net/netfilter/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 #ifdef CONFIG_IP_VS_IPV6
1334   out:
1335 #endif
1336
1337         if (old_sched)
1338                 ip_vs_scheduler_put(old_sched);
1339
1340         return ret;
1341 }
1342
1343
1344 /*
1345  *      Delete a service from the service list
1346  *      - The service must be unlinked, unlocked and not referenced!
1347  *      - We are called under _bh lock
1348  */
1349 static void __ip_vs_del_service(struct ip_vs_service *svc)
1350 {
1351         struct ip_vs_dest *dest, *nxt;
1352         struct ip_vs_scheduler *old_sched;
1353
1354         /* Count only IPv4 services for old get/setsockopt interface */
1355         if (svc->af == AF_INET)
1356                 ip_vs_num_services--;
1357
1358         ip_vs_kill_estimator(&svc->stats);
1359
1360         /* Unbind scheduler */
1361         old_sched = svc->scheduler;
1362         ip_vs_unbind_scheduler(svc);
1363         if (old_sched)
1364                 ip_vs_scheduler_put(old_sched);
1365
1366         /* Unbind app inc */
1367         if (svc->inc) {
1368                 ip_vs_app_inc_put(svc->inc);
1369                 svc->inc = NULL;
1370         }
1371
1372         /*
1373          *    Unlink the whole destination list
1374          */
1375         list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1376                 __ip_vs_unlink_dest(svc, dest, 0);
1377                 __ip_vs_del_dest(dest);
1378         }
1379
1380         /*
1381          *    Update the virtual service counters
1382          */
1383         if (svc->port == FTPPORT)
1384                 atomic_dec(&ip_vs_ftpsvc_counter);
1385         else if (svc->port == 0)
1386                 atomic_dec(&ip_vs_nullsvc_counter);
1387
1388         /*
1389          *    Free the service if nobody refers to it
1390          */
1391         if (atomic_read(&svc->refcnt) == 0)
1392                 kfree(svc);
1393
1394         /* decrease the module use count */
1395         ip_vs_use_count_dec();
1396 }
1397
1398 /*
1399  *      Delete a service from the service list
1400  */
1401 static int ip_vs_del_service(struct ip_vs_service *svc)
1402 {
1403         if (svc == NULL)
1404                 return -EEXIST;
1405
1406         /*
1407          * Unhash it from the service table
1408          */
1409         write_lock_bh(&__ip_vs_svc_lock);
1410
1411         ip_vs_svc_unhash(svc);
1412
1413         /*
1414          * Wait until all the svc users go away.
1415          */
1416         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
1417
1418         __ip_vs_del_service(svc);
1419
1420         write_unlock_bh(&__ip_vs_svc_lock);
1421
1422         return 0;
1423 }
1424
1425
1426 /*
1427  *      Flush all the virtual services
1428  */
1429 static int ip_vs_flush(void)
1430 {
1431         int idx;
1432         struct ip_vs_service *svc, *nxt;
1433
1434         /*
1435          * Flush the service table hashed by <protocol,addr,port>
1436          */
1437         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1438                 list_for_each_entry_safe(svc, nxt, &ip_vs_svc_table[idx], s_list) {
1439                         write_lock_bh(&__ip_vs_svc_lock);
1440                         ip_vs_svc_unhash(svc);
1441                         /*
1442                          * Wait until all the svc users go away.
1443                          */
1444                         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1445                         __ip_vs_del_service(svc);
1446                         write_unlock_bh(&__ip_vs_svc_lock);
1447                 }
1448         }
1449
1450         /*
1451          * Flush the service table hashed by fwmark
1452          */
1453         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1454                 list_for_each_entry_safe(svc, nxt,
1455                                          &ip_vs_svc_fwm_table[idx], f_list) {
1456                         write_lock_bh(&__ip_vs_svc_lock);
1457                         ip_vs_svc_unhash(svc);
1458                         /*
1459                          * Wait until all the svc users go away.
1460                          */
1461                         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1462                         __ip_vs_del_service(svc);
1463                         write_unlock_bh(&__ip_vs_svc_lock);
1464                 }
1465         }
1466
1467         return 0;
1468 }
1469
1470
1471 /*
1472  *      Zero counters in a service or all services
1473  */
1474 static int ip_vs_zero_service(struct ip_vs_service *svc)
1475 {
1476         struct ip_vs_dest *dest;
1477
1478         write_lock_bh(&__ip_vs_svc_lock);
1479         list_for_each_entry(dest, &svc->destinations, n_list) {
1480                 ip_vs_zero_stats(&dest->stats);
1481         }
1482         ip_vs_zero_stats(&svc->stats);
1483         write_unlock_bh(&__ip_vs_svc_lock);
1484         return 0;
1485 }
1486
1487 static int ip_vs_zero_all(void)
1488 {
1489         int idx;
1490         struct ip_vs_service *svc;
1491
1492         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1493                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1494                         ip_vs_zero_service(svc);
1495                 }
1496         }
1497
1498         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1499                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1500                         ip_vs_zero_service(svc);
1501                 }
1502         }
1503
1504         ip_vs_zero_stats(&ip_vs_stats);
1505         return 0;
1506 }
1507
1508
1509 static int
1510 proc_do_defense_mode(ctl_table *table, int write, struct file * filp,
1511                      void __user *buffer, size_t *lenp, loff_t *ppos)
1512 {
1513         int *valp = table->data;
1514         int val = *valp;
1515         int rc;
1516
1517         rc = proc_dointvec(table, write, filp, buffer, lenp, ppos);
1518         if (write && (*valp != val)) {
1519                 if ((*valp < 0) || (*valp > 3)) {
1520                         /* Restore the correct value */
1521                         *valp = val;
1522                 } else {
1523                         update_defense_level();
1524                 }
1525         }
1526         return rc;
1527 }
1528
1529
1530 static int
1531 proc_do_sync_threshold(ctl_table *table, int write, struct file *filp,
1532                        void __user *buffer, size_t *lenp, loff_t *ppos)
1533 {
1534         int *valp = table->data;
1535         int val[2];
1536         int rc;
1537
1538         /* backup the value first */
1539         memcpy(val, valp, sizeof(val));
1540
1541         rc = proc_dointvec(table, write, filp, buffer, lenp, ppos);
1542         if (write && (valp[0] < 0 || valp[1] < 0 || valp[0] >= valp[1])) {
1543                 /* Restore the correct value */
1544                 memcpy(valp, val, sizeof(val));
1545         }
1546         return rc;
1547 }
1548
1549
1550 /*
1551  *      IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1552  */
1553
1554 static struct ctl_table vs_vars[] = {
1555         {
1556                 .procname       = "amemthresh",
1557                 .data           = &sysctl_ip_vs_amemthresh,
1558                 .maxlen         = sizeof(int),
1559                 .mode           = 0644,
1560                 .proc_handler   = &proc_dointvec,
1561         },
1562 #ifdef CONFIG_IP_VS_DEBUG
1563         {
1564                 .procname       = "debug_level",
1565                 .data           = &sysctl_ip_vs_debug_level,
1566                 .maxlen         = sizeof(int),
1567                 .mode           = 0644,
1568                 .proc_handler   = &proc_dointvec,
1569         },
1570 #endif
1571         {
1572                 .procname       = "am_droprate",
1573                 .data           = &sysctl_ip_vs_am_droprate,
1574                 .maxlen         = sizeof(int),
1575                 .mode           = 0644,
1576                 .proc_handler   = &proc_dointvec,
1577         },
1578         {
1579                 .procname       = "drop_entry",
1580                 .data           = &sysctl_ip_vs_drop_entry,
1581                 .maxlen         = sizeof(int),
1582                 .mode           = 0644,
1583                 .proc_handler   = &proc_do_defense_mode,
1584         },
1585         {
1586                 .procname       = "drop_packet",
1587                 .data           = &sysctl_ip_vs_drop_packet,
1588                 .maxlen         = sizeof(int),
1589                 .mode           = 0644,
1590                 .proc_handler   = &proc_do_defense_mode,
1591         },
1592         {
1593                 .procname       = "secure_tcp",
1594                 .data           = &sysctl_ip_vs_secure_tcp,
1595                 .maxlen         = sizeof(int),
1596                 .mode           = 0644,
1597                 .proc_handler   = &proc_do_defense_mode,
1598         },
1599 #if 0
1600         {
1601                 .procname       = "timeout_established",
1602                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1603                 .maxlen         = sizeof(int),
1604                 .mode           = 0644,
1605                 .proc_handler   = &proc_dointvec_jiffies,
1606         },
1607         {
1608                 .procname       = "timeout_synsent",
1609                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1610                 .maxlen         = sizeof(int),
1611                 .mode           = 0644,
1612                 .proc_handler   = &proc_dointvec_jiffies,
1613         },
1614         {
1615                 .procname       = "timeout_synrecv",
1616                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1617                 .maxlen         = sizeof(int),
1618                 .mode           = 0644,
1619                 .proc_handler   = &proc_dointvec_jiffies,
1620         },
1621         {
1622                 .procname       = "timeout_finwait",
1623                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1624                 .maxlen         = sizeof(int),
1625                 .mode           = 0644,
1626                 .proc_handler   = &proc_dointvec_jiffies,
1627         },
1628         {
1629                 .procname       = "timeout_timewait",
1630                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1631                 .maxlen         = sizeof(int),
1632                 .mode           = 0644,
1633                 .proc_handler   = &proc_dointvec_jiffies,
1634         },
1635         {
1636                 .procname       = "timeout_close",
1637                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1638                 .maxlen         = sizeof(int),
1639                 .mode           = 0644,
1640                 .proc_handler   = &proc_dointvec_jiffies,
1641         },
1642         {
1643                 .procname       = "timeout_closewait",
1644                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1645                 .maxlen         = sizeof(int),
1646                 .mode           = 0644,
1647                 .proc_handler   = &proc_dointvec_jiffies,
1648         },
1649         {
1650                 .procname       = "timeout_lastack",
1651                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1652                 .maxlen         = sizeof(int),
1653                 .mode           = 0644,
1654                 .proc_handler   = &proc_dointvec_jiffies,
1655         },
1656         {
1657                 .procname       = "timeout_listen",
1658                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1659                 .maxlen         = sizeof(int),
1660                 .mode           = 0644,
1661                 .proc_handler   = &proc_dointvec_jiffies,
1662         },
1663         {
1664                 .procname       = "timeout_synack",
1665                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1666                 .maxlen         = sizeof(int),
1667                 .mode           = 0644,
1668                 .proc_handler   = &proc_dointvec_jiffies,
1669         },
1670         {
1671                 .procname       = "timeout_udp",
1672                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1673                 .maxlen         = sizeof(int),
1674                 .mode           = 0644,
1675                 .proc_handler   = &proc_dointvec_jiffies,
1676         },
1677         {
1678                 .procname       = "timeout_icmp",
1679                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1680                 .maxlen         = sizeof(int),
1681                 .mode           = 0644,
1682                 .proc_handler   = &proc_dointvec_jiffies,
1683         },
1684 #endif
1685         {
1686                 .procname       = "cache_bypass",
1687                 .data           = &sysctl_ip_vs_cache_bypass,
1688                 .maxlen         = sizeof(int),
1689                 .mode           = 0644,
1690                 .proc_handler   = &proc_dointvec,
1691         },
1692         {
1693                 .procname       = "expire_nodest_conn",
1694                 .data           = &sysctl_ip_vs_expire_nodest_conn,
1695                 .maxlen         = sizeof(int),
1696                 .mode           = 0644,
1697                 .proc_handler   = &proc_dointvec,
1698         },
1699         {
1700                 .procname       = "expire_quiescent_template",
1701                 .data           = &sysctl_ip_vs_expire_quiescent_template,
1702                 .maxlen         = sizeof(int),
1703                 .mode           = 0644,
1704                 .proc_handler   = &proc_dointvec,
1705         },
1706         {
1707                 .procname       = "sync_threshold",
1708                 .data           = &sysctl_ip_vs_sync_threshold,
1709                 .maxlen         = sizeof(sysctl_ip_vs_sync_threshold),
1710                 .mode           = 0644,
1711                 .proc_handler   = &proc_do_sync_threshold,
1712         },
1713         {
1714                 .procname       = "nat_icmp_send",
1715                 .data           = &sysctl_ip_vs_nat_icmp_send,
1716                 .maxlen         = sizeof(int),
1717                 .mode           = 0644,
1718                 .proc_handler   = &proc_dointvec,
1719         },
1720         { .ctl_name = 0 }
1721 };
1722
1723 const struct ctl_path net_vs_ctl_path[] = {
1724         { .procname = "net", .ctl_name = CTL_NET, },
1725         { .procname = "ipv4", .ctl_name = NET_IPV4, },
1726         { .procname = "vs", },
1727         { }
1728 };
1729 EXPORT_SYMBOL_GPL(net_vs_ctl_path);
1730
1731 static struct ctl_table_header * sysctl_header;
1732
1733 #ifdef CONFIG_PROC_FS
1734
1735 struct ip_vs_iter {
1736         struct list_head *table;
1737         int bucket;
1738 };
1739
1740 /*
1741  *      Write the contents of the VS rule table to a PROCfs file.
1742  *      (It is kept just for backward compatibility)
1743  */
1744 static inline const char *ip_vs_fwd_name(unsigned flags)
1745 {
1746         switch (flags & IP_VS_CONN_F_FWD_MASK) {
1747         case IP_VS_CONN_F_LOCALNODE:
1748                 return "Local";
1749         case IP_VS_CONN_F_TUNNEL:
1750                 return "Tunnel";
1751         case IP_VS_CONN_F_DROUTE:
1752                 return "Route";
1753         default:
1754                 return "Masq";
1755         }
1756 }
1757
1758
1759 /* Get the Nth entry in the two lists */
1760 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1761 {
1762         struct ip_vs_iter *iter = seq->private;
1763         int idx;
1764         struct ip_vs_service *svc;
1765
1766         /* look in hash by protocol */
1767         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1768                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1769                         if (pos-- == 0){
1770                                 iter->table = ip_vs_svc_table;
1771                                 iter->bucket = idx;
1772                                 return svc;
1773                         }
1774                 }
1775         }
1776
1777         /* keep looking in fwmark */
1778         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1779                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1780                         if (pos-- == 0) {
1781                                 iter->table = ip_vs_svc_fwm_table;
1782                                 iter->bucket = idx;
1783                                 return svc;
1784                         }
1785                 }
1786         }
1787
1788         return NULL;
1789 }
1790
1791 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1792 __acquires(__ip_vs_svc_lock)
1793 {
1794
1795         read_lock_bh(&__ip_vs_svc_lock);
1796         return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1797 }
1798
1799
1800 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1801 {
1802         struct list_head *e;
1803         struct ip_vs_iter *iter;
1804         struct ip_vs_service *svc;
1805
1806         ++*pos;
1807         if (v == SEQ_START_TOKEN)
1808                 return ip_vs_info_array(seq,0);
1809
1810         svc = v;
1811         iter = seq->private;
1812
1813         if (iter->table == ip_vs_svc_table) {
1814                 /* next service in table hashed by protocol */
1815                 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
1816                         return list_entry(e, struct ip_vs_service, s_list);
1817
1818
1819                 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1820                         list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
1821                                             s_list) {
1822                                 return svc;
1823                         }
1824                 }
1825
1826                 iter->table = ip_vs_svc_fwm_table;
1827                 iter->bucket = -1;
1828                 goto scan_fwmark;
1829         }
1830
1831         /* next service in hashed by fwmark */
1832         if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
1833                 return list_entry(e, struct ip_vs_service, f_list);
1834
1835  scan_fwmark:
1836         while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1837                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
1838                                     f_list)
1839                         return svc;
1840         }
1841
1842         return NULL;
1843 }
1844
1845 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1846 __releases(__ip_vs_svc_lock)
1847 {
1848         read_unlock_bh(&__ip_vs_svc_lock);
1849 }
1850
1851
1852 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1853 {
1854         if (v == SEQ_START_TOKEN) {
1855                 seq_printf(seq,
1856                         "IP Virtual Server version %d.%d.%d (size=%d)\n",
1857                         NVERSION(IP_VS_VERSION_CODE), IP_VS_CONN_TAB_SIZE);
1858                 seq_puts(seq,
1859                          "Prot LocalAddress:Port Scheduler Flags\n");
1860                 seq_puts(seq,
1861                          "  -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1862         } else {
1863                 const struct ip_vs_service *svc = v;
1864                 const struct ip_vs_iter *iter = seq->private;
1865                 const struct ip_vs_dest *dest;
1866
1867                 if (iter->table == ip_vs_svc_table) {
1868 #ifdef CONFIG_IP_VS_IPV6
1869                         if (svc->af == AF_INET6)
1870                                 seq_printf(seq, "%s  [%pI6]:%04X %s ",
1871                                            ip_vs_proto_name(svc->protocol),
1872                                            &svc->addr.in6,
1873                                            ntohs(svc->port),
1874                                            svc->scheduler->name);
1875                         else
1876 #endif
1877                                 seq_printf(seq, "%s  %08X:%04X %s ",
1878                                            ip_vs_proto_name(svc->protocol),
1879                                            ntohl(svc->addr.ip),
1880                                            ntohs(svc->port),
1881                                            svc->scheduler->name);
1882                 } else {
1883                         seq_printf(seq, "FWM  %08X %s ",
1884                                    svc->fwmark, svc->scheduler->name);
1885                 }
1886
1887                 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
1888                         seq_printf(seq, "persistent %d %08X\n",
1889                                 svc->timeout,
1890                                 ntohl(svc->netmask));
1891                 else
1892                         seq_putc(seq, '\n');
1893
1894                 list_for_each_entry(dest, &svc->destinations, n_list) {
1895 #ifdef CONFIG_IP_VS_IPV6
1896                         if (dest->af == AF_INET6)
1897                                 seq_printf(seq,
1898                                            "  -> [%pI6]:%04X"
1899                                            "      %-7s %-6d %-10d %-10d\n",
1900                                            &dest->addr.in6,
1901                                            ntohs(dest->port),
1902                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1903                                            atomic_read(&dest->weight),
1904                                            atomic_read(&dest->activeconns),
1905                                            atomic_read(&dest->inactconns));
1906                         else
1907 #endif
1908                                 seq_printf(seq,
1909                                            "  -> %08X:%04X      "
1910                                            "%-7s %-6d %-10d %-10d\n",
1911                                            ntohl(dest->addr.ip),
1912                                            ntohs(dest->port),
1913                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1914                                            atomic_read(&dest->weight),
1915                                            atomic_read(&dest->activeconns),
1916                                            atomic_read(&dest->inactconns));
1917
1918                 }
1919         }
1920         return 0;
1921 }
1922
1923 static const struct seq_operations ip_vs_info_seq_ops = {
1924         .start = ip_vs_info_seq_start,
1925         .next  = ip_vs_info_seq_next,
1926         .stop  = ip_vs_info_seq_stop,
1927         .show  = ip_vs_info_seq_show,
1928 };
1929
1930 static int ip_vs_info_open(struct inode *inode, struct file *file)
1931 {
1932         return seq_open_private(file, &ip_vs_info_seq_ops,
1933                         sizeof(struct ip_vs_iter));
1934 }
1935
1936 static const struct file_operations ip_vs_info_fops = {
1937         .owner   = THIS_MODULE,
1938         .open    = ip_vs_info_open,
1939         .read    = seq_read,
1940         .llseek  = seq_lseek,
1941         .release = seq_release_private,
1942 };
1943
1944 #endif
1945
1946 struct ip_vs_stats ip_vs_stats = {
1947         .lock = __SPIN_LOCK_UNLOCKED(ip_vs_stats.lock),
1948 };
1949
1950 #ifdef CONFIG_PROC_FS
1951 static int ip_vs_stats_show(struct seq_file *seq, void *v)
1952 {
1953
1954 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
1955         seq_puts(seq,
1956                  "   Total Incoming Outgoing         Incoming         Outgoing\n");
1957         seq_printf(seq,
1958                    "   Conns  Packets  Packets            Bytes            Bytes\n");
1959
1960         spin_lock_bh(&ip_vs_stats.lock);
1961         seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", ip_vs_stats.ustats.conns,
1962                    ip_vs_stats.ustats.inpkts, ip_vs_stats.ustats.outpkts,
1963                    (unsigned long long) ip_vs_stats.ustats.inbytes,
1964                    (unsigned long long) ip_vs_stats.ustats.outbytes);
1965
1966 /*                 01234567 01234567 01234567 0123456701234567 0123456701234567 */
1967         seq_puts(seq,
1968                    " Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
1969         seq_printf(seq,"%8X %8X %8X %16X %16X\n",
1970                         ip_vs_stats.ustats.cps,
1971                         ip_vs_stats.ustats.inpps,
1972                         ip_vs_stats.ustats.outpps,
1973                         ip_vs_stats.ustats.inbps,
1974                         ip_vs_stats.ustats.outbps);
1975         spin_unlock_bh(&ip_vs_stats.lock);
1976
1977         return 0;
1978 }
1979
1980 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
1981 {
1982         return single_open(file, ip_vs_stats_show, NULL);
1983 }
1984
1985 static const struct file_operations ip_vs_stats_fops = {
1986         .owner = THIS_MODULE,
1987         .open = ip_vs_stats_seq_open,
1988         .read = seq_read,
1989         .llseek = seq_lseek,
1990         .release = single_release,
1991 };
1992
1993 #endif
1994
1995 /*
1996  *      Set timeout values for tcp tcpfin udp in the timeout_table.
1997  */
1998 static int ip_vs_set_timeout(struct ip_vs_timeout_user *u)
1999 {
2000         IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2001                   u->tcp_timeout,
2002                   u->tcp_fin_timeout,
2003                   u->udp_timeout);
2004
2005 #ifdef CONFIG_IP_VS_PROTO_TCP
2006         if (u->tcp_timeout) {
2007                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED]
2008                         = u->tcp_timeout * HZ;
2009         }
2010
2011         if (u->tcp_fin_timeout) {
2012                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT]
2013                         = u->tcp_fin_timeout * HZ;
2014         }
2015 #endif
2016
2017 #ifdef CONFIG_IP_VS_PROTO_UDP
2018         if (u->udp_timeout) {
2019                 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL]
2020                         = u->udp_timeout * HZ;
2021         }
2022 #endif
2023         return 0;
2024 }
2025
2026
2027 #define SET_CMDID(cmd)          (cmd - IP_VS_BASE_CTL)
2028 #define SERVICE_ARG_LEN         (sizeof(struct ip_vs_service_user))
2029 #define SVCDEST_ARG_LEN         (sizeof(struct ip_vs_service_user) +    \
2030                                  sizeof(struct ip_vs_dest_user))
2031 #define TIMEOUT_ARG_LEN         (sizeof(struct ip_vs_timeout_user))
2032 #define DAEMON_ARG_LEN          (sizeof(struct ip_vs_daemon_user))
2033 #define MAX_ARG_LEN             SVCDEST_ARG_LEN
2034
2035 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2036         [SET_CMDID(IP_VS_SO_SET_ADD)]           = SERVICE_ARG_LEN,
2037         [SET_CMDID(IP_VS_SO_SET_EDIT)]          = SERVICE_ARG_LEN,
2038         [SET_CMDID(IP_VS_SO_SET_DEL)]           = SERVICE_ARG_LEN,
2039         [SET_CMDID(IP_VS_SO_SET_FLUSH)]         = 0,
2040         [SET_CMDID(IP_VS_SO_SET_ADDDEST)]       = SVCDEST_ARG_LEN,
2041         [SET_CMDID(IP_VS_SO_SET_DELDEST)]       = SVCDEST_ARG_LEN,
2042         [SET_CMDID(IP_VS_SO_SET_EDITDEST)]      = SVCDEST_ARG_LEN,
2043         [SET_CMDID(IP_VS_SO_SET_TIMEOUT)]       = TIMEOUT_ARG_LEN,
2044         [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)]   = DAEMON_ARG_LEN,
2045         [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)]    = DAEMON_ARG_LEN,
2046         [SET_CMDID(IP_VS_SO_SET_ZERO)]          = SERVICE_ARG_LEN,
2047 };
2048
2049 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2050                                   struct ip_vs_service_user *usvc_compat)
2051 {
2052         usvc->af                = AF_INET;
2053         usvc->protocol          = usvc_compat->protocol;
2054         usvc->addr.ip           = usvc_compat->addr;
2055         usvc->port              = usvc_compat->port;
2056         usvc->fwmark            = usvc_compat->fwmark;
2057
2058         /* Deep copy of sched_name is not needed here */
2059         usvc->sched_name        = usvc_compat->sched_name;
2060
2061         usvc->flags             = usvc_compat->flags;
2062         usvc->timeout           = usvc_compat->timeout;
2063         usvc->netmask           = usvc_compat->netmask;
2064 }
2065
2066 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2067                                    struct ip_vs_dest_user *udest_compat)
2068 {
2069         udest->addr.ip          = udest_compat->addr;
2070         udest->port             = udest_compat->port;
2071         udest->conn_flags       = udest_compat->conn_flags;
2072         udest->weight           = udest_compat->weight;
2073         udest->u_threshold      = udest_compat->u_threshold;
2074         udest->l_threshold      = udest_compat->l_threshold;
2075 }
2076
2077 static int
2078 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2079 {
2080         int ret;
2081         unsigned char arg[MAX_ARG_LEN];
2082         struct ip_vs_service_user *usvc_compat;
2083         struct ip_vs_service_user_kern usvc;
2084         struct ip_vs_service *svc;
2085         struct ip_vs_dest_user *udest_compat;
2086         struct ip_vs_dest_user_kern udest;
2087
2088         if (!capable(CAP_NET_ADMIN))
2089                 return -EPERM;
2090
2091         if (len != set_arglen[SET_CMDID(cmd)]) {
2092                 IP_VS_ERR("set_ctl: len %u != %u\n",
2093                           len, set_arglen[SET_CMDID(cmd)]);
2094                 return -EINVAL;
2095         }
2096
2097         if (copy_from_user(arg, user, len) != 0)
2098                 return -EFAULT;
2099
2100         /* increase the module use count */
2101         ip_vs_use_count_inc();
2102
2103         if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2104                 ret = -ERESTARTSYS;
2105                 goto out_dec;
2106         }
2107
2108         if (cmd == IP_VS_SO_SET_FLUSH) {
2109                 /* Flush the virtual service */
2110                 ret = ip_vs_flush();
2111                 goto out_unlock;
2112         } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2113                 /* Set timeout values for (tcp tcpfin udp) */
2114                 ret = ip_vs_set_timeout((struct ip_vs_timeout_user *)arg);
2115                 goto out_unlock;
2116         } else if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2117                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2118                 ret = start_sync_thread(dm->state, dm->mcast_ifn, dm->syncid);
2119                 goto out_unlock;
2120         } else if (cmd == IP_VS_SO_SET_STOPDAEMON) {
2121                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2122                 ret = stop_sync_thread(dm->state);
2123                 goto out_unlock;
2124         }
2125
2126         usvc_compat = (struct ip_vs_service_user *)arg;
2127         udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2128
2129         /* We only use the new structs internally, so copy userspace compat
2130          * structs to extended internal versions */
2131         ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2132         ip_vs_copy_udest_compat(&udest, udest_compat);
2133
2134         if (cmd == IP_VS_SO_SET_ZERO) {
2135                 /* if no service address is set, zero counters in all */
2136                 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2137                         ret = ip_vs_zero_all();
2138                         goto out_unlock;
2139                 }
2140         }
2141
2142         /* Check for valid protocol: TCP or UDP, even for fwmark!=0 */
2143         if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP) {
2144                 IP_VS_ERR("set_ctl: invalid protocol: %d %d.%d.%d.%d:%d %s\n",
2145                           usvc.protocol, NIPQUAD(usvc.addr.ip),
2146                           ntohs(usvc.port), usvc.sched_name);
2147                 ret = -EFAULT;
2148                 goto out_unlock;
2149         }
2150
2151         /* Lookup the exact service by <protocol, addr, port> or fwmark */
2152         if (usvc.fwmark == 0)
2153                 svc = __ip_vs_service_get(usvc.af, usvc.protocol,
2154                                           &usvc.addr, usvc.port);
2155         else
2156                 svc = __ip_vs_svc_fwm_get(usvc.af, usvc.fwmark);
2157
2158         if (cmd != IP_VS_SO_SET_ADD
2159             && (svc == NULL || svc->protocol != usvc.protocol)) {
2160                 ret = -ESRCH;
2161                 goto out_unlock;
2162         }
2163
2164         switch (cmd) {
2165         case IP_VS_SO_SET_ADD:
2166                 if (svc != NULL)
2167                         ret = -EEXIST;
2168                 else
2169                         ret = ip_vs_add_service(&usvc, &svc);
2170                 break;
2171         case IP_VS_SO_SET_EDIT:
2172                 ret = ip_vs_edit_service(svc, &usvc);
2173                 break;
2174         case IP_VS_SO_SET_DEL:
2175                 ret = ip_vs_del_service(svc);
2176                 if (!ret)
2177                         goto out_unlock;
2178                 break;
2179         case IP_VS_SO_SET_ZERO:
2180                 ret = ip_vs_zero_service(svc);
2181                 break;
2182         case IP_VS_SO_SET_ADDDEST:
2183                 ret = ip_vs_add_dest(svc, &udest);
2184                 break;
2185         case IP_VS_SO_SET_EDITDEST:
2186                 ret = ip_vs_edit_dest(svc, &udest);
2187                 break;
2188         case IP_VS_SO_SET_DELDEST:
2189                 ret = ip_vs_del_dest(svc, &udest);
2190                 break;
2191         default:
2192                 ret = -EINVAL;
2193         }
2194
2195         if (svc)
2196                 ip_vs_service_put(svc);
2197
2198   out_unlock:
2199         mutex_unlock(&__ip_vs_mutex);
2200   out_dec:
2201         /* decrease the module use count */
2202         ip_vs_use_count_dec();
2203
2204         return ret;
2205 }
2206
2207
2208 static void
2209 ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
2210 {
2211         spin_lock_bh(&src->lock);
2212         memcpy(dst, &src->ustats, sizeof(*dst));
2213         spin_unlock_bh(&src->lock);
2214 }
2215
2216 static void
2217 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2218 {
2219         dst->protocol = src->protocol;
2220         dst->addr = src->addr.ip;
2221         dst->port = src->port;
2222         dst->fwmark = src->fwmark;
2223         strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
2224         dst->flags = src->flags;
2225         dst->timeout = src->timeout / HZ;
2226         dst->netmask = src->netmask;
2227         dst->num_dests = src->num_dests;
2228         ip_vs_copy_stats(&dst->stats, &src->stats);
2229 }
2230
2231 static inline int
2232 __ip_vs_get_service_entries(const struct ip_vs_get_services *get,
2233                             struct ip_vs_get_services __user *uptr)
2234 {
2235         int idx, count=0;
2236         struct ip_vs_service *svc;
2237         struct ip_vs_service_entry entry;
2238         int ret = 0;
2239
2240         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2241                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2242                         /* Only expose IPv4 entries to old interface */
2243                         if (svc->af != AF_INET)
2244                                 continue;
2245
2246                         if (count >= get->num_services)
2247                                 goto out;
2248                         memset(&entry, 0, sizeof(entry));
2249                         ip_vs_copy_service(&entry, svc);
2250                         if (copy_to_user(&uptr->entrytable[count],
2251                                          &entry, sizeof(entry))) {
2252                                 ret = -EFAULT;
2253                                 goto out;
2254                         }
2255                         count++;
2256                 }
2257         }
2258
2259         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2260                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2261                         /* Only expose IPv4 entries to old interface */
2262                         if (svc->af != AF_INET)
2263                                 continue;
2264
2265                         if (count >= get->num_services)
2266                                 goto out;
2267                         memset(&entry, 0, sizeof(entry));
2268                         ip_vs_copy_service(&entry, svc);
2269                         if (copy_to_user(&uptr->entrytable[count],
2270                                          &entry, sizeof(entry))) {
2271                                 ret = -EFAULT;
2272                                 goto out;
2273                         }
2274                         count++;
2275                 }
2276         }
2277   out:
2278         return ret;
2279 }
2280
2281 static inline int
2282 __ip_vs_get_dest_entries(const struct ip_vs_get_dests *get,
2283                          struct ip_vs_get_dests __user *uptr)
2284 {
2285         struct ip_vs_service *svc;
2286         union nf_inet_addr addr = { .ip = get->addr };
2287         int ret = 0;
2288
2289         if (get->fwmark)
2290                 svc = __ip_vs_svc_fwm_get(AF_INET, get->fwmark);
2291         else
2292                 svc = __ip_vs_service_get(AF_INET, get->protocol, &addr,
2293                                           get->port);
2294
2295         if (svc) {
2296                 int count = 0;
2297                 struct ip_vs_dest *dest;
2298                 struct ip_vs_dest_entry entry;
2299
2300                 list_for_each_entry(dest, &svc->destinations, n_list) {
2301                         if (count >= get->num_dests)
2302                                 break;
2303
2304                         entry.addr = dest->addr.ip;
2305                         entry.port = dest->port;
2306                         entry.conn_flags = atomic_read(&dest->conn_flags);
2307                         entry.weight = atomic_read(&dest->weight);
2308                         entry.u_threshold = dest->u_threshold;
2309                         entry.l_threshold = dest->l_threshold;
2310                         entry.activeconns = atomic_read(&dest->activeconns);
2311                         entry.inactconns = atomic_read(&dest->inactconns);
2312                         entry.persistconns = atomic_read(&dest->persistconns);
2313                         ip_vs_copy_stats(&entry.stats, &dest->stats);
2314                         if (copy_to_user(&uptr->entrytable[count],
2315                                          &entry, sizeof(entry))) {
2316                                 ret = -EFAULT;
2317                                 break;
2318                         }
2319                         count++;
2320                 }
2321                 ip_vs_service_put(svc);
2322         } else
2323                 ret = -ESRCH;
2324         return ret;
2325 }
2326
2327 static inline void
2328 __ip_vs_get_timeouts(struct ip_vs_timeout_user *u)
2329 {
2330 #ifdef CONFIG_IP_VS_PROTO_TCP
2331         u->tcp_timeout =
2332                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2333         u->tcp_fin_timeout =
2334                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2335 #endif
2336 #ifdef CONFIG_IP_VS_PROTO_UDP
2337         u->udp_timeout =
2338                 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2339 #endif
2340 }
2341
2342
2343 #define GET_CMDID(cmd)          (cmd - IP_VS_BASE_CTL)
2344 #define GET_INFO_ARG_LEN        (sizeof(struct ip_vs_getinfo))
2345 #define GET_SERVICES_ARG_LEN    (sizeof(struct ip_vs_get_services))
2346 #define GET_SERVICE_ARG_LEN     (sizeof(struct ip_vs_service_entry))
2347 #define GET_DESTS_ARG_LEN       (sizeof(struct ip_vs_get_dests))
2348 #define GET_TIMEOUT_ARG_LEN     (sizeof(struct ip_vs_timeout_user))
2349 #define GET_DAEMON_ARG_LEN      (sizeof(struct ip_vs_daemon_user) * 2)
2350
2351 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2352         [GET_CMDID(IP_VS_SO_GET_VERSION)]       = 64,
2353         [GET_CMDID(IP_VS_SO_GET_INFO)]          = GET_INFO_ARG_LEN,
2354         [GET_CMDID(IP_VS_SO_GET_SERVICES)]      = GET_SERVICES_ARG_LEN,
2355         [GET_CMDID(IP_VS_SO_GET_SERVICE)]       = GET_SERVICE_ARG_LEN,
2356         [GET_CMDID(IP_VS_SO_GET_DESTS)]         = GET_DESTS_ARG_LEN,
2357         [GET_CMDID(IP_VS_SO_GET_TIMEOUT)]       = GET_TIMEOUT_ARG_LEN,
2358         [GET_CMDID(IP_VS_SO_GET_DAEMON)]        = GET_DAEMON_ARG_LEN,
2359 };
2360
2361 static int
2362 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2363 {
2364         unsigned char arg[128];
2365         int ret = 0;
2366
2367         if (!capable(CAP_NET_ADMIN))
2368                 return -EPERM;
2369
2370         if (*len < get_arglen[GET_CMDID(cmd)]) {
2371                 IP_VS_ERR("get_ctl: len %u < %u\n",
2372                           *len, get_arglen[GET_CMDID(cmd)]);
2373                 return -EINVAL;
2374         }
2375
2376         if (copy_from_user(arg, user, get_arglen[GET_CMDID(cmd)]) != 0)
2377                 return -EFAULT;
2378
2379         if (mutex_lock_interruptible(&__ip_vs_mutex))
2380                 return -ERESTARTSYS;
2381
2382         switch (cmd) {
2383         case IP_VS_SO_GET_VERSION:
2384         {
2385                 char buf[64];
2386
2387                 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2388                         NVERSION(IP_VS_VERSION_CODE), IP_VS_CONN_TAB_SIZE);
2389                 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2390                         ret = -EFAULT;
2391                         goto out;
2392                 }
2393                 *len = strlen(buf)+1;
2394         }
2395         break;
2396
2397         case IP_VS_SO_GET_INFO:
2398         {
2399                 struct ip_vs_getinfo info;
2400                 info.version = IP_VS_VERSION_CODE;
2401                 info.size = IP_VS_CONN_TAB_SIZE;
2402                 info.num_services = ip_vs_num_services;
2403                 if (copy_to_user(user, &info, sizeof(info)) != 0)
2404                         ret = -EFAULT;
2405         }
2406         break;
2407
2408         case IP_VS_SO_GET_SERVICES:
2409         {
2410                 struct ip_vs_get_services *get;
2411                 int size;
2412
2413                 get = (struct ip_vs_get_services *)arg;
2414                 size = sizeof(*get) +
2415                         sizeof(struct ip_vs_service_entry) * get->num_services;
2416                 if (*len != size) {
2417                         IP_VS_ERR("length: %u != %u\n", *len, size);
2418                         ret = -EINVAL;
2419                         goto out;
2420                 }
2421                 ret = __ip_vs_get_service_entries(get, user);
2422         }
2423         break;
2424
2425         case IP_VS_SO_GET_SERVICE:
2426         {
2427                 struct ip_vs_service_entry *entry;
2428                 struct ip_vs_service *svc;
2429                 union nf_inet_addr addr;
2430
2431                 entry = (struct ip_vs_service_entry *)arg;
2432                 addr.ip = entry->addr;
2433                 if (entry->fwmark)
2434                         svc = __ip_vs_svc_fwm_get(AF_INET, entry->fwmark);
2435                 else
2436                         svc = __ip_vs_service_get(AF_INET, entry->protocol,
2437                                                   &addr, entry->port);
2438                 if (svc) {
2439                         ip_vs_copy_service(entry, svc);
2440                         if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2441                                 ret = -EFAULT;
2442                         ip_vs_service_put(svc);
2443                 } else
2444                         ret = -ESRCH;
2445         }
2446         break;
2447
2448         case IP_VS_SO_GET_DESTS:
2449         {
2450                 struct ip_vs_get_dests *get;
2451                 int size;
2452
2453                 get = (struct ip_vs_get_dests *)arg;
2454                 size = sizeof(*get) +
2455                         sizeof(struct ip_vs_dest_entry) * get->num_dests;
2456                 if (*len != size) {
2457                         IP_VS_ERR("length: %u != %u\n", *len, size);
2458                         ret = -EINVAL;
2459                         goto out;
2460                 }
2461                 ret = __ip_vs_get_dest_entries(get, user);
2462         }
2463         break;
2464
2465         case IP_VS_SO_GET_TIMEOUT:
2466         {
2467                 struct ip_vs_timeout_user t;
2468
2469                 __ip_vs_get_timeouts(&t);
2470                 if (copy_to_user(user, &t, sizeof(t)) != 0)
2471                         ret = -EFAULT;
2472         }
2473         break;
2474
2475         case IP_VS_SO_GET_DAEMON:
2476         {
2477                 struct ip_vs_daemon_user d[2];
2478
2479                 memset(&d, 0, sizeof(d));
2480                 if (ip_vs_sync_state & IP_VS_STATE_MASTER) {
2481                         d[0].state = IP_VS_STATE_MASTER;
2482                         strlcpy(d[0].mcast_ifn, ip_vs_master_mcast_ifn, sizeof(d[0].mcast_ifn));
2483                         d[0].syncid = ip_vs_master_syncid;
2484                 }
2485                 if (ip_vs_sync_state & IP_VS_STATE_BACKUP) {
2486                         d[1].state = IP_VS_STATE_BACKUP;
2487                         strlcpy(d[1].mcast_ifn, ip_vs_backup_mcast_ifn, sizeof(d[1].mcast_ifn));
2488                         d[1].syncid = ip_vs_backup_syncid;
2489                 }
2490                 if (copy_to_user(user, &d, sizeof(d)) != 0)
2491                         ret = -EFAULT;
2492         }
2493         break;
2494
2495         default:
2496                 ret = -EINVAL;
2497         }
2498
2499   out:
2500         mutex_unlock(&__ip_vs_mutex);
2501         return ret;
2502 }
2503
2504
2505 static struct nf_sockopt_ops ip_vs_sockopts = {
2506         .pf             = PF_INET,
2507         .set_optmin     = IP_VS_BASE_CTL,
2508         .set_optmax     = IP_VS_SO_SET_MAX+1,
2509         .set            = do_ip_vs_set_ctl,
2510         .get_optmin     = IP_VS_BASE_CTL,
2511         .get_optmax     = IP_VS_SO_GET_MAX+1,
2512         .get            = do_ip_vs_get_ctl,
2513         .owner          = THIS_MODULE,
2514 };
2515
2516 /*
2517  * Generic Netlink interface
2518  */
2519
2520 /* IPVS genetlink family */
2521 static struct genl_family ip_vs_genl_family = {
2522         .id             = GENL_ID_GENERATE,
2523         .hdrsize        = 0,
2524         .name           = IPVS_GENL_NAME,
2525         .version        = IPVS_GENL_VERSION,
2526         .maxattr        = IPVS_CMD_MAX,
2527 };
2528
2529 /* Policy used for first-level command attributes */
2530 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2531         [IPVS_CMD_ATTR_SERVICE]         = { .type = NLA_NESTED },
2532         [IPVS_CMD_ATTR_DEST]            = { .type = NLA_NESTED },
2533         [IPVS_CMD_ATTR_DAEMON]          = { .type = NLA_NESTED },
2534         [IPVS_CMD_ATTR_TIMEOUT_TCP]     = { .type = NLA_U32 },
2535         [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2536         [IPVS_CMD_ATTR_TIMEOUT_UDP]     = { .type = NLA_U32 },
2537 };
2538
2539 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2540 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2541         [IPVS_DAEMON_ATTR_STATE]        = { .type = NLA_U32 },
2542         [IPVS_DAEMON_ATTR_MCAST_IFN]    = { .type = NLA_NUL_STRING,
2543                                             .len = IP_VS_IFNAME_MAXLEN },
2544         [IPVS_DAEMON_ATTR_SYNC_ID]      = { .type = NLA_U32 },
2545 };
2546
2547 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2548 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2549         [IPVS_SVC_ATTR_AF]              = { .type = NLA_U16 },
2550         [IPVS_SVC_ATTR_PROTOCOL]        = { .type = NLA_U16 },
2551         [IPVS_SVC_ATTR_ADDR]            = { .type = NLA_BINARY,
2552                                             .len = sizeof(union nf_inet_addr) },
2553         [IPVS_SVC_ATTR_PORT]            = { .type = NLA_U16 },
2554         [IPVS_SVC_ATTR_FWMARK]          = { .type = NLA_U32 },
2555         [IPVS_SVC_ATTR_SCHED_NAME]      = { .type = NLA_NUL_STRING,
2556                                             .len = IP_VS_SCHEDNAME_MAXLEN },
2557         [IPVS_SVC_ATTR_FLAGS]           = { .type = NLA_BINARY,
2558                                             .len = sizeof(struct ip_vs_flags) },
2559         [IPVS_SVC_ATTR_TIMEOUT]         = { .type = NLA_U32 },
2560         [IPVS_SVC_ATTR_NETMASK]         = { .type = NLA_U32 },
2561         [IPVS_SVC_ATTR_STATS]           = { .type = NLA_NESTED },
2562 };
2563
2564 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2565 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2566         [IPVS_DEST_ATTR_ADDR]           = { .type = NLA_BINARY,
2567                                             .len = sizeof(union nf_inet_addr) },
2568         [IPVS_DEST_ATTR_PORT]           = { .type = NLA_U16 },
2569         [IPVS_DEST_ATTR_FWD_METHOD]     = { .type = NLA_U32 },
2570         [IPVS_DEST_ATTR_WEIGHT]         = { .type = NLA_U32 },
2571         [IPVS_DEST_ATTR_U_THRESH]       = { .type = NLA_U32 },
2572         [IPVS_DEST_ATTR_L_THRESH]       = { .type = NLA_U32 },
2573         [IPVS_DEST_ATTR_ACTIVE_CONNS]   = { .type = NLA_U32 },
2574         [IPVS_DEST_ATTR_INACT_CONNS]    = { .type = NLA_U32 },
2575         [IPVS_DEST_ATTR_PERSIST_CONNS]  = { .type = NLA_U32 },
2576         [IPVS_DEST_ATTR_STATS]          = { .type = NLA_NESTED },
2577 };
2578
2579 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2580                                  struct ip_vs_stats *stats)
2581 {
2582         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2583         if (!nl_stats)
2584                 return -EMSGSIZE;
2585
2586         spin_lock_bh(&stats->lock);
2587
2588         NLA_PUT_U32(skb, IPVS_STATS_ATTR_CONNS, stats->ustats.conns);
2589         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPKTS, stats->ustats.inpkts);
2590         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPKTS, stats->ustats.outpkts);
2591         NLA_PUT_U64(skb, IPVS_STATS_ATTR_INBYTES, stats->ustats.inbytes);
2592         NLA_PUT_U64(skb, IPVS_STATS_ATTR_OUTBYTES, stats->ustats.outbytes);
2593         NLA_PUT_U32(skb, IPVS_STATS_ATTR_CPS, stats->ustats.cps);
2594         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPPS, stats->ustats.inpps);
2595         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPPS, stats->ustats.outpps);
2596         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INBPS, stats->ustats.inbps);
2597         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTBPS, stats->ustats.outbps);
2598
2599         spin_unlock_bh(&stats->lock);
2600
2601         nla_nest_end(skb, nl_stats);
2602
2603         return 0;
2604
2605 nla_put_failure:
2606         spin_unlock_bh(&stats->lock);
2607         nla_nest_cancel(skb, nl_stats);
2608         return -EMSGSIZE;
2609 }
2610
2611 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2612                                    struct ip_vs_service *svc)
2613 {
2614         struct nlattr *nl_service;
2615         struct ip_vs_flags flags = { .flags = svc->flags,
2616                                      .mask = ~0 };
2617
2618         nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2619         if (!nl_service)
2620                 return -EMSGSIZE;
2621
2622         NLA_PUT_U16(skb, IPVS_SVC_ATTR_AF, svc->af);
2623
2624         if (svc->fwmark) {
2625                 NLA_PUT_U32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark);
2626         } else {
2627                 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol);
2628                 NLA_PUT(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr);
2629                 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PORT, svc->port);
2630         }
2631
2632         NLA_PUT_STRING(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name);
2633         NLA_PUT(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags);
2634         NLA_PUT_U32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ);
2635         NLA_PUT_U32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask);
2636
2637         if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2638                 goto nla_put_failure;
2639
2640         nla_nest_end(skb, nl_service);
2641
2642         return 0;
2643
2644 nla_put_failure:
2645         nla_nest_cancel(skb, nl_service);
2646         return -EMSGSIZE;
2647 }
2648
2649 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2650                                    struct ip_vs_service *svc,
2651                                    struct netlink_callback *cb)
2652 {
2653         void *hdr;
2654
2655         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2656                           &ip_vs_genl_family, NLM_F_MULTI,
2657                           IPVS_CMD_NEW_SERVICE);
2658         if (!hdr)
2659                 return -EMSGSIZE;
2660
2661         if (ip_vs_genl_fill_service(skb, svc) < 0)
2662                 goto nla_put_failure;
2663
2664         return genlmsg_end(skb, hdr);
2665
2666 nla_put_failure:
2667         genlmsg_cancel(skb, hdr);
2668         return -EMSGSIZE;
2669 }
2670
2671 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2672                                     struct netlink_callback *cb)
2673 {
2674         int idx = 0, i;
2675         int start = cb->args[0];
2676         struct ip_vs_service *svc;
2677
2678         mutex_lock(&__ip_vs_mutex);
2679         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2680                 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2681                         if (++idx <= start)
2682                                 continue;
2683                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2684                                 idx--;
2685                                 goto nla_put_failure;
2686                         }
2687                 }
2688         }
2689
2690         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2691                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2692                         if (++idx <= start)
2693                                 continue;
2694                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2695                                 idx--;
2696                                 goto nla_put_failure;
2697                         }
2698                 }
2699         }
2700
2701 nla_put_failure:
2702         mutex_unlock(&__ip_vs_mutex);
2703         cb->args[0] = idx;
2704
2705         return skb->len;
2706 }
2707
2708 static int ip_vs_genl_parse_service(struct ip_vs_service_user_kern *usvc,
2709                                     struct nlattr *nla, int full_entry)
2710 {
2711         struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2712         struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2713
2714         /* Parse mandatory identifying service fields first */
2715         if (nla == NULL ||
2716             nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2717                 return -EINVAL;
2718
2719         nla_af          = attrs[IPVS_SVC_ATTR_AF];
2720         nla_protocol    = attrs[IPVS_SVC_ATTR_PROTOCOL];
2721         nla_addr        = attrs[IPVS_SVC_ATTR_ADDR];
2722         nla_port        = attrs[IPVS_SVC_ATTR_PORT];
2723         nla_fwmark      = attrs[IPVS_SVC_ATTR_FWMARK];
2724
2725         if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2726                 return -EINVAL;
2727
2728         usvc->af = nla_get_u16(nla_af);
2729 #ifdef CONFIG_IP_VS_IPV6
2730         if (usvc->af != AF_INET && usvc->af != AF_INET6)
2731 #else
2732         if (usvc->af != AF_INET)
2733 #endif
2734                 return -EAFNOSUPPORT;
2735
2736         if (nla_fwmark) {
2737                 usvc->protocol = IPPROTO_TCP;
2738                 usvc->fwmark = nla_get_u32(nla_fwmark);
2739         } else {
2740                 usvc->protocol = nla_get_u16(nla_protocol);
2741                 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
2742                 usvc->port = nla_get_u16(nla_port);
2743                 usvc->fwmark = 0;
2744         }
2745
2746         /* If a full entry was requested, check for the additional fields */
2747         if (full_entry) {
2748                 struct nlattr *nla_sched, *nla_flags, *nla_timeout,
2749                               *nla_netmask;
2750                 struct ip_vs_flags flags;
2751                 struct ip_vs_service *svc;
2752
2753                 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
2754                 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
2755                 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
2756                 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
2757
2758                 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
2759                         return -EINVAL;
2760
2761                 nla_memcpy(&flags, nla_flags, sizeof(flags));
2762
2763                 /* prefill flags from service if it already exists */
2764                 if (usvc->fwmark)
2765                         svc = __ip_vs_svc_fwm_get(usvc->af, usvc->fwmark);
2766                 else
2767                         svc = __ip_vs_service_get(usvc->af, usvc->protocol,
2768                                                   &usvc->addr, usvc->port);
2769                 if (svc) {
2770                         usvc->flags = svc->flags;
2771                         ip_vs_service_put(svc);
2772                 } else
2773                         usvc->flags = 0;
2774
2775                 /* set new flags from userland */
2776                 usvc->flags = (usvc->flags & ~flags.mask) |
2777                               (flags.flags & flags.mask);
2778                 usvc->sched_name = nla_data(nla_sched);
2779                 usvc->timeout = nla_get_u32(nla_timeout);
2780                 usvc->netmask = nla_get_u32(nla_netmask);
2781         }
2782
2783         return 0;
2784 }
2785
2786 static struct ip_vs_service *ip_vs_genl_find_service(struct nlattr *nla)
2787 {
2788         struct ip_vs_service_user_kern usvc;
2789         int ret;
2790
2791         ret = ip_vs_genl_parse_service(&usvc, nla, 0);
2792         if (ret)
2793                 return ERR_PTR(ret);
2794
2795         if (usvc.fwmark)
2796                 return __ip_vs_svc_fwm_get(usvc.af, usvc.fwmark);
2797         else
2798                 return __ip_vs_service_get(usvc.af, usvc.protocol,
2799                                            &usvc.addr, usvc.port);
2800 }
2801
2802 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
2803 {
2804         struct nlattr *nl_dest;
2805
2806         nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
2807         if (!nl_dest)
2808                 return -EMSGSIZE;
2809
2810         NLA_PUT(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr);
2811         NLA_PUT_U16(skb, IPVS_DEST_ATTR_PORT, dest->port);
2812
2813         NLA_PUT_U32(skb, IPVS_DEST_ATTR_FWD_METHOD,
2814                     atomic_read(&dest->conn_flags) & IP_VS_CONN_F_FWD_MASK);
2815         NLA_PUT_U32(skb, IPVS_DEST_ATTR_WEIGHT, atomic_read(&dest->weight));
2816         NLA_PUT_U32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold);
2817         NLA_PUT_U32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold);
2818         NLA_PUT_U32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
2819                     atomic_read(&dest->activeconns));
2820         NLA_PUT_U32(skb, IPVS_DEST_ATTR_INACT_CONNS,
2821                     atomic_read(&dest->inactconns));
2822         NLA_PUT_U32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
2823                     atomic_read(&dest->persistconns));
2824
2825         if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
2826                 goto nla_put_failure;
2827
2828         nla_nest_end(skb, nl_dest);
2829
2830         return 0;
2831
2832 nla_put_failure:
2833         nla_nest_cancel(skb, nl_dest);
2834         return -EMSGSIZE;
2835 }
2836
2837 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
2838                                 struct netlink_callback *cb)
2839 {
2840         void *hdr;
2841
2842         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2843                           &ip_vs_genl_family, NLM_F_MULTI,
2844                           IPVS_CMD_NEW_DEST);
2845         if (!hdr)
2846                 return -EMSGSIZE;
2847
2848         if (ip_vs_genl_fill_dest(skb, dest) < 0)
2849                 goto nla_put_failure;
2850
2851         return genlmsg_end(skb, hdr);
2852
2853 nla_put_failure:
2854         genlmsg_cancel(skb, hdr);
2855         return -EMSGSIZE;
2856 }
2857
2858 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
2859                                  struct netlink_callback *cb)
2860 {
2861         int idx = 0;
2862         int start = cb->args[0];
2863         struct ip_vs_service *svc;
2864         struct ip_vs_dest *dest;
2865         struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
2866
2867         mutex_lock(&__ip_vs_mutex);
2868
2869         /* Try to find the service for which to dump destinations */
2870         if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
2871                         IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
2872                 goto out_err;
2873
2874         svc = ip_vs_genl_find_service(attrs[IPVS_CMD_ATTR_SERVICE]);
2875         if (IS_ERR(svc) || svc == NULL)
2876                 goto out_err;
2877
2878         /* Dump the destinations */
2879         list_for_each_entry(dest, &svc->destinations, n_list) {
2880                 if (++idx <= start)
2881                         continue;
2882                 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
2883                         idx--;
2884                         goto nla_put_failure;
2885                 }
2886         }
2887
2888 nla_put_failure:
2889         cb->args[0] = idx;
2890         ip_vs_service_put(svc);
2891
2892 out_err:
2893         mutex_unlock(&__ip_vs_mutex);
2894
2895         return skb->len;
2896 }
2897
2898 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
2899                                  struct nlattr *nla, int full_entry)
2900 {
2901         struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
2902         struct nlattr *nla_addr, *nla_port;
2903
2904         /* Parse mandatory identifying destination fields first */
2905         if (nla == NULL ||
2906             nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
2907                 return -EINVAL;
2908
2909         nla_addr        = attrs[IPVS_DEST_ATTR_ADDR];
2910         nla_port        = attrs[IPVS_DEST_ATTR_PORT];
2911
2912         if (!(nla_addr && nla_port))
2913                 return -EINVAL;
2914
2915         nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
2916         udest->port = nla_get_u16(nla_port);
2917
2918         /* If a full entry was requested, check for the additional fields */
2919         if (full_entry) {
2920                 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
2921                               *nla_l_thresh;
2922
2923                 nla_fwd         = attrs[IPVS_DEST_ATTR_FWD_METHOD];
2924                 nla_weight      = attrs[IPVS_DEST_ATTR_WEIGHT];
2925                 nla_u_thresh    = attrs[IPVS_DEST_ATTR_U_THRESH];
2926                 nla_l_thresh    = attrs[IPVS_DEST_ATTR_L_THRESH];
2927
2928                 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
2929                         return -EINVAL;
2930
2931                 udest->conn_flags = nla_get_u32(nla_fwd)
2932                                     & IP_VS_CONN_F_FWD_MASK;
2933                 udest->weight = nla_get_u32(nla_weight);
2934                 udest->u_threshold = nla_get_u32(nla_u_thresh);
2935                 udest->l_threshold = nla_get_u32(nla_l_thresh);
2936         }
2937
2938         return 0;
2939 }
2940
2941 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
2942                                   const char *mcast_ifn, __be32 syncid)
2943 {
2944         struct nlattr *nl_daemon;
2945
2946         nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
2947         if (!nl_daemon)
2948                 return -EMSGSIZE;
2949
2950         NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_STATE, state);
2951         NLA_PUT_STRING(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn);
2952         NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid);
2953
2954         nla_nest_end(skb, nl_daemon);
2955
2956         return 0;
2957
2958 nla_put_failure:
2959         nla_nest_cancel(skb, nl_daemon);
2960         return -EMSGSIZE;
2961 }
2962
2963 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
2964                                   const char *mcast_ifn, __be32 syncid,
2965                                   struct netlink_callback *cb)
2966 {
2967         void *hdr;
2968         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2969                           &ip_vs_genl_family, NLM_F_MULTI,
2970                           IPVS_CMD_NEW_DAEMON);
2971         if (!hdr)
2972                 return -EMSGSIZE;
2973
2974         if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
2975                 goto nla_put_failure;
2976
2977         return genlmsg_end(skb, hdr);
2978
2979 nla_put_failure:
2980         genlmsg_cancel(skb, hdr);
2981         return -EMSGSIZE;
2982 }
2983
2984 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
2985                                    struct netlink_callback *cb)
2986 {
2987         mutex_lock(&__ip_vs_mutex);
2988         if ((ip_vs_sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
2989                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
2990                                            ip_vs_master_mcast_ifn,
2991                                            ip_vs_master_syncid, cb) < 0)
2992                         goto nla_put_failure;
2993
2994                 cb->args[0] = 1;
2995         }
2996
2997         if ((ip_vs_sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
2998                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
2999                                            ip_vs_backup_mcast_ifn,
3000                                            ip_vs_backup_syncid, cb) < 0)
3001                         goto nla_put_failure;
3002
3003                 cb->args[1] = 1;
3004         }
3005
3006 nla_put_failure:
3007         mutex_unlock(&__ip_vs_mutex);
3008
3009         return skb->len;
3010 }
3011
3012 static int ip_vs_genl_new_daemon(struct nlattr **attrs)
3013 {
3014         if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3015               attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3016               attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3017                 return -EINVAL;
3018
3019         return start_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3020                                  nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3021                                  nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3022 }
3023
3024 static int ip_vs_genl_del_daemon(struct nlattr **attrs)
3025 {
3026         if (!attrs[IPVS_DAEMON_ATTR_STATE])
3027                 return -EINVAL;
3028
3029         return stop_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3030 }
3031
3032 static int ip_vs_genl_set_config(struct nlattr **attrs)
3033 {
3034         struct ip_vs_timeout_user t;
3035
3036         __ip_vs_get_timeouts(&t);
3037
3038         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3039                 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3040
3041         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3042                 t.tcp_fin_timeout =
3043                         nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3044
3045         if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3046                 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3047
3048         return ip_vs_set_timeout(&t);
3049 }
3050
3051 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3052 {
3053         struct ip_vs_service *svc = NULL;
3054         struct ip_vs_service_user_kern usvc;
3055         struct ip_vs_dest_user_kern udest;
3056         int ret = 0, cmd;
3057         int need_full_svc = 0, need_full_dest = 0;
3058
3059         cmd = info->genlhdr->cmd;
3060
3061         mutex_lock(&__ip_vs_mutex);
3062
3063         if (cmd == IPVS_CMD_FLUSH) {
3064                 ret = ip_vs_flush();
3065                 goto out;
3066         } else if (cmd == IPVS_CMD_SET_CONFIG) {
3067                 ret = ip_vs_genl_set_config(info->attrs);
3068                 goto out;
3069         } else if (cmd == IPVS_CMD_NEW_DAEMON ||
3070                    cmd == IPVS_CMD_DEL_DAEMON) {
3071
3072                 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3073
3074                 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3075                     nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3076                                      info->attrs[IPVS_CMD_ATTR_DAEMON],
3077                                      ip_vs_daemon_policy)) {
3078                         ret = -EINVAL;
3079                         goto out;
3080                 }
3081
3082                 if (cmd == IPVS_CMD_NEW_DAEMON)
3083                         ret = ip_vs_genl_new_daemon(daemon_attrs);
3084                 else
3085                         ret = ip_vs_genl_del_daemon(daemon_attrs);
3086                 goto out;
3087         } else if (cmd == IPVS_CMD_ZERO &&
3088                    !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3089                 ret = ip_vs_zero_all();
3090                 goto out;
3091         }
3092
3093         /* All following commands require a service argument, so check if we
3094          * received a valid one. We need a full service specification when
3095          * adding / editing a service. Only identifying members otherwise. */
3096         if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3097                 need_full_svc = 1;
3098
3099         ret = ip_vs_genl_parse_service(&usvc,
3100                                        info->attrs[IPVS_CMD_ATTR_SERVICE],
3101                                        need_full_svc);
3102         if (ret)
3103                 goto out;
3104
3105         /* Lookup the exact service by <protocol, addr, port> or fwmark */
3106         if (usvc.fwmark == 0)
3107                 svc = __ip_vs_service_get(usvc.af, usvc.protocol,
3108                                           &usvc.addr, usvc.port);
3109         else
3110                 svc = __ip_vs_svc_fwm_get(usvc.af, usvc.fwmark);
3111
3112         /* Unless we're adding a new service, the service must already exist */
3113         if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3114                 ret = -ESRCH;
3115                 goto out;
3116         }
3117
3118         /* Destination commands require a valid destination argument. For
3119          * adding / editing a destination, we need a full destination
3120          * specification. */
3121         if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3122             cmd == IPVS_CMD_DEL_DEST) {
3123                 if (cmd != IPVS_CMD_DEL_DEST)
3124                         need_full_dest = 1;
3125
3126                 ret = ip_vs_genl_parse_dest(&udest,
3127                                             info->attrs[IPVS_CMD_ATTR_DEST],
3128                                             need_full_dest);
3129                 if (ret)
3130                         goto out;
3131         }
3132
3133         switch (cmd) {
3134         case IPVS_CMD_NEW_SERVICE:
3135                 if (svc == NULL)
3136                         ret = ip_vs_add_service(&usvc, &svc);
3137                 else
3138                         ret = -EEXIST;
3139                 break;
3140         case IPVS_CMD_SET_SERVICE:
3141                 ret = ip_vs_edit_service(svc, &usvc);
3142                 break;
3143         case IPVS_CMD_DEL_SERVICE:
3144                 ret = ip_vs_del_service(svc);
3145                 break;
3146         case IPVS_CMD_NEW_DEST:
3147                 ret = ip_vs_add_dest(svc, &udest);
3148                 break;
3149         case IPVS_CMD_SET_DEST:
3150                 ret = ip_vs_edit_dest(svc, &udest);
3151                 break;
3152         case IPVS_CMD_DEL_DEST:
3153                 ret = ip_vs_del_dest(svc, &udest);
3154                 break;
3155         case IPVS_CMD_ZERO:
3156                 ret = ip_vs_zero_service(svc);
3157                 break;
3158         default:
3159                 ret = -EINVAL;
3160         }
3161
3162 out:
3163         if (svc)
3164                 ip_vs_service_put(svc);
3165         mutex_unlock(&__ip_vs_mutex);
3166
3167         return ret;
3168 }
3169
3170 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3171 {
3172         struct sk_buff *msg;
3173         void *reply;
3174         int ret, cmd, reply_cmd;
3175
3176         cmd = info->genlhdr->cmd;
3177
3178         if (cmd == IPVS_CMD_GET_SERVICE)
3179                 reply_cmd = IPVS_CMD_NEW_SERVICE;
3180         else if (cmd == IPVS_CMD_GET_INFO)
3181                 reply_cmd = IPVS_CMD_SET_INFO;
3182         else if (cmd == IPVS_CMD_GET_CONFIG)
3183                 reply_cmd = IPVS_CMD_SET_CONFIG;
3184         else {
3185                 IP_VS_ERR("unknown Generic Netlink command\n");
3186                 return -EINVAL;
3187         }
3188
3189         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3190         if (!msg)
3191                 return -ENOMEM;
3192
3193         mutex_lock(&__ip_vs_mutex);
3194
3195         reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3196         if (reply == NULL)
3197                 goto nla_put_failure;
3198
3199         switch (cmd) {
3200         case IPVS_CMD_GET_SERVICE:
3201         {
3202                 struct ip_vs_service *svc;
3203
3204                 svc = ip_vs_genl_find_service(info->attrs[IPVS_CMD_ATTR_SERVICE]);
3205                 if (IS_ERR(svc)) {
3206                         ret = PTR_ERR(svc);
3207                         goto out_err;
3208                 } else if (svc) {
3209                         ret = ip_vs_genl_fill_service(msg, svc);
3210                         ip_vs_service_put(svc);
3211                         if (ret)
3212                                 goto nla_put_failure;
3213                 } else {
3214                         ret = -ESRCH;
3215                         goto out_err;
3216                 }
3217
3218                 break;
3219         }
3220
3221         case IPVS_CMD_GET_CONFIG:
3222         {
3223                 struct ip_vs_timeout_user t;
3224
3225                 __ip_vs_get_timeouts(&t);
3226 #ifdef CONFIG_IP_VS_PROTO_TCP
3227                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP, t.tcp_timeout);
3228                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3229                             t.tcp_fin_timeout);
3230 #endif
3231 #ifdef CONFIG_IP_VS_PROTO_UDP
3232                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout);
3233 #endif
3234
3235                 break;
3236         }
3237
3238         case IPVS_CMD_GET_INFO:
3239                 NLA_PUT_U32(msg, IPVS_INFO_ATTR_VERSION, IP_VS_VERSION_CODE);
3240                 NLA_PUT_U32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3241                             IP_VS_CONN_TAB_SIZE);
3242                 break;
3243         }
3244
3245         genlmsg_end(msg, reply);
3246         ret = genlmsg_unicast(msg, info->snd_pid);
3247         goto out;
3248
3249 nla_put_failure:
3250         IP_VS_ERR("not enough space in Netlink message\n");
3251         ret = -EMSGSIZE;
3252
3253 out_err:
3254         nlmsg_free(msg);
3255 out:
3256         mutex_unlock(&__ip_vs_mutex);
3257
3258         return ret;
3259 }
3260
3261
3262 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3263         {
3264                 .cmd    = IPVS_CMD_NEW_SERVICE,
3265                 .flags  = GENL_ADMIN_PERM,
3266                 .policy = ip_vs_cmd_policy,
3267                 .doit   = ip_vs_genl_set_cmd,
3268         },
3269         {
3270                 .cmd    = IPVS_CMD_SET_SERVICE,
3271                 .flags  = GENL_ADMIN_PERM,
3272                 .policy = ip_vs_cmd_policy,
3273                 .doit   = ip_vs_genl_set_cmd,
3274         },
3275         {
3276                 .cmd    = IPVS_CMD_DEL_SERVICE,
3277                 .flags  = GENL_ADMIN_PERM,
3278                 .policy = ip_vs_cmd_policy,
3279                 .doit   = ip_vs_genl_set_cmd,
3280         },
3281         {
3282                 .cmd    = IPVS_CMD_GET_SERVICE,
3283                 .flags  = GENL_ADMIN_PERM,
3284                 .doit   = ip_vs_genl_get_cmd,
3285                 .dumpit = ip_vs_genl_dump_services,
3286                 .policy = ip_vs_cmd_policy,
3287         },
3288         {
3289                 .cmd    = IPVS_CMD_NEW_DEST,
3290                 .flags  = GENL_ADMIN_PERM,
3291                 .policy = ip_vs_cmd_policy,
3292                 .doit   = ip_vs_genl_set_cmd,
3293         },
3294         {
3295                 .cmd    = IPVS_CMD_SET_DEST,
3296                 .flags  = GENL_ADMIN_PERM,
3297                 .policy = ip_vs_cmd_policy,
3298                 .doit   = ip_vs_genl_set_cmd,
3299         },
3300         {
3301                 .cmd    = IPVS_CMD_DEL_DEST,
3302                 .flags  = GENL_ADMIN_PERM,
3303                 .policy = ip_vs_cmd_policy,
3304                 .doit   = ip_vs_genl_set_cmd,
3305         },
3306         {
3307                 .cmd    = IPVS_CMD_GET_DEST,
3308                 .flags  = GENL_ADMIN_PERM,
3309                 .policy = ip_vs_cmd_policy,
3310                 .dumpit = ip_vs_genl_dump_dests,
3311         },
3312         {
3313                 .cmd    = IPVS_CMD_NEW_DAEMON,
3314                 .flags  = GENL_ADMIN_PERM,
3315                 .policy = ip_vs_cmd_policy,
3316                 .doit   = ip_vs_genl_set_cmd,
3317         },
3318         {
3319                 .cmd    = IPVS_CMD_DEL_DAEMON,
3320                 .flags  = GENL_ADMIN_PERM,
3321                 .policy = ip_vs_cmd_policy,
3322                 .doit   = ip_vs_genl_set_cmd,
3323         },
3324         {
3325                 .cmd    = IPVS_CMD_GET_DAEMON,
3326                 .flags  = GENL_ADMIN_PERM,
3327                 .dumpit = ip_vs_genl_dump_daemons,
3328         },
3329         {
3330                 .cmd    = IPVS_CMD_SET_CONFIG,
3331                 .flags  = GENL_ADMIN_PERM,
3332                 .policy = ip_vs_cmd_policy,
3333                 .doit   = ip_vs_genl_set_cmd,
3334         },
3335         {
3336                 .cmd    = IPVS_CMD_GET_CONFIG,
3337                 .flags  = GENL_ADMIN_PERM,
3338                 .doit   = ip_vs_genl_get_cmd,
3339         },
3340         {
3341                 .cmd    = IPVS_CMD_GET_INFO,
3342                 .flags  = GENL_ADMIN_PERM,
3343                 .doit   = ip_vs_genl_get_cmd,
3344         },
3345         {
3346                 .cmd    = IPVS_CMD_ZERO,
3347                 .flags  = GENL_ADMIN_PERM,
3348                 .policy = ip_vs_cmd_policy,
3349                 .doit   = ip_vs_genl_set_cmd,
3350         },
3351         {
3352                 .cmd    = IPVS_CMD_FLUSH,
3353                 .flags  = GENL_ADMIN_PERM,
3354                 .doit   = ip_vs_genl_set_cmd,
3355         },
3356 };
3357
3358 static int __init ip_vs_genl_register(void)
3359 {
3360         int ret, i;
3361
3362         ret = genl_register_family(&ip_vs_genl_family);
3363         if (ret)
3364                 return ret;
3365
3366         for (i = 0; i < ARRAY_SIZE(ip_vs_genl_ops); i++) {
3367                 ret = genl_register_ops(&ip_vs_genl_family, &ip_vs_genl_ops[i]);
3368                 if (ret)
3369                         goto err_out;
3370         }
3371         return 0;
3372
3373 err_out:
3374         genl_unregister_family(&ip_vs_genl_family);
3375         return ret;
3376 }
3377
3378 static void ip_vs_genl_unregister(void)
3379 {
3380         genl_unregister_family(&ip_vs_genl_family);
3381 }
3382
3383 /* End of Generic Netlink interface definitions */
3384
3385
3386 int __init ip_vs_control_init(void)
3387 {
3388         int ret;
3389         int idx;
3390
3391         EnterFunction(2);
3392
3393         ret = nf_register_sockopt(&ip_vs_sockopts);
3394         if (ret) {
3395                 IP_VS_ERR("cannot register sockopt.\n");
3396                 return ret;
3397         }
3398
3399         ret = ip_vs_genl_register();
3400         if (ret) {
3401                 IP_VS_ERR("cannot register Generic Netlink interface.\n");
3402                 nf_unregister_sockopt(&ip_vs_sockopts);
3403                 return ret;
3404         }
3405
3406         proc_net_fops_create(&init_net, "ip_vs", 0, &ip_vs_info_fops);
3407         proc_net_fops_create(&init_net, "ip_vs_stats",0, &ip_vs_stats_fops);
3408
3409         sysctl_header = register_sysctl_paths(net_vs_ctl_path, vs_vars);
3410
3411         /* Initialize ip_vs_svc_table, ip_vs_svc_fwm_table, ip_vs_rtable */
3412         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++)  {
3413                 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3414                 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3415         }
3416         for(idx = 0; idx < IP_VS_RTAB_SIZE; idx++)  {
3417                 INIT_LIST_HEAD(&ip_vs_rtable[idx]);
3418         }
3419
3420         ip_vs_new_estimator(&ip_vs_stats);
3421
3422         /* Hook the defense timer */
3423         schedule_delayed_work(&defense_work, DEFENSE_TIMER_PERIOD);
3424
3425         LeaveFunction(2);
3426         return 0;
3427 }
3428
3429
3430 void ip_vs_control_cleanup(void)
3431 {
3432         EnterFunction(2);
3433         ip_vs_trash_cleanup();
3434         cancel_rearming_delayed_work(&defense_work);
3435         cancel_work_sync(&defense_work.work);
3436         ip_vs_kill_estimator(&ip_vs_stats);
3437         unregister_sysctl_table(sysctl_header);
3438         proc_net_remove(&init_net, "ip_vs_stats");
3439         proc_net_remove(&init_net, "ip_vs");
3440         ip_vs_genl_unregister();
3441         nf_unregister_sockopt(&ip_vs_sockopts);
3442         LeaveFunction(2);
3443 }