5 * YOSHIFUJI Hideaki @USAGI
6 * Split up af-specific portion
10 #include <linux/slab.h>
11 #include <linux/module.h>
12 #include <linux/netdevice.h>
17 static struct kmem_cache
*secpath_cachep __read_mostly
;
19 static DEFINE_SPINLOCK(xfrm_input_afinfo_lock
);
20 static struct xfrm_input_afinfo __rcu
*xfrm_input_afinfo
[NPROTO
];
22 int xfrm_input_register_afinfo(struct xfrm_input_afinfo
*afinfo
)
26 if (unlikely(afinfo
== NULL
))
28 if (unlikely(afinfo
->family
>= NPROTO
))
30 spin_lock_bh(&xfrm_input_afinfo_lock
);
31 if (unlikely(xfrm_input_afinfo
[afinfo
->family
] != NULL
))
34 rcu_assign_pointer(xfrm_input_afinfo
[afinfo
->family
], afinfo
);
35 spin_unlock_bh(&xfrm_input_afinfo_lock
);
38 EXPORT_SYMBOL(xfrm_input_register_afinfo
);
40 int xfrm_input_unregister_afinfo(struct xfrm_input_afinfo
*afinfo
)
44 if (unlikely(afinfo
== NULL
))
46 if (unlikely(afinfo
->family
>= NPROTO
))
48 spin_lock_bh(&xfrm_input_afinfo_lock
);
49 if (likely(xfrm_input_afinfo
[afinfo
->family
] != NULL
)) {
50 if (unlikely(xfrm_input_afinfo
[afinfo
->family
] != afinfo
))
53 RCU_INIT_POINTER(xfrm_input_afinfo
[afinfo
->family
], NULL
);
55 spin_unlock_bh(&xfrm_input_afinfo_lock
);
59 EXPORT_SYMBOL(xfrm_input_unregister_afinfo
);
61 static struct xfrm_input_afinfo
*xfrm_input_get_afinfo(unsigned int family
)
63 struct xfrm_input_afinfo
*afinfo
;
65 if (unlikely(family
>= NPROTO
))
68 afinfo
= rcu_dereference(xfrm_input_afinfo
[family
]);
69 if (unlikely(!afinfo
))
74 static void xfrm_input_put_afinfo(struct xfrm_input_afinfo
*afinfo
)
79 static int xfrm_rcv_cb(struct sk_buff
*skb
, unsigned int family
, u8 protocol
,
83 struct xfrm_input_afinfo
*afinfo
= xfrm_input_get_afinfo(family
);
88 ret
= afinfo
->callback(skb
, protocol
, err
);
89 xfrm_input_put_afinfo(afinfo
);
94 void __secpath_destroy(struct sec_path
*sp
)
97 for (i
= 0; i
< sp
->len
; i
++)
98 xfrm_state_put(sp
->xvec
[i
]);
99 kmem_cache_free(secpath_cachep
, sp
);
101 EXPORT_SYMBOL(__secpath_destroy
);
103 struct sec_path
*secpath_dup(struct sec_path
*src
)
107 sp
= kmem_cache_alloc(secpath_cachep
, GFP_ATOMIC
);
115 memcpy(sp
, src
, sizeof(*sp
));
116 for (i
= 0; i
< sp
->len
; i
++)
117 xfrm_state_hold(sp
->xvec
[i
]);
119 atomic_set(&sp
->refcnt
, 1);
122 EXPORT_SYMBOL(secpath_dup
);
124 /* Fetch spi and seq from ipsec header */
126 int xfrm_parse_spi(struct sk_buff
*skb
, u8 nexthdr
, __be32
*spi
, __be32
*seq
)
128 int offset
, offset_seq
;
133 hlen
= sizeof(struct ip_auth_hdr
);
134 offset
= offsetof(struct ip_auth_hdr
, spi
);
135 offset_seq
= offsetof(struct ip_auth_hdr
, seq_no
);
138 hlen
= sizeof(struct ip_esp_hdr
);
139 offset
= offsetof(struct ip_esp_hdr
, spi
);
140 offset_seq
= offsetof(struct ip_esp_hdr
, seq_no
);
143 if (!pskb_may_pull(skb
, sizeof(struct ip_comp_hdr
)))
145 *spi
= htonl(ntohs(*(__be16
*)(skb_transport_header(skb
) + 2)));
152 if (!pskb_may_pull(skb
, hlen
))
155 *spi
= *(__be32
*)(skb_transport_header(skb
) + offset
);
156 *seq
= *(__be32
*)(skb_transport_header(skb
) + offset_seq
);
160 int xfrm_prepare_input(struct xfrm_state
*x
, struct sk_buff
*skb
)
162 struct xfrm_mode
*inner_mode
= x
->inner_mode
;
165 err
= x
->outer_mode
->afinfo
->extract_input(x
, skb
);
169 if (x
->sel
.family
== AF_UNSPEC
) {
170 inner_mode
= xfrm_ip2inner_mode(x
, XFRM_MODE_SKB_CB(skb
)->protocol
);
171 if (inner_mode
== NULL
)
172 return -EAFNOSUPPORT
;
175 skb
->protocol
= inner_mode
->afinfo
->eth_proto
;
176 return inner_mode
->input2(x
, skb
);
178 EXPORT_SYMBOL(xfrm_prepare_input
);
180 int xfrm_input(struct sk_buff
*skb
, int nexthdr
, __be32 spi
, int encap_type
)
182 struct net
*net
= dev_net(skb
->dev
);
186 struct xfrm_state
*x
= NULL
;
187 xfrm_address_t
*daddr
;
188 struct xfrm_mode
*inner_mode
;
193 /* A negative encap_type indicates async resumption. */
194 if (encap_type
< 0) {
196 x
= xfrm_input_state(skb
);
197 seq
= XFRM_SKB_CB(skb
)->seq
.input
.low
;
198 family
= x
->outer_mode
->afinfo
->family
;
202 daddr
= (xfrm_address_t
*)(skb_network_header(skb
) +
203 XFRM_SPI_SKB_CB(skb
)->daddroff
);
204 family
= XFRM_SPI_SKB_CB(skb
)->family
;
206 /* Allocate new secpath or COW existing one. */
207 if (!skb
->sp
|| atomic_read(&skb
->sp
->refcnt
) != 1) {
210 sp
= secpath_dup(skb
->sp
);
212 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINERROR
);
216 secpath_put(skb
->sp
);
221 if (!spi
&& (err
= xfrm_parse_spi(skb
, nexthdr
, &spi
, &seq
)) != 0) {
222 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINHDRERROR
);
227 if (skb
->sp
->len
== XFRM_MAX_DEPTH
) {
228 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINBUFFERERROR
);
232 x
= xfrm_state_lookup(net
, skb
->mark
, daddr
, spi
, nexthdr
, family
);
234 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINNOSTATES
);
235 xfrm_audit_state_notfound(skb
, family
, spi
, seq
);
239 skb
->sp
->xvec
[skb
->sp
->len
++] = x
;
241 if (xfrm_tunnel_check(skb
, x
, family
)) {
242 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEMODEERROR
);
247 if (unlikely(x
->km
.state
== XFRM_STATE_ACQ
)) {
248 XFRM_INC_STATS(net
, LINUX_MIB_XFRMACQUIREERROR
);
252 if (unlikely(x
->km
.state
!= XFRM_STATE_VALID
)) {
253 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEINVALID
);
257 if ((x
->encap
? x
->encap
->encap_type
: 0) != encap_type
) {
258 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEMISMATCH
);
262 if (x
->repl
->check(x
, skb
, seq
)) {
263 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATESEQERROR
);
267 if (xfrm_state_check_expire(x
)) {
268 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEEXPIRED
);
272 spin_unlock(&x
->lock
);
274 seq_hi
= htonl(xfrm_replay_seqhi(x
, seq
));
276 XFRM_SKB_CB(skb
)->seq
.input
.low
= seq
;
277 XFRM_SKB_CB(skb
)->seq
.input
.hi
= seq_hi
;
281 nexthdr
= x
->type
->input(x
, skb
);
283 if (nexthdr
== -EINPROGRESS
)
288 if (nexthdr
== -EBADMSG
) {
289 xfrm_audit_state_icvfail(x
, skb
,
291 x
->stats
.integrity_failed
++;
293 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEPROTOERROR
);
297 /* only the first xfrm gets the encap type */
300 if (async
&& x
->repl
->recheck(x
, skb
, seq
)) {
301 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATESEQERROR
);
305 x
->repl
->advance(x
, seq
);
307 x
->curlft
.bytes
+= skb
->len
;
310 spin_unlock(&x
->lock
);
312 XFRM_MODE_SKB_CB(skb
)->protocol
= nexthdr
;
314 inner_mode
= x
->inner_mode
;
316 if (x
->sel
.family
== AF_UNSPEC
) {
317 inner_mode
= xfrm_ip2inner_mode(x
, XFRM_MODE_SKB_CB(skb
)->protocol
);
318 if (inner_mode
== NULL
)
322 if (inner_mode
->input(x
, skb
)) {
323 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEMODEERROR
);
327 if (x
->outer_mode
->flags
& XFRM_MODE_FLAG_TUNNEL
) {
333 * We need the inner address. However, we only get here for
334 * transport mode so the outer address is identical.
336 daddr
= &x
->id
.daddr
;
337 family
= x
->outer_mode
->afinfo
->family
;
339 err
= xfrm_parse_spi(skb
, nexthdr
, &spi
, &seq
);
341 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINHDRERROR
);
346 err
= xfrm_rcv_cb(skb
, family
, x
->type
->proto
, 0);
357 return x
->inner_mode
->afinfo
->transport_finish(skb
, async
);
361 spin_unlock(&x
->lock
);
363 xfrm_rcv_cb(skb
, family
, x
&& x
->type
? x
->type
->proto
: nexthdr
, -1);
367 EXPORT_SYMBOL(xfrm_input
);
369 int xfrm_input_resume(struct sk_buff
*skb
, int nexthdr
)
371 return xfrm_input(skb
, nexthdr
, 0, -1);
373 EXPORT_SYMBOL(xfrm_input_resume
);
375 void __init
xfrm_input_init(void)
377 secpath_cachep
= kmem_cache_create("secpath_cache",
378 sizeof(struct sec_path
),
379 0, SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,