xref: /openbmc/linux/net/ipv6/ioam6_iptunnel.c (revision 2326f3cd)
1 // SPDX-License-Identifier: GPL-2.0+
2 /*
3  *  IPv6 IOAM Lightweight Tunnel implementation
4  *
5  *  Author:
6  *  Justin Iurman <justin.iurman@uliege.be>
7  */
8 
9 #include <linux/kernel.h>
10 #include <linux/skbuff.h>
11 #include <linux/net.h>
12 #include <linux/netlink.h>
13 #include <linux/in6.h>
14 #include <linux/ioam6.h>
15 #include <linux/ioam6_iptunnel.h>
16 #include <net/dst.h>
17 #include <net/sock.h>
18 #include <net/lwtunnel.h>
19 #include <net/ioam6.h>
20 
21 #define IOAM6_MASK_SHORT_FIELDS 0xff100000
22 #define IOAM6_MASK_WIDE_FIELDS 0xe00000
23 
24 struct ioam6_lwt_encap {
25 	struct ipv6_hopopt_hdr	eh;
26 	u8			pad[2];	/* 2-octet padding for 4n-alignment */
27 	struct ioam6_hdr	ioamh;
28 	struct ioam6_trace_hdr	traceh;
29 } __packed;
30 
31 struct ioam6_lwt {
32 	struct ioam6_lwt_encap	tuninfo;
33 };
34 
35 static struct ioam6_lwt *ioam6_lwt_state(struct lwtunnel_state *lwt)
36 {
37 	return (struct ioam6_lwt *)lwt->data;
38 }
39 
40 static struct ioam6_lwt_encap *ioam6_lwt_info(struct lwtunnel_state *lwt)
41 {
42 	return &ioam6_lwt_state(lwt)->tuninfo;
43 }
44 
45 static struct ioam6_trace_hdr *ioam6_trace(struct lwtunnel_state *lwt)
46 {
47 	return &(ioam6_lwt_state(lwt)->tuninfo.traceh);
48 }
49 
50 static const struct nla_policy ioam6_iptunnel_policy[IOAM6_IPTUNNEL_MAX + 1] = {
51 	[IOAM6_IPTUNNEL_TRACE]	= NLA_POLICY_EXACT_LEN(sizeof(struct ioam6_trace_hdr)),
52 };
53 
54 static int nla_put_ioam6_trace(struct sk_buff *skb, int attrtype,
55 			       struct ioam6_trace_hdr *trace)
56 {
57 	struct ioam6_trace_hdr *data;
58 	struct nlattr *nla;
59 	int len;
60 
61 	len = sizeof(*trace);
62 
63 	nla = nla_reserve(skb, attrtype, len);
64 	if (!nla)
65 		return -EMSGSIZE;
66 
67 	data = nla_data(nla);
68 	memcpy(data, trace, len);
69 
70 	return 0;
71 }
72 
73 static bool ioam6_validate_trace_hdr(struct ioam6_trace_hdr *trace)
74 {
75 	u32 fields;
76 
77 	if (!trace->type_be32 || !trace->remlen ||
78 	    trace->remlen > IOAM6_TRACE_DATA_SIZE_MAX / 4)
79 		return false;
80 
81 	trace->nodelen = 0;
82 	fields = be32_to_cpu(trace->type_be32);
83 
84 	trace->nodelen += hweight32(fields & IOAM6_MASK_SHORT_FIELDS)
85 				* (sizeof(__be32) / 4);
86 	trace->nodelen += hweight32(fields & IOAM6_MASK_WIDE_FIELDS)
87 				* (sizeof(__be64) / 4);
88 
89 	return true;
90 }
91 
92 static int ioam6_build_state(struct net *net, struct nlattr *nla,
93 			     unsigned int family, const void *cfg,
94 			     struct lwtunnel_state **ts,
95 			     struct netlink_ext_ack *extack)
96 {
97 	struct nlattr *tb[IOAM6_IPTUNNEL_MAX + 1];
98 	struct ioam6_lwt_encap *tuninfo;
99 	struct ioam6_trace_hdr *trace;
100 	struct lwtunnel_state *s;
101 	int len_aligned;
102 	int len, err;
103 
104 	if (family != AF_INET6)
105 		return -EINVAL;
106 
107 	err = nla_parse_nested(tb, IOAM6_IPTUNNEL_MAX, nla,
108 			       ioam6_iptunnel_policy, extack);
109 	if (err < 0)
110 		return err;
111 
112 	if (!tb[IOAM6_IPTUNNEL_TRACE]) {
113 		NL_SET_ERR_MSG(extack, "missing trace");
114 		return -EINVAL;
115 	}
116 
117 	trace = nla_data(tb[IOAM6_IPTUNNEL_TRACE]);
118 	if (!ioam6_validate_trace_hdr(trace)) {
119 		NL_SET_ERR_MSG_ATTR(extack, tb[IOAM6_IPTUNNEL_TRACE],
120 				    "invalid trace validation");
121 		return -EINVAL;
122 	}
123 
124 	len = sizeof(*tuninfo) + trace->remlen * 4;
125 	len_aligned = ALIGN(len, 8);
126 
127 	s = lwtunnel_state_alloc(len_aligned);
128 	if (!s)
129 		return -ENOMEM;
130 
131 	tuninfo = ioam6_lwt_info(s);
132 	tuninfo->eh.hdrlen = (len_aligned >> 3) - 1;
133 	tuninfo->pad[0] = IPV6_TLV_PADN;
134 	tuninfo->ioamh.type = IOAM6_TYPE_PREALLOC;
135 	tuninfo->ioamh.opt_type = IPV6_TLV_IOAM;
136 	tuninfo->ioamh.opt_len = sizeof(tuninfo->ioamh) - 2 + sizeof(*trace)
137 					+ trace->remlen * 4;
138 
139 	memcpy(&tuninfo->traceh, trace, sizeof(*trace));
140 
141 	len = len_aligned - len;
142 	if (len == 1) {
143 		tuninfo->traceh.data[trace->remlen * 4] = IPV6_TLV_PAD1;
144 	} else if (len > 0) {
145 		tuninfo->traceh.data[trace->remlen * 4] = IPV6_TLV_PADN;
146 		tuninfo->traceh.data[trace->remlen * 4 + 1] = len - 2;
147 	}
148 
149 	s->type = LWTUNNEL_ENCAP_IOAM6;
150 	s->flags |= LWTUNNEL_STATE_OUTPUT_REDIRECT;
151 
152 	*ts = s;
153 
154 	return 0;
155 }
156 
157 static int ioam6_do_inline(struct sk_buff *skb, struct ioam6_lwt_encap *tuninfo)
158 {
159 	struct ioam6_trace_hdr *trace;
160 	struct ipv6hdr *oldhdr, *hdr;
161 	struct ioam6_namespace *ns;
162 	int hdrlen, err;
163 
164 	hdrlen = (tuninfo->eh.hdrlen + 1) << 3;
165 
166 	err = skb_cow_head(skb, hdrlen + skb->mac_len);
167 	if (unlikely(err))
168 		return err;
169 
170 	oldhdr = ipv6_hdr(skb);
171 	skb_pull(skb, sizeof(*oldhdr));
172 	skb_postpull_rcsum(skb, skb_network_header(skb), sizeof(*oldhdr));
173 
174 	skb_push(skb, sizeof(*oldhdr) + hdrlen);
175 	skb_reset_network_header(skb);
176 	skb_mac_header_rebuild(skb);
177 
178 	hdr = ipv6_hdr(skb);
179 	memmove(hdr, oldhdr, sizeof(*oldhdr));
180 	tuninfo->eh.nexthdr = hdr->nexthdr;
181 
182 	skb_set_transport_header(skb, sizeof(*hdr));
183 	skb_postpush_rcsum(skb, hdr, sizeof(*hdr) + hdrlen);
184 
185 	memcpy(skb_transport_header(skb), (u8 *)tuninfo, hdrlen);
186 
187 	hdr->nexthdr = NEXTHDR_HOP;
188 	hdr->payload_len = cpu_to_be16(skb->len - sizeof(*hdr));
189 
190 	trace = (struct ioam6_trace_hdr *)(skb_transport_header(skb)
191 					   + sizeof(struct ipv6_hopopt_hdr) + 2
192 					   + sizeof(struct ioam6_hdr));
193 
194 	ns = ioam6_namespace(dev_net(skb_dst(skb)->dev), trace->namespace_id);
195 	if (ns)
196 		ioam6_fill_trace_data(skb, ns, trace);
197 
198 	return 0;
199 }
200 
201 static int ioam6_output(struct net *net, struct sock *sk, struct sk_buff *skb)
202 {
203 	struct lwtunnel_state *lwt = skb_dst(skb)->lwtstate;
204 	int err = -EINVAL;
205 
206 	if (skb->protocol != htons(ETH_P_IPV6))
207 		goto drop;
208 
209 	/* Only for packets we send and
210 	 * that do not contain a Hop-by-Hop yet
211 	 */
212 	if (skb->dev || ipv6_hdr(skb)->nexthdr == NEXTHDR_HOP)
213 		goto out;
214 
215 	err = ioam6_do_inline(skb, ioam6_lwt_info(lwt));
216 	if (unlikely(err))
217 		goto drop;
218 
219 	err = skb_cow_head(skb, LL_RESERVED_SPACE(skb_dst(skb)->dev));
220 	if (unlikely(err))
221 		goto drop;
222 
223 out:
224 	return lwt->orig_output(net, sk, skb);
225 
226 drop:
227 	kfree_skb(skb);
228 	return err;
229 }
230 
231 static int ioam6_fill_encap_info(struct sk_buff *skb,
232 				 struct lwtunnel_state *lwtstate)
233 {
234 	struct ioam6_trace_hdr *trace = ioam6_trace(lwtstate);
235 
236 	if (nla_put_ioam6_trace(skb, IOAM6_IPTUNNEL_TRACE, trace))
237 		return -EMSGSIZE;
238 
239 	return 0;
240 }
241 
242 static int ioam6_encap_nlsize(struct lwtunnel_state *lwtstate)
243 {
244 	struct ioam6_trace_hdr *trace = ioam6_trace(lwtstate);
245 
246 	return nla_total_size(sizeof(*trace));
247 }
248 
249 static int ioam6_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
250 {
251 	struct ioam6_trace_hdr *a_hdr = ioam6_trace(a);
252 	struct ioam6_trace_hdr *b_hdr = ioam6_trace(b);
253 
254 	return (a_hdr->namespace_id != b_hdr->namespace_id);
255 }
256 
257 static const struct lwtunnel_encap_ops ioam6_iptun_ops = {
258 	.build_state	= ioam6_build_state,
259 	.output		= ioam6_output,
260 	.fill_encap	= ioam6_fill_encap_info,
261 	.get_encap_size	= ioam6_encap_nlsize,
262 	.cmp_encap	= ioam6_encap_cmp,
263 	.owner		= THIS_MODULE,
264 };
265 
266 int __init ioam6_iptunnel_init(void)
267 {
268 	return lwtunnel_encap_add_ops(&ioam6_iptun_ops, LWTUNNEL_ENCAP_IOAM6);
269 }
270 
271 void ioam6_iptunnel_exit(void)
272 {
273 	lwtunnel_encap_del_ops(&ioam6_iptun_ops, LWTUNNEL_ENCAP_IOAM6);
274 }
275