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);
111 struct xfrm_state *x;
112 xfrm_address_t *daddr;
113 struct xfrm_mode *inner_mode;
118 /* A negative encap_type indicates async resumption. */
119 if (encap_type < 0) {
121 x = xfrm_input_state(skb);
122 seq = XFRM_SKB_CB(skb)->seq.input.low;
126 /* Allocate new secpath or COW existing one. */
127 if (!skb->sp || atomic_read(&skb->sp->refcnt) != 1) {
130 sp = secpath_dup(skb->sp);
132 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
136 secpath_put(skb->sp);
140 daddr = (xfrm_address_t *)(skb_network_header(skb) +
141 XFRM_SPI_SKB_CB(skb)->daddroff);
142 family = XFRM_SPI_SKB_CB(skb)->family;
145 if (!spi && (err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
146 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
151 if (skb->sp->len == XFRM_MAX_DEPTH) {
152 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
156 x = xfrm_state_lookup(net, skb->mark, daddr, spi, nexthdr, family);
158 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
159 xfrm_audit_state_notfound(skb, family, spi, seq);
163 skb->sp->xvec[skb->sp->len++] = x;
166 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
167 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEINVALID);
171 if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
172 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
176 if (x->repl->check(x, skb, seq)) {
177 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
181 if (xfrm_state_check_expire(x)) {
182 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEEXPIRED);
186 spin_unlock(&x->lock);
188 seq_hi = htonl(xfrm_replay_seqhi(x, seq));
190 XFRM_SKB_CB(skb)->seq.input.low = seq;
191 XFRM_SKB_CB(skb)->seq.input.hi = seq_hi;
195 nexthdr = x->type->input(x, skb);
197 if (nexthdr == -EINPROGRESS)
203 if (nexthdr == -EBADMSG) {
204 xfrm_audit_state_icvfail(x, skb,
206 x->stats.integrity_failed++;
208 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
212 /* only the first xfrm gets the encap type */
215 if (async && x->repl->check(x, skb, seq)) {
216 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
220 x->repl->advance(x, seq);
222 x->curlft.bytes += skb->len;
225 spin_unlock(&x->lock);
227 XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
229 inner_mode = x->inner_mode;
231 if (x->sel.family == AF_UNSPEC) {
232 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
233 if (inner_mode == NULL)
237 if (inner_mode->input(x, skb)) {
238 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
242 if (x->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) {
248 * We need the inner address. However, we only get here for
249 * transport mode so the outer address is identical.
251 daddr = &x->id.daddr;
252 family = x->outer_mode->afinfo->family;
254 err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
256 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
268 return x->inner_mode->afinfo->transport_finish(skb, async);
272 spin_unlock(&x->lock);
277 EXPORT_SYMBOL(xfrm_input);
279 int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
281 return xfrm_input(skb, nexthdr, 0, -1);
283 EXPORT_SYMBOL(xfrm_input_resume);
285 void __init xfrm_input_init(void)
287 secpath_cachep = kmem_cache_create("secpath_cache",
288 sizeof(struct sec_path),
289 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,