1 /* 2 * IPVS: Locality-Based Least-Connection scheduling module 3 * 4 * Authors: Wensong Zhang <wensong@gnuchina.org> 5 * 6 * This program is free software; you can redistribute it and/or 7 * modify it under the terms of the GNU General Public License 8 * as published by the Free Software Foundation; either version 9 * 2 of the License, or (at your option) any later version. 10 * 11 * Changes: 12 * Martin Hamilton : fixed the terrible locking bugs 13 * *lock(tbl->lock) ==> *lock(&tbl->lock) 14 * Wensong Zhang : fixed the uninitialized tbl->lock bug 15 * Wensong Zhang : added doing full expiration check to 16 * collect stale entries of 24+ hours when 17 * no partial expire check in a half hour 18 * Julian Anastasov : replaced del_timer call with del_timer_sync 19 * to avoid the possible race between timer 20 * handler and del_timer thread in SMP 21 * 22 */ 23 24 /* 25 * The lblc algorithm is as follows (pseudo code): 26 * 27 * if cachenode[dest_ip] is null then 28 * n, cachenode[dest_ip] <- {weighted least-conn node}; 29 * else 30 * n <- cachenode[dest_ip]; 31 * if (n is dead) OR 32 * (n.conns>n.weight AND 33 * there is a node m with m.conns<m.weight/2) then 34 * n, cachenode[dest_ip] <- {weighted least-conn node}; 35 * 36 * return n; 37 * 38 * Thanks must go to Wenzhuo Zhang for talking WCCP to me and pushing 39 * me to write this module. 40 */ 41 42 #define KMSG_COMPONENT "IPVS" 43 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt 44 45 #include <linux/ip.h> 46 #include <linux/slab.h> 47 #include <linux/module.h> 48 #include <linux/kernel.h> 49 #include <linux/skbuff.h> 50 #include <linux/jiffies.h> 51 52 /* for sysctl */ 53 #include <linux/fs.h> 54 #include <linux/sysctl.h> 55 56 #include <net/ip_vs.h> 57 58 59 /* 60 * It is for garbage collection of stale IPVS lblc entries, 61 * when the table is full. 62 */ 63 #define CHECK_EXPIRE_INTERVAL (60*HZ) 64 #define ENTRY_TIMEOUT (6*60*HZ) 65 66 #define DEFAULT_EXPIRATION (24*60*60*HZ) 67 68 /* 69 * It is for full expiration check. 70 * When there is no partial expiration check (garbage collection) 71 * in a half hour, do a full expiration check to collect stale 72 * entries that haven't been touched for a day. 73 */ 74 #define COUNT_FOR_FULL_EXPIRATION 30 75 76 77 /* 78 * for IPVS lblc entry hash table 79 */ 80 #ifndef CONFIG_IP_VS_LBLC_TAB_BITS 81 #define CONFIG_IP_VS_LBLC_TAB_BITS 10 82 #endif 83 #define IP_VS_LBLC_TAB_BITS CONFIG_IP_VS_LBLC_TAB_BITS 84 #define IP_VS_LBLC_TAB_SIZE (1 << IP_VS_LBLC_TAB_BITS) 85 #define IP_VS_LBLC_TAB_MASK (IP_VS_LBLC_TAB_SIZE - 1) 86 87 88 /* 89 * IPVS lblc entry represents an association between destination 90 * IP address and its destination server 91 */ 92 struct ip_vs_lblc_entry { 93 struct list_head list; 94 int af; /* address family */ 95 union nf_inet_addr addr; /* destination IP address */ 96 struct ip_vs_dest *dest; /* real server (cache) */ 97 unsigned long lastuse; /* last used time */ 98 }; 99 100 101 /* 102 * IPVS lblc hash table 103 */ 104 struct ip_vs_lblc_table { 105 struct list_head bucket[IP_VS_LBLC_TAB_SIZE]; /* hash bucket */ 106 atomic_t entries; /* number of entries */ 107 int max_size; /* maximum size of entries */ 108 struct timer_list periodic_timer; /* collect stale entries */ 109 int rover; /* rover for expire check */ 110 int counter; /* counter for no expire */ 111 }; 112 113 114 /* 115 * IPVS LBLC sysctl table 116 */ 117 #ifdef CONFIG_SYSCTL 118 static ctl_table vs_vars_table[] = { 119 { 120 .procname = "lblc_expiration", 121 .data = NULL, 122 .maxlen = sizeof(int), 123 .mode = 0644, 124 .proc_handler = proc_dointvec_jiffies, 125 }, 126 { } 127 }; 128 #endif 129 130 static inline void ip_vs_lblc_free(struct ip_vs_lblc_entry *en) 131 { 132 list_del(&en->list); 133 /* 134 * We don't kfree dest because it is referred either by its service 135 * or the trash dest list. 136 */ 137 atomic_dec(&en->dest->refcnt); 138 kfree(en); 139 } 140 141 142 /* 143 * Returns hash value for IPVS LBLC entry 144 */ 145 static inline unsigned 146 ip_vs_lblc_hashkey(int af, const union nf_inet_addr *addr) 147 { 148 __be32 addr_fold = addr->ip; 149 150 #ifdef CONFIG_IP_VS_IPV6 151 if (af == AF_INET6) 152 addr_fold = addr->ip6[0]^addr->ip6[1]^ 153 addr->ip6[2]^addr->ip6[3]; 154 #endif 155 return (ntohl(addr_fold)*2654435761UL) & IP_VS_LBLC_TAB_MASK; 156 } 157 158 159 /* 160 * Hash an entry in the ip_vs_lblc_table. 161 * returns bool success. 162 */ 163 static void 164 ip_vs_lblc_hash(struct ip_vs_lblc_table *tbl, struct ip_vs_lblc_entry *en) 165 { 166 unsigned hash = ip_vs_lblc_hashkey(en->af, &en->addr); 167 168 list_add(&en->list, &tbl->bucket[hash]); 169 atomic_inc(&tbl->entries); 170 } 171 172 173 /* 174 * Get ip_vs_lblc_entry associated with supplied parameters. Called under read 175 * lock 176 */ 177 static inline struct ip_vs_lblc_entry * 178 ip_vs_lblc_get(int af, struct ip_vs_lblc_table *tbl, 179 const union nf_inet_addr *addr) 180 { 181 unsigned hash = ip_vs_lblc_hashkey(af, addr); 182 struct ip_vs_lblc_entry *en; 183 184 list_for_each_entry(en, &tbl->bucket[hash], list) 185 if (ip_vs_addr_equal(af, &en->addr, addr)) 186 return en; 187 188 return NULL; 189 } 190 191 192 /* 193 * Create or update an ip_vs_lblc_entry, which is a mapping of a destination IP 194 * address to a server. Called under write lock. 195 */ 196 static inline struct ip_vs_lblc_entry * 197 ip_vs_lblc_new(struct ip_vs_lblc_table *tbl, const union nf_inet_addr *daddr, 198 struct ip_vs_dest *dest) 199 { 200 struct ip_vs_lblc_entry *en; 201 202 en = ip_vs_lblc_get(dest->af, tbl, daddr); 203 if (!en) { 204 en = kmalloc(sizeof(*en), GFP_ATOMIC); 205 if (!en) 206 return NULL; 207 208 en->af = dest->af; 209 ip_vs_addr_copy(dest->af, &en->addr, daddr); 210 en->lastuse = jiffies; 211 212 atomic_inc(&dest->refcnt); 213 en->dest = dest; 214 215 ip_vs_lblc_hash(tbl, en); 216 } else if (en->dest != dest) { 217 atomic_dec(&en->dest->refcnt); 218 atomic_inc(&dest->refcnt); 219 en->dest = dest; 220 } 221 222 return en; 223 } 224 225 226 /* 227 * Flush all the entries of the specified table. 228 */ 229 static void ip_vs_lblc_flush(struct ip_vs_lblc_table *tbl) 230 { 231 struct ip_vs_lblc_entry *en, *nxt; 232 int i; 233 234 for (i=0; i<IP_VS_LBLC_TAB_SIZE; i++) { 235 list_for_each_entry_safe(en, nxt, &tbl->bucket[i], list) { 236 ip_vs_lblc_free(en); 237 atomic_dec(&tbl->entries); 238 } 239 } 240 } 241 242 static int sysctl_lblc_expiration(struct ip_vs_service *svc) 243 { 244 #ifdef CONFIG_SYSCTL 245 struct netns_ipvs *ipvs = net_ipvs(svc->net); 246 return ipvs->sysctl_lblc_expiration; 247 #else 248 return DEFAULT_EXPIRATION; 249 #endif 250 } 251 252 static inline void ip_vs_lblc_full_check(struct ip_vs_service *svc) 253 { 254 struct ip_vs_lblc_table *tbl = svc->sched_data; 255 struct ip_vs_lblc_entry *en, *nxt; 256 unsigned long now = jiffies; 257 int i, j; 258 259 for (i=0, j=tbl->rover; i<IP_VS_LBLC_TAB_SIZE; i++) { 260 j = (j + 1) & IP_VS_LBLC_TAB_MASK; 261 262 write_lock(&svc->sched_lock); 263 list_for_each_entry_safe(en, nxt, &tbl->bucket[j], list) { 264 if (time_before(now, 265 en->lastuse + 266 sysctl_lblc_expiration(svc))) 267 continue; 268 269 ip_vs_lblc_free(en); 270 atomic_dec(&tbl->entries); 271 } 272 write_unlock(&svc->sched_lock); 273 } 274 tbl->rover = j; 275 } 276 277 278 /* 279 * Periodical timer handler for IPVS lblc table 280 * It is used to collect stale entries when the number of entries 281 * exceeds the maximum size of the table. 282 * 283 * Fixme: we probably need more complicated algorithm to collect 284 * entries that have not been used for a long time even 285 * if the number of entries doesn't exceed the maximum size 286 * of the table. 287 * The full expiration check is for this purpose now. 288 */ 289 static void ip_vs_lblc_check_expire(unsigned long data) 290 { 291 struct ip_vs_service *svc = (struct ip_vs_service *) data; 292 struct ip_vs_lblc_table *tbl = svc->sched_data; 293 unsigned long now = jiffies; 294 int goal; 295 int i, j; 296 struct ip_vs_lblc_entry *en, *nxt; 297 298 if ((tbl->counter % COUNT_FOR_FULL_EXPIRATION) == 0) { 299 /* do full expiration check */ 300 ip_vs_lblc_full_check(svc); 301 tbl->counter = 1; 302 goto out; 303 } 304 305 if (atomic_read(&tbl->entries) <= tbl->max_size) { 306 tbl->counter++; 307 goto out; 308 } 309 310 goal = (atomic_read(&tbl->entries) - tbl->max_size)*4/3; 311 if (goal > tbl->max_size/2) 312 goal = tbl->max_size/2; 313 314 for (i=0, j=tbl->rover; i<IP_VS_LBLC_TAB_SIZE; i++) { 315 j = (j + 1) & IP_VS_LBLC_TAB_MASK; 316 317 write_lock(&svc->sched_lock); 318 list_for_each_entry_safe(en, nxt, &tbl->bucket[j], list) { 319 if (time_before(now, en->lastuse + ENTRY_TIMEOUT)) 320 continue; 321 322 ip_vs_lblc_free(en); 323 atomic_dec(&tbl->entries); 324 goal--; 325 } 326 write_unlock(&svc->sched_lock); 327 if (goal <= 0) 328 break; 329 } 330 tbl->rover = j; 331 332 out: 333 mod_timer(&tbl->periodic_timer, jiffies+CHECK_EXPIRE_INTERVAL); 334 } 335 336 337 static int ip_vs_lblc_init_svc(struct ip_vs_service *svc) 338 { 339 int i; 340 struct ip_vs_lblc_table *tbl; 341 342 /* 343 * Allocate the ip_vs_lblc_table for this service 344 */ 345 tbl = kmalloc(sizeof(*tbl), GFP_ATOMIC); 346 if (tbl == NULL) 347 return -ENOMEM; 348 349 svc->sched_data = tbl; 350 IP_VS_DBG(6, "LBLC hash table (memory=%Zdbytes) allocated for " 351 "current service\n", sizeof(*tbl)); 352 353 /* 354 * Initialize the hash buckets 355 */ 356 for (i=0; i<IP_VS_LBLC_TAB_SIZE; i++) { 357 INIT_LIST_HEAD(&tbl->bucket[i]); 358 } 359 tbl->max_size = IP_VS_LBLC_TAB_SIZE*16; 360 tbl->rover = 0; 361 tbl->counter = 1; 362 363 /* 364 * Hook periodic timer for garbage collection 365 */ 366 setup_timer(&tbl->periodic_timer, ip_vs_lblc_check_expire, 367 (unsigned long)svc); 368 mod_timer(&tbl->periodic_timer, jiffies + CHECK_EXPIRE_INTERVAL); 369 370 return 0; 371 } 372 373 374 static int ip_vs_lblc_done_svc(struct ip_vs_service *svc) 375 { 376 struct ip_vs_lblc_table *tbl = svc->sched_data; 377 378 /* remove periodic timer */ 379 del_timer_sync(&tbl->periodic_timer); 380 381 /* got to clean up table entries here */ 382 ip_vs_lblc_flush(tbl); 383 384 /* release the table itself */ 385 kfree(tbl); 386 IP_VS_DBG(6, "LBLC hash table (memory=%Zdbytes) released\n", 387 sizeof(*tbl)); 388 389 return 0; 390 } 391 392 393 static inline struct ip_vs_dest * 394 __ip_vs_lblc_schedule(struct ip_vs_service *svc) 395 { 396 struct ip_vs_dest *dest, *least; 397 int loh, doh; 398 399 /* 400 * We use the following formula to estimate the load: 401 * (dest overhead) / dest->weight 402 * 403 * Remember -- no floats in kernel mode!!! 404 * The comparison of h1*w2 > h2*w1 is equivalent to that of 405 * h1/w1 > h2/w2 406 * if every weight is larger than zero. 407 * 408 * The server with weight=0 is quiesced and will not receive any 409 * new connection. 410 */ 411 list_for_each_entry(dest, &svc->destinations, n_list) { 412 if (dest->flags & IP_VS_DEST_F_OVERLOAD) 413 continue; 414 if (atomic_read(&dest->weight) > 0) { 415 least = dest; 416 loh = ip_vs_dest_conn_overhead(least); 417 goto nextstage; 418 } 419 } 420 return NULL; 421 422 /* 423 * Find the destination with the least load. 424 */ 425 nextstage: 426 list_for_each_entry_continue(dest, &svc->destinations, n_list) { 427 if (dest->flags & IP_VS_DEST_F_OVERLOAD) 428 continue; 429 430 doh = ip_vs_dest_conn_overhead(dest); 431 if (loh * atomic_read(&dest->weight) > 432 doh * atomic_read(&least->weight)) { 433 least = dest; 434 loh = doh; 435 } 436 } 437 438 IP_VS_DBG_BUF(6, "LBLC: server %s:%d " 439 "activeconns %d refcnt %d weight %d overhead %d\n", 440 IP_VS_DBG_ADDR(least->af, &least->addr), 441 ntohs(least->port), 442 atomic_read(&least->activeconns), 443 atomic_read(&least->refcnt), 444 atomic_read(&least->weight), loh); 445 446 return least; 447 } 448 449 450 /* 451 * If this destination server is overloaded and there is a less loaded 452 * server, then return true. 453 */ 454 static inline int 455 is_overloaded(struct ip_vs_dest *dest, struct ip_vs_service *svc) 456 { 457 if (atomic_read(&dest->activeconns) > atomic_read(&dest->weight)) { 458 struct ip_vs_dest *d; 459 460 list_for_each_entry(d, &svc->destinations, n_list) { 461 if (atomic_read(&d->activeconns)*2 462 < atomic_read(&d->weight)) { 463 return 1; 464 } 465 } 466 } 467 return 0; 468 } 469 470 471 /* 472 * Locality-Based (weighted) Least-Connection scheduling 473 */ 474 static struct ip_vs_dest * 475 ip_vs_lblc_schedule(struct ip_vs_service *svc, const struct sk_buff *skb) 476 { 477 struct ip_vs_lblc_table *tbl = svc->sched_data; 478 struct ip_vs_iphdr iph; 479 struct ip_vs_dest *dest = NULL; 480 struct ip_vs_lblc_entry *en; 481 482 ip_vs_fill_iphdr(svc->af, skb_network_header(skb), &iph); 483 484 IP_VS_DBG(6, "%s(): Scheduling...\n", __func__); 485 486 /* First look in our cache */ 487 read_lock(&svc->sched_lock); 488 en = ip_vs_lblc_get(svc->af, tbl, &iph.daddr); 489 if (en) { 490 /* We only hold a read lock, but this is atomic */ 491 en->lastuse = jiffies; 492 493 /* 494 * If the destination is not available, i.e. it's in the trash, 495 * we must ignore it, as it may be removed from under our feet, 496 * if someone drops our reference count. Our caller only makes 497 * sure that destinations, that are not in the trash, are not 498 * moved to the trash, while we are scheduling. But anyone can 499 * free up entries from the trash at any time. 500 */ 501 502 if (en->dest->flags & IP_VS_DEST_F_AVAILABLE) 503 dest = en->dest; 504 } 505 read_unlock(&svc->sched_lock); 506 507 /* If the destination has a weight and is not overloaded, use it */ 508 if (dest && atomic_read(&dest->weight) > 0 && !is_overloaded(dest, svc)) 509 goto out; 510 511 /* No cache entry or it is invalid, time to schedule */ 512 dest = __ip_vs_lblc_schedule(svc); 513 if (!dest) { 514 ip_vs_scheduler_err(svc, "no destination available"); 515 return NULL; 516 } 517 518 /* If we fail to create a cache entry, we'll just use the valid dest */ 519 write_lock(&svc->sched_lock); 520 ip_vs_lblc_new(tbl, &iph.daddr, dest); 521 write_unlock(&svc->sched_lock); 522 523 out: 524 IP_VS_DBG_BUF(6, "LBLC: destination IP address %s --> server %s:%d\n", 525 IP_VS_DBG_ADDR(svc->af, &iph.daddr), 526 IP_VS_DBG_ADDR(svc->af, &dest->addr), ntohs(dest->port)); 527 528 return dest; 529 } 530 531 532 /* 533 * IPVS LBLC Scheduler structure 534 */ 535 static struct ip_vs_scheduler ip_vs_lblc_scheduler = 536 { 537 .name = "lblc", 538 .refcnt = ATOMIC_INIT(0), 539 .module = THIS_MODULE, 540 .n_list = LIST_HEAD_INIT(ip_vs_lblc_scheduler.n_list), 541 .init_service = ip_vs_lblc_init_svc, 542 .done_service = ip_vs_lblc_done_svc, 543 .schedule = ip_vs_lblc_schedule, 544 }; 545 546 /* 547 * per netns init. 548 */ 549 #ifdef CONFIG_SYSCTL 550 static int __net_init __ip_vs_lblc_init(struct net *net) 551 { 552 struct netns_ipvs *ipvs = net_ipvs(net); 553 554 if (!net_eq(net, &init_net)) { 555 ipvs->lblc_ctl_table = kmemdup(vs_vars_table, 556 sizeof(vs_vars_table), 557 GFP_KERNEL); 558 if (ipvs->lblc_ctl_table == NULL) 559 return -ENOMEM; 560 } else 561 ipvs->lblc_ctl_table = vs_vars_table; 562 ipvs->sysctl_lblc_expiration = DEFAULT_EXPIRATION; 563 ipvs->lblc_ctl_table[0].data = &ipvs->sysctl_lblc_expiration; 564 565 ipvs->lblc_ctl_header = 566 register_net_sysctl_table(net, net_vs_ctl_path, 567 ipvs->lblc_ctl_table); 568 if (!ipvs->lblc_ctl_header) { 569 if (!net_eq(net, &init_net)) 570 kfree(ipvs->lblc_ctl_table); 571 return -ENOMEM; 572 } 573 574 return 0; 575 } 576 577 static void __net_exit __ip_vs_lblc_exit(struct net *net) 578 { 579 struct netns_ipvs *ipvs = net_ipvs(net); 580 581 unregister_net_sysctl_table(ipvs->lblc_ctl_header); 582 583 if (!net_eq(net, &init_net)) 584 kfree(ipvs->lblc_ctl_table); 585 } 586 587 #else 588 589 static int __net_init __ip_vs_lblc_init(struct net *net) { return 0; } 590 static void __net_exit __ip_vs_lblc_exit(struct net *net) { } 591 592 #endif 593 594 static struct pernet_operations ip_vs_lblc_ops = { 595 .init = __ip_vs_lblc_init, 596 .exit = __ip_vs_lblc_exit, 597 }; 598 599 static int __init ip_vs_lblc_init(void) 600 { 601 int ret; 602 603 ret = register_pernet_subsys(&ip_vs_lblc_ops); 604 if (ret) 605 return ret; 606 607 ret = register_ip_vs_scheduler(&ip_vs_lblc_scheduler); 608 if (ret) 609 unregister_pernet_subsys(&ip_vs_lblc_ops); 610 return ret; 611 } 612 613 static void __exit ip_vs_lblc_cleanup(void) 614 { 615 unregister_ip_vs_scheduler(&ip_vs_lblc_scheduler); 616 unregister_pernet_subsys(&ip_vs_lblc_ops); 617 } 618 619 620 module_init(ip_vs_lblc_init); 621 module_exit(ip_vs_lblc_cleanup); 622 MODULE_LICENSE("GPL"); 623