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 #if IS_ENABLED(CONFIG_IPV6_MIP6)
mip6_rthdr_offset(struct sk_buff * skb,u8 ** nexthdr,int type)81 static int mip6_rthdr_offset(struct sk_buff *skb, u8 **nexthdr, int type)
82 {
83 const unsigned char *nh = skb_network_header(skb);
84 unsigned int offset = sizeof(struct ipv6hdr);
85 unsigned int packet_len;
86 int found_rhdr = 0;
87
88 packet_len = skb_tail_pointer(skb) - nh;
89 *nexthdr = &ipv6_hdr(skb)->nexthdr;
90
91 while (offset <= packet_len) {
92 struct ipv6_opt_hdr *exthdr;
93
94 switch (**nexthdr) {
95 case NEXTHDR_HOP:
96 break;
97 case NEXTHDR_ROUTING:
98 if (type == IPPROTO_ROUTING && offset + 3 <= packet_len) {
99 struct ipv6_rt_hdr *rt;
100
101 rt = (struct ipv6_rt_hdr *)(nh + offset);
102 if (rt->type != 0)
103 return offset;
104 }
105 found_rhdr = 1;
106 break;
107 case NEXTHDR_DEST:
108 /* HAO MUST NOT appear more than once.
109 * XXX: It is better to try to find by the end of
110 * XXX: packet if HAO exists.
111 */
112 if (ipv6_find_tlv(skb, offset, IPV6_TLV_HAO) >= 0) {
113 net_dbg_ratelimited("mip6: hao exists already, override\n");
114 return offset;
115 }
116
117 if (found_rhdr)
118 return offset;
119
120 break;
121 default:
122 return offset;
123 }
124
125 if (offset + sizeof(struct ipv6_opt_hdr) > packet_len)
126 return -EINVAL;
127
128 exthdr = (struct ipv6_opt_hdr *)(skb_network_header(skb) +
129 offset);
130 offset += ipv6_optlen(exthdr);
131 if (offset > IPV6_MAXPLEN)
132 return -EINVAL;
133 *nexthdr = &exthdr->nexthdr;
134 }
135
136 return -EINVAL;
137 }
138 #endif
139
140 #if IS_ENABLED(CONFIG_IPV6)
xfrm6_hdr_offset(struct xfrm_state * x,struct sk_buff * skb,u8 ** prevhdr)141 static int xfrm6_hdr_offset(struct xfrm_state *x, struct sk_buff *skb, u8 **prevhdr)
142 {
143 switch (x->type->proto) {
144 #if IS_ENABLED(CONFIG_IPV6_MIP6)
145 case IPPROTO_DSTOPTS:
146 case IPPROTO_ROUTING:
147 return mip6_rthdr_offset(skb, prevhdr, x->type->proto);
148 #endif
149 default:
150 break;
151 }
152
153 return ip6_find_1stfragopt(skb, prevhdr);
154 }
155 #endif
156
157 /* Add encapsulation header.
158 *
159 * The IP header and mutable extension headers will be moved forward to make
160 * space for the encapsulation header.
161 */
xfrm6_transport_output(struct xfrm_state * x,struct sk_buff * skb)162 static int xfrm6_transport_output(struct xfrm_state *x, struct sk_buff *skb)
163 {
164 #if IS_ENABLED(CONFIG_IPV6)
165 struct ipv6hdr *iph;
166 u8 *prevhdr;
167 int hdr_len;
168
169 iph = ipv6_hdr(skb);
170 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
171
172 hdr_len = xfrm6_hdr_offset(x, skb, &prevhdr);
173 if (hdr_len < 0)
174 return hdr_len;
175 skb_set_mac_header(skb,
176 (prevhdr - x->props.header_len) - skb->data);
177 skb_set_network_header(skb, -x->props.header_len);
178 skb->transport_header = skb->network_header + hdr_len;
179 __skb_pull(skb, hdr_len);
180 memmove(ipv6_hdr(skb), iph, hdr_len);
181 return 0;
182 #else
183 WARN_ON_ONCE(1);
184 return -EAFNOSUPPORT;
185 #endif
186 }
187
188 /* Add route optimization header space.
189 *
190 * The IP header and mutable extension headers will be moved forward to make
191 * space for the route optimization header.
192 */
xfrm6_ro_output(struct xfrm_state * x,struct sk_buff * skb)193 static int xfrm6_ro_output(struct xfrm_state *x, struct sk_buff *skb)
194 {
195 #if IS_ENABLED(CONFIG_IPV6)
196 struct ipv6hdr *iph;
197 u8 *prevhdr;
198 int hdr_len;
199
200 iph = ipv6_hdr(skb);
201
202 hdr_len = xfrm6_hdr_offset(x, skb, &prevhdr);
203 if (hdr_len < 0)
204 return hdr_len;
205 skb_set_mac_header(skb,
206 (prevhdr - x->props.header_len) - skb->data);
207 skb_set_network_header(skb, -x->props.header_len);
208 skb->transport_header = skb->network_header + hdr_len;
209 __skb_pull(skb, hdr_len);
210 memmove(ipv6_hdr(skb), iph, hdr_len);
211
212 x->lastused = ktime_get_real_seconds();
213
214 return 0;
215 #else
216 WARN_ON_ONCE(1);
217 return -EAFNOSUPPORT;
218 #endif
219 }
220
221 /* Add encapsulation header.
222 *
223 * The top IP header will be constructed per draft-nikander-esp-beet-mode-06.txt.
224 */
xfrm4_beet_encap_add(struct xfrm_state * x,struct sk_buff * skb)225 static int xfrm4_beet_encap_add(struct xfrm_state *x, struct sk_buff *skb)
226 {
227 struct ip_beet_phdr *ph;
228 struct iphdr *top_iph;
229 int hdrlen, optlen;
230
231 hdrlen = 0;
232 optlen = XFRM_MODE_SKB_CB(skb)->optlen;
233 if (unlikely(optlen))
234 hdrlen += IPV4_BEET_PHMAXLEN - (optlen & 4);
235
236 skb_set_network_header(skb, -x->props.header_len - hdrlen +
237 (XFRM_MODE_SKB_CB(skb)->ihl - sizeof(*top_iph)));
238 if (x->sel.family != AF_INET6)
239 skb->network_header += IPV4_BEET_PHMAXLEN;
240 skb->mac_header = skb->network_header +
241 offsetof(struct iphdr, protocol);
242 skb->transport_header = skb->network_header + sizeof(*top_iph);
243
244 xfrm4_beet_make_header(skb);
245
246 ph = __skb_pull(skb, XFRM_MODE_SKB_CB(skb)->ihl - hdrlen);
247
248 top_iph = ip_hdr(skb);
249
250 if (unlikely(optlen)) {
251 if (WARN_ON(optlen < 0))
252 return -EINVAL;
253
254 ph->padlen = 4 - (optlen & 4);
255 ph->hdrlen = optlen / 8;
256 ph->nexthdr = top_iph->protocol;
257 if (ph->padlen)
258 memset(ph + 1, IPOPT_NOP, ph->padlen);
259
260 top_iph->protocol = IPPROTO_BEETPH;
261 top_iph->ihl = sizeof(struct iphdr) / 4;
262 }
263
264 top_iph->saddr = x->props.saddr.a4;
265 top_iph->daddr = x->id.daddr.a4;
266
267 return 0;
268 }
269
270 /* Add encapsulation header.
271 *
272 * The top IP header will be constructed per RFC 2401.
273 */
xfrm4_tunnel_encap_add(struct xfrm_state * x,struct sk_buff * skb)274 static int xfrm4_tunnel_encap_add(struct xfrm_state *x, struct sk_buff *skb)
275 {
276 bool small_ipv6 = (skb->protocol == htons(ETH_P_IPV6)) && (skb->len <= IPV6_MIN_MTU);
277 struct dst_entry *dst = skb_dst(skb);
278 struct iphdr *top_iph;
279 int flags;
280
281 skb_set_inner_network_header(skb, skb_network_offset(skb));
282 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
283
284 skb_set_network_header(skb, -x->props.header_len);
285 skb->mac_header = skb->network_header +
286 offsetof(struct iphdr, protocol);
287 skb->transport_header = skb->network_header + sizeof(*top_iph);
288 top_iph = ip_hdr(skb);
289
290 top_iph->ihl = 5;
291 top_iph->version = 4;
292
293 top_iph->protocol = xfrm_af2proto(skb_dst(skb)->ops->family);
294
295 /* DS disclosing depends on XFRM_SA_XFLAG_DONT_ENCAP_DSCP */
296 if (x->props.extra_flags & XFRM_SA_XFLAG_DONT_ENCAP_DSCP)
297 top_iph->tos = 0;
298 else
299 top_iph->tos = XFRM_MODE_SKB_CB(skb)->tos;
300 top_iph->tos = INET_ECN_encapsulate(top_iph->tos,
301 XFRM_MODE_SKB_CB(skb)->tos);
302
303 flags = x->props.flags;
304 if (flags & XFRM_STATE_NOECN)
305 IP_ECN_clear(top_iph);
306
307 top_iph->frag_off = (flags & XFRM_STATE_NOPMTUDISC) || small_ipv6 ?
308 0 : (XFRM_MODE_SKB_CB(skb)->frag_off & htons(IP_DF));
309
310 top_iph->ttl = ip4_dst_hoplimit(xfrm_dst_child(dst));
311
312 top_iph->saddr = x->props.saddr.a4;
313 top_iph->daddr = x->id.daddr.a4;
314 ip_select_ident(dev_net(dst->dev), skb, NULL);
315
316 return 0;
317 }
318
319 #if IS_ENABLED(CONFIG_IPV6)
xfrm6_tunnel_encap_add(struct xfrm_state * x,struct sk_buff * skb)320 static int xfrm6_tunnel_encap_add(struct xfrm_state *x, struct sk_buff *skb)
321 {
322 struct dst_entry *dst = skb_dst(skb);
323 struct ipv6hdr *top_iph;
324 int dsfield;
325
326 skb_set_inner_network_header(skb, skb_network_offset(skb));
327 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
328
329 skb_set_network_header(skb, -x->props.header_len);
330 skb->mac_header = skb->network_header +
331 offsetof(struct ipv6hdr, nexthdr);
332 skb->transport_header = skb->network_header + sizeof(*top_iph);
333 top_iph = ipv6_hdr(skb);
334
335 top_iph->version = 6;
336
337 memcpy(top_iph->flow_lbl, XFRM_MODE_SKB_CB(skb)->flow_lbl,
338 sizeof(top_iph->flow_lbl));
339 top_iph->nexthdr = xfrm_af2proto(skb_dst(skb)->ops->family);
340
341 if (x->props.extra_flags & XFRM_SA_XFLAG_DONT_ENCAP_DSCP)
342 dsfield = 0;
343 else
344 dsfield = XFRM_MODE_SKB_CB(skb)->tos;
345 dsfield = INET_ECN_encapsulate(dsfield, XFRM_MODE_SKB_CB(skb)->tos);
346 if (x->props.flags & XFRM_STATE_NOECN)
347 dsfield &= ~INET_ECN_MASK;
348 ipv6_change_dsfield(top_iph, 0, dsfield);
349 top_iph->hop_limit = ip6_dst_hoplimit(xfrm_dst_child(dst));
350 top_iph->saddr = *(struct in6_addr *)&x->props.saddr;
351 top_iph->daddr = *(struct in6_addr *)&x->id.daddr;
352 return 0;
353 }
354
xfrm6_beet_encap_add(struct xfrm_state * x,struct sk_buff * skb)355 static int xfrm6_beet_encap_add(struct xfrm_state *x, struct sk_buff *skb)
356 {
357 struct ipv6hdr *top_iph;
358 struct ip_beet_phdr *ph;
359 int optlen, hdr_len;
360
361 hdr_len = 0;
362 optlen = XFRM_MODE_SKB_CB(skb)->optlen;
363 if (unlikely(optlen))
364 hdr_len += IPV4_BEET_PHMAXLEN - (optlen & 4);
365
366 skb_set_network_header(skb, -x->props.header_len - hdr_len);
367 if (x->sel.family != AF_INET6)
368 skb->network_header += IPV4_BEET_PHMAXLEN;
369 skb->mac_header = skb->network_header +
370 offsetof(struct ipv6hdr, nexthdr);
371 skb->transport_header = skb->network_header + sizeof(*top_iph);
372 ph = __skb_pull(skb, XFRM_MODE_SKB_CB(skb)->ihl - hdr_len);
373
374 xfrm6_beet_make_header(skb);
375
376 top_iph = ipv6_hdr(skb);
377 if (unlikely(optlen)) {
378 if (WARN_ON(optlen < 0))
379 return -EINVAL;
380
381 ph->padlen = 4 - (optlen & 4);
382 ph->hdrlen = optlen / 8;
383 ph->nexthdr = top_iph->nexthdr;
384 if (ph->padlen)
385 memset(ph + 1, IPOPT_NOP, ph->padlen);
386
387 top_iph->nexthdr = IPPROTO_BEETPH;
388 }
389
390 top_iph->saddr = *(struct in6_addr *)&x->props.saddr;
391 top_iph->daddr = *(struct in6_addr *)&x->id.daddr;
392 return 0;
393 }
394 #endif
395
396 /* Add encapsulation header.
397 *
398 * On exit, the transport header will be set to the start of the
399 * encapsulation header to be filled in by x->type->output and the mac
400 * header will be set to the nextheader (protocol for IPv4) field of the
401 * extension header directly preceding the encapsulation header, or in
402 * its absence, that of the top IP header.
403 * The value of the network header will always point to the top IP header
404 * while skb->data will point to the payload.
405 */
xfrm4_prepare_output(struct xfrm_state * x,struct sk_buff * skb)406 static int xfrm4_prepare_output(struct xfrm_state *x, struct sk_buff *skb)
407 {
408 int err;
409
410 err = xfrm_inner_extract_output(x, skb);
411 if (err)
412 return err;
413
414 IPCB(skb)->flags |= IPSKB_XFRM_TUNNEL_SIZE;
415 skb->protocol = htons(ETH_P_IP);
416
417 switch (x->outer_mode.encap) {
418 case XFRM_MODE_BEET:
419 return xfrm4_beet_encap_add(x, skb);
420 case XFRM_MODE_TUNNEL:
421 return xfrm4_tunnel_encap_add(x, skb);
422 }
423
424 WARN_ON_ONCE(1);
425 return -EOPNOTSUPP;
426 }
427
xfrm6_prepare_output(struct xfrm_state * x,struct sk_buff * skb)428 static int xfrm6_prepare_output(struct xfrm_state *x, struct sk_buff *skb)
429 {
430 #if IS_ENABLED(CONFIG_IPV6)
431 int err;
432
433 err = xfrm_inner_extract_output(x, skb);
434 if (err)
435 return err;
436
437 skb->ignore_df = 1;
438 skb->protocol = htons(ETH_P_IPV6);
439
440 switch (x->outer_mode.encap) {
441 case XFRM_MODE_BEET:
442 return xfrm6_beet_encap_add(x, skb);
443 case XFRM_MODE_TUNNEL:
444 return xfrm6_tunnel_encap_add(x, skb);
445 default:
446 WARN_ON_ONCE(1);
447 return -EOPNOTSUPP;
448 }
449 #endif
450 WARN_ON_ONCE(1);
451 return -EAFNOSUPPORT;
452 }
453
xfrm_outer_mode_output(struct xfrm_state * x,struct sk_buff * skb)454 static int xfrm_outer_mode_output(struct xfrm_state *x, struct sk_buff *skb)
455 {
456 switch (x->outer_mode.encap) {
457 case XFRM_MODE_BEET:
458 case XFRM_MODE_TUNNEL:
459 if (x->outer_mode.family == AF_INET)
460 return xfrm4_prepare_output(x, skb);
461 if (x->outer_mode.family == AF_INET6)
462 return xfrm6_prepare_output(x, skb);
463 break;
464 case XFRM_MODE_TRANSPORT:
465 if (x->outer_mode.family == AF_INET)
466 return xfrm4_transport_output(x, skb);
467 if (x->outer_mode.family == AF_INET6)
468 return xfrm6_transport_output(x, skb);
469 break;
470 case XFRM_MODE_ROUTEOPTIMIZATION:
471 if (x->outer_mode.family == AF_INET6)
472 return xfrm6_ro_output(x, skb);
473 WARN_ON_ONCE(1);
474 break;
475 default:
476 WARN_ON_ONCE(1);
477 break;
478 }
479
480 return -EOPNOTSUPP;
481 }
482
pktgen_xfrm_outer_mode_output(struct xfrm_state * x,struct sk_buff * skb)483 int pktgen_xfrm_outer_mode_output(struct xfrm_state *x, struct sk_buff *skb)
484 {
485 return xfrm_outer_mode_output(x, skb);
486 }
487 EXPORT_SYMBOL_GPL(pktgen_xfrm_outer_mode_output);
488
xfrm_output_one(struct sk_buff * skb,int err)489 static int xfrm_output_one(struct sk_buff *skb, int err)
490 {
491 struct dst_entry *dst = skb_dst(skb);
492 struct xfrm_state *x = dst->xfrm;
493 struct net *net = xs_net(x);
494
495 if (err <= 0)
496 goto resume;
497
498 do {
499 err = xfrm_skb_check_space(skb);
500 if (err) {
501 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
502 goto error_nolock;
503 }
504
505 skb->mark = xfrm_smark_get(skb->mark, x);
506
507 err = xfrm_outer_mode_output(x, skb);
508 if (err) {
509 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEMODEERROR);
510 goto error_nolock;
511 }
512
513 spin_lock_bh(&x->lock);
514
515 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
516 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEINVALID);
517 err = -EINVAL;
518 goto error;
519 }
520
521 err = xfrm_state_check_expire(x);
522 if (err) {
523 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEEXPIRED);
524 goto error;
525 }
526
527 err = xfrm_replay_overflow(x, skb);
528 if (err) {
529 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATESEQERROR);
530 goto error;
531 }
532
533 x->curlft.bytes += skb->len;
534 x->curlft.packets++;
535
536 spin_unlock_bh(&x->lock);
537
538 skb_dst_force(skb);
539 if (!skb_dst(skb)) {
540 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
541 err = -EHOSTUNREACH;
542 goto error_nolock;
543 }
544
545 if (xfrm_offload(skb)) {
546 x->type_offload->encap(x, skb);
547 } else {
548 /* Inner headers are invalid now. */
549 skb->encapsulation = 0;
550
551 err = x->type->output(x, skb);
552 if (err == -EINPROGRESS)
553 goto out;
554 }
555
556 resume:
557 if (err) {
558 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEPROTOERROR);
559 goto error_nolock;
560 }
561
562 dst = skb_dst_pop(skb);
563 if (!dst) {
564 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
565 err = -EHOSTUNREACH;
566 goto error_nolock;
567 }
568 skb_dst_set(skb, dst);
569 x = dst->xfrm;
570 } while (x && !(x->outer_mode.flags & XFRM_MODE_FLAG_TUNNEL));
571
572 return 0;
573
574 error:
575 spin_unlock_bh(&x->lock);
576 error_nolock:
577 kfree_skb(skb);
578 out:
579 return err;
580 }
581
xfrm_output_resume(struct sk_buff * skb,int err)582 int xfrm_output_resume(struct sk_buff *skb, int err)
583 {
584 struct net *net = xs_net(skb_dst(skb)->xfrm);
585
586 while (likely((err = xfrm_output_one(skb, err)) == 0)) {
587 nf_reset_ct(skb);
588
589 err = skb_dst(skb)->ops->local_out(net, skb->sk, skb);
590 if (unlikely(err != 1))
591 goto out;
592
593 if (!skb_dst(skb)->xfrm)
594 return dst_output(net, skb->sk, skb);
595
596 err = nf_hook(skb_dst(skb)->ops->family,
597 NF_INET_POST_ROUTING, net, skb->sk, skb,
598 NULL, skb_dst(skb)->dev, xfrm_output2);
599 if (unlikely(err != 1))
600 goto out;
601 }
602
603 if (err == -EINPROGRESS)
604 err = 0;
605
606 out:
607 return err;
608 }
609 EXPORT_SYMBOL_GPL(xfrm_output_resume);
610
xfrm_output2(struct net * net,struct sock * sk,struct sk_buff * skb)611 static int xfrm_output2(struct net *net, struct sock *sk, struct sk_buff *skb)
612 {
613 return xfrm_output_resume(skb, 1);
614 }
615
xfrm_output_gso(struct net * net,struct sock * sk,struct sk_buff * skb)616 static int xfrm_output_gso(struct net *net, struct sock *sk, struct sk_buff *skb)
617 {
618 struct sk_buff *segs, *nskb;
619
620 BUILD_BUG_ON(sizeof(*IPCB(skb)) > SKB_GSO_CB_OFFSET);
621 BUILD_BUG_ON(sizeof(*IP6CB(skb)) > SKB_GSO_CB_OFFSET);
622 segs = skb_gso_segment(skb, 0);
623 kfree_skb(skb);
624 if (IS_ERR(segs))
625 return PTR_ERR(segs);
626 if (segs == NULL)
627 return -EINVAL;
628
629 skb_list_walk_safe(segs, segs, nskb) {
630 int err;
631
632 skb_mark_not_on_list(segs);
633 err = xfrm_output2(net, sk, segs);
634
635 if (unlikely(err)) {
636 kfree_skb_list(nskb);
637 return err;
638 }
639 }
640
641 return 0;
642 }
643
644 /* For partial checksum offload, the outer header checksum is calculated
645 * by software and the inner header checksum is calculated by hardware.
646 * This requires hardware to know the inner packet type to calculate
647 * the inner header checksum. Save inner ip protocol here to avoid
648 * traversing the packet in the vendor's xmit code.
649 * For IPsec tunnel mode save the ip protocol from the IP header of the
650 * plain text packet. Otherwise If the encap type is IPIP, just save
651 * skb->inner_ipproto in any other case get the ip protocol from the IP
652 * header.
653 */
xfrm_get_inner_ipproto(struct sk_buff * skb,struct xfrm_state * x)654 static void xfrm_get_inner_ipproto(struct sk_buff *skb, struct xfrm_state *x)
655 {
656 struct xfrm_offload *xo = xfrm_offload(skb);
657 const struct ethhdr *eth;
658
659 if (!xo)
660 return;
661
662 if (x->outer_mode.encap == XFRM_MODE_TUNNEL) {
663 switch (x->outer_mode.family) {
664 case AF_INET:
665 xo->inner_ipproto = ip_hdr(skb)->protocol;
666 break;
667 case AF_INET6:
668 xo->inner_ipproto = ipv6_hdr(skb)->nexthdr;
669 break;
670 default:
671 break;
672 }
673
674 return;
675 }
676
677 /* non-Tunnel Mode */
678 if (!skb->encapsulation)
679 return;
680
681 if (skb->inner_protocol_type == ENCAP_TYPE_IPPROTO) {
682 xo->inner_ipproto = skb->inner_ipproto;
683 return;
684 }
685
686 if (skb->inner_protocol_type != ENCAP_TYPE_ETHER)
687 return;
688
689 eth = (struct ethhdr *)skb_inner_mac_header(skb);
690
691 switch (ntohs(eth->h_proto)) {
692 case ETH_P_IPV6:
693 xo->inner_ipproto = inner_ipv6_hdr(skb)->nexthdr;
694 break;
695 case ETH_P_IP:
696 xo->inner_ipproto = inner_ip_hdr(skb)->protocol;
697 break;
698 }
699 }
700
xfrm_output(struct sock * sk,struct sk_buff * skb)701 int xfrm_output(struct sock *sk, struct sk_buff *skb)
702 {
703 struct net *net = dev_net(skb_dst(skb)->dev);
704 struct xfrm_state *x = skb_dst(skb)->xfrm;
705 int err;
706
707 switch (x->outer_mode.family) {
708 case AF_INET:
709 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
710 IPCB(skb)->flags |= IPSKB_XFRM_TRANSFORMED;
711 break;
712 case AF_INET6:
713 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
714
715 IP6CB(skb)->flags |= IP6SKB_XFRM_TRANSFORMED;
716 break;
717 }
718
719 secpath_reset(skb);
720
721 if (xfrm_dev_offload_ok(skb, x)) {
722 struct sec_path *sp;
723
724 sp = secpath_set(skb);
725 if (!sp) {
726 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
727 kfree_skb(skb);
728 return -ENOMEM;
729 }
730
731 sp->olen++;
732 sp->xvec[sp->len++] = x;
733 xfrm_state_hold(x);
734
735 xfrm_get_inner_ipproto(skb, x);
736 skb->encapsulation = 1;
737
738 if (skb_is_gso(skb)) {
739 if (skb->inner_protocol)
740 return xfrm_output_gso(net, sk, skb);
741
742 skb_shinfo(skb)->gso_type |= SKB_GSO_ESP;
743 goto out;
744 }
745
746 if (x->xso.dev && x->xso.dev->features & NETIF_F_HW_ESP_TX_CSUM)
747 goto out;
748 } else {
749 if (skb_is_gso(skb))
750 return xfrm_output_gso(net, sk, skb);
751 }
752
753 if (skb->ip_summed == CHECKSUM_PARTIAL) {
754 err = skb_checksum_help(skb);
755 if (err) {
756 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
757 kfree_skb(skb);
758 return err;
759 }
760 }
761
762 out:
763 return xfrm_output2(net, sk, skb);
764 }
765 EXPORT_SYMBOL_GPL(xfrm_output);
766
xfrm4_tunnel_check_size(struct sk_buff * skb)767 static int xfrm4_tunnel_check_size(struct sk_buff *skb)
768 {
769 int mtu, ret = 0;
770
771 if (IPCB(skb)->flags & IPSKB_XFRM_TUNNEL_SIZE)
772 goto out;
773
774 if (!(ip_hdr(skb)->frag_off & htons(IP_DF)) || skb->ignore_df)
775 goto out;
776
777 mtu = dst_mtu(skb_dst(skb));
778 if ((!skb_is_gso(skb) && skb->len > mtu) ||
779 (skb_is_gso(skb) &&
780 !skb_gso_validate_network_len(skb, ip_skb_dst_mtu(skb->sk, skb)))) {
781 skb->protocol = htons(ETH_P_IP);
782
783 if (skb->sk)
784 xfrm_local_error(skb, mtu);
785 else
786 icmp_send(skb, ICMP_DEST_UNREACH,
787 ICMP_FRAG_NEEDED, htonl(mtu));
788 ret = -EMSGSIZE;
789 }
790 out:
791 return ret;
792 }
793
xfrm4_extract_output(struct xfrm_state * x,struct sk_buff * skb)794 static int xfrm4_extract_output(struct xfrm_state *x, struct sk_buff *skb)
795 {
796 int err;
797
798 if (x->outer_mode.encap == XFRM_MODE_BEET &&
799 ip_is_fragment(ip_hdr(skb))) {
800 net_warn_ratelimited("BEET mode doesn't support inner IPv4 fragments\n");
801 return -EAFNOSUPPORT;
802 }
803
804 err = xfrm4_tunnel_check_size(skb);
805 if (err)
806 return err;
807
808 XFRM_MODE_SKB_CB(skb)->protocol = ip_hdr(skb)->protocol;
809
810 xfrm4_extract_header(skb);
811 return 0;
812 }
813
814 #if IS_ENABLED(CONFIG_IPV6)
xfrm6_tunnel_check_size(struct sk_buff * skb)815 static int xfrm6_tunnel_check_size(struct sk_buff *skb)
816 {
817 int mtu, ret = 0;
818 struct dst_entry *dst = skb_dst(skb);
819
820 if (skb->ignore_df)
821 goto out;
822
823 mtu = dst_mtu(dst);
824 if (mtu < IPV6_MIN_MTU)
825 mtu = IPV6_MIN_MTU;
826
827 if ((!skb_is_gso(skb) && skb->len > mtu) ||
828 (skb_is_gso(skb) &&
829 !skb_gso_validate_network_len(skb, ip6_skb_dst_mtu(skb)))) {
830 skb->dev = dst->dev;
831 skb->protocol = htons(ETH_P_IPV6);
832
833 if (xfrm6_local_dontfrag(skb->sk))
834 ipv6_stub->xfrm6_local_rxpmtu(skb, mtu);
835 else if (skb->sk)
836 xfrm_local_error(skb, mtu);
837 else
838 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
839 ret = -EMSGSIZE;
840 }
841 out:
842 return ret;
843 }
844 #endif
845
xfrm6_extract_output(struct xfrm_state * x,struct sk_buff * skb)846 static int xfrm6_extract_output(struct xfrm_state *x, struct sk_buff *skb)
847 {
848 #if IS_ENABLED(CONFIG_IPV6)
849 int err;
850
851 err = xfrm6_tunnel_check_size(skb);
852 if (err)
853 return err;
854
855 XFRM_MODE_SKB_CB(skb)->protocol = ipv6_hdr(skb)->nexthdr;
856
857 xfrm6_extract_header(skb);
858 return 0;
859 #else
860 WARN_ON_ONCE(1);
861 return -EAFNOSUPPORT;
862 #endif
863 }
864
xfrm_inner_extract_output(struct xfrm_state * x,struct sk_buff * skb)865 static int xfrm_inner_extract_output(struct xfrm_state *x, struct sk_buff *skb)
866 {
867 const struct xfrm_mode *inner_mode;
868
869 if (x->sel.family == AF_UNSPEC)
870 inner_mode = xfrm_ip2inner_mode(x,
871 xfrm_af2proto(skb_dst(skb)->ops->family));
872 else
873 inner_mode = &x->inner_mode;
874
875 if (inner_mode == NULL)
876 return -EAFNOSUPPORT;
877
878 switch (inner_mode->family) {
879 case AF_INET:
880 return xfrm4_extract_output(x, skb);
881 case AF_INET6:
882 return xfrm6_extract_output(x, skb);
883 }
884
885 return -EAFNOSUPPORT;
886 }
887
xfrm_local_error(struct sk_buff * skb,int mtu)888 void xfrm_local_error(struct sk_buff *skb, int mtu)
889 {
890 unsigned int proto;
891 struct xfrm_state_afinfo *afinfo;
892
893 if (skb->protocol == htons(ETH_P_IP))
894 proto = AF_INET;
895 else if (skb->protocol == htons(ETH_P_IPV6) &&
896 skb->sk->sk_family == AF_INET6)
897 proto = AF_INET6;
898 else
899 return;
900
901 afinfo = xfrm_state_get_afinfo(proto);
902 if (afinfo) {
903 afinfo->local_error(skb, mtu);
904 rcu_read_unlock();
905 }
906 }
907 EXPORT_SYMBOL_GPL(xfrm_local_error);
908