1 /* Copyright (c) 2016,2017 Facebook
2  *
3  * This program is free software; you can redistribute it and/or
4  * modify it under the terms of version 2 of the GNU General Public
5  * License as published by the Free Software Foundation.
6  */
7 #include <stddef.h>
8 #include <string.h>
9 #include <linux/bpf.h>
10 #include <linux/if_ether.h>
11 #include <linux/if_packet.h>
12 #include <linux/ip.h>
13 #include <linux/ipv6.h>
14 #include <linux/in.h>
15 #include <linux/udp.h>
16 #include <linux/tcp.h>
17 #include <linux/pkt_cls.h>
18 #include <sys/socket.h>
19 #include "bpf_helpers.h"
20 #include "bpf_endian.h"
21 #include "test_iptunnel_common.h"
22 
23 int _version SEC("version") = 1;
24 
25 struct bpf_map_def SEC("maps") rxcnt = {
26 	.type = BPF_MAP_TYPE_PERCPU_ARRAY,
27 	.key_size = sizeof(__u32),
28 	.value_size = sizeof(__u64),
29 	.max_entries = 256,
30 };
31 
32 struct bpf_map_def SEC("maps") vip2tnl = {
33 	.type = BPF_MAP_TYPE_HASH,
34 	.key_size = sizeof(struct vip),
35 	.value_size = sizeof(struct iptnl_info),
36 	.max_entries = MAX_IPTNL_ENTRIES,
37 };
38 
39 static __always_inline void count_tx(__u32 protocol)
40 {
41 	__u64 *rxcnt_count;
42 
43 	rxcnt_count = bpf_map_lookup_elem(&rxcnt, &protocol);
44 	if (rxcnt_count)
45 		*rxcnt_count += 1;
46 }
47 
48 static __always_inline int get_dport(void *trans_data, void *data_end,
49 				     __u8 protocol)
50 {
51 	struct tcphdr *th;
52 	struct udphdr *uh;
53 
54 	switch (protocol) {
55 	case IPPROTO_TCP:
56 		th = (struct tcphdr *)trans_data;
57 		if (th + 1 > data_end)
58 			return -1;
59 		return th->dest;
60 	case IPPROTO_UDP:
61 		uh = (struct udphdr *)trans_data;
62 		if (uh + 1 > data_end)
63 			return -1;
64 		return uh->dest;
65 	default:
66 		return 0;
67 	}
68 }
69 
70 static __always_inline void set_ethhdr(struct ethhdr *new_eth,
71 				       const struct ethhdr *old_eth,
72 				       const struct iptnl_info *tnl,
73 				       __be16 h_proto)
74 {
75 	memcpy(new_eth->h_source, old_eth->h_dest, sizeof(new_eth->h_source));
76 	memcpy(new_eth->h_dest, tnl->dmac, sizeof(new_eth->h_dest));
77 	new_eth->h_proto = h_proto;
78 }
79 
80 static __always_inline int handle_ipv4(struct xdp_md *xdp)
81 {
82 	void *data_end = (void *)(long)xdp->data_end;
83 	void *data = (void *)(long)xdp->data;
84 	struct iptnl_info *tnl;
85 	struct ethhdr *new_eth;
86 	struct ethhdr *old_eth;
87 	struct iphdr *iph = data + sizeof(struct ethhdr);
88 	__u16 *next_iph;
89 	__u16 payload_len;
90 	struct vip vip = {};
91 	int dport;
92 	__u32 csum = 0;
93 	int i;
94 
95 	if (iph + 1 > data_end)
96 		return XDP_DROP;
97 
98 	dport = get_dport(iph + 1, data_end, iph->protocol);
99 	if (dport == -1)
100 		return XDP_DROP;
101 
102 	vip.protocol = iph->protocol;
103 	vip.family = AF_INET;
104 	vip.daddr.v4 = iph->daddr;
105 	vip.dport = dport;
106 	payload_len = bpf_ntohs(iph->tot_len);
107 
108 	tnl = bpf_map_lookup_elem(&vip2tnl, &vip);
109 	/* It only does v4-in-v4 */
110 	if (!tnl || tnl->family != AF_INET)
111 		return XDP_PASS;
112 
113 	if (bpf_xdp_adjust_head(xdp, 0 - (int)sizeof(struct iphdr)))
114 		return XDP_DROP;
115 
116 	data = (void *)(long)xdp->data;
117 	data_end = (void *)(long)xdp->data_end;
118 
119 	new_eth = data;
120 	iph = data + sizeof(*new_eth);
121 	old_eth = data + sizeof(*iph);
122 
123 	if (new_eth + 1 > data_end ||
124 	    old_eth + 1 > data_end ||
125 	    iph + 1 > data_end)
126 		return XDP_DROP;
127 
128 	set_ethhdr(new_eth, old_eth, tnl, bpf_htons(ETH_P_IP));
129 
130 	iph->version = 4;
131 	iph->ihl = sizeof(*iph) >> 2;
132 	iph->frag_off =	0;
133 	iph->protocol = IPPROTO_IPIP;
134 	iph->check = 0;
135 	iph->tos = 0;
136 	iph->tot_len = bpf_htons(payload_len + sizeof(*iph));
137 	iph->daddr = tnl->daddr.v4;
138 	iph->saddr = tnl->saddr.v4;
139 	iph->ttl = 8;
140 
141 	next_iph = (__u16 *)iph;
142 #pragma clang loop unroll(full)
143 	for (i = 0; i < sizeof(*iph) >> 1; i++)
144 		csum += *next_iph++;
145 
146 	iph->check = ~((csum & 0xffff) + (csum >> 16));
147 
148 	count_tx(vip.protocol);
149 
150 	return XDP_TX;
151 }
152 
153 static __always_inline int handle_ipv6(struct xdp_md *xdp)
154 {
155 	void *data_end = (void *)(long)xdp->data_end;
156 	void *data = (void *)(long)xdp->data;
157 	struct iptnl_info *tnl;
158 	struct ethhdr *new_eth;
159 	struct ethhdr *old_eth;
160 	struct ipv6hdr *ip6h = data + sizeof(struct ethhdr);
161 	__u16 payload_len;
162 	struct vip vip = {};
163 	int dport;
164 
165 	if (ip6h + 1 > data_end)
166 		return XDP_DROP;
167 
168 	dport = get_dport(ip6h + 1, data_end, ip6h->nexthdr);
169 	if (dport == -1)
170 		return XDP_DROP;
171 
172 	vip.protocol = ip6h->nexthdr;
173 	vip.family = AF_INET6;
174 	memcpy(vip.daddr.v6, ip6h->daddr.s6_addr32, sizeof(vip.daddr));
175 	vip.dport = dport;
176 	payload_len = ip6h->payload_len;
177 
178 	tnl = bpf_map_lookup_elem(&vip2tnl, &vip);
179 	/* It only does v6-in-v6 */
180 	if (!tnl || tnl->family != AF_INET6)
181 		return XDP_PASS;
182 
183 	if (bpf_xdp_adjust_head(xdp, 0 - (int)sizeof(struct ipv6hdr)))
184 		return XDP_DROP;
185 
186 	data = (void *)(long)xdp->data;
187 	data_end = (void *)(long)xdp->data_end;
188 
189 	new_eth = data;
190 	ip6h = data + sizeof(*new_eth);
191 	old_eth = data + sizeof(*ip6h);
192 
193 	if (new_eth + 1 > data_end || old_eth + 1 > data_end ||
194 	    ip6h + 1 > data_end)
195 		return XDP_DROP;
196 
197 	set_ethhdr(new_eth, old_eth, tnl, bpf_htons(ETH_P_IPV6));
198 
199 	ip6h->version = 6;
200 	ip6h->priority = 0;
201 	memset(ip6h->flow_lbl, 0, sizeof(ip6h->flow_lbl));
202 	ip6h->payload_len = bpf_htons(bpf_ntohs(payload_len) + sizeof(*ip6h));
203 	ip6h->nexthdr = IPPROTO_IPV6;
204 	ip6h->hop_limit = 8;
205 	memcpy(ip6h->saddr.s6_addr32, tnl->saddr.v6, sizeof(tnl->saddr.v6));
206 	memcpy(ip6h->daddr.s6_addr32, tnl->daddr.v6, sizeof(tnl->daddr.v6));
207 
208 	count_tx(vip.protocol);
209 
210 	return XDP_TX;
211 }
212 
213 SEC("xdp_tx_iptunnel")
214 int _xdp_tx_iptunnel(struct xdp_md *xdp)
215 {
216 	void *data_end = (void *)(long)xdp->data_end;
217 	void *data = (void *)(long)xdp->data;
218 	struct ethhdr *eth = data;
219 	__u16 h_proto;
220 
221 	if (eth + 1 > data_end)
222 		return XDP_DROP;
223 
224 	h_proto = eth->h_proto;
225 
226 	if (h_proto == bpf_htons(ETH_P_IP))
227 		return handle_ipv4(xdp);
228 	else if (h_proto == bpf_htons(ETH_P_IPV6))
229 
230 		return handle_ipv6(xdp);
231 	else
232 		return XDP_DROP;
233 }
234 
235 char _license[] SEC("license") = "GPL";
236