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 * The options processing module for ip.c
8 * Authors: A.N.Kuznetsov
12 #define pr_fmt(fmt) "IPv4: " fmt
14 #include <linux/capability.h>
15 #include <linux/module.h>
16 #include <linux/slab.h>
17 #include <linux/types.h>
18 #include <asm/uaccess.h>
19 #include <asm/unaligned.h>
20 #include <linux/skbuff.h>
22 #include <linux/icmp.h>
23 #include <linux/netdevice.h>
24 #include <linux/rtnetlink.h>
28 #include <net/route.h>
29 #include <net/cipso_ipv4.h>
30 #include <net/ip_fib.h>
33 * Write options to IP header, record destination address to
34 * source route option, address of outgoing interface
35 * (we should already know it, so that this function is allowed be
36 * called only after routing decision) and timestamp,
37 * if we originate this datagram.
39 * daddr is real destination address, next hop is recorded in IP header.
40 * saddr is address of outgoing interface.
43 void ip_options_build(struct sk_buff
*skb
, struct ip_options
*opt
,
44 __be32 daddr
, struct rtable
*rt
, int is_frag
)
46 unsigned char *iph
= skb_network_header(skb
);
48 memcpy(&(IPCB(skb
)->opt
), opt
, sizeof(struct ip_options
));
49 memcpy(iph
+sizeof(struct iphdr
), opt
->__data
, opt
->optlen
);
50 opt
= &(IPCB(skb
)->opt
);
53 memcpy(iph
+opt
->srr
+iph
[opt
->srr
+1]-4, &daddr
, 4);
57 ip_rt_get_source(iph
+opt
->rr
+iph
[opt
->rr
+2]-5, skb
, rt
);
59 ip_rt_get_source(iph
+opt
->ts
+iph
[opt
->ts
+2]-9, skb
, rt
);
60 if (opt
->ts_needtime
) {
63 midtime
= inet_current_timestamp();
64 memcpy(iph
+opt
->ts
+iph
[opt
->ts
+2]-5, &midtime
, 4);
69 memset(iph
+opt
->rr
, IPOPT_NOP
, iph
[opt
->rr
+1]);
74 memset(iph
+opt
->ts
, IPOPT_NOP
, iph
[opt
->ts
+1]);
76 opt
->ts_needaddr
= opt
->ts_needtime
= 0;
81 * Provided (sopt, skb) points to received options,
82 * build in dopt compiled option set appropriate for answering.
83 * i.e. invert SRR option, copy anothers,
84 * and grab room in RR/TS options.
86 * NOTE: dopt cannot point to skb.
89 int __ip_options_echo(struct ip_options
*dopt
, struct sk_buff
*skb
,
90 const struct ip_options
*sopt
)
92 unsigned char *sptr
, *dptr
;
96 memset(dopt
, 0, sizeof(struct ip_options
));
98 if (sopt
->optlen
== 0)
101 sptr
= skb_network_header(skb
);
105 optlen
= sptr
[sopt
->rr
+1];
106 soffset
= sptr
[sopt
->rr
+2];
107 dopt
->rr
= dopt
->optlen
+ sizeof(struct iphdr
);
108 memcpy(dptr
, sptr
+sopt
->rr
, optlen
);
109 if (sopt
->rr_needaddr
&& soffset
<= optlen
) {
110 if (soffset
+ 3 > optlen
)
112 dptr
[2] = soffset
+ 4;
113 dopt
->rr_needaddr
= 1;
116 dopt
->optlen
+= optlen
;
119 optlen
= sptr
[sopt
->ts
+1];
120 soffset
= sptr
[sopt
->ts
+2];
121 dopt
->ts
= dopt
->optlen
+ sizeof(struct iphdr
);
122 memcpy(dptr
, sptr
+sopt
->ts
, optlen
);
123 if (soffset
<= optlen
) {
124 if (sopt
->ts_needaddr
) {
125 if (soffset
+ 3 > optlen
)
127 dopt
->ts_needaddr
= 1;
130 if (sopt
->ts_needtime
) {
131 if (soffset
+ 3 > optlen
)
133 if ((dptr
[3]&0xF) != IPOPT_TS_PRESPEC
) {
134 dopt
->ts_needtime
= 1;
137 dopt
->ts_needtime
= 0;
139 if (soffset
+ 7 <= optlen
) {
142 memcpy(&addr
, dptr
+soffset
-1, 4);
143 if (inet_addr_type(dev_net(skb_dst(skb
)->dev
), addr
) != RTN_UNICAST
) {
144 dopt
->ts_needtime
= 1;
153 dopt
->optlen
+= optlen
;
156 unsigned char *start
= sptr
+sopt
->srr
;
162 if (soffset
> optlen
)
163 soffset
= optlen
+ 1;
166 memcpy(&faddr
, &start
[soffset
-1], 4);
167 for (soffset
-= 4, doffset
= 4; soffset
> 3; soffset
-= 4, doffset
+= 4)
168 memcpy(&dptr
[doffset
-1], &start
[soffset
-1], 4);
170 * RFC1812 requires to fix illegal source routes.
172 if (memcmp(&ip_hdr(skb
)->saddr
,
173 &start
[soffset
+ 3], 4) == 0)
177 __be32 daddr
= fib_compute_spec_dst(skb
);
179 memcpy(&start
[doffset
-1], &daddr
, 4);
185 dopt
->srr
= dopt
->optlen
+ sizeof(struct iphdr
);
186 dopt
->optlen
+= doffset
+3;
187 dopt
->is_strictroute
= sopt
->is_strictroute
;
191 optlen
= sptr
[sopt
->cipso
+1];
192 dopt
->cipso
= dopt
->optlen
+sizeof(struct iphdr
);
193 memcpy(dptr
, sptr
+sopt
->cipso
, optlen
);
195 dopt
->optlen
+= optlen
;
197 while (dopt
->optlen
& 3) {
205 * Options "fragmenting", just fill options not
206 * allowed in fragments with NOOPs.
207 * Simple and stupid 8), but the most efficient way.
210 void ip_options_fragment(struct sk_buff
*skb
)
212 unsigned char *optptr
= skb_network_header(skb
) + sizeof(struct iphdr
);
213 struct ip_options
*opt
= &(IPCB(skb
)->opt
);
227 if (optlen
< 2 || optlen
> l
)
229 if (!IPOPT_COPIED(*optptr
))
230 memset(optptr
, IPOPT_NOOP
, optlen
);
236 opt
->rr_needaddr
= 0;
237 opt
->ts_needaddr
= 0;
238 opt
->ts_needtime
= 0;
241 /* helper used by ip_options_compile() to call fib_compute_spec_dst()
244 static void spec_dst_fill(__be32
*spec_dst
, struct sk_buff
*skb
)
246 if (*spec_dst
== htonl(INADDR_ANY
))
247 *spec_dst
= fib_compute_spec_dst(skb
);
251 * Verify options and fill pointers in struct options.
252 * Caller should clear *opt, and set opt->data.
253 * If opt == NULL, then skb->data should point to IP header.
256 int ip_options_compile(struct net
*net
,
257 struct ip_options
*opt
, struct sk_buff
*skb
)
259 __be32 spec_dst
= htonl(INADDR_ANY
);
260 unsigned char *pp_ptr
= NULL
;
261 struct rtable
*rt
= NULL
;
262 unsigned char *optptr
;
267 rt
= skb_rtable(skb
);
268 optptr
= (unsigned char *)&(ip_hdr(skb
)[1]);
270 optptr
= opt
->__data
;
271 iph
= optptr
- sizeof(struct iphdr
);
273 for (l
= opt
->optlen
; l
> 0; ) {
276 for (optptr
++, l
--; l
> 0; optptr
++, l
--) {
277 if (*optptr
!= IPOPT_END
) {
288 if (unlikely(l
< 2)) {
293 if (optlen
< 2 || optlen
> l
) {
308 /* NB: cf RFC-1812 5.2.4.1 */
314 if (optptr
[2] != 4 || optlen
< 7 || ((optlen
-3) & 3)) {
318 memcpy(&opt
->faddr
, &optptr
[3], 4);
320 memmove(&optptr
[3], &optptr
[7], optlen
-7);
322 opt
->is_strictroute
= (optptr
[0] == IPOPT_SSRR
);
323 opt
->srr
= optptr
- iph
;
338 if (optptr
[2] <= optlen
) {
339 if (optptr
[2]+3 > optlen
) {
344 spec_dst_fill(&spec_dst
, skb
);
345 memcpy(&optptr
[optptr
[2]-1], &spec_dst
, 4);
349 opt
->rr_needaddr
= 1;
351 opt
->rr
= optptr
- iph
;
353 case IPOPT_TIMESTAMP
:
366 if (optptr
[2] <= optlen
) {
367 unsigned char *timeptr
= NULL
;
368 if (optptr
[2]+3 > optlen
) {
372 switch (optptr
[3]&0xF) {
373 case IPOPT_TS_TSONLY
:
375 timeptr
= &optptr
[optptr
[2]-1];
376 opt
->ts_needtime
= 1;
379 case IPOPT_TS_TSANDADDR
:
380 if (optptr
[2]+7 > optlen
) {
385 spec_dst_fill(&spec_dst
, skb
);
386 memcpy(&optptr
[optptr
[2]-1], &spec_dst
, 4);
387 timeptr
= &optptr
[optptr
[2]+3];
389 opt
->ts_needaddr
= 1;
390 opt
->ts_needtime
= 1;
393 case IPOPT_TS_PRESPEC
:
394 if (optptr
[2]+7 > optlen
) {
400 memcpy(&addr
, &optptr
[optptr
[2]-1], 4);
401 if (inet_addr_type(net
, addr
) == RTN_UNICAST
)
404 timeptr
= &optptr
[optptr
[2]+3];
406 opt
->ts_needtime
= 1;
410 if (!skb
&& !ns_capable(net
->user_ns
, CAP_NET_RAW
)) {
419 midtime
= inet_current_timestamp();
420 memcpy(timeptr
, &midtime
, 4);
423 } else if ((optptr
[3]&0xF) != IPOPT_TS_PRESPEC
) {
424 unsigned int overflow
= optptr
[3]>>4;
425 if (overflow
== 15) {
430 optptr
[3] = (optptr
[3]&0xF)|((overflow
+1)<<4);
434 opt
->ts
= optptr
- iph
;
441 if (optptr
[2] == 0 && optptr
[3] == 0)
442 opt
->router_alert
= optptr
- iph
;
445 if ((!skb
&& !ns_capable(net
->user_ns
, CAP_NET_RAW
)) || opt
->cipso
) {
449 opt
->cipso
= optptr
- iph
;
450 if (cipso_v4_validate(skb
, &optptr
)) {
458 if (!skb
&& !ns_capable(net
->user_ns
, CAP_NET_RAW
)) {
474 icmp_send(skb
, ICMP_PARAMETERPROB
, 0, htonl((pp_ptr
-iph
)<<24));
478 EXPORT_SYMBOL(ip_options_compile
);
481 * Undo all the changes done by ip_options_compile().
484 void ip_options_undo(struct ip_options
*opt
)
487 unsigned char *optptr
= opt
->__data
+opt
->srr
-sizeof(struct iphdr
);
488 memmove(optptr
+7, optptr
+3, optptr
[1]-7);
489 memcpy(optptr
+3, &opt
->faddr
, 4);
491 if (opt
->rr_needaddr
) {
492 unsigned char *optptr
= opt
->__data
+opt
->rr
-sizeof(struct iphdr
);
494 memset(&optptr
[optptr
[2]-1], 0, 4);
497 unsigned char *optptr
= opt
->__data
+opt
->ts
-sizeof(struct iphdr
);
498 if (opt
->ts_needtime
) {
500 memset(&optptr
[optptr
[2]-1], 0, 4);
501 if ((optptr
[3]&0xF) == IPOPT_TS_PRESPEC
)
504 if (opt
->ts_needaddr
) {
506 memset(&optptr
[optptr
[2]-1], 0, 4);
511 static struct ip_options_rcu
*ip_options_get_alloc(const int optlen
)
513 return kzalloc(sizeof(struct ip_options_rcu
) + ((optlen
+ 3) & ~3),
517 static int ip_options_get_finish(struct net
*net
, struct ip_options_rcu
**optp
,
518 struct ip_options_rcu
*opt
, int optlen
)
521 opt
->opt
.__data
[optlen
++] = IPOPT_END
;
522 opt
->opt
.optlen
= optlen
;
523 if (optlen
&& ip_options_compile(net
, &opt
->opt
, NULL
)) {
532 int ip_options_get_from_user(struct net
*net
, struct ip_options_rcu
**optp
,
533 unsigned char __user
*data
, int optlen
)
535 struct ip_options_rcu
*opt
= ip_options_get_alloc(optlen
);
539 if (optlen
&& copy_from_user(opt
->opt
.__data
, data
, optlen
)) {
543 return ip_options_get_finish(net
, optp
, opt
, optlen
);
546 int ip_options_get(struct net
*net
, struct ip_options_rcu
**optp
,
547 unsigned char *data
, int optlen
)
549 struct ip_options_rcu
*opt
= ip_options_get_alloc(optlen
);
554 memcpy(opt
->opt
.__data
, data
, optlen
);
555 return ip_options_get_finish(net
, optp
, opt
, optlen
);
558 void ip_forward_options(struct sk_buff
*skb
)
560 struct ip_options
*opt
= &(IPCB(skb
)->opt
);
561 unsigned char *optptr
;
562 struct rtable
*rt
= skb_rtable(skb
);
563 unsigned char *raw
= skb_network_header(skb
);
565 if (opt
->rr_needaddr
) {
566 optptr
= (unsigned char *)raw
+ opt
->rr
;
567 ip_rt_get_source(&optptr
[optptr
[2]-5], skb
, rt
);
570 if (opt
->srr_is_hit
) {
571 int srrptr
, srrspace
;
573 optptr
= raw
+ opt
->srr
;
575 for ( srrptr
= optptr
[2], srrspace
= optptr
[1];
579 if (srrptr
+ 3 > srrspace
)
581 if (memcmp(&opt
->nexthop
, &optptr
[srrptr
-1], 4) == 0)
584 if (srrptr
+ 3 <= srrspace
) {
586 ip_hdr(skb
)->daddr
= opt
->nexthop
;
587 ip_rt_get_source(&optptr
[srrptr
-1], skb
, rt
);
588 optptr
[2] = srrptr
+4;
590 net_crit_ratelimited("%s(): Argh! Destination lost!\n",
593 if (opt
->ts_needaddr
) {
594 optptr
= raw
+ opt
->ts
;
595 ip_rt_get_source(&optptr
[optptr
[2]-9], skb
, rt
);
599 if (opt
->is_changed
) {
601 ip_send_check(ip_hdr(skb
));
605 int ip_options_rcv_srr(struct sk_buff
*skb
)
607 struct ip_options
*opt
= &(IPCB(skb
)->opt
);
608 int srrspace
, srrptr
;
610 struct iphdr
*iph
= ip_hdr(skb
);
611 unsigned char *optptr
= skb_network_header(skb
) + opt
->srr
;
612 struct rtable
*rt
= skb_rtable(skb
);
614 unsigned long orefdst
;
620 if (skb
->pkt_type
!= PACKET_HOST
)
622 if (rt
->rt_type
== RTN_UNICAST
) {
623 if (!opt
->is_strictroute
)
625 icmp_send(skb
, ICMP_PARAMETERPROB
, 0, htonl(16<<24));
628 if (rt
->rt_type
!= RTN_LOCAL
)
631 for (srrptr
= optptr
[2], srrspace
= optptr
[1]; srrptr
<= srrspace
; srrptr
+= 4) {
632 if (srrptr
+ 3 > srrspace
) {
633 icmp_send(skb
, ICMP_PARAMETERPROB
, 0, htonl((opt
->srr
+2)<<24));
636 memcpy(&nexthop
, &optptr
[srrptr
-1], 4);
638 orefdst
= skb
->_skb_refdst
;
639 skb_dst_set(skb
, NULL
);
640 err
= ip_route_input(skb
, nexthop
, iph
->saddr
, iph
->tos
, skb
->dev
);
641 rt2
= skb_rtable(skb
);
642 if (err
|| (rt2
->rt_type
!= RTN_UNICAST
&& rt2
->rt_type
!= RTN_LOCAL
)) {
644 skb
->_skb_refdst
= orefdst
;
647 refdst_drop(orefdst
);
648 if (rt2
->rt_type
!= RTN_LOCAL
)
650 /* Superfast 8) loopback forward */
651 iph
->daddr
= nexthop
;
654 if (srrptr
<= srrspace
) {
656 opt
->nexthop
= nexthop
;
661 EXPORT_SYMBOL(ip_options_rcv_srr
);