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.
11 * Many : Split from ip.c , see ip.c for history.
12 * Martin Mares : TOS setting fixed.
13 * Alan Cox : Fixed a couple of oopses in Martin's
15 * Mike McLagan : Routing by source
18 #include <linux/module.h>
19 #include <linux/types.h>
21 #include <linux/skbuff.h>
23 #include <linux/icmp.h>
24 #include <linux/inetdevice.h>
25 #include <linux/netdevice.h>
26 #include <linux/slab.h>
30 #include <net/tcp_states.h>
31 #include <linux/udp.h>
32 #include <linux/igmp.h>
33 #include <linux/netfilter.h>
34 #include <linux/route.h>
35 #include <linux/mroute.h>
36 #include <net/inet_ecn.h>
37 #include <net/route.h>
39 #include <net/compat.h>
40 #if IS_ENABLED(CONFIG_IPV6)
41 #include <net/transp_v6.h>
43 #include <net/ip_fib.h>
45 #include <linux/errqueue.h>
46 #include <asm/uaccess.h>
48 #define IP_CMSG_PKTINFO 1
51 #define IP_CMSG_RECVOPTS 8
52 #define IP_CMSG_RETOPTS 16
53 #define IP_CMSG_PASSSEC 32
54 #define IP_CMSG_ORIGDSTADDR 64
57 * SOL_IP control messages.
60 static void ip_cmsg_recv_pktinfo(struct msghdr
*msg
, struct sk_buff
*skb
)
62 struct in_pktinfo info
= *PKTINFO_SKB_CB(skb
);
64 info
.ipi_addr
.s_addr
= ip_hdr(skb
)->daddr
;
66 put_cmsg(msg
, SOL_IP
, IP_PKTINFO
, sizeof(info
), &info
);
69 static void ip_cmsg_recv_ttl(struct msghdr
*msg
, struct sk_buff
*skb
)
71 int ttl
= ip_hdr(skb
)->ttl
;
72 put_cmsg(msg
, SOL_IP
, IP_TTL
, sizeof(int), &ttl
);
75 static void ip_cmsg_recv_tos(struct msghdr
*msg
, struct sk_buff
*skb
)
77 put_cmsg(msg
, SOL_IP
, IP_TOS
, 1, &ip_hdr(skb
)->tos
);
80 static void ip_cmsg_recv_opts(struct msghdr
*msg
, struct sk_buff
*skb
)
82 if (IPCB(skb
)->opt
.optlen
== 0)
85 put_cmsg(msg
, SOL_IP
, IP_RECVOPTS
, IPCB(skb
)->opt
.optlen
,
90 static void ip_cmsg_recv_retopts(struct msghdr
*msg
, struct sk_buff
*skb
)
92 unsigned char optbuf
[sizeof(struct ip_options
) + 40];
93 struct ip_options
*opt
= (struct ip_options
*)optbuf
;
95 if (IPCB(skb
)->opt
.optlen
== 0)
98 if (ip_options_echo(opt
, skb
)) {
99 msg
->msg_flags
|= MSG_CTRUNC
;
102 ip_options_undo(opt
);
104 put_cmsg(msg
, SOL_IP
, IP_RETOPTS
, opt
->optlen
, opt
->__data
);
107 static void ip_cmsg_recv_security(struct msghdr
*msg
, struct sk_buff
*skb
)
113 err
= security_socket_getpeersec_dgram(NULL
, skb
, &secid
);
117 err
= security_secid_to_secctx(secid
, &secdata
, &seclen
);
121 put_cmsg(msg
, SOL_IP
, SCM_SECURITY
, seclen
, secdata
);
122 security_release_secctx(secdata
, seclen
);
125 static void ip_cmsg_recv_dstaddr(struct msghdr
*msg
, struct sk_buff
*skb
)
127 struct sockaddr_in sin
;
128 const struct iphdr
*iph
= ip_hdr(skb
);
129 __be16
*ports
= (__be16
*)skb_transport_header(skb
);
131 if (skb_transport_offset(skb
) + 4 > skb
->len
)
134 /* All current transport protocols have the port numbers in the
135 * first four bytes of the transport header and this function is
136 * written with this assumption in mind.
139 sin
.sin_family
= AF_INET
;
140 sin
.sin_addr
.s_addr
= iph
->daddr
;
141 sin
.sin_port
= ports
[1];
142 memset(sin
.sin_zero
, 0, sizeof(sin
.sin_zero
));
144 put_cmsg(msg
, SOL_IP
, IP_ORIGDSTADDR
, sizeof(sin
), &sin
);
147 void ip_cmsg_recv(struct msghdr
*msg
, struct sk_buff
*skb
)
149 struct inet_sock
*inet
= inet_sk(skb
->sk
);
150 unsigned int flags
= inet
->cmsg_flags
;
152 /* Ordered by supposed usage frequency */
154 ip_cmsg_recv_pktinfo(msg
, skb
);
155 if ((flags
>>= 1) == 0)
159 ip_cmsg_recv_ttl(msg
, skb
);
160 if ((flags
>>= 1) == 0)
164 ip_cmsg_recv_tos(msg
, skb
);
165 if ((flags
>>= 1) == 0)
169 ip_cmsg_recv_opts(msg
, skb
);
170 if ((flags
>>= 1) == 0)
174 ip_cmsg_recv_retopts(msg
, skb
);
175 if ((flags
>>= 1) == 0)
179 ip_cmsg_recv_security(msg
, skb
);
181 if ((flags
>>= 1) == 0)
184 ip_cmsg_recv_dstaddr(msg
, skb
);
187 EXPORT_SYMBOL(ip_cmsg_recv
);
189 int ip_cmsg_send(struct net
*net
, struct msghdr
*msg
, struct ipcm_cookie
*ipc
,
193 struct cmsghdr
*cmsg
;
195 for (cmsg
= CMSG_FIRSTHDR(msg
); cmsg
; cmsg
= CMSG_NXTHDR(msg
, cmsg
)) {
196 if (!CMSG_OK(msg
, cmsg
))
198 #if defined(CONFIG_IPV6)
200 cmsg
->cmsg_level
== SOL_IPV6
&&
201 cmsg
->cmsg_type
== IPV6_PKTINFO
) {
202 struct in6_pktinfo
*src_info
;
204 if (cmsg
->cmsg_len
< CMSG_LEN(sizeof(*src_info
)))
206 src_info
= (struct in6_pktinfo
*)CMSG_DATA(cmsg
);
207 if (!ipv6_addr_v4mapped(&src_info
->ipi6_addr
))
209 ipc
->oif
= src_info
->ipi6_ifindex
;
210 ipc
->addr
= src_info
->ipi6_addr
.s6_addr32
[3];
214 if (cmsg
->cmsg_level
!= SOL_IP
)
216 switch (cmsg
->cmsg_type
) {
218 err
= cmsg
->cmsg_len
- CMSG_ALIGN(sizeof(struct cmsghdr
));
219 err
= ip_options_get(net
, &ipc
->opt
, CMSG_DATA(cmsg
),
220 err
< 40 ? err
: 40);
226 struct in_pktinfo
*info
;
227 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct in_pktinfo
)))
229 info
= (struct in_pktinfo
*)CMSG_DATA(cmsg
);
230 ipc
->oif
= info
->ipi_ifindex
;
231 ipc
->addr
= info
->ipi_spec_dst
.s_addr
;
235 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(int)))
237 val
= *(int *)CMSG_DATA(cmsg
);
238 if (val
< 1 || val
> 255)
243 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(int)))
245 val
= *(int *)CMSG_DATA(cmsg
);
246 if (val
< 0 || val
> 255)
249 ipc
->priority
= rt_tos2priority(ipc
->tos
);
260 /* Special input handler for packets caught by router alert option.
261 They are selected only by protocol field, and then processed likely
262 local ones; but only if someone wants them! Otherwise, router
263 not running rsvpd will kill RSVP.
265 It is user level problem, what it will make with them.
266 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
267 but receiver should be enough clever f.e. to forward mtrace requests,
268 sent to multicast group to reach destination designated router.
270 struct ip_ra_chain __rcu
*ip_ra_chain
;
271 static DEFINE_SPINLOCK(ip_ra_lock
);
274 static void ip_ra_destroy_rcu(struct rcu_head
*head
)
276 struct ip_ra_chain
*ra
= container_of(head
, struct ip_ra_chain
, rcu
);
278 sock_put(ra
->saved_sk
);
282 int ip_ra_control(struct sock
*sk
, unsigned char on
,
283 void (*destructor
)(struct sock
*))
285 struct ip_ra_chain
*ra
, *new_ra
;
286 struct ip_ra_chain __rcu
**rap
;
288 if (sk
->sk_type
!= SOCK_RAW
|| inet_sk(sk
)->inet_num
== IPPROTO_RAW
)
291 new_ra
= on
? kmalloc(sizeof(*new_ra
), GFP_KERNEL
) : NULL
;
293 spin_lock_bh(&ip_ra_lock
);
294 for (rap
= &ip_ra_chain
;
295 (ra
= rcu_dereference_protected(*rap
,
296 lockdep_is_held(&ip_ra_lock
))) != NULL
;
300 spin_unlock_bh(&ip_ra_lock
);
304 /* dont let ip_call_ra_chain() use sk again */
306 RCU_INIT_POINTER(*rap
, ra
->next
);
307 spin_unlock_bh(&ip_ra_lock
);
312 * Delay sock_put(sk) and kfree(ra) after one rcu grace
313 * period. This guarantee ip_call_ra_chain() dont need
314 * to mess with socket refcounts.
317 call_rcu(&ra
->rcu
, ip_ra_destroy_rcu
);
321 if (new_ra
== NULL
) {
322 spin_unlock_bh(&ip_ra_lock
);
326 new_ra
->destructor
= destructor
;
328 RCU_INIT_POINTER(new_ra
->next
, ra
);
329 rcu_assign_pointer(*rap
, new_ra
);
331 spin_unlock_bh(&ip_ra_lock
);
336 void ip_icmp_error(struct sock
*sk
, struct sk_buff
*skb
, int err
,
337 __be16 port
, u32 info
, u8
*payload
)
339 struct sock_exterr_skb
*serr
;
341 skb
= skb_clone(skb
, GFP_ATOMIC
);
345 serr
= SKB_EXT_ERR(skb
);
346 serr
->ee
.ee_errno
= err
;
347 serr
->ee
.ee_origin
= SO_EE_ORIGIN_ICMP
;
348 serr
->ee
.ee_type
= icmp_hdr(skb
)->type
;
349 serr
->ee
.ee_code
= icmp_hdr(skb
)->code
;
351 serr
->ee
.ee_info
= info
;
352 serr
->ee
.ee_data
= 0;
353 serr
->addr_offset
= (u8
*)&(((struct iphdr
*)(icmp_hdr(skb
) + 1))->daddr
) -
354 skb_network_header(skb
);
357 if (skb_pull(skb
, payload
- skb
->data
) != NULL
) {
358 skb_reset_transport_header(skb
);
359 if (sock_queue_err_skb(sk
, skb
) == 0)
365 void ip_local_error(struct sock
*sk
, int err
, __be32 daddr
, __be16 port
, u32 info
)
367 struct inet_sock
*inet
= inet_sk(sk
);
368 struct sock_exterr_skb
*serr
;
375 skb
= alloc_skb(sizeof(struct iphdr
), GFP_ATOMIC
);
379 skb_put(skb
, sizeof(struct iphdr
));
380 skb_reset_network_header(skb
);
384 serr
= SKB_EXT_ERR(skb
);
385 serr
->ee
.ee_errno
= err
;
386 serr
->ee
.ee_origin
= SO_EE_ORIGIN_LOCAL
;
387 serr
->ee
.ee_type
= 0;
388 serr
->ee
.ee_code
= 0;
390 serr
->ee
.ee_info
= info
;
391 serr
->ee
.ee_data
= 0;
392 serr
->addr_offset
= (u8
*)&iph
->daddr
- skb_network_header(skb
);
395 __skb_pull(skb
, skb_tail_pointer(skb
) - skb
->data
);
396 skb_reset_transport_header(skb
);
398 if (sock_queue_err_skb(sk
, skb
))
403 * Handle MSG_ERRQUEUE
405 int ip_recv_error(struct sock
*sk
, struct msghdr
*msg
, int len
, int *addr_len
)
407 struct sock_exterr_skb
*serr
;
409 DECLARE_SOCKADDR(struct sockaddr_in
*, sin
, msg
->msg_name
);
411 struct sock_extended_err ee
;
412 struct sockaddr_in offender
;
418 skb
= sock_dequeue_err_skb(sk
);
424 msg
->msg_flags
|= MSG_TRUNC
;
427 err
= skb_copy_datagram_msg(skb
, 0, msg
, copied
);
431 sock_recv_timestamp(msg
, sk
, skb
);
433 serr
= SKB_EXT_ERR(skb
);
436 sin
->sin_family
= AF_INET
;
437 sin
->sin_addr
.s_addr
= *(__be32
*)(skb_network_header(skb
) +
439 sin
->sin_port
= serr
->port
;
440 memset(&sin
->sin_zero
, 0, sizeof(sin
->sin_zero
));
441 *addr_len
= sizeof(*sin
);
444 memcpy(&errhdr
.ee
, &serr
->ee
, sizeof(struct sock_extended_err
));
445 sin
= &errhdr
.offender
;
446 sin
->sin_family
= AF_UNSPEC
;
447 if (serr
->ee
.ee_origin
== SO_EE_ORIGIN_ICMP
) {
448 struct inet_sock
*inet
= inet_sk(sk
);
450 sin
->sin_family
= AF_INET
;
451 sin
->sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
453 memset(&sin
->sin_zero
, 0, sizeof(sin
->sin_zero
));
454 if (inet
->cmsg_flags
)
455 ip_cmsg_recv(msg
, skb
);
458 put_cmsg(msg
, SOL_IP
, IP_RECVERR
, sizeof(errhdr
), &errhdr
);
460 /* Now we could try to dump offended packet options */
462 msg
->msg_flags
|= MSG_ERRQUEUE
;
473 * Socket option code for IP. This is the end of the line after any
474 * TCP,UDP etc options on an IP socket.
477 static int do_ip_setsockopt(struct sock
*sk
, int level
,
478 int optname
, char __user
*optval
, unsigned int optlen
)
480 struct inet_sock
*inet
= inet_sk(sk
);
492 case IP_MTU_DISCOVER
:
494 case IP_ROUTER_ALERT
:
501 case IP_MULTICAST_TTL
:
502 case IP_MULTICAST_ALL
:
503 case IP_MULTICAST_LOOP
:
504 case IP_RECVORIGDSTADDR
:
505 if (optlen
>= sizeof(int)) {
506 if (get_user(val
, (int __user
*) optval
))
508 } else if (optlen
>= sizeof(char)) {
511 if (get_user(ucval
, (unsigned char __user
*) optval
))
517 /* If optlen==0, it is equivalent to val == 0 */
519 if (ip_mroute_opt(optname
))
520 return ip_mroute_setsockopt(sk
, optname
, optval
, optlen
);
528 struct ip_options_rcu
*old
, *opt
= NULL
;
532 err
= ip_options_get_from_user(sock_net(sk
), &opt
,
536 old
= rcu_dereference_protected(inet
->inet_opt
,
537 sock_owned_by_user(sk
));
539 struct inet_connection_sock
*icsk
= inet_csk(sk
);
540 #if IS_ENABLED(CONFIG_IPV6)
541 if (sk
->sk_family
== PF_INET
||
542 (!((1 << sk
->sk_state
) &
543 (TCPF_LISTEN
| TCPF_CLOSE
)) &&
544 inet
->inet_daddr
!= LOOPBACK4_IPV6
)) {
547 icsk
->icsk_ext_hdr_len
-= old
->opt
.optlen
;
549 icsk
->icsk_ext_hdr_len
+= opt
->opt
.optlen
;
550 icsk
->icsk_sync_mss(sk
, icsk
->icsk_pmtu_cookie
);
551 #if IS_ENABLED(CONFIG_IPV6)
555 rcu_assign_pointer(inet
->inet_opt
, opt
);
562 inet
->cmsg_flags
|= IP_CMSG_PKTINFO
;
564 inet
->cmsg_flags
&= ~IP_CMSG_PKTINFO
;
568 inet
->cmsg_flags
|= IP_CMSG_TTL
;
570 inet
->cmsg_flags
&= ~IP_CMSG_TTL
;
574 inet
->cmsg_flags
|= IP_CMSG_TOS
;
576 inet
->cmsg_flags
&= ~IP_CMSG_TOS
;
580 inet
->cmsg_flags
|= IP_CMSG_RECVOPTS
;
582 inet
->cmsg_flags
&= ~IP_CMSG_RECVOPTS
;
586 inet
->cmsg_flags
|= IP_CMSG_RETOPTS
;
588 inet
->cmsg_flags
&= ~IP_CMSG_RETOPTS
;
592 inet
->cmsg_flags
|= IP_CMSG_PASSSEC
;
594 inet
->cmsg_flags
&= ~IP_CMSG_PASSSEC
;
596 case IP_RECVORIGDSTADDR
:
598 inet
->cmsg_flags
|= IP_CMSG_ORIGDSTADDR
;
600 inet
->cmsg_flags
&= ~IP_CMSG_ORIGDSTADDR
;
602 case IP_TOS
: /* This sets both TOS and Precedence */
603 if (sk
->sk_type
== SOCK_STREAM
) {
604 val
&= ~INET_ECN_MASK
;
605 val
|= inet
->tos
& INET_ECN_MASK
;
607 if (inet
->tos
!= val
) {
609 sk
->sk_priority
= rt_tos2priority(val
);
616 if (val
!= -1 && (val
< 1 || val
> 255))
621 if (sk
->sk_type
!= SOCK_RAW
) {
625 inet
->hdrincl
= val
? 1 : 0;
628 if (sk
->sk_type
!= SOCK_RAW
) {
632 inet
->nodefrag
= val
? 1 : 0;
634 case IP_MTU_DISCOVER
:
635 if (val
< IP_PMTUDISC_DONT
|| val
> IP_PMTUDISC_OMIT
)
637 inet
->pmtudisc
= val
;
640 inet
->recverr
= !!val
;
642 skb_queue_purge(&sk
->sk_error_queue
);
644 case IP_MULTICAST_TTL
:
645 if (sk
->sk_type
== SOCK_STREAM
)
651 if (val
< 0 || val
> 255)
655 case IP_MULTICAST_LOOP
:
658 inet
->mc_loop
= !!val
;
662 struct net_device
*dev
= NULL
;
665 if (optlen
!= sizeof(int))
668 ifindex
= (__force
int)ntohl((__force __be32
)val
);
675 dev
= dev_get_by_index(sock_net(sk
), ifindex
);
676 err
= -EADDRNOTAVAIL
;
682 if (sk
->sk_bound_dev_if
)
685 inet
->uc_index
= ifindex
;
689 case IP_MULTICAST_IF
:
691 struct ip_mreqn mreq
;
692 struct net_device
*dev
= NULL
;
694 if (sk
->sk_type
== SOCK_STREAM
)
697 * Check the arguments are allowable
700 if (optlen
< sizeof(struct in_addr
))
704 if (optlen
>= sizeof(struct ip_mreqn
)) {
705 if (copy_from_user(&mreq
, optval
, sizeof(mreq
)))
708 memset(&mreq
, 0, sizeof(mreq
));
709 if (optlen
>= sizeof(struct ip_mreq
)) {
710 if (copy_from_user(&mreq
, optval
,
711 sizeof(struct ip_mreq
)))
713 } else if (optlen
>= sizeof(struct in_addr
)) {
714 if (copy_from_user(&mreq
.imr_address
, optval
,
715 sizeof(struct in_addr
)))
720 if (!mreq
.imr_ifindex
) {
721 if (mreq
.imr_address
.s_addr
== htonl(INADDR_ANY
)) {
727 dev
= ip_dev_find(sock_net(sk
), mreq
.imr_address
.s_addr
);
729 mreq
.imr_ifindex
= dev
->ifindex
;
731 dev
= dev_get_by_index(sock_net(sk
), mreq
.imr_ifindex
);
734 err
= -EADDRNOTAVAIL
;
740 if (sk
->sk_bound_dev_if
&&
741 mreq
.imr_ifindex
!= sk
->sk_bound_dev_if
)
744 inet
->mc_index
= mreq
.imr_ifindex
;
745 inet
->mc_addr
= mreq
.imr_address
.s_addr
;
750 case IP_ADD_MEMBERSHIP
:
751 case IP_DROP_MEMBERSHIP
:
753 struct ip_mreqn mreq
;
756 if (inet_sk(sk
)->is_icsk
)
759 if (optlen
< sizeof(struct ip_mreq
))
762 if (optlen
>= sizeof(struct ip_mreqn
)) {
763 if (copy_from_user(&mreq
, optval
, sizeof(mreq
)))
766 memset(&mreq
, 0, sizeof(mreq
));
767 if (copy_from_user(&mreq
, optval
, sizeof(struct ip_mreq
)))
771 if (optname
== IP_ADD_MEMBERSHIP
)
772 err
= ip_mc_join_group(sk
, &mreq
);
774 err
= ip_mc_leave_group(sk
, &mreq
);
779 struct ip_msfilter
*msf
;
781 if (optlen
< IP_MSFILTER_SIZE(0))
783 if (optlen
> sysctl_optmem_max
) {
787 msf
= kmalloc(optlen
, GFP_KERNEL
);
793 if (copy_from_user(msf
, optval
, optlen
)) {
797 /* numsrc >= (1G-4) overflow in 32 bits */
798 if (msf
->imsf_numsrc
>= 0x3ffffffcU
||
799 msf
->imsf_numsrc
> sysctl_igmp_max_msf
) {
804 if (IP_MSFILTER_SIZE(msf
->imsf_numsrc
) > optlen
) {
809 err
= ip_mc_msfilter(sk
, msf
, 0);
813 case IP_BLOCK_SOURCE
:
814 case IP_UNBLOCK_SOURCE
:
815 case IP_ADD_SOURCE_MEMBERSHIP
:
816 case IP_DROP_SOURCE_MEMBERSHIP
:
818 struct ip_mreq_source mreqs
;
821 if (optlen
!= sizeof(struct ip_mreq_source
))
823 if (copy_from_user(&mreqs
, optval
, sizeof(mreqs
))) {
827 if (optname
== IP_BLOCK_SOURCE
) {
828 omode
= MCAST_EXCLUDE
;
830 } else if (optname
== IP_UNBLOCK_SOURCE
) {
831 omode
= MCAST_EXCLUDE
;
833 } else if (optname
== IP_ADD_SOURCE_MEMBERSHIP
) {
834 struct ip_mreqn mreq
;
836 mreq
.imr_multiaddr
.s_addr
= mreqs
.imr_multiaddr
;
837 mreq
.imr_address
.s_addr
= mreqs
.imr_interface
;
838 mreq
.imr_ifindex
= 0;
839 err
= ip_mc_join_group(sk
, &mreq
);
840 if (err
&& err
!= -EADDRINUSE
)
842 omode
= MCAST_INCLUDE
;
844 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
845 omode
= MCAST_INCLUDE
;
848 err
= ip_mc_source(add
, omode
, sk
, &mreqs
, 0);
851 case MCAST_JOIN_GROUP
:
852 case MCAST_LEAVE_GROUP
:
854 struct group_req greq
;
855 struct sockaddr_in
*psin
;
856 struct ip_mreqn mreq
;
858 if (optlen
< sizeof(struct group_req
))
861 if (copy_from_user(&greq
, optval
, sizeof(greq
)))
863 psin
= (struct sockaddr_in
*)&greq
.gr_group
;
864 if (psin
->sin_family
!= AF_INET
)
866 memset(&mreq
, 0, sizeof(mreq
));
867 mreq
.imr_multiaddr
= psin
->sin_addr
;
868 mreq
.imr_ifindex
= greq
.gr_interface
;
870 if (optname
== MCAST_JOIN_GROUP
)
871 err
= ip_mc_join_group(sk
, &mreq
);
873 err
= ip_mc_leave_group(sk
, &mreq
);
876 case MCAST_JOIN_SOURCE_GROUP
:
877 case MCAST_LEAVE_SOURCE_GROUP
:
878 case MCAST_BLOCK_SOURCE
:
879 case MCAST_UNBLOCK_SOURCE
:
881 struct group_source_req greqs
;
882 struct ip_mreq_source mreqs
;
883 struct sockaddr_in
*psin
;
886 if (optlen
!= sizeof(struct group_source_req
))
888 if (copy_from_user(&greqs
, optval
, sizeof(greqs
))) {
892 if (greqs
.gsr_group
.ss_family
!= AF_INET
||
893 greqs
.gsr_source
.ss_family
!= AF_INET
) {
894 err
= -EADDRNOTAVAIL
;
897 psin
= (struct sockaddr_in
*)&greqs
.gsr_group
;
898 mreqs
.imr_multiaddr
= psin
->sin_addr
.s_addr
;
899 psin
= (struct sockaddr_in
*)&greqs
.gsr_source
;
900 mreqs
.imr_sourceaddr
= psin
->sin_addr
.s_addr
;
901 mreqs
.imr_interface
= 0; /* use index for mc_source */
903 if (optname
== MCAST_BLOCK_SOURCE
) {
904 omode
= MCAST_EXCLUDE
;
906 } else if (optname
== MCAST_UNBLOCK_SOURCE
) {
907 omode
= MCAST_EXCLUDE
;
909 } else if (optname
== MCAST_JOIN_SOURCE_GROUP
) {
910 struct ip_mreqn mreq
;
912 psin
= (struct sockaddr_in
*)&greqs
.gsr_group
;
913 mreq
.imr_multiaddr
= psin
->sin_addr
;
914 mreq
.imr_address
.s_addr
= 0;
915 mreq
.imr_ifindex
= greqs
.gsr_interface
;
916 err
= ip_mc_join_group(sk
, &mreq
);
917 if (err
&& err
!= -EADDRINUSE
)
919 greqs
.gsr_interface
= mreq
.imr_ifindex
;
920 omode
= MCAST_INCLUDE
;
922 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
923 omode
= MCAST_INCLUDE
;
926 err
= ip_mc_source(add
, omode
, sk
, &mreqs
,
927 greqs
.gsr_interface
);
932 struct sockaddr_in
*psin
;
933 struct ip_msfilter
*msf
= NULL
;
934 struct group_filter
*gsf
= NULL
;
935 int msize
, i
, ifindex
;
937 if (optlen
< GROUP_FILTER_SIZE(0))
939 if (optlen
> sysctl_optmem_max
) {
943 gsf
= kmalloc(optlen
, GFP_KERNEL
);
949 if (copy_from_user(gsf
, optval
, optlen
))
952 /* numsrc >= (4G-140)/128 overflow in 32 bits */
953 if (gsf
->gf_numsrc
>= 0x1ffffff ||
954 gsf
->gf_numsrc
> sysctl_igmp_max_msf
) {
958 if (GROUP_FILTER_SIZE(gsf
->gf_numsrc
) > optlen
) {
962 msize
= IP_MSFILTER_SIZE(gsf
->gf_numsrc
);
963 msf
= kmalloc(msize
, GFP_KERNEL
);
968 ifindex
= gsf
->gf_interface
;
969 psin
= (struct sockaddr_in
*)&gsf
->gf_group
;
970 if (psin
->sin_family
!= AF_INET
) {
971 err
= -EADDRNOTAVAIL
;
974 msf
->imsf_multiaddr
= psin
->sin_addr
.s_addr
;
975 msf
->imsf_interface
= 0;
976 msf
->imsf_fmode
= gsf
->gf_fmode
;
977 msf
->imsf_numsrc
= gsf
->gf_numsrc
;
978 err
= -EADDRNOTAVAIL
;
979 for (i
= 0; i
< gsf
->gf_numsrc
; ++i
) {
980 psin
= (struct sockaddr_in
*)&gsf
->gf_slist
[i
];
982 if (psin
->sin_family
!= AF_INET
)
984 msf
->imsf_slist
[i
] = psin
->sin_addr
.s_addr
;
989 err
= ip_mc_msfilter(sk
, msf
, ifindex
);
995 case IP_MULTICAST_ALL
:
998 if (val
!= 0 && val
!= 1)
1002 case IP_ROUTER_ALERT
:
1003 err
= ip_ra_control(sk
, val
? 1 : 0, NULL
);
1009 inet
->freebind
= !!val
;
1012 case IP_IPSEC_POLICY
:
1013 case IP_XFRM_POLICY
:
1015 if (!ns_capable(sock_net(sk
)->user_ns
, CAP_NET_ADMIN
))
1017 err
= xfrm_user_policy(sk
, optname
, optval
, optlen
);
1020 case IP_TRANSPARENT
:
1021 if (!!val
&& !ns_capable(sock_net(sk
)->user_ns
, CAP_NET_RAW
) &&
1022 !ns_capable(sock_net(sk
)->user_ns
, CAP_NET_ADMIN
)) {
1028 inet
->transparent
= !!val
;
1034 if (val
< 0 || val
> 255)
1036 inet
->min_ttl
= val
;
1052 * ipv4_pktinfo_prepare - transfert some info from rtable to skb
1056 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1057 * destination in skb->cb[] before dst drop.
1058 * This way, receiver doesn't make cache line misses to read rtable.
1060 void ipv4_pktinfo_prepare(const struct sock
*sk
, struct sk_buff
*skb
)
1062 struct in_pktinfo
*pktinfo
= PKTINFO_SKB_CB(skb
);
1063 bool prepare
= (inet_sk(sk
)->cmsg_flags
& IP_CMSG_PKTINFO
) ||
1066 if (prepare
&& skb_rtable(skb
)) {
1067 pktinfo
->ipi_ifindex
= inet_iif(skb
);
1068 pktinfo
->ipi_spec_dst
.s_addr
= fib_compute_spec_dst(skb
);
1070 pktinfo
->ipi_ifindex
= 0;
1071 pktinfo
->ipi_spec_dst
.s_addr
= 0;
1076 int ip_setsockopt(struct sock
*sk
, int level
,
1077 int optname
, char __user
*optval
, unsigned int optlen
)
1081 if (level
!= SOL_IP
)
1082 return -ENOPROTOOPT
;
1084 err
= do_ip_setsockopt(sk
, level
, optname
, optval
, optlen
);
1085 #ifdef CONFIG_NETFILTER
1086 /* we need to exclude all possible ENOPROTOOPTs except default case */
1087 if (err
== -ENOPROTOOPT
&& optname
!= IP_HDRINCL
&&
1088 optname
!= IP_IPSEC_POLICY
&&
1089 optname
!= IP_XFRM_POLICY
&&
1090 !ip_mroute_opt(optname
)) {
1092 err
= nf_setsockopt(sk
, PF_INET
, optname
, optval
, optlen
);
1098 EXPORT_SYMBOL(ip_setsockopt
);
1100 #ifdef CONFIG_COMPAT
1101 int compat_ip_setsockopt(struct sock
*sk
, int level
, int optname
,
1102 char __user
*optval
, unsigned int optlen
)
1106 if (level
!= SOL_IP
)
1107 return -ENOPROTOOPT
;
1109 if (optname
>= MCAST_JOIN_GROUP
&& optname
<= MCAST_MSFILTER
)
1110 return compat_mc_setsockopt(sk
, level
, optname
, optval
, optlen
,
1113 err
= do_ip_setsockopt(sk
, level
, optname
, optval
, optlen
);
1114 #ifdef CONFIG_NETFILTER
1115 /* we need to exclude all possible ENOPROTOOPTs except default case */
1116 if (err
== -ENOPROTOOPT
&& optname
!= IP_HDRINCL
&&
1117 optname
!= IP_IPSEC_POLICY
&&
1118 optname
!= IP_XFRM_POLICY
&&
1119 !ip_mroute_opt(optname
)) {
1121 err
= compat_nf_setsockopt(sk
, PF_INET
, optname
,
1128 EXPORT_SYMBOL(compat_ip_setsockopt
);
1132 * Get the options. Note for future reference. The GET of IP options gets
1133 * the _received_ ones. The set sets the _sent_ ones.
1136 static int do_ip_getsockopt(struct sock
*sk
, int level
, int optname
,
1137 char __user
*optval
, int __user
*optlen
, unsigned int flags
)
1139 struct inet_sock
*inet
= inet_sk(sk
);
1143 if (level
!= SOL_IP
)
1146 if (ip_mroute_opt(optname
))
1147 return ip_mroute_getsockopt(sk
, optname
, optval
, optlen
);
1149 if (get_user(len
, optlen
))
1159 unsigned char optbuf
[sizeof(struct ip_options
)+40];
1160 struct ip_options
*opt
= (struct ip_options
*)optbuf
;
1161 struct ip_options_rcu
*inet_opt
;
1163 inet_opt
= rcu_dereference_protected(inet
->inet_opt
,
1164 sock_owned_by_user(sk
));
1167 memcpy(optbuf
, &inet_opt
->opt
,
1168 sizeof(struct ip_options
) +
1169 inet_opt
->opt
.optlen
);
1172 if (opt
->optlen
== 0)
1173 return put_user(0, optlen
);
1175 ip_options_undo(opt
);
1177 len
= min_t(unsigned int, len
, opt
->optlen
);
1178 if (put_user(len
, optlen
))
1180 if (copy_to_user(optval
, opt
->__data
, len
))
1185 val
= (inet
->cmsg_flags
& IP_CMSG_PKTINFO
) != 0;
1188 val
= (inet
->cmsg_flags
& IP_CMSG_TTL
) != 0;
1191 val
= (inet
->cmsg_flags
& IP_CMSG_TOS
) != 0;
1194 val
= (inet
->cmsg_flags
& IP_CMSG_RECVOPTS
) != 0;
1197 val
= (inet
->cmsg_flags
& IP_CMSG_RETOPTS
) != 0;
1200 val
= (inet
->cmsg_flags
& IP_CMSG_PASSSEC
) != 0;
1202 case IP_RECVORIGDSTADDR
:
1203 val
= (inet
->cmsg_flags
& IP_CMSG_ORIGDSTADDR
) != 0;
1209 val
= (inet
->uc_ttl
== -1 ?
1210 sysctl_ip_default_ttl
:
1214 val
= inet
->hdrincl
;
1217 val
= inet
->nodefrag
;
1219 case IP_MTU_DISCOVER
:
1220 val
= inet
->pmtudisc
;
1224 struct dst_entry
*dst
;
1226 dst
= sk_dst_get(sk
);
1238 val
= inet
->recverr
;
1240 case IP_MULTICAST_TTL
:
1243 case IP_MULTICAST_LOOP
:
1244 val
= inet
->mc_loop
;
1247 val
= (__force
int)htonl((__u32
) inet
->uc_index
);
1249 case IP_MULTICAST_IF
:
1251 struct in_addr addr
;
1252 len
= min_t(unsigned int, len
, sizeof(struct in_addr
));
1253 addr
.s_addr
= inet
->mc_addr
;
1256 if (put_user(len
, optlen
))
1258 if (copy_to_user(optval
, &addr
, len
))
1264 struct ip_msfilter msf
;
1267 if (len
< IP_MSFILTER_SIZE(0)) {
1271 if (copy_from_user(&msf
, optval
, IP_MSFILTER_SIZE(0))) {
1275 err
= ip_mc_msfget(sk
, &msf
,
1276 (struct ip_msfilter __user
*)optval
, optlen
);
1280 case MCAST_MSFILTER
:
1282 struct group_filter gsf
;
1285 if (len
< GROUP_FILTER_SIZE(0)) {
1289 if (copy_from_user(&gsf
, optval
, GROUP_FILTER_SIZE(0))) {
1293 err
= ip_mc_gsfget(sk
, &gsf
,
1294 (struct group_filter __user
*)optval
,
1299 case IP_MULTICAST_ALL
:
1308 if (sk
->sk_type
!= SOCK_STREAM
)
1309 return -ENOPROTOOPT
;
1311 msg
.msg_control
= (__force
void *) optval
;
1312 msg
.msg_controllen
= len
;
1313 msg
.msg_flags
= flags
;
1315 if (inet
->cmsg_flags
& IP_CMSG_PKTINFO
) {
1316 struct in_pktinfo info
;
1318 info
.ipi_addr
.s_addr
= inet
->inet_rcv_saddr
;
1319 info
.ipi_spec_dst
.s_addr
= inet
->inet_rcv_saddr
;
1320 info
.ipi_ifindex
= inet
->mc_index
;
1321 put_cmsg(&msg
, SOL_IP
, IP_PKTINFO
, sizeof(info
), &info
);
1323 if (inet
->cmsg_flags
& IP_CMSG_TTL
) {
1324 int hlim
= inet
->mc_ttl
;
1325 put_cmsg(&msg
, SOL_IP
, IP_TTL
, sizeof(hlim
), &hlim
);
1327 if (inet
->cmsg_flags
& IP_CMSG_TOS
) {
1328 int tos
= inet
->rcv_tos
;
1329 put_cmsg(&msg
, SOL_IP
, IP_TOS
, sizeof(tos
), &tos
);
1331 len
-= msg
.msg_controllen
;
1332 return put_user(len
, optlen
);
1335 val
= inet
->freebind
;
1337 case IP_TRANSPARENT
:
1338 val
= inet
->transparent
;
1341 val
= inet
->min_ttl
;
1345 return -ENOPROTOOPT
;
1349 if (len
< sizeof(int) && len
> 0 && val
>= 0 && val
<= 255) {
1350 unsigned char ucval
= (unsigned char)val
;
1352 if (put_user(len
, optlen
))
1354 if (copy_to_user(optval
, &ucval
, 1))
1357 len
= min_t(unsigned int, sizeof(int), len
);
1358 if (put_user(len
, optlen
))
1360 if (copy_to_user(optval
, &val
, len
))
1366 int ip_getsockopt(struct sock
*sk
, int level
,
1367 int optname
, char __user
*optval
, int __user
*optlen
)
1371 err
= do_ip_getsockopt(sk
, level
, optname
, optval
, optlen
, 0);
1372 #ifdef CONFIG_NETFILTER
1373 /* we need to exclude all possible ENOPROTOOPTs except default case */
1374 if (err
== -ENOPROTOOPT
&& optname
!= IP_PKTOPTIONS
&&
1375 !ip_mroute_opt(optname
)) {
1378 if (get_user(len
, optlen
))
1382 err
= nf_getsockopt(sk
, PF_INET
, optname
, optval
,
1386 err
= put_user(len
, optlen
);
1392 EXPORT_SYMBOL(ip_getsockopt
);
1394 #ifdef CONFIG_COMPAT
1395 int compat_ip_getsockopt(struct sock
*sk
, int level
, int optname
,
1396 char __user
*optval
, int __user
*optlen
)
1400 if (optname
== MCAST_MSFILTER
)
1401 return compat_mc_getsockopt(sk
, level
, optname
, optval
, optlen
,
1404 err
= do_ip_getsockopt(sk
, level
, optname
, optval
, optlen
,
1407 #ifdef CONFIG_NETFILTER
1408 /* we need to exclude all possible ENOPROTOOPTs except default case */
1409 if (err
== -ENOPROTOOPT
&& optname
!= IP_PKTOPTIONS
&&
1410 !ip_mroute_opt(optname
)) {
1413 if (get_user(len
, optlen
))
1417 err
= compat_nf_getsockopt(sk
, PF_INET
, optname
, optval
, &len
);
1420 err
= put_user(len
, optlen
);
1426 EXPORT_SYMBOL(compat_ip_getsockopt
);