xref: /linux/tools/testing/selftests/bpf/progs/test_xdp.c (revision a4eb44a6435d6d8f9e642407a4a06f65eb90ca04)
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/bpf_helpers.h>
20 #include <bpf/bpf_endian.h>
21 #include "test_iptunnel_common.h"
22 
23 struct {
24 	__uint(type, BPF_MAP_TYPE_PERCPU_ARRAY);
25 	__uint(max_entries, 256);
26 	__type(key, __u32);
27 	__type(value, __u64);
28 } rxcnt SEC(".maps");
29 
30 struct {
31 	__uint(type, BPF_MAP_TYPE_HASH);
32 	__uint(max_entries, MAX_IPTNL_ENTRIES);
33 	__type(key, struct vip);
34 	__type(value, struct iptnl_info);
35 } vip2tnl SEC(".maps");
36 
37 static __always_inline void count_tx(__u32 protocol)
38 {
39 	__u64 *rxcnt_count;
40 
41 	rxcnt_count = bpf_map_lookup_elem(&rxcnt, &protocol);
42 	if (rxcnt_count)
43 		*rxcnt_count += 1;
44 }
45 
46 static __always_inline int get_dport(void *trans_data, void *data_end,
47 				     __u8 protocol)
48 {
49 	struct tcphdr *th;
50 	struct udphdr *uh;
51 
52 	switch (protocol) {
53 	case IPPROTO_TCP:
54 		th = (struct tcphdr *)trans_data;
55 		if (th + 1 > data_end)
56 			return -1;
57 		return th->dest;
58 	case IPPROTO_UDP:
59 		uh = (struct udphdr *)trans_data;
60 		if (uh + 1 > data_end)
61 			return -1;
62 		return uh->dest;
63 	default:
64 		return 0;
65 	}
66 }
67 
68 static __always_inline void set_ethhdr(struct ethhdr *new_eth,
69 				       const struct ethhdr *old_eth,
70 				       const struct iptnl_info *tnl,
71 				       __be16 h_proto)
72 {
73 	memcpy(new_eth->h_source, old_eth->h_dest, sizeof(new_eth->h_source));
74 	memcpy(new_eth->h_dest, tnl->dmac, sizeof(new_eth->h_dest));
75 	new_eth->h_proto = h_proto;
76 }
77 
78 static __always_inline int handle_ipv4(struct xdp_md *xdp)
79 {
80 	void *data_end = (void *)(long)xdp->data_end;
81 	void *data = (void *)(long)xdp->data;
82 	struct iptnl_info *tnl;
83 	struct ethhdr *new_eth;
84 	struct ethhdr *old_eth;
85 	struct iphdr *iph = data + sizeof(struct ethhdr);
86 	__u16 *next_iph;
87 	__u16 payload_len;
88 	struct vip vip = {};
89 	int dport;
90 	__u32 csum = 0;
91 	int i;
92 
93 	if (iph + 1 > data_end)
94 		return XDP_DROP;
95 
96 	dport = get_dport(iph + 1, data_end, iph->protocol);
97 	if (dport == -1)
98 		return XDP_DROP;
99 
100 	vip.protocol = iph->protocol;
101 	vip.family = AF_INET;
102 	vip.daddr.v4 = iph->daddr;
103 	vip.dport = dport;
104 	payload_len = bpf_ntohs(iph->tot_len);
105 
106 	tnl = bpf_map_lookup_elem(&vip2tnl, &vip);
107 	/* It only does v4-in-v4 */
108 	if (!tnl || tnl->family != AF_INET)
109 		return XDP_PASS;
110 
111 	if (bpf_xdp_adjust_head(xdp, 0 - (int)sizeof(struct iphdr)))
112 		return XDP_DROP;
113 
114 	data = (void *)(long)xdp->data;
115 	data_end = (void *)(long)xdp->data_end;
116 
117 	new_eth = data;
118 	iph = data + sizeof(*new_eth);
119 	old_eth = data + sizeof(*iph);
120 
121 	if (new_eth + 1 > data_end ||
122 	    old_eth + 1 > data_end ||
123 	    iph + 1 > data_end)
124 		return XDP_DROP;
125 
126 	set_ethhdr(new_eth, old_eth, tnl, bpf_htons(ETH_P_IP));
127 
128 	iph->version = 4;
129 	iph->ihl = sizeof(*iph) >> 2;
130 	iph->frag_off =	0;
131 	iph->protocol = IPPROTO_IPIP;
132 	iph->check = 0;
133 	iph->tos = 0;
134 	iph->tot_len = bpf_htons(payload_len + sizeof(*iph));
135 	iph->daddr = tnl->daddr.v4;
136 	iph->saddr = tnl->saddr.v4;
137 	iph->ttl = 8;
138 
139 	next_iph = (__u16 *)iph;
140 #pragma clang loop unroll(full)
141 	for (i = 0; i < sizeof(*iph) >> 1; i++)
142 		csum += *next_iph++;
143 
144 	iph->check = ~((csum & 0xffff) + (csum >> 16));
145 
146 	count_tx(vip.protocol);
147 
148 	return XDP_TX;
149 }
150 
151 static __always_inline int handle_ipv6(struct xdp_md *xdp)
152 {
153 	void *data_end = (void *)(long)xdp->data_end;
154 	void *data = (void *)(long)xdp->data;
155 	struct iptnl_info *tnl;
156 	struct ethhdr *new_eth;
157 	struct ethhdr *old_eth;
158 	struct ipv6hdr *ip6h = data + sizeof(struct ethhdr);
159 	__u16 payload_len;
160 	struct vip vip = {};
161 	int dport;
162 
163 	if (ip6h + 1 > data_end)
164 		return XDP_DROP;
165 
166 	dport = get_dport(ip6h + 1, data_end, ip6h->nexthdr);
167 	if (dport == -1)
168 		return XDP_DROP;
169 
170 	vip.protocol = ip6h->nexthdr;
171 	vip.family = AF_INET6;
172 	memcpy(vip.daddr.v6, ip6h->daddr.s6_addr32, sizeof(vip.daddr));
173 	vip.dport = dport;
174 	payload_len = ip6h->payload_len;
175 
176 	tnl = bpf_map_lookup_elem(&vip2tnl, &vip);
177 	/* It only does v6-in-v6 */
178 	if (!tnl || tnl->family != AF_INET6)
179 		return XDP_PASS;
180 
181 	if (bpf_xdp_adjust_head(xdp, 0 - (int)sizeof(struct ipv6hdr)))
182 		return XDP_DROP;
183 
184 	data = (void *)(long)xdp->data;
185 	data_end = (void *)(long)xdp->data_end;
186 
187 	new_eth = data;
188 	ip6h = data + sizeof(*new_eth);
189 	old_eth = data + sizeof(*ip6h);
190 
191 	if (new_eth + 1 > data_end || old_eth + 1 > data_end ||
192 	    ip6h + 1 > data_end)
193 		return XDP_DROP;
194 
195 	set_ethhdr(new_eth, old_eth, tnl, bpf_htons(ETH_P_IPV6));
196 
197 	ip6h->version = 6;
198 	ip6h->priority = 0;
199 	memset(ip6h->flow_lbl, 0, sizeof(ip6h->flow_lbl));
200 	ip6h->payload_len = bpf_htons(bpf_ntohs(payload_len) + sizeof(*ip6h));
201 	ip6h->nexthdr = IPPROTO_IPV6;
202 	ip6h->hop_limit = 8;
203 	memcpy(ip6h->saddr.s6_addr32, tnl->saddr.v6, sizeof(tnl->saddr.v6));
204 	memcpy(ip6h->daddr.s6_addr32, tnl->daddr.v6, sizeof(tnl->daddr.v6));
205 
206 	count_tx(vip.protocol);
207 
208 	return XDP_TX;
209 }
210 
211 SEC("xdp")
212 int _xdp_tx_iptunnel(struct xdp_md *xdp)
213 {
214 	void *data_end = (void *)(long)xdp->data_end;
215 	void *data = (void *)(long)xdp->data;
216 	struct ethhdr *eth = data;
217 	__u16 h_proto;
218 
219 	if (eth + 1 > data_end)
220 		return XDP_DROP;
221 
222 	h_proto = eth->h_proto;
223 
224 	if (h_proto == bpf_htons(ETH_P_IP))
225 		return handle_ipv4(xdp);
226 	else if (h_proto == bpf_htons(ETH_P_IPV6))
227 
228 		return handle_ipv6(xdp);
229 	else
230 		return XDP_DROP;
231 }
232 
233 char _license[] SEC("license") = "GPL";
234