int)))
307 val
= *(int *)CMSG_DATA(cmsg
);
308 else if (cmsg
->cmsg_len
== CMSG_LEN(sizeof(u8
)))
309 val
= *(u8
*)CMSG_DATA(cmsg
);
312 if (val
< 0 || val
> 255)
315 ipc
->priority
= rt_tos2priority(ipc
->tos
);
326 /* Special input handler for packets caught by router alert option.
327 They are selected only by protocol field, and then processed likely
328 local ones; but only if someone wants them! Otherwise, router
329 not running rsvpd will kill RSVP.
331 It is user level problem, what it will make with them.
332 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
333 but receiver should be enough clever f.e. to forward mtrace requests,
334 sent to multicast group to reach destination designated router.
336 struct ip_ra_chain __rcu
*ip_ra_chain
;
339 static void ip_ra_destroy_rcu(struct rcu_head
*head
)
341 struct ip_ra_chain
*ra
= container_of(head
, struct ip_ra_chain
, rcu
);
343 sock_put(ra
->saved_sk
);
347 int ip_ra_control(struct sock
*sk
, unsigned char on
,
348 void (*destructor
)(struct sock
*))
350 struct ip_ra_chain
*ra
, *new_ra
;
351 struct ip_ra_chain __rcu
**rap
;
353 if (sk
->sk_type
!= SOCK_RAW
|| inet_sk(sk
)->inet_num
== IPPROTO_RAW
)
356 new_ra
= on
? kmalloc(sizeof(*new_ra
), GFP_KERNEL
) : NULL
;
358 for (rap
= &ip_ra_chain
;
359 (ra
= rtnl_dereference(*rap
)) != NULL
;
366 /* dont let ip_call_ra_chain() use sk again */
368 RCU_INIT_POINTER(*rap
, ra
->next
);
373 * Delay sock_put(sk) and kfree(ra) after one rcu grace
374 * period. This guarantee ip_call_ra_chain() dont need
375 * to mess with socket refcounts.
378 call_rcu(&ra
->rcu
, ip_ra_destroy_rcu
);
385 new_ra
->destructor
= destructor
;
387 RCU_INIT_POINTER(new_ra
->next
, ra
);
388 rcu_assign_pointer(*rap
, new_ra
);
394 void ip_icmp_error(struct sock
*sk
, struct sk_buff
*skb
, int err
,
395 __be16 port
, u32 info
, u8
*payload
)
397 struct sock_exterr_skb
*serr
;
399 skb
= skb_clone(skb
, GFP_ATOMIC
);
403 serr
= SKB_EXT_ERR(skb
);
404 serr
->ee
.ee_errno
= err
;
405 serr
->ee
.ee_origin
= SO_EE_ORIGIN_ICMP
;
406 serr
->ee
.ee_type
= icmp_hdr(skb
)->type
;
407 serr
->ee
.ee_code
= icmp_hdr(skb
)->code
;
409 serr
->ee
.ee_info
= info
;
410 serr
->ee
.ee_data
= 0;
411 serr
->addr_offset
= (u8
*)&(((struct iphdr
*)(icmp_hdr(skb
) + 1))->daddr
) -
412 skb_network_header(skb
);
415 if (skb_pull(skb
, payload
- skb
->data
)) {
416 skb_reset_transport_header(skb
);
417 if (sock_queue_err_skb(sk
, skb
) == 0)
423 void ip_local_error(struct sock
*sk
, int err
, __be32 daddr
, __be16 port
, u32 info
)
425 struct inet_sock
*inet
= inet_sk(sk
);
426 struct sock_exterr_skb
*serr
;
433 skb
= alloc_skb(sizeof(struct iphdr
), GFP_ATOMIC
);
437 skb_put(skb
, sizeof(struct iphdr
));
438 skb_reset_network_header(skb
);
442 serr
= SKB_EXT_ERR(skb
);
443 serr
->ee
.ee_errno
= err
;
444 serr
->ee
.ee_origin
= SO_EE_ORIGIN_LOCAL
;
445 serr
->ee
.ee_type
= 0;
446 serr
->ee
.ee_code
= 0;
448 serr
->ee
.ee_info
= info
;
449 serr
->ee
.ee_data
= 0;
450 serr
->addr_offset
= (u8
*)&iph
->daddr
- skb_network_header(skb
);
453 __skb_pull(skb
, skb_tail_pointer(skb
) - skb
->data
);
454 skb_reset_transport_header(skb
);
456 if (sock_queue_err_skb(sk
, skb
))
460 /* For some errors we have valid addr_offset even with zero payload and
461 * zero port. Also, addr_offset should be supported if port is set.
463 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb
*serr
)
465 return serr
->ee
.ee_origin
== SO_EE_ORIGIN_ICMP
||
466 serr
->ee
.ee_origin
== SO_EE_ORIGIN_LOCAL
|| serr
->port
;
469 /* IPv4 supports cmsg on all imcp errors and some timestamps
471 * Timestamp code paths do not initialize the fields expected by cmsg:
472 * the PKTINFO fields in skb->cb[]. Fill those in here.
474 static bool ipv4_datagram_support_cmsg(const struct sock
*sk
,
478 struct in_pktinfo
*info
;
480 if (ee_origin
== SO_EE_ORIGIN_ICMP
)
483 if (ee_origin
== SO_EE_ORIGIN_LOCAL
)
486 /* Support IP_PKTINFO on tstamp packets if requested, to correlate
487 * timestamp with egress dev. Not possible for packets without iif
488 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
490 info
= PKTINFO_SKB_CB(skb
);
491 if (!(sk
->sk_tsflags
& SOF_TIMESTAMPING_OPT_CMSG
) ||
495 info
->ipi_spec_dst
.s_addr
= ip_hdr(skb
)->saddr
;
500 * Handle MSG_ERRQUEUE
502 int ip_recv_error(struct sock
*sk
, struct msghdr
*msg
, int len
, int *addr_len
)
504 struct sock_exterr_skb
*serr
;
506 DECLARE_SOCKADDR(struct sockaddr_in
*, sin
, msg
->msg_name
);
508 struct sock_extended_err ee
;
509 struct sockaddr_in offender
;
515 skb
= sock_dequeue_err_skb(sk
);
521 msg
->msg_flags
|= MSG_TRUNC
;
524 err
= skb_copy_datagram_msg(skb
, 0, msg
, copied
);
529 sock_recv_timestamp(msg
, sk
, skb
);
531 serr
= SKB_EXT_ERR(skb
);
533 if (sin
&& ipv4_datagram_support_addr(serr
)) {
534 sin
->sin_family
= AF_INET
;
535 sin
->sin_addr
.s_addr
= *(__be32
*)(skb_network_header(skb
) +
537 sin
->sin_port
= serr
->port
;
538 memset(&sin
->sin_zero
, 0, sizeof(sin
->sin_zero
));
539 *addr_len
= sizeof(*sin
);
542 memcpy(&errhdr
.ee
, &serr
->ee
, sizeof(struct sock_extended_err
));
543 sin
= &errhdr
.offender
;
544 memset(sin
, 0, sizeof(*sin
));
546 if (ipv4_datagram_support_cmsg(sk
, skb
, serr
->ee
.ee_origin
)) {
547 sin
->sin_family
= AF_INET
;
548 sin
->sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
549 if (inet_sk(sk
)->cmsg_flags
)
550 ip_cmsg_recv(msg
, skb
);
553 put_cmsg(msg
, SOL_IP
, IP_RECVERR
, sizeof(errhdr
), &errhdr
);
555 /* Now we could try to dump offended packet options */
557 msg
->msg_flags
|= MSG_ERRQUEUE
;
567 * Socket option code for IP. This is the end of the line after any
568 * TCP,UDP etc options on an IP socket.
570 static bool setsockopt_needs_rtnl(int optname
)
573 case IP_ADD_MEMBERSHIP
:
574 case IP_ADD_SOURCE_MEMBERSHIP
:
575 case IP_BLOCK_SOURCE
:
576 case IP_DROP_MEMBERSHIP
:
577 case IP_DROP_SOURCE_MEMBERSHIP
:
579 case IP_UNBLOCK_SOURCE
:
580 case MCAST_BLOCK_SOURCE
:
582 case MCAST_JOIN_GROUP
:
583 case MCAST_JOIN_SOURCE_GROUP
:
584 case MCAST_LEAVE_GROUP
:
585 case MCAST_LEAVE_SOURCE_GROUP
:
586 case MCAST_UNBLOCK_SOURCE
:
587 case IP_ROUTER_ALERT
:
593 static int do_ip_setsockopt(struct sock
*sk
, int level
,
594 int optname
, char __user
*optval
, unsigned int optlen
)
596 struct inet_sock
*inet
= inet_sk(sk
);
597 struct net
*net
= sock_net(sk
);
599 bool needs_rtnl
= setsockopt_needs_rtnl(optname
);
610 case IP_MTU_DISCOVER
:
612 case IP_ROUTER_ALERT
:
618 case IP_BIND_ADDRESS_NO_PORT
:
620 case IP_MULTICAST_TTL
:
621 case IP_MULTICAST_ALL
:
622 case IP_MULTICAST_LOOP
:
623 case IP_RECVORIGDSTADDR
:
625 case IP_RECVFRAGSIZE
:
626 if (optlen
>= sizeof(int)) {
627 if (get_user(val
, (int __user
*) optval
))
629 } else if (optlen
>= sizeof(char)) {
632 if (get_user(ucval
, (unsigned char __user
*) optval
))
638 /* If optlen==0, it is equivalent to val == 0 */
640 if (ip_mroute_opt(optname
))
641 return ip_mroute_setsockopt(sk
, optname
, optval
, optlen
);
651 struct ip_options_rcu
*old
, *opt
= NULL
;
655 err
= ip_options_get_from_user(sock_net(sk
), &opt
,
659 old
= rcu_dereference_protected(inet
->inet_opt
,
660 lockdep_sock_is_held(sk
));
662 struct inet_connection_sock
*icsk
= inet_csk(sk
);
663 #if IS_ENABLED(CONFIG_IPV6)
664 if (sk
->sk_family
== PF_INET
||
665 (!((1 << sk
->sk_state
) &
666 (TCPF_LISTEN
| TCPF_CLOSE
)) &&
667 inet
->inet_daddr
!= LOOPBACK4_IPV6
)) {
670 icsk
->icsk_ext_hdr_len
-= old
->opt
.optlen
;
672 icsk
->icsk_ext_hdr_len
+= opt
->opt
.optlen
;
673 icsk
->icsk_sync_mss(sk
, icsk
->icsk_pmtu_cookie
);
674 #if IS_ENABLED(CONFIG_IPV6)
678 rcu_assign_pointer(inet
->inet_opt
, opt
);
685 inet
->cmsg_flags
|= IP_CMSG_PKTINFO
;
687 inet
->cmsg_flags
&= ~IP_CMSG_PKTINFO
;
691 inet
->cmsg_flags
|= IP_CMSG_TTL
;
693 inet
->cmsg_flags
&= ~IP_CMSG_TTL
;
697 inet
->cmsg_flags
|= IP_CMSG_TOS
;
699 inet
->cmsg_flags
&= ~IP_CMSG_TOS
;
703 inet
->cmsg_flags
|= IP_CMSG_RECVOPTS
;
705 inet
->cmsg_flags
&= ~IP_CMSG_RECVOPTS
;
709 inet
->cmsg_flags
|= IP_CMSG_RETOPTS
;
711 inet
->cmsg_flags
&= ~IP_CMSG_RETOPTS
;
715 inet
->cmsg_flags
|= IP_CMSG_PASSSEC
;
717 inet
->cmsg_flags
&= ~IP_CMSG_PASSSEC
;
719 case IP_RECVORIGDSTADDR
:
721 inet
->cmsg_flags
|= IP_CMSG_ORIGDSTADDR
;
723 inet
->cmsg_flags
&= ~IP_CMSG_ORIGDSTADDR
;
727 if (!(inet
->cmsg_flags
& IP_CMSG_CHECKSUM
)) {
728 inet_inc_convert_csum(sk
);
729 inet
->cmsg_flags
|= IP_CMSG_CHECKSUM
;
732 if (inet
->cmsg_flags
& IP_CMSG_CHECKSUM
) {
733 inet_dec_convert_csum(sk
);
734 inet
->cmsg_flags
&= ~IP_CMSG_CHECKSUM
;
738 case IP_RECVFRAGSIZE
:
739 if (sk
->sk_type
!= SOCK_RAW
&& sk
->sk_type
!= SOCK_DGRAM
)
742 inet
->cmsg_flags
|= IP_CMSG_RECVFRAGSIZE
;
744 inet
->cmsg_flags
&= ~IP_CMSG_RECVFRAGSIZE
;
746 case IP_TOS
: /* This sets both TOS and Precedence */
747 if (sk
->sk_type
== SOCK_STREAM
) {
748 val
&= ~INET_ECN_MASK
;
749 val
|= inet
->tos
& INET_ECN_MASK
;
751 if (inet
->tos
!= val
) {
753 sk
->sk_priority
= rt_tos2priority(val
);
760 if (val
!= -1 && (val
< 1 || val
> 255))
765 if (sk
->sk_type
!= SOCK_RAW
) {
769 inet
->hdrincl
= val
? 1 : 0;
772 if (sk
->sk_type
!= SOCK_RAW
) {
776 inet
->nodefrag
= val
? 1 : 0;
778 case IP_BIND_ADDRESS_NO_PORT
:
779 inet
->bind_address_no_port
= val
? 1 : 0;
781 case IP_MTU_DISCOVER
:
782 if (val
< IP_PMTUDISC_DONT
|| val
> IP_PMTUDISC_OMIT
)
784 inet
->pmtudisc
= val
;
787 inet
->recverr
= !!val
;
789 skb_queue_purge(&sk
->sk_error_queue
);
791 case IP_MULTICAST_TTL
:
792 if (sk
->sk_type
== SOCK_STREAM
)
798 if (val
< 0 || val
> 255)
802 case IP_MULTICAST_LOOP
:
805 inet
->mc_loop
= !!val
;
809 struct net_device
*dev
= NULL
;
812 if (optlen
!= sizeof(int))
815 ifindex
= (__force
int)ntohl((__force __be32
)val
);
822 dev
= dev_get_by_index(sock_net(sk
), ifindex
);
823 err
= -EADDRNOTAVAIL
;
829 if (sk
->sk_bound_dev_if
)
832 inet
->uc_index
= ifindex
;
836 case IP_MULTICAST_IF
:
838 struct ip_mreqn mreq
;
839 struct net_device
*dev
= NULL
;
842 if (sk
->sk_type
== SOCK_STREAM
)
845 * Check the arguments are allowable
848 if (optlen
< sizeof(struct in_addr
))
852 if (optlen
>= sizeof(struct ip_mreqn
)) {
853 if (copy_from_user(&mreq
, optval
, sizeof(mreq
)))
856 memset(&mreq
, 0, sizeof(mreq
));
857 if (optlen
>= sizeof(struct ip_mreq
)) {
858 if (copy_from_user(&mreq
, optval
,
859 sizeof(struct ip_mreq
)))
861 } else if (optlen
>= sizeof(struct in_addr
)) {
862 if (copy_from_user(&mreq
.imr_address
, optval
,
863 sizeof(struct in_addr
)))
868 if (!mreq
.imr_ifindex
) {
869 if (mreq
.imr_address
.s_addr
== htonl(INADDR_ANY
)) {
875 dev
= ip_dev_find(sock_net(sk
), mreq
.imr_address
.s_addr
);
877 mreq
.imr_ifindex
= dev
->ifindex
;
879 dev
= dev_get_by_index(sock_net(sk
), mreq
.imr_ifindex
);
882 err
= -EADDRNOTAVAIL
;
886 midx
= l3mdev_master_ifindex(dev
);
891 if (sk
->sk_bound_dev_if
&&
892 mreq
.imr_ifindex
!= sk
->sk_bound_dev_if
&&
893 (!midx
|| midx
!= sk
->sk_bound_dev_if
))
896 inet
->mc_index
= mreq
.imr_ifindex
;
897 inet
->mc_addr
= mreq
.imr_address
.s_addr
;
902 case IP_ADD_MEMBERSHIP
:
903 case IP_DROP_MEMBERSHIP
:
905 struct ip_mreqn mreq
;
908 if (inet_sk(sk
)->is_icsk
)
911 if (optlen
< sizeof(struct ip_mreq
))
914 if (optlen
>= sizeof(struct ip_mreqn
)) {
915 if (copy_from_user(&mreq
, optval
, sizeof(mreq
)))
918 memset(&mreq
, 0, sizeof(mreq
));
919 if (copy_from_user(&mreq
, optval
, sizeof(struct ip_mreq
)))
923 if (optname
== IP_ADD_MEMBERSHIP
)
924 err
= ip_mc_join_group(sk
, &mreq
);
926 err
= ip_mc_leave_group(sk
, &mreq
);
931 struct ip_msfilter
*msf
;
933 if (optlen
< IP_MSFILTER_SIZE(0))
935 if (optlen
> sysctl_optmem_max
) {
939 msf
= memdup_user(optval
, optlen
);
944 /* numsrc >= (1G-4) overflow in 32 bits */
945 if (msf
->imsf_numsrc
>= 0x3ffffffcU
||
946 msf
->imsf_numsrc
> net
->ipv4
.sysctl_igmp_max_msf
) {
951 if (IP_MSFILTER_SIZE(msf
->imsf_numsrc
) > optlen
) {
956 err
= ip_mc_msfilter(sk
, msf
, 0);
960 case IP_BLOCK_SOURCE
:
961 case IP_UNBLOCK_SOURCE
:
962 case IP_ADD_SOURCE_MEMBERSHIP
:
963 case IP_DROP_SOURCE_MEMBERSHIP
:
965 struct ip_mreq_source mreqs
;
968 if (optlen
!= sizeof(struct ip_mreq_source
))
970 if (copy_from_user(&mreqs
, optval
, sizeof(mreqs
))) {
974 if (optname
== IP_BLOCK_SOURCE
) {
975 omode
= MCAST_EXCLUDE
;
977 } else if (optname
== IP_UNBLOCK_SOURCE
) {
978 omode
= MCAST_EXCLUDE
;
980 } else if (optname
== IP_ADD_SOURCE_MEMBERSHIP
) {
981 struct ip_mreqn mreq
;
983 mreq
.imr_multiaddr
.s_addr
= mreqs
.imr_multiaddr
;
984 mreq
.imr_address
.s_addr
= mreqs
.imr_interface
;
985 mreq
.imr_ifindex
= 0;
986 err
= ip_mc_join_group(sk
, &mreq
);
987 if (err
&& err
!= -EADDRINUSE
)
989 omode
= MCAST_INCLUDE
;
991 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
992 omode
= MCAST_INCLUDE
;
995 err
= ip_mc_source(add
, omode
, sk
, &mreqs
, 0);
998 case MCAST_JOIN_GROUP
:
999 case MCAST_LEAVE_GROUP
:
1001 struct group_req greq
;
1002 struct sockaddr_in
*psin
;
1003 struct ip_mreqn mreq
;
1005 if (optlen
< sizeof(struct group_req
))
1008 if (copy_from_user(&greq
, optval
, sizeof(greq
)))
1010 psin
= (struct sockaddr_in
*)&greq
.gr_group
;
1011 if (psin
->sin_family
!= AF_INET
)
1013 memset(&mreq
, 0, sizeof(mreq
));
1014 mreq
.imr_multiaddr
= psin
->sin_addr
;
1015 mreq
.imr_ifindex
= greq
.gr_interface
;
1017 if (optname
== MCAST_JOIN_GROUP
)
1018 err
= ip_mc_join_group(sk
, &mreq
);
1020 err
= ip_mc_leave_group(sk
, &mreq
);
1023 case MCAST_JOIN_SOURCE_GROUP
:
1024 case MCAST_LEAVE_SOURCE_GROUP
:
1025 case MCAST_BLOCK_SOURCE
:
1026 case MCAST_UNBLOCK_SOURCE
:
1028 struct group_source_req greqs
;
1029 struct ip_mreq_source mreqs
;
1030 struct sockaddr_in
*psin
;
1033 if (optlen
!= sizeof(struct group_source_req
))
1035 if (copy_from_user(&greqs
, optval
, sizeof(greqs
))) {
1039 if (greqs
.gsr_group
.ss_family
!= AF_INET
||
1040 greqs
.gsr_source
.ss_family
!= AF_INET
) {
1041 err
= -EADDRNOTAVAIL
;
1044 psin
= (struct sockaddr_in
*)&greqs
.gsr_group
;
1045 mreqs
.imr_multiaddr
= psin
->sin_addr
.s_addr
;
1046 psin
= (struct sockaddr_in
*)&greqs
.gsr_source
;
1047 mreqs
.imr_sourceaddr
= psin
->sin_addr
.s_addr
;
1048 mreqs
.imr_interface
= 0; /* use index for mc_source */
1050 if (optname
== MCAST_BLOCK_SOURCE
) {
1051 omode
= MCAST_EXCLUDE
;
1053 } else if (optname
== MCAST_UNBLOCK_SOURCE
) {
1054 omode
= MCAST_EXCLUDE
;
1056 } else if (optname
== MCAST_JOIN_SOURCE_GROUP
) {
1057 struct ip_mreqn mreq
;
1059 psin
= (struct sockaddr_in
*)&greqs
.gsr_group
;
1060 mreq
.imr_multiaddr
= psin
->sin_addr
;
1061 mreq
.imr_address
.s_addr
= 0;
1062 mreq
.imr_ifindex
= greqs
.gsr_interface
;
1063 err
= ip_mc_join_group(sk
, &mreq
);
1064 if (err
&& err
!= -EADDRINUSE
)
1066 greqs
.gsr_interface
= mreq
.imr_ifindex
;
1067 omode
= MCAST_INCLUDE
;
1069 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1070 omode
= MCAST_INCLUDE
;
1073 err
= ip_mc_source(add
, omode
, sk
, &mreqs
,
1074 greqs
.gsr_interface
);
1077 case MCAST_MSFILTER
:
1079 struct sockaddr_in
*psin
;
1080 struct ip_msfilter
*msf
= NULL
;
1081 struct group_filter
*gsf
= NULL
;
1082 int msize
, i
, ifindex
;
1084 if (optlen
< GROUP_FILTER_SIZE(0))
1086 if (optlen
> sysctl_optmem_max
) {
1090 gsf
= memdup_user(optval
, optlen
);
1096 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1097 if (gsf
->gf_numsrc
>= 0x1ffffff ||
1098 gsf
->gf_numsrc
> net
->ipv4
.sysctl_igmp_max_msf
) {
1102 if (GROUP_FILTER_SIZE(gsf
->gf_numsrc
) > optlen
) {
1106 msize
= IP_MSFILTER_SIZE(gsf
->gf_numsrc
);
1107 msf
= kmalloc(msize
, GFP_KERNEL
);
1112 ifindex
= gsf
->gf_interface
;
1113 psin
= (struct sockaddr_in
*)&gsf
->gf_group
;
1114 if (psin
->sin_family
!= AF_INET
) {
1115 err
= -EADDRNOTAVAIL
;
1118 msf
->imsf_multiaddr
= psin
->sin_addr
.s_addr
;
1119 msf
->imsf_interface
= 0;
1120 msf
->imsf_fmode
= gsf
->gf_fmode
;
1121 msf
->imsf_numsrc
= gsf
->gf_numsrc
;
1122 err
= -EADDRNOTAVAIL
;
1123 for (i
= 0; i
< gsf
->gf_numsrc
; ++i
) {
1124 psin
= (struct sockaddr_in
*)&gsf
->gf_slist
[i
];
1126 if (psin
->sin_family
!= AF_INET
)
1128 msf
->imsf_slist
[i
] = psin
->sin_addr
.s_addr
;
1133 err
= ip_mc_msfilter(sk
, msf
, ifindex
);
1139 case IP_MULTICAST_ALL
:
1142 if (val
!= 0 && val
!= 1)
1146 case IP_ROUTER_ALERT
:
1147 err
= ip_ra_control(sk
, val
? 1 : 0, NULL
);
1153 inet
->freebind
= !!val
;
1156 case IP_IPSEC_POLICY
:
1157 case IP_XFRM_POLICY
:
1159 if (!ns_capable(sock_net(sk
)->user_ns
, CAP_NET_ADMIN
))
1161 err
= xfrm_user_policy(sk
, optname
, optval
, optlen
);
1164 case IP_TRANSPARENT
:
1165 if (!!val
&& !ns_capable(sock_net(sk
)->user_ns
, CAP_NET_RAW
) &&
1166 !ns_capable(sock_net(sk
)->user_ns
, CAP_NET_ADMIN
)) {
1172 inet
->transparent
= !!val
;
1178 if (val
< 0 || val
> 255)
1180 inet
->min_ttl
= val
;
1200 * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1204 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1205 * destination in skb->cb[] before dst drop.
1206 * This way, receiver doesn't make cache line misses to read rtable.
1208 void ipv4_pktinfo_prepare(const struct sock
*sk
, struct sk_buff
*skb
)
1210 struct in_pktinfo
*pktinfo
= PKTINFO_SKB_CB(skb
);
1211 bool prepare
= (inet_sk(sk
)->cmsg_flags
& IP_CMSG_PKTINFO
) ||
1214 if (prepare
&& skb_rtable(skb
)) {
1215 /* skb->cb is overloaded: prior to this point it is IP{6}CB
1216 * which has interface index (iif) as the first member of the
1217 * underlying inet{6}_skb_parm struct. This code then overlays
1218 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1219 * element so the iif is picked up from the prior IPCB. If iif
1220 * is the loopback interface, then return the sending interface
1221 * (e.g., process binds socket to eth0 for Tx which is
1222 * redirected to loopback in the rtable/dst).
1224 struct rtable
*rt
= skb_rtable(skb
);
1225 bool l3slave
= ipv4_l3mdev_skb(IPCB(skb
)->flags
);
1227 if (pktinfo
->ipi_ifindex
== LOOPBACK_IFINDEX
)
1228 pktinfo
->ipi_ifindex
= inet_iif(skb
);
1229 else if (l3slave
&& rt
&& rt
->rt_iif
)
1230 pktinfo
->ipi_ifindex
= rt
->rt_iif
;
1232 pktinfo
->ipi_spec_dst
.s_addr
= fib_compute_spec_dst(skb
);
1234 pktinfo
->ipi_ifindex
= 0;
1235 pktinfo
->ipi_spec_dst
.s_addr
= 0;
1240 int ip_setsockopt(struct sock
*sk
, int level
,
1241 int optname
, char __user
*optval
, unsigned int optlen
)
1245 if (level
!= SOL_IP
)
1246 return -ENOPROTOOPT
;
1248 err
= do_ip_setsockopt(sk
, level
, optname
, optval
, optlen
);
1249 #ifdef CONFIG_NETFILTER
1250 /* we need to exclude all possible ENOPROTOOPTs except default case */
1251 if (err
== -ENOPROTOOPT
&& optname
!= IP_HDRINCL
&&
1252 optname
!= IP_IPSEC_POLICY
&&
1253 optname
!= IP_XFRM_POLICY
&&
1254 !ip_mroute_opt(optname
))
1255 err
= nf_setsockopt(sk
, PF_INET
, optname
, optval
, optlen
);
1259 EXPORT_SYMBOL(ip_setsockopt
);
1261 #ifdef CONFIG_COMPAT
1262 int compat_ip_setsockopt(struct sock
*sk
, int level
, int optname
,
1263 char __user
*optval
, unsigned int optlen
)
1267 if (level
!= SOL_IP
)
1268 return -ENOPROTOOPT
;
1270 if (optname
>= MCAST_JOIN_GROUP
&& optname
<= MCAST_MSFILTER
)
1271 return compat_mc_setsockopt(sk
, level
, optname
, optval
, optlen
,
1274 err
= do_ip_setsockopt(sk
, level
, optname
, optval
, optlen
);
1275 #ifdef CONFIG_NETFILTER
1276 /* we need to exclude all possible ENOPROTOOPTs except default case */
1277 if (err
== -ENOPROTOOPT
&& optname
!= IP_HDRINCL
&&
1278 optname
!= IP_IPSEC_POLICY
&&
1279 optname
!= IP_XFRM_POLICY
&&
1280 !ip_mroute_opt(optname
))
1281 err
= compat_nf_setsockopt(sk
, PF_INET
, optname
, optval
,
1286 EXPORT_SYMBOL(compat_ip_setsockopt
);
1290 * Get the options. Note for future reference. The GET of IP options gets
1291 * the _received_ ones. The set sets the _sent_ ones.
1294 static bool getsockopt_needs_rtnl(int optname
)
1298 case MCAST_MSFILTER
:
1304 static int do_ip_getsockopt(struct sock
*sk
, int level
, int optname
,
1305 char __user
*optval
, int __user
*optlen
, unsigned int flags
)
1307 struct inet_sock
*inet
= inet_sk(sk
);
1308 bool needs_rtnl
= getsockopt_needs_rtnl(optname
);
1312 if (level
!= SOL_IP
)
1315 if (ip_mroute_opt(optname
))
1316 return ip_mroute_getsockopt(sk
, optname
, optval
, optlen
);
1318 if (get_user(len
, optlen
))
1330 unsigned char optbuf
[sizeof(struct ip_options
)+40];
1331 struct ip_options
*opt
= (struct ip_options
*)optbuf
;
1332 struct ip_options_rcu
*inet_opt
;
1334 inet_opt
= rcu_dereference_protected(inet
->inet_opt
,
1335 lockdep_sock_is_held(sk
));
1338 memcpy(optbuf
, &inet_opt
->opt
,
1339 sizeof(struct ip_options
) +
1340 inet_opt
->opt
.optlen
);
1343 if (opt
->optlen
== 0)
1344 return put_user(0, optlen
);
1346 ip_options_undo(opt
);
1348 len
= min_t(unsigned int, len
, opt
->optlen
);
1349 if (put_user(len
, optlen
))
1351 if (copy_to_user(optval
, opt
->__data
, len
))
1356 val
= (inet
->cmsg_flags
& IP_CMSG_PKTINFO
) != 0;
1359 val
= (inet
->cmsg_flags
& IP_CMSG_TTL
) != 0;
1362 val
= (inet
->cmsg_flags
& IP_CMSG_TOS
) != 0;
1365 val
= (inet
->cmsg_flags
& IP_CMSG_RECVOPTS
) != 0;
1368 val
= (inet
->cmsg_flags
& IP_CMSG_RETOPTS
) != 0;
1371 val
= (inet
->cmsg_flags
& IP_CMSG_PASSSEC
) != 0;
1373 case IP_RECVORIGDSTADDR
:
1374 val
= (inet
->cmsg_flags
& IP_CMSG_ORIGDSTADDR
) != 0;
1377 val
= (inet
->cmsg_flags
& IP_CMSG_CHECKSUM
) != 0;
1379 case IP_RECVFRAGSIZE
:
1380 val
= (inet
->cmsg_flags
& IP_CMSG_RECVFRAGSIZE
) != 0;
1387 struct net
*net
= sock_net(sk
);
1388 val
= (inet
->uc_ttl
== -1 ?
1389 net
->ipv4
.sysctl_ip_default_ttl
:
1394 val
= inet
->hdrincl
;
1397 val
= inet
->nodefrag
;
1399 case IP_BIND_ADDRESS_NO_PORT
:
1400 val
= inet
->bind_address_no_port
;
1402 case IP_MTU_DISCOVER
:
1403 val
= inet
->pmtudisc
;
1407 struct dst_entry
*dst
;
1409 dst
= sk_dst_get(sk
);
1421 val
= inet
->recverr
;
1423 case IP_MULTICAST_TTL
:
1426 case IP_MULTICAST_LOOP
:
1427 val
= inet
->mc_loop
;
1430 val
= (__force
int)htonl((__u32
) inet
->uc_index
);
1432 case IP_MULTICAST_IF
:
1434 struct in_addr addr
;
1435 len
= min_t(unsigned int, len
, sizeof(struct in_addr
));
1436 addr
.s_addr
= inet
->mc_addr
;
1439 if (put_user(len
, optlen
))
1441 if (copy_to_user(optval
, &addr
, len
))
1447 struct ip_msfilter msf
;
1449 if (len
< IP_MSFILTER_SIZE(0)) {
1453 if (copy_from_user(&msf
, optval
, IP_MSFILTER_SIZE(0))) {
1457 err
= ip_mc_msfget(sk
, &msf
,
1458 (struct ip_msfilter __user
*)optval
, optlen
);
1461 case MCAST_MSFILTER
:
1463 struct group_filter gsf
;
1465 if (len
< GROUP_FILTER_SIZE(0)) {
1469 if (copy_from_user(&gsf
, optval
, GROUP_FILTER_SIZE(0))) {
1473 err
= ip_mc_gsfget(sk
, &gsf
,
1474 (struct group_filter __user
*)optval
,
1478 case IP_MULTICAST_ALL
:
1487 if (sk
->sk_type
!= SOCK_STREAM
)
1488 return -ENOPROTOOPT
;
1490 msg
.msg_control
= (__force
void *) optval
;
1491 msg
.msg_controllen
= len
;
1492 msg
.msg_flags
= flags
;
1494 if (inet
->cmsg_flags
& IP_CMSG_PKTINFO
) {
1495 struct in_pktinfo info
;
1497 info
.ipi_addr
.s_addr
= inet
->inet_rcv_saddr
;
1498 info
.ipi_spec_dst
.s_addr
= inet
->inet_rcv_saddr
;
1499 info
.ipi_ifindex
= inet
->mc_index
;
1500 put_cmsg(&msg
, SOL_IP
, IP_PKTINFO
, sizeof(info
), &info
);
1502 if (inet
->cmsg_flags
& IP_CMSG_TTL
) {
1503 int hlim
= inet
->mc_ttl
;
1504 put_cmsg(&msg
, SOL_IP
, IP_TTL
, sizeof(hlim
), &hlim
);
1506 if (inet
->cmsg_flags
& IP_CMSG_TOS
) {
1507 int tos
= inet
->rcv_tos
;
1508 put_cmsg(&msg
, SOL_IP
, IP_TOS
, sizeof(tos
), &tos
);
1510 len
-= msg
.msg_controllen
;
1511 return put_user(len
, optlen
);
1514 val
= inet
->freebind
;
1516 case IP_TRANSPARENT
:
1517 val
= inet
->transparent
;
1520 val
= inet
->min_ttl
;
1524 return -ENOPROTOOPT
;
1528 if (len
< sizeof(int) && len
> 0 && val
>= 0 && val
<= 255) {
1529 unsigned char ucval
= (unsigned char)val
;
1531 if (put_user(len
, optlen
))
1533 if (copy_to_user(optval
, &ucval
, 1))
1536 len
= min_t(unsigned int, sizeof(int), len
);
1537 if (put_user(len
, optlen
))
1539 if (copy_to_user(optval
, &val
, len
))
1551 int ip_getsockopt(struct sock
*sk
, int level
,
1552 int optname
, char __user
*optval
, int __user
*optlen
)
1556 err
= do_ip_getsockopt(sk
, level
, optname
, optval
, optlen
, 0);
1557 #ifdef CONFIG_NETFILTER
1558 /* we need to exclude all possible ENOPROTOOPTs except default case */
1559 if (err
== -ENOPROTOOPT
&& optname
!= IP_PKTOPTIONS
&&
1560 !ip_mroute_opt(optname
)) {
1563 if (get_user(len
, optlen
))
1566 err
= nf_getsockopt(sk
, PF_INET
, optname
, optval
, &len
);
1568 err
= put_user(len
, optlen
);
1574 EXPORT_SYMBOL(ip_getsockopt
);
1576 #ifdef CONFIG_COMPAT
1577 int compat_ip_getsockopt(struct sock
*sk
, int level
, int optname
,
1578 char __user
*optval
, int __user
*optlen
)
1582 if (optname
== MCAST_MSFILTER
)
1583 return compat_mc_getsockopt(sk
, level
, optname
, optval
, optlen
,
1586 err
= do_ip_getsockopt(sk
, level
, optname
, optval
, optlen
,
1589 #ifdef CONFIG_NETFILTER
1590 /* we need to exclude all possible ENOPROTOOPTs except default case */
1591 if (err
== -ENOPROTOOPT
&& optname
!= IP_PKTOPTIONS
&&
1592 !ip_mroute_opt(optname
)) {
1595 if (get_user(len
, optlen
))
1598 err
= compat_nf_getsockopt(sk
, PF_INET
, optname
, optval
, &len
);
1600 err
= put_user(len
, optlen
);
1606 EXPORT_SYMBOL(compat_ip_getsockopt
);