xref: /openbmc/linux/net/ipv6/rpl_iptunnel.c (revision 8e2a46a4)
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 
15 struct rpl_iptunnel_encap {
16 	struct ipv6_rpl_sr_hdr srh[0];
17 };
18 
19 struct rpl_lwt {
20 	struct dst_cache cache;
21 	struct rpl_iptunnel_encap tuninfo;
22 };
23 
24 static inline struct rpl_lwt *rpl_lwt_lwtunnel(struct lwtunnel_state *lwt)
25 {
26 	return (struct rpl_lwt *)lwt->data;
27 }
28 
29 static inline struct rpl_iptunnel_encap *
30 rpl_encap_lwtunnel(struct lwtunnel_state *lwt)
31 {
32 	return &rpl_lwt_lwtunnel(lwt)->tuninfo;
33 }
34 
35 static const struct nla_policy rpl_iptunnel_policy[RPL_IPTUNNEL_MAX + 1] = {
36 	[RPL_IPTUNNEL_SRH]	= { .type = NLA_BINARY },
37 };
38 
39 static 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 
67 static 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 
122 static void rpl_destroy_state(struct lwtunnel_state *lwt)
123 {
124 	dst_cache_destroy(&rpl_lwt_lwtunnel(lwt)->cache);
125 }
126 
127 static 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 = kzalloc(ipv6_rpl_srh_alloc_size(srh->segments_left - 1) * 2,
140 		      GFP_ATOMIC);
141 	if (!buf)
142 		return -ENOMEM;
143 
144 	isrh = (struct ipv6_rpl_sr_hdr *)buf;
145 	csrh = (struct ipv6_rpl_sr_hdr *)(buf + ((srh->hdrlen + 1) << 3));
146 
147 	memcpy(isrh, srh, sizeof(*isrh));
148 	memcpy(isrh->rpl_segaddr, &srh->rpl_segaddr[1],
149 	       (srh->segments_left - 1) * 16);
150 	isrh->rpl_segaddr[srh->segments_left - 1] = oldhdr->daddr;
151 
152 	ipv6_rpl_srh_compress(csrh, isrh, &srh->rpl_segaddr[0],
153 			      isrh->segments_left - 1);
154 
155 	hdrlen = ((csrh->hdrlen + 1) << 3);
156 
157 	err = skb_cow_head(skb, hdrlen + skb->mac_len);
158 	if (unlikely(err)) {
159 		kfree(buf);
160 		return err;
161 	}
162 
163 	skb_pull(skb, sizeof(struct ipv6hdr));
164 	skb_postpull_rcsum(skb, skb_network_header(skb),
165 			   sizeof(struct ipv6hdr));
166 
167 	skb_push(skb, sizeof(struct ipv6hdr) + hdrlen);
168 	skb_reset_network_header(skb);
169 	skb_mac_header_rebuild(skb);
170 
171 	hdr = ipv6_hdr(skb);
172 	memmove(hdr, oldhdr, sizeof(*hdr));
173 	isrh = (void *)hdr + sizeof(*hdr);
174 	memcpy(isrh, csrh, hdrlen);
175 
176 	isrh->nexthdr = hdr->nexthdr;
177 	hdr->nexthdr = NEXTHDR_ROUTING;
178 	hdr->daddr = srh->rpl_segaddr[0];
179 
180 	ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
181 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
182 
183 	skb_postpush_rcsum(skb, hdr, sizeof(struct ipv6hdr) + hdrlen);
184 
185 	kfree(buf);
186 
187 	return 0;
188 }
189 
190 static int rpl_do_srh(struct sk_buff *skb, const struct rpl_lwt *rlwt)
191 {
192 	struct dst_entry *dst = skb_dst(skb);
193 	struct rpl_iptunnel_encap *tinfo;
194 	int err = 0;
195 
196 	if (skb->protocol != htons(ETH_P_IPV6))
197 		return -EINVAL;
198 
199 	tinfo = rpl_encap_lwtunnel(dst->lwtstate);
200 
201 	err = rpl_do_srh_inline(skb, rlwt, tinfo->srh);
202 	if (err)
203 		return err;
204 
205 	return 0;
206 }
207 
208 static int rpl_output(struct net *net, struct sock *sk, struct sk_buff *skb)
209 {
210 	struct dst_entry *orig_dst = skb_dst(skb);
211 	struct dst_entry *dst = NULL;
212 	struct rpl_lwt *rlwt;
213 	int err;
214 
215 	rlwt = rpl_lwt_lwtunnel(orig_dst->lwtstate);
216 
217 	err = rpl_do_srh(skb, rlwt);
218 	if (unlikely(err))
219 		goto drop;
220 
221 	preempt_disable();
222 	dst = dst_cache_get(&rlwt->cache);
223 	preempt_enable();
224 
225 	if (unlikely(!dst)) {
226 		struct ipv6hdr *hdr = ipv6_hdr(skb);
227 		struct flowi6 fl6;
228 
229 		memset(&fl6, 0, sizeof(fl6));
230 		fl6.daddr = hdr->daddr;
231 		fl6.saddr = hdr->saddr;
232 		fl6.flowlabel = ip6_flowinfo(hdr);
233 		fl6.flowi6_mark = skb->mark;
234 		fl6.flowi6_proto = hdr->nexthdr;
235 
236 		dst = ip6_route_output(net, NULL, &fl6);
237 		if (dst->error) {
238 			err = dst->error;
239 			dst_release(dst);
240 			goto drop;
241 		}
242 
243 		preempt_disable();
244 		dst_cache_set_ip6(&rlwt->cache, dst, &fl6.saddr);
245 		preempt_enable();
246 	}
247 
248 	skb_dst_drop(skb);
249 	skb_dst_set(skb, dst);
250 
251 	err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
252 	if (unlikely(err))
253 		goto drop;
254 
255 	return dst_output(net, sk, skb);
256 
257 drop:
258 	kfree_skb(skb);
259 	return err;
260 }
261 
262 static int rpl_input(struct sk_buff *skb)
263 {
264 	struct dst_entry *orig_dst = skb_dst(skb);
265 	struct dst_entry *dst = NULL;
266 	struct rpl_lwt *rlwt;
267 	int err;
268 
269 	rlwt = rpl_lwt_lwtunnel(orig_dst->lwtstate);
270 
271 	err = rpl_do_srh(skb, rlwt);
272 	if (unlikely(err)) {
273 		kfree_skb(skb);
274 		return err;
275 	}
276 
277 	preempt_disable();
278 	dst = dst_cache_get(&rlwt->cache);
279 	preempt_enable();
280 
281 	skb_dst_drop(skb);
282 
283 	if (!dst) {
284 		ip6_route_input(skb);
285 		dst = skb_dst(skb);
286 		if (!dst->error) {
287 			preempt_disable();
288 			dst_cache_set_ip6(&rlwt->cache, dst,
289 					  &ipv6_hdr(skb)->saddr);
290 			preempt_enable();
291 		}
292 	} else {
293 		skb_dst_set(skb, dst);
294 	}
295 
296 	err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
297 	if (unlikely(err))
298 		return err;
299 
300 	return dst_input(skb);
301 }
302 
303 static int nla_put_rpl_srh(struct sk_buff *skb, int attrtype,
304 			   struct rpl_iptunnel_encap *tuninfo)
305 {
306 	struct rpl_iptunnel_encap *data;
307 	struct nlattr *nla;
308 	int len;
309 
310 	len = RPL_IPTUNNEL_SRH_SIZE(tuninfo->srh);
311 
312 	nla = nla_reserve(skb, attrtype, len);
313 	if (!nla)
314 		return -EMSGSIZE;
315 
316 	data = nla_data(nla);
317 	memcpy(data, tuninfo->srh, len);
318 
319 	return 0;
320 }
321 
322 static int rpl_fill_encap_info(struct sk_buff *skb,
323 			       struct lwtunnel_state *lwtstate)
324 {
325 	struct rpl_iptunnel_encap *tuninfo = rpl_encap_lwtunnel(lwtstate);
326 
327 	if (nla_put_rpl_srh(skb, RPL_IPTUNNEL_SRH, tuninfo))
328 		return -EMSGSIZE;
329 
330 	return 0;
331 }
332 
333 static int rpl_encap_nlsize(struct lwtunnel_state *lwtstate)
334 {
335 	struct rpl_iptunnel_encap *tuninfo = rpl_encap_lwtunnel(lwtstate);
336 
337 	return nla_total_size(RPL_IPTUNNEL_SRH_SIZE(tuninfo->srh));
338 }
339 
340 static int rpl_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
341 {
342 	struct rpl_iptunnel_encap *a_hdr = rpl_encap_lwtunnel(a);
343 	struct rpl_iptunnel_encap *b_hdr = rpl_encap_lwtunnel(b);
344 	int len = RPL_IPTUNNEL_SRH_SIZE(a_hdr->srh);
345 
346 	if (len != RPL_IPTUNNEL_SRH_SIZE(b_hdr->srh))
347 		return 1;
348 
349 	return memcmp(a_hdr, b_hdr, len);
350 }
351 
352 static const struct lwtunnel_encap_ops rpl_ops = {
353 	.build_state	= rpl_build_state,
354 	.destroy_state	= rpl_destroy_state,
355 	.output		= rpl_output,
356 	.input		= rpl_input,
357 	.fill_encap	= rpl_fill_encap_info,
358 	.get_encap_size	= rpl_encap_nlsize,
359 	.cmp_encap	= rpl_encap_cmp,
360 	.owner		= THIS_MODULE,
361 };
362 
363 int __init rpl_init(void)
364 {
365 	int err;
366 
367 	err = lwtunnel_encap_add_ops(&rpl_ops, LWTUNNEL_ENCAP_RPL);
368 	if (err)
369 		goto out;
370 
371 	pr_info("RPL Segment Routing with IPv6\n");
372 
373 	return 0;
374 
375 out:
376 	return err;
377 }
378 
379 void rpl_exit(void)
380 {
381 	lwtunnel_encap_del_ops(&rpl_ops, LWTUNNEL_ENCAP_RPL);
382 }
383