xref: /openbmc/linux/net/mpls/mpls_iptunnel.c (revision be709d48)
1 /*
2  * mpls tunnels	An implementation mpls tunnels using the light weight tunnel
3  *		infrastructure
4  *
5  * Authors:	Roopa Prabhu, <roopa@cumulusnetworks.com>
6  *
7  *		This program is free software; you can redistribute it and/or
8  *		modify it under the terms of the GNU General Public License
9  *		as published by the Free Software Foundation; either version
10  *		2 of the License, or (at your option) any later version.
11  *
12  */
13 #include <linux/types.h>
14 #include <linux/skbuff.h>
15 #include <linux/net.h>
16 #include <linux/module.h>
17 #include <linux/mpls.h>
18 #include <linux/vmalloc.h>
19 #include <net/ip.h>
20 #include <net/dst.h>
21 #include <net/lwtunnel.h>
22 #include <net/netevent.h>
23 #include <net/netns/generic.h>
24 #include <net/ip6_fib.h>
25 #include <net/route.h>
26 #include <net/mpls_iptunnel.h>
27 #include <linux/mpls_iptunnel.h>
28 #include "internal.h"
29 
30 static const struct nla_policy mpls_iptunnel_policy[MPLS_IPTUNNEL_MAX + 1] = {
31 	[MPLS_IPTUNNEL_DST]	= { .type = NLA_U32 },
32 	[MPLS_IPTUNNEL_TTL]	= { .type = NLA_U8 },
33 };
34 
35 static unsigned int mpls_encap_size(struct mpls_iptunnel_encap *en)
36 {
37 	/* The size of the layer 2.5 labels to be added for this route */
38 	return en->labels * sizeof(struct mpls_shim_hdr);
39 }
40 
41 static int mpls_xmit(struct sk_buff *skb)
42 {
43 	struct mpls_iptunnel_encap *tun_encap_info;
44 	struct mpls_shim_hdr *hdr;
45 	struct net_device *out_dev;
46 	unsigned int hh_len;
47 	unsigned int new_header_size;
48 	unsigned int mtu;
49 	struct dst_entry *dst = skb_dst(skb);
50 	struct rtable *rt = NULL;
51 	struct rt6_info *rt6 = NULL;
52 	struct mpls_dev *out_mdev;
53 	struct net *net;
54 	int err = 0;
55 	bool bos;
56 	int i;
57 	unsigned int ttl;
58 
59 	/* Find the output device */
60 	out_dev = dst->dev;
61 	net = dev_net(out_dev);
62 
63 	skb_orphan(skb);
64 
65 	if (!mpls_output_possible(out_dev) ||
66 	    !dst->lwtstate || skb_warn_if_lro(skb))
67 		goto drop;
68 
69 	skb_forward_csum(skb);
70 
71 	tun_encap_info = mpls_lwtunnel_encap(dst->lwtstate);
72 
73 	/* Obtain the ttl using the following set of rules.
74 	 *
75 	 * LWT ttl propagation setting:
76 	 *  - disabled => use default TTL value from LWT
77 	 *  - enabled  => use TTL value from IPv4/IPv6 header
78 	 *  - default  =>
79 	 *   Global ttl propagation setting:
80 	 *    - disabled => use default TTL value from global setting
81 	 *    - enabled => use TTL value from IPv4/IPv6 header
82 	 */
83 	if (dst->ops->family == AF_INET) {
84 		if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DISABLED)
85 			ttl = tun_encap_info->default_ttl;
86 		else if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
87 			 !net->mpls.ip_ttl_propagate)
88 			ttl = net->mpls.default_ttl;
89 		else
90 			ttl = ip_hdr(skb)->ttl;
91 		rt = (struct rtable *)dst;
92 	} else if (dst->ops->family == AF_INET6) {
93 		if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DISABLED)
94 			ttl = tun_encap_info->default_ttl;
95 		else if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
96 			 !net->mpls.ip_ttl_propagate)
97 			ttl = net->mpls.default_ttl;
98 		else
99 			ttl = ipv6_hdr(skb)->hop_limit;
100 		rt6 = (struct rt6_info *)dst;
101 	} else {
102 		goto drop;
103 	}
104 
105 	/* Verify the destination can hold the packet */
106 	new_header_size = mpls_encap_size(tun_encap_info);
107 	mtu = mpls_dev_mtu(out_dev);
108 	if (mpls_pkt_too_big(skb, mtu - new_header_size))
109 		goto drop;
110 
111 	hh_len = LL_RESERVED_SPACE(out_dev);
112 	if (!out_dev->header_ops)
113 		hh_len = 0;
114 
115 	/* Ensure there is enough space for the headers in the skb */
116 	if (skb_cow(skb, hh_len + new_header_size))
117 		goto drop;
118 
119 	skb_set_inner_protocol(skb, skb->protocol);
120 	skb_reset_inner_network_header(skb);
121 
122 	skb_push(skb, new_header_size);
123 
124 	skb_reset_network_header(skb);
125 
126 	skb->dev = out_dev;
127 	skb->protocol = htons(ETH_P_MPLS_UC);
128 
129 	/* Push the new labels */
130 	hdr = mpls_hdr(skb);
131 	bos = true;
132 	for (i = tun_encap_info->labels - 1; i >= 0; i--) {
133 		hdr[i] = mpls_entry_encode(tun_encap_info->label[i],
134 					   ttl, 0, bos);
135 		bos = false;
136 	}
137 
138 	mpls_stats_inc_outucastpkts(out_dev, skb);
139 
140 	if (rt)
141 		err = neigh_xmit(NEIGH_ARP_TABLE, out_dev, &rt->rt_gateway,
142 				 skb);
143 	else if (rt6) {
144 		if (ipv6_addr_v4mapped(&rt6->rt6i_gateway)) {
145 			/* 6PE (RFC 4798) */
146 			err = neigh_xmit(NEIGH_ARP_TABLE, out_dev, &rt6->rt6i_gateway.s6_addr32[3],
147 					 skb);
148 		} else
149 			err = neigh_xmit(NEIGH_ND_TABLE, out_dev, &rt6->rt6i_gateway,
150 					 skb);
151 	}
152 	if (err)
153 		net_dbg_ratelimited("%s: packet transmission failed: %d\n",
154 				    __func__, err);
155 
156 	return LWTUNNEL_XMIT_DONE;
157 
158 drop:
159 	out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL;
160 	if (out_mdev)
161 		MPLS_INC_STATS(out_mdev, tx_errors);
162 	kfree_skb(skb);
163 	return -EINVAL;
164 }
165 
166 static int mpls_build_state(struct nlattr *nla,
167 			    unsigned int family, const void *cfg,
168 			    struct lwtunnel_state **ts,
169 			    struct netlink_ext_ack *extack)
170 {
171 	struct mpls_iptunnel_encap *tun_encap_info;
172 	struct nlattr *tb[MPLS_IPTUNNEL_MAX + 1];
173 	struct lwtunnel_state *newts;
174 	u8 n_labels;
175 	int ret;
176 
177 	ret = nla_parse_nested(tb, MPLS_IPTUNNEL_MAX, nla,
178 			       mpls_iptunnel_policy, extack);
179 	if (ret < 0)
180 		return ret;
181 
182 	if (!tb[MPLS_IPTUNNEL_DST]) {
183 		NL_SET_ERR_MSG(extack, "MPLS_IPTUNNEL_DST attribute is missing");
184 		return -EINVAL;
185 	}
186 
187 	/* determine number of labels */
188 	if (nla_get_labels(tb[MPLS_IPTUNNEL_DST], MAX_NEW_LABELS,
189 			   &n_labels, NULL, extack))
190 		return -EINVAL;
191 
192 	newts = lwtunnel_state_alloc(struct_size(tun_encap_info, label,
193 						 n_labels));
194 	if (!newts)
195 		return -ENOMEM;
196 
197 	tun_encap_info = mpls_lwtunnel_encap(newts);
198 	ret = nla_get_labels(tb[MPLS_IPTUNNEL_DST], n_labels,
199 			     &tun_encap_info->labels, tun_encap_info->label,
200 			     extack);
201 	if (ret)
202 		goto errout;
203 
204 	tun_encap_info->ttl_propagate = MPLS_TTL_PROP_DEFAULT;
205 
206 	if (tb[MPLS_IPTUNNEL_TTL]) {
207 		tun_encap_info->default_ttl = nla_get_u8(tb[MPLS_IPTUNNEL_TTL]);
208 		/* TTL 0 implies propagate from IP header */
209 		tun_encap_info->ttl_propagate = tun_encap_info->default_ttl ?
210 			MPLS_TTL_PROP_DISABLED :
211 			MPLS_TTL_PROP_ENABLED;
212 	}
213 
214 	newts->type = LWTUNNEL_ENCAP_MPLS;
215 	newts->flags |= LWTUNNEL_STATE_XMIT_REDIRECT;
216 	newts->headroom = mpls_encap_size(tun_encap_info);
217 
218 	*ts = newts;
219 
220 	return 0;
221 
222 errout:
223 	kfree(newts);
224 	*ts = NULL;
225 
226 	return ret;
227 }
228 
229 static int mpls_fill_encap_info(struct sk_buff *skb,
230 				struct lwtunnel_state *lwtstate)
231 {
232 	struct mpls_iptunnel_encap *tun_encap_info;
233 
234 	tun_encap_info = mpls_lwtunnel_encap(lwtstate);
235 
236 	if (nla_put_labels(skb, MPLS_IPTUNNEL_DST, tun_encap_info->labels,
237 			   tun_encap_info->label))
238 		goto nla_put_failure;
239 
240 	if (tun_encap_info->ttl_propagate != MPLS_TTL_PROP_DEFAULT &&
241 	    nla_put_u8(skb, MPLS_IPTUNNEL_TTL, tun_encap_info->default_ttl))
242 		goto nla_put_failure;
243 
244 	return 0;
245 
246 nla_put_failure:
247 	return -EMSGSIZE;
248 }
249 
250 static int mpls_encap_nlsize(struct lwtunnel_state *lwtstate)
251 {
252 	struct mpls_iptunnel_encap *tun_encap_info;
253 	int nlsize;
254 
255 	tun_encap_info = mpls_lwtunnel_encap(lwtstate);
256 
257 	nlsize = nla_total_size(tun_encap_info->labels * 4);
258 
259 	if (tun_encap_info->ttl_propagate != MPLS_TTL_PROP_DEFAULT)
260 		nlsize += nla_total_size(1);
261 
262 	return nlsize;
263 }
264 
265 static int mpls_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
266 {
267 	struct mpls_iptunnel_encap *a_hdr = mpls_lwtunnel_encap(a);
268 	struct mpls_iptunnel_encap *b_hdr = mpls_lwtunnel_encap(b);
269 	int l;
270 
271 	if (a_hdr->labels != b_hdr->labels ||
272 	    a_hdr->ttl_propagate != b_hdr->ttl_propagate ||
273 	    a_hdr->default_ttl != b_hdr->default_ttl)
274 		return 1;
275 
276 	for (l = 0; l < a_hdr->labels; l++)
277 		if (a_hdr->label[l] != b_hdr->label[l])
278 			return 1;
279 	return 0;
280 }
281 
282 static const struct lwtunnel_encap_ops mpls_iptun_ops = {
283 	.build_state = mpls_build_state,
284 	.xmit = mpls_xmit,
285 	.fill_encap = mpls_fill_encap_info,
286 	.get_encap_size = mpls_encap_nlsize,
287 	.cmp_encap = mpls_encap_cmp,
288 	.owner = THIS_MODULE,
289 };
290 
291 static int __init mpls_iptunnel_init(void)
292 {
293 	return lwtunnel_encap_add_ops(&mpls_iptun_ops, LWTUNNEL_ENCAP_MPLS);
294 }
295 module_init(mpls_iptunnel_init);
296 
297 static void __exit mpls_iptunnel_exit(void)
298 {
299 	lwtunnel_encap_del_ops(&mpls_iptun_ops, LWTUNNEL_ENCAP_MPLS);
300 }
301 module_exit(mpls_iptunnel_exit);
302 
303 MODULE_ALIAS_RTNL_LWT(MPLS);
304 MODULE_DESCRIPTION("MultiProtocol Label Switching IP Tunnels");
305 MODULE_LICENSE("GPL v2");
306