1 /* 2 * ip_vs_proto.c: transport protocol load balancing support for IPVS 3 * 4 * Authors: Wensong Zhang <wensong@linuxvirtualserver.org> 5 * Julian Anastasov <ja@ssi.bg> 6 * 7 * This program is free software; you can redistribute it and/or 8 * modify it under the terms of the GNU General Public License 9 * as published by the Free Software Foundation; either version 10 * 2 of the License, or (at your option) any later version. 11 * 12 * Changes: 13 * 14 */ 15 16 #define KMSG_COMPONENT "IPVS" 17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt 18 19 #include <linux/module.h> 20 #include <linux/kernel.h> 21 #include <linux/skbuff.h> 22 #include <linux/gfp.h> 23 #include <linux/in.h> 24 #include <linux/ip.h> 25 #include <net/protocol.h> 26 #include <net/tcp.h> 27 #include <net/udp.h> 28 #include <linux/stat.h> 29 #include <linux/proc_fs.h> 30 31 #include <net/ip_vs.h> 32 33 34 /* 35 * IPVS protocols can only be registered/unregistered when the ipvs 36 * module is loaded/unloaded, so no lock is needed in accessing the 37 * ipvs protocol table. 38 */ 39 40 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */ 41 #define IP_VS_PROTO_HASH(proto) ((proto) & (IP_VS_PROTO_TAB_SIZE-1)) 42 43 static struct ip_vs_protocol *ip_vs_proto_table[IP_VS_PROTO_TAB_SIZE]; 44 45 46 /* 47 * register an ipvs protocol 48 */ 49 static int __used __init register_ip_vs_protocol(struct ip_vs_protocol *pp) 50 { 51 unsigned int hash = IP_VS_PROTO_HASH(pp->protocol); 52 53 pp->next = ip_vs_proto_table[hash]; 54 ip_vs_proto_table[hash] = pp; 55 56 if (pp->init != NULL) 57 pp->init(pp); 58 59 return 0; 60 } 61 62 /* 63 * register an ipvs protocols netns related data 64 */ 65 static int 66 register_ip_vs_proto_netns(struct netns_ipvs *ipvs, struct ip_vs_protocol *pp) 67 { 68 unsigned int hash = IP_VS_PROTO_HASH(pp->protocol); 69 struct ip_vs_proto_data *pd = 70 kzalloc(sizeof(struct ip_vs_proto_data), GFP_KERNEL); 71 72 if (!pd) 73 return -ENOMEM; 74 75 pd->pp = pp; /* For speed issues */ 76 pd->next = ipvs->proto_data_table[hash]; 77 ipvs->proto_data_table[hash] = pd; 78 atomic_set(&pd->appcnt, 0); /* Init app counter */ 79 80 if (pp->init_netns != NULL) { 81 int ret = pp->init_netns(ipvs, pd); 82 if (ret) { 83 /* unlink an free proto data */ 84 ipvs->proto_data_table[hash] = pd->next; 85 kfree(pd); 86 return ret; 87 } 88 } 89 90 return 0; 91 } 92 93 /* 94 * unregister an ipvs protocol 95 */ 96 static int unregister_ip_vs_protocol(struct ip_vs_protocol *pp) 97 { 98 struct ip_vs_protocol **pp_p; 99 unsigned int hash = IP_VS_PROTO_HASH(pp->protocol); 100 101 pp_p = &ip_vs_proto_table[hash]; 102 for (; *pp_p; pp_p = &(*pp_p)->next) { 103 if (*pp_p == pp) { 104 *pp_p = pp->next; 105 if (pp->exit != NULL) 106 pp->exit(pp); 107 return 0; 108 } 109 } 110 111 return -ESRCH; 112 } 113 114 /* 115 * unregister an ipvs protocols netns data 116 */ 117 static int 118 unregister_ip_vs_proto_netns(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd) 119 { 120 struct ip_vs_proto_data **pd_p; 121 unsigned int hash = IP_VS_PROTO_HASH(pd->pp->protocol); 122 123 pd_p = &ipvs->proto_data_table[hash]; 124 for (; *pd_p; pd_p = &(*pd_p)->next) { 125 if (*pd_p == pd) { 126 *pd_p = pd->next; 127 if (pd->pp->exit_netns != NULL) 128 pd->pp->exit_netns(ipvs, pd); 129 kfree(pd); 130 return 0; 131 } 132 } 133 134 return -ESRCH; 135 } 136 137 /* 138 * get ip_vs_protocol object by its proto. 139 */ 140 struct ip_vs_protocol * ip_vs_proto_get(unsigned short proto) 141 { 142 struct ip_vs_protocol *pp; 143 unsigned int hash = IP_VS_PROTO_HASH(proto); 144 145 for (pp = ip_vs_proto_table[hash]; pp; pp = pp->next) { 146 if (pp->protocol == proto) 147 return pp; 148 } 149 150 return NULL; 151 } 152 EXPORT_SYMBOL(ip_vs_proto_get); 153 154 /* 155 * get ip_vs_protocol object data by netns and proto 156 */ 157 struct ip_vs_proto_data * 158 ip_vs_proto_data_get(struct netns_ipvs *ipvs, unsigned short proto) 159 { 160 struct ip_vs_proto_data *pd; 161 unsigned int hash = IP_VS_PROTO_HASH(proto); 162 163 for (pd = ipvs->proto_data_table[hash]; pd; pd = pd->next) { 164 if (pd->pp->protocol == proto) 165 return pd; 166 } 167 168 return NULL; 169 } 170 EXPORT_SYMBOL(ip_vs_proto_data_get); 171 172 /* 173 * Propagate event for state change to all protocols 174 */ 175 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags) 176 { 177 struct ip_vs_proto_data *pd; 178 int i; 179 180 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) { 181 for (pd = ipvs->proto_data_table[i]; pd; pd = pd->next) { 182 if (pd->pp->timeout_change) 183 pd->pp->timeout_change(pd, flags); 184 } 185 } 186 } 187 188 189 int * 190 ip_vs_create_timeout_table(int *table, int size) 191 { 192 return kmemdup(table, size, GFP_KERNEL); 193 } 194 195 196 /* 197 * Set timeout value for state specified by name 198 */ 199 int 200 ip_vs_set_state_timeout(int *table, int num, const char *const *names, 201 const char *name, int to) 202 { 203 int i; 204 205 if (!table || !name || !to) 206 return -EINVAL; 207 208 for (i = 0; i < num; i++) { 209 if (strcmp(names[i], name)) 210 continue; 211 table[i] = to * HZ; 212 return 0; 213 } 214 return -ENOENT; 215 } 216 217 218 const char * ip_vs_state_name(__u16 proto, int state) 219 { 220 struct ip_vs_protocol *pp = ip_vs_proto_get(proto); 221 222 if (pp == NULL || pp->state_name == NULL) 223 return (IPPROTO_IP == proto) ? "NONE" : "ERR!"; 224 return pp->state_name(state); 225 } 226 227 228 static void 229 ip_vs_tcpudp_debug_packet_v4(struct ip_vs_protocol *pp, 230 const struct sk_buff *skb, 231 int offset, 232 const char *msg) 233 { 234 char buf[128]; 235 struct iphdr _iph, *ih; 236 237 ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph); 238 if (ih == NULL) 239 sprintf(buf, "TRUNCATED"); 240 else if (ih->frag_off & htons(IP_OFFSET)) 241 sprintf(buf, "%pI4->%pI4 frag", &ih->saddr, &ih->daddr); 242 else { 243 __be16 _ports[2], *pptr; 244 245 pptr = skb_header_pointer(skb, offset + ih->ihl*4, 246 sizeof(_ports), _ports); 247 if (pptr == NULL) 248 sprintf(buf, "TRUNCATED %pI4->%pI4", 249 &ih->saddr, &ih->daddr); 250 else 251 sprintf(buf, "%pI4:%u->%pI4:%u", 252 &ih->saddr, ntohs(pptr[0]), 253 &ih->daddr, ntohs(pptr[1])); 254 } 255 256 pr_debug("%s: %s %s\n", msg, pp->name, buf); 257 } 258 259 #ifdef CONFIG_IP_VS_IPV6 260 static void 261 ip_vs_tcpudp_debug_packet_v6(struct ip_vs_protocol *pp, 262 const struct sk_buff *skb, 263 int offset, 264 const char *msg) 265 { 266 char buf[192]; 267 struct ipv6hdr _iph, *ih; 268 269 ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph); 270 if (ih == NULL) 271 sprintf(buf, "TRUNCATED"); 272 else if (ih->nexthdr == IPPROTO_FRAGMENT) 273 sprintf(buf, "%pI6c->%pI6c frag", &ih->saddr, &ih->daddr); 274 else { 275 __be16 _ports[2], *pptr; 276 277 pptr = skb_header_pointer(skb, offset + sizeof(struct ipv6hdr), 278 sizeof(_ports), _ports); 279 if (pptr == NULL) 280 sprintf(buf, "TRUNCATED %pI6c->%pI6c", 281 &ih->saddr, &ih->daddr); 282 else 283 sprintf(buf, "%pI6c:%u->%pI6c:%u", 284 &ih->saddr, ntohs(pptr[0]), 285 &ih->daddr, ntohs(pptr[1])); 286 } 287 288 pr_debug("%s: %s %s\n", msg, pp->name, buf); 289 } 290 #endif 291 292 293 void 294 ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp, 295 const struct sk_buff *skb, 296 int offset, 297 const char *msg) 298 { 299 #ifdef CONFIG_IP_VS_IPV6 300 if (af == AF_INET6) 301 ip_vs_tcpudp_debug_packet_v6(pp, skb, offset, msg); 302 else 303 #endif 304 ip_vs_tcpudp_debug_packet_v4(pp, skb, offset, msg); 305 } 306 307 /* 308 * per network name-space init 309 */ 310 int __net_init ip_vs_protocol_net_init(struct netns_ipvs *ipvs) 311 { 312 int i, ret; 313 static struct ip_vs_protocol *protos[] = { 314 #ifdef CONFIG_IP_VS_PROTO_TCP 315 &ip_vs_protocol_tcp, 316 #endif 317 #ifdef CONFIG_IP_VS_PROTO_UDP 318 &ip_vs_protocol_udp, 319 #endif 320 #ifdef CONFIG_IP_VS_PROTO_SCTP 321 &ip_vs_protocol_sctp, 322 #endif 323 #ifdef CONFIG_IP_VS_PROTO_AH 324 &ip_vs_protocol_ah, 325 #endif 326 #ifdef CONFIG_IP_VS_PROTO_ESP 327 &ip_vs_protocol_esp, 328 #endif 329 }; 330 331 for (i = 0; i < ARRAY_SIZE(protos); i++) { 332 ret = register_ip_vs_proto_netns(ipvs, protos[i]); 333 if (ret < 0) 334 goto cleanup; 335 } 336 return 0; 337 338 cleanup: 339 ip_vs_protocol_net_cleanup(ipvs); 340 return ret; 341 } 342 343 void __net_exit ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs) 344 { 345 struct ip_vs_proto_data *pd; 346 int i; 347 348 /* unregister all the ipvs proto data for this netns */ 349 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) { 350 while ((pd = ipvs->proto_data_table[i]) != NULL) 351 unregister_ip_vs_proto_netns(ipvs, pd); 352 } 353 } 354 355 int __init ip_vs_protocol_init(void) 356 { 357 char protocols[64]; 358 #define REGISTER_PROTOCOL(p) \ 359 do { \ 360 register_ip_vs_protocol(p); \ 361 strcat(protocols, ", "); \ 362 strcat(protocols, (p)->name); \ 363 } while (0) 364 365 protocols[0] = '\0'; 366 protocols[2] = '\0'; 367 #ifdef CONFIG_IP_VS_PROTO_TCP 368 REGISTER_PROTOCOL(&ip_vs_protocol_tcp); 369 #endif 370 #ifdef CONFIG_IP_VS_PROTO_UDP 371 REGISTER_PROTOCOL(&ip_vs_protocol_udp); 372 #endif 373 #ifdef CONFIG_IP_VS_PROTO_SCTP 374 REGISTER_PROTOCOL(&ip_vs_protocol_sctp); 375 #endif 376 #ifdef CONFIG_IP_VS_PROTO_AH 377 REGISTER_PROTOCOL(&ip_vs_protocol_ah); 378 #endif 379 #ifdef CONFIG_IP_VS_PROTO_ESP 380 REGISTER_PROTOCOL(&ip_vs_protocol_esp); 381 #endif 382 pr_info("Registered protocols (%s)\n", &protocols[2]); 383 384 return 0; 385 } 386 387 388 void ip_vs_protocol_cleanup(void) 389 { 390 struct ip_vs_protocol *pp; 391 int i; 392 393 /* unregister all the ipvs protocols */ 394 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) { 395 while ((pp = ip_vs_proto_table[i]) != NULL) 396 unregister_ip_vs_protocol(pp); 397 } 398 } 399