1 // SPDX-License-Identifier: GPL-2.0 2 /* IPVS: Maglev Hashing scheduling module 3 * 4 * Authors: Inju Song <inju.song@navercorp.com> 5 * 6 */ 7 8 /* The mh algorithm is to assign a preference list of all the lookup 9 * table positions to each destination and populate the table with 10 * the most-preferred position of destinations. Then it is to select 11 * destination with the hash key of source IP address through looking 12 * up a the lookup table. 13 * 14 * The algorithm is detailed in: 15 * [3.4 Consistent Hasing] 16 https://www.usenix.org/system/files/conference/nsdi16/nsdi16-paper-eisenbud.pdf 17 * 18 */ 19 20 #define KMSG_COMPONENT "IPVS" 21 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt 22 23 #include <linux/ip.h> 24 #include <linux/slab.h> 25 #include <linux/module.h> 26 #include <linux/kernel.h> 27 #include <linux/skbuff.h> 28 29 #include <net/ip_vs.h> 30 31 #include <linux/siphash.h> 32 #include <linux/bitops.h> 33 #include <linux/gcd.h> 34 35 #define IP_VS_SVC_F_SCHED_MH_FALLBACK IP_VS_SVC_F_SCHED1 /* MH fallback */ 36 #define IP_VS_SVC_F_SCHED_MH_PORT IP_VS_SVC_F_SCHED2 /* MH use port */ 37 38 struct ip_vs_mh_lookup { 39 struct ip_vs_dest __rcu *dest; /* real server (cache) */ 40 }; 41 42 struct ip_vs_mh_dest_setup { 43 unsigned int offset; /* starting offset */ 44 unsigned int skip; /* skip */ 45 unsigned int perm; /* next_offset */ 46 int turns; /* weight / gcd() and rshift */ 47 }; 48 49 /* Available prime numbers for MH table */ 50 static int primes[] = {251, 509, 1021, 2039, 4093, 51 8191, 16381, 32749, 65521, 131071}; 52 53 /* For IPVS MH entry hash table */ 54 #ifndef CONFIG_IP_VS_MH_TAB_INDEX 55 #define CONFIG_IP_VS_MH_TAB_INDEX 12 56 #endif 57 #define IP_VS_MH_TAB_BITS (CONFIG_IP_VS_MH_TAB_INDEX / 2) 58 #define IP_VS_MH_TAB_INDEX (CONFIG_IP_VS_MH_TAB_INDEX - 8) 59 #define IP_VS_MH_TAB_SIZE primes[IP_VS_MH_TAB_INDEX] 60 61 struct ip_vs_mh_state { 62 struct rcu_head rcu_head; 63 struct ip_vs_mh_lookup *lookup; 64 struct ip_vs_mh_dest_setup *dest_setup; 65 hsiphash_key_t hash1, hash2; 66 int gcd; 67 int rshift; 68 }; 69 70 static inline void generate_hash_secret(hsiphash_key_t *hash1, 71 hsiphash_key_t *hash2) 72 { 73 hash1->key[0] = 2654435761UL; 74 hash1->key[1] = 2654435761UL; 75 76 hash2->key[0] = 2654446892UL; 77 hash2->key[1] = 2654446892UL; 78 } 79 80 /* Helper function to determine if server is unavailable */ 81 static inline bool is_unavailable(struct ip_vs_dest *dest) 82 { 83 return atomic_read(&dest->weight) <= 0 || 84 dest->flags & IP_VS_DEST_F_OVERLOAD; 85 } 86 87 /* Returns hash value for IPVS MH entry */ 88 static inline unsigned int 89 ip_vs_mh_hashkey(int af, const union nf_inet_addr *addr, 90 __be16 port, hsiphash_key_t *key, unsigned int offset) 91 { 92 unsigned int v; 93 __be32 addr_fold = addr->ip; 94 95 #ifdef CONFIG_IP_VS_IPV6 96 if (af == AF_INET6) 97 addr_fold = addr->ip6[0] ^ addr->ip6[1] ^ 98 addr->ip6[2] ^ addr->ip6[3]; 99 #endif 100 v = (offset + ntohs(port) + ntohl(addr_fold)); 101 return hsiphash(&v, sizeof(v), key); 102 } 103 104 /* Reset all the hash buckets of the specified table. */ 105 static void ip_vs_mh_reset(struct ip_vs_mh_state *s) 106 { 107 int i; 108 struct ip_vs_mh_lookup *l; 109 struct ip_vs_dest *dest; 110 111 l = &s->lookup[0]; 112 for (i = 0; i < IP_VS_MH_TAB_SIZE; i++) { 113 dest = rcu_dereference_protected(l->dest, 1); 114 if (dest) { 115 ip_vs_dest_put(dest); 116 RCU_INIT_POINTER(l->dest, NULL); 117 } 118 l++; 119 } 120 } 121 122 static int ip_vs_mh_permutate(struct ip_vs_mh_state *s, 123 struct ip_vs_service *svc) 124 { 125 struct list_head *p; 126 struct ip_vs_mh_dest_setup *ds; 127 struct ip_vs_dest *dest; 128 int lw; 129 130 /* If gcd is smaller then 1, number of dests or 131 * all last_weight of dests are zero. So, skip 132 * permutation for the dests. 133 */ 134 if (s->gcd < 1) 135 return 0; 136 137 /* Set dest_setup for the dests permutation */ 138 p = &svc->destinations; 139 ds = &s->dest_setup[0]; 140 while ((p = p->next) != &svc->destinations) { 141 dest = list_entry(p, struct ip_vs_dest, n_list); 142 143 ds->offset = ip_vs_mh_hashkey(svc->af, &dest->addr, 144 dest->port, &s->hash1, 0) % 145 IP_VS_MH_TAB_SIZE; 146 ds->skip = ip_vs_mh_hashkey(svc->af, &dest->addr, 147 dest->port, &s->hash2, 0) % 148 (IP_VS_MH_TAB_SIZE - 1) + 1; 149 ds->perm = ds->offset; 150 151 lw = atomic_read(&dest->last_weight); 152 ds->turns = ((lw / s->gcd) >> s->rshift) ? : (lw != 0); 153 ds++; 154 } 155 156 return 0; 157 } 158 159 static int ip_vs_mh_populate(struct ip_vs_mh_state *s, 160 struct ip_vs_service *svc) 161 { 162 int n, c, dt_count; 163 unsigned long *table; 164 struct list_head *p; 165 struct ip_vs_mh_dest_setup *ds; 166 struct ip_vs_dest *dest, *new_dest; 167 168 /* If gcd is smaller then 1, number of dests or 169 * all last_weight of dests are zero. So, skip 170 * the population for the dests and reset lookup table. 171 */ 172 if (s->gcd < 1) { 173 ip_vs_mh_reset(s); 174 return 0; 175 } 176 177 table = bitmap_zalloc(IP_VS_MH_TAB_SIZE, GFP_KERNEL); 178 if (!table) 179 return -ENOMEM; 180 181 p = &svc->destinations; 182 n = 0; 183 dt_count = 0; 184 while (n < IP_VS_MH_TAB_SIZE) { 185 if (p == &svc->destinations) 186 p = p->next; 187 188 ds = &s->dest_setup[0]; 189 while (p != &svc->destinations) { 190 /* Ignore added server with zero weight */ 191 if (ds->turns < 1) { 192 p = p->next; 193 ds++; 194 continue; 195 } 196 197 c = ds->perm; 198 while (test_bit(c, table)) { 199 /* Add skip, mod IP_VS_MH_TAB_SIZE */ 200 ds->perm += ds->skip; 201 if (ds->perm >= IP_VS_MH_TAB_SIZE) 202 ds->perm -= IP_VS_MH_TAB_SIZE; 203 c = ds->perm; 204 } 205 206 __set_bit(c, table); 207 208 dest = rcu_dereference_protected(s->lookup[c].dest, 1); 209 new_dest = list_entry(p, struct ip_vs_dest, n_list); 210 if (dest != new_dest) { 211 if (dest) 212 ip_vs_dest_put(dest); 213 ip_vs_dest_hold(new_dest); 214 RCU_INIT_POINTER(s->lookup[c].dest, new_dest); 215 } 216 217 if (++n == IP_VS_MH_TAB_SIZE) 218 goto out; 219 220 if (++dt_count >= ds->turns) { 221 dt_count = 0; 222 p = p->next; 223 ds++; 224 } 225 } 226 } 227 228 out: 229 bitmap_free(table); 230 return 0; 231 } 232 233 /* Get ip_vs_dest associated with supplied parameters. */ 234 static inline struct ip_vs_dest * 235 ip_vs_mh_get(struct ip_vs_service *svc, struct ip_vs_mh_state *s, 236 const union nf_inet_addr *addr, __be16 port) 237 { 238 unsigned int hash = ip_vs_mh_hashkey(svc->af, addr, port, &s->hash1, 0) 239 % IP_VS_MH_TAB_SIZE; 240 struct ip_vs_dest *dest = rcu_dereference(s->lookup[hash].dest); 241 242 return (!dest || is_unavailable(dest)) ? NULL : dest; 243 } 244 245 /* As ip_vs_mh_get, but with fallback if selected server is unavailable */ 246 static inline struct ip_vs_dest * 247 ip_vs_mh_get_fallback(struct ip_vs_service *svc, struct ip_vs_mh_state *s, 248 const union nf_inet_addr *addr, __be16 port) 249 { 250 unsigned int offset, roffset; 251 unsigned int hash, ihash; 252 struct ip_vs_dest *dest; 253 254 /* First try the dest it's supposed to go to */ 255 ihash = ip_vs_mh_hashkey(svc->af, addr, port, 256 &s->hash1, 0) % IP_VS_MH_TAB_SIZE; 257 dest = rcu_dereference(s->lookup[ihash].dest); 258 if (!dest) 259 return NULL; 260 if (!is_unavailable(dest)) 261 return dest; 262 263 IP_VS_DBG_BUF(6, "MH: selected unavailable server %s:%u, reselecting", 264 IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port)); 265 266 /* If the original dest is unavailable, loop around the table 267 * starting from ihash to find a new dest 268 */ 269 for (offset = 0; offset < IP_VS_MH_TAB_SIZE; offset++) { 270 roffset = (offset + ihash) % IP_VS_MH_TAB_SIZE; 271 hash = ip_vs_mh_hashkey(svc->af, addr, port, &s->hash1, 272 roffset) % IP_VS_MH_TAB_SIZE; 273 dest = rcu_dereference(s->lookup[hash].dest); 274 if (!dest) 275 break; 276 if (!is_unavailable(dest)) 277 return dest; 278 IP_VS_DBG_BUF(6, 279 "MH: selected unavailable server %s:%u (offset %u), reselecting", 280 IP_VS_DBG_ADDR(dest->af, &dest->addr), 281 ntohs(dest->port), roffset); 282 } 283 284 return NULL; 285 } 286 287 /* Assign all the hash buckets of the specified table with the service. */ 288 static int ip_vs_mh_reassign(struct ip_vs_mh_state *s, 289 struct ip_vs_service *svc) 290 { 291 int ret; 292 293 if (svc->num_dests > IP_VS_MH_TAB_SIZE) 294 return -EINVAL; 295 296 if (svc->num_dests >= 1) { 297 s->dest_setup = kcalloc(svc->num_dests, 298 sizeof(struct ip_vs_mh_dest_setup), 299 GFP_KERNEL); 300 if (!s->dest_setup) 301 return -ENOMEM; 302 } 303 304 ip_vs_mh_permutate(s, svc); 305 306 ret = ip_vs_mh_populate(s, svc); 307 if (ret < 0) 308 goto out; 309 310 IP_VS_DBG_BUF(6, "MH: reassign lookup table of %s:%u\n", 311 IP_VS_DBG_ADDR(svc->af, &svc->addr), 312 ntohs(svc->port)); 313 314 out: 315 if (svc->num_dests >= 1) { 316 kfree(s->dest_setup); 317 s->dest_setup = NULL; 318 } 319 return ret; 320 } 321 322 static int ip_vs_mh_gcd_weight(struct ip_vs_service *svc) 323 { 324 struct ip_vs_dest *dest; 325 int weight; 326 int g = 0; 327 328 list_for_each_entry(dest, &svc->destinations, n_list) { 329 weight = atomic_read(&dest->last_weight); 330 if (weight > 0) { 331 if (g > 0) 332 g = gcd(weight, g); 333 else 334 g = weight; 335 } 336 } 337 return g; 338 } 339 340 /* To avoid assigning huge weight for the MH table, 341 * calculate shift value with gcd. 342 */ 343 static int ip_vs_mh_shift_weight(struct ip_vs_service *svc, int gcd) 344 { 345 struct ip_vs_dest *dest; 346 int new_weight, weight = 0; 347 int mw, shift; 348 349 /* If gcd is smaller then 1, number of dests or 350 * all last_weight of dests are zero. So, return 351 * shift value as zero. 352 */ 353 if (gcd < 1) 354 return 0; 355 356 list_for_each_entry(dest, &svc->destinations, n_list) { 357 new_weight = atomic_read(&dest->last_weight); 358 if (new_weight > weight) 359 weight = new_weight; 360 } 361 362 /* Because gcd is greater than zero, 363 * the maximum weight and gcd are always greater than zero 364 */ 365 mw = weight / gcd; 366 367 /* shift = occupied bits of weight/gcd - MH highest bits */ 368 shift = fls(mw) - IP_VS_MH_TAB_BITS; 369 return (shift >= 0) ? shift : 0; 370 } 371 372 static void ip_vs_mh_state_free(struct rcu_head *head) 373 { 374 struct ip_vs_mh_state *s; 375 376 s = container_of(head, struct ip_vs_mh_state, rcu_head); 377 kfree(s->lookup); 378 kfree(s); 379 } 380 381 static int ip_vs_mh_init_svc(struct ip_vs_service *svc) 382 { 383 int ret; 384 struct ip_vs_mh_state *s; 385 386 /* Allocate the MH table for this service */ 387 s = kzalloc(sizeof(*s), GFP_KERNEL); 388 if (!s) 389 return -ENOMEM; 390 391 s->lookup = kcalloc(IP_VS_MH_TAB_SIZE, sizeof(struct ip_vs_mh_lookup), 392 GFP_KERNEL); 393 if (!s->lookup) { 394 kfree(s); 395 return -ENOMEM; 396 } 397 398 generate_hash_secret(&s->hash1, &s->hash2); 399 s->gcd = ip_vs_mh_gcd_weight(svc); 400 s->rshift = ip_vs_mh_shift_weight(svc, s->gcd); 401 402 IP_VS_DBG(6, 403 "MH lookup table (memory=%zdbytes) allocated for current service\n", 404 sizeof(struct ip_vs_mh_lookup) * IP_VS_MH_TAB_SIZE); 405 406 /* Assign the lookup table with current dests */ 407 ret = ip_vs_mh_reassign(s, svc); 408 if (ret < 0) { 409 ip_vs_mh_reset(s); 410 ip_vs_mh_state_free(&s->rcu_head); 411 return ret; 412 } 413 414 /* No more failures, attach state */ 415 svc->sched_data = s; 416 return 0; 417 } 418 419 static void ip_vs_mh_done_svc(struct ip_vs_service *svc) 420 { 421 struct ip_vs_mh_state *s = svc->sched_data; 422 423 /* Got to clean up lookup entry here */ 424 ip_vs_mh_reset(s); 425 426 call_rcu(&s->rcu_head, ip_vs_mh_state_free); 427 IP_VS_DBG(6, "MH lookup table (memory=%zdbytes) released\n", 428 sizeof(struct ip_vs_mh_lookup) * IP_VS_MH_TAB_SIZE); 429 } 430 431 static int ip_vs_mh_dest_changed(struct ip_vs_service *svc, 432 struct ip_vs_dest *dest) 433 { 434 struct ip_vs_mh_state *s = svc->sched_data; 435 436 s->gcd = ip_vs_mh_gcd_weight(svc); 437 s->rshift = ip_vs_mh_shift_weight(svc, s->gcd); 438 439 /* Assign the lookup table with the updated service */ 440 return ip_vs_mh_reassign(s, svc); 441 } 442 443 /* Helper function to get port number */ 444 static inline __be16 445 ip_vs_mh_get_port(const struct sk_buff *skb, struct ip_vs_iphdr *iph) 446 { 447 __be16 _ports[2], *ports; 448 449 /* At this point we know that we have a valid packet of some kind. 450 * Because ICMP packets are only guaranteed to have the first 8 451 * bytes, let's just grab the ports. Fortunately they're in the 452 * same position for all three of the protocols we care about. 453 */ 454 switch (iph->protocol) { 455 case IPPROTO_TCP: 456 case IPPROTO_UDP: 457 case IPPROTO_SCTP: 458 ports = skb_header_pointer(skb, iph->len, sizeof(_ports), 459 &_ports); 460 if (unlikely(!ports)) 461 return 0; 462 463 if (likely(!ip_vs_iph_inverse(iph))) 464 return ports[0]; 465 else 466 return ports[1]; 467 default: 468 return 0; 469 } 470 } 471 472 /* Maglev Hashing scheduling */ 473 static struct ip_vs_dest * 474 ip_vs_mh_schedule(struct ip_vs_service *svc, const struct sk_buff *skb, 475 struct ip_vs_iphdr *iph) 476 { 477 struct ip_vs_dest *dest; 478 struct ip_vs_mh_state *s; 479 __be16 port = 0; 480 const union nf_inet_addr *hash_addr; 481 482 hash_addr = ip_vs_iph_inverse(iph) ? &iph->daddr : &iph->saddr; 483 484 IP_VS_DBG(6, "%s : Scheduling...\n", __func__); 485 486 if (svc->flags & IP_VS_SVC_F_SCHED_MH_PORT) 487 port = ip_vs_mh_get_port(skb, iph); 488 489 s = (struct ip_vs_mh_state *)svc->sched_data; 490 491 if (svc->flags & IP_VS_SVC_F_SCHED_MH_FALLBACK) 492 dest = ip_vs_mh_get_fallback(svc, s, hash_addr, port); 493 else 494 dest = ip_vs_mh_get(svc, s, hash_addr, port); 495 496 if (!dest) { 497 ip_vs_scheduler_err(svc, "no destination available"); 498 return NULL; 499 } 500 501 IP_VS_DBG_BUF(6, "MH: source IP address %s:%u --> server %s:%u\n", 502 IP_VS_DBG_ADDR(svc->af, hash_addr), 503 ntohs(port), 504 IP_VS_DBG_ADDR(dest->af, &dest->addr), 505 ntohs(dest->port)); 506 507 return dest; 508 } 509 510 /* IPVS MH Scheduler structure */ 511 static struct ip_vs_scheduler ip_vs_mh_scheduler = { 512 .name = "mh", 513 .refcnt = ATOMIC_INIT(0), 514 .module = THIS_MODULE, 515 .n_list = LIST_HEAD_INIT(ip_vs_mh_scheduler.n_list), 516 .init_service = ip_vs_mh_init_svc, 517 .done_service = ip_vs_mh_done_svc, 518 .add_dest = ip_vs_mh_dest_changed, 519 .del_dest = ip_vs_mh_dest_changed, 520 .upd_dest = ip_vs_mh_dest_changed, 521 .schedule = ip_vs_mh_schedule, 522 }; 523 524 static int __init ip_vs_mh_init(void) 525 { 526 return register_ip_vs_scheduler(&ip_vs_mh_scheduler); 527 } 528 529 static void __exit ip_vs_mh_cleanup(void) 530 { 531 unregister_ip_vs_scheduler(&ip_vs_mh_scheduler); 532 rcu_barrier(); 533 } 534 535 module_init(ip_vs_mh_init); 536 module_exit(ip_vs_mh_cleanup); 537 MODULE_DESCRIPTION("Maglev hashing ipvs scheduler"); 538 MODULE_LICENSE("GPL v2"); 539 MODULE_AUTHOR("Inju Song <inju.song@navercorp.com>"); 540