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);
113 memcpy(sp, src, sizeof(*sp));
114 for (i = 0; i < sp->len; i++)
115 xfrm_state_hold(sp->xvec[i]);
117 atomic_set(&sp->refcnt, 1);
120 EXPORT_SYMBOL(secpath_dup);
122 int secpath_set(struct sk_buff *skb)
126 /* Allocate new secpath or COW existing one. */
127 if (!skb->sp || atomic_read(&skb->sp->refcnt) != 1) {
128 sp = secpath_dup(skb->sp);
133 secpath_put(skb->sp);
138 EXPORT_SYMBOL(secpath_set);
140 /* Fetch spi and seq from ipsec header */
142 int xfrm_parse_spi(struct sk_buff *skb, u8 nexthdr, __be32 *spi, __be32 *seq)
144 int offset, offset_seq;
149 hlen = sizeof(struct ip_auth_hdr);
150 offset = offsetof(struct ip_auth_hdr, spi);
151 offset_seq = offsetof(struct ip_auth_hdr, seq_no);
154 hlen = sizeof(struct ip_esp_hdr);
155 offset = offsetof(struct ip_esp_hdr, spi);
156 offset_seq = offsetof(struct ip_esp_hdr, seq_no);
159 if (!pskb_may_pull(skb, sizeof(struct ip_comp_hdr)))
161 *spi = htonl(ntohs(*(__be16 *)(skb_transport_header(skb) + 2)));
168 if (!pskb_may_pull(skb, hlen))
171 *spi = *(__be32 *)(skb_transport_header(skb) + offset);
172 *seq = *(__be32 *)(skb_transport_header(skb) + offset_seq);
175 EXPORT_SYMBOL(xfrm_parse_spi);
177 int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
179 struct xfrm_mode *inner_mode = x->inner_mode;
182 err = x->outer_mode->afinfo->extract_input(x, skb);
186 if (x->sel.family == AF_UNSPEC) {
187 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
188 if (inner_mode == NULL)
189 return -EAFNOSUPPORT;
192 skb->protocol = inner_mode->afinfo->eth_proto;
193 return inner_mode->input2(x, skb);
195 EXPORT_SYMBOL(xfrm_prepare_input);
197 int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
199 struct net *net = dev_net(skb->dev);
203 struct xfrm_state *x = NULL;
204 xfrm_address_t *daddr;
205 struct xfrm_mode *inner_mode;
206 u32 mark = skb->mark;
210 struct xfrm_offload *xo;
211 bool xfrm_gro = false;
213 if (encap_type < 0) {
214 x = xfrm_input_state(skb);
215 family = x->outer_mode->afinfo->family;
217 /* An encap_type of -1 indicates async resumption. */
218 if (encap_type == -1) {
220 seq = XFRM_SKB_CB(skb)->seq.input.low;
223 /* encap_type < -1 indicates a GRO call. */
225 seq = XFRM_SPI_SKB_CB(skb)->seq;
229 daddr = (xfrm_address_t *)(skb_network_header(skb) +
230 XFRM_SPI_SKB_CB(skb)->daddroff);
231 family = XFRM_SPI_SKB_CB(skb)->family;
233 /* if tunnel is present override skb->mark value with tunnel i_key */
236 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4)
237 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4->parms.i_key);
240 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6)
241 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6->parms.i_key);
245 err = secpath_set(skb);
247 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
252 if (!spi && (err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
253 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
258 if (skb->sp->len == XFRM_MAX_DEPTH) {
259 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
263 x = xfrm_state_lookup(net, mark, daddr, spi, nexthdr, family);
265 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
266 xfrm_audit_state_notfound(skb, family, spi, seq);
270 skb->sp->xvec[skb->sp->len++] = x;
275 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
276 if (x->km.state == XFRM_STATE_ACQ)
277 XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
280 LINUX_MIB_XFRMINSTATEINVALID);
284 if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
285 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
289 if (x->repl->check(x, skb, seq)) {
290 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
294 if (xfrm_state_check_expire(x)) {
295 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEEXPIRED);
299 spin_unlock(&x->lock);
301 if (xfrm_tunnel_check(skb, x, family)) {
302 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
306 seq_hi = htonl(xfrm_replay_seqhi(x, seq));
308 XFRM_SKB_CB(skb)->seq.input.low = seq;
309 XFRM_SKB_CB(skb)->seq.input.hi = seq_hi;
314 nexthdr = x->type->input(x, skb);
316 if (nexthdr == -EINPROGRESS)
323 if (nexthdr == -EBADMSG) {
324 xfrm_audit_state_icvfail(x, skb,
326 x->stats.integrity_failed++;
328 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
332 /* only the first xfrm gets the encap type */
335 if (async && x->repl->recheck(x, skb, seq)) {
336 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
340 x->repl->advance(x, seq);
342 x->curlft.bytes += skb->len;
345 spin_unlock(&x->lock);
347 XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
349 inner_mode = x->inner_mode;
351 if (x->sel.family == AF_UNSPEC) {
352 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
353 if (inner_mode == NULL) {
354 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
359 if (inner_mode->input(x, skb)) {
360 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
364 if (x->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) {
370 * We need the inner address. However, we only get here for
371 * transport mode so the outer address is identical.
373 daddr = &x->id.daddr;
374 family = x->outer_mode->afinfo->family;
376 err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
378 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
383 err = xfrm_rcv_cb(skb, family, x->type->proto, 0);
391 gro_cells_receive(&gro_cells, skb);
394 xo = xfrm_offload(skb);
396 xfrm_gro = xo->flags & XFRM_GRO;
398 err = x->inner_mode->afinfo->transport_finish(skb, async);
401 gro_cells_receive(&gro_cells, skb);
409 spin_unlock(&x->lock);
411 xfrm_rcv_cb(skb, family, x && x->type ? x->type->proto : nexthdr, -1);
415 EXPORT_SYMBOL(xfrm_input);
417 int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
419 return xfrm_input(skb, nexthdr, 0, -1);
421 EXPORT_SYMBOL(xfrm_input_resume);
423 void __init xfrm_input_init(void)
427 init_dummy_netdev(&xfrm_napi_dev);
428 err = gro_cells_init(&gro_cells, &xfrm_napi_dev);
430 gro_cells.cells = NULL;
432 secpath_cachep = kmem_cache_create("secpath_cache",
433 sizeof(struct sec_path),
434 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,