1 // SPDX-License-Identifier: GPL-2.0
6 * YOSHIFUJI Hideaki @USAGI
7 * Split up af-specific portion
11 #include <linux/slab.h>
12 #include <linux/module.h>
13 #include <linux/netdevice.h>
17 #include <net/ip_tunnels.h>
18 #include <net/ip6_tunnel.h>
20 static struct kmem_cache *secpath_cachep __read_mostly;
22 static DEFINE_SPINLOCK(xfrm_input_afinfo_lock);
23 static struct xfrm_input_afinfo const __rcu *xfrm_input_afinfo[AF_INET6 + 1];
25 static struct gro_cells gro_cells;
26 static struct net_device xfrm_napi_dev;
28 int xfrm_input_register_afinfo(const struct xfrm_input_afinfo *afinfo)
32 if (WARN_ON(afinfo->family >= ARRAY_SIZE(xfrm_input_afinfo)))
35 spin_lock_bh(&xfrm_input_afinfo_lock);
36 if (unlikely(xfrm_input_afinfo[afinfo->family] != NULL))
39 rcu_assign_pointer(xfrm_input_afinfo[afinfo->family], afinfo);
40 spin_unlock_bh(&xfrm_input_afinfo_lock);
43 EXPORT_SYMBOL(xfrm_input_register_afinfo);
45 int xfrm_input_unregister_afinfo(const struct xfrm_input_afinfo *afinfo)
49 spin_lock_bh(&xfrm_input_afinfo_lock);
50 if (likely(xfrm_input_afinfo[afinfo->family] != NULL)) {
51 if (unlikely(xfrm_input_afinfo[afinfo->family] != afinfo))
54 RCU_INIT_POINTER(xfrm_input_afinfo[afinfo->family], NULL);
56 spin_unlock_bh(&xfrm_input_afinfo_lock);
60 EXPORT_SYMBOL(xfrm_input_unregister_afinfo);
62 static const struct xfrm_input_afinfo *xfrm_input_get_afinfo(unsigned int family)
64 const struct xfrm_input_afinfo *afinfo;
66 if (WARN_ON_ONCE(family >= ARRAY_SIZE(xfrm_input_afinfo)))
70 afinfo = rcu_dereference(xfrm_input_afinfo[family]);
71 if (unlikely(!afinfo))
76 static int xfrm_rcv_cb(struct sk_buff *skb, unsigned int family, u8 protocol,
80 const struct xfrm_input_afinfo *afinfo = xfrm_input_get_afinfo(family);
85 ret = afinfo->callback(skb, protocol, err);
91 void __secpath_destroy(struct sec_path *sp)
94 for (i = 0; i < sp->len; i++)
95 xfrm_state_put(sp->xvec[i]);
96 kmem_cache_free(secpath_cachep, sp);
98 EXPORT_SYMBOL(__secpath_destroy);
100 struct sec_path *secpath_dup(struct sec_path *src)
104 sp = kmem_cache_alloc(secpath_cachep, GFP_ATOMIC);
111 memset(sp->ovec, 0, sizeof(sp->ovec[XFRM_MAX_OFFLOAD_DEPTH]));
116 memcpy(sp, src, sizeof(*sp));
117 for (i = 0; i < sp->len; i++)
118 xfrm_state_hold(sp->xvec[i]);
120 refcount_set(&sp->refcnt, 1);
123 EXPORT_SYMBOL(secpath_dup);
125 int secpath_set(struct sk_buff *skb)
129 /* Allocate new secpath or COW existing one. */
130 if (!skb->sp || refcount_read(&skb->sp->refcnt) != 1) {
131 sp = secpath_dup(skb->sp);
136 secpath_put(skb->sp);
141 EXPORT_SYMBOL(secpath_set);
143 /* Fetch spi and seq from ipsec header */
145 int xfrm_parse_spi(struct sk_buff *skb, u8 nexthdr, __be32 *spi, __be32 *seq)
147 int offset, offset_seq;
152 hlen = sizeof(struct ip_auth_hdr);
153 offset = offsetof(struct ip_auth_hdr, spi);
154 offset_seq = offsetof(struct ip_auth_hdr, seq_no);
157 hlen = sizeof(struct ip_esp_hdr);
158 offset = offsetof(struct ip_esp_hdr, spi);
159 offset_seq = offsetof(struct ip_esp_hdr, seq_no);
162 if (!pskb_may_pull(skb, sizeof(struct ip_comp_hdr)))
164 *spi = htonl(ntohs(*(__be16 *)(skb_transport_header(skb) + 2)));
171 if (!pskb_may_pull(skb, hlen))
174 *spi = *(__be32 *)(skb_transport_header(skb) + offset);
175 *seq = *(__be32 *)(skb_transport_header(skb) + offset_seq);
178 EXPORT_SYMBOL(xfrm_parse_spi);
180 int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
182 struct xfrm_mode *inner_mode = x->inner_mode;
185 err = x->outer_mode->afinfo->extract_input(x, skb);
189 if (x->sel.family == AF_UNSPEC) {
190 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
191 if (inner_mode == NULL)
192 return -EAFNOSUPPORT;
195 skb->protocol = inner_mode->afinfo->eth_proto;
196 return inner_mode->input2(x, skb);
198 EXPORT_SYMBOL(xfrm_prepare_input);
200 int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
202 struct net *net = dev_net(skb->dev);
206 struct xfrm_state *x = NULL;
207 xfrm_address_t *daddr;
208 struct xfrm_mode *inner_mode;
209 u32 mark = skb->mark;
213 bool xfrm_gro = false;
214 bool crypto_done = false;
215 struct xfrm_offload *xo = xfrm_offload(skb);
217 if (encap_type < 0) {
218 x = xfrm_input_state(skb);
219 family = x->outer_mode->afinfo->family;
221 /* An encap_type of -1 indicates async resumption. */
222 if (encap_type == -1) {
224 seq = XFRM_SKB_CB(skb)->seq.input.low;
228 /* encap_type < -1 indicates a GRO call. */
230 seq = XFRM_SPI_SKB_CB(skb)->seq;
232 if (xo && (xo->flags & CRYPTO_DONE)) {
234 x = xfrm_input_state(skb);
235 family = XFRM_SPI_SKB_CB(skb)->family;
237 if (!(xo->status & CRYPTO_SUCCESS)) {
239 (CRYPTO_TRANSPORT_AH_AUTH_FAILED |
240 CRYPTO_TRANSPORT_ESP_AUTH_FAILED |
241 CRYPTO_TUNNEL_AH_AUTH_FAILED |
242 CRYPTO_TUNNEL_ESP_AUTH_FAILED)) {
244 xfrm_audit_state_icvfail(x, skb,
246 x->stats.integrity_failed++;
247 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
251 if (xo->status & CRYPTO_INVALID_PROTOCOL) {
252 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
256 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
260 if ((err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
261 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
269 family = XFRM_SPI_SKB_CB(skb)->family;
271 /* if tunnel is present override skb->mark value with tunnel i_key */
274 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4)
275 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4->parms.i_key);
278 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6)
279 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6->parms.i_key);
283 err = secpath_set(skb);
285 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
290 if (!spi && (err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
291 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
295 daddr = (xfrm_address_t *)(skb_network_header(skb) +
296 XFRM_SPI_SKB_CB(skb)->daddroff);
298 if (skb->sp->len == XFRM_MAX_DEPTH) {
299 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
303 x = xfrm_state_lookup(net, mark, daddr, spi, nexthdr, family);
305 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
306 xfrm_audit_state_notfound(skb, family, spi, seq);
310 skb->sp->xvec[skb->sp->len++] = x;
315 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
316 if (x->km.state == XFRM_STATE_ACQ)
317 XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
320 LINUX_MIB_XFRMINSTATEINVALID);
324 if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
325 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
329 if (x->repl->check(x, skb, seq)) {
330 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
334 if (xfrm_state_check_expire(x)) {
335 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEEXPIRED);
339 spin_unlock(&x->lock);
341 if (xfrm_tunnel_check(skb, x, family)) {
342 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
346 seq_hi = htonl(xfrm_replay_seqhi(x, seq));
348 XFRM_SKB_CB(skb)->seq.input.low = seq;
349 XFRM_SKB_CB(skb)->seq.input.hi = seq_hi;
355 nexthdr = x->type_offload->input_tail(x, skb);
357 nexthdr = x->type->input(x, skb);
359 if (nexthdr == -EINPROGRESS)
366 if (nexthdr == -EBADMSG) {
367 xfrm_audit_state_icvfail(x, skb,
369 x->stats.integrity_failed++;
371 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
375 /* only the first xfrm gets the encap type */
378 if (async && x->repl->recheck(x, skb, seq)) {
379 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
383 x->repl->advance(x, seq);
385 x->curlft.bytes += skb->len;
388 spin_unlock(&x->lock);
390 XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
392 inner_mode = x->inner_mode;
394 if (x->sel.family == AF_UNSPEC) {
395 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
396 if (inner_mode == NULL) {
397 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
402 if (inner_mode->input(x, skb)) {
403 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
407 if (x->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) {
413 * We need the inner address. However, we only get here for
414 * transport mode so the outer address is identical.
416 daddr = &x->id.daddr;
417 family = x->outer_mode->afinfo->family;
419 err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
421 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
426 err = xfrm_rcv_cb(skb, family, x->type->proto, 0);
436 gro_cells_receive(&gro_cells, skb);
439 xo = xfrm_offload(skb);
441 xfrm_gro = xo->flags & XFRM_GRO;
443 err = x->inner_mode->afinfo->transport_finish(skb, xfrm_gro || async);
448 gro_cells_receive(&gro_cells, skb);
456 spin_unlock(&x->lock);
458 xfrm_rcv_cb(skb, family, x && x->type ? x->type->proto : nexthdr, -1);
462 EXPORT_SYMBOL(xfrm_input);
464 int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
466 return xfrm_input(skb, nexthdr, 0, -1);
468 EXPORT_SYMBOL(xfrm_input_resume);
470 void __init xfrm_input_init(void)
474 init_dummy_netdev(&xfrm_napi_dev);
475 err = gro_cells_init(&gro_cells, &xfrm_napi_dev);
477 gro_cells.cells = NULL;
479 secpath_cachep = kmem_cache_create("secpath_cache",
480 sizeof(struct sec_path),
481 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,