1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * xfrm_output.c - Common IPsec encapsulation code.
4 *
5 * Copyright (c) 2007 Herbert Xu <herbert@gondor.apana.org.au>
6 */
7
8 #include <linux/errno.h>
9 #include <linux/module.h>
10 #include <linux/netdevice.h>
11 #include <linux/netfilter.h>
12 #include <linux/skbuff.h>
13 #include <linux/slab.h>
14 #include <linux/spinlock.h>
15 #include <net/dst.h>
16 #include <net/icmp.h>
17 #include <net/inet_ecn.h>
18 #include <net/xfrm.h>
19
20 #if IS_ENABLED(CONFIG_IPV6)
21 #include <net/ip6_route.h>
22 #include <net/ipv6_stubs.h>
23 #endif
24
25 #include "xfrm_inout.h"
26
27 static int xfrm_output2(struct net *net, struct sock *sk, struct sk_buff *skb);
28 static int xfrm_inner_extract_output(struct xfrm_state *x, struct sk_buff *skb);
29
xfrm_skb_check_space(struct sk_buff * skb)30 static int xfrm_skb_check_space(struct sk_buff *skb)
31 {
32 struct dst_entry *dst = skb_dst(skb);
33 int nhead = dst->header_len + LL_RESERVED_SPACE(dst->dev)
34 - skb_headroom(skb);
35 int ntail = dst->dev->needed_tailroom - skb_tailroom(skb);
36
37 if (nhead <= 0) {
38 if (ntail <= 0)
39 return 0;
40 nhead = 0;
41 } else if (ntail < 0)
42 ntail = 0;
43
44 return pskb_expand_head(skb, nhead, ntail, GFP_ATOMIC);
45 }
46
47 /* Children define the path of the packet through the
48 * Linux networking. Thus, destinations are stackable.
49 */
50
skb_dst_pop(struct sk_buff * skb)51 static struct dst_entry *skb_dst_pop(struct sk_buff *skb)
52 {
53 struct dst_entry *child = dst_clone(xfrm_dst_child(skb_dst(skb)));
54
55 skb_dst_drop(skb);
56 return child;
57 }
58
59 /* Add encapsulation header.
60 *
61 * The IP header will be moved forward to make space for the encapsulation
62 * header.
63 */
xfrm4_transport_output(struct xfrm_state * x,struct sk_buff * skb)64 static int xfrm4_transport_output(struct xfrm_state *x, struct sk_buff *skb)
65 {
66 struct iphdr *iph = ip_hdr(skb);
67 int ihl = iph->ihl * 4;
68
69 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
70
71 skb_set_network_header(skb, -x->props.header_len);
72 skb->mac_header = skb->network_header +
73 offsetof(struct iphdr, protocol);
74 skb->transport_header = skb->network_header + ihl;
75 __skb_pull(skb, ihl);
76 memmove(skb_network_header(skb), iph, ihl);
77 return 0;
78 }
79
80 /* Add encapsulation header.
81 *
82 * The IP header and mutable extension headers will be moved forward to make
83 * space for the encapsulation header.
84 */
xfrm6_transport_output(struct xfrm_state * x,struct sk_buff * skb)85 static int xfrm6_transport_output(struct xfrm_state *x, struct sk_buff *skb)
86 {
87 #if IS_ENABLED(CONFIG_IPV6)
88 struct ipv6hdr *iph;
89 u8 *prevhdr;
90 int hdr_len;
91
92 iph = ipv6_hdr(skb);
93 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
94
95 hdr_len = x->type->hdr_offset(x, skb, &prevhdr);
96 if (hdr_len < 0)
97 return hdr_len;
98 skb_set_mac_header(skb,
99 (prevhdr - x->props.header_len) - skb->data);
100 skb_set_network_header(skb, -x->props.header_len);
101 skb->transport_header = skb->network_header + hdr_len;
102 __skb_pull(skb, hdr_len);
103 memmove(ipv6_hdr(skb), iph, hdr_len);
104 return 0;
105 #else
106 WARN_ON_ONCE(1);
107 return -EAFNOSUPPORT;
108 #endif
109 }
110
111 /* Add route optimization header space.
112 *
113 * The IP header and mutable extension headers will be moved forward to make
114 * space for the route optimization header.
115 */
xfrm6_ro_output(struct xfrm_state * x,struct sk_buff * skb)116 static int xfrm6_ro_output(struct xfrm_state *x, struct sk_buff *skb)
117 {
118 #if IS_ENABLED(CONFIG_IPV6)
119 struct ipv6hdr *iph;
120 u8 *prevhdr;
121 int hdr_len;
122
123 iph = ipv6_hdr(skb);
124
125 hdr_len = x->type->hdr_offset(x, skb, &prevhdr);
126 if (hdr_len < 0)
127 return hdr_len;
128 skb_set_mac_header(skb,
129 (prevhdr - x->props.header_len) - skb->data);
130 skb_set_network_header(skb, -x->props.header_len);
131 skb->transport_header = skb->network_header + hdr_len;
132 __skb_pull(skb, hdr_len);
133 memmove(ipv6_hdr(skb), iph, hdr_len);
134
135 x->lastused = ktime_get_real_seconds();
136
137 return 0;
138 #else
139 WARN_ON_ONCE(1);
140 return -EAFNOSUPPORT;
141 #endif
142 }
143
144 /* Add encapsulation header.
145 *
146 * The top IP header will be constructed per draft-nikander-esp-beet-mode-06.txt.
147 */
xfrm4_beet_encap_add(struct xfrm_state * x,struct sk_buff * skb)148 static int xfrm4_beet_encap_add(struct xfrm_state *x, struct sk_buff *skb)
149 {
150 struct ip_beet_phdr *ph;
151 struct iphdr *top_iph;
152 int hdrlen, optlen;
153
154 hdrlen = 0;
155 optlen = XFRM_MODE_SKB_CB(skb)->optlen;
156 if (unlikely(optlen))
157 hdrlen += IPV4_BEET_PHMAXLEN - (optlen & 4);
158
159 skb_set_network_header(skb, -x->props.header_len - hdrlen +
160 (XFRM_MODE_SKB_CB(skb)->ihl - sizeof(*top_iph)));
161 if (x->sel.family != AF_INET6)
162 skb->network_header += IPV4_BEET_PHMAXLEN;
163 skb->mac_header = skb->network_header +
164 offsetof(struct iphdr, protocol);
165 skb->transport_header = skb->network_header + sizeof(*top_iph);
166
167 xfrm4_beet_make_header(skb);
168
169 ph = __skb_pull(skb, XFRM_MODE_SKB_CB(skb)->ihl - hdrlen);
170
171 top_iph = ip_hdr(skb);
172
173 if (unlikely(optlen)) {
174 if (WARN_ON(optlen < 0))
175 return -EINVAL;
176
177 ph->padlen = 4 - (optlen & 4);
178 ph->hdrlen = optlen / 8;
179 ph->nexthdr = top_iph->protocol;
180 if (ph->padlen)
181 memset(ph + 1, IPOPT_NOP, ph->padlen);
182
183 top_iph->protocol = IPPROTO_BEETPH;
184 top_iph->ihl = sizeof(struct iphdr) / 4;
185 }
186
187 top_iph->saddr = x->props.saddr.a4;
188 top_iph->daddr = x->id.daddr.a4;
189
190 return 0;
191 }
192
193 /* Add encapsulation header.
194 *
195 * The top IP header will be constructed per RFC 2401.
196 */
xfrm4_tunnel_encap_add(struct xfrm_state * x,struct sk_buff * skb)197 static int xfrm4_tunnel_encap_add(struct xfrm_state *x, struct sk_buff *skb)
198 {
199 struct dst_entry *dst = skb_dst(skb);
200 struct iphdr *top_iph;
201 int flags;
202
203 skb_set_inner_network_header(skb, skb_network_offset(skb));
204 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
205
206 skb_set_network_header(skb, -x->props.header_len);
207 skb->mac_header = skb->network_header +
208 offsetof(struct iphdr, protocol);
209 skb->transport_header = skb->network_header + sizeof(*top_iph);
210 top_iph = ip_hdr(skb);
211
212 top_iph->ihl = 5;
213 top_iph->version = 4;
214
215 top_iph->protocol = xfrm_af2proto(skb_dst(skb)->ops->family);
216
217 /* DS disclosing depends on XFRM_SA_XFLAG_DONT_ENCAP_DSCP */
218 if (x->props.extra_flags & XFRM_SA_XFLAG_DONT_ENCAP_DSCP)
219 top_iph->tos = 0;
220 else
221 top_iph->tos = XFRM_MODE_SKB_CB(skb)->tos;
222 top_iph->tos = INET_ECN_encapsulate(top_iph->tos,
223 XFRM_MODE_SKB_CB(skb)->tos);
224
225 flags = x->props.flags;
226 if (flags & XFRM_STATE_NOECN)
227 IP_ECN_clear(top_iph);
228
229 top_iph->frag_off = (flags & XFRM_STATE_NOPMTUDISC) ?
230 0 : (XFRM_MODE_SKB_CB(skb)->frag_off & htons(IP_DF));
231
232 top_iph->ttl = ip4_dst_hoplimit(xfrm_dst_child(dst));
233
234 top_iph->saddr = x->props.saddr.a4;
235 top_iph->daddr = x->id.daddr.a4;
236 ip_select_ident(dev_net(dst->dev), skb, NULL);
237
238 return 0;
239 }
240
241 #if IS_ENABLED(CONFIG_IPV6)
xfrm6_tunnel_encap_add(struct xfrm_state * x,struct sk_buff * skb)242 static int xfrm6_tunnel_encap_add(struct xfrm_state *x, struct sk_buff *skb)
243 {
244 struct dst_entry *dst = skb_dst(skb);
245 struct ipv6hdr *top_iph;
246 int dsfield;
247
248 skb_set_inner_network_header(skb, skb_network_offset(skb));
249 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
250
251 skb_set_network_header(skb, -x->props.header_len);
252 skb->mac_header = skb->network_header +
253 offsetof(struct ipv6hdr, nexthdr);
254 skb->transport_header = skb->network_header + sizeof(*top_iph);
255 top_iph = ipv6_hdr(skb);
256
257 top_iph->version = 6;
258
259 memcpy(top_iph->flow_lbl, XFRM_MODE_SKB_CB(skb)->flow_lbl,
260 sizeof(top_iph->flow_lbl));
261 top_iph->nexthdr = xfrm_af2proto(skb_dst(skb)->ops->family);
262
263 if (x->props.extra_flags & XFRM_SA_XFLAG_DONT_ENCAP_DSCP)
264 dsfield = 0;
265 else
266 dsfield = XFRM_MODE_SKB_CB(skb)->tos;
267 dsfield = INET_ECN_encapsulate(dsfield, XFRM_MODE_SKB_CB(skb)->tos);
268 if (x->props.flags & XFRM_STATE_NOECN)
269 dsfield &= ~INET_ECN_MASK;
270 ipv6_change_dsfield(top_iph, 0, dsfield);
271 top_iph->hop_limit = ip6_dst_hoplimit(xfrm_dst_child(dst));
272 top_iph->saddr = *(struct in6_addr *)&x->props.saddr;
273 top_iph->daddr = *(struct in6_addr *)&x->id.daddr;
274 return 0;
275 }
276
xfrm6_beet_encap_add(struct xfrm_state * x,struct sk_buff * skb)277 static int xfrm6_beet_encap_add(struct xfrm_state *x, struct sk_buff *skb)
278 {
279 struct ipv6hdr *top_iph;
280 struct ip_beet_phdr *ph;
281 int optlen, hdr_len;
282
283 hdr_len = 0;
284 optlen = XFRM_MODE_SKB_CB(skb)->optlen;
285 if (unlikely(optlen))
286 hdr_len += IPV4_BEET_PHMAXLEN - (optlen & 4);
287
288 skb_set_network_header(skb, -x->props.header_len - hdr_len);
289 if (x->sel.family != AF_INET6)
290 skb->network_header += IPV4_BEET_PHMAXLEN;
291 skb->mac_header = skb->network_header +
292 offsetof(struct ipv6hdr, nexthdr);
293 skb->transport_header = skb->network_header + sizeof(*top_iph);
294 ph = __skb_pull(skb, XFRM_MODE_SKB_CB(skb)->ihl - hdr_len);
295
296 xfrm6_beet_make_header(skb);
297
298 top_iph = ipv6_hdr(skb);
299 if (unlikely(optlen)) {
300 if (WARN_ON(optlen < 0))
301 return -EINVAL;
302
303 ph->padlen = 4 - (optlen & 4);
304 ph->hdrlen = optlen / 8;
305 ph->nexthdr = top_iph->nexthdr;
306 if (ph->padlen)
307 memset(ph + 1, IPOPT_NOP, ph->padlen);
308
309 top_iph->nexthdr = IPPROTO_BEETPH;
310 }
311
312 top_iph->saddr = *(struct in6_addr *)&x->props.saddr;
313 top_iph->daddr = *(struct in6_addr *)&x->id.daddr;
314 return 0;
315 }
316 #endif
317
318 /* Add encapsulation header.
319 *
320 * On exit, the transport header will be set to the start of the
321 * encapsulation header to be filled in by x->type->output and the mac
322 * header will be set to the nextheader (protocol for IPv4) field of the
323 * extension header directly preceding the encapsulation header, or in
324 * its absence, that of the top IP header.
325 * The value of the network header will always point to the top IP header
326 * while skb->data will point to the payload.
327 */
xfrm4_prepare_output(struct xfrm_state * x,struct sk_buff * skb)328 static int xfrm4_prepare_output(struct xfrm_state *x, struct sk_buff *skb)
329 {
330 int err;
331
332 err = xfrm_inner_extract_output(x, skb);
333 if (err)
334 return err;
335
336 IPCB(skb)->flags |= IPSKB_XFRM_TUNNEL_SIZE;
337 skb->protocol = htons(ETH_P_IP);
338
339 switch (x->outer_mode.encap) {
340 case XFRM_MODE_BEET:
341 return xfrm4_beet_encap_add(x, skb);
342 case XFRM_MODE_TUNNEL:
343 return xfrm4_tunnel_encap_add(x, skb);
344 }
345
346 WARN_ON_ONCE(1);
347 return -EOPNOTSUPP;
348 }
349
xfrm6_prepare_output(struct xfrm_state * x,struct sk_buff * skb)350 static int xfrm6_prepare_output(struct xfrm_state *x, struct sk_buff *skb)
351 {
352 #if IS_ENABLED(CONFIG_IPV6)
353 int err;
354
355 err = xfrm_inner_extract_output(x, skb);
356 if (err)
357 return err;
358
359 skb->ignore_df = 1;
360 skb->protocol = htons(ETH_P_IPV6);
361
362 switch (x->outer_mode.encap) {
363 case XFRM_MODE_BEET:
364 return xfrm6_beet_encap_add(x, skb);
365 case XFRM_MODE_TUNNEL:
366 return xfrm6_tunnel_encap_add(x, skb);
367 default:
368 WARN_ON_ONCE(1);
369 return -EOPNOTSUPP;
370 }
371 #endif
372 WARN_ON_ONCE(1);
373 return -EAFNOSUPPORT;
374 }
375
xfrm_outer_mode_output(struct xfrm_state * x,struct sk_buff * skb)376 static int xfrm_outer_mode_output(struct xfrm_state *x, struct sk_buff *skb)
377 {
378 switch (x->outer_mode.encap) {
379 case XFRM_MODE_BEET:
380 case XFRM_MODE_TUNNEL:
381 if (x->outer_mode.family == AF_INET)
382 return xfrm4_prepare_output(x, skb);
383 if (x->outer_mode.family == AF_INET6)
384 return xfrm6_prepare_output(x, skb);
385 break;
386 case XFRM_MODE_TRANSPORT:
387 if (x->outer_mode.family == AF_INET)
388 return xfrm4_transport_output(x, skb);
389 if (x->outer_mode.family == AF_INET6)
390 return xfrm6_transport_output(x, skb);
391 break;
392 case XFRM_MODE_ROUTEOPTIMIZATION:
393 if (x->outer_mode.family == AF_INET6)
394 return xfrm6_ro_output(x, skb);
395 WARN_ON_ONCE(1);
396 break;
397 default:
398 WARN_ON_ONCE(1);
399 break;
400 }
401
402 return -EOPNOTSUPP;
403 }
404
pktgen_xfrm_outer_mode_output(struct xfrm_state * x,struct sk_buff * skb)405 int pktgen_xfrm_outer_mode_output(struct xfrm_state *x, struct sk_buff *skb)
406 {
407 return xfrm_outer_mode_output(x, skb);
408 }
409 EXPORT_SYMBOL_GPL(pktgen_xfrm_outer_mode_output);
410
xfrm_output_one(struct sk_buff * skb,int err)411 static int xfrm_output_one(struct sk_buff *skb, int err)
412 {
413 struct dst_entry *dst = skb_dst(skb);
414 struct xfrm_state *x = dst->xfrm;
415 struct net *net = xs_net(x);
416
417 if (err <= 0)
418 goto resume;
419
420 do {
421 err = xfrm_skb_check_space(skb);
422 if (err) {
423 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
424 goto error_nolock;
425 }
426
427 skb->mark = xfrm_smark_get(skb->mark, x);
428
429 err = xfrm_outer_mode_output(x, skb);
430 if (err) {
431 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEMODEERROR);
432 goto error_nolock;
433 }
434
435 spin_lock_bh(&x->lock);
436
437 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
438 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEINVALID);
439 err = -EINVAL;
440 goto error;
441 }
442
443 err = xfrm_state_check_expire(x);
444 if (err) {
445 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEEXPIRED);
446 goto error;
447 }
448
449 err = x->repl->overflow(x, skb);
450 if (err) {
451 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATESEQERROR);
452 goto error;
453 }
454
455 x->curlft.bytes += skb->len;
456 x->curlft.packets++;
457
458 spin_unlock_bh(&x->lock);
459
460 skb_dst_force(skb);
461 if (!skb_dst(skb)) {
462 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
463 err = -EHOSTUNREACH;
464 goto error_nolock;
465 }
466
467 if (xfrm_offload(skb)) {
468 x->type_offload->encap(x, skb);
469 } else {
470 /* Inner headers are invalid now. */
471 skb->encapsulation = 0;
472
473 err = x->type->output(x, skb);
474 if (err == -EINPROGRESS)
475 goto out;
476 }
477
478 resume:
479 if (err) {
480 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEPROTOERROR);
481 goto error_nolock;
482 }
483
484 dst = skb_dst_pop(skb);
485 if (!dst) {
486 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
487 err = -EHOSTUNREACH;
488 goto error_nolock;
489 }
490 skb_dst_set(skb, dst);
491 x = dst->xfrm;
492 } while (x && !(x->outer_mode.flags & XFRM_MODE_FLAG_TUNNEL));
493
494 return 0;
495
496 error:
497 spin_unlock_bh(&x->lock);
498 error_nolock:
499 kfree_skb(skb);
500 out:
501 return err;
502 }
503
xfrm_output_resume(struct sk_buff * skb,int err)504 int xfrm_output_resume(struct sk_buff *skb, int err)
505 {
506 struct net *net = xs_net(skb_dst(skb)->xfrm);
507
508 while (likely((err = xfrm_output_one(skb, err)) == 0)) {
509 nf_reset_ct(skb);
510
511 err = skb_dst(skb)->ops->local_out(net, skb->sk, skb);
512 if (unlikely(err != 1))
513 goto out;
514
515 if (!skb_dst(skb)->xfrm)
516 return dst_output(net, skb->sk, skb);
517
518 err = nf_hook(skb_dst(skb)->ops->family,
519 NF_INET_POST_ROUTING, net, skb->sk, skb,
520 NULL, skb_dst(skb)->dev, xfrm_output2);
521 if (unlikely(err != 1))
522 goto out;
523 }
524
525 if (err == -EINPROGRESS)
526 err = 0;
527
528 out:
529 return err;
530 }
531 EXPORT_SYMBOL_GPL(xfrm_output_resume);
532
xfrm_output2(struct net * net,struct sock * sk,struct sk_buff * skb)533 static int xfrm_output2(struct net *net, struct sock *sk, struct sk_buff *skb)
534 {
535 return xfrm_output_resume(skb, 1);
536 }
537
xfrm_output_gso(struct net * net,struct sock * sk,struct sk_buff * skb)538 static int xfrm_output_gso(struct net *net, struct sock *sk, struct sk_buff *skb)
539 {
540 struct sk_buff *segs, *nskb;
541
542 BUILD_BUG_ON(sizeof(*IPCB(skb)) > SKB_GSO_CB_OFFSET);
543 BUILD_BUG_ON(sizeof(*IP6CB(skb)) > SKB_GSO_CB_OFFSET);
544 segs = skb_gso_segment(skb, 0);
545 kfree_skb(skb);
546 if (IS_ERR(segs))
547 return PTR_ERR(segs);
548 if (segs == NULL)
549 return -EINVAL;
550
551 skb_list_walk_safe(segs, segs, nskb) {
552 int err;
553
554 skb_mark_not_on_list(segs);
555 err = xfrm_output2(net, sk, segs);
556
557 if (unlikely(err)) {
558 kfree_skb_list(nskb);
559 return err;
560 }
561 }
562
563 return 0;
564 }
565
xfrm_output(struct sock * sk,struct sk_buff * skb)566 int xfrm_output(struct sock *sk, struct sk_buff *skb)
567 {
568 struct net *net = dev_net(skb_dst(skb)->dev);
569 struct xfrm_state *x = skb_dst(skb)->xfrm;
570 int err;
571
572 switch (x->outer_mode.family) {
573 case AF_INET:
574 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
575 IPCB(skb)->flags |= IPSKB_XFRM_TRANSFORMED;
576 break;
577 case AF_INET6:
578 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
579
580 IP6CB(skb)->flags |= IP6SKB_XFRM_TRANSFORMED;
581 break;
582 }
583
584 secpath_reset(skb);
585
586 if (xfrm_dev_offload_ok(skb, x)) {
587 struct sec_path *sp;
588
589 sp = secpath_set(skb);
590 if (!sp) {
591 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
592 kfree_skb(skb);
593 return -ENOMEM;
594 }
595 skb->encapsulation = 1;
596
597 sp->olen++;
598 sp->xvec[sp->len++] = x;
599 xfrm_state_hold(x);
600
601 if (skb_is_gso(skb)) {
602 if (skb->inner_protocol)
603 return xfrm_output_gso(net, sk, skb);
604
605 skb_shinfo(skb)->gso_type |= SKB_GSO_ESP;
606 goto out;
607 }
608
609 if (x->xso.dev && x->xso.dev->features & NETIF_F_HW_ESP_TX_CSUM)
610 goto out;
611 } else {
612 if (skb_is_gso(skb))
613 return xfrm_output_gso(net, sk, skb);
614 }
615
616 if (skb->ip_summed == CHECKSUM_PARTIAL) {
617 err = skb_checksum_help(skb);
618 if (err) {
619 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
620 kfree_skb(skb);
621 return err;
622 }
623 }
624
625 out:
626 return xfrm_output2(net, sk, skb);
627 }
628 EXPORT_SYMBOL_GPL(xfrm_output);
629
xfrm4_tunnel_check_size(struct sk_buff * skb)630 static int xfrm4_tunnel_check_size(struct sk_buff *skb)
631 {
632 int mtu, ret = 0;
633
634 if (IPCB(skb)->flags & IPSKB_XFRM_TUNNEL_SIZE)
635 goto out;
636
637 if (!(ip_hdr(skb)->frag_off & htons(IP_DF)) || skb->ignore_df)
638 goto out;
639
640 mtu = dst_mtu(skb_dst(skb));
641 if ((!skb_is_gso(skb) && skb->len > mtu) ||
642 (skb_is_gso(skb) &&
643 !skb_gso_validate_network_len(skb, ip_skb_dst_mtu(skb->sk, skb)))) {
644 skb->protocol = htons(ETH_P_IP);
645
646 if (skb->sk)
647 xfrm_local_error(skb, mtu);
648 else
649 icmp_send(skb, ICMP_DEST_UNREACH,
650 ICMP_FRAG_NEEDED, htonl(mtu));
651 ret = -EMSGSIZE;
652 }
653 out:
654 return ret;
655 }
656
xfrm4_extract_output(struct xfrm_state * x,struct sk_buff * skb)657 static int xfrm4_extract_output(struct xfrm_state *x, struct sk_buff *skb)
658 {
659 int err;
660
661 if (x->outer_mode.encap == XFRM_MODE_BEET &&
662 ip_is_fragment(ip_hdr(skb))) {
663 net_warn_ratelimited("BEET mode doesn't support inner IPv4 fragments\n");
664 return -EAFNOSUPPORT;
665 }
666
667 err = xfrm4_tunnel_check_size(skb);
668 if (err)
669 return err;
670
671 XFRM_MODE_SKB_CB(skb)->protocol = ip_hdr(skb)->protocol;
672
673 xfrm4_extract_header(skb);
674 return 0;
675 }
676
677 #if IS_ENABLED(CONFIG_IPV6)
xfrm6_tunnel_check_size(struct sk_buff * skb)678 static int xfrm6_tunnel_check_size(struct sk_buff *skb)
679 {
680 int mtu, ret = 0;
681 struct dst_entry *dst = skb_dst(skb);
682
683 if (skb->ignore_df)
684 goto out;
685
686 mtu = dst_mtu(dst);
687 if (mtu < IPV6_MIN_MTU)
688 mtu = IPV6_MIN_MTU;
689
690 if ((!skb_is_gso(skb) && skb->len > mtu) ||
691 (skb_is_gso(skb) &&
692 !skb_gso_validate_network_len(skb, ip6_skb_dst_mtu(skb)))) {
693 skb->dev = dst->dev;
694 skb->protocol = htons(ETH_P_IPV6);
695
696 if (xfrm6_local_dontfrag(skb->sk))
697 ipv6_stub->xfrm6_local_rxpmtu(skb, mtu);
698 else if (skb->sk)
699 xfrm_local_error(skb, mtu);
700 else
701 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
702 ret = -EMSGSIZE;
703 }
704 out:
705 return ret;
706 }
707 #endif
708
xfrm6_extract_output(struct xfrm_state * x,struct sk_buff * skb)709 static int xfrm6_extract_output(struct xfrm_state *x, struct sk_buff *skb)
710 {
711 #if IS_ENABLED(CONFIG_IPV6)
712 int err;
713
714 err = xfrm6_tunnel_check_size(skb);
715 if (err)
716 return err;
717
718 XFRM_MODE_SKB_CB(skb)->protocol = ipv6_hdr(skb)->nexthdr;
719
720 xfrm6_extract_header(skb);
721 return 0;
722 #else
723 WARN_ON_ONCE(1);
724 return -EAFNOSUPPORT;
725 #endif
726 }
727
xfrm_inner_extract_output(struct xfrm_state * x,struct sk_buff * skb)728 static int xfrm_inner_extract_output(struct xfrm_state *x, struct sk_buff *skb)
729 {
730 const struct xfrm_mode *inner_mode;
731
732 if (x->sel.family == AF_UNSPEC)
733 inner_mode = xfrm_ip2inner_mode(x,
734 xfrm_af2proto(skb_dst(skb)->ops->family));
735 else
736 inner_mode = &x->inner_mode;
737
738 if (inner_mode == NULL)
739 return -EAFNOSUPPORT;
740
741 switch (inner_mode->family) {
742 case AF_INET:
743 return xfrm4_extract_output(x, skb);
744 case AF_INET6:
745 return xfrm6_extract_output(x, skb);
746 }
747
748 return -EAFNOSUPPORT;
749 }
750
xfrm_local_error(struct sk_buff * skb,int mtu)751 void xfrm_local_error(struct sk_buff *skb, int mtu)
752 {
753 unsigned int proto;
754 struct xfrm_state_afinfo *afinfo;
755
756 if (skb->protocol == htons(ETH_P_IP))
757 proto = AF_INET;
758 else if (skb->protocol == htons(ETH_P_IPV6) &&
759 skb->sk->sk_family == AF_INET6)
760 proto = AF_INET6;
761 else
762 return;
763
764 afinfo = xfrm_state_get_afinfo(proto);
765 if (afinfo) {
766 afinfo->local_error(skb, mtu);
767 rcu_read_unlock();
768 }
769 }
770 EXPORT_SYMBOL_GPL(xfrm_local_error);
771