1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * xfrm_output.c - Common IPsec encapsulation code.
5 * Copyright (c) 2007 Herbert Xu <herbert@gondor.apana.org.au>
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>
17 #include <net/inet_ecn.h>
20 #if IS_ENABLED(CONFIG_IPV6)
21 #include <net/ip6_route.h>
22 #include <net/ipv6_stubs.h>
25 #include "xfrm_inout.h"
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);
30 static int xfrm_skb_check_space(struct sk_buff *skb)
32 struct dst_entry *dst = skb_dst(skb);
33 int nhead = dst->header_len + LL_RESERVED_SPACE(dst->dev)
35 int ntail = dst->dev->needed_tailroom - skb_tailroom(skb);
44 return pskb_expand_head(skb, nhead, ntail, GFP_ATOMIC);
47 /* Children define the path of the packet through the
48 * Linux networking. Thus, destinations are stackable.
51 static struct dst_entry *skb_dst_pop(struct sk_buff *skb)
53 struct dst_entry *child = dst_clone(xfrm_dst_child(skb_dst(skb)));
59 /* Add encapsulation header.
61 * The IP header will be moved forward to make space for the encapsulation
64 static int xfrm4_transport_output(struct xfrm_state *x, struct sk_buff *skb)
66 struct iphdr *iph = ip_hdr(skb);
67 int ihl = iph->ihl * 4;
69 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
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;
76 memmove(skb_network_header(skb), iph, ihl);
80 /* Add encapsulation header.
82 * The IP header and mutable extension headers will be moved forward to make
83 * space for the encapsulation header.
85 static int xfrm6_transport_output(struct xfrm_state *x, struct sk_buff *skb)
87 #if IS_ENABLED(CONFIG_IPV6)
93 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
95 hdr_len = x->type->hdr_offset(x, skb, &prevhdr);
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);
107 return -EAFNOSUPPORT;
111 /* Add route optimization header space.
113 * The IP header and mutable extension headers will be moved forward to make
114 * space for the route optimization header.
116 static int xfrm6_ro_output(struct xfrm_state *x, struct sk_buff *skb)
118 #if IS_ENABLED(CONFIG_IPV6)
125 hdr_len = x->type->hdr_offset(x, skb, &prevhdr);
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);
135 x->lastused = ktime_get_real_seconds();
140 return -EAFNOSUPPORT;
144 /* Add encapsulation header.
146 * The top IP header will be constructed per draft-nikander-esp-beet-mode-06.txt.
148 static int xfrm4_beet_encap_add(struct xfrm_state *x, struct sk_buff *skb)
150 struct ip_beet_phdr *ph;
151 struct iphdr *top_iph;
155 optlen = XFRM_MODE_SKB_CB(skb)->optlen;
156 if (unlikely(optlen))
157 hdrlen += IPV4_BEET_PHMAXLEN - (optlen & 4);
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);
167 xfrm4_beet_make_header(skb);
169 ph = __skb_pull(skb, XFRM_MODE_SKB_CB(skb)->ihl - hdrlen);
171 top_iph = ip_hdr(skb);
173 if (unlikely(optlen)) {
174 if (WARN_ON(optlen < 0))
177 ph->padlen = 4 - (optlen & 4);
178 ph->hdrlen = optlen / 8;
179 ph->nexthdr = top_iph->protocol;
181 memset(ph + 1, IPOPT_NOP, ph->padlen);
183 top_iph->protocol = IPPROTO_BEETPH;
184 top_iph->ihl = sizeof(struct iphdr) / 4;
187 top_iph->saddr = x->props.saddr.a4;
188 top_iph->daddr = x->id.daddr.a4;
193 /* Add encapsulation header.
195 * The top IP header will be constructed per RFC 2401.
197 static int xfrm4_tunnel_encap_add(struct xfrm_state *x, struct sk_buff *skb)
199 struct dst_entry *dst = skb_dst(skb);
200 struct iphdr *top_iph;
203 skb_set_inner_network_header(skb, skb_network_offset(skb));
204 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
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);
213 top_iph->version = 4;
215 top_iph->protocol = xfrm_af2proto(skb_dst(skb)->ops->family);
217 /* DS disclosing depends on XFRM_SA_XFLAG_DONT_ENCAP_DSCP */
218 if (x->props.extra_flags & XFRM_SA_XFLAG_DONT_ENCAP_DSCP)
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);
225 flags = x->props.flags;
226 if (flags & XFRM_STATE_NOECN)
227 IP_ECN_clear(top_iph);
229 top_iph->frag_off = (flags & XFRM_STATE_NOPMTUDISC) ?
230 0 : (XFRM_MODE_SKB_CB(skb)->frag_off & htons(IP_DF));
232 top_iph->ttl = ip4_dst_hoplimit(xfrm_dst_child(dst));
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);
241 #if IS_ENABLED(CONFIG_IPV6)
242 static int xfrm6_tunnel_encap_add(struct xfrm_state *x, struct sk_buff *skb)
244 struct dst_entry *dst = skb_dst(skb);
245 struct ipv6hdr *top_iph;
248 skb_set_inner_network_header(skb, skb_network_offset(skb));
249 skb_set_inner_transport_header(skb, skb_transport_offset(skb));
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);
257 top_iph->version = 6;
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);
263 if (x->props.extra_flags & XFRM_SA_XFLAG_DONT_ENCAP_DSCP)
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;
277 static int xfrm6_beet_encap_add(struct xfrm_state *x, struct sk_buff *skb)
279 struct ipv6hdr *top_iph;
280 struct ip_beet_phdr *ph;
284 optlen = XFRM_MODE_SKB_CB(skb)->optlen;
285 if (unlikely(optlen))
286 hdr_len += IPV4_BEET_PHMAXLEN - (optlen & 4);
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);
296 xfrm6_beet_make_header(skb);
298 top_iph = ipv6_hdr(skb);
299 if (unlikely(optlen)) {
300 if (WARN_ON(optlen < 0))
303 ph->padlen = 4 - (optlen & 4);
304 ph->hdrlen = optlen / 8;
305 ph->nexthdr = top_iph->nexthdr;
307 memset(ph + 1, IPOPT_NOP, ph->padlen);
309 top_iph->nexthdr = IPPROTO_BEETPH;
312 top_iph->saddr = *(struct in6_addr *)&x->props.saddr;
313 top_iph->daddr = *(struct in6_addr *)&x->id.daddr;
318 /* Add encapsulation header.
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.
328 static int xfrm4_prepare_output(struct xfrm_state *x, struct sk_buff *skb)
332 err = xfrm_inner_extract_output(x, skb);
336 IPCB(skb)->flags |= IPSKB_XFRM_TUNNEL_SIZE;
337 skb->protocol = htons(ETH_P_IP);
339 switch (x->outer_mode.encap) {
341 return xfrm4_beet_encap_add(x, skb);
342 case XFRM_MODE_TUNNEL:
343 return xfrm4_tunnel_encap_add(x, skb);
350 static int xfrm6_prepare_output(struct xfrm_state *x, struct sk_buff *skb)
352 #if IS_ENABLED(CONFIG_IPV6)
355 err = xfrm_inner_extract_output(x, skb);
360 skb->protocol = htons(ETH_P_IPV6);
362 switch (x->outer_mode.encap) {
364 return xfrm6_beet_encap_add(x, skb);
365 case XFRM_MODE_TUNNEL:
366 return xfrm6_tunnel_encap_add(x, skb);
373 return -EAFNOSUPPORT;
376 static int xfrm_outer_mode_output(struct xfrm_state *x, struct sk_buff *skb)
378 switch (x->outer_mode.encap) {
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);
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);
392 case XFRM_MODE_ROUTEOPTIMIZATION:
393 if (x->outer_mode.family == AF_INET6)
394 return xfrm6_ro_output(x, skb);
405 #if IS_ENABLED(CONFIG_NET_PKTGEN)
406 int pktgen_xfrm_outer_mode_output(struct xfrm_state *x, struct sk_buff *skb)
408 return xfrm_outer_mode_output(x, skb);
410 EXPORT_SYMBOL_GPL(pktgen_xfrm_outer_mode_output);
413 static int xfrm_output_one(struct sk_buff *skb, int err)
415 struct dst_entry *dst = skb_dst(skb);
416 struct xfrm_state *x = dst->xfrm;
417 struct net *net = xs_net(x);
423 err = xfrm_skb_check_space(skb);
425 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
429 skb->mark = xfrm_smark_get(skb->mark, x);
431 err = xfrm_outer_mode_output(x, skb);
433 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEMODEERROR);
437 spin_lock_bh(&x->lock);
439 if (unlikely(x->km.state != XFRM_STATE_VALID)) {
440 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEINVALID);
445 err = xfrm_state_check_expire(x);
447 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEEXPIRED);
451 err = x->repl->overflow(x, skb);
453 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATESEQERROR);
457 x->curlft.bytes += skb->len;
460 spin_unlock_bh(&x->lock);
464 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
469 if (xfrm_offload(skb)) {
470 x->type_offload->encap(x, skb);
472 /* Inner headers are invalid now. */
473 skb->encapsulation = 0;
475 err = x->type->output(x, skb);
476 if (err == -EINPROGRESS)
482 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTSTATEPROTOERROR);
486 dst = skb_dst_pop(skb);
488 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
492 skb_dst_set(skb, dst);
494 } while (x && !(x->outer_mode.flags & XFRM_MODE_FLAG_TUNNEL));
499 spin_unlock_bh(&x->lock);
506 int xfrm_output_resume(struct sk_buff *skb, int err)
508 struct net *net = xs_net(skb_dst(skb)->xfrm);
510 while (likely((err = xfrm_output_one(skb, err)) == 0)) {
513 err = skb_dst(skb)->ops->local_out(net, skb->sk, skb);
514 if (unlikely(err != 1))
517 if (!skb_dst(skb)->xfrm)
518 return dst_output(net, skb->sk, skb);
520 err = nf_hook(skb_dst(skb)->ops->family,
521 NF_INET_POST_ROUTING, net, skb->sk, skb,
522 NULL, skb_dst(skb)->dev, xfrm_output2);
523 if (unlikely(err != 1))
527 if (err == -EINPROGRESS)
533 EXPORT_SYMBOL_GPL(xfrm_output_resume);
535 static int xfrm_output2(struct net *net, struct sock *sk, struct sk_buff *skb)
537 return xfrm_output_resume(skb, 1);
540 static int xfrm_output_gso(struct net *net, struct sock *sk, struct sk_buff *skb)
542 struct sk_buff *segs, *nskb;
544 BUILD_BUG_ON(sizeof(*IPCB(skb)) > SKB_GSO_CB_OFFSET);
545 BUILD_BUG_ON(sizeof(*IP6CB(skb)) > SKB_GSO_CB_OFFSET);
546 segs = skb_gso_segment(skb, 0);
549 return PTR_ERR(segs);
553 skb_list_walk_safe(segs, segs, nskb) {
556 skb_mark_not_on_list(segs);
557 err = xfrm_output2(net, sk, segs);
560 kfree_skb_list(nskb);
568 int xfrm_output(struct sock *sk, struct sk_buff *skb)
570 struct net *net = dev_net(skb_dst(skb)->dev);
571 struct xfrm_state *x = skb_dst(skb)->xfrm;
574 switch (x->outer_mode.family) {
576 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
577 IPCB(skb)->flags |= IPSKB_XFRM_TRANSFORMED;
580 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
582 IP6CB(skb)->flags |= IP6SKB_XFRM_TRANSFORMED;
588 if (xfrm_dev_offload_ok(skb, x)) {
591 sp = secpath_set(skb);
593 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
597 skb->encapsulation = 1;
600 sp->xvec[sp->len++] = x;
603 if (skb_is_gso(skb)) {
604 if (skb->inner_protocol)
605 return xfrm_output_gso(net, sk, skb);
607 skb_shinfo(skb)->gso_type |= SKB_GSO_ESP;
611 if (x->xso.dev && x->xso.dev->features & NETIF_F_HW_ESP_TX_CSUM)
615 return xfrm_output_gso(net, sk, skb);
618 if (skb->ip_summed == CHECKSUM_PARTIAL) {
619 err = skb_checksum_help(skb);
621 XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
628 return xfrm_output2(net, sk, skb);
630 EXPORT_SYMBOL_GPL(xfrm_output);
632 static int xfrm4_tunnel_check_size(struct sk_buff *skb)
636 if (IPCB(skb)->flags & IPSKB_XFRM_TUNNEL_SIZE)
639 if (!(ip_hdr(skb)->frag_off & htons(IP_DF)) || skb->ignore_df)
642 mtu = dst_mtu(skb_dst(skb));
643 if ((!skb_is_gso(skb) && skb->len > mtu) ||
645 !skb_gso_validate_network_len(skb, ip_skb_dst_mtu(skb->sk, skb)))) {
646 skb->protocol = htons(ETH_P_IP);
649 xfrm_local_error(skb, mtu);
651 icmp_send(skb, ICMP_DEST_UNREACH,
652 ICMP_FRAG_NEEDED, htonl(mtu));
659 static int xfrm4_extract_output(struct xfrm_state *x, struct sk_buff *skb)
663 err = xfrm4_tunnel_check_size(skb);
667 XFRM_MODE_SKB_CB(skb)->protocol = ip_hdr(skb)->protocol;
669 xfrm4_extract_header(skb);
673 #if IS_ENABLED(CONFIG_IPV6)
674 static int xfrm6_tunnel_check_size(struct sk_buff *skb)
677 struct dst_entry *dst = skb_dst(skb);
683 if (mtu < IPV6_MIN_MTU)
686 if ((!skb_is_gso(skb) && skb->len > mtu) ||
688 !skb_gso_validate_network_len(skb, ip6_skb_dst_mtu(skb)))) {
690 skb->protocol = htons(ETH_P_IPV6);
692 if (xfrm6_local_dontfrag(skb->sk))
693 ipv6_stub->xfrm6_local_rxpmtu(skb, mtu);
695 xfrm_local_error(skb, mtu);
697 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
705 static int xfrm6_extract_output(struct xfrm_state *x, struct sk_buff *skb)
707 #if IS_ENABLED(CONFIG_IPV6)
710 err = xfrm6_tunnel_check_size(skb);
714 XFRM_MODE_SKB_CB(skb)->protocol = ipv6_hdr(skb)->nexthdr;
716 xfrm6_extract_header(skb);
720 return -EAFNOSUPPORT;
724 static int xfrm_inner_extract_output(struct xfrm_state *x, struct sk_buff *skb)
726 const struct xfrm_mode *inner_mode;
728 if (x->sel.family == AF_UNSPEC)
729 inner_mode = xfrm_ip2inner_mode(x,
730 xfrm_af2proto(skb_dst(skb)->ops->family));
732 inner_mode = &x->inner_mode;
734 if (inner_mode == NULL)
735 return -EAFNOSUPPORT;
737 switch (inner_mode->family) {
739 return xfrm4_extract_output(x, skb);
741 return xfrm6_extract_output(x, skb);
744 return -EAFNOSUPPORT;
747 void xfrm_local_error(struct sk_buff *skb, int mtu)
750 struct xfrm_state_afinfo *afinfo;
752 if (skb->protocol == htons(ETH_P_IP))
754 else if (skb->protocol == htons(ETH_P_IPV6) &&
755 skb->sk->sk_family == AF_INET6)
760 afinfo = xfrm_state_get_afinfo(proto);
762 afinfo->local_error(skb, mtu);
766 EXPORT_SYMBOL_GPL(xfrm_local_error);