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]	= { .len = sizeof(u32) },
27 	[MPLS_IPTUNNEL_TTL]	= { .type = NLA_U8 },
28 };
29 
mpls_encap_size(struct mpls_iptunnel_encap * en)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 
mpls_xmit(struct sk_buff * skb)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 	if (!mpls_output_possible(out_dev) ||
59 	    !dst->lwtstate || skb_warn_if_lro(skb))
60 		goto drop;
61 
62 	skb_forward_csum(skb);
63 
64 	tun_encap_info = mpls_lwtunnel_encap(dst->lwtstate);
65 
66 	/* Obtain the ttl using the following set of rules.
67 	 *
68 	 * LWT ttl propagation setting:
69 	 *  - disabled => use default TTL value from LWT
70 	 *  - enabled  => use TTL value from IPv4/IPv6 header
71 	 *  - default  =>
72 	 *   Global ttl propagation setting:
73 	 *    - disabled => use default TTL value from global setting
74 	 *    - enabled => use TTL value from IPv4/IPv6 header
75 	 */
76 	if (dst->ops->family == AF_INET) {
77 		if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DISABLED)
78 			ttl = tun_encap_info->default_ttl;
79 		else if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
80 			 !net->mpls.ip_ttl_propagate)
81 			ttl = net->mpls.default_ttl;
82 		else
83 			ttl = ip_hdr(skb)->ttl;
84 		rt = dst_rtable(dst);
85 	} else if (dst->ops->family == AF_INET6) {
86 		if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DISABLED)
87 			ttl = tun_encap_info->default_ttl;
88 		else if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
89 			 !net->mpls.ip_ttl_propagate)
90 			ttl = net->mpls.default_ttl;
91 		else
92 			ttl = ipv6_hdr(skb)->hop_limit;
93 		rt6 = dst_rt6_info(dst);
94 	} else {
95 		goto drop;
96 	}
97 
98 	/* Verify the destination can hold the packet */
99 	new_header_size = mpls_encap_size(tun_encap_info);
100 	mtu = mpls_dev_mtu(out_dev);
101 	if (mpls_pkt_too_big(skb, mtu - new_header_size))
102 		goto drop;
103 
104 	hh_len = LL_RESERVED_SPACE(out_dev);
105 	if (!out_dev->header_ops)
106 		hh_len = 0;
107 
108 	/* Ensure there is enough space for the headers in the skb */
109 	if (skb_cow_head(skb, hh_len + new_header_size))
110 		goto drop;
111 
112 	skb_set_inner_protocol(skb, skb->protocol);
113 	skb_reset_inner_network_header(skb);
114 
115 	skb_push(skb, new_header_size);
116 
117 	skb_reset_network_header(skb);
118 
119 	skb->dev = out_dev;
120 	skb->protocol = htons(ETH_P_MPLS_UC);
121 
122 	/* Push the new labels */
123 	hdr = mpls_hdr(skb);
124 	bos = true;
125 	for (i = tun_encap_info->labels - 1; i >= 0; i--) {
126 		hdr[i] = mpls_entry_encode(tun_encap_info->label[i],
127 					   ttl, 0, bos);
128 		bos = false;
129 	}
130 
131 	mpls_stats_inc_outucastpkts(out_dev, skb);
132 
133 	if (rt) {
134 		if (rt->rt_gw_family == AF_INET6)
135 			err = neigh_xmit(NEIGH_ND_TABLE, out_dev, &rt->rt_gw6,
136 					 skb);
137 		else
138 			err = neigh_xmit(NEIGH_ARP_TABLE, out_dev, &rt->rt_gw4,
139 					 skb);
140 	} else if (rt6) {
141 		if (ipv6_addr_v4mapped(&rt6->rt6i_gateway)) {
142 			/* 6PE (RFC 4798) */
143 			err = neigh_xmit(NEIGH_ARP_TABLE, out_dev, &rt6->rt6i_gateway.s6_addr32[3],
144 					 skb);
145 		} else
146 			err = neigh_xmit(NEIGH_ND_TABLE, out_dev, &rt6->rt6i_gateway,
147 					 skb);
148 	}
149 	if (err)
150 		net_dbg_ratelimited("%s: packet transmission failed: %d\n",
151 				    __func__, err);
152 
153 	return LWTUNNEL_XMIT_DONE;
154 
155 drop:
156 	out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL;
157 	if (out_mdev)
158 		MPLS_INC_STATS(out_mdev, tx_errors);
159 	kfree_skb(skb);
160 	return -EINVAL;
161 }
162 
mpls_build_state(struct net * net,struct nlattr * nla,unsigned int family,const void * cfg,struct lwtunnel_state ** ts,struct netlink_ext_ack * extack)163 static int mpls_build_state(struct net *net, struct nlattr *nla,
164 			    unsigned int family, const void *cfg,
165 			    struct lwtunnel_state **ts,
166 			    struct netlink_ext_ack *extack)
167 {
168 	struct mpls_iptunnel_encap *tun_encap_info;
169 	struct nlattr *tb[MPLS_IPTUNNEL_MAX + 1];
170 	struct lwtunnel_state *newts;
171 	u8 n_labels;
172 	int ret;
173 
174 	ret = nla_parse_nested_deprecated(tb, MPLS_IPTUNNEL_MAX, nla,
175 					  mpls_iptunnel_policy, extack);
176 	if (ret < 0)
177 		return ret;
178 
179 	if (!tb[MPLS_IPTUNNEL_DST]) {
180 		NL_SET_ERR_MSG(extack, "MPLS_IPTUNNEL_DST attribute is missing");
181 		return -EINVAL;
182 	}
183 
184 	/* determine number of labels */
185 	if (nla_get_labels(tb[MPLS_IPTUNNEL_DST], MAX_NEW_LABELS,
186 			   &n_labels, NULL, extack))
187 		return -EINVAL;
188 
189 	newts = lwtunnel_state_alloc(struct_size(tun_encap_info, label,
190 						 n_labels));
191 	if (!newts)
192 		return -ENOMEM;
193 
194 	tun_encap_info = mpls_lwtunnel_encap(newts);
195 	ret = nla_get_labels(tb[MPLS_IPTUNNEL_DST], n_labels,
196 			     &tun_encap_info->labels, tun_encap_info->label,
197 			     extack);
198 	if (ret)
199 		goto errout;
200 
201 	tun_encap_info->ttl_propagate = MPLS_TTL_PROP_DEFAULT;
202 
203 	if (tb[MPLS_IPTUNNEL_TTL]) {
204 		tun_encap_info->default_ttl = nla_get_u8(tb[MPLS_IPTUNNEL_TTL]);
205 		/* TTL 0 implies propagate from IP header */
206 		tun_encap_info->ttl_propagate = tun_encap_info->default_ttl ?
207 			MPLS_TTL_PROP_DISABLED :
208 			MPLS_TTL_PROP_ENABLED;
209 	}
210 
211 	newts->type = LWTUNNEL_ENCAP_MPLS;
212 	newts->flags |= LWTUNNEL_STATE_XMIT_REDIRECT;
213 	newts->headroom = mpls_encap_size(tun_encap_info);
214 
215 	*ts = newts;
216 
217 	return 0;
218 
219 errout:
220 	kfree(newts);
221 	*ts = NULL;
222 
223 	return ret;
224 }
225 
mpls_fill_encap_info(struct sk_buff * skb,struct lwtunnel_state * lwtstate)226 static int mpls_fill_encap_info(struct sk_buff *skb,
227 				struct lwtunnel_state *lwtstate)
228 {
229 	struct mpls_iptunnel_encap *tun_encap_info;
230 
231 	tun_encap_info = mpls_lwtunnel_encap(lwtstate);
232 
233 	if (nla_put_labels(skb, MPLS_IPTUNNEL_DST, tun_encap_info->labels,
234 			   tun_encap_info->label))
235 		goto nla_put_failure;
236 
237 	if (tun_encap_info->ttl_propagate != MPLS_TTL_PROP_DEFAULT &&
238 	    nla_put_u8(skb, MPLS_IPTUNNEL_TTL, tun_encap_info->default_ttl))
239 		goto nla_put_failure;
240 
241 	return 0;
242 
243 nla_put_failure:
244 	return -EMSGSIZE;
245 }
246 
mpls_encap_nlsize(struct lwtunnel_state * lwtstate)247 static int mpls_encap_nlsize(struct lwtunnel_state *lwtstate)
248 {
249 	struct mpls_iptunnel_encap *tun_encap_info;
250 	int nlsize;
251 
252 	tun_encap_info = mpls_lwtunnel_encap(lwtstate);
253 
254 	nlsize = nla_total_size(tun_encap_info->labels * 4);
255 
256 	if (tun_encap_info->ttl_propagate != MPLS_TTL_PROP_DEFAULT)
257 		nlsize += nla_total_size(1);
258 
259 	return nlsize;
260 }
261 
mpls_encap_cmp(struct lwtunnel_state * a,struct lwtunnel_state * b)262 static int mpls_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
263 {
264 	struct mpls_iptunnel_encap *a_hdr = mpls_lwtunnel_encap(a);
265 	struct mpls_iptunnel_encap *b_hdr = mpls_lwtunnel_encap(b);
266 	int l;
267 
268 	if (a_hdr->labels != b_hdr->labels ||
269 	    a_hdr->ttl_propagate != b_hdr->ttl_propagate ||
270 	    a_hdr->default_ttl != b_hdr->default_ttl)
271 		return 1;
272 
273 	for (l = 0; l < a_hdr->labels; l++)
274 		if (a_hdr->label[l] != b_hdr->label[l])
275 			return 1;
276 	return 0;
277 }
278 
279 static const struct lwtunnel_encap_ops mpls_iptun_ops = {
280 	.build_state = mpls_build_state,
281 	.xmit = mpls_xmit,
282 	.fill_encap = mpls_fill_encap_info,
283 	.get_encap_size = mpls_encap_nlsize,
284 	.cmp_encap = mpls_encap_cmp,
285 	.owner = THIS_MODULE,
286 };
287 
mpls_iptunnel_init(void)288 static int __init mpls_iptunnel_init(void)
289 {
290 	return lwtunnel_encap_add_ops(&mpls_iptun_ops, LWTUNNEL_ENCAP_MPLS);
291 }
292 module_init(mpls_iptunnel_init);
293 
mpls_iptunnel_exit(void)294 static void __exit mpls_iptunnel_exit(void)
295 {
296 	lwtunnel_encap_del_ops(&mpls_iptun_ops, LWTUNNEL_ENCAP_MPLS);
297 }
298 module_exit(mpls_iptunnel_exit);
299 
300 MODULE_ALIAS_RTNL_LWT(MPLS);
301 MODULE_SOFTDEP("post: mpls_gso");
302 MODULE_DESCRIPTION("MultiProtocol Label Switching IP Tunnels");
303 MODULE_LICENSE("GPL v2");
304