SAFE public projects git trees. - safe/jmp/linux-2.6/blob - net/ipv4/ipvs/ip_vs_ctl.c

   1 /*
   2  * IPVS         An implementation of the IP virtual server support for the
   3  *              LINUX operating system.  IPVS is now implemented as a module
   4  *              over the NetFilter framework. IPVS can be used to build a
   5  *              high-performance and highly available server based on a
   6  *              cluster of servers.
   7  *
   8  * Authors:     Wensong Zhang <wensong@linuxvirtualserver.org>
   9  *              Peter Kese <peter.kese@ijs.si>
  10  *              Julian Anastasov <ja@ssi.bg>
  11  *
  12  *              This program is free software; you can redistribute it and/or
  13  *              modify it under the terms of the GNU General Public License
  14  *              as published by the Free Software Foundation; either version
  15  *              2 of the License, or (at your option) any later version.
  16  *
  17  * Changes:
  18  *
  19  */
  20
  21 #include <linux/module.h>
  22 #include <linux/init.h>
  23 #include <linux/types.h>
  24 #include <linux/capability.h>
  25 #include <linux/fs.h>
  26 #include <linux/sysctl.h>
  27 #include <linux/proc_fs.h>
  28 #include <linux/workqueue.h>
  29 #include <linux/swap.h>
  30 #include <linux/seq_file.h>
  31
  32 #include <linux/netfilter.h>
  33 #include <linux/netfilter_ipv4.h>
  34 #include <linux/mutex.h>
  35
  36 #include <net/net_namespace.h>
  37 #include <net/ip.h>
  38 #ifdef CONFIG_IP_VS_IPV6
  39 #include <net/ipv6.h>
  40 #include <net/ip6_route.h>
  41 #endif
  42 #include <net/route.h>
  43 #include <net/sock.h>
  44 #include <net/genetlink.h>
  45
  46 #include <asm/uaccess.h>
  47
  48 #include <net/ip_vs.h>
  49
  50 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
  51 static DEFINE_MUTEX(__ip_vs_mutex);
  52
  53 /* lock for service table */
  54 static DEFINE_RWLOCK(__ip_vs_svc_lock);
  55
  56 /* lock for table with the real services */
  57 static DEFINE_RWLOCK(__ip_vs_rs_lock);
  58
  59 /* lock for state and timeout tables */
  60 static DEFINE_RWLOCK(__ip_vs_securetcp_lock);
  61
  62 /* lock for drop entry handling */
  63 static DEFINE_SPINLOCK(__ip_vs_dropentry_lock);
  64
  65 /* lock for drop packet handling */
  66 static DEFINE_SPINLOCK(__ip_vs_droppacket_lock);
  67
  68 /* 1/rate drop and drop-entry variables */
  69 int ip_vs_drop_rate = 0;
  70 int ip_vs_drop_counter = 0;
  71 static atomic_t ip_vs_dropentry = ATOMIC_INIT(0);
  72
  73 /* number of virtual services */
  74 static int ip_vs_num_services = 0;
  75
  76 /* sysctl variables */
  77 static int sysctl_ip_vs_drop_entry = 0;
  78 static int sysctl_ip_vs_drop_packet = 0;
  79 static int sysctl_ip_vs_secure_tcp = 0;
  80 static int sysctl_ip_vs_amemthresh = 1024;
  81 static int sysctl_ip_vs_am_droprate = 10;
  82 int sysctl_ip_vs_cache_bypass = 0;
  83 int sysctl_ip_vs_expire_nodest_conn = 0;
  84 int sysctl_ip_vs_expire_quiescent_template = 0;
  85 int sysctl_ip_vs_sync_threshold[2] = { 3, 50 };
  86 int sysctl_ip_vs_nat_icmp_send = 0;
  87
  88
  89 #ifdef CONFIG_IP_VS_DEBUG
  90 static int sysctl_ip_vs_debug_level = 0;
  91
  92 int ip_vs_get_debug_level(void)
  93 {
  94         return sysctl_ip_vs_debug_level;
  95 }
  96 #endif
  97
  98 #ifdef CONFIG_IP_VS_IPV6
  99 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
 100 static int __ip_vs_addr_is_local_v6(const struct in6_addr *addr)
 101 {
 102         struct rt6_info *rt;
 103         struct flowi fl = {
 104                 .oif = 0,
 105                 .nl_u = {
 106                         .ip6_u = {
 107                                 .daddr = *addr,
 108                                 .saddr = { .s6_addr32 = {0, 0, 0, 0} }, } },
 109         };
 110
 111         rt = (struct rt6_info *)ip6_route_output(&init_net, NULL, &fl);
 112         if (rt && rt->rt6i_dev && (rt->rt6i_dev->flags & IFF_LOOPBACK))
 113                         return 1;
 114
 115         return 0;
 116 }
 117 #endif
 118 /*
 119  *      update_defense_level is called from keventd and from sysctl,
 120  *      so it needs to protect itself from softirqs
 121  */
 122 static void update_defense_level(void)
 123 {
 124         struct sysinfo i;
 125         static int old_secure_tcp = 0;
 126         int availmem;
 127         int nomem;
 128         int to_change = -1;
 129
 130         /* we only count free and buffered memory (in pages) */
 131         si_meminfo(&i);
 132         availmem = i.freeram + i.bufferram;
 133         /* however in linux 2.5 the i.bufferram is total page cache size,
 134            we need adjust it */
 135         /* si_swapinfo(&i); */
 136         /* availmem = availmem - (i.totalswap - i.freeswap); */
 137
 138         nomem = (availmem < sysctl_ip_vs_amemthresh);
 139
 140         local_bh_disable();
 141
 142         /* drop_entry */
 143         spin_lock(&__ip_vs_dropentry_lock);
 144         switch (sysctl_ip_vs_drop_entry) {
 145         case 0:
 146                 atomic_set(&ip_vs_dropentry, 0);
 147                 break;
 148         case 1:
 149                 if (nomem) {
 150                         atomic_set(&ip_vs_dropentry, 1);
 151                         sysctl_ip_vs_drop_entry = 2;
 152                 } else {
 153                         atomic_set(&ip_vs_dropentry, 0);
 154                 }
 155                 break;
 156         case 2:
 157                 if (nomem) {
 158                         atomic_set(&ip_vs_dropentry, 1);
 159                 } else {
 160                         atomic_set(&ip_vs_dropentry, 0);
 161                         sysctl_ip_vs_drop_entry = 1;
 162                 };
 163                 break;
 164         case 3:
 165                 atomic_set(&ip_vs_dropentry, 1);
 166                 break;
 167         }
 168         spin_unlock(&__ip_vs_dropentry_lock);
 169
 170         /* drop_packet */
 171         spin_lock(&__ip_vs_droppacket_lock);
 172         switch (sysctl_ip_vs_drop_packet) {
 173         case 0:
 174                 ip_vs_drop_rate = 0;
 175                 break;
 176         case 1:
 177                 if (nomem) {
 178                         ip_vs_drop_rate = ip_vs_drop_counter
 179                                 = sysctl_ip_vs_amemthresh /
 180                                 (sysctl_ip_vs_amemthresh-availmem);
 181                         sysctl_ip_vs_drop_packet = 2;
 182                 } else {
 183                         ip_vs_drop_rate = 0;
 184                 }
 185                 break;
 186         case 2:
 187                 if (nomem) {
 188                         ip_vs_drop_rate = ip_vs_drop_counter
 189                                 = sysctl_ip_vs_amemthresh /
 190                                 (sysctl_ip_vs_amemthresh-availmem);
 191                 } else {
 192                         ip_vs_drop_rate = 0;
 193                         sysctl_ip_vs_drop_packet = 1;
 194                 }
 195                 break;
 196         case 3:
 197                 ip_vs_drop_rate = sysctl_ip_vs_am_droprate;
 198                 break;
 199         }
 200         spin_unlock(&__ip_vs_droppacket_lock);
 201
 202         /* secure_tcp */
 203         write_lock(&__ip_vs_securetcp_lock);
 204         switch (sysctl_ip_vs_secure_tcp) {
 205         case 0:
 206                 if (old_secure_tcp >= 2)
 207                         to_change = 0;
 208                 break;
 209         case 1:
 210                 if (nomem) {
 211                         if (old_secure_tcp < 2)
 212                                 to_change = 1;
 213                         sysctl_ip_vs_secure_tcp = 2;
 214                 } else {
 215                         if (old_secure_tcp >= 2)
 216                                 to_change = 0;
 217                 }
 218                 break;
 219         case 2:
 220                 if (nomem) {
 221                         if (old_secure_tcp < 2)
 222                                 to_change = 1;
 223                 } else {
 224                         if (old_secure_tcp >= 2)
 225                                 to_change = 0;
 226                         sysctl_ip_vs_secure_tcp = 1;
 227                 }
 228                 break;
 229         case 3:
 230                 if (old_secure_tcp < 2)
 231                         to_change = 1;
 232                 break;
 233         }
 234         old_secure_tcp = sysctl_ip_vs_secure_tcp;
 235         if (to_change >= 0)
 236                 ip_vs_protocol_timeout_change(sysctl_ip_vs_secure_tcp>1);
 237         write_unlock(&__ip_vs_securetcp_lock);
 238
 239         local_bh_enable();
 240 }
 241
 242
 243 /*
 244  *      Timer for checking the defense
 245  */
 246 #define DEFENSE_TIMER_PERIOD    1*HZ
 247 static void defense_work_handler(struct work_struct *work);
 248 static DECLARE_DELAYED_WORK(defense_work, defense_work_handler);
 249
 250 static void defense_work_handler(struct work_struct *work)
 251 {
 252         update_defense_level();
 253         if (atomic_read(&ip_vs_dropentry))
 254                 ip_vs_random_dropentry();
 255
 256         schedule_delayed_work(&defense_work, DEFENSE_TIMER_PERIOD);
 257 }
 258
 259 int
 260 ip_vs_use_count_inc(void)
 261 {
 262         return try_module_get(THIS_MODULE);
 263 }
 264
 265 void
 266 ip_vs_use_count_dec(void)
 267 {
 268         module_put(THIS_MODULE);
 269 }
 270
 271
 272 /*
 273  *      Hash table: for virtual service lookups
 274  */
 275 #define IP_VS_SVC_TAB_BITS 8
 276 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
 277 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
 278
 279 /* the service table hashed by <protocol, addr, port> */
 280 static struct list_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
 281 /* the service table hashed by fwmark */
 282 static struct list_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
 283
 284 /*
 285  *      Hash table: for real service lookups
 286  */
 287 #define IP_VS_RTAB_BITS 4
 288 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
 289 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
 290
 291 static struct list_head ip_vs_rtable[IP_VS_RTAB_SIZE];
 292
 293 /*
 294  *      Trash for destinations
 295  */
 296 static LIST_HEAD(ip_vs_dest_trash);
 297
 298 /*
 299  *      FTP & NULL virtual service counters
 300  */
 301 static atomic_t ip_vs_ftpsvc_counter = ATOMIC_INIT(0);
 302 static atomic_t ip_vs_nullsvc_counter = ATOMIC_INIT(0);
 303
 304
 305 /*
 306  *      Returns hash value for virtual service
 307  */
 308 static __inline__ unsigned
 309 ip_vs_svc_hashkey(int af, unsigned proto, const union nf_inet_addr *addr,
 310                   __be16 port)
 311 {
 312         register unsigned porth = ntohs(port);
 313         __be32 addr_fold = addr->ip;
 314
 315 #ifdef CONFIG_IP_VS_IPV6
 316         if (af == AF_INET6)
 317                 addr_fold = addr->ip6[0]^addr->ip6[1]^
 318                             addr->ip6[2]^addr->ip6[3];
 319 #endif
 320
 321         return (proto^ntohl(addr_fold)^(porth>>IP_VS_SVC_TAB_BITS)^porth)
 322                 & IP_VS_SVC_TAB_MASK;
 323 }
 324
 325 /*
 326  *      Returns hash value of fwmark for virtual service lookup
 327  */
 328 static __inline__ unsigned ip_vs_svc_fwm_hashkey(__u32 fwmark)
 329 {
 330         return fwmark & IP_VS_SVC_TAB_MASK;
 331 }
 332
 333 /*
 334  *      Hashes a service in the ip_vs_svc_table by <proto,addr,port>
 335  *      or in the ip_vs_svc_fwm_table by fwmark.
 336  *      Should be called with locked tables.
 337  */
 338 static int ip_vs_svc_hash(struct ip_vs_service *svc)
 339 {
 340         unsigned hash;
 341
 342         if (svc->flags & IP_VS_SVC_F_HASHED) {
 343                 IP_VS_ERR("ip_vs_svc_hash(): request for already hashed, "
 344                           "called from %p\n", __builtin_return_address(0));
 345                 return 0;
 346         }
 347
 348         if (svc->fwmark == 0) {
 349                 /*
 350                  *  Hash it by <protocol,addr,port> in ip_vs_svc_table
 351                  */
 352                 hash = ip_vs_svc_hashkey(svc->af, svc->protocol, &svc->addr,
 353                                          svc->port);
 354                 list_add(&svc->s_list, &ip_vs_svc_table[hash]);
 355         } else {
 356                 /*
 357                  *  Hash it by fwmark in ip_vs_svc_fwm_table
 358                  */
 359                 hash = ip_vs_svc_fwm_hashkey(svc->fwmark);
 360                 list_add(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
 361         }
 362
 363         svc->flags |= IP_VS_SVC_F_HASHED;
 364         /* increase its refcnt because it is referenced by the svc table */
 365         atomic_inc(&svc->refcnt);
 366         return 1;
 367 }
 368
 369
 370 /*
 371  *      Unhashes a service from ip_vs_svc_table/ip_vs_svc_fwm_table.
 372  *      Should be called with locked tables.
 373  */
 374 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
 375 {
 376         if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
 377                 IP_VS_ERR("ip_vs_svc_unhash(): request for unhash flagged, "
 378                           "called from %p\n", __builtin_return_address(0));
 379                 return 0;
 380         }
 381
 382         if (svc->fwmark == 0) {
 383                 /* Remove it from the ip_vs_svc_table table */
 384                 list_del(&svc->s_list);
 385         } else {
 386                 /* Remove it from the ip_vs_svc_fwm_table table */
 387                 list_del(&svc->f_list);
 388         }
 389
 390         svc->flags &= ~IP_VS_SVC_F_HASHED;
 391         atomic_dec(&svc->refcnt);
 392         return 1;
 393 }
 394
 395
 396 /*
 397  *      Get service by {proto,addr,port} in the service table.
 398  */
 399 static inline struct ip_vs_service *
 400 __ip_vs_service_get(int af, __u16 protocol, const union nf_inet_addr *vaddr,
 401                     __be16 vport)
 402 {
 403         unsigned hash;
 404         struct ip_vs_service *svc;
 405
 406         /* Check for "full" addressed entries */
 407         hash = ip_vs_svc_hashkey(af, protocol, vaddr, vport);
 408
 409         list_for_each_entry(svc, &ip_vs_svc_table[hash], s_list){
 410                 if ((svc->af == af)
 411                     && ip_vs_addr_equal(af, &svc->addr, vaddr)
 412                     && (svc->port == vport)
 413                     && (svc->protocol == protocol)) {
 414                         /* HIT */
 415                         atomic_inc(&svc->usecnt);
 416                         return svc;
 417                 }
 418         }
 419
 420         return NULL;
 421 }
 422
 423
 424 /*
 425  *      Get service by {fwmark} in the service table.
 426  */
 427 static inline struct ip_vs_service *
 428 __ip_vs_svc_fwm_get(int af, __u32 fwmark)
 429 {
 430         unsigned hash;
 431         struct ip_vs_service *svc;
 432
 433         /* Check for fwmark addressed entries */
 434         hash = ip_vs_svc_fwm_hashkey(fwmark);
 435
 436         list_for_each_entry(svc, &ip_vs_svc_fwm_table[hash], f_list) {
 437                 if (svc->fwmark == fwmark && svc->af == af) {
 438                         /* HIT */
 439                         atomic_inc(&svc->usecnt);
 440                         return svc;
 441                 }
 442         }
 443
 444         return NULL;
 445 }
 446
 447 struct ip_vs_service *
 448 ip_vs_service_get(int af, __u32 fwmark, __u16 protocol,
 449                   const union nf_inet_addr *vaddr, __be16 vport)
 450 {
 451         struct ip_vs_service *svc;
 452
 453         read_lock(&__ip_vs_svc_lock);
 454
 455         /*
 456          *      Check the table hashed by fwmark first
 457          */
 458         if (fwmark && (svc = __ip_vs_svc_fwm_get(af, fwmark)))
 459                 goto out;
 460
 461         /*
 462          *      Check the table hashed by <protocol,addr,port>
 463          *      for "full" addressed entries
 464          */
 465         svc = __ip_vs_service_get(af, protocol, vaddr, vport);
 466
 467         if (svc == NULL
 468             && protocol == IPPROTO_TCP
 469             && atomic_read(&ip_vs_ftpsvc_counter)
 470             && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
 471                 /*
 472                  * Check if ftp service entry exists, the packet
 473                  * might belong to FTP data connections.
 474                  */
 475                 svc = __ip_vs_service_get(af, protocol, vaddr, FTPPORT);
 476         }
 477
 478         if (svc == NULL
 479             && atomic_read(&ip_vs_nullsvc_counter)) {
 480                 /*
 481                  * Check if the catch-all port (port zero) exists
 482                  */
 483                 svc = __ip_vs_service_get(af, protocol, vaddr, 0);
 484         }
 485
 486   out:
 487         read_unlock(&__ip_vs_svc_lock);
 488
 489         IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
 490                       fwmark, ip_vs_proto_name(protocol),
 491                       IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
 492                       svc ? "hit" : "not hit");
 493
 494         return svc;
 495 }
 496
 497
 498 static inline void
 499 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
 500 {
 501         atomic_inc(&svc->refcnt);
 502         dest->svc = svc;
 503 }
 504
 505 static inline void
 506 __ip_vs_unbind_svc(struct ip_vs_dest *dest)
 507 {
 508         struct ip_vs_service *svc = dest->svc;
 509
 510         dest->svc = NULL;
 511         if (atomic_dec_and_test(&svc->refcnt))
 512                 kfree(svc);
 513 }
 514
 515
 516 /*
 517  *      Returns hash value for real service
 518  */
 519 static inline unsigned ip_vs_rs_hashkey(int af,
 520                                             const union nf_inet_addr *addr,
 521                                             __be16 port)
 522 {
 523         register unsigned porth = ntohs(port);
 524         __be32 addr_fold = addr->ip;
 525
 526 #ifdef CONFIG_IP_VS_IPV6
 527         if (af == AF_INET6)
 528                 addr_fold = addr->ip6[0]^addr->ip6[1]^
 529                             addr->ip6[2]^addr->ip6[3];
 530 #endif
 531
 532         return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
 533                 & IP_VS_RTAB_MASK;
 534 }
 535
 536 /*
 537  *      Hashes ip_vs_dest in ip_vs_rtable by <proto,addr,port>.
 538  *      should be called with locked tables.
 539  */
 540 static int ip_vs_rs_hash(struct ip_vs_dest *dest)
 541 {
 542         unsigned hash;
 543
 544         if (!list_empty(&dest->d_list)) {
 545                 return 0;
 546         }
 547
 548         /*
 549          *      Hash by proto,addr,port,
 550          *      which are the parameters of the real service.
 551          */
 552         hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
 553
 554         list_add(&dest->d_list, &ip_vs_rtable[hash]);
 555
 556         return 1;
 557 }
 558
 559 /*
 560  *      UNhashes ip_vs_dest from ip_vs_rtable.
 561  *      should be called with locked tables.
 562  */
 563 static int ip_vs_rs_unhash(struct ip_vs_dest *dest)
 564 {
 565         /*
 566          * Remove it from the ip_vs_rtable table.
 567          */
 568         if (!list_empty(&dest->d_list)) {
 569                 list_del(&dest->d_list);
 570                 INIT_LIST_HEAD(&dest->d_list);
 571         }
 572
 573         return 1;
 574 }
 575
 576 /*
 577  *      Lookup real service by <proto,addr,port> in the real service table.
 578  */
 579 struct ip_vs_dest *
 580 ip_vs_lookup_real_service(int af, __u16 protocol,
 581                           const union nf_inet_addr *daddr,
 582                           __be16 dport)
 583 {
 584         unsigned hash;
 585         struct ip_vs_dest *dest;
 586
 587         /*
 588          *      Check for "full" addressed entries
 589          *      Return the first found entry
 590          */
 591         hash = ip_vs_rs_hashkey(af, daddr, dport);
 592
 593         read_lock(&__ip_vs_rs_lock);
 594         list_for_each_entry(dest, &ip_vs_rtable[hash], d_list) {
 595                 if ((dest->af == af)
 596                     && ip_vs_addr_equal(af, &dest->addr, daddr)
 597                     && (dest->port == dport)
 598                     && ((dest->protocol == protocol) ||
 599                         dest->vfwmark)) {
 600                         /* HIT */
 601                         read_unlock(&__ip_vs_rs_lock);
 602                         return dest;
 603                 }
 604         }
 605         read_unlock(&__ip_vs_rs_lock);
 606
 607         return NULL;
 608 }
 609
 610 /*
 611  *      Lookup destination by {addr,port} in the given service
 612  */
 613 static struct ip_vs_dest *
 614 ip_vs_lookup_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
 615                   __be16 dport)
 616 {
 617         struct ip_vs_dest *dest;
 618
 619         /*
 620          * Find the destination for the given service
 621          */
 622         list_for_each_entry(dest, &svc->destinations, n_list) {
 623                 if ((dest->af == svc->af)
 624                     && ip_vs_addr_equal(svc->af, &dest->addr, daddr)
 625                     && (dest->port == dport)) {
 626                         /* HIT */
 627                         return dest;
 628                 }
 629         }
 630
 631         return NULL;
 632 }
 633
 634 /*
 635  * Find destination by {daddr,dport,vaddr,protocol}
 636  * Cretaed to be used in ip_vs_process_message() in
 637  * the backup synchronization daemon. It finds the
 638  * destination to be bound to the received connection
 639  * on the backup.
 640  *
 641  * ip_vs_lookup_real_service() looked promissing, but
 642  * seems not working as expected.
 643  */
 644 struct ip_vs_dest *ip_vs_find_dest(int af, const union nf_inet_addr *daddr,
 645                                    __be16 dport,
 646                                    const union nf_inet_addr *vaddr,
 647                                    __be16 vport, __u16 protocol)
 648 {
 649         struct ip_vs_dest *dest;
 650         struct ip_vs_service *svc;
 651
 652         svc = ip_vs_service_get(af, 0, protocol, vaddr, vport);
 653         if (!svc)
 654                 return NULL;
 655         dest = ip_vs_lookup_dest(svc, daddr, dport);
 656         if (dest)
 657                 atomic_inc(&dest->refcnt);
 658         ip_vs_service_put(svc);
 659         return dest;
 660 }
 661
 662 /*
 663  *  Lookup dest by {svc,addr,port} in the destination trash.
 664  *  The destination trash is used to hold the destinations that are removed
 665  *  from the service table but are still referenced by some conn entries.
 666  *  The reason to add the destination trash is when the dest is temporary
 667  *  down (either by administrator or by monitor program), the dest can be
 668  *  picked back from the trash, the remaining connections to the dest can
 669  *  continue, and the counting information of the dest is also useful for
 670  *  scheduling.
 671  */
 672 static struct ip_vs_dest *
 673 ip_vs_trash_get_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
 674                      __be16 dport)
 675 {
 676         struct ip_vs_dest *dest, *nxt;
 677
 678         /*
 679          * Find the destination in trash
 680          */
 681         list_for_each_entry_safe(dest, nxt, &ip_vs_dest_trash, n_list) {
 682                 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
 683                               "dest->refcnt=%d\n",
 684                               dest->vfwmark,
 685                               IP_VS_DBG_ADDR(svc->af, &dest->addr),
 686                               ntohs(dest->port),
 687                               atomic_read(&dest->refcnt));
 688                 if (dest->af == svc->af &&
 689                     ip_vs_addr_equal(svc->af, &dest->addr, daddr) &&
 690                     dest->port == dport &&
 691                     dest->vfwmark == svc->fwmark &&
 692                     dest->protocol == svc->protocol &&
 693                     (svc->fwmark ||
 694                      (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
 695                       dest->vport == svc->port))) {
 696                         /* HIT */
 697                         return dest;
 698                 }
 699
 700                 /*
 701                  * Try to purge the destination from trash if not referenced
 702                  */
 703                 if (atomic_read(&dest->refcnt) == 1) {
 704                         IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u "
 705                                       "from trash\n",
 706                                       dest->vfwmark,
 707                                       IP_VS_DBG_ADDR(svc->af, &dest->addr),
 708                                       ntohs(dest->port));
 709                         list_del(&dest->n_list);
 710                         ip_vs_dst_reset(dest);
 711                         __ip_vs_unbind_svc(dest);
 712                         kfree(dest);
 713                 }
 714         }
 715
 716         return NULL;
 717 }
 718
 719
 720 /*
 721  *  Clean up all the destinations in the trash
 722  *  Called by the ip_vs_control_cleanup()
 723  *
 724  *  When the ip_vs_control_clearup is activated by ipvs module exit,
 725  *  the service tables must have been flushed and all the connections
 726  *  are expired, and the refcnt of each destination in the trash must
 727  *  be 1, so we simply release them here.
 728  */
 729 static void ip_vs_trash_cleanup(void)
 730 {
 731         struct ip_vs_dest *dest, *nxt;
 732
 733         list_for_each_entry_safe(dest, nxt, &ip_vs_dest_trash, n_list) {
 734                 list_del(&dest->n_list);
 735                 ip_vs_dst_reset(dest);
 736                 __ip_vs_unbind_svc(dest);
 737                 kfree(dest);
 738         }
 739 }
 740
 741
 742 static void
 743 ip_vs_zero_stats(struct ip_vs_stats *stats)
 744 {
 745         spin_lock_bh(&stats->lock);
 746
 747         memset(&stats->ustats, 0, sizeof(stats->ustats));
 748         ip_vs_zero_estimator(stats);
 749
 750         spin_unlock_bh(&stats->lock);
 751 }
 752
 753 /*
 754  *      Update a destination in the given service
 755  */
 756 static void
 757 __ip_vs_update_dest(struct ip_vs_service *svc,
 758                     struct ip_vs_dest *dest, struct ip_vs_dest_user_kern *udest)
 759 {
 760         int conn_flags;
 761
 762         /* set the weight and the flags */
 763         atomic_set(&dest->weight, udest->weight);
 764         conn_flags = udest->conn_flags | IP_VS_CONN_F_INACTIVE;
 765
 766         /* check if local node and update the flags */
 767 #ifdef CONFIG_IP_VS_IPV6
 768         if (svc->af == AF_INET6) {
 769                 if (__ip_vs_addr_is_local_v6(&udest->addr.in6)) {
 770                         conn_flags = (conn_flags & ~IP_VS_CONN_F_FWD_MASK)
 771                                 | IP_VS_CONN_F_LOCALNODE;
 772                 }
 773         } else
 774 #endif
 775                 if (inet_addr_type(&init_net, udest->addr.ip) == RTN_LOCAL) {
 776                         conn_flags = (conn_flags & ~IP_VS_CONN_F_FWD_MASK)
 777                                 | IP_VS_CONN_F_LOCALNODE;
 778                 }
 779
 780         /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
 781         if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != 0) {
 782                 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
 783         } else {
 784                 /*
 785                  *    Put the real service in ip_vs_rtable if not present.
 786                  *    For now only for NAT!
 787                  */
 788                 write_lock_bh(&__ip_vs_rs_lock);
 789                 ip_vs_rs_hash(dest);
 790                 write_unlock_bh(&__ip_vs_rs_lock);
 791         }
 792         atomic_set(&dest->conn_flags, conn_flags);
 793
 794         /* bind the service */
 795         if (!dest->svc) {
 796                 __ip_vs_bind_svc(dest, svc);
 797         } else {
 798                 if (dest->svc != svc) {
 799                         __ip_vs_unbind_svc(dest);
 800                         ip_vs_zero_stats(&dest->stats);
 801                         __ip_vs_bind_svc(dest, svc);
 802                 }
 803         }
 804
 805         /* set the dest status flags */
 806         dest->flags |= IP_VS_DEST_F_AVAILABLE;
 807
 808         if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
 809                 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
 810         dest->u_threshold = udest->u_threshold;
 811         dest->l_threshold = udest->l_threshold;
 812 }
 813
 814
 815 /*
 816  *      Create a destination for the given service
 817  */
 818 static int
 819 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
 820                struct ip_vs_dest **dest_p)
 821 {
 822         struct ip_vs_dest *dest;
 823         unsigned atype;
 824
 825         EnterFunction(2);
 826
 827 #ifdef CONFIG_IP_VS_IPV6
 828         if (svc->af == AF_INET6) {
 829                 atype = ipv6_addr_type(&udest->addr.in6);
 830                 if ((!(atype & IPV6_ADDR_UNICAST) ||
 831                         atype & IPV6_ADDR_LINKLOCAL) &&
 832                         !__ip_vs_addr_is_local_v6(&udest->addr.in6))
 833                         return -EINVAL;
 834         } else
 835 #endif
 836         {
 837                 atype = inet_addr_type(&init_net, udest->addr.ip);
 838                 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
 839                         return -EINVAL;
 840         }
 841
 842         dest = kzalloc(sizeof(struct ip_vs_dest), GFP_ATOMIC);
 843         if (dest == NULL) {
 844                 IP_VS_ERR("ip_vs_new_dest: kmalloc failed.\n");
 845                 return -ENOMEM;
 846         }
 847
 848         dest->af = svc->af;
 849         dest->protocol = svc->protocol;
 850         dest->vaddr = svc->addr;
 851         dest->vport = svc->port;
 852         dest->vfwmark = svc->fwmark;
 853         ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
 854         dest->port = udest->port;
 855
 856         atomic_set(&dest->activeconns, 0);
 857         atomic_set(&dest->inactconns, 0);
 858         atomic_set(&dest->persistconns, 0);
 859         atomic_set(&dest->refcnt, 0);
 860
 861         INIT_LIST_HEAD(&dest->d_list);
 862         spin_lock_init(&dest->dst_lock);
 863         spin_lock_init(&dest->stats.lock);
 864         __ip_vs_update_dest(svc, dest, udest);
 865         ip_vs_new_estimator(&dest->stats);
 866
 867         *dest_p = dest;
 868
 869         LeaveFunction(2);
 870         return 0;
 871 }
 872
 873
 874 /*
 875  *      Add a destination into an existing service
 876  */
 877 static int
 878 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
 879 {
 880         struct ip_vs_dest *dest;
 881         union nf_inet_addr daddr;
 882         __be16 dport = udest->port;
 883         int ret;
 884
 885         EnterFunction(2);
 886
 887         if (udest->weight < 0) {
 888                 IP_VS_ERR("ip_vs_add_dest(): server weight less than zero\n");
 889                 return -ERANGE;
 890         }
 891
 892         if (udest->l_threshold > udest->u_threshold) {
 893                 IP_VS_ERR("ip_vs_add_dest(): lower threshold is higher than "
 894                           "upper threshold\n");
 895                 return -ERANGE;
 896         }
 897
 898         ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
 899
 900         /*
 901          * Check if the dest already exists in the list
 902          */
 903         dest = ip_vs_lookup_dest(svc, &daddr, dport);
 904
 905         if (dest != NULL) {
 906                 IP_VS_DBG(1, "ip_vs_add_dest(): dest already exists\n");
 907                 return -EEXIST;
 908         }
 909
 910         /*
 911          * Check if the dest already exists in the trash and
 912          * is from the same service
 913          */
 914         dest = ip_vs_trash_get_dest(svc, &daddr, dport);
 915
 916         if (dest != NULL) {
 917                 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
 918                               "dest->refcnt=%d, service %u/%s:%u\n",
 919                               IP_VS_DBG_ADDR(svc->af, &daddr), ntohs(dport),
 920                               atomic_read(&dest->refcnt),
 921                               dest->vfwmark,
 922                               IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
 923                               ntohs(dest->vport));
 924
 925                 __ip_vs_update_dest(svc, dest, udest);
 926
 927                 /*
 928                  * Get the destination from the trash
 929                  */
 930                 list_del(&dest->n_list);
 931
 932                 ip_vs_new_estimator(&dest->stats);
 933
 934                 write_lock_bh(&__ip_vs_svc_lock);
 935
 936                 /*
 937                  * Wait until all other svc users go away.
 938                  */
 939                 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
 940
 941                 list_add(&dest->n_list, &svc->destinations);
 942                 svc->num_dests++;
 943
 944                 /* call the update_service function of its scheduler */
 945                 if (svc->scheduler->update_service)
 946                         svc->scheduler->update_service(svc);
 947
 948                 write_unlock_bh(&__ip_vs_svc_lock);
 949                 return 0;
 950         }
 951
 952         /*
 953          * Allocate and initialize the dest structure
 954          */
 955         ret = ip_vs_new_dest(svc, udest, &dest);
 956         if (ret) {
 957                 return ret;
 958         }
 959
 960         /*
 961          * Add the dest entry into the list
 962          */
 963         atomic_inc(&dest->refcnt);
 964
 965         write_lock_bh(&__ip_vs_svc_lock);
 966
 967         /*
 968          * Wait until all other svc users go away.
 969          */
 970         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
 971
 972         list_add(&dest->n_list, &svc->destinations);
 973         svc->num_dests++;
 974
 975         /* call the update_service function of its scheduler */
 976         if (svc->scheduler->update_service)
 977                 svc->scheduler->update_service(svc);
 978
 979         write_unlock_bh(&__ip_vs_svc_lock);
 980
 981         LeaveFunction(2);
 982
 983         return 0;
 984 }
 985
 986
 987 /*
 988  *      Edit a destination in the given service
 989  */
 990 static int
 991 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
 992 {
 993         struct ip_vs_dest *dest;
 994         union nf_inet_addr daddr;
 995         __be16 dport = udest->port;
 996
 997         EnterFunction(2);
 998
 999         if (udest->weight < 0) {
1000                 IP_VS_ERR("ip_vs_edit_dest(): server weight less than zero\n");
1001                 return -ERANGE;
1002         }
1003
1004         if (udest->l_threshold > udest->u_threshold) {
1005                 IP_VS_ERR("ip_vs_edit_dest(): lower threshold is higher than "
1006                           "upper threshold\n");
1007                 return -ERANGE;
1008         }
1009
1010         ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
1011
1012         /*
1013          *  Lookup the destination list
1014          */
1015         dest = ip_vs_lookup_dest(svc, &daddr, dport);
1016
1017         if (dest == NULL) {
1018                 IP_VS_DBG(1, "ip_vs_edit_dest(): dest doesn't exist\n");
1019                 return -ENOENT;
1020         }
1021
1022         __ip_vs_update_dest(svc, dest, udest);
1023
1024         write_lock_bh(&__ip_vs_svc_lock);
1025
1026         /* Wait until all other svc users go away */
1027         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
1028
1029         /* call the update_service, because server weight may be changed */
1030         if (svc->scheduler->update_service)
1031                 svc->scheduler->update_service(svc);
1032
1033         write_unlock_bh(&__ip_vs_svc_lock);
1034
1035         LeaveFunction(2);
1036
1037         return 0;
1038 }
1039
1040
1041 /*
1042  *      Delete a destination (must be already unlinked from the service)
1043  */
1044 static void __ip_vs_del_dest(struct ip_vs_dest *dest)
1045 {
1046         ip_vs_kill_estimator(&dest->stats);
1047
1048         /*
1049          *  Remove it from the d-linked list with the real services.
1050          */
1051         write_lock_bh(&__ip_vs_rs_lock);
1052         ip_vs_rs_unhash(dest);
1053         write_unlock_bh(&__ip_vs_rs_lock);
1054
1055         /*
1056          *  Decrease the refcnt of the dest, and free the dest
1057          *  if nobody refers to it (refcnt=0). Otherwise, throw
1058          *  the destination into the trash.
1059          */
1060         if (atomic_dec_and_test(&dest->refcnt)) {
1061                 ip_vs_dst_reset(dest);
1062                 /* simply decrease svc->refcnt here, let the caller check
1063                    and release the service if nobody refers to it.
1064                    Only user context can release destination and service,
1065                    and only one user context can update virtual service at a
1066                    time, so the operation here is OK */
1067                 atomic_dec(&dest->svc->refcnt);
1068                 kfree(dest);
1069         } else {
1070                 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, "
1071                               "dest->refcnt=%d\n",
1072                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1073                               ntohs(dest->port),
1074                               atomic_read(&dest->refcnt));
1075                 list_add(&dest->n_list, &ip_vs_dest_trash);
1076                 atomic_inc(&dest->refcnt);
1077         }
1078 }
1079
1080
1081 /*
1082  *      Unlink a destination from the given service
1083  */
1084 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1085                                 struct ip_vs_dest *dest,
1086                                 int svcupd)
1087 {
1088         dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1089
1090         /*
1091          *  Remove it from the d-linked destination list.
1092          */
1093         list_del(&dest->n_list);
1094         svc->num_dests--;
1095
1096         /*
1097          *  Call the update_service function of its scheduler
1098          */
1099         if (svcupd && svc->scheduler->update_service)
1100                         svc->scheduler->update_service(svc);
1101 }
1102
1103
1104 /*
1105  *      Delete a destination server in the given service
1106  */
1107 static int
1108 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1109 {
1110         struct ip_vs_dest *dest;
1111         __be16 dport = udest->port;
1112
1113         EnterFunction(2);
1114
1115         dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
1116
1117         if (dest == NULL) {
1118                 IP_VS_DBG(1, "ip_vs_del_dest(): destination not found!\n");
1119                 return -ENOENT;
1120         }
1121
1122         write_lock_bh(&__ip_vs_svc_lock);
1123
1124         /*
1125          *      Wait until all other svc users go away.
1126          */
1127         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
1128
1129         /*
1130          *      Unlink dest from the service
1131          */
1132         __ip_vs_unlink_dest(svc, dest, 1);
1133
1134         write_unlock_bh(&__ip_vs_svc_lock);
1135
1136         /*
1137          *      Delete the destination
1138          */
1139         __ip_vs_del_dest(dest);
1140
1141         LeaveFunction(2);
1142
1143         return 0;
1144 }
1145
1146
1147 /*
1148  *      Add a service into the service hash table
1149  */
1150 static int
1151 ip_vs_add_service(struct ip_vs_service_user_kern *u,
1152                   struct ip_vs_service **svc_p)
1153 {
1154         int ret = 0;
1155         struct ip_vs_scheduler *sched = NULL;
1156         struct ip_vs_service *svc = NULL;
1157
1158         /* increase the module use count */
1159         ip_vs_use_count_inc();
1160
1161         /* Lookup the scheduler by 'u->sched_name' */
1162         sched = ip_vs_scheduler_get(u->sched_name);
1163         if (sched == NULL) {
1164                 IP_VS_INFO("Scheduler module ip_vs_%s not found\n",
1165                            u->sched_name);
1166                 ret = -ENOENT;
1167                 goto out_mod_dec;
1168         }
1169
1170 #ifdef CONFIG_IP_VS_IPV6
1171         if (u->af == AF_INET6) {
1172                 if (!sched->supports_ipv6) {
1173                         ret = -EAFNOSUPPORT;
1174                         goto out_err;
1175                 }
1176                 if ((u->netmask < 1) || (u->netmask > 128)) {
1177                         ret = -EINVAL;
1178                         goto out_err;
1179                 }
1180         }
1181 #endif
1182
1183         svc = kzalloc(sizeof(struct ip_vs_service), GFP_ATOMIC);
1184         if (svc == NULL) {
1185                 IP_VS_DBG(1, "ip_vs_add_service: kmalloc failed.\n");
1186                 ret = -ENOMEM;
1187                 goto out_err;
1188         }
1189
1190         /* I'm the first user of the service */
1191         atomic_set(&svc->usecnt, 1);
1192         atomic_set(&svc->refcnt, 0);
1193
1194         svc->af = u->af;
1195         svc->protocol = u->protocol;
1196         ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1197         svc->port = u->port;
1198         svc->fwmark = u->fwmark;
1199         svc->flags = u->flags;
1200         svc->timeout = u->timeout * HZ;
1201         svc->netmask = u->netmask;
1202
1203         INIT_LIST_HEAD(&svc->destinations);
1204         rwlock_init(&svc->sched_lock);
1205         spin_lock_init(&svc->stats.lock);
1206
1207         /* Bind the scheduler */
1208         ret = ip_vs_bind_scheduler(svc, sched);
1209         if (ret)
1210                 goto out_err;
1211         sched = NULL;
1212
1213         /* Update the virtual service counters */
1214         if (svc->port == FTPPORT)
1215                 atomic_inc(&ip_vs_ftpsvc_counter);
1216         else if (svc->port == 0)
1217                 atomic_inc(&ip_vs_nullsvc_counter);
1218
1219         ip_vs_new_estimator(&svc->stats);
1220
1221         /* Count only IPv4 services for old get/setsockopt interface */
1222         if (svc->af == AF_INET)
1223                 ip_vs_num_services++;
1224
1225         /* Hash the service into the service table */
1226         write_lock_bh(&__ip_vs_svc_lock);
1227         ip_vs_svc_hash(svc);
1228         write_unlock_bh(&__ip_vs_svc_lock);
1229
1230         *svc_p = svc;
1231         return 0;
1232
1233   out_err:
1234         if (svc != NULL) {
1235                 if (svc->scheduler)
1236                         ip_vs_unbind_scheduler(svc);
1237                 if (svc->inc) {
1238                         local_bh_disable();
1239                         ip_vs_app_inc_put(svc->inc);
1240                         local_bh_enable();
1241                 }
1242                 kfree(svc);
1243         }
1244         ip_vs_scheduler_put(sched);
1245
1246   out_mod_dec:
1247         /* decrease the module use count */
1248         ip_vs_use_count_dec();
1249
1250         return ret;
1251 }
1252
1253
1254 /*
1255  *      Edit a service and bind it with a new scheduler
1256  */
1257 static int
1258 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1259 {
1260         struct ip_vs_scheduler *sched, *old_sched;
1261         int ret = 0;
1262
1263         /*
1264          * Lookup the scheduler, by 'u->sched_name'
1265          */
1266         sched = ip_vs_scheduler_get(u->sched_name);
1267         if (sched == NULL) {
1268                 IP_VS_INFO("Scheduler module ip_vs_%s not found\n",
1269                            u->sched_name);
1270                 return -ENOENT;
1271         }
1272         old_sched = sched;
1273
1274 #ifdef CONFIG_IP_VS_IPV6
1275         if (u->af == AF_INET6) {
1276                 if (!sched->supports_ipv6) {
1277                         ret = -EAFNOSUPPORT;
1278                         goto out;
1279                 }
1280                 if ((u->netmask < 1) || (u->netmask > 128)) {
1281                         ret = -EINVAL;
1282                         goto out;
1283                 }
1284         }
1285 #endif
1286
1287         write_lock_bh(&__ip_vs_svc_lock);
1288
1289         /*
1290          * Wait until all other svc users go away.
1291          */
1292         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
1293
1294         /*
1295          * Set the flags and timeout value
1296          */
1297         svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1298         svc->timeout = u->timeout * HZ;
1299         svc->netmask = u->netmask;
1300
1301         old_sched = svc->scheduler;
1302         if (sched != old_sched) {
1303                 /*
1304                  * Unbind the old scheduler
1305                  */
1306                 if ((ret = ip_vs_unbind_scheduler(svc))) {
1307                         old_sched = sched;
1308                         goto out_unlock;
1309                 }
1310
1311                 /*
1312                  * Bind the new scheduler
1313                  */
1314                 if ((ret = ip_vs_bind_scheduler(svc, sched))) {
1315                         /*
1316                          * If ip_vs_bind_scheduler fails, restore the old
1317                          * scheduler.
1318                          * The main reason of failure is out of memory.
1319                          *
1320                          * The question is if the old scheduler can be
1321                          * restored all the time. TODO: if it cannot be
1322                          * restored some time, we must delete the service,
1323                          * otherwise the system may crash.
1324                          */
1325                         ip_vs_bind_scheduler(svc, old_sched);
1326                         old_sched = sched;
1327                         goto out_unlock;
1328                 }
1329         }
1330
1331   out_unlock:
1332         write_unlock_bh(&__ip_vs_svc_lock);
1333   out:
1334
1335         if (old_sched)
1336                 ip_vs_scheduler_put(old_sched);
1337
1338         return ret;
1339 }
1340
1341
1342 /*
1343  *      Delete a service from the service list
1344  *      - The service must be unlinked, unlocked and not referenced!
1345  *      - We are called under _bh lock
1346  */
1347 static void __ip_vs_del_service(struct ip_vs_service *svc)
1348 {
1349         struct ip_vs_dest *dest, *nxt;
1350         struct ip_vs_scheduler *old_sched;
1351
1352         /* Count only IPv4 services for old get/setsockopt interface */
1353         if (svc->af == AF_INET)
1354                 ip_vs_num_services--;
1355
1356         ip_vs_kill_estimator(&svc->stats);
1357
1358         /* Unbind scheduler */
1359         old_sched = svc->scheduler;
1360         ip_vs_unbind_scheduler(svc);
1361         if (old_sched)
1362                 ip_vs_scheduler_put(old_sched);
1363
1364         /* Unbind app inc */
1365         if (svc->inc) {
1366                 ip_vs_app_inc_put(svc->inc);
1367                 svc->inc = NULL;
1368         }
1369
1370         /*
1371          *    Unlink the whole destination list
1372          */
1373         list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1374                 __ip_vs_unlink_dest(svc, dest, 0);
1375                 __ip_vs_del_dest(dest);
1376         }
1377
1378         /*
1379          *    Update the virtual service counters
1380          */
1381         if (svc->port == FTPPORT)
1382                 atomic_dec(&ip_vs_ftpsvc_counter);
1383         else if (svc->port == 0)
1384                 atomic_dec(&ip_vs_nullsvc_counter);
1385
1386         /*
1387          *    Free the service if nobody refers to it
1388          */
1389         if (atomic_read(&svc->refcnt) == 0)
1390                 kfree(svc);
1391
1392         /* decrease the module use count */
1393         ip_vs_use_count_dec();
1394 }
1395
1396 /*
1397  *      Delete a service from the service list
1398  */
1399 static int ip_vs_del_service(struct ip_vs_service *svc)
1400 {
1401         if (svc == NULL)
1402                 return -EEXIST;
1403
1404         /*
1405          * Unhash it from the service table
1406          */
1407         write_lock_bh(&__ip_vs_svc_lock);
1408
1409         ip_vs_svc_unhash(svc);
1410
1411         /*
1412          * Wait until all the svc users go away.
1413          */
1414         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
1415
1416         __ip_vs_del_service(svc);
1417
1418         write_unlock_bh(&__ip_vs_svc_lock);
1419
1420         return 0;
1421 }
1422
1423
1424 /*
1425  *      Flush all the virtual services
1426  */
1427 static int ip_vs_flush(void)
1428 {
1429         int idx;
1430         struct ip_vs_service *svc, *nxt;
1431
1432         /*
1433          * Flush the service table hashed by <protocol,addr,port>
1434          */
1435         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1436                 list_for_each_entry_safe(svc, nxt, &ip_vs_svc_table[idx], s_list) {
1437                         write_lock_bh(&__ip_vs_svc_lock);
1438                         ip_vs_svc_unhash(svc);
1439                         /*
1440                          * Wait until all the svc users go away.
1441                          */
1442                         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1443                         __ip_vs_del_service(svc);
1444                         write_unlock_bh(&__ip_vs_svc_lock);
1445                 }
1446         }
1447
1448         /*
1449          * Flush the service table hashed by fwmark
1450          */
1451         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1452                 list_for_each_entry_safe(svc, nxt,
1453                                          &ip_vs_svc_fwm_table[idx], f_list) {
1454                         write_lock_bh(&__ip_vs_svc_lock);
1455                         ip_vs_svc_unhash(svc);
1456                         /*
1457                          * Wait until all the svc users go away.
1458                          */
1459                         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1460                         __ip_vs_del_service(svc);
1461                         write_unlock_bh(&__ip_vs_svc_lock);
1462                 }
1463         }
1464
1465         return 0;
1466 }
1467
1468
1469 /*
1470  *      Zero counters in a service or all services
1471  */
1472 static int ip_vs_zero_service(struct ip_vs_service *svc)
1473 {
1474         struct ip_vs_dest *dest;
1475
1476         write_lock_bh(&__ip_vs_svc_lock);
1477         list_for_each_entry(dest, &svc->destinations, n_list) {
1478                 ip_vs_zero_stats(&dest->stats);
1479         }
1480         ip_vs_zero_stats(&svc->stats);
1481         write_unlock_bh(&__ip_vs_svc_lock);
1482         return 0;
1483 }
1484
1485 static int ip_vs_zero_all(void)
1486 {
1487         int idx;
1488         struct ip_vs_service *svc;
1489
1490         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1491                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1492                         ip_vs_zero_service(svc);
1493                 }
1494         }
1495
1496         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1497                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1498                         ip_vs_zero_service(svc);
1499                 }
1500         }
1501
1502         ip_vs_zero_stats(&ip_vs_stats);
1503         return 0;
1504 }
1505
1506
1507 static int
1508 proc_do_defense_mode(ctl_table *table, int write, struct file * filp,
1509                      void __user *buffer, size_t *lenp, loff_t *ppos)
1510 {
1511         int *valp = table->data;
1512         int val = *valp;
1513         int rc;
1514
1515         rc = proc_dointvec(table, write, filp, buffer, lenp, ppos);
1516         if (write && (*valp != val)) {
1517                 if ((*valp < 0) || (*valp > 3)) {
1518                         /* Restore the correct value */
1519                         *valp = val;
1520                 } else {
1521                         update_defense_level();
1522                 }
1523         }
1524         return rc;
1525 }
1526
1527
1528 static int
1529 proc_do_sync_threshold(ctl_table *table, int write, struct file *filp,
1530                        void __user *buffer, size_t *lenp, loff_t *ppos)
1531 {
1532         int *valp = table->data;
1533         int val[2];
1534         int rc;
1535
1536         /* backup the value first */
1537         memcpy(val, valp, sizeof(val));
1538
1539         rc = proc_dointvec(table, write, filp, buffer, lenp, ppos);
1540         if (write && (valp[0] < 0 || valp[1] < 0 || valp[0] >= valp[1])) {
1541                 /* Restore the correct value */
1542                 memcpy(valp, val, sizeof(val));
1543         }
1544         return rc;
1545 }
1546
1547
1548 /*
1549  *      IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1550  */
1551
1552 static struct ctl_table vs_vars[] = {
1553         {
1554                 .procname       = "amemthresh",
1555                 .data           = &sysctl_ip_vs_amemthresh,
1556                 .maxlen         = sizeof(int),
1557                 .mode           = 0644,
1558                 .proc_handler   = &proc_dointvec,
1559         },
1560 #ifdef CONFIG_IP_VS_DEBUG
1561         {
1562                 .procname       = "debug_level",
1563                 .data           = &sysctl_ip_vs_debug_level,
1564                 .maxlen         = sizeof(int),
1565                 .mode           = 0644,
1566                 .proc_handler   = &proc_dointvec,
1567         },
1568 #endif
1569         {
1570                 .procname       = "am_droprate",
1571                 .data           = &sysctl_ip_vs_am_droprate,
1572                 .maxlen         = sizeof(int),
1573                 .mode           = 0644,
1574                 .proc_handler   = &proc_dointvec,
1575         },
1576         {
1577                 .procname       = "drop_entry",
1578                 .data           = &sysctl_ip_vs_drop_entry,
1579                 .maxlen         = sizeof(int),
1580                 .mode           = 0644,
1581                 .proc_handler   = &proc_do_defense_mode,
1582         },
1583         {
1584                 .procname       = "drop_packet",
1585                 .data           = &sysctl_ip_vs_drop_packet,
1586                 .maxlen         = sizeof(int),
1587                 .mode           = 0644,
1588                 .proc_handler   = &proc_do_defense_mode,
1589         },
1590         {
1591                 .procname       = "secure_tcp",
1592                 .data           = &sysctl_ip_vs_secure_tcp,
1593                 .maxlen         = sizeof(int),
1594                 .mode           = 0644,
1595                 .proc_handler   = &proc_do_defense_mode,
1596         },
1597 #if 0
1598         {
1599                 .procname       = "timeout_established",
1600                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1601                 .maxlen         = sizeof(int),
1602                 .mode           = 0644,
1603                 .proc_handler   = &proc_dointvec_jiffies,
1604         },
1605         {
1606                 .procname       = "timeout_synsent",
1607                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1608                 .maxlen         = sizeof(int),
1609                 .mode           = 0644,
1610                 .proc_handler   = &proc_dointvec_jiffies,
1611         },
1612         {
1613                 .procname       = "timeout_synrecv",
1614                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1615                 .maxlen         = sizeof(int),
1616                 .mode           = 0644,
1617                 .proc_handler   = &proc_dointvec_jiffies,
1618         },
1619         {
1620                 .procname       = "timeout_finwait",
1621                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1622                 .maxlen         = sizeof(int),
1623                 .mode           = 0644,
1624                 .proc_handler   = &proc_dointvec_jiffies,
1625         },
1626         {
1627                 .procname       = "timeout_timewait",
1628                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1629                 .maxlen         = sizeof(int),
1630                 .mode           = 0644,
1631                 .proc_handler   = &proc_dointvec_jiffies,
1632         },
1633         {
1634                 .procname       = "timeout_close",
1635                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1636                 .maxlen         = sizeof(int),
1637                 .mode           = 0644,
1638                 .proc_handler   = &proc_dointvec_jiffies,
1639         },
1640         {
1641                 .procname       = "timeout_closewait",
1642                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1643                 .maxlen         = sizeof(int),
1644                 .mode           = 0644,
1645                 .proc_handler   = &proc_dointvec_jiffies,
1646         },
1647         {
1648                 .procname       = "timeout_lastack",
1649                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1650                 .maxlen         = sizeof(int),
1651                 .mode           = 0644,
1652                 .proc_handler   = &proc_dointvec_jiffies,
1653         },
1654         {
1655                 .procname       = "timeout_listen",
1656                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1657                 .maxlen         = sizeof(int),
1658                 .mode           = 0644,
1659                 .proc_handler   = &proc_dointvec_jiffies,
1660         },
1661         {
1662                 .procname       = "timeout_synack",
1663                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1664                 .maxlen         = sizeof(int),
1665                 .mode           = 0644,
1666                 .proc_handler   = &proc_dointvec_jiffies,
1667         },
1668         {
1669                 .procname       = "timeout_udp",
1670                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1671                 .maxlen         = sizeof(int),
1672                 .mode           = 0644,
1673                 .proc_handler   = &proc_dointvec_jiffies,
1674         },
1675         {
1676                 .procname       = "timeout_icmp",
1677                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1678                 .maxlen         = sizeof(int),
1679                 .mode           = 0644,
1680                 .proc_handler   = &proc_dointvec_jiffies,
1681         },
1682 #endif
1683         {
1684                 .procname       = "cache_bypass",
1685                 .data           = &sysctl_ip_vs_cache_bypass,
1686                 .maxlen         = sizeof(int),
1687                 .mode           = 0644,
1688                 .proc_handler   = &proc_dointvec,
1689         },
1690         {
1691                 .procname       = "expire_nodest_conn",
1692                 .data           = &sysctl_ip_vs_expire_nodest_conn,
1693                 .maxlen         = sizeof(int),
1694                 .mode           = 0644,
1695                 .proc_handler   = &proc_dointvec,
1696         },
1697         {
1698                 .procname       = "expire_quiescent_template",
1699                 .data           = &sysctl_ip_vs_expire_quiescent_template,
1700                 .maxlen         = sizeof(int),
1701                 .mode           = 0644,
1702                 .proc_handler   = &proc_dointvec,
1703         },
1704         {
1705                 .procname       = "sync_threshold",
1706                 .data           = &sysctl_ip_vs_sync_threshold,
1707                 .maxlen         = sizeof(sysctl_ip_vs_sync_threshold),
1708                 .mode           = 0644,
1709                 .proc_handler   = &proc_do_sync_threshold,
1710         },
1711         {
1712                 .procname       = "nat_icmp_send",
1713                 .data           = &sysctl_ip_vs_nat_icmp_send,
1714                 .maxlen         = sizeof(int),
1715                 .mode           = 0644,
1716                 .proc_handler   = &proc_dointvec,
1717         },
1718         { .ctl_name = 0 }
1719 };
1720
1721 const struct ctl_path net_vs_ctl_path[] = {
1722         { .procname = "net", .ctl_name = CTL_NET, },
1723         { .procname = "ipv4", .ctl_name = NET_IPV4, },
1724         { .procname = "vs", },
1725         { }
1726 };
1727 EXPORT_SYMBOL_GPL(net_vs_ctl_path);
1728
1729 static struct ctl_table_header * sysctl_header;
1730
1731 #ifdef CONFIG_PROC_FS
1732
1733 struct ip_vs_iter {
1734         struct list_head *table;
1735         int bucket;
1736 };
1737
1738 /*
1739  *      Write the contents of the VS rule table to a PROCfs file.
1740  *      (It is kept just for backward compatibility)
1741  */
1742 static inline const char *ip_vs_fwd_name(unsigned flags)
1743 {
1744         switch (flags & IP_VS_CONN_F_FWD_MASK) {
1745         case IP_VS_CONN_F_LOCALNODE:
1746                 return "Local";
1747         case IP_VS_CONN_F_TUNNEL:
1748                 return "Tunnel";
1749         case IP_VS_CONN_F_DROUTE:
1750                 return "Route";
1751         default:
1752                 return "Masq";
1753         }
1754 }
1755
1756
1757 /* Get the Nth entry in the two lists */
1758 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1759 {
1760         struct ip_vs_iter *iter = seq->private;
1761         int idx;
1762         struct ip_vs_service *svc;
1763
1764         /* look in hash by protocol */
1765         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1766                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1767                         if (pos-- == 0){
1768                                 iter->table = ip_vs_svc_table;
1769                                 iter->bucket = idx;
1770                                 return svc;
1771                         }
1772                 }
1773         }
1774
1775         /* keep looking in fwmark */
1776         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1777                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1778                         if (pos-- == 0) {
1779                                 iter->table = ip_vs_svc_fwm_table;
1780                                 iter->bucket = idx;
1781                                 return svc;
1782                         }
1783                 }
1784         }
1785
1786         return NULL;
1787 }
1788
1789 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1790 __acquires(__ip_vs_svc_lock)
1791 {
1792
1793         read_lock_bh(&__ip_vs_svc_lock);
1794         return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1795 }
1796
1797
1798 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1799 {
1800         struct list_head *e;
1801         struct ip_vs_iter *iter;
1802         struct ip_vs_service *svc;
1803
1804         ++*pos;
1805         if (v == SEQ_START_TOKEN)
1806                 return ip_vs_info_array(seq,0);
1807
1808         svc = v;
1809         iter = seq->private;
1810
1811         if (iter->table == ip_vs_svc_table) {
1812                 /* next service in table hashed by protocol */
1813                 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
1814                         return list_entry(e, struct ip_vs_service, s_list);
1815
1816
1817                 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1818                         list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
1819                                             s_list) {
1820                                 return svc;
1821                         }
1822                 }
1823
1824                 iter->table = ip_vs_svc_fwm_table;
1825                 iter->bucket = -1;
1826                 goto scan_fwmark;
1827         }
1828
1829         /* next service in hashed by fwmark */
1830         if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
1831                 return list_entry(e, struct ip_vs_service, f_list);
1832
1833  scan_fwmark:
1834         while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1835                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
1836                                     f_list)
1837                         return svc;
1838         }
1839
1840         return NULL;
1841 }
1842
1843 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1844 __releases(__ip_vs_svc_lock)
1845 {
1846         read_unlock_bh(&__ip_vs_svc_lock);
1847 }
1848
1849
1850 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1851 {
1852         if (v == SEQ_START_TOKEN) {
1853                 seq_printf(seq,
1854                         "IP Virtual Server version %d.%d.%d (size=%d)\n",
1855                         NVERSION(IP_VS_VERSION_CODE), IP_VS_CONN_TAB_SIZE);
1856                 seq_puts(seq,
1857                          "Prot LocalAddress:Port Scheduler Flags\n");
1858                 seq_puts(seq,
1859                          "  -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1860         } else {
1861                 const struct ip_vs_service *svc = v;
1862                 const struct ip_vs_iter *iter = seq->private;
1863                 const struct ip_vs_dest *dest;
1864
1865                 if (iter->table == ip_vs_svc_table) {
1866 #ifdef CONFIG_IP_VS_IPV6
1867                         if (svc->af == AF_INET6)
1868                                 seq_printf(seq, "%s  [" NIP6_FMT "]:%04X %s ",
1869                                            ip_vs_proto_name(svc->protocol),
1870                                            NIP6(svc->addr.in6),
1871                                            ntohs(svc->port),
1872                                            svc->scheduler->name);
1873                         else
1874 #endif
1875                                 seq_printf(seq, "%s  %08X:%04X %s ",
1876                                            ip_vs_proto_name(svc->protocol),
1877                                            ntohl(svc->addr.ip),
1878                                            ntohs(svc->port),
1879                                            svc->scheduler->name);
1880                 } else {
1881                         seq_printf(seq, "FWM  %08X %s ",
1882                                    svc->fwmark, svc->scheduler->name);
1883                 }
1884
1885                 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
1886                         seq_printf(seq, "persistent %d %08X\n",
1887                                 svc->timeout,
1888                                 ntohl(svc->netmask));
1889                 else
1890                         seq_putc(seq, '\n');
1891
1892                 list_for_each_entry(dest, &svc->destinations, n_list) {
1893 #ifdef CONFIG_IP_VS_IPV6
1894                         if (dest->af == AF_INET6)
1895                                 seq_printf(seq,
1896                                            "  -> [" NIP6_FMT "]:%04X"
1897                                            "      %-7s %-6d %-10d %-10d\n",
1898                                            NIP6(dest->addr.in6),
1899                                            ntohs(dest->port),
1900                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1901                                            atomic_read(&dest->weight),
1902                                            atomic_read(&dest->activeconns),
1903                                            atomic_read(&dest->inactconns));
1904                         else
1905 #endif
1906                                 seq_printf(seq,
1907                                            "  -> %08X:%04X      "
1908                                            "%-7s %-6d %-10d %-10d\n",
1909                                            ntohl(dest->addr.ip),
1910                                            ntohs(dest->port),
1911                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1912                                            atomic_read(&dest->weight),
1913                                            atomic_read(&dest->activeconns),
1914                                            atomic_read(&dest->inactconns));
1915
1916                 }
1917         }
1918         return 0;
1919 }
1920
1921 static const struct seq_operations ip_vs_info_seq_ops = {
1922         .start = ip_vs_info_seq_start,
1923         .next  = ip_vs_info_seq_next,
1924         .stop  = ip_vs_info_seq_stop,
1925         .show  = ip_vs_info_seq_show,
1926 };
1927
1928 static int ip_vs_info_open(struct inode *inode, struct file *file)
1929 {
1930         return seq_open_private(file, &ip_vs_info_seq_ops,
1931                         sizeof(struct ip_vs_iter));
1932 }
1933
1934 static const struct file_operations ip_vs_info_fops = {
1935         .owner   = THIS_MODULE,
1936         .open    = ip_vs_info_open,
1937         .read    = seq_read,
1938         .llseek  = seq_lseek,
1939         .release = seq_release_private,
1940 };
1941
1942 #endif
1943
1944 struct ip_vs_stats ip_vs_stats = {
1945         .lock = __SPIN_LOCK_UNLOCKED(ip_vs_stats.lock),
1946 };
1947
1948 #ifdef CONFIG_PROC_FS
1949 static int ip_vs_stats_show(struct seq_file *seq, void *v)
1950 {
1951
1952 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
1953         seq_puts(seq,
1954                  "   Total Incoming Outgoing         Incoming         Outgoing\n");
1955         seq_printf(seq,
1956                    "   Conns  Packets  Packets            Bytes            Bytes\n");
1957
1958         spin_lock_bh(&ip_vs_stats.lock);
1959         seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", ip_vs_stats.ustats.conns,
1960                    ip_vs_stats.ustats.inpkts, ip_vs_stats.ustats.outpkts,
1961                    (unsigned long long) ip_vs_stats.ustats.inbytes,
1962                    (unsigned long long) ip_vs_stats.ustats.outbytes);
1963
1964 /*                 01234567 01234567 01234567 0123456701234567 0123456701234567 */
1965         seq_puts(seq,
1966                    " Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
1967         seq_printf(seq,"%8X %8X %8X %16X %16X\n",
1968                         ip_vs_stats.ustats.cps,
1969                         ip_vs_stats.ustats.inpps,
1970                         ip_vs_stats.ustats.outpps,
1971                         ip_vs_stats.ustats.inbps,
1972                         ip_vs_stats.ustats.outbps);
1973         spin_unlock_bh(&ip_vs_stats.lock);
1974
1975         return 0;
1976 }
1977
1978 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
1979 {
1980         return single_open(file, ip_vs_stats_show, NULL);
1981 }
1982
1983 static const struct file_operations ip_vs_stats_fops = {
1984         .owner = THIS_MODULE,
1985         .open = ip_vs_stats_seq_open,
1986         .read = seq_read,
1987         .llseek = seq_lseek,
1988         .release = single_release,
1989 };
1990
1991 #endif
1992
1993 /*
1994  *      Set timeout values for tcp tcpfin udp in the timeout_table.
1995  */
1996 static int ip_vs_set_timeout(struct ip_vs_timeout_user *u)
1997 {
1998         IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
1999                   u->tcp_timeout,
2000                   u->tcp_fin_timeout,
2001                   u->udp_timeout);
2002
2003 #ifdef CONFIG_IP_VS_PROTO_TCP
2004         if (u->tcp_timeout) {
2005                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED]
2006                         = u->tcp_timeout * HZ;
2007         }
2008
2009         if (u->tcp_fin_timeout) {
2010                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT]
2011                         = u->tcp_fin_timeout * HZ;
2012         }
2013 #endif
2014
2015 #ifdef CONFIG_IP_VS_PROTO_UDP
2016         if (u->udp_timeout) {
2017                 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL]
2018                         = u->udp_timeout * HZ;
2019         }
2020 #endif
2021         return 0;
2022 }
2023
2024
2025 #define SET_CMDID(cmd)          (cmd - IP_VS_BASE_CTL)
2026 #define SERVICE_ARG_LEN         (sizeof(struct ip_vs_service_user))
2027 #define SVCDEST_ARG_LEN         (sizeof(struct ip_vs_service_user) +    \
2028                                  sizeof(struct ip_vs_dest_user))
2029 #define TIMEOUT_ARG_LEN         (sizeof(struct ip_vs_timeout_user))
2030 #define DAEMON_ARG_LEN          (sizeof(struct ip_vs_daemon_user))
2031 #define MAX_ARG_LEN             SVCDEST_ARG_LEN
2032
2033 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2034         [SET_CMDID(IP_VS_SO_SET_ADD)]           = SERVICE_ARG_LEN,
2035         [SET_CMDID(IP_VS_SO_SET_EDIT)]          = SERVICE_ARG_LEN,
2036         [SET_CMDID(IP_VS_SO_SET_DEL)]           = SERVICE_ARG_LEN,
2037         [SET_CMDID(IP_VS_SO_SET_FLUSH)]         = 0,
2038         [SET_CMDID(IP_VS_SO_SET_ADDDEST)]       = SVCDEST_ARG_LEN,
2039         [SET_CMDID(IP_VS_SO_SET_DELDEST)]       = SVCDEST_ARG_LEN,
2040         [SET_CMDID(IP_VS_SO_SET_EDITDEST)]      = SVCDEST_ARG_LEN,
2041         [SET_CMDID(IP_VS_SO_SET_TIMEOUT)]       = TIMEOUT_ARG_LEN,
2042         [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)]   = DAEMON_ARG_LEN,
2043         [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)]    = DAEMON_ARG_LEN,
2044         [SET_CMDID(IP_VS_SO_SET_ZERO)]          = SERVICE_ARG_LEN,
2045 };
2046
2047 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2048                                   struct ip_vs_service_user *usvc_compat)
2049 {
2050         usvc->af                = AF_INET;
2051         usvc->protocol          = usvc_compat->protocol;
2052         usvc->addr.ip           = usvc_compat->addr;
2053         usvc->port              = usvc_compat->port;
2054         usvc->fwmark            = usvc_compat->fwmark;
2055
2056         /* Deep copy of sched_name is not needed here */
2057         usvc->sched_name        = usvc_compat->sched_name;
2058
2059         usvc->flags             = usvc_compat->flags;
2060         usvc->timeout           = usvc_compat->timeout;
2061         usvc->netmask           = usvc_compat->netmask;
2062 }
2063
2064 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2065                                    struct ip_vs_dest_user *udest_compat)
2066 {
2067         udest->addr.ip          = udest_compat->addr;
2068         udest->port             = udest_compat->port;
2069         udest->conn_flags       = udest_compat->conn_flags;
2070         udest->weight           = udest_compat->weight;
2071         udest->u_threshold      = udest_compat->u_threshold;
2072         udest->l_threshold      = udest_compat->l_threshold;
2073 }
2074
2075 static int
2076 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2077 {
2078         int ret;
2079         unsigned char arg[MAX_ARG_LEN];
2080         struct ip_vs_service_user *usvc_compat;
2081         struct ip_vs_service_user_kern usvc;
2082         struct ip_vs_service *svc;
2083         struct ip_vs_dest_user *udest_compat;
2084         struct ip_vs_dest_user_kern udest;
2085
2086         if (!capable(CAP_NET_ADMIN))
2087                 return -EPERM;
2088
2089         if (len != set_arglen[SET_CMDID(cmd)]) {
2090                 IP_VS_ERR("set_ctl: len %u != %u\n",
2091                           len, set_arglen[SET_CMDID(cmd)]);
2092                 return -EINVAL;
2093         }
2094
2095         if (copy_from_user(arg, user, len) != 0)
2096                 return -EFAULT;
2097
2098         /* increase the module use count */
2099         ip_vs_use_count_inc();
2100
2101         if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2102                 ret = -ERESTARTSYS;
2103                 goto out_dec;
2104         }
2105
2106         if (cmd == IP_VS_SO_SET_FLUSH) {
2107                 /* Flush the virtual service */
2108                 ret = ip_vs_flush();
2109                 goto out_unlock;
2110         } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2111                 /* Set timeout values for (tcp tcpfin udp) */
2112                 ret = ip_vs_set_timeout((struct ip_vs_timeout_user *)arg);
2113                 goto out_unlock;
2114         } else if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2115                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2116                 ret = start_sync_thread(dm->state, dm->mcast_ifn, dm->syncid);
2117                 goto out_unlock;
2118         } else if (cmd == IP_VS_SO_SET_STOPDAEMON) {
2119                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2120                 ret = stop_sync_thread(dm->state);
2121                 goto out_unlock;
2122         }
2123
2124         usvc_compat = (struct ip_vs_service_user *)arg;
2125         udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2126
2127         /* We only use the new structs internally, so copy userspace compat
2128          * structs to extended internal versions */
2129         ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2130         ip_vs_copy_udest_compat(&udest, udest_compat);
2131
2132         if (cmd == IP_VS_SO_SET_ZERO) {
2133                 /* if no service address is set, zero counters in all */
2134                 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2135                         ret = ip_vs_zero_all();
2136                         goto out_unlock;
2137                 }
2138         }
2139
2140         /* Check for valid protocol: TCP or UDP, even for fwmark!=0 */
2141         if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP) {
2142                 IP_VS_ERR("set_ctl: invalid protocol: %d %d.%d.%d.%d:%d %s\n",
2143                           usvc.protocol, NIPQUAD(usvc.addr.ip),
2144                           ntohs(usvc.port), usvc.sched_name);
2145                 ret = -EFAULT;
2146                 goto out_unlock;
2147         }
2148
2149         /* Lookup the exact service by <protocol, addr, port> or fwmark */
2150         if (usvc.fwmark == 0)
2151                 svc = __ip_vs_service_get(usvc.af, usvc.protocol,
2152                                           &usvc.addr, usvc.port);
2153         else
2154                 svc = __ip_vs_svc_fwm_get(usvc.af, usvc.fwmark);
2155
2156         if (cmd != IP_VS_SO_SET_ADD
2157             && (svc == NULL || svc->protocol != usvc.protocol)) {
2158                 ret = -ESRCH;
2159                 goto out_unlock;
2160         }
2161
2162         switch (cmd) {
2163         case IP_VS_SO_SET_ADD:
2164                 if (svc != NULL)
2165                         ret = -EEXIST;
2166                 else
2167                         ret = ip_vs_add_service(&usvc, &svc);
2168                 break;
2169         case IP_VS_SO_SET_EDIT:
2170                 ret = ip_vs_edit_service(svc, &usvc);
2171                 break;
2172         case IP_VS_SO_SET_DEL:
2173                 ret = ip_vs_del_service(svc);
2174                 if (!ret)
2175                         goto out_unlock;
2176                 break;
2177         case IP_VS_SO_SET_ZERO:
2178                 ret = ip_vs_zero_service(svc);
2179                 break;
2180         case IP_VS_SO_SET_ADDDEST:
2181                 ret = ip_vs_add_dest(svc, &udest);
2182                 break;
2183         case IP_VS_SO_SET_EDITDEST:
2184                 ret = ip_vs_edit_dest(svc, &udest);
2185                 break;
2186         case IP_VS_SO_SET_DELDEST:
2187                 ret = ip_vs_del_dest(svc, &udest);
2188                 break;
2189         default:
2190                 ret = -EINVAL;
2191         }
2192
2193         if (svc)
2194                 ip_vs_service_put(svc);
2195
2196   out_unlock:
2197         mutex_unlock(&__ip_vs_mutex);
2198   out_dec:
2199         /* decrease the module use count */
2200         ip_vs_use_count_dec();
2201
2202         return ret;
2203 }
2204
2205
2206 static void
2207 ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
2208 {
2209         spin_lock_bh(&src->lock);
2210         memcpy(dst, &src->ustats, sizeof(*dst));
2211         spin_unlock_bh(&src->lock);
2212 }
2213
2214 static void
2215 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2216 {
2217         dst->protocol = src->protocol;
2218         dst->addr = src->addr.ip;
2219         dst->port = src->port;
2220         dst->fwmark = src->fwmark;
2221         strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
2222         dst->flags = src->flags;
2223         dst->timeout = src->timeout / HZ;
2224         dst->netmask = src->netmask;
2225         dst->num_dests = src->num_dests;
2226         ip_vs_copy_stats(&dst->stats, &src->stats);
2227 }
2228
2229 static inline int
2230 __ip_vs_get_service_entries(const struct ip_vs_get_services *get,
2231                             struct ip_vs_get_services __user *uptr)
2232 {
2233         int idx, count=0;
2234         struct ip_vs_service *svc;
2235         struct ip_vs_service_entry entry;
2236         int ret = 0;
2237
2238         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2239                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2240                         /* Only expose IPv4 entries to old interface */
2241                         if (svc->af != AF_INET)
2242                                 continue;
2243
2244                         if (count >= get->num_services)
2245                                 goto out;
2246                         memset(&entry, 0, sizeof(entry));
2247                         ip_vs_copy_service(&entry, svc);
2248                         if (copy_to_user(&uptr->entrytable[count],
2249                                          &entry, sizeof(entry))) {
2250                                 ret = -EFAULT;
2251                                 goto out;
2252                         }
2253                         count++;
2254                 }
2255         }
2256
2257         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2258                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2259                         /* Only expose IPv4 entries to old interface */
2260                         if (svc->af != AF_INET)
2261                                 continue;
2262
2263                         if (count >= get->num_services)
2264                                 goto out;
2265                         memset(&entry, 0, sizeof(entry));
2266                         ip_vs_copy_service(&entry, svc);
2267                         if (copy_to_user(&uptr->entrytable[count],
2268                                          &entry, sizeof(entry))) {
2269                                 ret = -EFAULT;
2270                                 goto out;
2271                         }
2272                         count++;
2273                 }
2274         }
2275   out:
2276         return ret;
2277 }
2278
2279 static inline int
2280 __ip_vs_get_dest_entries(const struct ip_vs_get_dests *get,
2281                          struct ip_vs_get_dests __user *uptr)
2282 {
2283         struct ip_vs_service *svc;
2284         union nf_inet_addr addr = { .ip = get->addr };
2285         int ret = 0;
2286
2287         if (get->fwmark)
2288                 svc = __ip_vs_svc_fwm_get(AF_INET, get->fwmark);
2289         else
2290                 svc = __ip_vs_service_get(AF_INET, get->protocol, &addr,
2291                                           get->port);
2292
2293         if (svc) {
2294                 int count = 0;
2295                 struct ip_vs_dest *dest;
2296                 struct ip_vs_dest_entry entry;
2297
2298                 list_for_each_entry(dest, &svc->destinations, n_list) {
2299                         if (count >= get->num_dests)
2300                                 break;
2301
2302                         entry.addr = dest->addr.ip;
2303                         entry.port = dest->port;
2304                         entry.conn_flags = atomic_read(&dest->conn_flags);
2305                         entry.weight = atomic_read(&dest->weight);
2306                         entry.u_threshold = dest->u_threshold;
2307                         entry.l_threshold = dest->l_threshold;
2308                         entry.activeconns = atomic_read(&dest->activeconns);
2309                         entry.inactconns = atomic_read(&dest->inactconns);
2310                         entry.persistconns = atomic_read(&dest->persistconns);
2311                         ip_vs_copy_stats(&entry.stats, &dest->stats);
2312                         if (copy_to_user(&uptr->entrytable[count],
2313                                          &entry, sizeof(entry))) {
2314                                 ret = -EFAULT;
2315                                 break;
2316                         }
2317                         count++;
2318                 }
2319                 ip_vs_service_put(svc);
2320         } else
2321                 ret = -ESRCH;
2322         return ret;
2323 }
2324
2325 static inline void
2326 __ip_vs_get_timeouts(struct ip_vs_timeout_user *u)
2327 {
2328 #ifdef CONFIG_IP_VS_PROTO_TCP
2329         u->tcp_timeout =
2330                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2331         u->tcp_fin_timeout =
2332                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2333 #endif
2334 #ifdef CONFIG_IP_VS_PROTO_UDP
2335         u->udp_timeout =
2336                 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2337 #endif
2338 }
2339
2340
2341 #define GET_CMDID(cmd)          (cmd - IP_VS_BASE_CTL)
2342 #define GET_INFO_ARG_LEN        (sizeof(struct ip_vs_getinfo))
2343 #define GET_SERVICES_ARG_LEN    (sizeof(struct ip_vs_get_services))
2344 #define GET_SERVICE_ARG_LEN     (sizeof(struct ip_vs_service_entry))
2345 #define GET_DESTS_ARG_LEN       (sizeof(struct ip_vs_get_dests))
2346 #define GET_TIMEOUT_ARG_LEN     (sizeof(struct ip_vs_timeout_user))
2347 #define GET_DAEMON_ARG_LEN      (sizeof(struct ip_vs_daemon_user) * 2)
2348
2349 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2350         [GET_CMDID(IP_VS_SO_GET_VERSION)]       = 64,
2351         [GET_CMDID(IP_VS_SO_GET_INFO)]          = GET_INFO_ARG_LEN,
2352         [GET_CMDID(IP_VS_SO_GET_SERVICES)]      = GET_SERVICES_ARG_LEN,
2353         [GET_CMDID(IP_VS_SO_GET_SERVICE)]       = GET_SERVICE_ARG_LEN,
2354         [GET_CMDID(IP_VS_SO_GET_DESTS)]         = GET_DESTS_ARG_LEN,
2355         [GET_CMDID(IP_VS_SO_GET_TIMEOUT)]       = GET_TIMEOUT_ARG_LEN,
2356         [GET_CMDID(IP_VS_SO_GET_DAEMON)]        = GET_DAEMON_ARG_LEN,
2357 };
2358
2359 static int
2360 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2361 {
2362         unsigned char arg[128];
2363         int ret = 0;
2364
2365         if (!capable(CAP_NET_ADMIN))
2366                 return -EPERM;
2367
2368         if (*len < get_arglen[GET_CMDID(cmd)]) {
2369                 IP_VS_ERR("get_ctl: len %u < %u\n",
2370                           *len, get_arglen[GET_CMDID(cmd)]);
2371                 return -EINVAL;
2372         }
2373
2374         if (copy_from_user(arg, user, get_arglen[GET_CMDID(cmd)]) != 0)
2375                 return -EFAULT;
2376
2377         if (mutex_lock_interruptible(&__ip_vs_mutex))
2378                 return -ERESTARTSYS;
2379
2380         switch (cmd) {
2381         case IP_VS_SO_GET_VERSION:
2382         {
2383                 char buf[64];
2384
2385                 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2386                         NVERSION(IP_VS_VERSION_CODE), IP_VS_CONN_TAB_SIZE);
2387                 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2388                         ret = -EFAULT;
2389                         goto out;
2390                 }
2391                 *len = strlen(buf)+1;
2392         }
2393         break;
2394
2395         case IP_VS_SO_GET_INFO:
2396         {
2397                 struct ip_vs_getinfo info;
2398                 info.version = IP_VS_VERSION_CODE;
2399                 info.size = IP_VS_CONN_TAB_SIZE;
2400                 info.num_services = ip_vs_num_services;
2401                 if (copy_to_user(user, &info, sizeof(info)) != 0)
2402                         ret = -EFAULT;
2403         }
2404         break;
2405
2406         case IP_VS_SO_GET_SERVICES:
2407         {
2408                 struct ip_vs_get_services *get;
2409                 int size;
2410
2411                 get = (struct ip_vs_get_services *)arg;
2412                 size = sizeof(*get) +
2413                         sizeof(struct ip_vs_service_entry) * get->num_services;
2414                 if (*len != size) {
2415                         IP_VS_ERR("length: %u != %u\n", *len, size);
2416                         ret = -EINVAL;
2417                         goto out;
2418                 }
2419                 ret = __ip_vs_get_service_entries(get, user);
2420         }
2421         break;
2422
2423         case IP_VS_SO_GET_SERVICE:
2424         {
2425                 struct ip_vs_service_entry *entry;
2426                 struct ip_vs_service *svc;
2427                 union nf_inet_addr addr;
2428
2429                 entry = (struct ip_vs_service_entry *)arg;
2430                 addr.ip = entry->addr;
2431                 if (entry->fwmark)
2432                         svc = __ip_vs_svc_fwm_get(AF_INET, entry->fwmark);
2433                 else
2434                         svc = __ip_vs_service_get(AF_INET, entry->protocol,
2435                                                   &addr, entry->port);
2436                 if (svc) {
2437                         ip_vs_copy_service(entry, svc);
2438                         if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2439                                 ret = -EFAULT;
2440                         ip_vs_service_put(svc);
2441                 } else
2442                         ret = -ESRCH;
2443         }
2444         break;
2445
2446         case IP_VS_SO_GET_DESTS:
2447         {
2448                 struct ip_vs_get_dests *get;
2449                 int size;
2450
2451                 get = (struct ip_vs_get_dests *)arg;
2452                 size = sizeof(*get) +
2453                         sizeof(struct ip_vs_dest_entry) * get->num_dests;
2454                 if (*len != size) {
2455                         IP_VS_ERR("length: %u != %u\n", *len, size);
2456                         ret = -EINVAL;
2457                         goto out;
2458                 }
2459                 ret = __ip_vs_get_dest_entries(get, user);
2460         }
2461         break;
2462
2463         case IP_VS_SO_GET_TIMEOUT:
2464         {
2465                 struct ip_vs_timeout_user t;
2466
2467                 __ip_vs_get_timeouts(&t);
2468                 if (copy_to_user(user, &t, sizeof(t)) != 0)
2469                         ret = -EFAULT;
2470         }
2471         break;
2472
2473         case IP_VS_SO_GET_DAEMON:
2474         {
2475                 struct ip_vs_daemon_user d[2];
2476
2477                 memset(&d, 0, sizeof(d));
2478                 if (ip_vs_sync_state & IP_VS_STATE_MASTER) {
2479                         d[0].state = IP_VS_STATE_MASTER;
2480                         strlcpy(d[0].mcast_ifn, ip_vs_master_mcast_ifn, sizeof(d[0].mcast_ifn));
2481                         d[0].syncid = ip_vs_master_syncid;
2482                 }
2483                 if (ip_vs_sync_state & IP_VS_STATE_BACKUP) {
2484                         d[1].state = IP_VS_STATE_BACKUP;
2485                         strlcpy(d[1].mcast_ifn, ip_vs_backup_mcast_ifn, sizeof(d[1].mcast_ifn));
2486                         d[1].syncid = ip_vs_backup_syncid;
2487                 }
2488                 if (copy_to_user(user, &d, sizeof(d)) != 0)
2489                         ret = -EFAULT;
2490         }
2491         break;
2492
2493         default:
2494                 ret = -EINVAL;
2495         }
2496
2497   out:
2498         mutex_unlock(&__ip_vs_mutex);
2499         return ret;
2500 }
2501
2502
2503 static struct nf_sockopt_ops ip_vs_sockopts = {
2504         .pf             = PF_INET,
2505         .set_optmin     = IP_VS_BASE_CTL,
2506         .set_optmax     = IP_VS_SO_SET_MAX+1,
2507         .set            = do_ip_vs_set_ctl,
2508         .get_optmin     = IP_VS_BASE_CTL,
2509         .get_optmax     = IP_VS_SO_GET_MAX+1,
2510         .get            = do_ip_vs_get_ctl,
2511         .owner          = THIS_MODULE,
2512 };
2513
2514 /*
2515  * Generic Netlink interface
2516  */
2517
2518 /* IPVS genetlink family */
2519 static struct genl_family ip_vs_genl_family = {
2520         .id             = GENL_ID_GENERATE,
2521         .hdrsize        = 0,
2522         .name           = IPVS_GENL_NAME,
2523         .version        = IPVS_GENL_VERSION,
2524         .maxattr        = IPVS_CMD_MAX,
2525 };
2526
2527 /* Policy used for first-level command attributes */
2528 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2529         [IPVS_CMD_ATTR_SERVICE]         = { .type = NLA_NESTED },
2530         [IPVS_CMD_ATTR_DEST]            = { .type = NLA_NESTED },
2531         [IPVS_CMD_ATTR_DAEMON]          = { .type = NLA_NESTED },
2532         [IPVS_CMD_ATTR_TIMEOUT_TCP]     = { .type = NLA_U32 },
2533         [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2534         [IPVS_CMD_ATTR_TIMEOUT_UDP]     = { .type = NLA_U32 },
2535 };
2536
2537 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2538 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2539         [IPVS_DAEMON_ATTR_STATE]        = { .type = NLA_U32 },
2540         [IPVS_DAEMON_ATTR_MCAST_IFN]    = { .type = NLA_NUL_STRING,
2541                                             .len = IP_VS_IFNAME_MAXLEN },
2542         [IPVS_DAEMON_ATTR_SYNC_ID]      = { .type = NLA_U32 },
2543 };
2544
2545 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2546 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2547         [IPVS_SVC_ATTR_AF]              = { .type = NLA_U16 },
2548         [IPVS_SVC_ATTR_PROTOCOL]        = { .type = NLA_U16 },
2549         [IPVS_SVC_ATTR_ADDR]            = { .type = NLA_BINARY,
2550                                             .len = sizeof(union nf_inet_addr) },
2551         [IPVS_SVC_ATTR_PORT]            = { .type = NLA_U16 },
2552         [IPVS_SVC_ATTR_FWMARK]          = { .type = NLA_U32 },
2553         [IPVS_SVC_ATTR_SCHED_NAME]      = { .type = NLA_NUL_STRING,
2554                                             .len = IP_VS_SCHEDNAME_MAXLEN },
2555         [IPVS_SVC_ATTR_FLAGS]           = { .type = NLA_BINARY,
2556                                             .len = sizeof(struct ip_vs_flags) },
2557         [IPVS_SVC_ATTR_TIMEOUT]         = { .type = NLA_U32 },
2558         [IPVS_SVC_ATTR_NETMASK]         = { .type = NLA_U32 },
2559         [IPVS_SVC_ATTR_STATS]           = { .type = NLA_NESTED },
2560 };
2561
2562 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2563 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2564         [IPVS_DEST_ATTR_ADDR]           = { .type = NLA_BINARY,
2565                                             .len = sizeof(union nf_inet_addr) },
2566         [IPVS_DEST_ATTR_PORT]           = { .type = NLA_U16 },
2567         [IPVS_DEST_ATTR_FWD_METHOD]     = { .type = NLA_U32 },
2568         [IPVS_DEST_ATTR_WEIGHT]         = { .type = NLA_U32 },
2569         [IPVS_DEST_ATTR_U_THRESH]       = { .type = NLA_U32 },
2570         [IPVS_DEST_ATTR_L_THRESH]       = { .type = NLA_U32 },
2571         [IPVS_DEST_ATTR_ACTIVE_CONNS]   = { .type = NLA_U32 },
2572         [IPVS_DEST_ATTR_INACT_CONNS]    = { .type = NLA_U32 },
2573         [IPVS_DEST_ATTR_PERSIST_CONNS]  = { .type = NLA_U32 },
2574         [IPVS_DEST_ATTR_STATS]          = { .type = NLA_NESTED },
2575 };
2576
2577 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2578                                  struct ip_vs_stats *stats)
2579 {
2580         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2581         if (!nl_stats)
2582                 return -EMSGSIZE;
2583
2584         spin_lock_bh(&stats->lock);
2585
2586         NLA_PUT_U32(skb, IPVS_STATS_ATTR_CONNS, stats->ustats.conns);
2587         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPKTS, stats->ustats.inpkts);
2588         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPKTS, stats->ustats.outpkts);
2589         NLA_PUT_U64(skb, IPVS_STATS_ATTR_INBYTES, stats->ustats.inbytes);
2590         NLA_PUT_U64(skb, IPVS_STATS_ATTR_OUTBYTES, stats->ustats.outbytes);
2591         NLA_PUT_U32(skb, IPVS_STATS_ATTR_CPS, stats->ustats.cps);
2592         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPPS, stats->ustats.inpps);
2593         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPPS, stats->ustats.outpps);
2594         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INBPS, stats->ustats.inbps);
2595         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTBPS, stats->ustats.outbps);
2596
2597         spin_unlock_bh(&stats->lock);
2598
2599         nla_nest_end(skb, nl_stats);
2600
2601         return 0;
2602
2603 nla_put_failure:
2604         spin_unlock_bh(&stats->lock);
2605         nla_nest_cancel(skb, nl_stats);
2606         return -EMSGSIZE;
2607 }
2608
2609 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2610                                    struct ip_vs_service *svc)
2611 {
2612         struct nlattr *nl_service;
2613         struct ip_vs_flags flags = { .flags = svc->flags,
2614                                      .mask = ~0 };
2615
2616         nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2617         if (!nl_service)
2618                 return -EMSGSIZE;
2619
2620         NLA_PUT_U16(skb, IPVS_SVC_ATTR_AF, svc->af);
2621
2622         if (svc->fwmark) {
2623                 NLA_PUT_U32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark);
2624         } else {
2625                 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol);
2626                 NLA_PUT(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr);
2627                 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PORT, svc->port);
2628         }
2629
2630         NLA_PUT_STRING(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name);
2631         NLA_PUT(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags);
2632         NLA_PUT_U32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ);
2633         NLA_PUT_U32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask);
2634
2635         if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2636                 goto nla_put_failure;
2637
2638         nla_nest_end(skb, nl_service);
2639
2640         return 0;
2641
2642 nla_put_failure:
2643         nla_nest_cancel(skb, nl_service);
2644         return -EMSGSIZE;
2645 }
2646
2647 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2648                                    struct ip_vs_service *svc,
2649                                    struct netlink_callback *cb)
2650 {
2651         void *hdr;
2652
2653         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2654                           &ip_vs_genl_family, NLM_F_MULTI,
2655                           IPVS_CMD_NEW_SERVICE);
2656         if (!hdr)
2657                 return -EMSGSIZE;
2658
2659         if (ip_vs_genl_fill_service(skb, svc) < 0)
2660                 goto nla_put_failure;
2661
2662         return genlmsg_end(skb, hdr);
2663
2664 nla_put_failure:
2665         genlmsg_cancel(skb, hdr);
2666         return -EMSGSIZE;
2667 }
2668
2669 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2670                                     struct netlink_callback *cb)
2671 {
2672         int idx = 0, i;
2673         int start = cb->args[0];
2674         struct ip_vs_service *svc;
2675
2676         mutex_lock(&__ip_vs_mutex);
2677         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2678                 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2679                         if (++idx <= start)
2680                                 continue;
2681                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2682                                 idx--;
2683                                 goto nla_put_failure;
2684                         }
2685                 }
2686         }
2687
2688         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2689                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2690                         if (++idx <= start)
2691                                 continue;
2692                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2693                                 idx--;
2694                                 goto nla_put_failure;
2695                         }
2696                 }
2697         }
2698
2699 nla_put_failure:
2700         mutex_unlock(&__ip_vs_mutex);
2701         cb->args[0] = idx;
2702
2703         return skb->len;
2704 }
2705
2706 static int ip_vs_genl_parse_service(struct ip_vs_service_user_kern *usvc,
2707                                     struct nlattr *nla, int full_entry)
2708 {
2709         struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2710         struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2711
2712         /* Parse mandatory identifying service fields first */
2713         if (nla == NULL ||
2714             nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2715                 return -EINVAL;
2716
2717         nla_af          = attrs[IPVS_SVC_ATTR_AF];
2718         nla_protocol    = attrs[IPVS_SVC_ATTR_PROTOCOL];
2719         nla_addr        = attrs[IPVS_SVC_ATTR_ADDR];
2720         nla_port        = attrs[IPVS_SVC_ATTR_PORT];
2721         nla_fwmark      = attrs[IPVS_SVC_ATTR_FWMARK];
2722
2723         if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2724                 return -EINVAL;
2725
2726         usvc->af = nla_get_u16(nla_af);
2727 #ifdef CONFIG_IP_VS_IPV6
2728         if (usvc->af != AF_INET && usvc->af != AF_INET6)
2729 #else
2730         if (usvc->af != AF_INET)
2731 #endif
2732                 return -EAFNOSUPPORT;
2733
2734         if (nla_fwmark) {
2735                 usvc->protocol = IPPROTO_TCP;
2736                 usvc->fwmark = nla_get_u32(nla_fwmark);
2737         } else {
2738                 usvc->protocol = nla_get_u16(nla_protocol);
2739                 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
2740                 usvc->port = nla_get_u16(nla_port);
2741                 usvc->fwmark = 0;
2742         }
2743
2744         /* If a full entry was requested, check for the additional fields */
2745         if (full_entry) {
2746                 struct nlattr *nla_sched, *nla_flags, *nla_timeout,
2747                               *nla_netmask;
2748                 struct ip_vs_flags flags;
2749                 struct ip_vs_service *svc;
2750
2751                 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
2752                 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
2753                 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
2754                 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
2755
2756                 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
2757                         return -EINVAL;
2758
2759                 nla_memcpy(&flags, nla_flags, sizeof(flags));
2760
2761                 /* prefill flags from service if it already exists */
2762                 if (usvc->fwmark)
2763                         svc = __ip_vs_svc_fwm_get(usvc->af, usvc->fwmark);
2764                 else
2765                         svc = __ip_vs_service_get(usvc->af, usvc->protocol,
2766                                                   &usvc->addr, usvc->port);
2767                 if (svc) {
2768                         usvc->flags = svc->flags;
2769                         ip_vs_service_put(svc);
2770                 } else
2771                         usvc->flags = 0;
2772
2773                 /* set new flags from userland */
2774                 usvc->flags = (usvc->flags & ~flags.mask) |
2775                               (flags.flags & flags.mask);
2776                 usvc->sched_name = nla_data(nla_sched);
2777                 usvc->timeout = nla_get_u32(nla_timeout);
2778                 usvc->netmask = nla_get_u32(nla_netmask);
2779         }
2780
2781         return 0;
2782 }
2783
2784 static struct ip_vs_service *ip_vs_genl_find_service(struct nlattr *nla)
2785 {
2786         struct ip_vs_service_user_kern usvc;
2787         int ret;
2788
2789         ret = ip_vs_genl_parse_service(&usvc, nla, 0);
2790         if (ret)
2791                 return ERR_PTR(ret);
2792
2793         if (usvc.fwmark)
2794                 return __ip_vs_svc_fwm_get(usvc.af, usvc.fwmark);
2795         else
2796                 return __ip_vs_service_get(usvc.af, usvc.protocol,
2797                                            &usvc.addr, usvc.port);
2798 }
2799
2800 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
2801 {
2802         struct nlattr *nl_dest;
2803
2804         nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
2805         if (!nl_dest)
2806                 return -EMSGSIZE;
2807
2808         NLA_PUT(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr);
2809         NLA_PUT_U16(skb, IPVS_DEST_ATTR_PORT, dest->port);
2810
2811         NLA_PUT_U32(skb, IPVS_DEST_ATTR_FWD_METHOD,
2812                     atomic_read(&dest->conn_flags) & IP_VS_CONN_F_FWD_MASK);
2813         NLA_PUT_U32(skb, IPVS_DEST_ATTR_WEIGHT, atomic_read(&dest->weight));
2814         NLA_PUT_U32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold);
2815         NLA_PUT_U32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold);
2816         NLA_PUT_U32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
2817                     atomic_read(&dest->activeconns));
2818         NLA_PUT_U32(skb, IPVS_DEST_ATTR_INACT_CONNS,
2819                     atomic_read(&dest->inactconns));
2820         NLA_PUT_U32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
2821                     atomic_read(&dest->persistconns));
2822
2823         if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
2824                 goto nla_put_failure;
2825
2826         nla_nest_end(skb, nl_dest);
2827
2828         return 0;
2829
2830 nla_put_failure:
2831         nla_nest_cancel(skb, nl_dest);
2832         return -EMSGSIZE;
2833 }
2834
2835 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
2836                                 struct netlink_callback *cb)
2837 {
2838         void *hdr;
2839
2840         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2841                           &ip_vs_genl_family, NLM_F_MULTI,
2842                           IPVS_CMD_NEW_DEST);
2843         if (!hdr)
2844                 return -EMSGSIZE;
2845
2846         if (ip_vs_genl_fill_dest(skb, dest) < 0)
2847                 goto nla_put_failure;
2848
2849         return genlmsg_end(skb, hdr);
2850
2851 nla_put_failure:
2852         genlmsg_cancel(skb, hdr);
2853         return -EMSGSIZE;
2854 }
2855
2856 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
2857                                  struct netlink_callback *cb)
2858 {
2859         int idx = 0;
2860         int start = cb->args[0];
2861         struct ip_vs_service *svc;
2862         struct ip_vs_dest *dest;
2863         struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
2864
2865         mutex_lock(&__ip_vs_mutex);
2866
2867         /* Try to find the service for which to dump destinations */
2868         if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
2869                         IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
2870                 goto out_err;
2871
2872         svc = ip_vs_genl_find_service(attrs[IPVS_CMD_ATTR_SERVICE]);
2873         if (IS_ERR(svc) || svc == NULL)
2874                 goto out_err;
2875
2876         /* Dump the destinations */
2877         list_for_each_entry(dest, &svc->destinations, n_list) {
2878                 if (++idx <= start)
2879                         continue;
2880                 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
2881                         idx--;
2882                         goto nla_put_failure;
2883                 }
2884         }
2885
2886 nla_put_failure:
2887         cb->args[0] = idx;
2888         ip_vs_service_put(svc);
2889
2890 out_err:
2891         mutex_unlock(&__ip_vs_mutex);
2892
2893         return skb->len;
2894 }
2895
2896 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
2897                                  struct nlattr *nla, int full_entry)
2898 {
2899         struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
2900         struct nlattr *nla_addr, *nla_port;
2901
2902         /* Parse mandatory identifying destination fields first */
2903         if (nla == NULL ||
2904             nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
2905                 return -EINVAL;
2906
2907         nla_addr        = attrs[IPVS_DEST_ATTR_ADDR];
2908         nla_port        = attrs[IPVS_DEST_ATTR_PORT];
2909
2910         if (!(nla_addr && nla_port))
2911                 return -EINVAL;
2912
2913         nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
2914         udest->port = nla_get_u16(nla_port);
2915
2916         /* If a full entry was requested, check for the additional fields */
2917         if (full_entry) {
2918                 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
2919                               *nla_l_thresh;
2920
2921                 nla_fwd         = attrs[IPVS_DEST_ATTR_FWD_METHOD];
2922                 nla_weight      = attrs[IPVS_DEST_ATTR_WEIGHT];
2923                 nla_u_thresh    = attrs[IPVS_DEST_ATTR_U_THRESH];
2924                 nla_l_thresh    = attrs[IPVS_DEST_ATTR_L_THRESH];
2925
2926                 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
2927                         return -EINVAL;
2928
2929                 udest->conn_flags = nla_get_u32(nla_fwd)
2930                                     & IP_VS_CONN_F_FWD_MASK;
2931                 udest->weight = nla_get_u32(nla_weight);
2932                 udest->u_threshold = nla_get_u32(nla_u_thresh);
2933                 udest->l_threshold = nla_get_u32(nla_l_thresh);
2934         }
2935
2936         return 0;
2937 }
2938
2939 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
2940                                   const char *mcast_ifn, __be32 syncid)
2941 {
2942         struct nlattr *nl_daemon;
2943
2944         nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
2945         if (!nl_daemon)
2946                 return -EMSGSIZE;
2947
2948         NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_STATE, state);
2949         NLA_PUT_STRING(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn);
2950         NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid);
2951
2952         nla_nest_end(skb, nl_daemon);
2953
2954         return 0;
2955
2956 nla_put_failure:
2957         nla_nest_cancel(skb, nl_daemon);
2958         return -EMSGSIZE;
2959 }
2960
2961 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
2962                                   const char *mcast_ifn, __be32 syncid,
2963                                   struct netlink_callback *cb)
2964 {
2965         void *hdr;
2966         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2967                           &ip_vs_genl_family, NLM_F_MULTI,
2968                           IPVS_CMD_NEW_DAEMON);
2969         if (!hdr)
2970                 return -EMSGSIZE;
2971
2972         if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
2973                 goto nla_put_failure;
2974
2975         return genlmsg_end(skb, hdr);
2976
2977 nla_put_failure:
2978         genlmsg_cancel(skb, hdr);
2979         return -EMSGSIZE;
2980 }
2981
2982 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
2983                                    struct netlink_callback *cb)
2984 {
2985         mutex_lock(&__ip_vs_mutex);
2986         if ((ip_vs_sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
2987                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
2988                                            ip_vs_master_mcast_ifn,
2989                                            ip_vs_master_syncid, cb) < 0)
2990                         goto nla_put_failure;
2991
2992                 cb->args[0] = 1;
2993         }
2994
2995         if ((ip_vs_sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
2996                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
2997                                            ip_vs_backup_mcast_ifn,
2998                                            ip_vs_backup_syncid, cb) < 0)
2999                         goto nla_put_failure;
3000
3001                 cb->args[1] = 1;
3002         }
3003
3004 nla_put_failure:
3005         mutex_unlock(&__ip_vs_mutex);
3006
3007         return skb->len;
3008 }
3009
3010 static int ip_vs_genl_new_daemon(struct nlattr **attrs)
3011 {
3012         if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3013               attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3014               attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3015                 return -EINVAL;
3016
3017         return start_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3018                                  nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3019                                  nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3020 }
3021
3022 static int ip_vs_genl_del_daemon(struct nlattr **attrs)
3023 {
3024         if (!attrs[IPVS_DAEMON_ATTR_STATE])
3025                 return -EINVAL;
3026
3027         return stop_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3028 }
3029
3030 static int ip_vs_genl_set_config(struct nlattr **attrs)
3031 {
3032         struct ip_vs_timeout_user t;
3033
3034         __ip_vs_get_timeouts(&t);
3035
3036         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3037                 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3038
3039         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3040                 t.tcp_fin_timeout =
3041                         nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3042
3043         if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3044                 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3045
3046         return ip_vs_set_timeout(&t);
3047 }
3048
3049 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3050 {
3051         struct ip_vs_service *svc = NULL;
3052         struct ip_vs_service_user_kern usvc;
3053         struct ip_vs_dest_user_kern udest;
3054         int ret = 0, cmd;
3055         int need_full_svc = 0, need_full_dest = 0;
3056
3057         cmd = info->genlhdr->cmd;
3058
3059         mutex_lock(&__ip_vs_mutex);
3060
3061         if (cmd == IPVS_CMD_FLUSH) {
3062                 ret = ip_vs_flush();
3063                 goto out;
3064         } else if (cmd == IPVS_CMD_SET_CONFIG) {
3065                 ret = ip_vs_genl_set_config(info->attrs);
3066                 goto out;
3067         } else if (cmd == IPVS_CMD_NEW_DAEMON ||
3068                    cmd == IPVS_CMD_DEL_DAEMON) {
3069
3070                 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3071
3072                 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3073                     nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3074                                      info->attrs[IPVS_CMD_ATTR_DAEMON],
3075                                      ip_vs_daemon_policy)) {
3076                         ret = -EINVAL;
3077                         goto out;
3078                 }
3079
3080                 if (cmd == IPVS_CMD_NEW_DAEMON)
3081                         ret = ip_vs_genl_new_daemon(daemon_attrs);
3082                 else
3083                         ret = ip_vs_genl_del_daemon(daemon_attrs);
3084                 goto out;
3085         } else if (cmd == IPVS_CMD_ZERO &&
3086                    !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3087                 ret = ip_vs_zero_all();
3088                 goto out;
3089         }
3090
3091         /* All following commands require a service argument, so check if we
3092          * received a valid one. We need a full service specification when
3093          * adding / editing a service. Only identifying members otherwise. */
3094         if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3095                 need_full_svc = 1;
3096
3097         ret = ip_vs_genl_parse_service(&usvc,
3098                                        info->attrs[IPVS_CMD_ATTR_SERVICE],
3099                                        need_full_svc);
3100         if (ret)
3101                 goto out;
3102
3103         /* Lookup the exact service by <protocol, addr, port> or fwmark */
3104         if (usvc.fwmark == 0)
3105                 svc = __ip_vs_service_get(usvc.af, usvc.protocol,
3106                                           &usvc.addr, usvc.port);
3107         else
3108                 svc = __ip_vs_svc_fwm_get(usvc.af, usvc.fwmark);
3109
3110         /* Unless we're adding a new service, the service must already exist */
3111         if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3112                 ret = -ESRCH;
3113                 goto out;
3114         }
3115
3116         /* Destination commands require a valid destination argument. For
3117          * adding / editing a destination, we need a full destination
3118          * specification. */
3119         if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3120             cmd == IPVS_CMD_DEL_DEST) {
3121                 if (cmd != IPVS_CMD_DEL_DEST)
3122                         need_full_dest = 1;
3123
3124                 ret = ip_vs_genl_parse_dest(&udest,
3125                                             info->attrs[IPVS_CMD_ATTR_DEST],
3126                                             need_full_dest);
3127                 if (ret)
3128                         goto out;
3129         }
3130
3131         switch (cmd) {
3132         case IPVS_CMD_NEW_SERVICE:
3133                 if (svc == NULL)
3134                         ret = ip_vs_add_service(&usvc, &svc);
3135                 else
3136                         ret = -EEXIST;
3137                 break;
3138         case IPVS_CMD_SET_SERVICE:
3139                 ret = ip_vs_edit_service(svc, &usvc);
3140                 break;
3141         case IPVS_CMD_DEL_SERVICE:
3142                 ret = ip_vs_del_service(svc);
3143                 break;
3144         case IPVS_CMD_NEW_DEST:
3145                 ret = ip_vs_add_dest(svc, &udest);
3146                 break;
3147         case IPVS_CMD_SET_DEST:
3148                 ret = ip_vs_edit_dest(svc, &udest);
3149                 break;
3150         case IPVS_CMD_DEL_DEST:
3151                 ret = ip_vs_del_dest(svc, &udest);
3152                 break;
3153         case IPVS_CMD_ZERO:
3154                 ret = ip_vs_zero_service(svc);
3155                 break;
3156         default:
3157                 ret = -EINVAL;
3158         }
3159
3160 out:
3161         if (svc)
3162                 ip_vs_service_put(svc);
3163         mutex_unlock(&__ip_vs_mutex);
3164
3165         return ret;
3166 }
3167
3168 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3169 {
3170         struct sk_buff *msg;
3171         void *reply;
3172         int ret, cmd, reply_cmd;
3173
3174         cmd = info->genlhdr->cmd;
3175
3176         if (cmd == IPVS_CMD_GET_SERVICE)
3177                 reply_cmd = IPVS_CMD_NEW_SERVICE;
3178         else if (cmd == IPVS_CMD_GET_INFO)
3179                 reply_cmd = IPVS_CMD_SET_INFO;
3180         else if (cmd == IPVS_CMD_GET_CONFIG)
3181                 reply_cmd = IPVS_CMD_SET_CONFIG;
3182         else {
3183                 IP_VS_ERR("unknown Generic Netlink command\n");
3184                 return -EINVAL;
3185         }
3186
3187         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3188         if (!msg)
3189                 return -ENOMEM;
3190
3191         mutex_lock(&__ip_vs_mutex);
3192
3193         reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3194         if (reply == NULL)
3195                 goto nla_put_failure;
3196
3197         switch (cmd) {
3198         case IPVS_CMD_GET_SERVICE:
3199         {
3200                 struct ip_vs_service *svc;
3201
3202                 svc = ip_vs_genl_find_service(info->attrs[IPVS_CMD_ATTR_SERVICE]);
3203                 if (IS_ERR(svc)) {
3204                         ret = PTR_ERR(svc);
3205                         goto out_err;
3206                 } else if (svc) {
3207                         ret = ip_vs_genl_fill_service(msg, svc);
3208                         ip_vs_service_put(svc);
3209                         if (ret)
3210                                 goto nla_put_failure;
3211                 } else {
3212                         ret = -ESRCH;
3213                         goto out_err;
3214                 }
3215
3216                 break;
3217         }
3218
3219         case IPVS_CMD_GET_CONFIG:
3220         {
3221                 struct ip_vs_timeout_user t;
3222
3223                 __ip_vs_get_timeouts(&t);
3224 #ifdef CONFIG_IP_VS_PROTO_TCP
3225                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP, t.tcp_timeout);
3226                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3227                             t.tcp_fin_timeout);
3228 #endif
3229 #ifdef CONFIG_IP_VS_PROTO_UDP
3230                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout);
3231 #endif
3232
3233                 break;
3234         }
3235
3236         case IPVS_CMD_GET_INFO:
3237                 NLA_PUT_U32(msg, IPVS_INFO_ATTR_VERSION, IP_VS_VERSION_CODE);
3238                 NLA_PUT_U32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3239                             IP_VS_CONN_TAB_SIZE);
3240                 break;
3241         }
3242
3243         genlmsg_end(msg, reply);
3244         ret = genlmsg_unicast(msg, info->snd_pid);
3245         goto out;
3246
3247 nla_put_failure:
3248         IP_VS_ERR("not enough space in Netlink message\n");
3249         ret = -EMSGSIZE;
3250
3251 out_err:
3252         nlmsg_free(msg);
3253 out:
3254         mutex_unlock(&__ip_vs_mutex);
3255
3256         return ret;
3257 }
3258
3259
3260 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3261         {
3262                 .cmd    = IPVS_CMD_NEW_SERVICE,
3263                 .flags  = GENL_ADMIN_PERM,
3264                 .policy = ip_vs_cmd_policy,
3265                 .doit   = ip_vs_genl_set_cmd,
3266         },
3267         {
3268                 .cmd    = IPVS_CMD_SET_SERVICE,
3269                 .flags  = GENL_ADMIN_PERM,
3270                 .policy = ip_vs_cmd_policy,
3271                 .doit   = ip_vs_genl_set_cmd,
3272         },
3273         {
3274                 .cmd    = IPVS_CMD_DEL_SERVICE,
3275                 .flags  = GENL_ADMIN_PERM,
3276                 .policy = ip_vs_cmd_policy,
3277                 .doit   = ip_vs_genl_set_cmd,
3278         },
3279         {
3280                 .cmd    = IPVS_CMD_GET_SERVICE,
3281                 .flags  = GENL_ADMIN_PERM,
3282                 .doit   = ip_vs_genl_get_cmd,
3283                 .dumpit = ip_vs_genl_dump_services,
3284                 .policy = ip_vs_cmd_policy,
3285         },
3286         {
3287                 .cmd    = IPVS_CMD_NEW_DEST,
3288                 .flags  = GENL_ADMIN_PERM,
3289                 .policy = ip_vs_cmd_policy,
3290                 .doit   = ip_vs_genl_set_cmd,
3291         },
3292         {
3293                 .cmd    = IPVS_CMD_SET_DEST,
3294                 .flags  = GENL_ADMIN_PERM,
3295                 .policy = ip_vs_cmd_policy,
3296                 .doit   = ip_vs_genl_set_cmd,
3297         },
3298         {
3299                 .cmd    = IPVS_CMD_DEL_DEST,
3300                 .flags  = GENL_ADMIN_PERM,
3301                 .policy = ip_vs_cmd_policy,
3302                 .doit   = ip_vs_genl_set_cmd,
3303         },
3304         {
3305                 .cmd    = IPVS_CMD_GET_DEST,
3306                 .flags  = GENL_ADMIN_PERM,
3307                 .policy = ip_vs_cmd_policy,
3308                 .dumpit = ip_vs_genl_dump_dests,
3309         },
3310         {
3311                 .cmd    = IPVS_CMD_NEW_DAEMON,
3312                 .flags  = GENL_ADMIN_PERM,
3313                 .policy = ip_vs_cmd_policy,
3314                 .doit   = ip_vs_genl_set_cmd,
3315         },
3316         {
3317                 .cmd    = IPVS_CMD_DEL_DAEMON,
3318                 .flags  = GENL_ADMIN_PERM,
3319                 .policy = ip_vs_cmd_policy,
3320                 .doit   = ip_vs_genl_set_cmd,
3321         },
3322         {
3323                 .cmd    = IPVS_CMD_GET_DAEMON,
3324                 .flags  = GENL_ADMIN_PERM,
3325                 .dumpit = ip_vs_genl_dump_daemons,
3326         },
3327         {
3328                 .cmd    = IPVS_CMD_SET_CONFIG,
3329                 .flags  = GENL_ADMIN_PERM,
3330                 .policy = ip_vs_cmd_policy,
3331                 .doit   = ip_vs_genl_set_cmd,
3332         },
3333         {
3334                 .cmd    = IPVS_CMD_GET_CONFIG,
3335                 .flags  = GENL_ADMIN_PERM,
3336                 .doit   = ip_vs_genl_get_cmd,
3337         },
3338         {
3339                 .cmd    = IPVS_CMD_GET_INFO,
3340                 .flags  = GENL_ADMIN_PERM,
3341                 .doit   = ip_vs_genl_get_cmd,
3342         },
3343         {
3344                 .cmd    = IPVS_CMD_ZERO,
3345                 .flags  = GENL_ADMIN_PERM,
3346                 .policy = ip_vs_cmd_policy,
3347                 .doit   = ip_vs_genl_set_cmd,
3348         },
3349         {
3350                 .cmd    = IPVS_CMD_FLUSH,
3351                 .flags  = GENL_ADMIN_PERM,
3352                 .doit   = ip_vs_genl_set_cmd,
3353         },
3354 };
3355
3356 static int __init ip_vs_genl_register(void)
3357 {
3358         int ret, i;
3359
3360         ret = genl_register_family(&ip_vs_genl_family);
3361         if (ret)
3362                 return ret;
3363
3364         for (i = 0; i < ARRAY_SIZE(ip_vs_genl_ops); i++) {
3365                 ret = genl_register_ops(&ip_vs_genl_family, &ip_vs_genl_ops[i]);
3366                 if (ret)
3367                         goto err_out;
3368         }
3369         return 0;
3370
3371 err_out:
3372         genl_unregister_family(&ip_vs_genl_family);
3373         return ret;
3374 }
3375
3376 static void ip_vs_genl_unregister(void)
3377 {
3378         genl_unregister_family(&ip_vs_genl_family);
3379 }
3380
3381 /* End of Generic Netlink interface definitions */
3382
3383
3384 int __init ip_vs_control_init(void)
3385 {
3386         int ret;
3387         int idx;
3388
3389         EnterFunction(2);
3390
3391         ret = nf_register_sockopt(&ip_vs_sockopts);
3392         if (ret) {
3393                 IP_VS_ERR("cannot register sockopt.\n");
3394                 return ret;
3395         }
3396
3397         ret = ip_vs_genl_register();
3398         if (ret) {
3399                 IP_VS_ERR("cannot register Generic Netlink interface.\n");
3400                 nf_unregister_sockopt(&ip_vs_sockopts);
3401                 return ret;
3402         }
3403
3404         proc_net_fops_create(&init_net, "ip_vs", 0, &ip_vs_info_fops);
3405         proc_net_fops_create(&init_net, "ip_vs_stats",0, &ip_vs_stats_fops);
3406
3407         sysctl_header = register_sysctl_paths(net_vs_ctl_path, vs_vars);
3408
3409         /* Initialize ip_vs_svc_table, ip_vs_svc_fwm_table, ip_vs_rtable */
3410         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++)  {
3411                 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3412                 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3413         }
3414         for(idx = 0; idx < IP_VS_RTAB_SIZE; idx++)  {
3415                 INIT_LIST_HEAD(&ip_vs_rtable[idx]);
3416         }
3417
3418         ip_vs_new_estimator(&ip_vs_stats);
3419
3420         /* Hook the defense timer */
3421         schedule_delayed_work(&defense_work, DEFENSE_TIMER_PERIOD);
3422
3423         LeaveFunction(2);
3424         return 0;
3425 }
3426
3427
3428 void ip_vs_control_cleanup(void)
3429 {
3430         EnterFunction(2);
3431         ip_vs_trash_cleanup();
3432         cancel_rearming_delayed_work(&defense_work);
3433         cancel_work_sync(&defense_work.work);
3434         ip_vs_kill_estimator(&ip_vs_stats);
3435         unregister_sysctl_table(sysctl_header);
3436         proc_net_remove(&init_net, "ip_vs_stats");
3437         proc_net_remove(&init_net, "ip_vs");
3438         ip_vs_genl_unregister();
3439         nf_unregister_sockopt(&ip_vs_sockopts);
3440         LeaveFunction(2);
3441 }