xref: /linux/net/mpls/mpls_iptunnel.c (revision 63307d015b91e626c97bb82e88054af3d0b74643)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * mpls tunnels	An implementation mpls tunnels using the light weight tunnel
4  *		infrastructure
5  *
6  * Authors:	Roopa Prabhu, <roopa@cumulusnetworks.com>
7  */
8 #include <linux/types.h>
9 #include <linux/skbuff.h>
10 #include <linux/net.h>
11 #include <linux/module.h>
12 #include <linux/mpls.h>
13 #include <linux/vmalloc.h>
14 #include <net/ip.h>
15 #include <net/dst.h>
16 #include <net/lwtunnel.h>
17 #include <net/netevent.h>
18 #include <net/netns/generic.h>
19 #include <net/ip6_fib.h>
20 #include <net/route.h>
21 #include <net/mpls_iptunnel.h>
22 #include <linux/mpls_iptunnel.h>
23 #include "internal.h"
24 
25 static const struct nla_policy mpls_iptunnel_policy[MPLS_IPTUNNEL_MAX + 1] = {
26 	[MPLS_IPTUNNEL_DST]	= { .type = NLA_U32 },
27 	[MPLS_IPTUNNEL_TTL]	= { .type = NLA_U8 },
28 };
29 
30 static unsigned int mpls_encap_size(struct mpls_iptunnel_encap *en)
31 {
32 	/* The size of the layer 2.5 labels to be added for this route */
33 	return en->labels * sizeof(struct mpls_shim_hdr);
34 }
35 
36 static int mpls_xmit(struct sk_buff *skb)
37 {
38 	struct mpls_iptunnel_encap *tun_encap_info;
39 	struct mpls_shim_hdr *hdr;
40 	struct net_device *out_dev;
41 	unsigned int hh_len;
42 	unsigned int new_header_size;
43 	unsigned int mtu;
44 	struct dst_entry *dst = skb_dst(skb);
45 	struct rtable *rt = NULL;
46 	struct rt6_info *rt6 = NULL;
47 	struct mpls_dev *out_mdev;
48 	struct net *net;
49 	int err = 0;
50 	bool bos;
51 	int i;
52 	unsigned int ttl;
53 
54 	/* Find the output device */
55 	out_dev = dst->dev;
56 	net = dev_net(out_dev);
57 
58 	skb_orphan(skb);
59 
60 	if (!mpls_output_possible(out_dev) ||
61 	    !dst->lwtstate || skb_warn_if_lro(skb))
62 		goto drop;
63 
64 	skb_forward_csum(skb);
65 
66 	tun_encap_info = mpls_lwtunnel_encap(dst->lwtstate);
67 
68 	/* Obtain the ttl using the following set of rules.
69 	 *
70 	 * LWT ttl propagation setting:
71 	 *  - disabled => use default TTL value from LWT
72 	 *  - enabled  => use TTL value from IPv4/IPv6 header
73 	 *  - default  =>
74 	 *   Global ttl propagation setting:
75 	 *    - disabled => use default TTL value from global setting
76 	 *    - enabled => use TTL value from IPv4/IPv6 header
77 	 */
78 	if (dst->ops->family == AF_INET) {
79 		if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DISABLED)
80 			ttl = tun_encap_info->default_ttl;
81 		else if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
82 			 !net->mpls.ip_ttl_propagate)
83 			ttl = net->mpls.default_ttl;
84 		else
85 			ttl = ip_hdr(skb)->ttl;
86 		rt = (struct rtable *)dst;
87 	} else if (dst->ops->family == AF_INET6) {
88 		if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DISABLED)
89 			ttl = tun_encap_info->default_ttl;
90 		else if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
91 			 !net->mpls.ip_ttl_propagate)
92 			ttl = net->mpls.default_ttl;
93 		else
94 			ttl = ipv6_hdr(skb)->hop_limit;
95 		rt6 = (struct rt6_info *)dst;
96 	} else {
97 		goto drop;
98 	}
99 
100 	/* Verify the destination can hold the packet */
101 	new_header_size = mpls_encap_size(tun_encap_info);
102 	mtu = mpls_dev_mtu(out_dev);
103 	if (mpls_pkt_too_big(skb, mtu - new_header_size))
104 		goto drop;
105 
106 	hh_len = LL_RESERVED_SPACE(out_dev);
107 	if (!out_dev->header_ops)
108 		hh_len = 0;
109 
110 	/* Ensure there is enough space for the headers in the skb */
111 	if (skb_cow(skb, hh_len + new_header_size))
112 		goto drop;
113 
114 	skb_set_inner_protocol(skb, skb->protocol);
115 	skb_reset_inner_network_header(skb);
116 
117 	skb_push(skb, new_header_size);
118 
119 	skb_reset_network_header(skb);
120 
121 	skb->dev = out_dev;
122 	skb->protocol = htons(ETH_P_MPLS_UC);
123 
124 	/* Push the new labels */
125 	hdr = mpls_hdr(skb);
126 	bos = true;
127 	for (i = tun_encap_info->labels - 1; i >= 0; i--) {
128 		hdr[i] = mpls_entry_encode(tun_encap_info->label[i],
129 					   ttl, 0, bos);
130 		bos = false;
131 	}
132 
133 	mpls_stats_inc_outucastpkts(out_dev, skb);
134 
135 	if (rt) {
136 		if (rt->rt_gw_family == AF_INET)
137 			err = neigh_xmit(NEIGH_ARP_TABLE, out_dev, &rt->rt_gw4,
138 					 skb);
139 		else if (rt->rt_gw_family == AF_INET6)
140 			err = neigh_xmit(NEIGH_ND_TABLE, out_dev, &rt->rt_gw6,
141 					 skb);
142 	} else if (rt6) {
143 		if (ipv6_addr_v4mapped(&rt6->rt6i_gateway)) {
144 			/* 6PE (RFC 4798) */
145 			err = neigh_xmit(NEIGH_ARP_TABLE, out_dev, &rt6->rt6i_gateway.s6_addr32[3],
146 					 skb);
147 		} else
148 			err = neigh_xmit(NEIGH_ND_TABLE, out_dev, &rt6->rt6i_gateway,
149 					 skb);
150 	}
151 	if (err)
152 		net_dbg_ratelimited("%s: packet transmission failed: %d\n",
153 				    __func__, err);
154 
155 	return LWTUNNEL_XMIT_DONE;
156 
157 drop:
158 	out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL;
159 	if (out_mdev)
160 		MPLS_INC_STATS(out_mdev, tx_errors);
161 	kfree_skb(skb);
162 	return -EINVAL;
163 }
164 
165 static int mpls_build_state(struct nlattr *nla,
166 			    unsigned int family, const void *cfg,
167 			    struct lwtunnel_state **ts,
168 			    struct netlink_ext_ack *extack)
169 {
170 	struct mpls_iptunnel_encap *tun_encap_info;
171 	struct nlattr *tb[MPLS_IPTUNNEL_MAX + 1];
172 	struct lwtunnel_state *newts;
173 	u8 n_labels;
174 	int ret;
175 
176 	ret = nla_parse_nested_deprecated(tb, MPLS_IPTUNNEL_MAX, nla,
177 					  mpls_iptunnel_policy, extack);
178 	if (ret < 0)
179 		return ret;
180 
181 	if (!tb[MPLS_IPTUNNEL_DST]) {
182 		NL_SET_ERR_MSG(extack, "MPLS_IPTUNNEL_DST attribute is missing");
183 		return -EINVAL;
184 	}
185 
186 	/* determine number of labels */
187 	if (nla_get_labels(tb[MPLS_IPTUNNEL_DST], MAX_NEW_LABELS,
188 			   &n_labels, NULL, extack))
189 		return -EINVAL;
190 
191 	newts = lwtunnel_state_alloc(struct_size(tun_encap_info, label,
192 						 n_labels));
193 	if (!newts)
194 		return -ENOMEM;
195 
196 	tun_encap_info = mpls_lwtunnel_encap(newts);
197 	ret = nla_get_labels(tb[MPLS_IPTUNNEL_DST], n_labels,
198 			     &tun_encap_info->labels, tun_encap_info->label,
199 			     extack);
200 	if (ret)
201 		goto errout;
202 
203 	tun_encap_info->ttl_propagate = MPLS_TTL_PROP_DEFAULT;
204 
205 	if (tb[MPLS_IPTUNNEL_TTL]) {
206 		tun_encap_info->default_ttl = nla_get_u8(tb[MPLS_IPTUNNEL_TTL]);
207 		/* TTL 0 implies propagate from IP header */
208 		tun_encap_info->ttl_propagate = tun_encap_info->default_ttl ?
209 			MPLS_TTL_PROP_DISABLED :
210 			MPLS_TTL_PROP_ENABLED;
211 	}
212 
213 	newts->type = LWTUNNEL_ENCAP_MPLS;
214 	newts->flags |= LWTUNNEL_STATE_XMIT_REDIRECT;
215 	newts->headroom = mpls_encap_size(tun_encap_info);
216 
217 	*ts = newts;
218 
219 	return 0;
220 
221 errout:
222 	kfree(newts);
223 	*ts = NULL;
224 
225 	return ret;
226 }
227 
228 static int mpls_fill_encap_info(struct sk_buff *skb,
229 				struct lwtunnel_state *lwtstate)
230 {
231 	struct mpls_iptunnel_encap *tun_encap_info;
232 
233 	tun_encap_info = mpls_lwtunnel_encap(lwtstate);
234 
235 	if (nla_put_labels(skb, MPLS_IPTUNNEL_DST, tun_encap_info->labels,
236 			   tun_encap_info->label))
237 		goto nla_put_failure;
238 
239 	if (tun_encap_info->ttl_propagate != MPLS_TTL_PROP_DEFAULT &&
240 	    nla_put_u8(skb, MPLS_IPTUNNEL_TTL, tun_encap_info->default_ttl))
241 		goto nla_put_failure;
242 
243 	return 0;
244 
245 nla_put_failure:
246 	return -EMSGSIZE;
247 }
248 
249 static int mpls_encap_nlsize(struct lwtunnel_state *lwtstate)
250 {
251 	struct mpls_iptunnel_encap *tun_encap_info;
252 	int nlsize;
253 
254 	tun_encap_info = mpls_lwtunnel_encap(lwtstate);
255 
256 	nlsize = nla_total_size(tun_encap_info->labels * 4);
257 
258 	if (tun_encap_info->ttl_propagate != MPLS_TTL_PROP_DEFAULT)
259 		nlsize += nla_total_size(1);
260 
261 	return nlsize;
262 }
263 
264 static int mpls_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
265 {
266 	struct mpls_iptunnel_encap *a_hdr = mpls_lwtunnel_encap(a);
267 	struct mpls_iptunnel_encap *b_hdr = mpls_lwtunnel_encap(b);
268 	int l;
269 
270 	if (a_hdr->labels != b_hdr->labels ||
271 	    a_hdr->ttl_propagate != b_hdr->ttl_propagate ||
272 	    a_hdr->default_ttl != b_hdr->default_ttl)
273 		return 1;
274 
275 	for (l = 0; l < a_hdr->labels; l++)
276 		if (a_hdr->label[l] != b_hdr->label[l])
277 			return 1;
278 	return 0;
279 }
280 
281 static const struct lwtunnel_encap_ops mpls_iptun_ops = {
282 	.build_state = mpls_build_state,
283 	.xmit = mpls_xmit,
284 	.fill_encap = mpls_fill_encap_info,
285 	.get_encap_size = mpls_encap_nlsize,
286 	.cmp_encap = mpls_encap_cmp,
287 	.owner = THIS_MODULE,
288 };
289 
290 static int __init mpls_iptunnel_init(void)
291 {
292 	return lwtunnel_encap_add_ops(&mpls_iptun_ops, LWTUNNEL_ENCAP_MPLS);
293 }
294 module_init(mpls_iptunnel_init);
295 
296 static void __exit mpls_iptunnel_exit(void)
297 {
298 	lwtunnel_encap_del_ops(&mpls_iptun_ops, LWTUNNEL_ENCAP_MPLS);
299 }
300 module_exit(mpls_iptunnel_exit);
301 
302 MODULE_ALIAS_RTNL_LWT(MPLS);
303 MODULE_DESCRIPTION("MultiProtocol Label Switching IP Tunnels");
304 MODULE_LICENSE("GPL v2");
305