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>
16 #include <net/ip_tunnels.h>
17 #include <net/ip6_tunnel.h>
19 static struct kmem_cache
*secpath_cachep __read_mostly
;
21 static DEFINE_SPINLOCK(xfrm_input_afinfo_lock
);
22 static struct xfrm_input_afinfo
const __rcu
*xfrm_input_afinfo
[AF_INET6
+ 1];
24 static struct gro_cells gro_cells
;
25 static struct net_device xfrm_napi_dev
;
27 int xfrm_input_register_afinfo(const struct xfrm_input_afinfo
*afinfo
)
31 if (WARN_ON(afinfo
->family
>= ARRAY_SIZE(xfrm_input_afinfo
)))
34 spin_lock_bh(&xfrm_input_afinfo_lock
);
35 if (unlikely(xfrm_input_afinfo
[afinfo
->family
] != NULL
))
38 rcu_assign_pointer(xfrm_input_afinfo
[afinfo
->family
], afinfo
);
39 spin_unlock_bh(&xfrm_input_afinfo_lock
);
42 EXPORT_SYMBOL(xfrm_input_register_afinfo
);
44 int xfrm_input_unregister_afinfo(const struct xfrm_input_afinfo
*afinfo
)
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 const struct xfrm_input_afinfo
*xfrm_input_get_afinfo(unsigned int family
)
63 const struct xfrm_input_afinfo
*afinfo
;
65 if (WARN_ON_ONCE(family
>= ARRAY_SIZE(xfrm_input_afinfo
)))
69 afinfo
= rcu_dereference(xfrm_input_afinfo
[family
]);
70 if (unlikely(!afinfo
))
75 static int xfrm_rcv_cb(struct sk_buff
*skb
, unsigned int family
, u8 protocol
,
79 const struct xfrm_input_afinfo
*afinfo
= xfrm_input_get_afinfo(family
);
84 ret
= afinfo
->callback(skb
, protocol
, err
);
90 void __secpath_destroy(struct sec_path
*sp
)
93 for (i
= 0; i
< sp
->len
; i
++)
94 xfrm_state_put(sp
->xvec
[i
]);
95 kmem_cache_free(secpath_cachep
, sp
);
97 EXPORT_SYMBOL(__secpath_destroy
);
99 struct sec_path
*secpath_dup(struct sec_path
*src
)
103 sp
= kmem_cache_alloc(secpath_cachep
, GFP_ATOMIC
);
110 memset(sp
->ovec
, 0, sizeof(sp
->ovec
[XFRM_MAX_OFFLOAD_DEPTH
]));
115 memcpy(sp
, src
, sizeof(*sp
));
116 for (i
= 0; i
< sp
->len
; i
++)
117 xfrm_state_hold(sp
->xvec
[i
]);
119 refcount_set(&sp
->refcnt
, 1);
122 EXPORT_SYMBOL(secpath_dup
);
124 int secpath_set(struct sk_buff
*skb
)
128 /* Allocate new secpath or COW existing one. */
129 if (!skb
->sp
|| refcount_read(&skb
->sp
->refcnt
) != 1) {
130 sp
= secpath_dup(skb
->sp
);
135 secpath_put(skb
->sp
);
140 EXPORT_SYMBOL(secpath_set
);
142 /* Fetch spi and seq from ipsec header */
144 int xfrm_parse_spi(struct sk_buff
*skb
, u8 nexthdr
, __be32
*spi
, __be32
*seq
)
146 int offset
, offset_seq
;
151 hlen
= sizeof(struct ip_auth_hdr
);
152 offset
= offsetof(struct ip_auth_hdr
, spi
);
153 offset_seq
= offsetof(struct ip_auth_hdr
, seq_no
);
156 hlen
= sizeof(struct ip_esp_hdr
);
157 offset
= offsetof(struct ip_esp_hdr
, spi
);
158 offset_seq
= offsetof(struct ip_esp_hdr
, seq_no
);
161 if (!pskb_may_pull(skb
, sizeof(struct ip_comp_hdr
)))
163 *spi
= htonl(ntohs(*(__be16
*)(skb_transport_header(skb
) + 2)));
170 if (!pskb_may_pull(skb
, hlen
))
173 *spi
= *(__be32
*)(skb_transport_header(skb
) + offset
);
174 *seq
= *(__be32
*)(skb_transport_header(skb
) + offset_seq
);
177 EXPORT_SYMBOL(xfrm_parse_spi
);
179 int xfrm_prepare_input(struct xfrm_state
*x
, struct sk_buff
*skb
)
181 struct xfrm_mode
*inner_mode
= x
->inner_mode
;
184 err
= x
->outer_mode
->afinfo
->extract_input(x
, skb
);
188 if (x
->sel
.family
== AF_UNSPEC
) {
189 inner_mode
= xfrm_ip2inner_mode(x
, XFRM_MODE_SKB_CB(skb
)->protocol
);
190 if (inner_mode
== NULL
)
191 return -EAFNOSUPPORT
;
194 skb
->protocol
= inner_mode
->afinfo
->eth_proto
;
195 return inner_mode
->input2(x
, skb
);
197 EXPORT_SYMBOL(xfrm_prepare_input
);
199 int xfrm_input(struct sk_buff
*skb
, int nexthdr
, __be32 spi
, int encap_type
)
201 struct net
*net
= dev_net(skb
->dev
);
205 struct xfrm_state
*x
= NULL
;
206 xfrm_address_t
*daddr
;
207 struct xfrm_mode
*inner_mode
;
208 u32 mark
= skb
->mark
;
212 bool xfrm_gro
= false;
213 bool crypto_done
= false;
214 struct xfrm_offload
*xo
= xfrm_offload(skb
);
216 if (encap_type
< 0) {
217 x
= xfrm_input_state(skb
);
218 family
= x
->outer_mode
->afinfo
->family
;
220 /* An encap_type of -1 indicates async resumption. */
221 if (encap_type
== -1) {
223 seq
= XFRM_SKB_CB(skb
)->seq
.input
.low
;
227 /* encap_type < -1 indicates a GRO call. */
229 seq
= XFRM_SPI_SKB_CB(skb
)->seq
;
231 if (xo
&& (xo
->flags
& CRYPTO_DONE
)) {
233 x
= xfrm_input_state(skb
);
234 family
= XFRM_SPI_SKB_CB(skb
)->family
;
236 if (!(xo
->status
& CRYPTO_SUCCESS
)) {
238 (CRYPTO_TRANSPORT_AH_AUTH_FAILED
|
239 CRYPTO_TRANSPORT_ESP_AUTH_FAILED
|
240 CRYPTO_TUNNEL_AH_AUTH_FAILED
|
241 CRYPTO_TUNNEL_ESP_AUTH_FAILED
)) {
243 xfrm_audit_state_icvfail(x
, skb
,
245 x
->stats
.integrity_failed
++;
246 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEPROTOERROR
);
250 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINBUFFERERROR
);
254 if ((err
= xfrm_parse_spi(skb
, nexthdr
, &spi
, &seq
)) != 0) {
255 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINHDRERROR
);
263 daddr
= (xfrm_address_t
*)(skb_network_header(skb
) +
264 XFRM_SPI_SKB_CB(skb
)->daddroff
);
265 family
= XFRM_SPI_SKB_CB(skb
)->family
;
267 /* if tunnel is present override skb->mark value with tunnel i_key */
270 if (XFRM_TUNNEL_SKB_CB(skb
)->tunnel
.ip4
)
271 mark
= be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb
)->tunnel
.ip4
->parms
.i_key
);
274 if (XFRM_TUNNEL_SKB_CB(skb
)->tunnel
.ip6
)
275 mark
= be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb
)->tunnel
.ip6
->parms
.i_key
);
279 err
= secpath_set(skb
);
281 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINERROR
);
286 if (!spi
&& (err
= xfrm_parse_spi(skb
, nexthdr
, &spi
, &seq
)) != 0) {
287 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINHDRERROR
);
292 if (skb
->sp
->len
== XFRM_MAX_DEPTH
) {
293 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINBUFFERERROR
);
297 x
= xfrm_state_lookup(net
, mark
, daddr
, spi
, nexthdr
, family
);
299 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINNOSTATES
);
300 xfrm_audit_state_notfound(skb
, family
, spi
, seq
);
304 skb
->sp
->xvec
[skb
->sp
->len
++] = x
;
309 if (unlikely(x
->km
.state
!= XFRM_STATE_VALID
)) {
310 if (x
->km
.state
== XFRM_STATE_ACQ
)
311 XFRM_INC_STATS(net
, LINUX_MIB_XFRMACQUIREERROR
);
314 LINUX_MIB_XFRMINSTATEINVALID
);
318 if ((x
->encap
? x
->encap
->encap_type
: 0) != encap_type
) {
319 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEMISMATCH
);
323 if (x
->repl
->check(x
, skb
, seq
)) {
324 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATESEQERROR
);
328 if (xfrm_state_check_expire(x
)) {
329 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEEXPIRED
);
333 spin_unlock(&x
->lock
);
335 if (xfrm_tunnel_check(skb
, x
, family
)) {
336 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEMODEERROR
);
340 seq_hi
= htonl(xfrm_replay_seqhi(x
, seq
));
342 XFRM_SKB_CB(skb
)->seq
.input
.low
= seq
;
343 XFRM_SKB_CB(skb
)->seq
.input
.hi
= seq_hi
;
349 nexthdr
= x
->type_offload
->input_tail(x
, skb
);
351 nexthdr
= x
->type
->input(x
, skb
);
353 if (nexthdr
== -EINPROGRESS
)
360 if (nexthdr
== -EBADMSG
) {
361 xfrm_audit_state_icvfail(x
, skb
,
363 x
->stats
.integrity_failed
++;
365 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEPROTOERROR
);
369 /* only the first xfrm gets the encap type */
372 if (async
&& x
->repl
->recheck(x
, skb
, seq
)) {
373 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATESEQERROR
);
377 x
->repl
->advance(x
, seq
);
379 x
->curlft
.bytes
+= skb
->len
;
382 spin_unlock(&x
->lock
);
384 XFRM_MODE_SKB_CB(skb
)->protocol
= nexthdr
;
386 inner_mode
= x
->inner_mode
;
388 if (x
->sel
.family
== AF_UNSPEC
) {
389 inner_mode
= xfrm_ip2inner_mode(x
, XFRM_MODE_SKB_CB(skb
)->protocol
);
390 if (inner_mode
== NULL
) {
391 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEMODEERROR
);
396 if (inner_mode
->input(x
, skb
)) {
397 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINSTATEMODEERROR
);
401 if (x
->outer_mode
->flags
& XFRM_MODE_FLAG_TUNNEL
) {
407 * We need the inner address. However, we only get here for
408 * transport mode so the outer address is identical.
410 daddr
= &x
->id
.daddr
;
411 family
= x
->outer_mode
->afinfo
->family
;
413 err
= xfrm_parse_spi(skb
, nexthdr
, &spi
, &seq
);
415 XFRM_INC_STATS(net
, LINUX_MIB_XFRMINHDRERROR
);
420 err
= xfrm_rcv_cb(skb
, family
, x
->type
->proto
, 0);
428 gro_cells_receive(&gro_cells
, skb
);
431 xo
= xfrm_offload(skb
);
433 xfrm_gro
= xo
->flags
& XFRM_GRO
;
435 err
= x
->inner_mode
->afinfo
->transport_finish(skb
, xfrm_gro
|| async
);
438 gro_cells_receive(&gro_cells
, skb
);
446 spin_unlock(&x
->lock
);
448 xfrm_rcv_cb(skb
, family
, x
&& x
->type
? x
->type
->proto
: nexthdr
, -1);
452 EXPORT_SYMBOL(xfrm_input
);
454 int xfrm_input_resume(struct sk_buff
*skb
, int nexthdr
)
456 return xfrm_input(skb
, nexthdr
, 0, -1);
458 EXPORT_SYMBOL(xfrm_input_resume
);
460 void __init
xfrm_input_init(void)
464 init_dummy_netdev(&xfrm_napi_dev
);
465 err
= gro_cells_init(&gro_cells
, &xfrm_napi_dev
);
467 gro_cells
.cells
= NULL
;
469 secpath_cachep
= kmem_cache_create("secpath_cache",
470 sizeof(struct sec_path
),
471 0, SLAB_HWCACHE_ALIGN
|SLAB_PANIC
,