1 /* Copyright (c) 2016 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 <linux/bpf.h> 8 #include <linux/if_link.h> 9 #include <assert.h> 10 #include <errno.h> 11 #include <signal.h> 12 #include <stdio.h> 13 #include <stdlib.h> 14 #include <string.h> 15 #include <sys/resource.h> 16 #include <arpa/inet.h> 17 #include <netinet/ether.h> 18 #include <unistd.h> 19 #include <time.h> 20 #include "bpf/libbpf.h" 21 #include <bpf/bpf.h> 22 #include "bpf_util.h" 23 #include "xdp_tx_iptunnel_common.h" 24 25 #define STATS_INTERVAL_S 2U 26 27 static int ifindex = -1; 28 static __u32 xdp_flags = XDP_FLAGS_UPDATE_IF_NOEXIST; 29 static int rxcnt_map_fd; 30 static __u32 prog_id; 31 32 static void int_exit(int sig) 33 { 34 __u32 curr_prog_id = 0; 35 36 if (ifindex > -1) { 37 if (bpf_get_link_xdp_id(ifindex, &curr_prog_id, xdp_flags)) { 38 printf("bpf_get_link_xdp_id failed\n"); 39 exit(1); 40 } 41 if (prog_id == curr_prog_id) 42 bpf_set_link_xdp_fd(ifindex, -1, xdp_flags); 43 else if (!curr_prog_id) 44 printf("couldn't find a prog id on a given iface\n"); 45 else 46 printf("program on interface changed, not removing\n"); 47 } 48 exit(0); 49 } 50 51 /* simple per-protocol drop counter 52 */ 53 static void poll_stats(unsigned int kill_after_s) 54 { 55 const unsigned int nr_protos = 256; 56 unsigned int nr_cpus = bpf_num_possible_cpus(); 57 time_t started_at = time(NULL); 58 __u64 values[nr_cpus], prev[nr_protos][nr_cpus]; 59 __u32 proto; 60 int i; 61 62 memset(prev, 0, sizeof(prev)); 63 64 while (!kill_after_s || time(NULL) - started_at <= kill_after_s) { 65 sleep(STATS_INTERVAL_S); 66 67 for (proto = 0; proto < nr_protos; proto++) { 68 __u64 sum = 0; 69 70 assert(bpf_map_lookup_elem(rxcnt_map_fd, &proto, 71 values) == 0); 72 for (i = 0; i < nr_cpus; i++) 73 sum += (values[i] - prev[proto][i]); 74 75 if (sum) 76 printf("proto %u: sum:%10llu pkts, rate:%10llu pkts/s\n", 77 proto, sum, sum / STATS_INTERVAL_S); 78 memcpy(prev[proto], values, sizeof(values)); 79 } 80 } 81 } 82 83 static void usage(const char *cmd) 84 { 85 printf("Start a XDP prog which encapsulates incoming packets\n" 86 "in an IPv4/v6 header and XDP_TX it out. The dst <VIP:PORT>\n" 87 "is used to select packets to encapsulate\n\n"); 88 printf("Usage: %s [...]\n", cmd); 89 printf(" -i <ifindex> Interface Index\n"); 90 printf(" -a <vip-service-address> IPv4 or IPv6\n"); 91 printf(" -p <vip-service-port> A port range (e.g. 433-444) is also allowed\n"); 92 printf(" -s <source-ip> Used in the IPTunnel header\n"); 93 printf(" -d <dest-ip> Used in the IPTunnel header\n"); 94 printf(" -m <dest-MAC> Used in sending the IP Tunneled pkt\n"); 95 printf(" -T <stop-after-X-seconds> Default: 0 (forever)\n"); 96 printf(" -P <IP-Protocol> Default is TCP\n"); 97 printf(" -S use skb-mode\n"); 98 printf(" -N enforce native mode\n"); 99 printf(" -F Force loading the XDP prog\n"); 100 printf(" -h Display this help\n"); 101 } 102 103 static int parse_ipstr(const char *ipstr, unsigned int *addr) 104 { 105 if (inet_pton(AF_INET6, ipstr, addr) == 1) { 106 return AF_INET6; 107 } else if (inet_pton(AF_INET, ipstr, addr) == 1) { 108 addr[1] = addr[2] = addr[3] = 0; 109 return AF_INET; 110 } 111 112 fprintf(stderr, "%s is an invalid IP\n", ipstr); 113 return AF_UNSPEC; 114 } 115 116 static int parse_ports(const char *port_str, int *min_port, int *max_port) 117 { 118 char *end; 119 long tmp_min_port; 120 long tmp_max_port; 121 122 tmp_min_port = strtol(optarg, &end, 10); 123 if (tmp_min_port < 1 || tmp_min_port > 65535) { 124 fprintf(stderr, "Invalid port(s):%s\n", optarg); 125 return 1; 126 } 127 128 if (*end == '-') { 129 end++; 130 tmp_max_port = strtol(end, NULL, 10); 131 if (tmp_max_port < 1 || tmp_max_port > 65535) { 132 fprintf(stderr, "Invalid port(s):%s\n", optarg); 133 return 1; 134 } 135 } else { 136 tmp_max_port = tmp_min_port; 137 } 138 139 if (tmp_min_port > tmp_max_port) { 140 fprintf(stderr, "Invalid port(s):%s\n", optarg); 141 return 1; 142 } 143 144 if (tmp_max_port - tmp_min_port + 1 > MAX_IPTNL_ENTRIES) { 145 fprintf(stderr, "Port range (%s) is larger than %u\n", 146 port_str, MAX_IPTNL_ENTRIES); 147 return 1; 148 } 149 *min_port = tmp_min_port; 150 *max_port = tmp_max_port; 151 152 return 0; 153 } 154 155 int main(int argc, char **argv) 156 { 157 struct bpf_prog_load_attr prog_load_attr = { 158 .prog_type = BPF_PROG_TYPE_XDP, 159 }; 160 struct rlimit r = {RLIM_INFINITY, RLIM_INFINITY}; 161 int min_port = 0, max_port = 0, vip2tnl_map_fd; 162 const char *optstr = "i:a:p:s:d:m:T:P:FSNh"; 163 unsigned char opt_flags[256] = {}; 164 struct bpf_prog_info info = {}; 165 __u32 info_len = sizeof(info); 166 unsigned int kill_after_s = 0; 167 struct iptnl_info tnl = {}; 168 struct bpf_object *obj; 169 struct vip vip = {}; 170 char filename[256]; 171 int opt, prog_fd; 172 int i, err; 173 174 tnl.family = AF_UNSPEC; 175 vip.protocol = IPPROTO_TCP; 176 177 for (i = 0; i < strlen(optstr); i++) 178 if (optstr[i] != 'h' && 'a' <= optstr[i] && optstr[i] <= 'z') 179 opt_flags[(unsigned char)optstr[i]] = 1; 180 181 while ((opt = getopt(argc, argv, optstr)) != -1) { 182 unsigned short family; 183 unsigned int *v6; 184 185 switch (opt) { 186 case 'i': 187 ifindex = atoi(optarg); 188 break; 189 case 'a': 190 vip.family = parse_ipstr(optarg, vip.daddr.v6); 191 if (vip.family == AF_UNSPEC) 192 return 1; 193 break; 194 case 'p': 195 if (parse_ports(optarg, &min_port, &max_port)) 196 return 1; 197 break; 198 case 'P': 199 vip.protocol = atoi(optarg); 200 break; 201 case 's': 202 case 'd': 203 if (opt == 's') 204 v6 = tnl.saddr.v6; 205 else 206 v6 = tnl.daddr.v6; 207 208 family = parse_ipstr(optarg, v6); 209 if (family == AF_UNSPEC) 210 return 1; 211 if (tnl.family == AF_UNSPEC) { 212 tnl.family = family; 213 } else if (tnl.family != family) { 214 fprintf(stderr, 215 "The IP version of the src and dst addresses used in the IP encapsulation does not match\n"); 216 return 1; 217 } 218 break; 219 case 'm': 220 if (!ether_aton_r(optarg, 221 (struct ether_addr *)tnl.dmac)) { 222 fprintf(stderr, "Invalid mac address:%s\n", 223 optarg); 224 return 1; 225 } 226 break; 227 case 'T': 228 kill_after_s = atoi(optarg); 229 break; 230 case 'S': 231 xdp_flags |= XDP_FLAGS_SKB_MODE; 232 break; 233 case 'N': 234 xdp_flags |= XDP_FLAGS_DRV_MODE; 235 break; 236 case 'F': 237 xdp_flags &= ~XDP_FLAGS_UPDATE_IF_NOEXIST; 238 break; 239 default: 240 usage(argv[0]); 241 return 1; 242 } 243 opt_flags[opt] = 0; 244 } 245 246 for (i = 0; i < strlen(optstr); i++) { 247 if (opt_flags[(unsigned int)optstr[i]]) { 248 fprintf(stderr, "Missing argument -%c\n", optstr[i]); 249 usage(argv[0]); 250 return 1; 251 } 252 } 253 254 if (setrlimit(RLIMIT_MEMLOCK, &r)) { 255 perror("setrlimit(RLIMIT_MEMLOCK, RLIM_INFINITY)"); 256 return 1; 257 } 258 259 snprintf(filename, sizeof(filename), "%s_kern.o", argv[0]); 260 prog_load_attr.file = filename; 261 262 if (bpf_prog_load_xattr(&prog_load_attr, &obj, &prog_fd)) 263 return 1; 264 265 if (!prog_fd) { 266 printf("load_bpf_file: %s\n", strerror(errno)); 267 return 1; 268 } 269 270 rxcnt_map_fd = bpf_object__find_map_fd_by_name(obj, "rxcnt"); 271 vip2tnl_map_fd = bpf_object__find_map_fd_by_name(obj, "vip2tnl"); 272 if (vip2tnl_map_fd < 0 || rxcnt_map_fd < 0) { 273 printf("bpf_object__find_map_fd_by_name failed\n"); 274 return 1; 275 } 276 277 signal(SIGINT, int_exit); 278 signal(SIGTERM, int_exit); 279 280 while (min_port <= max_port) { 281 vip.dport = htons(min_port++); 282 if (bpf_map_update_elem(vip2tnl_map_fd, &vip, &tnl, 283 BPF_NOEXIST)) { 284 perror("bpf_map_update_elem(&vip2tnl)"); 285 return 1; 286 } 287 } 288 289 if (bpf_set_link_xdp_fd(ifindex, prog_fd, xdp_flags) < 0) { 290 printf("link set xdp fd failed\n"); 291 return 1; 292 } 293 294 err = bpf_obj_get_info_by_fd(prog_fd, &info, &info_len); 295 if (err) { 296 printf("can't get prog info - %s\n", strerror(errno)); 297 return err; 298 } 299 prog_id = info.id; 300 301 poll_stats(kill_after_s); 302 303 bpf_set_link_xdp_fd(ifindex, -1, xdp_flags); 304 305 return 0; 306 } 307