Loading...
1/*
2 * xfrm_input.c
3 *
4 * Changes:
5 * YOSHIFUJI Hideaki @USAGI
6 * Split up af-specific portion
7 *
8 */
9
10#include <linux/slab.h>
11#include <linux/module.h>
12#include <linux/netdevice.h>
13#include <net/dst.h>
14#include <net/ip.h>
15#include <net/xfrm.h>
16
17static struct kmem_cache *secpath_cachep __read_mostly;
18
19static DEFINE_SPINLOCK(xfrm_input_afinfo_lock);
20static struct xfrm_input_afinfo __rcu *xfrm_input_afinfo[NPROTO];
21
22int xfrm_input_register_afinfo(struct xfrm_input_afinfo *afinfo)
23{
24 int err = 0;
25
26 if (unlikely(afinfo == NULL))
27 return -EINVAL;
28 if (unlikely(afinfo->family >= NPROTO))
29 return -EAFNOSUPPORT;
30 spin_lock_bh(&xfrm_input_afinfo_lock);
31 if (unlikely(xfrm_input_afinfo[afinfo->family] != NULL))
32 err = -ENOBUFS;
33 else
34 rcu_assign_pointer(xfrm_input_afinfo[afinfo->family], afinfo);
35 spin_unlock_bh(&xfrm_input_afinfo_lock);
36 return err;
37}
38EXPORT_SYMBOL(xfrm_input_register_afinfo);
39
40int xfrm_input_unregister_afinfo(struct xfrm_input_afinfo *afinfo)
41{
42 int err = 0;
43
44 if (unlikely(afinfo == NULL))
45 return -EINVAL;
46 if (unlikely(afinfo->family >= NPROTO))
47 return -EAFNOSUPPORT;
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))
51 err = -EINVAL;
52 else
53 RCU_INIT_POINTER(xfrm_input_afinfo[afinfo->family], NULL);
54 }
55 spin_unlock_bh(&xfrm_input_afinfo_lock);
56 synchronize_rcu();
57 return err;
58}
59EXPORT_SYMBOL(xfrm_input_unregister_afinfo);
60
61static struct xfrm_input_afinfo *xfrm_input_get_afinfo(unsigned int family)
62{
63 struct xfrm_input_afinfo *afinfo;
64
65 if (unlikely(family >= NPROTO))
66 return NULL;
67 rcu_read_lock();
68 afinfo = rcu_dereference(xfrm_input_afinfo[family]);
69 if (unlikely(!afinfo))
70 rcu_read_unlock();
71 return afinfo;
72}
73
74static void xfrm_input_put_afinfo(struct xfrm_input_afinfo *afinfo)
75{
76 rcu_read_unlock();
77}
78
79static int xfrm_rcv_cb(struct sk_buff *skb, unsigned int family, u8 protocol,
80 int err)
81{
82 int ret;
83 struct xfrm_input_afinfo *afinfo = xfrm_input_get_afinfo(family);
84
85 if (!afinfo)
86 return -EAFNOSUPPORT;
87
88 ret = afinfo->callback(skb, protocol, err);
89 xfrm_input_put_afinfo(afinfo);
90
91 return ret;
92}
93
94void __secpath_destroy(struct sec_path *sp)
95{
96 int i;
97 for (i = 0; i < sp->len; i++)
98 xfrm_state_put(sp->xvec[i]);
99 kmem_cache_free(secpath_cachep, sp);
100}
101EXPORT_SYMBOL(__secpath_destroy);
102
103struct sec_path *secpath_dup(struct sec_path *src)
104{
105 struct sec_path *sp;
106
107 sp = kmem_cache_alloc(secpath_cachep, GFP_ATOMIC);
108 if (!sp)
109 return NULL;
110
111 sp->len = 0;
112 if (src) {
113 int i;
114
115 memcpy(sp, src, sizeof(*sp));
116 for (i = 0; i < sp->len; i++)
117 xfrm_state_hold(sp->xvec[i]);
118 }
119 atomic_set(&sp->refcnt, 1);
120 return sp;
121}
122EXPORT_SYMBOL(secpath_dup);
123
124/* Fetch spi and seq from ipsec header */
125
126int xfrm_parse_spi(struct sk_buff *skb, u8 nexthdr, __be32 *spi, __be32 *seq)
127{
128 int offset, offset_seq;
129 int hlen;
130
131 switch (nexthdr) {
132 case IPPROTO_AH:
133 hlen = sizeof(struct ip_auth_hdr);
134 offset = offsetof(struct ip_auth_hdr, spi);
135 offset_seq = offsetof(struct ip_auth_hdr, seq_no);
136 break;
137 case IPPROTO_ESP:
138 hlen = sizeof(struct ip_esp_hdr);
139 offset = offsetof(struct ip_esp_hdr, spi);
140 offset_seq = offsetof(struct ip_esp_hdr, seq_no);
141 break;
142 case IPPROTO_COMP:
143 if (!pskb_may_pull(skb, sizeof(struct ip_comp_hdr)))
144 return -EINVAL;
145 *spi = htonl(ntohs(*(__be16 *)(skb_transport_header(skb) + 2)));
146 *seq = 0;
147 return 0;
148 default:
149 return 1;
150 }
151
152 if (!pskb_may_pull(skb, hlen))
153 return -EINVAL;
154
155 *spi = *(__be32 *)(skb_transport_header(skb) + offset);
156 *seq = *(__be32 *)(skb_transport_header(skb) + offset_seq);
157 return 0;
158}
159
160int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
161{
162 struct xfrm_mode *inner_mode = x->inner_mode;
163 int err;
164
165 err = x->outer_mode->afinfo->extract_input(x, skb);
166 if (err)
167 return err;
168
169 if (x->sel.family == AF_UNSPEC) {
170 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
171 if (inner_mode == NULL)
172 return -EAFNOSUPPORT;
173 }
174
175 skb->protocol = inner_mode->afinfo->eth_proto;
176 return inner_mode->input2(x, skb);
177}
178EXPORT_SYMBOL(xfrm_prepare_input);
179
180int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
181{
182 struct net *net = dev_net(skb->dev);
183 int err;
184 __be32 seq;
185 __be32 seq_hi;
186 struct xfrm_state *x = NULL;
187 xfrm_address_t *daddr;
188 struct xfrm_mode *inner_mode;
189 unsigned int family;
190 int decaps = 0;
191 int async = 0;
192
193 /* A negative encap_type indicates async resumption. */
194 if (encap_type < 0) {
195 async = 1;
196 x = xfrm_input_state(skb);
197 seq = XFRM_SKB_CB(skb)->seq.input.low;
198 family = x->outer_mode->afinfo->family;
199 goto resume;
200 }
201
202 daddr = (xfrm_address_t *)(skb_network_header(skb) +
203 XFRM_SPI_SKB_CB(skb)->daddroff);
204 family = XFRM_SPI_SKB_CB(skb)->family;
205
206 /* Allocate new secpath or COW existing one. */
207 if (!skb->sp || atomic_read(&skb->sp->refcnt) != 1) {
208 struct sec_path *sp;
209
210 sp = secpath_dup(skb->sp);
211 if (!sp) {
212 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
213 goto drop;
214 }
215 if (skb->sp)
216 secpath_put(skb->sp);
217 skb->sp = sp;
218 }
219
220 seq = 0;
221 if (!spi && (err = xfrm_parse_spi(skb, nexthdr, &spi, &seq)) != 0) {
222 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
223 goto drop;
224 }
225
226 do {
227 if (skb->sp->len == XFRM_MAX_DEPTH) {
228 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
229 goto drop;
230 }
231
232 x = xfrm_state_lookup(net, skb->mark, daddr, spi, nexthdr, family);
233 if (x == NULL) {
234 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
235 xfrm_audit_state_notfound(skb, family, spi, seq);
236 goto drop;
237 }
238
239 skb->sp->xvec[skb->sp->len++] = x;
240
241 if (xfrm_tunnel_check(skb, x, family)) {
242 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
243 goto drop;
244 }
245
246 spin_lock(&x->lock);
247 if (unlikely(x->km.state == XFRM_STATE_ACQ)) {
248 XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
249 goto drop_unlock;
250 }
251
252 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
253 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEINVALID);
254 goto drop_unlock;
255 }
256
257 if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
258 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
259 goto drop_unlock;
260 }
261
262 if (x->repl->check(x, skb, seq)) {
263 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
264 goto drop_unlock;
265 }
266
267 if (xfrm_state_check_expire(x)) {
268 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEEXPIRED);
269 goto drop_unlock;
270 }
271
272 spin_unlock(&x->lock);
273
274 seq_hi = htonl(xfrm_replay_seqhi(x, seq));
275
276 XFRM_SKB_CB(skb)->seq.input.low = seq;
277 XFRM_SKB_CB(skb)->seq.input.hi = seq_hi;
278
279 skb_dst_force(skb);
280
281 nexthdr = x->type->input(x, skb);
282
283 if (nexthdr == -EINPROGRESS)
284 return 0;
285resume:
286 spin_lock(&x->lock);
287 if (nexthdr <= 0) {
288 if (nexthdr == -EBADMSG) {
289 xfrm_audit_state_icvfail(x, skb,
290 x->type->proto);
291 x->stats.integrity_failed++;
292 }
293 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
294 goto drop_unlock;
295 }
296
297 /* only the first xfrm gets the encap type */
298 encap_type = 0;
299
300 if (async && x->repl->recheck(x, skb, seq)) {
301 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
302 goto drop_unlock;
303 }
304
305 x->repl->advance(x, seq);
306
307 x->curlft.bytes += skb->len;
308 x->curlft.packets++;
309
310 spin_unlock(&x->lock);
311
312 XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
313
314 inner_mode = x->inner_mode;
315
316 if (x->sel.family == AF_UNSPEC) {
317 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
318 if (inner_mode == NULL)
319 goto drop;
320 }
321
322 if (inner_mode->input(x, skb)) {
323 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
324 goto drop;
325 }
326
327 if (x->outer_mode->flags & XFRM_MODE_FLAG_TUNNEL) {
328 decaps = 1;
329 break;
330 }
331
332 /*
333 * We need the inner address. However, we only get here for
334 * transport mode so the outer address is identical.
335 */
336 daddr = &x->id.daddr;
337 family = x->outer_mode->afinfo->family;
338
339 err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
340 if (err < 0) {
341 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
342 goto drop;
343 }
344 } while (!err);
345
346 err = xfrm_rcv_cb(skb, family, x->type->proto, 0);
347 if (err)
348 goto drop;
349
350 nf_reset(skb);
351
352 if (decaps) {
353 skb_dst_drop(skb);
354 netif_rx(skb);
355 return 0;
356 } else {
357 return x->inner_mode->afinfo->transport_finish(skb, async);
358 }
359
360drop_unlock:
361 spin_unlock(&x->lock);
362drop:
363 xfrm_rcv_cb(skb, family, x && x->type ? x->type->proto : nexthdr, -1);
364 kfree_skb(skb);
365 return 0;
366}
367EXPORT_SYMBOL(xfrm_input);
368
369int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
370{
371 return xfrm_input(skb, nexthdr, 0, -1);
372}
373EXPORT_SYMBOL(xfrm_input_resume);
374
375void __init xfrm_input_init(void)
376{
377 secpath_cachep = kmem_cache_create("secpath_cache",
378 sizeof(struct sec_path),
379 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
380 NULL);
381}
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * xfrm_input.c
4 *
5 * Changes:
6 * YOSHIFUJI Hideaki @USAGI
7 * Split up af-specific portion
8 *
9 */
10
11#include <linux/bottom_half.h>
12#include <linux/cache.h>
13#include <linux/interrupt.h>
14#include <linux/slab.h>
15#include <linux/module.h>
16#include <linux/netdevice.h>
17#include <linux/percpu.h>
18#include <net/dst.h>
19#include <net/ip.h>
20#include <net/xfrm.h>
21#include <net/ip_tunnels.h>
22#include <net/ip6_tunnel.h>
23#include <net/dst_metadata.h>
24#include <net/hotdata.h>
25
26#include "xfrm_inout.h"
27
28struct xfrm_trans_tasklet {
29 struct work_struct work;
30 spinlock_t queue_lock;
31 struct sk_buff_head queue;
32};
33
34struct xfrm_trans_cb {
35 union {
36 struct inet_skb_parm h4;
37#if IS_ENABLED(CONFIG_IPV6)
38 struct inet6_skb_parm h6;
39#endif
40 } header;
41 int (*finish)(struct net *net, struct sock *sk, struct sk_buff *skb);
42 struct net *net;
43};
44
45#define XFRM_TRANS_SKB_CB(__skb) ((struct xfrm_trans_cb *)&((__skb)->cb[0]))
46
47static DEFINE_SPINLOCK(xfrm_input_afinfo_lock);
48static struct xfrm_input_afinfo const __rcu *xfrm_input_afinfo[2][AF_INET6 + 1];
49
50static struct gro_cells gro_cells;
51static struct net_device xfrm_napi_dev;
52
53static DEFINE_PER_CPU(struct xfrm_trans_tasklet, xfrm_trans_tasklet);
54
55int xfrm_input_register_afinfo(const struct xfrm_input_afinfo *afinfo)
56{
57 int err = 0;
58
59 if (WARN_ON(afinfo->family > AF_INET6))
60 return -EAFNOSUPPORT;
61
62 spin_lock_bh(&xfrm_input_afinfo_lock);
63 if (unlikely(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family]))
64 err = -EEXIST;
65 else
66 rcu_assign_pointer(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family], afinfo);
67 spin_unlock_bh(&xfrm_input_afinfo_lock);
68 return err;
69}
70EXPORT_SYMBOL(xfrm_input_register_afinfo);
71
72int xfrm_input_unregister_afinfo(const struct xfrm_input_afinfo *afinfo)
73{
74 int err = 0;
75
76 spin_lock_bh(&xfrm_input_afinfo_lock);
77 if (likely(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family])) {
78 if (unlikely(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family] != afinfo))
79 err = -EINVAL;
80 else
81 RCU_INIT_POINTER(xfrm_input_afinfo[afinfo->is_ipip][afinfo->family], NULL);
82 }
83 spin_unlock_bh(&xfrm_input_afinfo_lock);
84 synchronize_rcu();
85 return err;
86}
87EXPORT_SYMBOL(xfrm_input_unregister_afinfo);
88
89static const struct xfrm_input_afinfo *xfrm_input_get_afinfo(u8 family, bool is_ipip)
90{
91 const struct xfrm_input_afinfo *afinfo;
92
93 if (WARN_ON_ONCE(family > AF_INET6))
94 return NULL;
95
96 rcu_read_lock();
97 afinfo = rcu_dereference(xfrm_input_afinfo[is_ipip][family]);
98 if (unlikely(!afinfo))
99 rcu_read_unlock();
100 return afinfo;
101}
102
103static int xfrm_rcv_cb(struct sk_buff *skb, unsigned int family, u8 protocol,
104 int err)
105{
106 bool is_ipip = (protocol == IPPROTO_IPIP || protocol == IPPROTO_IPV6);
107 const struct xfrm_input_afinfo *afinfo;
108 int ret;
109
110 afinfo = xfrm_input_get_afinfo(family, is_ipip);
111 if (!afinfo)
112 return -EAFNOSUPPORT;
113
114 ret = afinfo->callback(skb, protocol, err);
115 rcu_read_unlock();
116
117 return ret;
118}
119
120struct sec_path *secpath_set(struct sk_buff *skb)
121{
122 struct sec_path *sp, *tmp = skb_ext_find(skb, SKB_EXT_SEC_PATH);
123
124 sp = skb_ext_add(skb, SKB_EXT_SEC_PATH);
125 if (!sp)
126 return NULL;
127
128 if (tmp) /* reused existing one (was COW'd if needed) */
129 return sp;
130
131 /* allocated new secpath */
132 memset(sp->ovec, 0, sizeof(sp->ovec));
133 sp->olen = 0;
134 sp->len = 0;
135 sp->verified_cnt = 0;
136
137 return sp;
138}
139EXPORT_SYMBOL(secpath_set);
140
141/* Fetch spi and seq from ipsec header */
142
143int xfrm_parse_spi(struct sk_buff *skb, u8 nexthdr, __be32 *spi, __be32 *seq)
144{
145 int offset, offset_seq;
146 int hlen;
147
148 switch (nexthdr) {
149 case IPPROTO_AH:
150 hlen = sizeof(struct ip_auth_hdr);
151 offset = offsetof(struct ip_auth_hdr, spi);
152 offset_seq = offsetof(struct ip_auth_hdr, seq_no);
153 break;
154 case IPPROTO_ESP:
155 hlen = sizeof(struct ip_esp_hdr);
156 offset = offsetof(struct ip_esp_hdr, spi);
157 offset_seq = offsetof(struct ip_esp_hdr, seq_no);
158 break;
159 case IPPROTO_COMP:
160 if (!pskb_may_pull(skb, sizeof(struct ip_comp_hdr)))
161 return -EINVAL;
162 *spi = htonl(ntohs(*(__be16 *)(skb_transport_header(skb) + 2)));
163 *seq = 0;
164 return 0;
165 default:
166 return 1;
167 }
168
169 if (!pskb_may_pull(skb, hlen))
170 return -EINVAL;
171
172 *spi = *(__be32 *)(skb_transport_header(skb) + offset);
173 *seq = *(__be32 *)(skb_transport_header(skb) + offset_seq);
174 return 0;
175}
176EXPORT_SYMBOL(xfrm_parse_spi);
177
178static int xfrm4_remove_beet_encap(struct xfrm_state *x, struct sk_buff *skb)
179{
180 struct iphdr *iph;
181 int optlen = 0;
182 int err = -EINVAL;
183
184 skb->protocol = htons(ETH_P_IP);
185
186 if (unlikely(XFRM_MODE_SKB_CB(skb)->protocol == IPPROTO_BEETPH)) {
187 struct ip_beet_phdr *ph;
188 int phlen;
189
190 if (!pskb_may_pull(skb, sizeof(*ph)))
191 goto out;
192
193 ph = (struct ip_beet_phdr *)skb->data;
194
195 phlen = sizeof(*ph) + ph->padlen;
196 optlen = ph->hdrlen * 8 + (IPV4_BEET_PHMAXLEN - phlen);
197 if (optlen < 0 || optlen & 3 || optlen > 250)
198 goto out;
199
200 XFRM_MODE_SKB_CB(skb)->protocol = ph->nexthdr;
201
202 if (!pskb_may_pull(skb, phlen))
203 goto out;
204 __skb_pull(skb, phlen);
205 }
206
207 skb_push(skb, sizeof(*iph));
208 skb_reset_network_header(skb);
209 skb_mac_header_rebuild(skb);
210
211 xfrm4_beet_make_header(skb);
212
213 iph = ip_hdr(skb);
214
215 iph->ihl += optlen / 4;
216 iph->tot_len = htons(skb->len);
217 iph->daddr = x->sel.daddr.a4;
218 iph->saddr = x->sel.saddr.a4;
219 iph->check = 0;
220 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
221 err = 0;
222out:
223 return err;
224}
225
226static void ipip_ecn_decapsulate(struct sk_buff *skb)
227{
228 struct iphdr *inner_iph = ipip_hdr(skb);
229
230 if (INET_ECN_is_ce(XFRM_MODE_SKB_CB(skb)->tos))
231 IP_ECN_set_ce(inner_iph);
232}
233
234static int xfrm4_remove_tunnel_encap(struct xfrm_state *x, struct sk_buff *skb)
235{
236 int err = -EINVAL;
237
238 skb->protocol = htons(ETH_P_IP);
239
240 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
241 goto out;
242
243 err = skb_unclone(skb, GFP_ATOMIC);
244 if (err)
245 goto out;
246
247 if (x->props.flags & XFRM_STATE_DECAP_DSCP)
248 ipv4_copy_dscp(XFRM_MODE_SKB_CB(skb)->tos, ipip_hdr(skb));
249 if (!(x->props.flags & XFRM_STATE_NOECN))
250 ipip_ecn_decapsulate(skb);
251
252 skb_reset_network_header(skb);
253 skb_mac_header_rebuild(skb);
254 if (skb->mac_len)
255 eth_hdr(skb)->h_proto = skb->protocol;
256
257 err = 0;
258
259out:
260 return err;
261}
262
263static void ipip6_ecn_decapsulate(struct sk_buff *skb)
264{
265 struct ipv6hdr *inner_iph = ipipv6_hdr(skb);
266
267 if (INET_ECN_is_ce(XFRM_MODE_SKB_CB(skb)->tos))
268 IP6_ECN_set_ce(skb, inner_iph);
269}
270
271static int xfrm6_remove_tunnel_encap(struct xfrm_state *x, struct sk_buff *skb)
272{
273 int err = -EINVAL;
274
275 skb->protocol = htons(ETH_P_IPV6);
276
277 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
278 goto out;
279
280 err = skb_unclone(skb, GFP_ATOMIC);
281 if (err)
282 goto out;
283
284 if (x->props.flags & XFRM_STATE_DECAP_DSCP)
285 ipv6_copy_dscp(XFRM_MODE_SKB_CB(skb)->tos, ipipv6_hdr(skb));
286 if (!(x->props.flags & XFRM_STATE_NOECN))
287 ipip6_ecn_decapsulate(skb);
288
289 skb_reset_network_header(skb);
290 skb_mac_header_rebuild(skb);
291 if (skb->mac_len)
292 eth_hdr(skb)->h_proto = skb->protocol;
293
294 err = 0;
295
296out:
297 return err;
298}
299
300static int xfrm6_remove_beet_encap(struct xfrm_state *x, struct sk_buff *skb)
301{
302 struct ipv6hdr *ip6h;
303 int size = sizeof(struct ipv6hdr);
304 int err;
305
306 skb->protocol = htons(ETH_P_IPV6);
307
308 err = skb_cow_head(skb, size + skb->mac_len);
309 if (err)
310 goto out;
311
312 __skb_push(skb, size);
313 skb_reset_network_header(skb);
314 skb_mac_header_rebuild(skb);
315
316 xfrm6_beet_make_header(skb);
317
318 ip6h = ipv6_hdr(skb);
319 ip6h->payload_len = htons(skb->len - size);
320 ip6h->daddr = x->sel.daddr.in6;
321 ip6h->saddr = x->sel.saddr.in6;
322 err = 0;
323out:
324 return err;
325}
326
327/* Remove encapsulation header.
328 *
329 * The IP header will be moved over the top of the encapsulation
330 * header.
331 *
332 * On entry, the transport header shall point to where the IP header
333 * should be and the network header shall be set to where the IP
334 * header currently is. skb->data shall point to the start of the
335 * payload.
336 */
337static int
338xfrm_inner_mode_encap_remove(struct xfrm_state *x,
339 struct sk_buff *skb)
340{
341 switch (x->props.mode) {
342 case XFRM_MODE_BEET:
343 switch (x->sel.family) {
344 case AF_INET:
345 return xfrm4_remove_beet_encap(x, skb);
346 case AF_INET6:
347 return xfrm6_remove_beet_encap(x, skb);
348 }
349 break;
350 case XFRM_MODE_TUNNEL:
351 switch (XFRM_MODE_SKB_CB(skb)->protocol) {
352 case IPPROTO_IPIP:
353 return xfrm4_remove_tunnel_encap(x, skb);
354 case IPPROTO_IPV6:
355 return xfrm6_remove_tunnel_encap(x, skb);
356 break;
357 }
358 return -EINVAL;
359 }
360
361 WARN_ON_ONCE(1);
362 return -EOPNOTSUPP;
363}
364
365static int xfrm_prepare_input(struct xfrm_state *x, struct sk_buff *skb)
366{
367 switch (x->props.family) {
368 case AF_INET:
369 xfrm4_extract_header(skb);
370 break;
371 case AF_INET6:
372 xfrm6_extract_header(skb);
373 break;
374 default:
375 WARN_ON_ONCE(1);
376 return -EAFNOSUPPORT;
377 }
378
379 return xfrm_inner_mode_encap_remove(x, skb);
380}
381
382/* Remove encapsulation header.
383 *
384 * The IP header will be moved over the top of the encapsulation header.
385 *
386 * On entry, skb_transport_header() shall point to where the IP header
387 * should be and skb_network_header() shall be set to where the IP header
388 * currently is. skb->data shall point to the start of the payload.
389 */
390static int xfrm4_transport_input(struct xfrm_state *x, struct sk_buff *skb)
391{
392 struct xfrm_offload *xo = xfrm_offload(skb);
393 int ihl = skb->data - skb_transport_header(skb);
394
395 if (skb->transport_header != skb->network_header) {
396 memmove(skb_transport_header(skb),
397 skb_network_header(skb), ihl);
398 if (xo)
399 xo->orig_mac_len =
400 skb_mac_header_was_set(skb) ? skb_mac_header_len(skb) : 0;
401 skb->network_header = skb->transport_header;
402 }
403 ip_hdr(skb)->tot_len = htons(skb->len + ihl);
404 skb_reset_transport_header(skb);
405 return 0;
406}
407
408static int xfrm6_transport_input(struct xfrm_state *x, struct sk_buff *skb)
409{
410#if IS_ENABLED(CONFIG_IPV6)
411 struct xfrm_offload *xo = xfrm_offload(skb);
412 int ihl = skb->data - skb_transport_header(skb);
413
414 if (skb->transport_header != skb->network_header) {
415 memmove(skb_transport_header(skb),
416 skb_network_header(skb), ihl);
417 if (xo)
418 xo->orig_mac_len =
419 skb_mac_header_was_set(skb) ? skb_mac_header_len(skb) : 0;
420 skb->network_header = skb->transport_header;
421 }
422 ipv6_hdr(skb)->payload_len = htons(skb->len + ihl -
423 sizeof(struct ipv6hdr));
424 skb_reset_transport_header(skb);
425 return 0;
426#else
427 WARN_ON_ONCE(1);
428 return -EAFNOSUPPORT;
429#endif
430}
431
432static int xfrm_inner_mode_input(struct xfrm_state *x,
433 struct sk_buff *skb)
434{
435 switch (x->props.mode) {
436 case XFRM_MODE_BEET:
437 case XFRM_MODE_TUNNEL:
438 return xfrm_prepare_input(x, skb);
439 case XFRM_MODE_TRANSPORT:
440 if (x->props.family == AF_INET)
441 return xfrm4_transport_input(x, skb);
442 if (x->props.family == AF_INET6)
443 return xfrm6_transport_input(x, skb);
444 break;
445 case XFRM_MODE_ROUTEOPTIMIZATION:
446 WARN_ON_ONCE(1);
447 break;
448 default:
449 WARN_ON_ONCE(1);
450 break;
451 }
452
453 return -EOPNOTSUPP;
454}
455
456int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
457{
458 const struct xfrm_state_afinfo *afinfo;
459 struct net *net = dev_net(skb->dev);
460 int err;
461 __be32 seq;
462 __be32 seq_hi;
463 struct xfrm_state *x = NULL;
464 xfrm_address_t *daddr;
465 u32 mark = skb->mark;
466 unsigned int family = AF_UNSPEC;
467 int decaps = 0;
468 int async = 0;
469 bool xfrm_gro = false;
470 bool crypto_done = false;
471 struct xfrm_offload *xo = xfrm_offload(skb);
472 struct sec_path *sp;
473
474 if (encap_type < 0 || (xo && xo->flags & XFRM_GRO)) {
475 x = xfrm_input_state(skb);
476
477 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
478 if (x->km.state == XFRM_STATE_ACQ)
479 XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
480 else
481 XFRM_INC_STATS(net,
482 LINUX_MIB_XFRMINSTATEINVALID);
483
484 if (encap_type == -1)
485 dev_put(skb->dev);
486 goto drop;
487 }
488
489 family = x->props.family;
490
491 /* An encap_type of -1 indicates async resumption. */
492 if (encap_type == -1) {
493 async = 1;
494 seq = XFRM_SKB_CB(skb)->seq.input.low;
495 goto resume;
496 }
497 /* GRO call */
498 seq = XFRM_SPI_SKB_CB(skb)->seq;
499
500 if (xo && (xo->flags & CRYPTO_DONE)) {
501 crypto_done = true;
502 family = XFRM_SPI_SKB_CB(skb)->family;
503
504 if (!(xo->status & CRYPTO_SUCCESS)) {
505 if (xo->status &
506 (CRYPTO_TRANSPORT_AH_AUTH_FAILED |
507 CRYPTO_TRANSPORT_ESP_AUTH_FAILED |
508 CRYPTO_TUNNEL_AH_AUTH_FAILED |
509 CRYPTO_TUNNEL_ESP_AUTH_FAILED)) {
510
511 xfrm_audit_state_icvfail(x, skb,
512 x->type->proto);
513 x->stats.integrity_failed++;
514 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
515 goto drop;
516 }
517
518 if (xo->status & CRYPTO_INVALID_PROTOCOL) {
519 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
520 goto drop;
521 }
522
523 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
524 goto drop;
525 }
526
527 if (xfrm_parse_spi(skb, nexthdr, &spi, &seq)) {
528 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
529 goto drop;
530 }
531 }
532
533 goto lock;
534 }
535
536 family = XFRM_SPI_SKB_CB(skb)->family;
537
538 /* if tunnel is present override skb->mark value with tunnel i_key */
539 switch (family) {
540 case AF_INET:
541 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4)
542 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4->parms.i_key);
543 break;
544 case AF_INET6:
545 if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6)
546 mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6->parms.i_key);
547 break;
548 }
549
550 sp = secpath_set(skb);
551 if (!sp) {
552 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
553 goto drop;
554 }
555
556 seq = 0;
557 if (!spi && xfrm_parse_spi(skb, nexthdr, &spi, &seq)) {
558 secpath_reset(skb);
559 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
560 goto drop;
561 }
562
563 daddr = (xfrm_address_t *)(skb_network_header(skb) +
564 XFRM_SPI_SKB_CB(skb)->daddroff);
565 do {
566 sp = skb_sec_path(skb);
567
568 if (sp->len == XFRM_MAX_DEPTH) {
569 secpath_reset(skb);
570 XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
571 goto drop;
572 }
573
574 x = xfrm_state_lookup(net, mark, daddr, spi, nexthdr, family);
575 if (x == NULL) {
576 secpath_reset(skb);
577 XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
578 xfrm_audit_state_notfound(skb, family, spi, seq);
579 goto drop;
580 }
581
582 skb->mark = xfrm_smark_get(skb->mark, x);
583
584 sp->xvec[sp->len++] = x;
585
586 skb_dst_force(skb);
587 if (!skb_dst(skb)) {
588 XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
589 goto drop;
590 }
591
592lock:
593 spin_lock(&x->lock);
594
595 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
596 if (x->km.state == XFRM_STATE_ACQ)
597 XFRM_INC_STATS(net, LINUX_MIB_XFRMACQUIREERROR);
598 else
599 XFRM_INC_STATS(net,
600 LINUX_MIB_XFRMINSTATEINVALID);
601 goto drop_unlock;
602 }
603
604 if ((x->encap ? x->encap->encap_type : 0) != encap_type) {
605 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMISMATCH);
606 goto drop_unlock;
607 }
608
609 if (xfrm_replay_check(x, skb, seq)) {
610 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
611 goto drop_unlock;
612 }
613
614 if (xfrm_state_check_expire(x)) {
615 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEEXPIRED);
616 goto drop_unlock;
617 }
618
619 spin_unlock(&x->lock);
620
621 if (xfrm_tunnel_check(skb, x, family)) {
622 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
623 goto drop;
624 }
625
626 seq_hi = htonl(xfrm_replay_seqhi(x, seq));
627
628 XFRM_SKB_CB(skb)->seq.input.low = seq;
629 XFRM_SKB_CB(skb)->seq.input.hi = seq_hi;
630
631 dev_hold(skb->dev);
632
633 if (crypto_done)
634 nexthdr = x->type_offload->input_tail(x, skb);
635 else
636 nexthdr = x->type->input(x, skb);
637
638 if (nexthdr == -EINPROGRESS)
639 return 0;
640resume:
641 dev_put(skb->dev);
642
643 spin_lock(&x->lock);
644 if (nexthdr < 0) {
645 if (nexthdr == -EBADMSG) {
646 xfrm_audit_state_icvfail(x, skb,
647 x->type->proto);
648 x->stats.integrity_failed++;
649 }
650 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEPROTOERROR);
651 goto drop_unlock;
652 }
653
654 /* only the first xfrm gets the encap type */
655 encap_type = 0;
656
657 if (xfrm_replay_recheck(x, skb, seq)) {
658 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
659 goto drop_unlock;
660 }
661
662 xfrm_replay_advance(x, seq);
663
664 x->curlft.bytes += skb->len;
665 x->curlft.packets++;
666 x->lastused = ktime_get_real_seconds();
667
668 spin_unlock(&x->lock);
669
670 XFRM_MODE_SKB_CB(skb)->protocol = nexthdr;
671
672 if (xfrm_inner_mode_input(x, skb)) {
673 XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
674 goto drop;
675 }
676
677 if (x->outer_mode.flags & XFRM_MODE_FLAG_TUNNEL) {
678 decaps = 1;
679 break;
680 }
681
682 /*
683 * We need the inner address. However, we only get here for
684 * transport mode so the outer address is identical.
685 */
686 daddr = &x->id.daddr;
687 family = x->props.family;
688
689 err = xfrm_parse_spi(skb, nexthdr, &spi, &seq);
690 if (err < 0) {
691 XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
692 goto drop;
693 }
694 crypto_done = false;
695 } while (!err);
696
697 err = xfrm_rcv_cb(skb, family, x->type->proto, 0);
698 if (err)
699 goto drop;
700
701 nf_reset_ct(skb);
702
703 if (decaps) {
704 sp = skb_sec_path(skb);
705 if (sp)
706 sp->olen = 0;
707 if (skb_valid_dst(skb))
708 skb_dst_drop(skb);
709 gro_cells_receive(&gro_cells, skb);
710 return 0;
711 } else {
712 xo = xfrm_offload(skb);
713 if (xo)
714 xfrm_gro = xo->flags & XFRM_GRO;
715
716 err = -EAFNOSUPPORT;
717 rcu_read_lock();
718 afinfo = xfrm_state_afinfo_get_rcu(x->props.family);
719 if (likely(afinfo))
720 err = afinfo->transport_finish(skb, xfrm_gro || async);
721 rcu_read_unlock();
722 if (xfrm_gro) {
723 sp = skb_sec_path(skb);
724 if (sp)
725 sp->olen = 0;
726 if (skb_valid_dst(skb))
727 skb_dst_drop(skb);
728 gro_cells_receive(&gro_cells, skb);
729 return err;
730 }
731
732 return err;
733 }
734
735drop_unlock:
736 spin_unlock(&x->lock);
737drop:
738 xfrm_rcv_cb(skb, family, x && x->type ? x->type->proto : nexthdr, -1);
739 kfree_skb(skb);
740 return 0;
741}
742EXPORT_SYMBOL(xfrm_input);
743
744int xfrm_input_resume(struct sk_buff *skb, int nexthdr)
745{
746 return xfrm_input(skb, nexthdr, 0, -1);
747}
748EXPORT_SYMBOL(xfrm_input_resume);
749
750static void xfrm_trans_reinject(struct work_struct *work)
751{
752 struct xfrm_trans_tasklet *trans = container_of(work, struct xfrm_trans_tasklet, work);
753 struct sk_buff_head queue;
754 struct sk_buff *skb;
755
756 __skb_queue_head_init(&queue);
757 spin_lock_bh(&trans->queue_lock);
758 skb_queue_splice_init(&trans->queue, &queue);
759 spin_unlock_bh(&trans->queue_lock);
760
761 local_bh_disable();
762 while ((skb = __skb_dequeue(&queue)))
763 XFRM_TRANS_SKB_CB(skb)->finish(XFRM_TRANS_SKB_CB(skb)->net,
764 NULL, skb);
765 local_bh_enable();
766}
767
768int xfrm_trans_queue_net(struct net *net, struct sk_buff *skb,
769 int (*finish)(struct net *, struct sock *,
770 struct sk_buff *))
771{
772 struct xfrm_trans_tasklet *trans;
773
774 trans = this_cpu_ptr(&xfrm_trans_tasklet);
775
776 if (skb_queue_len(&trans->queue) >= READ_ONCE(net_hotdata.max_backlog))
777 return -ENOBUFS;
778
779 BUILD_BUG_ON(sizeof(struct xfrm_trans_cb) > sizeof(skb->cb));
780
781 XFRM_TRANS_SKB_CB(skb)->finish = finish;
782 XFRM_TRANS_SKB_CB(skb)->net = net;
783 spin_lock_bh(&trans->queue_lock);
784 __skb_queue_tail(&trans->queue, skb);
785 spin_unlock_bh(&trans->queue_lock);
786 schedule_work(&trans->work);
787 return 0;
788}
789EXPORT_SYMBOL(xfrm_trans_queue_net);
790
791int xfrm_trans_queue(struct sk_buff *skb,
792 int (*finish)(struct net *, struct sock *,
793 struct sk_buff *))
794{
795 return xfrm_trans_queue_net(dev_net(skb->dev), skb, finish);
796}
797EXPORT_SYMBOL(xfrm_trans_queue);
798
799void __init xfrm_input_init(void)
800{
801 int err;
802 int i;
803
804 init_dummy_netdev(&xfrm_napi_dev);
805 err = gro_cells_init(&gro_cells, &xfrm_napi_dev);
806 if (err)
807 gro_cells.cells = NULL;
808
809 for_each_possible_cpu(i) {
810 struct xfrm_trans_tasklet *trans;
811
812 trans = &per_cpu(xfrm_trans_tasklet, i);
813 spin_lock_init(&trans->queue_lock);
814 __skb_queue_head_init(&trans->queue);
815 INIT_WORK(&trans->work, xfrm_trans_reinject);
816 }
817}