1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * IPVS An implementation of the IP virtual server support for the 4 * LINUX operating system. IPVS is now implemented as a module 5 * over the NetFilter framework. IPVS can be used to build a 6 * high-performance and highly available server based on a 7 * cluster of servers. 8 * 9 * Authors: Wensong Zhang <wensong@linuxvirtualserver.org> 10 * Peter Kese <peter.kese@ijs.si> 11 * Julian Anastasov <ja@ssi.bg> 12 * 13 * Changes: 14 */ 15 16 #define KMSG_COMPONENT "IPVS" 17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt 18 19 #include <linux/module.h> 20 #include <linux/init.h> 21 #include <linux/types.h> 22 #include <linux/capability.h> 23 #include <linux/fs.h> 24 #include <linux/sysctl.h> 25 #include <linux/proc_fs.h> 26 #include <linux/workqueue.h> 27 #include <linux/seq_file.h> 28 #include <linux/slab.h> 29 30 #include <linux/netfilter.h> 31 #include <linux/netfilter_ipv4.h> 32 #include <linux/mutex.h> 33 34 #include <net/net_namespace.h> 35 #include <linux/nsproxy.h> 36 #include <net/ip.h> 37 #ifdef CONFIG_IP_VS_IPV6 38 #include <net/ipv6.h> 39 #include <net/ip6_route.h> 40 #include <net/netfilter/ipv6/nf_defrag_ipv6.h> 41 #endif 42 #include <net/route.h> 43 #include <net/sock.h> 44 #include <net/genetlink.h> 45 46 #include <linux/uaccess.h> 47 48 #include <net/ip_vs.h> 49 50 MODULE_ALIAS_GENL_FAMILY(IPVS_GENL_NAME); 51 52 DEFINE_MUTEX(__ip_vs_mutex); /* Serialize configuration with sockopt/netlink */ 53 54 /* sysctl variables */ 55 56 #ifdef CONFIG_IP_VS_DEBUG 57 static int sysctl_ip_vs_debug_level = 0; 58 59 int ip_vs_get_debug_level(void) 60 { 61 return sysctl_ip_vs_debug_level; 62 } 63 #endif 64 65 66 /* Protos */ 67 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup); 68 69 70 #ifdef CONFIG_IP_VS_IPV6 71 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */ 72 static bool __ip_vs_addr_is_local_v6(struct net *net, 73 const struct in6_addr *addr) 74 { 75 struct flowi6 fl6 = { 76 .daddr = *addr, 77 }; 78 struct dst_entry *dst = ip6_route_output(net, NULL, &fl6); 79 bool is_local; 80 81 is_local = !dst->error && dst->dev && (dst->dev->flags & IFF_LOOPBACK); 82 83 dst_release(dst); 84 return is_local; 85 } 86 #endif 87 88 #ifdef CONFIG_SYSCTL 89 /* 90 * update_defense_level is called from keventd and from sysctl, 91 * so it needs to protect itself from softirqs 92 */ 93 static void update_defense_level(struct netns_ipvs *ipvs) 94 { 95 struct sysinfo i; 96 int availmem; 97 int nomem; 98 int to_change = -1; 99 100 /* we only count free and buffered memory (in pages) */ 101 si_meminfo(&i); 102 availmem = i.freeram + i.bufferram; 103 /* however in linux 2.5 the i.bufferram is total page cache size, 104 we need adjust it */ 105 /* si_swapinfo(&i); */ 106 /* availmem = availmem - (i.totalswap - i.freeswap); */ 107 108 nomem = (availmem < ipvs->sysctl_amemthresh); 109 110 local_bh_disable(); 111 112 /* drop_entry */ 113 spin_lock(&ipvs->dropentry_lock); 114 switch (ipvs->sysctl_drop_entry) { 115 case 0: 116 atomic_set(&ipvs->dropentry, 0); 117 break; 118 case 1: 119 if (nomem) { 120 atomic_set(&ipvs->dropentry, 1); 121 ipvs->sysctl_drop_entry = 2; 122 } else { 123 atomic_set(&ipvs->dropentry, 0); 124 } 125 break; 126 case 2: 127 if (nomem) { 128 atomic_set(&ipvs->dropentry, 1); 129 } else { 130 atomic_set(&ipvs->dropentry, 0); 131 ipvs->sysctl_drop_entry = 1; 132 } 133 break; 134 case 3: 135 atomic_set(&ipvs->dropentry, 1); 136 break; 137 } 138 spin_unlock(&ipvs->dropentry_lock); 139 140 /* drop_packet */ 141 spin_lock(&ipvs->droppacket_lock); 142 switch (ipvs->sysctl_drop_packet) { 143 case 0: 144 ipvs->drop_rate = 0; 145 break; 146 case 1: 147 if (nomem) { 148 ipvs->drop_rate = ipvs->drop_counter 149 = ipvs->sysctl_amemthresh / 150 (ipvs->sysctl_amemthresh-availmem); 151 ipvs->sysctl_drop_packet = 2; 152 } else { 153 ipvs->drop_rate = 0; 154 } 155 break; 156 case 2: 157 if (nomem) { 158 ipvs->drop_rate = ipvs->drop_counter 159 = ipvs->sysctl_amemthresh / 160 (ipvs->sysctl_amemthresh-availmem); 161 } else { 162 ipvs->drop_rate = 0; 163 ipvs->sysctl_drop_packet = 1; 164 } 165 break; 166 case 3: 167 ipvs->drop_rate = ipvs->sysctl_am_droprate; 168 break; 169 } 170 spin_unlock(&ipvs->droppacket_lock); 171 172 /* secure_tcp */ 173 spin_lock(&ipvs->securetcp_lock); 174 switch (ipvs->sysctl_secure_tcp) { 175 case 0: 176 if (ipvs->old_secure_tcp >= 2) 177 to_change = 0; 178 break; 179 case 1: 180 if (nomem) { 181 if (ipvs->old_secure_tcp < 2) 182 to_change = 1; 183 ipvs->sysctl_secure_tcp = 2; 184 } else { 185 if (ipvs->old_secure_tcp >= 2) 186 to_change = 0; 187 } 188 break; 189 case 2: 190 if (nomem) { 191 if (ipvs->old_secure_tcp < 2) 192 to_change = 1; 193 } else { 194 if (ipvs->old_secure_tcp >= 2) 195 to_change = 0; 196 ipvs->sysctl_secure_tcp = 1; 197 } 198 break; 199 case 3: 200 if (ipvs->old_secure_tcp < 2) 201 to_change = 1; 202 break; 203 } 204 ipvs->old_secure_tcp = ipvs->sysctl_secure_tcp; 205 if (to_change >= 0) 206 ip_vs_protocol_timeout_change(ipvs, 207 ipvs->sysctl_secure_tcp > 1); 208 spin_unlock(&ipvs->securetcp_lock); 209 210 local_bh_enable(); 211 } 212 213 /* Handler for delayed work for expiring no 214 * destination connections 215 */ 216 static void expire_nodest_conn_handler(struct work_struct *work) 217 { 218 struct netns_ipvs *ipvs; 219 220 ipvs = container_of(work, struct netns_ipvs, 221 expire_nodest_conn_work.work); 222 ip_vs_expire_nodest_conn_flush(ipvs); 223 } 224 225 /* 226 * Timer for checking the defense 227 */ 228 #define DEFENSE_TIMER_PERIOD 1*HZ 229 230 static void defense_work_handler(struct work_struct *work) 231 { 232 struct netns_ipvs *ipvs = 233 container_of(work, struct netns_ipvs, defense_work.work); 234 235 update_defense_level(ipvs); 236 if (atomic_read(&ipvs->dropentry)) 237 ip_vs_random_dropentry(ipvs); 238 queue_delayed_work(system_long_wq, &ipvs->defense_work, 239 DEFENSE_TIMER_PERIOD); 240 } 241 #endif 242 243 static void est_reload_work_handler(struct work_struct *work) 244 { 245 struct netns_ipvs *ipvs = 246 container_of(work, struct netns_ipvs, est_reload_work.work); 247 int genid_done = atomic_read(&ipvs->est_genid_done); 248 unsigned long delay = HZ / 10; /* repeat startups after failure */ 249 bool repeat = false; 250 int genid; 251 int id; 252 253 mutex_lock(&ipvs->est_mutex); 254 genid = atomic_read(&ipvs->est_genid); 255 for (id = 0; id < ipvs->est_kt_count; id++) { 256 struct ip_vs_est_kt_data *kd = ipvs->est_kt_arr[id]; 257 258 /* netns clean up started, abort delayed work */ 259 if (!ipvs->enable) 260 goto unlock; 261 if (!kd) 262 continue; 263 /* New config ? Stop kthread tasks */ 264 if (genid != genid_done) 265 ip_vs_est_kthread_stop(kd); 266 if (!kd->task && !ip_vs_est_stopped(ipvs)) { 267 /* Do not start kthreads above 0 in calc phase */ 268 if ((!id || !ipvs->est_calc_phase) && 269 ip_vs_est_kthread_start(ipvs, kd) < 0) 270 repeat = true; 271 } 272 } 273 274 atomic_set(&ipvs->est_genid_done, genid); 275 276 if (repeat) 277 queue_delayed_work(system_long_wq, &ipvs->est_reload_work, 278 delay); 279 280 unlock: 281 mutex_unlock(&ipvs->est_mutex); 282 } 283 284 int 285 ip_vs_use_count_inc(void) 286 { 287 return try_module_get(THIS_MODULE); 288 } 289 290 void 291 ip_vs_use_count_dec(void) 292 { 293 module_put(THIS_MODULE); 294 } 295 296 297 /* 298 * Hash table: for virtual service lookups 299 */ 300 #define IP_VS_SVC_TAB_BITS 8 301 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS) 302 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1) 303 304 /* the service table hashed by <protocol, addr, port> */ 305 static struct hlist_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE]; 306 /* the service table hashed by fwmark */ 307 static struct hlist_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE]; 308 309 310 /* 311 * Returns hash value for virtual service 312 */ 313 static inline unsigned int 314 ip_vs_svc_hashkey(struct netns_ipvs *ipvs, int af, unsigned int proto, 315 const union nf_inet_addr *addr, __be16 port) 316 { 317 unsigned int porth = ntohs(port); 318 __be32 addr_fold = addr->ip; 319 __u32 ahash; 320 321 #ifdef CONFIG_IP_VS_IPV6 322 if (af == AF_INET6) 323 addr_fold = addr->ip6[0]^addr->ip6[1]^ 324 addr->ip6[2]^addr->ip6[3]; 325 #endif 326 ahash = ntohl(addr_fold); 327 ahash ^= ((size_t) ipvs >> 8); 328 329 return (proto ^ ahash ^ (porth >> IP_VS_SVC_TAB_BITS) ^ porth) & 330 IP_VS_SVC_TAB_MASK; 331 } 332 333 /* 334 * Returns hash value of fwmark for virtual service lookup 335 */ 336 static inline unsigned int ip_vs_svc_fwm_hashkey(struct netns_ipvs *ipvs, __u32 fwmark) 337 { 338 return (((size_t)ipvs>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK; 339 } 340 341 /* 342 * Hashes a service in the ip_vs_svc_table by <netns,proto,addr,port> 343 * or in the ip_vs_svc_fwm_table by fwmark. 344 * Should be called with locked tables. 345 */ 346 static int ip_vs_svc_hash(struct ip_vs_service *svc) 347 { 348 unsigned int hash; 349 350 if (svc->flags & IP_VS_SVC_F_HASHED) { 351 pr_err("%s(): request for already hashed, called from %pS\n", 352 __func__, __builtin_return_address(0)); 353 return 0; 354 } 355 356 if (svc->fwmark == 0) { 357 /* 358 * Hash it by <netns,protocol,addr,port> in ip_vs_svc_table 359 */ 360 hash = ip_vs_svc_hashkey(svc->ipvs, svc->af, svc->protocol, 361 &svc->addr, svc->port); 362 hlist_add_head_rcu(&svc->s_list, &ip_vs_svc_table[hash]); 363 } else { 364 /* 365 * Hash it by fwmark in svc_fwm_table 366 */ 367 hash = ip_vs_svc_fwm_hashkey(svc->ipvs, svc->fwmark); 368 hlist_add_head_rcu(&svc->f_list, &ip_vs_svc_fwm_table[hash]); 369 } 370 371 svc->flags |= IP_VS_SVC_F_HASHED; 372 /* increase its refcnt because it is referenced by the svc table */ 373 atomic_inc(&svc->refcnt); 374 return 1; 375 } 376 377 378 /* 379 * Unhashes a service from svc_table / svc_fwm_table. 380 * Should be called with locked tables. 381 */ 382 static int ip_vs_svc_unhash(struct ip_vs_service *svc) 383 { 384 if (!(svc->flags & IP_VS_SVC_F_HASHED)) { 385 pr_err("%s(): request for unhash flagged, called from %pS\n", 386 __func__, __builtin_return_address(0)); 387 return 0; 388 } 389 390 if (svc->fwmark == 0) { 391 /* Remove it from the svc_table table */ 392 hlist_del_rcu(&svc->s_list); 393 } else { 394 /* Remove it from the svc_fwm_table table */ 395 hlist_del_rcu(&svc->f_list); 396 } 397 398 svc->flags &= ~IP_VS_SVC_F_HASHED; 399 atomic_dec(&svc->refcnt); 400 return 1; 401 } 402 403 404 /* 405 * Get service by {netns, proto,addr,port} in the service table. 406 */ 407 static inline struct ip_vs_service * 408 __ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u16 protocol, 409 const union nf_inet_addr *vaddr, __be16 vport) 410 { 411 unsigned int hash; 412 struct ip_vs_service *svc; 413 414 /* Check for "full" addressed entries */ 415 hash = ip_vs_svc_hashkey(ipvs, af, protocol, vaddr, vport); 416 417 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[hash], s_list) { 418 if ((svc->af == af) 419 && ip_vs_addr_equal(af, &svc->addr, vaddr) 420 && (svc->port == vport) 421 && (svc->protocol == protocol) 422 && (svc->ipvs == ipvs)) { 423 /* HIT */ 424 return svc; 425 } 426 } 427 428 return NULL; 429 } 430 431 432 /* 433 * Get service by {fwmark} in the service table. 434 */ 435 static inline struct ip_vs_service * 436 __ip_vs_svc_fwm_find(struct netns_ipvs *ipvs, int af, __u32 fwmark) 437 { 438 unsigned int hash; 439 struct ip_vs_service *svc; 440 441 /* Check for fwmark addressed entries */ 442 hash = ip_vs_svc_fwm_hashkey(ipvs, fwmark); 443 444 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[hash], f_list) { 445 if (svc->fwmark == fwmark && svc->af == af 446 && (svc->ipvs == ipvs)) { 447 /* HIT */ 448 return svc; 449 } 450 } 451 452 return NULL; 453 } 454 455 /* Find service, called under RCU lock */ 456 struct ip_vs_service * 457 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol, 458 const union nf_inet_addr *vaddr, __be16 vport) 459 { 460 struct ip_vs_service *svc; 461 462 /* 463 * Check the table hashed by fwmark first 464 */ 465 if (fwmark) { 466 svc = __ip_vs_svc_fwm_find(ipvs, af, fwmark); 467 if (svc) 468 goto out; 469 } 470 471 /* 472 * Check the table hashed by <protocol,addr,port> 473 * for "full" addressed entries 474 */ 475 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, vport); 476 477 if (!svc && protocol == IPPROTO_TCP && 478 atomic_read(&ipvs->ftpsvc_counter) && 479 (vport == FTPDATA || !inet_port_requires_bind_service(ipvs->net, ntohs(vport)))) { 480 /* 481 * Check if ftp service entry exists, the packet 482 * might belong to FTP data connections. 483 */ 484 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, FTPPORT); 485 } 486 487 if (svc == NULL 488 && atomic_read(&ipvs->nullsvc_counter)) { 489 /* 490 * Check if the catch-all port (port zero) exists 491 */ 492 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, 0); 493 } 494 495 out: 496 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n", 497 fwmark, ip_vs_proto_name(protocol), 498 IP_VS_DBG_ADDR(af, vaddr), ntohs(vport), 499 svc ? "hit" : "not hit"); 500 501 return svc; 502 } 503 504 505 static inline void 506 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc) 507 { 508 atomic_inc(&svc->refcnt); 509 rcu_assign_pointer(dest->svc, svc); 510 } 511 512 static void ip_vs_service_free(struct ip_vs_service *svc) 513 { 514 ip_vs_stats_release(&svc->stats); 515 kfree(svc); 516 } 517 518 static void ip_vs_service_rcu_free(struct rcu_head *head) 519 { 520 struct ip_vs_service *svc; 521 522 svc = container_of(head, struct ip_vs_service, rcu_head); 523 ip_vs_service_free(svc); 524 } 525 526 static void __ip_vs_svc_put(struct ip_vs_service *svc) 527 { 528 if (atomic_dec_and_test(&svc->refcnt)) { 529 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n", 530 svc->fwmark, 531 IP_VS_DBG_ADDR(svc->af, &svc->addr), 532 ntohs(svc->port)); 533 call_rcu(&svc->rcu_head, ip_vs_service_rcu_free); 534 } 535 } 536 537 538 /* 539 * Returns hash value for real service 540 */ 541 static inline unsigned int ip_vs_rs_hashkey(int af, 542 const union nf_inet_addr *addr, 543 __be16 port) 544 { 545 unsigned int porth = ntohs(port); 546 __be32 addr_fold = addr->ip; 547 548 #ifdef CONFIG_IP_VS_IPV6 549 if (af == AF_INET6) 550 addr_fold = addr->ip6[0]^addr->ip6[1]^ 551 addr->ip6[2]^addr->ip6[3]; 552 #endif 553 554 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth) 555 & IP_VS_RTAB_MASK; 556 } 557 558 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */ 559 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest) 560 { 561 unsigned int hash; 562 __be16 port; 563 564 if (dest->in_rs_table) 565 return; 566 567 switch (IP_VS_DFWD_METHOD(dest)) { 568 case IP_VS_CONN_F_MASQ: 569 port = dest->port; 570 break; 571 case IP_VS_CONN_F_TUNNEL: 572 switch (dest->tun_type) { 573 case IP_VS_CONN_F_TUNNEL_TYPE_GUE: 574 port = dest->tun_port; 575 break; 576 case IP_VS_CONN_F_TUNNEL_TYPE_IPIP: 577 case IP_VS_CONN_F_TUNNEL_TYPE_GRE: 578 port = 0; 579 break; 580 default: 581 return; 582 } 583 break; 584 default: 585 return; 586 } 587 588 /* 589 * Hash by proto,addr,port, 590 * which are the parameters of the real service. 591 */ 592 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, port); 593 594 hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]); 595 dest->in_rs_table = 1; 596 } 597 598 /* Unhash ip_vs_dest from rs_table. */ 599 static void ip_vs_rs_unhash(struct ip_vs_dest *dest) 600 { 601 /* 602 * Remove it from the rs_table table. 603 */ 604 if (dest->in_rs_table) { 605 hlist_del_rcu(&dest->d_list); 606 dest->in_rs_table = 0; 607 } 608 } 609 610 /* Check if real service by <proto,addr,port> is present */ 611 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol, 612 const union nf_inet_addr *daddr, __be16 dport) 613 { 614 unsigned int hash; 615 struct ip_vs_dest *dest; 616 617 /* Check for "full" addressed entries */ 618 hash = ip_vs_rs_hashkey(af, daddr, dport); 619 620 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) { 621 if (dest->port == dport && 622 dest->af == af && 623 ip_vs_addr_equal(af, &dest->addr, daddr) && 624 (dest->protocol == protocol || dest->vfwmark) && 625 IP_VS_DFWD_METHOD(dest) == IP_VS_CONN_F_MASQ) { 626 /* HIT */ 627 return true; 628 } 629 } 630 631 return false; 632 } 633 634 /* Find real service record by <proto,addr,port>. 635 * In case of multiple records with the same <proto,addr,port>, only 636 * the first found record is returned. 637 * 638 * To be called under RCU lock. 639 */ 640 struct ip_vs_dest *ip_vs_find_real_service(struct netns_ipvs *ipvs, int af, 641 __u16 protocol, 642 const union nf_inet_addr *daddr, 643 __be16 dport) 644 { 645 unsigned int hash; 646 struct ip_vs_dest *dest; 647 648 /* Check for "full" addressed entries */ 649 hash = ip_vs_rs_hashkey(af, daddr, dport); 650 651 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) { 652 if (dest->port == dport && 653 dest->af == af && 654 ip_vs_addr_equal(af, &dest->addr, daddr) && 655 (dest->protocol == protocol || dest->vfwmark) && 656 IP_VS_DFWD_METHOD(dest) == IP_VS_CONN_F_MASQ) { 657 /* HIT */ 658 return dest; 659 } 660 } 661 662 return NULL; 663 } 664 665 /* Find real service record by <af,addr,tun_port>. 666 * In case of multiple records with the same <af,addr,tun_port>, only 667 * the first found record is returned. 668 * 669 * To be called under RCU lock. 670 */ 671 struct ip_vs_dest *ip_vs_find_tunnel(struct netns_ipvs *ipvs, int af, 672 const union nf_inet_addr *daddr, 673 __be16 tun_port) 674 { 675 struct ip_vs_dest *dest; 676 unsigned int hash; 677 678 /* Check for "full" addressed entries */ 679 hash = ip_vs_rs_hashkey(af, daddr, tun_port); 680 681 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) { 682 if (dest->tun_port == tun_port && 683 dest->af == af && 684 ip_vs_addr_equal(af, &dest->addr, daddr) && 685 IP_VS_DFWD_METHOD(dest) == IP_VS_CONN_F_TUNNEL) { 686 /* HIT */ 687 return dest; 688 } 689 } 690 691 return NULL; 692 } 693 694 /* Lookup destination by {addr,port} in the given service 695 * Called under RCU lock. 696 */ 697 static struct ip_vs_dest * 698 ip_vs_lookup_dest(struct ip_vs_service *svc, int dest_af, 699 const union nf_inet_addr *daddr, __be16 dport) 700 { 701 struct ip_vs_dest *dest; 702 703 /* 704 * Find the destination for the given service 705 */ 706 list_for_each_entry_rcu(dest, &svc->destinations, n_list) { 707 if ((dest->af == dest_af) && 708 ip_vs_addr_equal(dest_af, &dest->addr, daddr) && 709 (dest->port == dport)) { 710 /* HIT */ 711 return dest; 712 } 713 } 714 715 return NULL; 716 } 717 718 /* 719 * Find destination by {daddr,dport,vaddr,protocol} 720 * Created to be used in ip_vs_process_message() in 721 * the backup synchronization daemon. It finds the 722 * destination to be bound to the received connection 723 * on the backup. 724 * Called under RCU lock, no refcnt is returned. 725 */ 726 struct ip_vs_dest *ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af, 727 const union nf_inet_addr *daddr, 728 __be16 dport, 729 const union nf_inet_addr *vaddr, 730 __be16 vport, __u16 protocol, __u32 fwmark, 731 __u32 flags) 732 { 733 struct ip_vs_dest *dest; 734 struct ip_vs_service *svc; 735 __be16 port = dport; 736 737 svc = ip_vs_service_find(ipvs, svc_af, fwmark, protocol, vaddr, vport); 738 if (!svc) 739 return NULL; 740 if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) 741 port = 0; 742 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port); 743 if (!dest) 744 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port ^ dport); 745 return dest; 746 } 747 748 void ip_vs_dest_dst_rcu_free(struct rcu_head *head) 749 { 750 struct ip_vs_dest_dst *dest_dst = container_of(head, 751 struct ip_vs_dest_dst, 752 rcu_head); 753 754 dst_release(dest_dst->dst_cache); 755 kfree(dest_dst); 756 } 757 758 /* Release dest_dst and dst_cache for dest in user context */ 759 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest) 760 { 761 struct ip_vs_dest_dst *old; 762 763 old = rcu_dereference_protected(dest->dest_dst, 1); 764 if (old) { 765 RCU_INIT_POINTER(dest->dest_dst, NULL); 766 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free); 767 } 768 } 769 770 /* 771 * Lookup dest by {svc,addr,port} in the destination trash. 772 * The destination trash is used to hold the destinations that are removed 773 * from the service table but are still referenced by some conn entries. 774 * The reason to add the destination trash is when the dest is temporary 775 * down (either by administrator or by monitor program), the dest can be 776 * picked back from the trash, the remaining connections to the dest can 777 * continue, and the counting information of the dest is also useful for 778 * scheduling. 779 */ 780 static struct ip_vs_dest * 781 ip_vs_trash_get_dest(struct ip_vs_service *svc, int dest_af, 782 const union nf_inet_addr *daddr, __be16 dport) 783 { 784 struct ip_vs_dest *dest; 785 struct netns_ipvs *ipvs = svc->ipvs; 786 787 /* 788 * Find the destination in trash 789 */ 790 spin_lock_bh(&ipvs->dest_trash_lock); 791 list_for_each_entry(dest, &ipvs->dest_trash, t_list) { 792 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, " 793 "dest->refcnt=%d\n", 794 dest->vfwmark, 795 IP_VS_DBG_ADDR(dest->af, &dest->addr), 796 ntohs(dest->port), 797 refcount_read(&dest->refcnt)); 798 if (dest->af == dest_af && 799 ip_vs_addr_equal(dest_af, &dest->addr, daddr) && 800 dest->port == dport && 801 dest->vfwmark == svc->fwmark && 802 dest->protocol == svc->protocol && 803 (svc->fwmark || 804 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) && 805 dest->vport == svc->port))) { 806 /* HIT */ 807 list_del(&dest->t_list); 808 goto out; 809 } 810 } 811 812 dest = NULL; 813 814 out: 815 spin_unlock_bh(&ipvs->dest_trash_lock); 816 817 return dest; 818 } 819 820 static void ip_vs_dest_rcu_free(struct rcu_head *head) 821 { 822 struct ip_vs_dest *dest; 823 824 dest = container_of(head, struct ip_vs_dest, rcu_head); 825 ip_vs_stats_release(&dest->stats); 826 ip_vs_dest_put_and_free(dest); 827 } 828 829 static void ip_vs_dest_free(struct ip_vs_dest *dest) 830 { 831 struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1); 832 833 __ip_vs_dst_cache_reset(dest); 834 __ip_vs_svc_put(svc); 835 call_rcu(&dest->rcu_head, ip_vs_dest_rcu_free); 836 } 837 838 /* 839 * Clean up all the destinations in the trash 840 * Called by the ip_vs_control_cleanup() 841 * 842 * When the ip_vs_control_clearup is activated by ipvs module exit, 843 * the service tables must have been flushed and all the connections 844 * are expired, and the refcnt of each destination in the trash must 845 * be 1, so we simply release them here. 846 */ 847 static void ip_vs_trash_cleanup(struct netns_ipvs *ipvs) 848 { 849 struct ip_vs_dest *dest, *nxt; 850 851 del_timer_sync(&ipvs->dest_trash_timer); 852 /* No need to use dest_trash_lock */ 853 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) { 854 list_del(&dest->t_list); 855 ip_vs_dest_free(dest); 856 } 857 } 858 859 static void ip_vs_stats_rcu_free(struct rcu_head *head) 860 { 861 struct ip_vs_stats_rcu *rs = container_of(head, 862 struct ip_vs_stats_rcu, 863 rcu_head); 864 865 ip_vs_stats_release(&rs->s); 866 kfree(rs); 867 } 868 869 static void 870 ip_vs_copy_stats(struct ip_vs_kstats *dst, struct ip_vs_stats *src) 871 { 872 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c 873 874 spin_lock(&src->lock); 875 876 IP_VS_SHOW_STATS_COUNTER(conns); 877 IP_VS_SHOW_STATS_COUNTER(inpkts); 878 IP_VS_SHOW_STATS_COUNTER(outpkts); 879 IP_VS_SHOW_STATS_COUNTER(inbytes); 880 IP_VS_SHOW_STATS_COUNTER(outbytes); 881 882 ip_vs_read_estimator(dst, src); 883 884 spin_unlock(&src->lock); 885 } 886 887 static void 888 ip_vs_export_stats_user(struct ip_vs_stats_user *dst, struct ip_vs_kstats *src) 889 { 890 dst->conns = (u32)src->conns; 891 dst->inpkts = (u32)src->inpkts; 892 dst->outpkts = (u32)src->outpkts; 893 dst->inbytes = src->inbytes; 894 dst->outbytes = src->outbytes; 895 dst->cps = (u32)src->cps; 896 dst->inpps = (u32)src->inpps; 897 dst->outpps = (u32)src->outpps; 898 dst->inbps = (u32)src->inbps; 899 dst->outbps = (u32)src->outbps; 900 } 901 902 static void 903 ip_vs_zero_stats(struct ip_vs_stats *stats) 904 { 905 spin_lock(&stats->lock); 906 907 /* get current counters as zero point, rates are zeroed */ 908 909 #define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c 910 911 IP_VS_ZERO_STATS_COUNTER(conns); 912 IP_VS_ZERO_STATS_COUNTER(inpkts); 913 IP_VS_ZERO_STATS_COUNTER(outpkts); 914 IP_VS_ZERO_STATS_COUNTER(inbytes); 915 IP_VS_ZERO_STATS_COUNTER(outbytes); 916 917 ip_vs_zero_estimator(stats); 918 919 spin_unlock(&stats->lock); 920 } 921 922 /* Allocate fields after kzalloc */ 923 int ip_vs_stats_init_alloc(struct ip_vs_stats *s) 924 { 925 int i; 926 927 spin_lock_init(&s->lock); 928 s->cpustats = alloc_percpu(struct ip_vs_cpu_stats); 929 if (!s->cpustats) 930 return -ENOMEM; 931 932 for_each_possible_cpu(i) { 933 struct ip_vs_cpu_stats *cs = per_cpu_ptr(s->cpustats, i); 934 935 u64_stats_init(&cs->syncp); 936 } 937 return 0; 938 } 939 940 struct ip_vs_stats *ip_vs_stats_alloc(void) 941 { 942 struct ip_vs_stats *s = kzalloc(sizeof(*s), GFP_KERNEL); 943 944 if (s && ip_vs_stats_init_alloc(s) >= 0) 945 return s; 946 kfree(s); 947 return NULL; 948 } 949 950 void ip_vs_stats_release(struct ip_vs_stats *stats) 951 { 952 free_percpu(stats->cpustats); 953 } 954 955 void ip_vs_stats_free(struct ip_vs_stats *stats) 956 { 957 if (stats) { 958 ip_vs_stats_release(stats); 959 kfree(stats); 960 } 961 } 962 963 /* 964 * Update a destination in the given service 965 */ 966 static void 967 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest, 968 struct ip_vs_dest_user_kern *udest, int add) 969 { 970 struct netns_ipvs *ipvs = svc->ipvs; 971 struct ip_vs_service *old_svc; 972 struct ip_vs_scheduler *sched; 973 int conn_flags; 974 975 /* We cannot modify an address and change the address family */ 976 BUG_ON(!add && udest->af != dest->af); 977 978 if (add && udest->af != svc->af) 979 ipvs->mixed_address_family_dests++; 980 981 /* keep the last_weight with latest non-0 weight */ 982 if (add || udest->weight != 0) 983 atomic_set(&dest->last_weight, udest->weight); 984 985 /* set the weight and the flags */ 986 atomic_set(&dest->weight, udest->weight); 987 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK; 988 conn_flags |= IP_VS_CONN_F_INACTIVE; 989 990 /* Need to rehash? */ 991 if ((udest->conn_flags & IP_VS_CONN_F_FWD_MASK) != 992 IP_VS_DFWD_METHOD(dest) || 993 udest->tun_type != dest->tun_type || 994 udest->tun_port != dest->tun_port) 995 ip_vs_rs_unhash(dest); 996 997 /* set the tunnel info */ 998 dest->tun_type = udest->tun_type; 999 dest->tun_port = udest->tun_port; 1000 dest->tun_flags = udest->tun_flags; 1001 1002 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */ 1003 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) { 1004 conn_flags |= IP_VS_CONN_F_NOOUTPUT; 1005 } else { 1006 /* FTP-NAT requires conntrack for mangling */ 1007 if (svc->port == FTPPORT) 1008 ip_vs_register_conntrack(svc); 1009 } 1010 atomic_set(&dest->conn_flags, conn_flags); 1011 /* Put the real service in rs_table if not present. */ 1012 ip_vs_rs_hash(ipvs, dest); 1013 1014 /* bind the service */ 1015 old_svc = rcu_dereference_protected(dest->svc, 1); 1016 if (!old_svc) { 1017 __ip_vs_bind_svc(dest, svc); 1018 } else { 1019 if (old_svc != svc) { 1020 ip_vs_zero_stats(&dest->stats); 1021 __ip_vs_bind_svc(dest, svc); 1022 __ip_vs_svc_put(old_svc); 1023 } 1024 } 1025 1026 /* set the dest status flags */ 1027 dest->flags |= IP_VS_DEST_F_AVAILABLE; 1028 1029 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold) 1030 dest->flags &= ~IP_VS_DEST_F_OVERLOAD; 1031 dest->u_threshold = udest->u_threshold; 1032 dest->l_threshold = udest->l_threshold; 1033 1034 dest->af = udest->af; 1035 1036 spin_lock_bh(&dest->dst_lock); 1037 __ip_vs_dst_cache_reset(dest); 1038 spin_unlock_bh(&dest->dst_lock); 1039 1040 if (add) { 1041 list_add_rcu(&dest->n_list, &svc->destinations); 1042 svc->num_dests++; 1043 sched = rcu_dereference_protected(svc->scheduler, 1); 1044 if (sched && sched->add_dest) 1045 sched->add_dest(svc, dest); 1046 } else { 1047 sched = rcu_dereference_protected(svc->scheduler, 1); 1048 if (sched && sched->upd_dest) 1049 sched->upd_dest(svc, dest); 1050 } 1051 } 1052 1053 1054 /* 1055 * Create a destination for the given service 1056 */ 1057 static int 1058 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest) 1059 { 1060 struct ip_vs_dest *dest; 1061 unsigned int atype; 1062 int ret; 1063 1064 #ifdef CONFIG_IP_VS_IPV6 1065 if (udest->af == AF_INET6) { 1066 atype = ipv6_addr_type(&udest->addr.in6); 1067 if ((!(atype & IPV6_ADDR_UNICAST) || 1068 atype & IPV6_ADDR_LINKLOCAL) && 1069 !__ip_vs_addr_is_local_v6(svc->ipvs->net, &udest->addr.in6)) 1070 return -EINVAL; 1071 1072 ret = nf_defrag_ipv6_enable(svc->ipvs->net); 1073 if (ret) 1074 return ret; 1075 } else 1076 #endif 1077 { 1078 atype = inet_addr_type(svc->ipvs->net, udest->addr.ip); 1079 if (atype != RTN_LOCAL && atype != RTN_UNICAST) 1080 return -EINVAL; 1081 } 1082 1083 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL); 1084 if (dest == NULL) 1085 return -ENOMEM; 1086 1087 ret = ip_vs_stats_init_alloc(&dest->stats); 1088 if (ret < 0) 1089 goto err_alloc; 1090 1091 ret = ip_vs_start_estimator(svc->ipvs, &dest->stats); 1092 if (ret < 0) 1093 goto err_stats; 1094 1095 dest->af = udest->af; 1096 dest->protocol = svc->protocol; 1097 dest->vaddr = svc->addr; 1098 dest->vport = svc->port; 1099 dest->vfwmark = svc->fwmark; 1100 ip_vs_addr_copy(udest->af, &dest->addr, &udest->addr); 1101 dest->port = udest->port; 1102 1103 atomic_set(&dest->activeconns, 0); 1104 atomic_set(&dest->inactconns, 0); 1105 atomic_set(&dest->persistconns, 0); 1106 refcount_set(&dest->refcnt, 1); 1107 1108 INIT_HLIST_NODE(&dest->d_list); 1109 spin_lock_init(&dest->dst_lock); 1110 __ip_vs_update_dest(svc, dest, udest, 1); 1111 1112 return 0; 1113 1114 err_stats: 1115 ip_vs_stats_release(&dest->stats); 1116 1117 err_alloc: 1118 kfree(dest); 1119 return ret; 1120 } 1121 1122 1123 /* 1124 * Add a destination into an existing service 1125 */ 1126 static int 1127 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest) 1128 { 1129 struct ip_vs_dest *dest; 1130 union nf_inet_addr daddr; 1131 __be16 dport = udest->port; 1132 int ret; 1133 1134 if (udest->weight < 0) { 1135 pr_err("%s(): server weight less than zero\n", __func__); 1136 return -ERANGE; 1137 } 1138 1139 if (udest->l_threshold > udest->u_threshold) { 1140 pr_err("%s(): lower threshold is higher than upper threshold\n", 1141 __func__); 1142 return -ERANGE; 1143 } 1144 1145 if (udest->tun_type == IP_VS_CONN_F_TUNNEL_TYPE_GUE) { 1146 if (udest->tun_port == 0) { 1147 pr_err("%s(): tunnel port is zero\n", __func__); 1148 return -EINVAL; 1149 } 1150 } 1151 1152 ip_vs_addr_copy(udest->af, &daddr, &udest->addr); 1153 1154 /* We use function that requires RCU lock */ 1155 rcu_read_lock(); 1156 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport); 1157 rcu_read_unlock(); 1158 1159 if (dest != NULL) { 1160 IP_VS_DBG(1, "%s(): dest already exists\n", __func__); 1161 return -EEXIST; 1162 } 1163 1164 /* 1165 * Check if the dest already exists in the trash and 1166 * is from the same service 1167 */ 1168 dest = ip_vs_trash_get_dest(svc, udest->af, &daddr, dport); 1169 1170 if (dest != NULL) { 1171 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, " 1172 "dest->refcnt=%d, service %u/%s:%u\n", 1173 IP_VS_DBG_ADDR(udest->af, &daddr), ntohs(dport), 1174 refcount_read(&dest->refcnt), 1175 dest->vfwmark, 1176 IP_VS_DBG_ADDR(svc->af, &dest->vaddr), 1177 ntohs(dest->vport)); 1178 1179 ret = ip_vs_start_estimator(svc->ipvs, &dest->stats); 1180 if (ret < 0) 1181 return ret; 1182 __ip_vs_update_dest(svc, dest, udest, 1); 1183 } else { 1184 /* 1185 * Allocate and initialize the dest structure 1186 */ 1187 ret = ip_vs_new_dest(svc, udest); 1188 } 1189 1190 return ret; 1191 } 1192 1193 1194 /* 1195 * Edit a destination in the given service 1196 */ 1197 static int 1198 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest) 1199 { 1200 struct ip_vs_dest *dest; 1201 union nf_inet_addr daddr; 1202 __be16 dport = udest->port; 1203 1204 if (udest->weight < 0) { 1205 pr_err("%s(): server weight less than zero\n", __func__); 1206 return -ERANGE; 1207 } 1208 1209 if (udest->l_threshold > udest->u_threshold) { 1210 pr_err("%s(): lower threshold is higher than upper threshold\n", 1211 __func__); 1212 return -ERANGE; 1213 } 1214 1215 if (udest->tun_type == IP_VS_CONN_F_TUNNEL_TYPE_GUE) { 1216 if (udest->tun_port == 0) { 1217 pr_err("%s(): tunnel port is zero\n", __func__); 1218 return -EINVAL; 1219 } 1220 } 1221 1222 ip_vs_addr_copy(udest->af, &daddr, &udest->addr); 1223 1224 /* We use function that requires RCU lock */ 1225 rcu_read_lock(); 1226 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport); 1227 rcu_read_unlock(); 1228 1229 if (dest == NULL) { 1230 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__); 1231 return -ENOENT; 1232 } 1233 1234 __ip_vs_update_dest(svc, dest, udest, 0); 1235 1236 return 0; 1237 } 1238 1239 /* 1240 * Delete a destination (must be already unlinked from the service) 1241 */ 1242 static void __ip_vs_del_dest(struct netns_ipvs *ipvs, struct ip_vs_dest *dest, 1243 bool cleanup) 1244 { 1245 ip_vs_stop_estimator(ipvs, &dest->stats); 1246 1247 /* 1248 * Remove it from the d-linked list with the real services. 1249 */ 1250 ip_vs_rs_unhash(dest); 1251 1252 spin_lock_bh(&ipvs->dest_trash_lock); 1253 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n", 1254 IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port), 1255 refcount_read(&dest->refcnt)); 1256 if (list_empty(&ipvs->dest_trash) && !cleanup) 1257 mod_timer(&ipvs->dest_trash_timer, 1258 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1)); 1259 /* dest lives in trash with reference */ 1260 list_add(&dest->t_list, &ipvs->dest_trash); 1261 dest->idle_start = 0; 1262 spin_unlock_bh(&ipvs->dest_trash_lock); 1263 1264 /* Queue up delayed work to expire all no destination connections. 1265 * No-op when CONFIG_SYSCTL is disabled. 1266 */ 1267 if (!cleanup) 1268 ip_vs_enqueue_expire_nodest_conns(ipvs); 1269 } 1270 1271 1272 /* 1273 * Unlink a destination from the given service 1274 */ 1275 static void __ip_vs_unlink_dest(struct ip_vs_service *svc, 1276 struct ip_vs_dest *dest, 1277 int svcupd) 1278 { 1279 dest->flags &= ~IP_VS_DEST_F_AVAILABLE; 1280 1281 /* 1282 * Remove it from the d-linked destination list. 1283 */ 1284 list_del_rcu(&dest->n_list); 1285 svc->num_dests--; 1286 1287 if (dest->af != svc->af) 1288 svc->ipvs->mixed_address_family_dests--; 1289 1290 if (svcupd) { 1291 struct ip_vs_scheduler *sched; 1292 1293 sched = rcu_dereference_protected(svc->scheduler, 1); 1294 if (sched && sched->del_dest) 1295 sched->del_dest(svc, dest); 1296 } 1297 } 1298 1299 1300 /* 1301 * Delete a destination server in the given service 1302 */ 1303 static int 1304 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest) 1305 { 1306 struct ip_vs_dest *dest; 1307 __be16 dport = udest->port; 1308 1309 /* We use function that requires RCU lock */ 1310 rcu_read_lock(); 1311 dest = ip_vs_lookup_dest(svc, udest->af, &udest->addr, dport); 1312 rcu_read_unlock(); 1313 1314 if (dest == NULL) { 1315 IP_VS_DBG(1, "%s(): destination not found!\n", __func__); 1316 return -ENOENT; 1317 } 1318 1319 /* 1320 * Unlink dest from the service 1321 */ 1322 __ip_vs_unlink_dest(svc, dest, 1); 1323 1324 /* 1325 * Delete the destination 1326 */ 1327 __ip_vs_del_dest(svc->ipvs, dest, false); 1328 1329 return 0; 1330 } 1331 1332 static void ip_vs_dest_trash_expire(struct timer_list *t) 1333 { 1334 struct netns_ipvs *ipvs = from_timer(ipvs, t, dest_trash_timer); 1335 struct ip_vs_dest *dest, *next; 1336 unsigned long now = jiffies; 1337 1338 spin_lock(&ipvs->dest_trash_lock); 1339 list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) { 1340 if (refcount_read(&dest->refcnt) > 1) 1341 continue; 1342 if (dest->idle_start) { 1343 if (time_before(now, dest->idle_start + 1344 IP_VS_DEST_TRASH_PERIOD)) 1345 continue; 1346 } else { 1347 dest->idle_start = max(1UL, now); 1348 continue; 1349 } 1350 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n", 1351 dest->vfwmark, 1352 IP_VS_DBG_ADDR(dest->af, &dest->addr), 1353 ntohs(dest->port)); 1354 list_del(&dest->t_list); 1355 ip_vs_dest_free(dest); 1356 } 1357 if (!list_empty(&ipvs->dest_trash)) 1358 mod_timer(&ipvs->dest_trash_timer, 1359 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1)); 1360 spin_unlock(&ipvs->dest_trash_lock); 1361 } 1362 1363 /* 1364 * Add a service into the service hash table 1365 */ 1366 static int 1367 ip_vs_add_service(struct netns_ipvs *ipvs, struct ip_vs_service_user_kern *u, 1368 struct ip_vs_service **svc_p) 1369 { 1370 int ret = 0; 1371 struct ip_vs_scheduler *sched = NULL; 1372 struct ip_vs_pe *pe = NULL; 1373 struct ip_vs_service *svc = NULL; 1374 int ret_hooks = -1; 1375 1376 /* increase the module use count */ 1377 if (!ip_vs_use_count_inc()) 1378 return -ENOPROTOOPT; 1379 1380 /* Lookup the scheduler by 'u->sched_name' */ 1381 if (strcmp(u->sched_name, "none")) { 1382 sched = ip_vs_scheduler_get(u->sched_name); 1383 if (!sched) { 1384 pr_info("Scheduler module ip_vs_%s not found\n", 1385 u->sched_name); 1386 ret = -ENOENT; 1387 goto out_err; 1388 } 1389 } 1390 1391 if (u->pe_name && *u->pe_name) { 1392 pe = ip_vs_pe_getbyname(u->pe_name); 1393 if (pe == NULL) { 1394 pr_info("persistence engine module ip_vs_pe_%s " 1395 "not found\n", u->pe_name); 1396 ret = -ENOENT; 1397 goto out_err; 1398 } 1399 } 1400 1401 #ifdef CONFIG_IP_VS_IPV6 1402 if (u->af == AF_INET6) { 1403 __u32 plen = (__force __u32) u->netmask; 1404 1405 if (plen < 1 || plen > 128) { 1406 ret = -EINVAL; 1407 goto out_err; 1408 } 1409 1410 ret = nf_defrag_ipv6_enable(ipvs->net); 1411 if (ret) 1412 goto out_err; 1413 } 1414 #endif 1415 1416 if ((u->af == AF_INET && !ipvs->num_services) || 1417 (u->af == AF_INET6 && !ipvs->num_services6)) { 1418 ret = ip_vs_register_hooks(ipvs, u->af); 1419 if (ret < 0) 1420 goto out_err; 1421 ret_hooks = ret; 1422 } 1423 1424 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL); 1425 if (svc == NULL) { 1426 IP_VS_DBG(1, "%s(): no memory\n", __func__); 1427 ret = -ENOMEM; 1428 goto out_err; 1429 } 1430 ret = ip_vs_stats_init_alloc(&svc->stats); 1431 if (ret < 0) 1432 goto out_err; 1433 1434 /* I'm the first user of the service */ 1435 atomic_set(&svc->refcnt, 0); 1436 1437 svc->af = u->af; 1438 svc->protocol = u->protocol; 1439 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr); 1440 svc->port = u->port; 1441 svc->fwmark = u->fwmark; 1442 svc->flags = u->flags & ~IP_VS_SVC_F_HASHED; 1443 svc->timeout = u->timeout * HZ; 1444 svc->netmask = u->netmask; 1445 svc->ipvs = ipvs; 1446 1447 INIT_LIST_HEAD(&svc->destinations); 1448 spin_lock_init(&svc->sched_lock); 1449 1450 /* Bind the scheduler */ 1451 if (sched) { 1452 ret = ip_vs_bind_scheduler(svc, sched); 1453 if (ret) 1454 goto out_err; 1455 sched = NULL; 1456 } 1457 1458 ret = ip_vs_start_estimator(ipvs, &svc->stats); 1459 if (ret < 0) 1460 goto out_err; 1461 1462 /* Bind the ct retriever */ 1463 RCU_INIT_POINTER(svc->pe, pe); 1464 pe = NULL; 1465 1466 /* Update the virtual service counters */ 1467 if (svc->port == FTPPORT) 1468 atomic_inc(&ipvs->ftpsvc_counter); 1469 else if (svc->port == 0) 1470 atomic_inc(&ipvs->nullsvc_counter); 1471 if (svc->pe && svc->pe->conn_out) 1472 atomic_inc(&ipvs->conn_out_counter); 1473 1474 /* Count only IPv4 services for old get/setsockopt interface */ 1475 if (svc->af == AF_INET) 1476 ipvs->num_services++; 1477 else if (svc->af == AF_INET6) 1478 ipvs->num_services6++; 1479 1480 /* Hash the service into the service table */ 1481 ip_vs_svc_hash(svc); 1482 1483 *svc_p = svc; 1484 1485 if (!ipvs->enable) { 1486 /* Now there is a service - full throttle */ 1487 ipvs->enable = 1; 1488 1489 /* Start estimation for first time */ 1490 ip_vs_est_reload_start(ipvs); 1491 } 1492 1493 return 0; 1494 1495 1496 out_err: 1497 if (ret_hooks >= 0) 1498 ip_vs_unregister_hooks(ipvs, u->af); 1499 if (svc != NULL) { 1500 ip_vs_unbind_scheduler(svc, sched); 1501 ip_vs_service_free(svc); 1502 } 1503 ip_vs_scheduler_put(sched); 1504 ip_vs_pe_put(pe); 1505 1506 /* decrease the module use count */ 1507 ip_vs_use_count_dec(); 1508 1509 return ret; 1510 } 1511 1512 1513 /* 1514 * Edit a service and bind it with a new scheduler 1515 */ 1516 static int 1517 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u) 1518 { 1519 struct ip_vs_scheduler *sched = NULL, *old_sched; 1520 struct ip_vs_pe *pe = NULL, *old_pe = NULL; 1521 int ret = 0; 1522 bool new_pe_conn_out, old_pe_conn_out; 1523 1524 /* 1525 * Lookup the scheduler, by 'u->sched_name' 1526 */ 1527 if (strcmp(u->sched_name, "none")) { 1528 sched = ip_vs_scheduler_get(u->sched_name); 1529 if (!sched) { 1530 pr_info("Scheduler module ip_vs_%s not found\n", 1531 u->sched_name); 1532 return -ENOENT; 1533 } 1534 } 1535 old_sched = sched; 1536 1537 if (u->pe_name && *u->pe_name) { 1538 pe = ip_vs_pe_getbyname(u->pe_name); 1539 if (pe == NULL) { 1540 pr_info("persistence engine module ip_vs_pe_%s " 1541 "not found\n", u->pe_name); 1542 ret = -ENOENT; 1543 goto out; 1544 } 1545 old_pe = pe; 1546 } 1547 1548 #ifdef CONFIG_IP_VS_IPV6 1549 if (u->af == AF_INET6) { 1550 __u32 plen = (__force __u32) u->netmask; 1551 1552 if (plen < 1 || plen > 128) { 1553 ret = -EINVAL; 1554 goto out; 1555 } 1556 } 1557 #endif 1558 1559 old_sched = rcu_dereference_protected(svc->scheduler, 1); 1560 if (sched != old_sched) { 1561 if (old_sched) { 1562 ip_vs_unbind_scheduler(svc, old_sched); 1563 RCU_INIT_POINTER(svc->scheduler, NULL); 1564 /* Wait all svc->sched_data users */ 1565 synchronize_rcu(); 1566 } 1567 /* Bind the new scheduler */ 1568 if (sched) { 1569 ret = ip_vs_bind_scheduler(svc, sched); 1570 if (ret) { 1571 ip_vs_scheduler_put(sched); 1572 goto out; 1573 } 1574 } 1575 } 1576 1577 /* 1578 * Set the flags and timeout value 1579 */ 1580 svc->flags = u->flags | IP_VS_SVC_F_HASHED; 1581 svc->timeout = u->timeout * HZ; 1582 svc->netmask = u->netmask; 1583 1584 old_pe = rcu_dereference_protected(svc->pe, 1); 1585 if (pe != old_pe) { 1586 rcu_assign_pointer(svc->pe, pe); 1587 /* check for optional methods in new pe */ 1588 new_pe_conn_out = (pe && pe->conn_out) ? true : false; 1589 old_pe_conn_out = (old_pe && old_pe->conn_out) ? true : false; 1590 if (new_pe_conn_out && !old_pe_conn_out) 1591 atomic_inc(&svc->ipvs->conn_out_counter); 1592 if (old_pe_conn_out && !new_pe_conn_out) 1593 atomic_dec(&svc->ipvs->conn_out_counter); 1594 } 1595 1596 out: 1597 ip_vs_scheduler_put(old_sched); 1598 ip_vs_pe_put(old_pe); 1599 return ret; 1600 } 1601 1602 /* 1603 * Delete a service from the service list 1604 * - The service must be unlinked, unlocked and not referenced! 1605 * - We are called under _bh lock 1606 */ 1607 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup) 1608 { 1609 struct ip_vs_dest *dest, *nxt; 1610 struct ip_vs_scheduler *old_sched; 1611 struct ip_vs_pe *old_pe; 1612 struct netns_ipvs *ipvs = svc->ipvs; 1613 1614 if (svc->af == AF_INET) { 1615 ipvs->num_services--; 1616 if (!ipvs->num_services) 1617 ip_vs_unregister_hooks(ipvs, svc->af); 1618 } else if (svc->af == AF_INET6) { 1619 ipvs->num_services6--; 1620 if (!ipvs->num_services6) 1621 ip_vs_unregister_hooks(ipvs, svc->af); 1622 } 1623 1624 ip_vs_stop_estimator(svc->ipvs, &svc->stats); 1625 1626 /* Unbind scheduler */ 1627 old_sched = rcu_dereference_protected(svc->scheduler, 1); 1628 ip_vs_unbind_scheduler(svc, old_sched); 1629 ip_vs_scheduler_put(old_sched); 1630 1631 /* Unbind persistence engine, keep svc->pe */ 1632 old_pe = rcu_dereference_protected(svc->pe, 1); 1633 if (old_pe && old_pe->conn_out) 1634 atomic_dec(&ipvs->conn_out_counter); 1635 ip_vs_pe_put(old_pe); 1636 1637 /* 1638 * Unlink the whole destination list 1639 */ 1640 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) { 1641 __ip_vs_unlink_dest(svc, dest, 0); 1642 __ip_vs_del_dest(svc->ipvs, dest, cleanup); 1643 } 1644 1645 /* 1646 * Update the virtual service counters 1647 */ 1648 if (svc->port == FTPPORT) 1649 atomic_dec(&ipvs->ftpsvc_counter); 1650 else if (svc->port == 0) 1651 atomic_dec(&ipvs->nullsvc_counter); 1652 1653 /* 1654 * Free the service if nobody refers to it 1655 */ 1656 __ip_vs_svc_put(svc); 1657 1658 /* decrease the module use count */ 1659 ip_vs_use_count_dec(); 1660 } 1661 1662 /* 1663 * Unlink a service from list and try to delete it if its refcnt reached 0 1664 */ 1665 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup) 1666 { 1667 ip_vs_unregister_conntrack(svc); 1668 /* Hold svc to avoid double release from dest_trash */ 1669 atomic_inc(&svc->refcnt); 1670 /* 1671 * Unhash it from the service table 1672 */ 1673 ip_vs_svc_unhash(svc); 1674 1675 __ip_vs_del_service(svc, cleanup); 1676 } 1677 1678 /* 1679 * Delete a service from the service list 1680 */ 1681 static int ip_vs_del_service(struct ip_vs_service *svc) 1682 { 1683 if (svc == NULL) 1684 return -EEXIST; 1685 ip_vs_unlink_service(svc, false); 1686 1687 return 0; 1688 } 1689 1690 1691 /* 1692 * Flush all the virtual services 1693 */ 1694 static int ip_vs_flush(struct netns_ipvs *ipvs, bool cleanup) 1695 { 1696 int idx; 1697 struct ip_vs_service *svc; 1698 struct hlist_node *n; 1699 1700 /* 1701 * Flush the service table hashed by <netns,protocol,addr,port> 1702 */ 1703 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) { 1704 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx], 1705 s_list) { 1706 if (svc->ipvs == ipvs) 1707 ip_vs_unlink_service(svc, cleanup); 1708 } 1709 } 1710 1711 /* 1712 * Flush the service table hashed by fwmark 1713 */ 1714 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) { 1715 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx], 1716 f_list) { 1717 if (svc->ipvs == ipvs) 1718 ip_vs_unlink_service(svc, cleanup); 1719 } 1720 } 1721 1722 return 0; 1723 } 1724 1725 /* 1726 * Delete service by {netns} in the service table. 1727 * Called by __ip_vs_batch_cleanup() 1728 */ 1729 void ip_vs_service_nets_cleanup(struct list_head *net_list) 1730 { 1731 struct netns_ipvs *ipvs; 1732 struct net *net; 1733 1734 /* Check for "full" addressed entries */ 1735 mutex_lock(&__ip_vs_mutex); 1736 list_for_each_entry(net, net_list, exit_list) { 1737 ipvs = net_ipvs(net); 1738 ip_vs_flush(ipvs, true); 1739 } 1740 mutex_unlock(&__ip_vs_mutex); 1741 } 1742 1743 /* Put all references for device (dst_cache) */ 1744 static inline void 1745 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev) 1746 { 1747 struct ip_vs_dest_dst *dest_dst; 1748 1749 spin_lock_bh(&dest->dst_lock); 1750 dest_dst = rcu_dereference_protected(dest->dest_dst, 1); 1751 if (dest_dst && dest_dst->dst_cache->dev == dev) { 1752 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n", 1753 dev->name, 1754 IP_VS_DBG_ADDR(dest->af, &dest->addr), 1755 ntohs(dest->port), 1756 refcount_read(&dest->refcnt)); 1757 __ip_vs_dst_cache_reset(dest); 1758 } 1759 spin_unlock_bh(&dest->dst_lock); 1760 1761 } 1762 /* Netdev event receiver 1763 * Currently only NETDEV_DOWN is handled to release refs to cached dsts 1764 */ 1765 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event, 1766 void *ptr) 1767 { 1768 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1769 struct net *net = dev_net(dev); 1770 struct netns_ipvs *ipvs = net_ipvs(net); 1771 struct ip_vs_service *svc; 1772 struct ip_vs_dest *dest; 1773 unsigned int idx; 1774 1775 if (event != NETDEV_DOWN || !ipvs) 1776 return NOTIFY_DONE; 1777 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name); 1778 mutex_lock(&__ip_vs_mutex); 1779 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) { 1780 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) { 1781 if (svc->ipvs == ipvs) { 1782 list_for_each_entry(dest, &svc->destinations, 1783 n_list) { 1784 ip_vs_forget_dev(dest, dev); 1785 } 1786 } 1787 } 1788 1789 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) { 1790 if (svc->ipvs == ipvs) { 1791 list_for_each_entry(dest, &svc->destinations, 1792 n_list) { 1793 ip_vs_forget_dev(dest, dev); 1794 } 1795 } 1796 1797 } 1798 } 1799 1800 spin_lock_bh(&ipvs->dest_trash_lock); 1801 list_for_each_entry(dest, &ipvs->dest_trash, t_list) { 1802 ip_vs_forget_dev(dest, dev); 1803 } 1804 spin_unlock_bh(&ipvs->dest_trash_lock); 1805 mutex_unlock(&__ip_vs_mutex); 1806 return NOTIFY_DONE; 1807 } 1808 1809 /* 1810 * Zero counters in a service or all services 1811 */ 1812 static int ip_vs_zero_service(struct ip_vs_service *svc) 1813 { 1814 struct ip_vs_dest *dest; 1815 1816 list_for_each_entry(dest, &svc->destinations, n_list) { 1817 ip_vs_zero_stats(&dest->stats); 1818 } 1819 ip_vs_zero_stats(&svc->stats); 1820 return 0; 1821 } 1822 1823 static int ip_vs_zero_all(struct netns_ipvs *ipvs) 1824 { 1825 int idx; 1826 struct ip_vs_service *svc; 1827 1828 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) { 1829 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) { 1830 if (svc->ipvs == ipvs) 1831 ip_vs_zero_service(svc); 1832 } 1833 } 1834 1835 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) { 1836 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) { 1837 if (svc->ipvs == ipvs) 1838 ip_vs_zero_service(svc); 1839 } 1840 } 1841 1842 ip_vs_zero_stats(&ipvs->tot_stats->s); 1843 return 0; 1844 } 1845 1846 #ifdef CONFIG_SYSCTL 1847 1848 static int 1849 proc_do_defense_mode(struct ctl_table *table, int write, 1850 void *buffer, size_t *lenp, loff_t *ppos) 1851 { 1852 struct netns_ipvs *ipvs = table->extra2; 1853 int *valp = table->data; 1854 int val = *valp; 1855 int rc; 1856 1857 struct ctl_table tmp = { 1858 .data = &val, 1859 .maxlen = sizeof(int), 1860 .mode = table->mode, 1861 }; 1862 1863 rc = proc_dointvec(&tmp, write, buffer, lenp, ppos); 1864 if (write && (*valp != val)) { 1865 if (val < 0 || val > 3) { 1866 rc = -EINVAL; 1867 } else { 1868 *valp = val; 1869 update_defense_level(ipvs); 1870 } 1871 } 1872 return rc; 1873 } 1874 1875 static int 1876 proc_do_sync_threshold(struct ctl_table *table, int write, 1877 void *buffer, size_t *lenp, loff_t *ppos) 1878 { 1879 int *valp = table->data; 1880 int val[2]; 1881 int rc; 1882 struct ctl_table tmp = { 1883 .data = &val, 1884 .maxlen = table->maxlen, 1885 .mode = table->mode, 1886 }; 1887 1888 memcpy(val, valp, sizeof(val)); 1889 rc = proc_dointvec(&tmp, write, buffer, lenp, ppos); 1890 if (write) { 1891 if (val[0] < 0 || val[1] < 0 || 1892 (val[0] >= val[1] && val[1])) 1893 rc = -EINVAL; 1894 else 1895 memcpy(valp, val, sizeof(val)); 1896 } 1897 return rc; 1898 } 1899 1900 static int 1901 proc_do_sync_ports(struct ctl_table *table, int write, 1902 void *buffer, size_t *lenp, loff_t *ppos) 1903 { 1904 int *valp = table->data; 1905 int val = *valp; 1906 int rc; 1907 1908 struct ctl_table tmp = { 1909 .data = &val, 1910 .maxlen = sizeof(int), 1911 .mode = table->mode, 1912 }; 1913 1914 rc = proc_dointvec(&tmp, write, buffer, lenp, ppos); 1915 if (write && (*valp != val)) { 1916 if (val < 1 || !is_power_of_2(val)) 1917 rc = -EINVAL; 1918 else 1919 *valp = val; 1920 } 1921 return rc; 1922 } 1923 1924 static int ipvs_proc_est_cpumask_set(struct ctl_table *table, void *buffer) 1925 { 1926 struct netns_ipvs *ipvs = table->extra2; 1927 cpumask_var_t *valp = table->data; 1928 cpumask_var_t newmask; 1929 int ret; 1930 1931 if (!zalloc_cpumask_var(&newmask, GFP_KERNEL)) 1932 return -ENOMEM; 1933 1934 ret = cpulist_parse(buffer, newmask); 1935 if (ret) 1936 goto out; 1937 1938 mutex_lock(&ipvs->est_mutex); 1939 1940 if (!ipvs->est_cpulist_valid) { 1941 if (!zalloc_cpumask_var(valp, GFP_KERNEL)) { 1942 ret = -ENOMEM; 1943 goto unlock; 1944 } 1945 ipvs->est_cpulist_valid = 1; 1946 } 1947 cpumask_and(newmask, newmask, ¤t->cpus_mask); 1948 cpumask_copy(*valp, newmask); 1949 /* est_max_threads may depend on cpulist size */ 1950 ipvs->est_max_threads = ip_vs_est_max_threads(ipvs); 1951 ipvs->est_calc_phase = 1; 1952 ip_vs_est_reload_start(ipvs); 1953 1954 unlock: 1955 mutex_unlock(&ipvs->est_mutex); 1956 1957 out: 1958 free_cpumask_var(newmask); 1959 return ret; 1960 } 1961 1962 static int ipvs_proc_est_cpumask_get(struct ctl_table *table, void *buffer, 1963 size_t size) 1964 { 1965 struct netns_ipvs *ipvs = table->extra2; 1966 cpumask_var_t *valp = table->data; 1967 struct cpumask *mask; 1968 int ret; 1969 1970 mutex_lock(&ipvs->est_mutex); 1971 1972 if (ipvs->est_cpulist_valid) 1973 mask = *valp; 1974 else 1975 mask = (struct cpumask *)housekeeping_cpumask(HK_TYPE_KTHREAD); 1976 ret = scnprintf(buffer, size, "%*pbl\n", cpumask_pr_args(mask)); 1977 1978 mutex_unlock(&ipvs->est_mutex); 1979 1980 return ret; 1981 } 1982 1983 static int ipvs_proc_est_cpulist(struct ctl_table *table, int write, 1984 void *buffer, size_t *lenp, loff_t *ppos) 1985 { 1986 int ret; 1987 1988 /* Ignore both read and write(append) if *ppos not 0 */ 1989 if (*ppos || !*lenp) { 1990 *lenp = 0; 1991 return 0; 1992 } 1993 if (write) { 1994 /* proc_sys_call_handler() appends terminator */ 1995 ret = ipvs_proc_est_cpumask_set(table, buffer); 1996 if (ret >= 0) 1997 *ppos += *lenp; 1998 } else { 1999 /* proc_sys_call_handler() allocates 1 byte for terminator */ 2000 ret = ipvs_proc_est_cpumask_get(table, buffer, *lenp + 1); 2001 if (ret >= 0) { 2002 *lenp = ret; 2003 *ppos += *lenp; 2004 ret = 0; 2005 } 2006 } 2007 return ret; 2008 } 2009 2010 static int ipvs_proc_est_nice(struct ctl_table *table, int write, 2011 void *buffer, size_t *lenp, loff_t *ppos) 2012 { 2013 struct netns_ipvs *ipvs = table->extra2; 2014 int *valp = table->data; 2015 int val = *valp; 2016 int ret; 2017 2018 struct ctl_table tmp_table = { 2019 .data = &val, 2020 .maxlen = sizeof(int), 2021 .mode = table->mode, 2022 }; 2023 2024 ret = proc_dointvec(&tmp_table, write, buffer, lenp, ppos); 2025 if (write && ret >= 0) { 2026 if (val < MIN_NICE || val > MAX_NICE) { 2027 ret = -EINVAL; 2028 } else { 2029 mutex_lock(&ipvs->est_mutex); 2030 if (*valp != val) { 2031 *valp = val; 2032 ip_vs_est_reload_start(ipvs); 2033 } 2034 mutex_unlock(&ipvs->est_mutex); 2035 } 2036 } 2037 return ret; 2038 } 2039 2040 static int ipvs_proc_run_estimation(struct ctl_table *table, int write, 2041 void *buffer, size_t *lenp, loff_t *ppos) 2042 { 2043 struct netns_ipvs *ipvs = table->extra2; 2044 int *valp = table->data; 2045 int val = *valp; 2046 int ret; 2047 2048 struct ctl_table tmp_table = { 2049 .data = &val, 2050 .maxlen = sizeof(int), 2051 .mode = table->mode, 2052 }; 2053 2054 ret = proc_dointvec(&tmp_table, write, buffer, lenp, ppos); 2055 if (write && ret >= 0) { 2056 mutex_lock(&ipvs->est_mutex); 2057 if (*valp != val) { 2058 *valp = val; 2059 ip_vs_est_reload_start(ipvs); 2060 } 2061 mutex_unlock(&ipvs->est_mutex); 2062 } 2063 return ret; 2064 } 2065 2066 /* 2067 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/) 2068 * Do not change order or insert new entries without 2069 * align with netns init in ip_vs_control_net_init() 2070 */ 2071 2072 static struct ctl_table vs_vars[] = { 2073 { 2074 .procname = "amemthresh", 2075 .maxlen = sizeof(int), 2076 .mode = 0644, 2077 .proc_handler = proc_dointvec, 2078 }, 2079 { 2080 .procname = "am_droprate", 2081 .maxlen = sizeof(int), 2082 .mode = 0644, 2083 .proc_handler = proc_dointvec, 2084 }, 2085 { 2086 .procname = "drop_entry", 2087 .maxlen = sizeof(int), 2088 .mode = 0644, 2089 .proc_handler = proc_do_defense_mode, 2090 }, 2091 { 2092 .procname = "drop_packet", 2093 .maxlen = sizeof(int), 2094 .mode = 0644, 2095 .proc_handler = proc_do_defense_mode, 2096 }, 2097 #ifdef CONFIG_IP_VS_NFCT 2098 { 2099 .procname = "conntrack", 2100 .maxlen = sizeof(int), 2101 .mode = 0644, 2102 .proc_handler = &proc_dointvec, 2103 }, 2104 #endif 2105 { 2106 .procname = "secure_tcp", 2107 .maxlen = sizeof(int), 2108 .mode = 0644, 2109 .proc_handler = proc_do_defense_mode, 2110 }, 2111 { 2112 .procname = "snat_reroute", 2113 .maxlen = sizeof(int), 2114 .mode = 0644, 2115 .proc_handler = &proc_dointvec, 2116 }, 2117 { 2118 .procname = "sync_version", 2119 .maxlen = sizeof(int), 2120 .mode = 0644, 2121 .proc_handler = proc_dointvec_minmax, 2122 .extra1 = SYSCTL_ZERO, 2123 .extra2 = SYSCTL_ONE, 2124 }, 2125 { 2126 .procname = "sync_ports", 2127 .maxlen = sizeof(int), 2128 .mode = 0644, 2129 .proc_handler = proc_do_sync_ports, 2130 }, 2131 { 2132 .procname = "sync_persist_mode", 2133 .maxlen = sizeof(int), 2134 .mode = 0644, 2135 .proc_handler = proc_dointvec, 2136 }, 2137 { 2138 .procname = "sync_qlen_max", 2139 .maxlen = sizeof(unsigned long), 2140 .mode = 0644, 2141 .proc_handler = proc_doulongvec_minmax, 2142 }, 2143 { 2144 .procname = "sync_sock_size", 2145 .maxlen = sizeof(int), 2146 .mode = 0644, 2147 .proc_handler = proc_dointvec, 2148 }, 2149 { 2150 .procname = "cache_bypass", 2151 .maxlen = sizeof(int), 2152 .mode = 0644, 2153 .proc_handler = proc_dointvec, 2154 }, 2155 { 2156 .procname = "expire_nodest_conn", 2157 .maxlen = sizeof(int), 2158 .mode = 0644, 2159 .proc_handler = proc_dointvec, 2160 }, 2161 { 2162 .procname = "sloppy_tcp", 2163 .maxlen = sizeof(int), 2164 .mode = 0644, 2165 .proc_handler = proc_dointvec, 2166 }, 2167 { 2168 .procname = "sloppy_sctp", 2169 .maxlen = sizeof(int), 2170 .mode = 0644, 2171 .proc_handler = proc_dointvec, 2172 }, 2173 { 2174 .procname = "expire_quiescent_template", 2175 .maxlen = sizeof(int), 2176 .mode = 0644, 2177 .proc_handler = proc_dointvec, 2178 }, 2179 { 2180 .procname = "sync_threshold", 2181 .maxlen = 2182 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold), 2183 .mode = 0644, 2184 .proc_handler = proc_do_sync_threshold, 2185 }, 2186 { 2187 .procname = "sync_refresh_period", 2188 .maxlen = sizeof(int), 2189 .mode = 0644, 2190 .proc_handler = proc_dointvec_jiffies, 2191 }, 2192 { 2193 .procname = "sync_retries", 2194 .maxlen = sizeof(int), 2195 .mode = 0644, 2196 .proc_handler = proc_dointvec_minmax, 2197 .extra1 = SYSCTL_ZERO, 2198 .extra2 = SYSCTL_THREE, 2199 }, 2200 { 2201 .procname = "nat_icmp_send", 2202 .maxlen = sizeof(int), 2203 .mode = 0644, 2204 .proc_handler = proc_dointvec, 2205 }, 2206 { 2207 .procname = "pmtu_disc", 2208 .maxlen = sizeof(int), 2209 .mode = 0644, 2210 .proc_handler = proc_dointvec, 2211 }, 2212 { 2213 .procname = "backup_only", 2214 .maxlen = sizeof(int), 2215 .mode = 0644, 2216 .proc_handler = proc_dointvec, 2217 }, 2218 { 2219 .procname = "conn_reuse_mode", 2220 .maxlen = sizeof(int), 2221 .mode = 0644, 2222 .proc_handler = proc_dointvec, 2223 }, 2224 { 2225 .procname = "schedule_icmp", 2226 .maxlen = sizeof(int), 2227 .mode = 0644, 2228 .proc_handler = proc_dointvec, 2229 }, 2230 { 2231 .procname = "ignore_tunneled", 2232 .maxlen = sizeof(int), 2233 .mode = 0644, 2234 .proc_handler = proc_dointvec, 2235 }, 2236 { 2237 .procname = "run_estimation", 2238 .maxlen = sizeof(int), 2239 .mode = 0644, 2240 .proc_handler = ipvs_proc_run_estimation, 2241 }, 2242 { 2243 .procname = "est_cpulist", 2244 .maxlen = NR_CPUS, /* unused */ 2245 .mode = 0644, 2246 .proc_handler = ipvs_proc_est_cpulist, 2247 }, 2248 { 2249 .procname = "est_nice", 2250 .maxlen = sizeof(int), 2251 .mode = 0644, 2252 .proc_handler = ipvs_proc_est_nice, 2253 }, 2254 #ifdef CONFIG_IP_VS_DEBUG 2255 { 2256 .procname = "debug_level", 2257 .data = &sysctl_ip_vs_debug_level, 2258 .maxlen = sizeof(int), 2259 .mode = 0644, 2260 .proc_handler = proc_dointvec, 2261 }, 2262 #endif 2263 { } 2264 }; 2265 2266 #endif 2267 2268 #ifdef CONFIG_PROC_FS 2269 2270 struct ip_vs_iter { 2271 struct seq_net_private p; /* Do not move this, netns depends upon it*/ 2272 struct hlist_head *table; 2273 int bucket; 2274 }; 2275 2276 /* 2277 * Write the contents of the VS rule table to a PROCfs file. 2278 * (It is kept just for backward compatibility) 2279 */ 2280 static inline const char *ip_vs_fwd_name(unsigned int flags) 2281 { 2282 switch (flags & IP_VS_CONN_F_FWD_MASK) { 2283 case IP_VS_CONN_F_LOCALNODE: 2284 return "Local"; 2285 case IP_VS_CONN_F_TUNNEL: 2286 return "Tunnel"; 2287 case IP_VS_CONN_F_DROUTE: 2288 return "Route"; 2289 default: 2290 return "Masq"; 2291 } 2292 } 2293 2294 2295 /* Get the Nth entry in the two lists */ 2296 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos) 2297 { 2298 struct net *net = seq_file_net(seq); 2299 struct netns_ipvs *ipvs = net_ipvs(net); 2300 struct ip_vs_iter *iter = seq->private; 2301 int idx; 2302 struct ip_vs_service *svc; 2303 2304 /* look in hash by protocol */ 2305 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) { 2306 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) { 2307 if ((svc->ipvs == ipvs) && pos-- == 0) { 2308 iter->table = ip_vs_svc_table; 2309 iter->bucket = idx; 2310 return svc; 2311 } 2312 } 2313 } 2314 2315 /* keep looking in fwmark */ 2316 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) { 2317 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx], 2318 f_list) { 2319 if ((svc->ipvs == ipvs) && pos-- == 0) { 2320 iter->table = ip_vs_svc_fwm_table; 2321 iter->bucket = idx; 2322 return svc; 2323 } 2324 } 2325 } 2326 2327 return NULL; 2328 } 2329 2330 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos) 2331 __acquires(RCU) 2332 { 2333 rcu_read_lock(); 2334 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN; 2335 } 2336 2337 2338 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos) 2339 { 2340 struct hlist_node *e; 2341 struct ip_vs_iter *iter; 2342 struct ip_vs_service *svc; 2343 2344 ++*pos; 2345 if (v == SEQ_START_TOKEN) 2346 return ip_vs_info_array(seq,0); 2347 2348 svc = v; 2349 iter = seq->private; 2350 2351 if (iter->table == ip_vs_svc_table) { 2352 /* next service in table hashed by protocol */ 2353 e = rcu_dereference(hlist_next_rcu(&svc->s_list)); 2354 if (e) 2355 return hlist_entry(e, struct ip_vs_service, s_list); 2356 2357 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) { 2358 hlist_for_each_entry_rcu(svc, 2359 &ip_vs_svc_table[iter->bucket], 2360 s_list) { 2361 return svc; 2362 } 2363 } 2364 2365 iter->table = ip_vs_svc_fwm_table; 2366 iter->bucket = -1; 2367 goto scan_fwmark; 2368 } 2369 2370 /* next service in hashed by fwmark */ 2371 e = rcu_dereference(hlist_next_rcu(&svc->f_list)); 2372 if (e) 2373 return hlist_entry(e, struct ip_vs_service, f_list); 2374 2375 scan_fwmark: 2376 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) { 2377 hlist_for_each_entry_rcu(svc, 2378 &ip_vs_svc_fwm_table[iter->bucket], 2379 f_list) 2380 return svc; 2381 } 2382 2383 return NULL; 2384 } 2385 2386 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v) 2387 __releases(RCU) 2388 { 2389 rcu_read_unlock(); 2390 } 2391 2392 2393 static int ip_vs_info_seq_show(struct seq_file *seq, void *v) 2394 { 2395 if (v == SEQ_START_TOKEN) { 2396 seq_printf(seq, 2397 "IP Virtual Server version %d.%d.%d (size=%d)\n", 2398 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size); 2399 seq_puts(seq, 2400 "Prot LocalAddress:Port Scheduler Flags\n"); 2401 seq_puts(seq, 2402 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n"); 2403 } else { 2404 struct net *net = seq_file_net(seq); 2405 struct netns_ipvs *ipvs = net_ipvs(net); 2406 const struct ip_vs_service *svc = v; 2407 const struct ip_vs_iter *iter = seq->private; 2408 const struct ip_vs_dest *dest; 2409 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler); 2410 char *sched_name = sched ? sched->name : "none"; 2411 2412 if (svc->ipvs != ipvs) 2413 return 0; 2414 if (iter->table == ip_vs_svc_table) { 2415 #ifdef CONFIG_IP_VS_IPV6 2416 if (svc->af == AF_INET6) 2417 seq_printf(seq, "%s [%pI6]:%04X %s ", 2418 ip_vs_proto_name(svc->protocol), 2419 &svc->addr.in6, 2420 ntohs(svc->port), 2421 sched_name); 2422 else 2423 #endif 2424 seq_printf(seq, "%s %08X:%04X %s %s ", 2425 ip_vs_proto_name(svc->protocol), 2426 ntohl(svc->addr.ip), 2427 ntohs(svc->port), 2428 sched_name, 2429 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":""); 2430 } else { 2431 seq_printf(seq, "FWM %08X %s %s", 2432 svc->fwmark, sched_name, 2433 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":""); 2434 } 2435 2436 if (svc->flags & IP_VS_SVC_F_PERSISTENT) 2437 seq_printf(seq, "persistent %d %08X\n", 2438 svc->timeout, 2439 ntohl(svc->netmask)); 2440 else 2441 seq_putc(seq, '\n'); 2442 2443 list_for_each_entry_rcu(dest, &svc->destinations, n_list) { 2444 #ifdef CONFIG_IP_VS_IPV6 2445 if (dest->af == AF_INET6) 2446 seq_printf(seq, 2447 " -> [%pI6]:%04X" 2448 " %-7s %-6d %-10d %-10d\n", 2449 &dest->addr.in6, 2450 ntohs(dest->port), 2451 ip_vs_fwd_name(atomic_read(&dest->conn_flags)), 2452 atomic_read(&dest->weight), 2453 atomic_read(&dest->activeconns), 2454 atomic_read(&dest->inactconns)); 2455 else 2456 #endif 2457 seq_printf(seq, 2458 " -> %08X:%04X " 2459 "%-7s %-6d %-10d %-10d\n", 2460 ntohl(dest->addr.ip), 2461 ntohs(dest->port), 2462 ip_vs_fwd_name(atomic_read(&dest->conn_flags)), 2463 atomic_read(&dest->weight), 2464 atomic_read(&dest->activeconns), 2465 atomic_read(&dest->inactconns)); 2466 2467 } 2468 } 2469 return 0; 2470 } 2471 2472 static const struct seq_operations ip_vs_info_seq_ops = { 2473 .start = ip_vs_info_seq_start, 2474 .next = ip_vs_info_seq_next, 2475 .stop = ip_vs_info_seq_stop, 2476 .show = ip_vs_info_seq_show, 2477 }; 2478 2479 static int ip_vs_stats_show(struct seq_file *seq, void *v) 2480 { 2481 struct net *net = seq_file_single_net(seq); 2482 struct ip_vs_kstats show; 2483 2484 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */ 2485 seq_puts(seq, 2486 " Total Incoming Outgoing Incoming Outgoing\n"); 2487 seq_puts(seq, 2488 " Conns Packets Packets Bytes Bytes\n"); 2489 2490 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats->s); 2491 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n", 2492 (unsigned long long)show.conns, 2493 (unsigned long long)show.inpkts, 2494 (unsigned long long)show.outpkts, 2495 (unsigned long long)show.inbytes, 2496 (unsigned long long)show.outbytes); 2497 2498 /* 01234567 01234567 01234567 0123456701234567 0123456701234567*/ 2499 seq_puts(seq, 2500 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n"); 2501 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n", 2502 (unsigned long long)show.cps, 2503 (unsigned long long)show.inpps, 2504 (unsigned long long)show.outpps, 2505 (unsigned long long)show.inbps, 2506 (unsigned long long)show.outbps); 2507 2508 return 0; 2509 } 2510 2511 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v) 2512 { 2513 struct net *net = seq_file_single_net(seq); 2514 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats->s; 2515 struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats; 2516 struct ip_vs_kstats kstats; 2517 int i; 2518 2519 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */ 2520 seq_puts(seq, 2521 " Total Incoming Outgoing Incoming Outgoing\n"); 2522 seq_puts(seq, 2523 "CPU Conns Packets Packets Bytes Bytes\n"); 2524 2525 for_each_possible_cpu(i) { 2526 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i); 2527 unsigned int start; 2528 u64 conns, inpkts, outpkts, inbytes, outbytes; 2529 2530 do { 2531 start = u64_stats_fetch_begin(&u->syncp); 2532 conns = u64_stats_read(&u->cnt.conns); 2533 inpkts = u64_stats_read(&u->cnt.inpkts); 2534 outpkts = u64_stats_read(&u->cnt.outpkts); 2535 inbytes = u64_stats_read(&u->cnt.inbytes); 2536 outbytes = u64_stats_read(&u->cnt.outbytes); 2537 } while (u64_stats_fetch_retry(&u->syncp, start)); 2538 2539 seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n", 2540 i, (u64)conns, (u64)inpkts, 2541 (u64)outpkts, (u64)inbytes, 2542 (u64)outbytes); 2543 } 2544 2545 ip_vs_copy_stats(&kstats, tot_stats); 2546 2547 seq_printf(seq, " ~ %8LX %8LX %8LX %16LX %16LX\n\n", 2548 (unsigned long long)kstats.conns, 2549 (unsigned long long)kstats.inpkts, 2550 (unsigned long long)kstats.outpkts, 2551 (unsigned long long)kstats.inbytes, 2552 (unsigned long long)kstats.outbytes); 2553 2554 /* ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */ 2555 seq_puts(seq, 2556 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n"); 2557 seq_printf(seq, " %8LX %8LX %8LX %16LX %16LX\n", 2558 kstats.cps, 2559 kstats.inpps, 2560 kstats.outpps, 2561 kstats.inbps, 2562 kstats.outbps); 2563 2564 return 0; 2565 } 2566 #endif 2567 2568 /* 2569 * Set timeout values for tcp tcpfin udp in the timeout_table. 2570 */ 2571 static int ip_vs_set_timeout(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u) 2572 { 2573 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP) 2574 struct ip_vs_proto_data *pd; 2575 #endif 2576 2577 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n", 2578 u->tcp_timeout, 2579 u->tcp_fin_timeout, 2580 u->udp_timeout); 2581 2582 #ifdef CONFIG_IP_VS_PROTO_TCP 2583 if (u->tcp_timeout < 0 || u->tcp_timeout > (INT_MAX / HZ) || 2584 u->tcp_fin_timeout < 0 || u->tcp_fin_timeout > (INT_MAX / HZ)) { 2585 return -EINVAL; 2586 } 2587 #endif 2588 2589 #ifdef CONFIG_IP_VS_PROTO_UDP 2590 if (u->udp_timeout < 0 || u->udp_timeout > (INT_MAX / HZ)) 2591 return -EINVAL; 2592 #endif 2593 2594 #ifdef CONFIG_IP_VS_PROTO_TCP 2595 if (u->tcp_timeout) { 2596 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP); 2597 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] 2598 = u->tcp_timeout * HZ; 2599 } 2600 2601 if (u->tcp_fin_timeout) { 2602 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP); 2603 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] 2604 = u->tcp_fin_timeout * HZ; 2605 } 2606 #endif 2607 2608 #ifdef CONFIG_IP_VS_PROTO_UDP 2609 if (u->udp_timeout) { 2610 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP); 2611 pd->timeout_table[IP_VS_UDP_S_NORMAL] 2612 = u->udp_timeout * HZ; 2613 } 2614 #endif 2615 return 0; 2616 } 2617 2618 #define CMDID(cmd) (cmd - IP_VS_BASE_CTL) 2619 2620 struct ip_vs_svcdest_user { 2621 struct ip_vs_service_user s; 2622 struct ip_vs_dest_user d; 2623 }; 2624 2625 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = { 2626 [CMDID(IP_VS_SO_SET_ADD)] = sizeof(struct ip_vs_service_user), 2627 [CMDID(IP_VS_SO_SET_EDIT)] = sizeof(struct ip_vs_service_user), 2628 [CMDID(IP_VS_SO_SET_DEL)] = sizeof(struct ip_vs_service_user), 2629 [CMDID(IP_VS_SO_SET_ADDDEST)] = sizeof(struct ip_vs_svcdest_user), 2630 [CMDID(IP_VS_SO_SET_DELDEST)] = sizeof(struct ip_vs_svcdest_user), 2631 [CMDID(IP_VS_SO_SET_EDITDEST)] = sizeof(struct ip_vs_svcdest_user), 2632 [CMDID(IP_VS_SO_SET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user), 2633 [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user), 2634 [CMDID(IP_VS_SO_SET_STOPDAEMON)] = sizeof(struct ip_vs_daemon_user), 2635 [CMDID(IP_VS_SO_SET_ZERO)] = sizeof(struct ip_vs_service_user), 2636 }; 2637 2638 union ip_vs_set_arglen { 2639 struct ip_vs_service_user field_IP_VS_SO_SET_ADD; 2640 struct ip_vs_service_user field_IP_VS_SO_SET_EDIT; 2641 struct ip_vs_service_user field_IP_VS_SO_SET_DEL; 2642 struct ip_vs_svcdest_user field_IP_VS_SO_SET_ADDDEST; 2643 struct ip_vs_svcdest_user field_IP_VS_SO_SET_DELDEST; 2644 struct ip_vs_svcdest_user field_IP_VS_SO_SET_EDITDEST; 2645 struct ip_vs_timeout_user field_IP_VS_SO_SET_TIMEOUT; 2646 struct ip_vs_daemon_user field_IP_VS_SO_SET_STARTDAEMON; 2647 struct ip_vs_daemon_user field_IP_VS_SO_SET_STOPDAEMON; 2648 struct ip_vs_service_user field_IP_VS_SO_SET_ZERO; 2649 }; 2650 2651 #define MAX_SET_ARGLEN sizeof(union ip_vs_set_arglen) 2652 2653 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc, 2654 struct ip_vs_service_user *usvc_compat) 2655 { 2656 memset(usvc, 0, sizeof(*usvc)); 2657 2658 usvc->af = AF_INET; 2659 usvc->protocol = usvc_compat->protocol; 2660 usvc->addr.ip = usvc_compat->addr; 2661 usvc->port = usvc_compat->port; 2662 usvc->fwmark = usvc_compat->fwmark; 2663 2664 /* Deep copy of sched_name is not needed here */ 2665 usvc->sched_name = usvc_compat->sched_name; 2666 2667 usvc->flags = usvc_compat->flags; 2668 usvc->timeout = usvc_compat->timeout; 2669 usvc->netmask = usvc_compat->netmask; 2670 } 2671 2672 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest, 2673 struct ip_vs_dest_user *udest_compat) 2674 { 2675 memset(udest, 0, sizeof(*udest)); 2676 2677 udest->addr.ip = udest_compat->addr; 2678 udest->port = udest_compat->port; 2679 udest->conn_flags = udest_compat->conn_flags; 2680 udest->weight = udest_compat->weight; 2681 udest->u_threshold = udest_compat->u_threshold; 2682 udest->l_threshold = udest_compat->l_threshold; 2683 udest->af = AF_INET; 2684 udest->tun_type = IP_VS_CONN_F_TUNNEL_TYPE_IPIP; 2685 } 2686 2687 static int 2688 do_ip_vs_set_ctl(struct sock *sk, int cmd, sockptr_t ptr, unsigned int len) 2689 { 2690 struct net *net = sock_net(sk); 2691 int ret; 2692 unsigned char arg[MAX_SET_ARGLEN]; 2693 struct ip_vs_service_user *usvc_compat; 2694 struct ip_vs_service_user_kern usvc; 2695 struct ip_vs_service *svc; 2696 struct ip_vs_dest_user *udest_compat; 2697 struct ip_vs_dest_user_kern udest; 2698 struct netns_ipvs *ipvs = net_ipvs(net); 2699 2700 BUILD_BUG_ON(sizeof(arg) > 255); 2701 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) 2702 return -EPERM; 2703 2704 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX) 2705 return -EINVAL; 2706 if (len != set_arglen[CMDID(cmd)]) { 2707 IP_VS_DBG(1, "set_ctl: len %u != %u\n", 2708 len, set_arglen[CMDID(cmd)]); 2709 return -EINVAL; 2710 } 2711 2712 if (copy_from_sockptr(arg, ptr, len) != 0) 2713 return -EFAULT; 2714 2715 /* Handle daemons since they have another lock */ 2716 if (cmd == IP_VS_SO_SET_STARTDAEMON || 2717 cmd == IP_VS_SO_SET_STOPDAEMON) { 2718 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg; 2719 2720 if (cmd == IP_VS_SO_SET_STARTDAEMON) { 2721 struct ipvs_sync_daemon_cfg cfg; 2722 2723 memset(&cfg, 0, sizeof(cfg)); 2724 ret = -EINVAL; 2725 if (strscpy(cfg.mcast_ifn, dm->mcast_ifn, 2726 sizeof(cfg.mcast_ifn)) <= 0) 2727 return ret; 2728 cfg.syncid = dm->syncid; 2729 ret = start_sync_thread(ipvs, &cfg, dm->state); 2730 } else { 2731 ret = stop_sync_thread(ipvs, dm->state); 2732 } 2733 return ret; 2734 } 2735 2736 mutex_lock(&__ip_vs_mutex); 2737 if (cmd == IP_VS_SO_SET_FLUSH) { 2738 /* Flush the virtual service */ 2739 ret = ip_vs_flush(ipvs, false); 2740 goto out_unlock; 2741 } else if (cmd == IP_VS_SO_SET_TIMEOUT) { 2742 /* Set timeout values for (tcp tcpfin udp) */ 2743 ret = ip_vs_set_timeout(ipvs, (struct ip_vs_timeout_user *)arg); 2744 goto out_unlock; 2745 } else if (!len) { 2746 /* No more commands with len == 0 below */ 2747 ret = -EINVAL; 2748 goto out_unlock; 2749 } 2750 2751 usvc_compat = (struct ip_vs_service_user *)arg; 2752 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1); 2753 2754 /* We only use the new structs internally, so copy userspace compat 2755 * structs to extended internal versions */ 2756 ip_vs_copy_usvc_compat(&usvc, usvc_compat); 2757 ip_vs_copy_udest_compat(&udest, udest_compat); 2758 2759 if (cmd == IP_VS_SO_SET_ZERO) { 2760 /* if no service address is set, zero counters in all */ 2761 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) { 2762 ret = ip_vs_zero_all(ipvs); 2763 goto out_unlock; 2764 } 2765 } 2766 2767 if ((cmd == IP_VS_SO_SET_ADD || cmd == IP_VS_SO_SET_EDIT) && 2768 strnlen(usvc.sched_name, IP_VS_SCHEDNAME_MAXLEN) == 2769 IP_VS_SCHEDNAME_MAXLEN) { 2770 ret = -EINVAL; 2771 goto out_unlock; 2772 } 2773 2774 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */ 2775 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP && 2776 usvc.protocol != IPPROTO_SCTP) { 2777 pr_err("set_ctl: invalid protocol: %d %pI4:%d\n", 2778 usvc.protocol, &usvc.addr.ip, 2779 ntohs(usvc.port)); 2780 ret = -EFAULT; 2781 goto out_unlock; 2782 } 2783 2784 /* Lookup the exact service by <protocol, addr, port> or fwmark */ 2785 rcu_read_lock(); 2786 if (usvc.fwmark == 0) 2787 svc = __ip_vs_service_find(ipvs, usvc.af, usvc.protocol, 2788 &usvc.addr, usvc.port); 2789 else 2790 svc = __ip_vs_svc_fwm_find(ipvs, usvc.af, usvc.fwmark); 2791 rcu_read_unlock(); 2792 2793 if (cmd != IP_VS_SO_SET_ADD 2794 && (svc == NULL || svc->protocol != usvc.protocol)) { 2795 ret = -ESRCH; 2796 goto out_unlock; 2797 } 2798 2799 switch (cmd) { 2800 case IP_VS_SO_SET_ADD: 2801 if (svc != NULL) 2802 ret = -EEXIST; 2803 else 2804 ret = ip_vs_add_service(ipvs, &usvc, &svc); 2805 break; 2806 case IP_VS_SO_SET_EDIT: 2807 ret = ip_vs_edit_service(svc, &usvc); 2808 break; 2809 case IP_VS_SO_SET_DEL: 2810 ret = ip_vs_del_service(svc); 2811 if (!ret) 2812 goto out_unlock; 2813 break; 2814 case IP_VS_SO_SET_ZERO: 2815 ret = ip_vs_zero_service(svc); 2816 break; 2817 case IP_VS_SO_SET_ADDDEST: 2818 ret = ip_vs_add_dest(svc, &udest); 2819 break; 2820 case IP_VS_SO_SET_EDITDEST: 2821 ret = ip_vs_edit_dest(svc, &udest); 2822 break; 2823 case IP_VS_SO_SET_DELDEST: 2824 ret = ip_vs_del_dest(svc, &udest); 2825 break; 2826 default: 2827 WARN_ON_ONCE(1); 2828 ret = -EINVAL; 2829 break; 2830 } 2831 2832 out_unlock: 2833 mutex_unlock(&__ip_vs_mutex); 2834 return ret; 2835 } 2836 2837 2838 static void 2839 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src) 2840 { 2841 struct ip_vs_scheduler *sched; 2842 struct ip_vs_kstats kstats; 2843 char *sched_name; 2844 2845 sched = rcu_dereference_protected(src->scheduler, 1); 2846 sched_name = sched ? sched->name : "none"; 2847 dst->protocol = src->protocol; 2848 dst->addr = src->addr.ip; 2849 dst->port = src->port; 2850 dst->fwmark = src->fwmark; 2851 strscpy(dst->sched_name, sched_name, sizeof(dst->sched_name)); 2852 dst->flags = src->flags; 2853 dst->timeout = src->timeout / HZ; 2854 dst->netmask = src->netmask; 2855 dst->num_dests = src->num_dests; 2856 ip_vs_copy_stats(&kstats, &src->stats); 2857 ip_vs_export_stats_user(&dst->stats, &kstats); 2858 } 2859 2860 static inline int 2861 __ip_vs_get_service_entries(struct netns_ipvs *ipvs, 2862 const struct ip_vs_get_services *get, 2863 struct ip_vs_get_services __user *uptr) 2864 { 2865 int idx, count=0; 2866 struct ip_vs_service *svc; 2867 struct ip_vs_service_entry entry; 2868 int ret = 0; 2869 2870 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) { 2871 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) { 2872 /* Only expose IPv4 entries to old interface */ 2873 if (svc->af != AF_INET || (svc->ipvs != ipvs)) 2874 continue; 2875 2876 if (count >= get->num_services) 2877 goto out; 2878 memset(&entry, 0, sizeof(entry)); 2879 ip_vs_copy_service(&entry, svc); 2880 if (copy_to_user(&uptr->entrytable[count], 2881 &entry, sizeof(entry))) { 2882 ret = -EFAULT; 2883 goto out; 2884 } 2885 count++; 2886 } 2887 } 2888 2889 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) { 2890 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) { 2891 /* Only expose IPv4 entries to old interface */ 2892 if (svc->af != AF_INET || (svc->ipvs != ipvs)) 2893 continue; 2894 2895 if (count >= get->num_services) 2896 goto out; 2897 memset(&entry, 0, sizeof(entry)); 2898 ip_vs_copy_service(&entry, svc); 2899 if (copy_to_user(&uptr->entrytable[count], 2900 &entry, sizeof(entry))) { 2901 ret = -EFAULT; 2902 goto out; 2903 } 2904 count++; 2905 } 2906 } 2907 out: 2908 return ret; 2909 } 2910 2911 static inline int 2912 __ip_vs_get_dest_entries(struct netns_ipvs *ipvs, const struct ip_vs_get_dests *get, 2913 struct ip_vs_get_dests __user *uptr) 2914 { 2915 struct ip_vs_service *svc; 2916 union nf_inet_addr addr = { .ip = get->addr }; 2917 int ret = 0; 2918 2919 rcu_read_lock(); 2920 if (get->fwmark) 2921 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, get->fwmark); 2922 else 2923 svc = __ip_vs_service_find(ipvs, AF_INET, get->protocol, &addr, 2924 get->port); 2925 rcu_read_unlock(); 2926 2927 if (svc) { 2928 int count = 0; 2929 struct ip_vs_dest *dest; 2930 struct ip_vs_dest_entry entry; 2931 struct ip_vs_kstats kstats; 2932 2933 memset(&entry, 0, sizeof(entry)); 2934 list_for_each_entry(dest, &svc->destinations, n_list) { 2935 if (count >= get->num_dests) 2936 break; 2937 2938 /* Cannot expose heterogeneous members via sockopt 2939 * interface 2940 */ 2941 if (dest->af != svc->af) 2942 continue; 2943 2944 entry.addr = dest->addr.ip; 2945 entry.port = dest->port; 2946 entry.conn_flags = atomic_read(&dest->conn_flags); 2947 entry.weight = atomic_read(&dest->weight); 2948 entry.u_threshold = dest->u_threshold; 2949 entry.l_threshold = dest->l_threshold; 2950 entry.activeconns = atomic_read(&dest->activeconns); 2951 entry.inactconns = atomic_read(&dest->inactconns); 2952 entry.persistconns = atomic_read(&dest->persistconns); 2953 ip_vs_copy_stats(&kstats, &dest->stats); 2954 ip_vs_export_stats_user(&entry.stats, &kstats); 2955 if (copy_to_user(&uptr->entrytable[count], 2956 &entry, sizeof(entry))) { 2957 ret = -EFAULT; 2958 break; 2959 } 2960 count++; 2961 } 2962 } else 2963 ret = -ESRCH; 2964 return ret; 2965 } 2966 2967 static inline void 2968 __ip_vs_get_timeouts(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u) 2969 { 2970 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP) 2971 struct ip_vs_proto_data *pd; 2972 #endif 2973 2974 memset(u, 0, sizeof (*u)); 2975 2976 #ifdef CONFIG_IP_VS_PROTO_TCP 2977 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP); 2978 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ; 2979 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ; 2980 #endif 2981 #ifdef CONFIG_IP_VS_PROTO_UDP 2982 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP); 2983 u->udp_timeout = 2984 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ; 2985 #endif 2986 } 2987 2988 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = { 2989 [CMDID(IP_VS_SO_GET_VERSION)] = 64, 2990 [CMDID(IP_VS_SO_GET_INFO)] = sizeof(struct ip_vs_getinfo), 2991 [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services), 2992 [CMDID(IP_VS_SO_GET_SERVICE)] = sizeof(struct ip_vs_service_entry), 2993 [CMDID(IP_VS_SO_GET_DESTS)] = sizeof(struct ip_vs_get_dests), 2994 [CMDID(IP_VS_SO_GET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user), 2995 [CMDID(IP_VS_SO_GET_DAEMON)] = 2 * sizeof(struct ip_vs_daemon_user), 2996 }; 2997 2998 union ip_vs_get_arglen { 2999 char field_IP_VS_SO_GET_VERSION[64]; 3000 struct ip_vs_getinfo field_IP_VS_SO_GET_INFO; 3001 struct ip_vs_get_services field_IP_VS_SO_GET_SERVICES; 3002 struct ip_vs_service_entry field_IP_VS_SO_GET_SERVICE; 3003 struct ip_vs_get_dests field_IP_VS_SO_GET_DESTS; 3004 struct ip_vs_timeout_user field_IP_VS_SO_GET_TIMEOUT; 3005 struct ip_vs_daemon_user field_IP_VS_SO_GET_DAEMON[2]; 3006 }; 3007 3008 #define MAX_GET_ARGLEN sizeof(union ip_vs_get_arglen) 3009 3010 static int 3011 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len) 3012 { 3013 unsigned char arg[MAX_GET_ARGLEN]; 3014 int ret = 0; 3015 unsigned int copylen; 3016 struct net *net = sock_net(sk); 3017 struct netns_ipvs *ipvs = net_ipvs(net); 3018 3019 BUG_ON(!net); 3020 BUILD_BUG_ON(sizeof(arg) > 255); 3021 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) 3022 return -EPERM; 3023 3024 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX) 3025 return -EINVAL; 3026 3027 copylen = get_arglen[CMDID(cmd)]; 3028 if (*len < (int) copylen) { 3029 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen); 3030 return -EINVAL; 3031 } 3032 3033 if (copy_from_user(arg, user, copylen) != 0) 3034 return -EFAULT; 3035 /* 3036 * Handle daemons first since it has its own locking 3037 */ 3038 if (cmd == IP_VS_SO_GET_DAEMON) { 3039 struct ip_vs_daemon_user d[2]; 3040 3041 memset(&d, 0, sizeof(d)); 3042 mutex_lock(&ipvs->sync_mutex); 3043 if (ipvs->sync_state & IP_VS_STATE_MASTER) { 3044 d[0].state = IP_VS_STATE_MASTER; 3045 strscpy(d[0].mcast_ifn, ipvs->mcfg.mcast_ifn, 3046 sizeof(d[0].mcast_ifn)); 3047 d[0].syncid = ipvs->mcfg.syncid; 3048 } 3049 if (ipvs->sync_state & IP_VS_STATE_BACKUP) { 3050 d[1].state = IP_VS_STATE_BACKUP; 3051 strscpy(d[1].mcast_ifn, ipvs->bcfg.mcast_ifn, 3052 sizeof(d[1].mcast_ifn)); 3053 d[1].syncid = ipvs->bcfg.syncid; 3054 } 3055 if (copy_to_user(user, &d, sizeof(d)) != 0) 3056 ret = -EFAULT; 3057 mutex_unlock(&ipvs->sync_mutex); 3058 return ret; 3059 } 3060 3061 mutex_lock(&__ip_vs_mutex); 3062 switch (cmd) { 3063 case IP_VS_SO_GET_VERSION: 3064 { 3065 char buf[64]; 3066 3067 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)", 3068 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size); 3069 if (copy_to_user(user, buf, strlen(buf)+1) != 0) { 3070 ret = -EFAULT; 3071 goto out; 3072 } 3073 *len = strlen(buf)+1; 3074 } 3075 break; 3076 3077 case IP_VS_SO_GET_INFO: 3078 { 3079 struct ip_vs_getinfo info; 3080 info.version = IP_VS_VERSION_CODE; 3081 info.size = ip_vs_conn_tab_size; 3082 info.num_services = ipvs->num_services; 3083 if (copy_to_user(user, &info, sizeof(info)) != 0) 3084 ret = -EFAULT; 3085 } 3086 break; 3087 3088 case IP_VS_SO_GET_SERVICES: 3089 { 3090 struct ip_vs_get_services *get; 3091 int size; 3092 3093 get = (struct ip_vs_get_services *)arg; 3094 size = struct_size(get, entrytable, get->num_services); 3095 if (*len != size) { 3096 pr_err("length: %u != %u\n", *len, size); 3097 ret = -EINVAL; 3098 goto out; 3099 } 3100 ret = __ip_vs_get_service_entries(ipvs, get, user); 3101 } 3102 break; 3103 3104 case IP_VS_SO_GET_SERVICE: 3105 { 3106 struct ip_vs_service_entry *entry; 3107 struct ip_vs_service *svc; 3108 union nf_inet_addr addr; 3109 3110 entry = (struct ip_vs_service_entry *)arg; 3111 addr.ip = entry->addr; 3112 rcu_read_lock(); 3113 if (entry->fwmark) 3114 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, entry->fwmark); 3115 else 3116 svc = __ip_vs_service_find(ipvs, AF_INET, 3117 entry->protocol, &addr, 3118 entry->port); 3119 rcu_read_unlock(); 3120 if (svc) { 3121 ip_vs_copy_service(entry, svc); 3122 if (copy_to_user(user, entry, sizeof(*entry)) != 0) 3123 ret = -EFAULT; 3124 } else 3125 ret = -ESRCH; 3126 } 3127 break; 3128 3129 case IP_VS_SO_GET_DESTS: 3130 { 3131 struct ip_vs_get_dests *get; 3132 int size; 3133 3134 get = (struct ip_vs_get_dests *)arg; 3135 size = struct_size(get, entrytable, get->num_dests); 3136 if (*len != size) { 3137 pr_err("length: %u != %u\n", *len, size); 3138 ret = -EINVAL; 3139 goto out; 3140 } 3141 ret = __ip_vs_get_dest_entries(ipvs, get, user); 3142 } 3143 break; 3144 3145 case IP_VS_SO_GET_TIMEOUT: 3146 { 3147 struct ip_vs_timeout_user t; 3148 3149 __ip_vs_get_timeouts(ipvs, &t); 3150 if (copy_to_user(user, &t, sizeof(t)) != 0) 3151 ret = -EFAULT; 3152 } 3153 break; 3154 3155 default: 3156 ret = -EINVAL; 3157 } 3158 3159 out: 3160 mutex_unlock(&__ip_vs_mutex); 3161 return ret; 3162 } 3163 3164 3165 static struct nf_sockopt_ops ip_vs_sockopts = { 3166 .pf = PF_INET, 3167 .set_optmin = IP_VS_BASE_CTL, 3168 .set_optmax = IP_VS_SO_SET_MAX+1, 3169 .set = do_ip_vs_set_ctl, 3170 .get_optmin = IP_VS_BASE_CTL, 3171 .get_optmax = IP_VS_SO_GET_MAX+1, 3172 .get = do_ip_vs_get_ctl, 3173 .owner = THIS_MODULE, 3174 }; 3175 3176 /* 3177 * Generic Netlink interface 3178 */ 3179 3180 /* IPVS genetlink family */ 3181 static struct genl_family ip_vs_genl_family; 3182 3183 /* Policy used for first-level command attributes */ 3184 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = { 3185 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED }, 3186 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED }, 3187 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED }, 3188 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 }, 3189 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 }, 3190 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 }, 3191 }; 3192 3193 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */ 3194 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = { 3195 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 }, 3196 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING, 3197 .len = IP_VS_IFNAME_MAXLEN - 1 }, 3198 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 }, 3199 [IPVS_DAEMON_ATTR_SYNC_MAXLEN] = { .type = NLA_U16 }, 3200 [IPVS_DAEMON_ATTR_MCAST_GROUP] = { .type = NLA_U32 }, 3201 [IPVS_DAEMON_ATTR_MCAST_GROUP6] = { .len = sizeof(struct in6_addr) }, 3202 [IPVS_DAEMON_ATTR_MCAST_PORT] = { .type = NLA_U16 }, 3203 [IPVS_DAEMON_ATTR_MCAST_TTL] = { .type = NLA_U8 }, 3204 }; 3205 3206 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */ 3207 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = { 3208 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 }, 3209 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 }, 3210 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY, 3211 .len = sizeof(union nf_inet_addr) }, 3212 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 }, 3213 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 }, 3214 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING, 3215 .len = IP_VS_SCHEDNAME_MAXLEN - 1 }, 3216 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING, 3217 .len = IP_VS_PENAME_MAXLEN }, 3218 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY, 3219 .len = sizeof(struct ip_vs_flags) }, 3220 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 }, 3221 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 }, 3222 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED }, 3223 }; 3224 3225 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */ 3226 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = { 3227 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY, 3228 .len = sizeof(union nf_inet_addr) }, 3229 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 }, 3230 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 }, 3231 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 }, 3232 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 }, 3233 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 }, 3234 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 }, 3235 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 }, 3236 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 }, 3237 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED }, 3238 [IPVS_DEST_ATTR_ADDR_FAMILY] = { .type = NLA_U16 }, 3239 [IPVS_DEST_ATTR_TUN_TYPE] = { .type = NLA_U8 }, 3240 [IPVS_DEST_ATTR_TUN_PORT] = { .type = NLA_U16 }, 3241 [IPVS_DEST_ATTR_TUN_FLAGS] = { .type = NLA_U16 }, 3242 }; 3243 3244 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type, 3245 struct ip_vs_kstats *kstats) 3246 { 3247 struct nlattr *nl_stats = nla_nest_start_noflag(skb, container_type); 3248 3249 if (!nl_stats) 3250 return -EMSGSIZE; 3251 3252 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) || 3253 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) || 3254 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) || 3255 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes, 3256 IPVS_STATS_ATTR_PAD) || 3257 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes, 3258 IPVS_STATS_ATTR_PAD) || 3259 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) || 3260 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) || 3261 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) || 3262 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) || 3263 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps)) 3264 goto nla_put_failure; 3265 nla_nest_end(skb, nl_stats); 3266 3267 return 0; 3268 3269 nla_put_failure: 3270 nla_nest_cancel(skb, nl_stats); 3271 return -EMSGSIZE; 3272 } 3273 3274 static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type, 3275 struct ip_vs_kstats *kstats) 3276 { 3277 struct nlattr *nl_stats = nla_nest_start_noflag(skb, container_type); 3278 3279 if (!nl_stats) 3280 return -EMSGSIZE; 3281 3282 if (nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CONNS, kstats->conns, 3283 IPVS_STATS_ATTR_PAD) || 3284 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts, 3285 IPVS_STATS_ATTR_PAD) || 3286 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts, 3287 IPVS_STATS_ATTR_PAD) || 3288 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes, 3289 IPVS_STATS_ATTR_PAD) || 3290 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes, 3291 IPVS_STATS_ATTR_PAD) || 3292 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CPS, kstats->cps, 3293 IPVS_STATS_ATTR_PAD) || 3294 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps, 3295 IPVS_STATS_ATTR_PAD) || 3296 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps, 3297 IPVS_STATS_ATTR_PAD) || 3298 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps, 3299 IPVS_STATS_ATTR_PAD) || 3300 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps, 3301 IPVS_STATS_ATTR_PAD)) 3302 goto nla_put_failure; 3303 nla_nest_end(skb, nl_stats); 3304 3305 return 0; 3306 3307 nla_put_failure: 3308 nla_nest_cancel(skb, nl_stats); 3309 return -EMSGSIZE; 3310 } 3311 3312 static int ip_vs_genl_fill_service(struct sk_buff *skb, 3313 struct ip_vs_service *svc) 3314 { 3315 struct ip_vs_scheduler *sched; 3316 struct ip_vs_pe *pe; 3317 struct nlattr *nl_service; 3318 struct ip_vs_flags flags = { .flags = svc->flags, 3319 .mask = ~0 }; 3320 struct ip_vs_kstats kstats; 3321 char *sched_name; 3322 3323 nl_service = nla_nest_start_noflag(skb, IPVS_CMD_ATTR_SERVICE); 3324 if (!nl_service) 3325 return -EMSGSIZE; 3326 3327 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af)) 3328 goto nla_put_failure; 3329 if (svc->fwmark) { 3330 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark)) 3331 goto nla_put_failure; 3332 } else { 3333 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) || 3334 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) || 3335 nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port)) 3336 goto nla_put_failure; 3337 } 3338 3339 sched = rcu_dereference_protected(svc->scheduler, 1); 3340 sched_name = sched ? sched->name : "none"; 3341 pe = rcu_dereference_protected(svc->pe, 1); 3342 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) || 3343 (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) || 3344 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) || 3345 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) || 3346 nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask)) 3347 goto nla_put_failure; 3348 ip_vs_copy_stats(&kstats, &svc->stats); 3349 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats)) 3350 goto nla_put_failure; 3351 if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats)) 3352 goto nla_put_failure; 3353 3354 nla_nest_end(skb, nl_service); 3355 3356 return 0; 3357 3358 nla_put_failure: 3359 nla_nest_cancel(skb, nl_service); 3360 return -EMSGSIZE; 3361 } 3362 3363 static int ip_vs_genl_dump_service(struct sk_buff *skb, 3364 struct ip_vs_service *svc, 3365 struct netlink_callback *cb) 3366 { 3367 void *hdr; 3368 3369 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, 3370 &ip_vs_genl_family, NLM_F_MULTI, 3371 IPVS_CMD_NEW_SERVICE); 3372 if (!hdr) 3373 return -EMSGSIZE; 3374 3375 if (ip_vs_genl_fill_service(skb, svc) < 0) 3376 goto nla_put_failure; 3377 3378 genlmsg_end(skb, hdr); 3379 return 0; 3380 3381 nla_put_failure: 3382 genlmsg_cancel(skb, hdr); 3383 return -EMSGSIZE; 3384 } 3385 3386 static int ip_vs_genl_dump_services(struct sk_buff *skb, 3387 struct netlink_callback *cb) 3388 { 3389 int idx = 0, i; 3390 int start = cb->args[0]; 3391 struct ip_vs_service *svc; 3392 struct net *net = sock_net(skb->sk); 3393 struct netns_ipvs *ipvs = net_ipvs(net); 3394 3395 mutex_lock(&__ip_vs_mutex); 3396 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) { 3397 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) { 3398 if (++idx <= start || (svc->ipvs != ipvs)) 3399 continue; 3400 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) { 3401 idx--; 3402 goto nla_put_failure; 3403 } 3404 } 3405 } 3406 3407 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) { 3408 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) { 3409 if (++idx <= start || (svc->ipvs != ipvs)) 3410 continue; 3411 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) { 3412 idx--; 3413 goto nla_put_failure; 3414 } 3415 } 3416 } 3417 3418 nla_put_failure: 3419 mutex_unlock(&__ip_vs_mutex); 3420 cb->args[0] = idx; 3421 3422 return skb->len; 3423 } 3424 3425 static bool ip_vs_is_af_valid(int af) 3426 { 3427 if (af == AF_INET) 3428 return true; 3429 #ifdef CONFIG_IP_VS_IPV6 3430 if (af == AF_INET6 && ipv6_mod_enabled()) 3431 return true; 3432 #endif 3433 return false; 3434 } 3435 3436 static int ip_vs_genl_parse_service(struct netns_ipvs *ipvs, 3437 struct ip_vs_service_user_kern *usvc, 3438 struct nlattr *nla, bool full_entry, 3439 struct ip_vs_service **ret_svc) 3440 { 3441 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1]; 3442 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr; 3443 struct ip_vs_service *svc; 3444 3445 /* Parse mandatory identifying service fields first */ 3446 if (nla == NULL || 3447 nla_parse_nested_deprecated(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy, NULL)) 3448 return -EINVAL; 3449 3450 nla_af = attrs[IPVS_SVC_ATTR_AF]; 3451 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL]; 3452 nla_addr = attrs[IPVS_SVC_ATTR_ADDR]; 3453 nla_port = attrs[IPVS_SVC_ATTR_PORT]; 3454 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK]; 3455 3456 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr)))) 3457 return -EINVAL; 3458 3459 memset(usvc, 0, sizeof(*usvc)); 3460 3461 usvc->af = nla_get_u16(nla_af); 3462 if (!ip_vs_is_af_valid(usvc->af)) 3463 return -EAFNOSUPPORT; 3464 3465 if (nla_fwmark) { 3466 usvc->protocol = IPPROTO_TCP; 3467 usvc->fwmark = nla_get_u32(nla_fwmark); 3468 } else { 3469 usvc->protocol = nla_get_u16(nla_protocol); 3470 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr)); 3471 usvc->port = nla_get_be16(nla_port); 3472 usvc->fwmark = 0; 3473 } 3474 3475 rcu_read_lock(); 3476 if (usvc->fwmark) 3477 svc = __ip_vs_svc_fwm_find(ipvs, usvc->af, usvc->fwmark); 3478 else 3479 svc = __ip_vs_service_find(ipvs, usvc->af, usvc->protocol, 3480 &usvc->addr, usvc->port); 3481 rcu_read_unlock(); 3482 *ret_svc = svc; 3483 3484 /* If a full entry was requested, check for the additional fields */ 3485 if (full_entry) { 3486 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout, 3487 *nla_netmask; 3488 struct ip_vs_flags flags; 3489 3490 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME]; 3491 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME]; 3492 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS]; 3493 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT]; 3494 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK]; 3495 3496 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask)) 3497 return -EINVAL; 3498 3499 nla_memcpy(&flags, nla_flags, sizeof(flags)); 3500 3501 /* prefill flags from service if it already exists */ 3502 if (svc) 3503 usvc->flags = svc->flags; 3504 3505 /* set new flags from userland */ 3506 usvc->flags = (usvc->flags & ~flags.mask) | 3507 (flags.flags & flags.mask); 3508 usvc->sched_name = nla_data(nla_sched); 3509 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL; 3510 usvc->timeout = nla_get_u32(nla_timeout); 3511 usvc->netmask = nla_get_be32(nla_netmask); 3512 } 3513 3514 return 0; 3515 } 3516 3517 static struct ip_vs_service *ip_vs_genl_find_service(struct netns_ipvs *ipvs, 3518 struct nlattr *nla) 3519 { 3520 struct ip_vs_service_user_kern usvc; 3521 struct ip_vs_service *svc; 3522 int ret; 3523 3524 ret = ip_vs_genl_parse_service(ipvs, &usvc, nla, false, &svc); 3525 return ret ? ERR_PTR(ret) : svc; 3526 } 3527 3528 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest) 3529 { 3530 struct nlattr *nl_dest; 3531 struct ip_vs_kstats kstats; 3532 3533 nl_dest = nla_nest_start_noflag(skb, IPVS_CMD_ATTR_DEST); 3534 if (!nl_dest) 3535 return -EMSGSIZE; 3536 3537 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) || 3538 nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) || 3539 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD, 3540 (atomic_read(&dest->conn_flags) & 3541 IP_VS_CONN_F_FWD_MASK)) || 3542 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT, 3543 atomic_read(&dest->weight)) || 3544 nla_put_u8(skb, IPVS_DEST_ATTR_TUN_TYPE, 3545 dest->tun_type) || 3546 nla_put_be16(skb, IPVS_DEST_ATTR_TUN_PORT, 3547 dest->tun_port) || 3548 nla_put_u16(skb, IPVS_DEST_ATTR_TUN_FLAGS, 3549 dest->tun_flags) || 3550 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) || 3551 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) || 3552 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS, 3553 atomic_read(&dest->activeconns)) || 3554 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS, 3555 atomic_read(&dest->inactconns)) || 3556 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS, 3557 atomic_read(&dest->persistconns)) || 3558 nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af)) 3559 goto nla_put_failure; 3560 ip_vs_copy_stats(&kstats, &dest->stats); 3561 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats)) 3562 goto nla_put_failure; 3563 if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats)) 3564 goto nla_put_failure; 3565 3566 nla_nest_end(skb, nl_dest); 3567 3568 return 0; 3569 3570 nla_put_failure: 3571 nla_nest_cancel(skb, nl_dest); 3572 return -EMSGSIZE; 3573 } 3574 3575 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest, 3576 struct netlink_callback *cb) 3577 { 3578 void *hdr; 3579 3580 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, 3581 &ip_vs_genl_family, NLM_F_MULTI, 3582 IPVS_CMD_NEW_DEST); 3583 if (!hdr) 3584 return -EMSGSIZE; 3585 3586 if (ip_vs_genl_fill_dest(skb, dest) < 0) 3587 goto nla_put_failure; 3588 3589 genlmsg_end(skb, hdr); 3590 return 0; 3591 3592 nla_put_failure: 3593 genlmsg_cancel(skb, hdr); 3594 return -EMSGSIZE; 3595 } 3596 3597 static int ip_vs_genl_dump_dests(struct sk_buff *skb, 3598 struct netlink_callback *cb) 3599 { 3600 int idx = 0; 3601 int start = cb->args[0]; 3602 struct ip_vs_service *svc; 3603 struct ip_vs_dest *dest; 3604 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1]; 3605 struct net *net = sock_net(skb->sk); 3606 struct netns_ipvs *ipvs = net_ipvs(net); 3607 3608 mutex_lock(&__ip_vs_mutex); 3609 3610 /* Try to find the service for which to dump destinations */ 3611 if (nlmsg_parse_deprecated(cb->nlh, GENL_HDRLEN, attrs, IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy, cb->extack)) 3612 goto out_err; 3613 3614 3615 svc = ip_vs_genl_find_service(ipvs, attrs[IPVS_CMD_ATTR_SERVICE]); 3616 if (IS_ERR_OR_NULL(svc)) 3617 goto out_err; 3618 3619 /* Dump the destinations */ 3620 list_for_each_entry(dest, &svc->destinations, n_list) { 3621 if (++idx <= start) 3622 continue; 3623 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) { 3624 idx--; 3625 goto nla_put_failure; 3626 } 3627 } 3628 3629 nla_put_failure: 3630 cb->args[0] = idx; 3631 3632 out_err: 3633 mutex_unlock(&__ip_vs_mutex); 3634 3635 return skb->len; 3636 } 3637 3638 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest, 3639 struct nlattr *nla, bool full_entry) 3640 { 3641 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1]; 3642 struct nlattr *nla_addr, *nla_port; 3643 struct nlattr *nla_addr_family; 3644 3645 /* Parse mandatory identifying destination fields first */ 3646 if (nla == NULL || 3647 nla_parse_nested_deprecated(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy, NULL)) 3648 return -EINVAL; 3649 3650 nla_addr = attrs[IPVS_DEST_ATTR_ADDR]; 3651 nla_port = attrs[IPVS_DEST_ATTR_PORT]; 3652 nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY]; 3653 3654 if (!(nla_addr && nla_port)) 3655 return -EINVAL; 3656 3657 memset(udest, 0, sizeof(*udest)); 3658 3659 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr)); 3660 udest->port = nla_get_be16(nla_port); 3661 3662 if (nla_addr_family) 3663 udest->af = nla_get_u16(nla_addr_family); 3664 else 3665 udest->af = 0; 3666 3667 /* If a full entry was requested, check for the additional fields */ 3668 if (full_entry) { 3669 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh, 3670 *nla_l_thresh, *nla_tun_type, *nla_tun_port, 3671 *nla_tun_flags; 3672 3673 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD]; 3674 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT]; 3675 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH]; 3676 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH]; 3677 nla_tun_type = attrs[IPVS_DEST_ATTR_TUN_TYPE]; 3678 nla_tun_port = attrs[IPVS_DEST_ATTR_TUN_PORT]; 3679 nla_tun_flags = attrs[IPVS_DEST_ATTR_TUN_FLAGS]; 3680 3681 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh)) 3682 return -EINVAL; 3683 3684 udest->conn_flags = nla_get_u32(nla_fwd) 3685 & IP_VS_CONN_F_FWD_MASK; 3686 udest->weight = nla_get_u32(nla_weight); 3687 udest->u_threshold = nla_get_u32(nla_u_thresh); 3688 udest->l_threshold = nla_get_u32(nla_l_thresh); 3689 3690 if (nla_tun_type) 3691 udest->tun_type = nla_get_u8(nla_tun_type); 3692 3693 if (nla_tun_port) 3694 udest->tun_port = nla_get_be16(nla_tun_port); 3695 3696 if (nla_tun_flags) 3697 udest->tun_flags = nla_get_u16(nla_tun_flags); 3698 } 3699 3700 return 0; 3701 } 3702 3703 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state, 3704 struct ipvs_sync_daemon_cfg *c) 3705 { 3706 struct nlattr *nl_daemon; 3707 3708 nl_daemon = nla_nest_start_noflag(skb, IPVS_CMD_ATTR_DAEMON); 3709 if (!nl_daemon) 3710 return -EMSGSIZE; 3711 3712 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) || 3713 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, c->mcast_ifn) || 3714 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, c->syncid) || 3715 nla_put_u16(skb, IPVS_DAEMON_ATTR_SYNC_MAXLEN, c->sync_maxlen) || 3716 nla_put_u16(skb, IPVS_DAEMON_ATTR_MCAST_PORT, c->mcast_port) || 3717 nla_put_u8(skb, IPVS_DAEMON_ATTR_MCAST_TTL, c->mcast_ttl)) 3718 goto nla_put_failure; 3719 #ifdef CONFIG_IP_VS_IPV6 3720 if (c->mcast_af == AF_INET6) { 3721 if (nla_put_in6_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP6, 3722 &c->mcast_group.in6)) 3723 goto nla_put_failure; 3724 } else 3725 #endif 3726 if (c->mcast_af == AF_INET && 3727 nla_put_in_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP, 3728 c->mcast_group.ip)) 3729 goto nla_put_failure; 3730 nla_nest_end(skb, nl_daemon); 3731 3732 return 0; 3733 3734 nla_put_failure: 3735 nla_nest_cancel(skb, nl_daemon); 3736 return -EMSGSIZE; 3737 } 3738 3739 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state, 3740 struct ipvs_sync_daemon_cfg *c, 3741 struct netlink_callback *cb) 3742 { 3743 void *hdr; 3744 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, 3745 &ip_vs_genl_family, NLM_F_MULTI, 3746 IPVS_CMD_NEW_DAEMON); 3747 if (!hdr) 3748 return -EMSGSIZE; 3749 3750 if (ip_vs_genl_fill_daemon(skb, state, c)) 3751 goto nla_put_failure; 3752 3753 genlmsg_end(skb, hdr); 3754 return 0; 3755 3756 nla_put_failure: 3757 genlmsg_cancel(skb, hdr); 3758 return -EMSGSIZE; 3759 } 3760 3761 static int ip_vs_genl_dump_daemons(struct sk_buff *skb, 3762 struct netlink_callback *cb) 3763 { 3764 struct net *net = sock_net(skb->sk); 3765 struct netns_ipvs *ipvs = net_ipvs(net); 3766 3767 mutex_lock(&ipvs->sync_mutex); 3768 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) { 3769 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER, 3770 &ipvs->mcfg, cb) < 0) 3771 goto nla_put_failure; 3772 3773 cb->args[0] = 1; 3774 } 3775 3776 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) { 3777 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP, 3778 &ipvs->bcfg, cb) < 0) 3779 goto nla_put_failure; 3780 3781 cb->args[1] = 1; 3782 } 3783 3784 nla_put_failure: 3785 mutex_unlock(&ipvs->sync_mutex); 3786 3787 return skb->len; 3788 } 3789 3790 static int ip_vs_genl_new_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs) 3791 { 3792 struct ipvs_sync_daemon_cfg c; 3793 struct nlattr *a; 3794 int ret; 3795 3796 memset(&c, 0, sizeof(c)); 3797 if (!(attrs[IPVS_DAEMON_ATTR_STATE] && 3798 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] && 3799 attrs[IPVS_DAEMON_ATTR_SYNC_ID])) 3800 return -EINVAL; 3801 strscpy(c.mcast_ifn, nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]), 3802 sizeof(c.mcast_ifn)); 3803 c.syncid = nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]); 3804 3805 a = attrs[IPVS_DAEMON_ATTR_SYNC_MAXLEN]; 3806 if (a) 3807 c.sync_maxlen = nla_get_u16(a); 3808 3809 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP]; 3810 if (a) { 3811 c.mcast_af = AF_INET; 3812 c.mcast_group.ip = nla_get_in_addr(a); 3813 if (!ipv4_is_multicast(c.mcast_group.ip)) 3814 return -EINVAL; 3815 } else { 3816 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP6]; 3817 if (a) { 3818 #ifdef CONFIG_IP_VS_IPV6 3819 int addr_type; 3820 3821 c.mcast_af = AF_INET6; 3822 c.mcast_group.in6 = nla_get_in6_addr(a); 3823 addr_type = ipv6_addr_type(&c.mcast_group.in6); 3824 if (!(addr_type & IPV6_ADDR_MULTICAST)) 3825 return -EINVAL; 3826 #else 3827 return -EAFNOSUPPORT; 3828 #endif 3829 } 3830 } 3831 3832 a = attrs[IPVS_DAEMON_ATTR_MCAST_PORT]; 3833 if (a) 3834 c.mcast_port = nla_get_u16(a); 3835 3836 a = attrs[IPVS_DAEMON_ATTR_MCAST_TTL]; 3837 if (a) 3838 c.mcast_ttl = nla_get_u8(a); 3839 3840 /* The synchronization protocol is incompatible with mixed family 3841 * services 3842 */ 3843 if (ipvs->mixed_address_family_dests > 0) 3844 return -EINVAL; 3845 3846 ret = start_sync_thread(ipvs, &c, 3847 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE])); 3848 return ret; 3849 } 3850 3851 static int ip_vs_genl_del_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs) 3852 { 3853 int ret; 3854 3855 if (!attrs[IPVS_DAEMON_ATTR_STATE]) 3856 return -EINVAL; 3857 3858 ret = stop_sync_thread(ipvs, 3859 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE])); 3860 return ret; 3861 } 3862 3863 static int ip_vs_genl_set_config(struct netns_ipvs *ipvs, struct nlattr **attrs) 3864 { 3865 struct ip_vs_timeout_user t; 3866 3867 __ip_vs_get_timeouts(ipvs, &t); 3868 3869 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]) 3870 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]); 3871 3872 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]) 3873 t.tcp_fin_timeout = 3874 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]); 3875 3876 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]) 3877 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]); 3878 3879 return ip_vs_set_timeout(ipvs, &t); 3880 } 3881 3882 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info) 3883 { 3884 int ret = -EINVAL, cmd; 3885 struct net *net = sock_net(skb->sk); 3886 struct netns_ipvs *ipvs = net_ipvs(net); 3887 3888 cmd = info->genlhdr->cmd; 3889 3890 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) { 3891 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1]; 3892 3893 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] || 3894 nla_parse_nested_deprecated(daemon_attrs, IPVS_DAEMON_ATTR_MAX, info->attrs[IPVS_CMD_ATTR_DAEMON], ip_vs_daemon_policy, info->extack)) 3895 goto out; 3896 3897 if (cmd == IPVS_CMD_NEW_DAEMON) 3898 ret = ip_vs_genl_new_daemon(ipvs, daemon_attrs); 3899 else 3900 ret = ip_vs_genl_del_daemon(ipvs, daemon_attrs); 3901 } 3902 3903 out: 3904 return ret; 3905 } 3906 3907 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info) 3908 { 3909 bool need_full_svc = false, need_full_dest = false; 3910 struct ip_vs_service *svc = NULL; 3911 struct ip_vs_service_user_kern usvc; 3912 struct ip_vs_dest_user_kern udest; 3913 int ret = 0, cmd; 3914 struct net *net = sock_net(skb->sk); 3915 struct netns_ipvs *ipvs = net_ipvs(net); 3916 3917 cmd = info->genlhdr->cmd; 3918 3919 mutex_lock(&__ip_vs_mutex); 3920 3921 if (cmd == IPVS_CMD_FLUSH) { 3922 ret = ip_vs_flush(ipvs, false); 3923 goto out; 3924 } else if (cmd == IPVS_CMD_SET_CONFIG) { 3925 ret = ip_vs_genl_set_config(ipvs, info->attrs); 3926 goto out; 3927 } else if (cmd == IPVS_CMD_ZERO && 3928 !info->attrs[IPVS_CMD_ATTR_SERVICE]) { 3929 ret = ip_vs_zero_all(ipvs); 3930 goto out; 3931 } 3932 3933 /* All following commands require a service argument, so check if we 3934 * received a valid one. We need a full service specification when 3935 * adding / editing a service. Only identifying members otherwise. */ 3936 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE) 3937 need_full_svc = true; 3938 3939 ret = ip_vs_genl_parse_service(ipvs, &usvc, 3940 info->attrs[IPVS_CMD_ATTR_SERVICE], 3941 need_full_svc, &svc); 3942 if (ret) 3943 goto out; 3944 3945 /* Unless we're adding a new service, the service must already exist */ 3946 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) { 3947 ret = -ESRCH; 3948 goto out; 3949 } 3950 3951 /* Destination commands require a valid destination argument. For 3952 * adding / editing a destination, we need a full destination 3953 * specification. */ 3954 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST || 3955 cmd == IPVS_CMD_DEL_DEST) { 3956 if (cmd != IPVS_CMD_DEL_DEST) 3957 need_full_dest = true; 3958 3959 ret = ip_vs_genl_parse_dest(&udest, 3960 info->attrs[IPVS_CMD_ATTR_DEST], 3961 need_full_dest); 3962 if (ret) 3963 goto out; 3964 3965 /* Old protocols did not allow the user to specify address 3966 * family, so we set it to zero instead. We also didn't 3967 * allow heterogeneous pools in the old code, so it's safe 3968 * to assume that this will have the same address family as 3969 * the service. 3970 */ 3971 if (udest.af == 0) 3972 udest.af = svc->af; 3973 3974 if (!ip_vs_is_af_valid(udest.af)) { 3975 ret = -EAFNOSUPPORT; 3976 goto out; 3977 } 3978 3979 if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) { 3980 /* The synchronization protocol is incompatible 3981 * with mixed family services 3982 */ 3983 if (ipvs->sync_state) { 3984 ret = -EINVAL; 3985 goto out; 3986 } 3987 3988 /* Which connection types do we support? */ 3989 switch (udest.conn_flags) { 3990 case IP_VS_CONN_F_TUNNEL: 3991 /* We are able to forward this */ 3992 break; 3993 default: 3994 ret = -EINVAL; 3995 goto out; 3996 } 3997 } 3998 } 3999 4000 switch (cmd) { 4001 case IPVS_CMD_NEW_SERVICE: 4002 if (svc == NULL) 4003 ret = ip_vs_add_service(ipvs, &usvc, &svc); 4004 else 4005 ret = -EEXIST; 4006 break; 4007 case IPVS_CMD_SET_SERVICE: 4008 ret = ip_vs_edit_service(svc, &usvc); 4009 break; 4010 case IPVS_CMD_DEL_SERVICE: 4011 ret = ip_vs_del_service(svc); 4012 /* do not use svc, it can be freed */ 4013 break; 4014 case IPVS_CMD_NEW_DEST: 4015 ret = ip_vs_add_dest(svc, &udest); 4016 break; 4017 case IPVS_CMD_SET_DEST: 4018 ret = ip_vs_edit_dest(svc, &udest); 4019 break; 4020 case IPVS_CMD_DEL_DEST: 4021 ret = ip_vs_del_dest(svc, &udest); 4022 break; 4023 case IPVS_CMD_ZERO: 4024 ret = ip_vs_zero_service(svc); 4025 break; 4026 default: 4027 ret = -EINVAL; 4028 } 4029 4030 out: 4031 mutex_unlock(&__ip_vs_mutex); 4032 4033 return ret; 4034 } 4035 4036 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info) 4037 { 4038 struct sk_buff *msg; 4039 void *reply; 4040 int ret, cmd, reply_cmd; 4041 struct net *net = sock_net(skb->sk); 4042 struct netns_ipvs *ipvs = net_ipvs(net); 4043 4044 cmd = info->genlhdr->cmd; 4045 4046 if (cmd == IPVS_CMD_GET_SERVICE) 4047 reply_cmd = IPVS_CMD_NEW_SERVICE; 4048 else if (cmd == IPVS_CMD_GET_INFO) 4049 reply_cmd = IPVS_CMD_SET_INFO; 4050 else if (cmd == IPVS_CMD_GET_CONFIG) 4051 reply_cmd = IPVS_CMD_SET_CONFIG; 4052 else { 4053 pr_err("unknown Generic Netlink command\n"); 4054 return -EINVAL; 4055 } 4056 4057 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 4058 if (!msg) 4059 return -ENOMEM; 4060 4061 mutex_lock(&__ip_vs_mutex); 4062 4063 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd); 4064 if (reply == NULL) 4065 goto nla_put_failure; 4066 4067 switch (cmd) { 4068 case IPVS_CMD_GET_SERVICE: 4069 { 4070 struct ip_vs_service *svc; 4071 4072 svc = ip_vs_genl_find_service(ipvs, 4073 info->attrs[IPVS_CMD_ATTR_SERVICE]); 4074 if (IS_ERR(svc)) { 4075 ret = PTR_ERR(svc); 4076 goto out_err; 4077 } else if (svc) { 4078 ret = ip_vs_genl_fill_service(msg, svc); 4079 if (ret) 4080 goto nla_put_failure; 4081 } else { 4082 ret = -ESRCH; 4083 goto out_err; 4084 } 4085 4086 break; 4087 } 4088 4089 case IPVS_CMD_GET_CONFIG: 4090 { 4091 struct ip_vs_timeout_user t; 4092 4093 __ip_vs_get_timeouts(ipvs, &t); 4094 #ifdef CONFIG_IP_VS_PROTO_TCP 4095 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP, 4096 t.tcp_timeout) || 4097 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN, 4098 t.tcp_fin_timeout)) 4099 goto nla_put_failure; 4100 #endif 4101 #ifdef CONFIG_IP_VS_PROTO_UDP 4102 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout)) 4103 goto nla_put_failure; 4104 #endif 4105 4106 break; 4107 } 4108 4109 case IPVS_CMD_GET_INFO: 4110 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION, 4111 IP_VS_VERSION_CODE) || 4112 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE, 4113 ip_vs_conn_tab_size)) 4114 goto nla_put_failure; 4115 break; 4116 } 4117 4118 genlmsg_end(msg, reply); 4119 ret = genlmsg_reply(msg, info); 4120 goto out; 4121 4122 nla_put_failure: 4123 pr_err("not enough space in Netlink message\n"); 4124 ret = -EMSGSIZE; 4125 4126 out_err: 4127 nlmsg_free(msg); 4128 out: 4129 mutex_unlock(&__ip_vs_mutex); 4130 4131 return ret; 4132 } 4133 4134 4135 static const struct genl_small_ops ip_vs_genl_ops[] = { 4136 { 4137 .cmd = IPVS_CMD_NEW_SERVICE, 4138 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, 4139 .flags = GENL_ADMIN_PERM, 4140 .doit = ip_vs_genl_set_cmd, 4141 }, 4142 { 4143 .cmd = IPVS_CMD_SET_SERVICE, 4144 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, 4145 .flags = GENL_ADMIN_PERM, 4146 .doit = ip_vs_genl_set_cmd, 4147 }, 4148 { 4149 .cmd = IPVS_CMD_DEL_SERVICE, 4150 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, 4151 .flags = GENL_ADMIN_PERM, 4152 .doit = ip_vs_genl_set_cmd, 4153 }, 4154 { 4155 .cmd = IPVS_CMD_GET_SERVICE, 4156 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, 4157 .flags = GENL_ADMIN_PERM, 4158 .doit = ip_vs_genl_get_cmd, 4159 .dumpit = ip_vs_genl_dump_services, 4160 }, 4161 { 4162 .cmd = IPVS_CMD_NEW_DEST, 4163 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, 4164 .flags = GENL_ADMIN_PERM, 4165 .doit = ip_vs_genl_set_cmd, 4166 }, 4167 { 4168 .cmd = IPVS_CMD_SET_DEST, 4169 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, 4170 .flags = GENL_ADMIN_PERM, 4171 .doit = ip_vs_genl_set_cmd, 4172 }, 4173 { 4174 .cmd = IPVS_CMD_DEL_DEST, 4175 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, 4176 .flags = GENL_ADMIN_PERM, 4177 .doit = ip_vs_genl_set_cmd, 4178 }, 4179 { 4180 .cmd = IPVS_CMD_GET_DEST, 4181 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, 4182 .flags = GENL_ADMIN_PERM, 4183 .dumpit = ip_vs_genl_dump_dests, 4184 }, 4185 { 4186 .cmd = IPVS_CMD_NEW_DAEMON, 4187 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, 4188 .flags = GENL_ADMIN_PERM, 4189 .doit = ip_vs_genl_set_daemon, 4190 }, 4191 { 4192 .cmd = IPVS_CMD_DEL_DAEMON, 4193 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, 4194 .flags = GENL_ADMIN_PERM, 4195 .doit = ip_vs_genl_set_daemon, 4196 }, 4197 { 4198 .cmd = IPVS_CMD_GET_DAEMON, 4199 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, 4200 .flags = GENL_ADMIN_PERM, 4201 .dumpit = ip_vs_genl_dump_daemons, 4202 }, 4203 { 4204 .cmd = IPVS_CMD_SET_CONFIG, 4205 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, 4206 .flags = GENL_ADMIN_PERM, 4207 .doit = ip_vs_genl_set_cmd, 4208 }, 4209 { 4210 .cmd = IPVS_CMD_GET_CONFIG, 4211 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, 4212 .flags = GENL_ADMIN_PERM, 4213 .doit = ip_vs_genl_get_cmd, 4214 }, 4215 { 4216 .cmd = IPVS_CMD_GET_INFO, 4217 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, 4218 .flags = GENL_ADMIN_PERM, 4219 .doit = ip_vs_genl_get_cmd, 4220 }, 4221 { 4222 .cmd = IPVS_CMD_ZERO, 4223 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, 4224 .flags = GENL_ADMIN_PERM, 4225 .doit = ip_vs_genl_set_cmd, 4226 }, 4227 { 4228 .cmd = IPVS_CMD_FLUSH, 4229 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, 4230 .flags = GENL_ADMIN_PERM, 4231 .doit = ip_vs_genl_set_cmd, 4232 }, 4233 }; 4234 4235 static struct genl_family ip_vs_genl_family __ro_after_init = { 4236 .hdrsize = 0, 4237 .name = IPVS_GENL_NAME, 4238 .version = IPVS_GENL_VERSION, 4239 .maxattr = IPVS_CMD_ATTR_MAX, 4240 .policy = ip_vs_cmd_policy, 4241 .netnsok = true, /* Make ipvsadm to work on netns */ 4242 .module = THIS_MODULE, 4243 .small_ops = ip_vs_genl_ops, 4244 .n_small_ops = ARRAY_SIZE(ip_vs_genl_ops), 4245 .resv_start_op = IPVS_CMD_FLUSH + 1, 4246 }; 4247 4248 static int __init ip_vs_genl_register(void) 4249 { 4250 return genl_register_family(&ip_vs_genl_family); 4251 } 4252 4253 static void ip_vs_genl_unregister(void) 4254 { 4255 genl_unregister_family(&ip_vs_genl_family); 4256 } 4257 4258 /* End of Generic Netlink interface definitions */ 4259 4260 /* 4261 * per netns intit/exit func. 4262 */ 4263 #ifdef CONFIG_SYSCTL 4264 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) 4265 { 4266 struct net *net = ipvs->net; 4267 struct ctl_table *tbl; 4268 int idx, ret; 4269 size_t ctl_table_size = ARRAY_SIZE(vs_vars); 4270 4271 atomic_set(&ipvs->dropentry, 0); 4272 spin_lock_init(&ipvs->dropentry_lock); 4273 spin_lock_init(&ipvs->droppacket_lock); 4274 spin_lock_init(&ipvs->securetcp_lock); 4275 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler); 4276 INIT_DELAYED_WORK(&ipvs->expire_nodest_conn_work, 4277 expire_nodest_conn_handler); 4278 ipvs->est_stopped = 0; 4279 4280 if (!net_eq(net, &init_net)) { 4281 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL); 4282 if (tbl == NULL) 4283 return -ENOMEM; 4284 4285 /* Don't export sysctls to unprivileged users */ 4286 if (net->user_ns != &init_user_ns) { 4287 tbl[0].procname = NULL; 4288 ctl_table_size = 0; 4289 } 4290 } else 4291 tbl = vs_vars; 4292 /* Initialize sysctl defaults */ 4293 for (idx = 0; idx < ARRAY_SIZE(vs_vars); idx++) { 4294 if (tbl[idx].proc_handler == proc_do_defense_mode) 4295 tbl[idx].extra2 = ipvs; 4296 } 4297 idx = 0; 4298 ipvs->sysctl_amemthresh = 1024; 4299 tbl[idx++].data = &ipvs->sysctl_amemthresh; 4300 ipvs->sysctl_am_droprate = 10; 4301 tbl[idx++].data = &ipvs->sysctl_am_droprate; 4302 tbl[idx++].data = &ipvs->sysctl_drop_entry; 4303 tbl[idx++].data = &ipvs->sysctl_drop_packet; 4304 #ifdef CONFIG_IP_VS_NFCT 4305 tbl[idx++].data = &ipvs->sysctl_conntrack; 4306 #endif 4307 tbl[idx++].data = &ipvs->sysctl_secure_tcp; 4308 ipvs->sysctl_snat_reroute = 1; 4309 tbl[idx++].data = &ipvs->sysctl_snat_reroute; 4310 ipvs->sysctl_sync_ver = 1; 4311 tbl[idx++].data = &ipvs->sysctl_sync_ver; 4312 ipvs->sysctl_sync_ports = 1; 4313 tbl[idx++].data = &ipvs->sysctl_sync_ports; 4314 tbl[idx++].data = &ipvs->sysctl_sync_persist_mode; 4315 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32; 4316 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max; 4317 ipvs->sysctl_sync_sock_size = 0; 4318 tbl[idx++].data = &ipvs->sysctl_sync_sock_size; 4319 tbl[idx++].data = &ipvs->sysctl_cache_bypass; 4320 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn; 4321 tbl[idx++].data = &ipvs->sysctl_sloppy_tcp; 4322 tbl[idx++].data = &ipvs->sysctl_sloppy_sctp; 4323 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template; 4324 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD; 4325 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD; 4326 tbl[idx].data = &ipvs->sysctl_sync_threshold; 4327 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold); 4328 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD; 4329 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period; 4330 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3); 4331 tbl[idx++].data = &ipvs->sysctl_sync_retries; 4332 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send; 4333 ipvs->sysctl_pmtu_disc = 1; 4334 tbl[idx++].data = &ipvs->sysctl_pmtu_disc; 4335 tbl[idx++].data = &ipvs->sysctl_backup_only; 4336 ipvs->sysctl_conn_reuse_mode = 1; 4337 tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode; 4338 tbl[idx++].data = &ipvs->sysctl_schedule_icmp; 4339 tbl[idx++].data = &ipvs->sysctl_ignore_tunneled; 4340 ipvs->sysctl_run_estimation = 1; 4341 tbl[idx].extra2 = ipvs; 4342 tbl[idx++].data = &ipvs->sysctl_run_estimation; 4343 4344 ipvs->est_cpulist_valid = 0; 4345 tbl[idx].extra2 = ipvs; 4346 tbl[idx++].data = &ipvs->sysctl_est_cpulist; 4347 4348 ipvs->sysctl_est_nice = IPVS_EST_NICE; 4349 tbl[idx].extra2 = ipvs; 4350 tbl[idx++].data = &ipvs->sysctl_est_nice; 4351 4352 #ifdef CONFIG_IP_VS_DEBUG 4353 /* Global sysctls must be ro in non-init netns */ 4354 if (!net_eq(net, &init_net)) 4355 tbl[idx++].mode = 0444; 4356 #endif 4357 4358 ret = -ENOMEM; 4359 ipvs->sysctl_hdr = register_net_sysctl_sz(net, "net/ipv4/vs", tbl, 4360 ctl_table_size); 4361 if (!ipvs->sysctl_hdr) 4362 goto err; 4363 ipvs->sysctl_tbl = tbl; 4364 4365 ret = ip_vs_start_estimator(ipvs, &ipvs->tot_stats->s); 4366 if (ret < 0) 4367 goto err; 4368 4369 /* Schedule defense work */ 4370 queue_delayed_work(system_long_wq, &ipvs->defense_work, 4371 DEFENSE_TIMER_PERIOD); 4372 4373 return 0; 4374 4375 err: 4376 unregister_net_sysctl_table(ipvs->sysctl_hdr); 4377 if (!net_eq(net, &init_net)) 4378 kfree(tbl); 4379 return ret; 4380 } 4381 4382 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs) 4383 { 4384 struct net *net = ipvs->net; 4385 4386 cancel_delayed_work_sync(&ipvs->expire_nodest_conn_work); 4387 cancel_delayed_work_sync(&ipvs->defense_work); 4388 cancel_work_sync(&ipvs->defense_work.work); 4389 unregister_net_sysctl_table(ipvs->sysctl_hdr); 4390 ip_vs_stop_estimator(ipvs, &ipvs->tot_stats->s); 4391 4392 if (ipvs->est_cpulist_valid) 4393 free_cpumask_var(ipvs->sysctl_est_cpulist); 4394 4395 if (!net_eq(net, &init_net)) 4396 kfree(ipvs->sysctl_tbl); 4397 } 4398 4399 #else 4400 4401 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) { return 0; } 4402 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs) { } 4403 4404 #endif 4405 4406 static struct notifier_block ip_vs_dst_notifier = { 4407 .notifier_call = ip_vs_dst_event, 4408 #ifdef CONFIG_IP_VS_IPV6 4409 .priority = ADDRCONF_NOTIFY_PRIORITY + 5, 4410 #endif 4411 }; 4412 4413 int __net_init ip_vs_control_net_init(struct netns_ipvs *ipvs) 4414 { 4415 int ret = -ENOMEM; 4416 int idx; 4417 4418 /* Initialize rs_table */ 4419 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++) 4420 INIT_HLIST_HEAD(&ipvs->rs_table[idx]); 4421 4422 INIT_LIST_HEAD(&ipvs->dest_trash); 4423 spin_lock_init(&ipvs->dest_trash_lock); 4424 timer_setup(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire, 0); 4425 atomic_set(&ipvs->ftpsvc_counter, 0); 4426 atomic_set(&ipvs->nullsvc_counter, 0); 4427 atomic_set(&ipvs->conn_out_counter, 0); 4428 4429 INIT_DELAYED_WORK(&ipvs->est_reload_work, est_reload_work_handler); 4430 4431 /* procfs stats */ 4432 ipvs->tot_stats = kzalloc(sizeof(*ipvs->tot_stats), GFP_KERNEL); 4433 if (!ipvs->tot_stats) 4434 goto out; 4435 if (ip_vs_stats_init_alloc(&ipvs->tot_stats->s) < 0) 4436 goto err_tot_stats; 4437 4438 #ifdef CONFIG_PROC_FS 4439 if (!proc_create_net("ip_vs", 0, ipvs->net->proc_net, 4440 &ip_vs_info_seq_ops, sizeof(struct ip_vs_iter))) 4441 goto err_vs; 4442 if (!proc_create_net_single("ip_vs_stats", 0, ipvs->net->proc_net, 4443 ip_vs_stats_show, NULL)) 4444 goto err_stats; 4445 if (!proc_create_net_single("ip_vs_stats_percpu", 0, 4446 ipvs->net->proc_net, 4447 ip_vs_stats_percpu_show, NULL)) 4448 goto err_percpu; 4449 #endif 4450 4451 ret = ip_vs_control_net_init_sysctl(ipvs); 4452 if (ret < 0) 4453 goto err; 4454 4455 return 0; 4456 4457 err: 4458 #ifdef CONFIG_PROC_FS 4459 remove_proc_entry("ip_vs_stats_percpu", ipvs->net->proc_net); 4460 4461 err_percpu: 4462 remove_proc_entry("ip_vs_stats", ipvs->net->proc_net); 4463 4464 err_stats: 4465 remove_proc_entry("ip_vs", ipvs->net->proc_net); 4466 4467 err_vs: 4468 #endif 4469 ip_vs_stats_release(&ipvs->tot_stats->s); 4470 4471 err_tot_stats: 4472 kfree(ipvs->tot_stats); 4473 4474 out: 4475 return ret; 4476 } 4477 4478 void __net_exit ip_vs_control_net_cleanup(struct netns_ipvs *ipvs) 4479 { 4480 ip_vs_trash_cleanup(ipvs); 4481 ip_vs_control_net_cleanup_sysctl(ipvs); 4482 cancel_delayed_work_sync(&ipvs->est_reload_work); 4483 #ifdef CONFIG_PROC_FS 4484 remove_proc_entry("ip_vs_stats_percpu", ipvs->net->proc_net); 4485 remove_proc_entry("ip_vs_stats", ipvs->net->proc_net); 4486 remove_proc_entry("ip_vs", ipvs->net->proc_net); 4487 #endif 4488 call_rcu(&ipvs->tot_stats->rcu_head, ip_vs_stats_rcu_free); 4489 } 4490 4491 int __init ip_vs_register_nl_ioctl(void) 4492 { 4493 int ret; 4494 4495 ret = nf_register_sockopt(&ip_vs_sockopts); 4496 if (ret) { 4497 pr_err("cannot register sockopt.\n"); 4498 goto err_sock; 4499 } 4500 4501 ret = ip_vs_genl_register(); 4502 if (ret) { 4503 pr_err("cannot register Generic Netlink interface.\n"); 4504 goto err_genl; 4505 } 4506 return 0; 4507 4508 err_genl: 4509 nf_unregister_sockopt(&ip_vs_sockopts); 4510 err_sock: 4511 return ret; 4512 } 4513 4514 void ip_vs_unregister_nl_ioctl(void) 4515 { 4516 ip_vs_genl_unregister(); 4517 nf_unregister_sockopt(&ip_vs_sockopts); 4518 } 4519 4520 int __init ip_vs_control_init(void) 4521 { 4522 int idx; 4523 int ret; 4524 4525 /* Initialize svc_table, ip_vs_svc_fwm_table */ 4526 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) { 4527 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]); 4528 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]); 4529 } 4530 4531 smp_wmb(); /* Do we really need it now ? */ 4532 4533 ret = register_netdevice_notifier(&ip_vs_dst_notifier); 4534 if (ret < 0) 4535 return ret; 4536 4537 return 0; 4538 } 4539 4540 4541 void ip_vs_control_cleanup(void) 4542 { 4543 unregister_netdevice_notifier(&ip_vs_dst_notifier); 4544 /* relying on common rcu_barrier() in ip_vs_cleanup() */ 4545 } 4546