1// SPDX-License-Identifier: GPL-2.0-only
  2/**
  3 * Authors:
  4 * (C) 2020 Alexander Aring <alex.aring@gmail.com>
  5 */
  6
  7#include <linux/rpl_iptunnel.h>
  8
  9#include <net/dst_cache.h>
 10#include <net/ip6_route.h>
 11#include <net/lwtunnel.h>
 12#include <net/ipv6.h>
 13#include <net/rpl.h>
 14
 15struct rpl_iptunnel_encap {
 16	struct ipv6_rpl_sr_hdr srh[0];
 17};
 18
 19struct rpl_lwt {
 20	struct dst_cache cache;
 21	struct rpl_iptunnel_encap tuninfo;
 22};
 23
 24static inline struct rpl_lwt *rpl_lwt_lwtunnel(struct lwtunnel_state *lwt)
 25{
 26	return (struct rpl_lwt *)lwt->data;
 27}
 28
 29static inline struct rpl_iptunnel_encap *
 30rpl_encap_lwtunnel(struct lwtunnel_state *lwt)
 31{
 32	return &rpl_lwt_lwtunnel(lwt)->tuninfo;
 33}
 34
 35static const struct nla_policy rpl_iptunnel_policy[RPL_IPTUNNEL_MAX + 1] = {
 36	[RPL_IPTUNNEL_SRH]	= { .type = NLA_BINARY },
 37};
 38
 39static bool rpl_validate_srh(struct net *net, struct ipv6_rpl_sr_hdr *srh,
 40			     size_t seglen)
 41{
 42	int err;
 43
 44	if ((srh->hdrlen << 3) != seglen)
 45		return false;
 46
 47	/* check at least one segment and seglen fit with segments_left */
 48	if (!srh->segments_left ||
 49	    (srh->segments_left * sizeof(struct in6_addr)) != seglen)
 50		return false;
 51
 52	if (srh->cmpri || srh->cmpre)
 53		return false;
 54
 55	err = ipv6_chk_rpl_srh_loop(net, srh->rpl_segaddr,
 56				    srh->segments_left);
 57	if (err)
 58		return false;
 59
 60	if (ipv6_addr_type(&srh->rpl_segaddr[srh->segments_left - 1]) &
 61	    IPV6_ADDR_MULTICAST)
 62		return false;
 63
 64	return true;
 65}
 66
 67static int rpl_build_state(struct net *net, struct nlattr *nla,
 68			   unsigned int family, const void *cfg,
 69			   struct lwtunnel_state **ts,
 70			   struct netlink_ext_ack *extack)
 71{
 72	struct nlattr *tb[RPL_IPTUNNEL_MAX + 1];
 73	struct lwtunnel_state *newts;
 74	struct ipv6_rpl_sr_hdr *srh;
 75	struct rpl_lwt *rlwt;
 76	int err, srh_len;
 77
 78	if (family != AF_INET6)
 79		return -EINVAL;
 80
 81	err = nla_parse_nested(tb, RPL_IPTUNNEL_MAX, nla,
 82			       rpl_iptunnel_policy, extack);
 83	if (err < 0)
 84		return err;
 85
 86	if (!tb[RPL_IPTUNNEL_SRH])
 87		return -EINVAL;
 88
 89	srh = nla_data(tb[RPL_IPTUNNEL_SRH]);
 90	srh_len = nla_len(tb[RPL_IPTUNNEL_SRH]);
 91
 92	if (srh_len < sizeof(*srh))
 93		return -EINVAL;
 94
 95	/* verify that SRH is consistent */
 96	if (!rpl_validate_srh(net, srh, srh_len - sizeof(*srh)))
 97		return -EINVAL;
 98
 99	newts = lwtunnel_state_alloc(srh_len + sizeof(*rlwt));
100	if (!newts)
101		return -ENOMEM;
102
103	rlwt = rpl_lwt_lwtunnel(newts);
104
105	err = dst_cache_init(&rlwt->cache, GFP_ATOMIC);
106	if (err) {
107		kfree(newts);
108		return err;
109	}
110
111	memcpy(&rlwt->tuninfo.srh, srh, srh_len);
112
113	newts->type = LWTUNNEL_ENCAP_RPL;
114	newts->flags |= LWTUNNEL_STATE_INPUT_REDIRECT;
115	newts->flags |= LWTUNNEL_STATE_OUTPUT_REDIRECT;
116
117	*ts = newts;
118
119	return 0;
120}
121
122static void rpl_destroy_state(struct lwtunnel_state *lwt)
123{
124	dst_cache_destroy(&rpl_lwt_lwtunnel(lwt)->cache);
125}
126
127static int rpl_do_srh_inline(struct sk_buff *skb, const struct rpl_lwt *rlwt,
128			     const struct ipv6_rpl_sr_hdr *srh)
129{
130	struct ipv6_rpl_sr_hdr *isrh, *csrh;
131	const struct ipv6hdr *oldhdr;
132	struct ipv6hdr *hdr;
133	unsigned char *buf;
134	size_t hdrlen;
135	int err;
136
137	oldhdr = ipv6_hdr(skb);
138
139	buf = kcalloc(struct_size(srh, segments.addr, srh->segments_left), 2, GFP_ATOMIC);
140	if (!buf)
141		return -ENOMEM;
142
143	isrh = (struct ipv6_rpl_sr_hdr *)buf;
144	csrh = (struct ipv6_rpl_sr_hdr *)(buf + ((srh->hdrlen + 1) << 3));
145
146	memcpy(isrh, srh, sizeof(*isrh));
147	memcpy(isrh->rpl_segaddr, &srh->rpl_segaddr[1],
148	       (srh->segments_left - 1) * 16);
149	isrh->rpl_segaddr[srh->segments_left - 1] = oldhdr->daddr;
150
151	ipv6_rpl_srh_compress(csrh, isrh, &srh->rpl_segaddr[0],
152			      isrh->segments_left - 1);
153
154	hdrlen = ((csrh->hdrlen + 1) << 3);
155
156	err = skb_cow_head(skb, hdrlen + skb->mac_len);
157	if (unlikely(err)) {
158		kfree(buf);
159		return err;
160	}
161
162	skb_pull(skb, sizeof(struct ipv6hdr));
163	skb_postpull_rcsum(skb, skb_network_header(skb),
164			   sizeof(struct ipv6hdr));
165
166	skb_push(skb, sizeof(struct ipv6hdr) + hdrlen);
167	skb_reset_network_header(skb);
168	skb_mac_header_rebuild(skb);
169
170	hdr = ipv6_hdr(skb);
171	memmove(hdr, oldhdr, sizeof(*hdr));
172	isrh = (void *)hdr + sizeof(*hdr);
173	memcpy(isrh, csrh, hdrlen);
174
175	isrh->nexthdr = hdr->nexthdr;
176	hdr->nexthdr = NEXTHDR_ROUTING;
177	hdr->daddr = srh->rpl_segaddr[0];
178
179	ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
180	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
181
182	skb_postpush_rcsum(skb, hdr, sizeof(struct ipv6hdr) + hdrlen);
183
184	kfree(buf);
185
186	return 0;
187}
188
189static int rpl_do_srh(struct sk_buff *skb, const struct rpl_lwt *rlwt)
190{
191	struct dst_entry *dst = skb_dst(skb);
192	struct rpl_iptunnel_encap *tinfo;
193	int err = 0;
194
195	if (skb->protocol != htons(ETH_P_IPV6))
196		return -EINVAL;
197
198	tinfo = rpl_encap_lwtunnel(dst->lwtstate);
199
200	err = rpl_do_srh_inline(skb, rlwt, tinfo->srh);
201	if (err)
202		return err;
203
204	return 0;
205}
206
207static int rpl_output(struct net *net, struct sock *sk, struct sk_buff *skb)
208{
209	struct dst_entry *orig_dst = skb_dst(skb);
210	struct dst_entry *dst = NULL;
211	struct rpl_lwt *rlwt;
212	int err;
213
214	rlwt = rpl_lwt_lwtunnel(orig_dst->lwtstate);
215
216	err = rpl_do_srh(skb, rlwt);
217	if (unlikely(err))
218		goto drop;
219
220	preempt_disable();
221	dst = dst_cache_get(&rlwt->cache);
222	preempt_enable();
223
224	if (unlikely(!dst)) {
225		struct ipv6hdr *hdr = ipv6_hdr(skb);
226		struct flowi6 fl6;
227
228		memset(&fl6, 0, sizeof(fl6));
229		fl6.daddr = hdr->daddr;
230		fl6.saddr = hdr->saddr;
231		fl6.flowlabel = ip6_flowinfo(hdr);
232		fl6.flowi6_mark = skb->mark;
233		fl6.flowi6_proto = hdr->nexthdr;
234
235		dst = ip6_route_output(net, NULL, &fl6);
236		if (dst->error) {
237			err = dst->error;
238			dst_release(dst);
239			goto drop;
240		}
241
242		preempt_disable();
243		dst_cache_set_ip6(&rlwt->cache, dst, &fl6.saddr);
244		preempt_enable();
245	}
246
247	skb_dst_drop(skb);
248	skb_dst_set(skb, dst);
249
250	err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
251	if (unlikely(err))
252		goto drop;
253
254	return dst_output(net, sk, skb);
255
256drop:
257	kfree_skb(skb);
258	return err;
259}
260
261static int rpl_input(struct sk_buff *skb)
262{
263	struct dst_entry *orig_dst = skb_dst(skb);
264	struct dst_entry *dst = NULL;
265	struct rpl_lwt *rlwt;
266	int err;
267
268	rlwt = rpl_lwt_lwtunnel(orig_dst->lwtstate);
269
270	err = rpl_do_srh(skb, rlwt);
271	if (unlikely(err)) {
272		kfree_skb(skb);
273		return err;
274	}
275
276	preempt_disable();
277	dst = dst_cache_get(&rlwt->cache);
278	preempt_enable();
279
280	skb_dst_drop(skb);
281
282	if (!dst) {
283		ip6_route_input(skb);
284		dst = skb_dst(skb);
285		if (!dst->error) {
286			preempt_disable();
287			dst_cache_set_ip6(&rlwt->cache, dst,
288					  &ipv6_hdr(skb)->saddr);
289			preempt_enable();
290		}
291	} else {
292		skb_dst_set(skb, dst);
293	}
294
295	err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
296	if (unlikely(err))
297		return err;
298
299	return dst_input(skb);
300}
301
302static int nla_put_rpl_srh(struct sk_buff *skb, int attrtype,
303			   struct rpl_iptunnel_encap *tuninfo)
304{
305	struct rpl_iptunnel_encap *data;
306	struct nlattr *nla;
307	int len;
308
309	len = RPL_IPTUNNEL_SRH_SIZE(tuninfo->srh);
310
311	nla = nla_reserve(skb, attrtype, len);
312	if (!nla)
313		return -EMSGSIZE;
314
315	data = nla_data(nla);
316	memcpy(data, tuninfo->srh, len);
317
318	return 0;
319}
320
321static int rpl_fill_encap_info(struct sk_buff *skb,
322			       struct lwtunnel_state *lwtstate)
323{
324	struct rpl_iptunnel_encap *tuninfo = rpl_encap_lwtunnel(lwtstate);
325
326	if (nla_put_rpl_srh(skb, RPL_IPTUNNEL_SRH, tuninfo))
327		return -EMSGSIZE;
328
329	return 0;
330}
331
332static int rpl_encap_nlsize(struct lwtunnel_state *lwtstate)
333{
334	struct rpl_iptunnel_encap *tuninfo = rpl_encap_lwtunnel(lwtstate);
335
336	return nla_total_size(RPL_IPTUNNEL_SRH_SIZE(tuninfo->srh));
337}
338
339static int rpl_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
340{
341	struct rpl_iptunnel_encap *a_hdr = rpl_encap_lwtunnel(a);
342	struct rpl_iptunnel_encap *b_hdr = rpl_encap_lwtunnel(b);
343	int len = RPL_IPTUNNEL_SRH_SIZE(a_hdr->srh);
344
345	if (len != RPL_IPTUNNEL_SRH_SIZE(b_hdr->srh))
346		return 1;
347
348	return memcmp(a_hdr, b_hdr, len);
349}
350
351static const struct lwtunnel_encap_ops rpl_ops = {
352	.build_state	= rpl_build_state,
353	.destroy_state	= rpl_destroy_state,
354	.output		= rpl_output,
355	.input		= rpl_input,
356	.fill_encap	= rpl_fill_encap_info,
357	.get_encap_size	= rpl_encap_nlsize,
358	.cmp_encap	= rpl_encap_cmp,
359	.owner		= THIS_MODULE,
360};
361
362int __init rpl_init(void)
363{
364	int err;
365
366	err = lwtunnel_encap_add_ops(&rpl_ops, LWTUNNEL_ENCAP_RPL);
367	if (err)
368		goto out;
369
370	pr_info("RPL Segment Routing with IPv6\n");
371
372	return 0;
373
374out:
375	return err;
376}
377
378void rpl_exit(void)
379{
380	lwtunnel_encap_del_ops(&rpl_ops, LWTUNNEL_ENCAP_RPL);
381}