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 void __secpath_destroy(struct sec_path *sp)
22 for (i = 0; i < sp->len; i++)
23 xfrm_state_put(sp->xvec[i]);
24 kmem_cache_free(secpath_cachep, sp);
26 EXPORT_SYMBOL(__secpath_destroy);
28 struct sec_path *secpath_dup(struct sec_path *src)
32 sp = kmem_cache_alloc(secpath_cachep, GFP_ATOMIC);
40 memcpy(sp, src, sizeof(*sp));
41 for (i = 0; i < sp->len; i++)
42 xfrm_state_hold(sp->xvec[i]);
44 atomic_set(&sp->refcnt, 1);
47 EXPORT_SYMBOL(secpath_dup);
49 /* Fetch spi and seq from ipsec header */
51 int xfrm_parse_spi(struct sk_buff *skb, u8 nexthdr, __be32 *spi, __be32 *seq)
53 int offset, offset_seq;
58 hlen = sizeof(struct ip_auth_hdr);
59 offset = offsetof(struct ip_auth_hdr, spi);
60 offset_seq = offsetof(struct ip_auth_hdr, seq_no);
63 hlen = sizeof(struct ip_esp_hdr);
64 offset = offsetof(struct ip_esp_hdr, spi);
65 offset_seq = offsetof(struct ip_esp_hdr, seq_no);
68 if (!pskb_may_pull(skb, sizeof(struct ip_comp_hdr)))
70 *spi = htonl(ntohs(*(__be16*)(skb_transport_header(skb) + 2)));
77 if (!pskb_may_pull(skb, hlen))
80 *spi = *(__be32*)(skb_transport_header(skb) + offset);
81 *seq = *(__be32*)(skb_transport_header(skb) + offset_seq);
85 int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
87 struct xfrm_mode *inner_mode = x->inner_mode;
90 err = x->outer_mode->afinfo->extract_input(x, skb);
94 if (x->sel.family == AF_UNSPEC) {
95 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
96 if (inner_mode == NULL)
100 skb->protocol = inner_mode->afinfo->eth_proto;
101 return inner_mode->input2(x, skb);
103 EXPORT_SYMBOL(xfrm_prepare_input);
105 int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
107 struct net *net = dev_net(skb->dev);
110 struct xfrm_state *x;
111 xfrm_address_t *daddr;
112 struct xfrm_mode *inner_mode;
117 /* A negative encap_type indicates async resumption. */
118 if (encap_type < 0) {
120 x = xfrm_input_state(skb);
121 seq = XFRM_SKB_CB(skb)->seq.input;
125 /* Allocate new secpath or COW existing one. */
126 if (!skb->sp || atomic_read(&skb->sp->refcnt) != 1) {
129 sp = secpath_dup(skb->sp);
131 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
135 secpath_put(skb->sp);
139 daddr = (xfrm_address_t *)(skb_network_header(skb) +
140 XFRM_SPI_SKB_CB(skb)->daddroff);
141 family = XFRM_SPI_SKB_CB(skb)->family;
144 if (!spi && (err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
145 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
150 if (skb->sp->len == XFRM_MAX_DEPTH) {
151 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
155 x = xfrm_state_lookup(net, daddr, spi, nexthdr, family);
157 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
158 xfrm_audit_state_notfound(skb, family, spi, seq);
162 skb->sp->xvec[skb->sp->len++] = x;
165 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
166 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEINVALID);
170 if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
171 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
175 if (x->props.replay_window && xfrm_replay_check(x, skb, seq)) {
176 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
180 if (xfrm_state_check_expire(x)) {
181 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEEXPIRED);
185 spin_unlock(&x->lock);
187 XFRM_SKB_CB(skb)->seq.input = seq;
189 nexthdr = x->type->input(x, skb);
191 if (nexthdr == -EINPROGRESS)
197 if (nexthdr == -EBADMSG) {
198 xfrm_audit_state_icvfail(x, skb,
200 x->stats.integrity_failed++;
202 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
206 /* only the first xfrm gets the encap type */
209 if (x->props.replay_window)
210 xfrm_replay_advance(x, seq);
212 x->curlft.bytes += skb->len;
215 spin_unlock(&x->lock);
217 XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
219 inner_mode = x->inner_mode;
221 if (x->sel.family == AF_UNSPEC) {
222 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
223 if (inner_mode == NULL)
227 if (inner_mode->input(x, skb)) {
228 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
232 if (x->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) {
238 * We need the inner address. However, we only get here for
239 * transport mode so the outer address is identical.
241 daddr = &x->id.daddr;
242 family = x->outer_mode->afinfo->family;
244 err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
246 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
258 return x->inner_mode->afinfo->transport_finish(skb, async);
262 spin_unlock(&x->lock);
267 EXPORT_SYMBOL(xfrm_input);
269 int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
271 return xfrm_input(skb, nexthdr, 0, -1);
273 EXPORT_SYMBOL(xfrm_input_resume);
275 void __init xfrm_input_init(void)
277 secpath_cachep = kmem_cache_create("secpath_cache",
278 sizeof(struct sec_path),
279 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,