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 && (u->netmask < 1 || u->netmask > 128)) {
1172                 ret = -EINVAL;
1173                 goto out_err;
1174         }
1175 #endif
1176
1177         svc = kzalloc(sizeof(struct ip_vs_service), GFP_ATOMIC);
1178         if (svc == NULL) {
1179                 IP_VS_DBG(1, "ip_vs_add_service: kmalloc failed.\n");
1180                 ret = -ENOMEM;
1181                 goto out_err;
1182         }
1183
1184         /* I'm the first user of the service */
1185         atomic_set(&svc->usecnt, 1);
1186         atomic_set(&svc->refcnt, 0);
1187
1188         svc->af = u->af;
1189         svc->protocol = u->protocol;
1190         ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1191         svc->port = u->port;
1192         svc->fwmark = u->fwmark;
1193         svc->flags = u->flags;
1194         svc->timeout = u->timeout * HZ;
1195         svc->netmask = u->netmask;
1196
1197         INIT_LIST_HEAD(&svc->destinations);
1198         rwlock_init(&svc->sched_lock);
1199         spin_lock_init(&svc->stats.lock);
1200
1201         /* Bind the scheduler */
1202         ret = ip_vs_bind_scheduler(svc, sched);
1203         if (ret)
1204                 goto out_err;
1205         sched = NULL;
1206
1207         /* Update the virtual service counters */
1208         if (svc->port == FTPPORT)
1209                 atomic_inc(&ip_vs_ftpsvc_counter);
1210         else if (svc->port == 0)
1211                 atomic_inc(&ip_vs_nullsvc_counter);
1212
1213         ip_vs_new_estimator(&svc->stats);
1214
1215         /* Count only IPv4 services for old get/setsockopt interface */
1216         if (svc->af == AF_INET)
1217                 ip_vs_num_services++;
1218
1219         /* Hash the service into the service table */
1220         write_lock_bh(&__ip_vs_svc_lock);
1221         ip_vs_svc_hash(svc);
1222         write_unlock_bh(&__ip_vs_svc_lock);
1223
1224         *svc_p = svc;
1225         return 0;
1226
1227   out_err:
1228         if (svc != NULL) {
1229                 if (svc->scheduler)
1230                         ip_vs_unbind_scheduler(svc);
1231                 if (svc->inc) {
1232                         local_bh_disable();
1233                         ip_vs_app_inc_put(svc->inc);
1234                         local_bh_enable();
1235                 }
1236                 kfree(svc);
1237         }
1238         ip_vs_scheduler_put(sched);
1239
1240   out_mod_dec:
1241         /* decrease the module use count */
1242         ip_vs_use_count_dec();
1243
1244         return ret;
1245 }
1246
1247
1248 /*
1249  *      Edit a service and bind it with a new scheduler
1250  */
1251 static int
1252 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1253 {
1254         struct ip_vs_scheduler *sched, *old_sched;
1255         int ret = 0;
1256
1257         /*
1258          * Lookup the scheduler, by 'u->sched_name'
1259          */
1260         sched = ip_vs_scheduler_get(u->sched_name);
1261         if (sched == NULL) {
1262                 IP_VS_INFO("Scheduler module ip_vs_%s not found\n",
1263                            u->sched_name);
1264                 return -ENOENT;
1265         }
1266         old_sched = sched;
1267
1268 #ifdef CONFIG_IP_VS_IPV6
1269         if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1270                 ret = -EINVAL;
1271                 goto out;
1272         }
1273 #endif
1274
1275         write_lock_bh(&__ip_vs_svc_lock);
1276
1277         /*
1278          * Wait until all other svc users go away.
1279          */
1280         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
1281
1282         /*
1283          * Set the flags and timeout value
1284          */
1285         svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1286         svc->timeout = u->timeout * HZ;
1287         svc->netmask = u->netmask;
1288
1289         old_sched = svc->scheduler;
1290         if (sched != old_sched) {
1291                 /*
1292                  * Unbind the old scheduler
1293                  */
1294                 if ((ret = ip_vs_unbind_scheduler(svc))) {
1295                         old_sched = sched;
1296                         goto out_unlock;
1297                 }
1298
1299                 /*
1300                  * Bind the new scheduler
1301                  */
1302                 if ((ret = ip_vs_bind_scheduler(svc, sched))) {
1303                         /*
1304                          * If ip_vs_bind_scheduler fails, restore the old
1305                          * scheduler.
1306                          * The main reason of failure is out of memory.
1307                          *
1308                          * The question is if the old scheduler can be
1309                          * restored all the time. TODO: if it cannot be
1310                          * restored some time, we must delete the service,
1311                          * otherwise the system may crash.
1312                          */
1313                         ip_vs_bind_scheduler(svc, old_sched);
1314                         old_sched = sched;
1315                         goto out_unlock;
1316                 }
1317         }
1318
1319   out_unlock:
1320         write_unlock_bh(&__ip_vs_svc_lock);
1321 #ifdef CONFIG_IP_VS_IPV6
1322   out:
1323 #endif
1324
1325         if (old_sched)
1326                 ip_vs_scheduler_put(old_sched);
1327
1328         return ret;
1329 }
1330
1331
1332 /*
1333  *      Delete a service from the service list
1334  *      - The service must be unlinked, unlocked and not referenced!
1335  *      - We are called under _bh lock
1336  */
1337 static void __ip_vs_del_service(struct ip_vs_service *svc)
1338 {
1339         struct ip_vs_dest *dest, *nxt;
1340         struct ip_vs_scheduler *old_sched;
1341
1342         /* Count only IPv4 services for old get/setsockopt interface */
1343         if (svc->af == AF_INET)
1344                 ip_vs_num_services--;
1345
1346         ip_vs_kill_estimator(&svc->stats);
1347
1348         /* Unbind scheduler */
1349         old_sched = svc->scheduler;
1350         ip_vs_unbind_scheduler(svc);
1351         if (old_sched)
1352                 ip_vs_scheduler_put(old_sched);
1353
1354         /* Unbind app inc */
1355         if (svc->inc) {
1356                 ip_vs_app_inc_put(svc->inc);
1357                 svc->inc = NULL;
1358         }
1359
1360         /*
1361          *    Unlink the whole destination list
1362          */
1363         list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1364                 __ip_vs_unlink_dest(svc, dest, 0);
1365                 __ip_vs_del_dest(dest);
1366         }
1367
1368         /*
1369          *    Update the virtual service counters
1370          */
1371         if (svc->port == FTPPORT)
1372                 atomic_dec(&ip_vs_ftpsvc_counter);
1373         else if (svc->port == 0)
1374                 atomic_dec(&ip_vs_nullsvc_counter);
1375
1376         /*
1377          *    Free the service if nobody refers to it
1378          */
1379         if (atomic_read(&svc->refcnt) == 0)
1380                 kfree(svc);
1381
1382         /* decrease the module use count */
1383         ip_vs_use_count_dec();
1384 }
1385
1386 /*
1387  *      Delete a service from the service list
1388  */
1389 static int ip_vs_del_service(struct ip_vs_service *svc)
1390 {
1391         if (svc == NULL)
1392                 return -EEXIST;
1393
1394         /*
1395          * Unhash it from the service table
1396          */
1397         write_lock_bh(&__ip_vs_svc_lock);
1398
1399         ip_vs_svc_unhash(svc);
1400
1401         /*
1402          * Wait until all the svc users go away.
1403          */
1404         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
1405
1406         __ip_vs_del_service(svc);
1407
1408         write_unlock_bh(&__ip_vs_svc_lock);
1409
1410         return 0;
1411 }
1412
1413
1414 /*
1415  *      Flush all the virtual services
1416  */
1417 static int ip_vs_flush(void)
1418 {
1419         int idx;
1420         struct ip_vs_service *svc, *nxt;
1421
1422         /*
1423          * Flush the service table hashed by <protocol,addr,port>
1424          */
1425         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1426                 list_for_each_entry_safe(svc, nxt, &ip_vs_svc_table[idx], s_list) {
1427                         write_lock_bh(&__ip_vs_svc_lock);
1428                         ip_vs_svc_unhash(svc);
1429                         /*
1430                          * Wait until all the svc users go away.
1431                          */
1432                         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1433                         __ip_vs_del_service(svc);
1434                         write_unlock_bh(&__ip_vs_svc_lock);
1435                 }
1436         }
1437
1438         /*
1439          * Flush the service table hashed by fwmark
1440          */
1441         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1442                 list_for_each_entry_safe(svc, nxt,
1443                                          &ip_vs_svc_fwm_table[idx], f_list) {
1444                         write_lock_bh(&__ip_vs_svc_lock);
1445                         ip_vs_svc_unhash(svc);
1446                         /*
1447                          * Wait until all the svc users go away.
1448                          */
1449                         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1450                         __ip_vs_del_service(svc);
1451                         write_unlock_bh(&__ip_vs_svc_lock);
1452                 }
1453         }
1454
1455         return 0;
1456 }
1457
1458
1459 /*
1460  *      Zero counters in a service or all services
1461  */
1462 static int ip_vs_zero_service(struct ip_vs_service *svc)
1463 {
1464         struct ip_vs_dest *dest;
1465
1466         write_lock_bh(&__ip_vs_svc_lock);
1467         list_for_each_entry(dest, &svc->destinations, n_list) {
1468                 ip_vs_zero_stats(&dest->stats);
1469         }
1470         ip_vs_zero_stats(&svc->stats);
1471         write_unlock_bh(&__ip_vs_svc_lock);
1472         return 0;
1473 }
1474
1475 static int ip_vs_zero_all(void)
1476 {
1477         int idx;
1478         struct ip_vs_service *svc;
1479
1480         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1481                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1482                         ip_vs_zero_service(svc);
1483                 }
1484         }
1485
1486         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1487                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1488                         ip_vs_zero_service(svc);
1489                 }
1490         }
1491
1492         ip_vs_zero_stats(&ip_vs_stats);
1493         return 0;
1494 }
1495
1496
1497 static int
1498 proc_do_defense_mode(ctl_table *table, int write, struct file * filp,
1499                      void __user *buffer, size_t *lenp, loff_t *ppos)
1500 {
1501         int *valp = table->data;
1502         int val = *valp;
1503         int rc;
1504
1505         rc = proc_dointvec(table, write, filp, buffer, lenp, ppos);
1506         if (write && (*valp != val)) {
1507                 if ((*valp < 0) || (*valp > 3)) {
1508                         /* Restore the correct value */
1509                         *valp = val;
1510                 } else {
1511                         update_defense_level();
1512                 }
1513         }
1514         return rc;
1515 }
1516
1517
1518 static int
1519 proc_do_sync_threshold(ctl_table *table, int write, struct file *filp,
1520                        void __user *buffer, size_t *lenp, loff_t *ppos)
1521 {
1522         int *valp = table->data;
1523         int val[2];
1524         int rc;
1525
1526         /* backup the value first */
1527         memcpy(val, valp, sizeof(val));
1528
1529         rc = proc_dointvec(table, write, filp, buffer, lenp, ppos);
1530         if (write && (valp[0] < 0 || valp[1] < 0 || valp[0] >= valp[1])) {
1531                 /* Restore the correct value */
1532                 memcpy(valp, val, sizeof(val));
1533         }
1534         return rc;
1535 }
1536
1537
1538 /*
1539  *      IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1540  */
1541
1542 static struct ctl_table vs_vars[] = {
1543         {
1544                 .procname       = "amemthresh",
1545                 .data           = &sysctl_ip_vs_amemthresh,
1546                 .maxlen         = sizeof(int),
1547                 .mode           = 0644,
1548                 .proc_handler   = proc_dointvec,
1549         },
1550 #ifdef CONFIG_IP_VS_DEBUG
1551         {
1552                 .procname       = "debug_level",
1553                 .data           = &sysctl_ip_vs_debug_level,
1554                 .maxlen         = sizeof(int),
1555                 .mode           = 0644,
1556                 .proc_handler   = proc_dointvec,
1557         },
1558 #endif
1559         {
1560                 .procname       = "am_droprate",
1561                 .data           = &sysctl_ip_vs_am_droprate,
1562                 .maxlen         = sizeof(int),
1563                 .mode           = 0644,
1564                 .proc_handler   = proc_dointvec,
1565         },
1566         {
1567                 .procname       = "drop_entry",
1568                 .data           = &sysctl_ip_vs_drop_entry,
1569                 .maxlen         = sizeof(int),
1570                 .mode           = 0644,
1571                 .proc_handler   = proc_do_defense_mode,
1572         },
1573         {
1574                 .procname       = "drop_packet",
1575                 .data           = &sysctl_ip_vs_drop_packet,
1576                 .maxlen         = sizeof(int),
1577                 .mode           = 0644,
1578                 .proc_handler   = proc_do_defense_mode,
1579         },
1580         {
1581                 .procname       = "secure_tcp",
1582                 .data           = &sysctl_ip_vs_secure_tcp,
1583                 .maxlen         = sizeof(int),
1584                 .mode           = 0644,
1585                 .proc_handler   = proc_do_defense_mode,
1586         },
1587 #if 0
1588         {
1589                 .procname       = "timeout_established",
1590                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1591                 .maxlen         = sizeof(int),
1592                 .mode           = 0644,
1593                 .proc_handler   = proc_dointvec_jiffies,
1594         },
1595         {
1596                 .procname       = "timeout_synsent",
1597                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1598                 .maxlen         = sizeof(int),
1599                 .mode           = 0644,
1600                 .proc_handler   = proc_dointvec_jiffies,
1601         },
1602         {
1603                 .procname       = "timeout_synrecv",
1604                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1605                 .maxlen         = sizeof(int),
1606                 .mode           = 0644,
1607                 .proc_handler   = proc_dointvec_jiffies,
1608         },
1609         {
1610                 .procname       = "timeout_finwait",
1611                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1612                 .maxlen         = sizeof(int),
1613                 .mode           = 0644,
1614                 .proc_handler   = proc_dointvec_jiffies,
1615         },
1616         {
1617                 .procname       = "timeout_timewait",
1618                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1619                 .maxlen         = sizeof(int),
1620                 .mode           = 0644,
1621                 .proc_handler   = proc_dointvec_jiffies,
1622         },
1623         {
1624                 .procname       = "timeout_close",
1625                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1626                 .maxlen         = sizeof(int),
1627                 .mode           = 0644,
1628                 .proc_handler   = proc_dointvec_jiffies,
1629         },
1630         {
1631                 .procname       = "timeout_closewait",
1632                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1633                 .maxlen         = sizeof(int),
1634                 .mode           = 0644,
1635                 .proc_handler   = proc_dointvec_jiffies,
1636         },
1637         {
1638                 .procname       = "timeout_lastack",
1639                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1640                 .maxlen         = sizeof(int),
1641                 .mode           = 0644,
1642                 .proc_handler   = proc_dointvec_jiffies,
1643         },
1644         {
1645                 .procname       = "timeout_listen",
1646                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1647                 .maxlen         = sizeof(int),
1648                 .mode           = 0644,
1649                 .proc_handler   = proc_dointvec_jiffies,
1650         },
1651         {
1652                 .procname       = "timeout_synack",
1653                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1654                 .maxlen         = sizeof(int),
1655                 .mode           = 0644,
1656                 .proc_handler   = proc_dointvec_jiffies,
1657         },
1658         {
1659                 .procname       = "timeout_udp",
1660                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1661                 .maxlen         = sizeof(int),
1662                 .mode           = 0644,
1663                 .proc_handler   = proc_dointvec_jiffies,
1664         },
1665         {
1666                 .procname       = "timeout_icmp",
1667                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1668                 .maxlen         = sizeof(int),
1669                 .mode           = 0644,
1670                 .proc_handler   = proc_dointvec_jiffies,
1671         },
1672 #endif
1673         {
1674                 .procname       = "cache_bypass",
1675                 .data           = &sysctl_ip_vs_cache_bypass,
1676                 .maxlen         = sizeof(int),
1677                 .mode           = 0644,
1678                 .proc_handler   = proc_dointvec,
1679         },
1680         {
1681                 .procname       = "expire_nodest_conn",
1682                 .data           = &sysctl_ip_vs_expire_nodest_conn,
1683                 .maxlen         = sizeof(int),
1684                 .mode           = 0644,
1685                 .proc_handler   = proc_dointvec,
1686         },
1687         {
1688                 .procname       = "expire_quiescent_template",
1689                 .data           = &sysctl_ip_vs_expire_quiescent_template,
1690                 .maxlen         = sizeof(int),
1691                 .mode           = 0644,
1692                 .proc_handler   = proc_dointvec,
1693         },
1694         {
1695                 .procname       = "sync_threshold",
1696                 .data           = &sysctl_ip_vs_sync_threshold,
1697                 .maxlen         = sizeof(sysctl_ip_vs_sync_threshold),
1698                 .mode           = 0644,
1699                 .proc_handler   = proc_do_sync_threshold,
1700         },
1701         {
1702                 .procname       = "nat_icmp_send",
1703                 .data           = &sysctl_ip_vs_nat_icmp_send,
1704                 .maxlen         = sizeof(int),
1705                 .mode           = 0644,
1706                 .proc_handler   = proc_dointvec,
1707         },
1708         { .ctl_name = 0 }
1709 };
1710
1711 const struct ctl_path net_vs_ctl_path[] = {
1712         { .procname = "net", .ctl_name = CTL_NET, },
1713         { .procname = "ipv4", .ctl_name = NET_IPV4, },
1714         { .procname = "vs", },
1715         { }
1716 };
1717 EXPORT_SYMBOL_GPL(net_vs_ctl_path);
1718
1719 static struct ctl_table_header * sysctl_header;
1720
1721 #ifdef CONFIG_PROC_FS
1722
1723 struct ip_vs_iter {
1724         struct list_head *table;
1725         int bucket;
1726 };
1727
1728 /*
1729  *      Write the contents of the VS rule table to a PROCfs file.
1730  *      (It is kept just for backward compatibility)
1731  */
1732 static inline const char *ip_vs_fwd_name(unsigned flags)
1733 {
1734         switch (flags & IP_VS_CONN_F_FWD_MASK) {
1735         case IP_VS_CONN_F_LOCALNODE:
1736                 return "Local";
1737         case IP_VS_CONN_F_TUNNEL:
1738                 return "Tunnel";
1739         case IP_VS_CONN_F_DROUTE:
1740                 return "Route";
1741         default:
1742                 return "Masq";
1743         }
1744 }
1745
1746
1747 /* Get the Nth entry in the two lists */
1748 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1749 {
1750         struct ip_vs_iter *iter = seq->private;
1751         int idx;
1752         struct ip_vs_service *svc;
1753
1754         /* look in hash by protocol */
1755         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1756                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1757                         if (pos-- == 0){
1758                                 iter->table = ip_vs_svc_table;
1759                                 iter->bucket = idx;
1760                                 return svc;
1761                         }
1762                 }
1763         }
1764
1765         /* keep looking in fwmark */
1766         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1767                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1768                         if (pos-- == 0) {
1769                                 iter->table = ip_vs_svc_fwm_table;
1770                                 iter->bucket = idx;
1771                                 return svc;
1772                         }
1773                 }
1774         }
1775
1776         return NULL;
1777 }
1778
1779 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1780 __acquires(__ip_vs_svc_lock)
1781 {
1782
1783         read_lock_bh(&__ip_vs_svc_lock);
1784         return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1785 }
1786
1787
1788 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1789 {
1790         struct list_head *e;
1791         struct ip_vs_iter *iter;
1792         struct ip_vs_service *svc;
1793
1794         ++*pos;
1795         if (v == SEQ_START_TOKEN)
1796                 return ip_vs_info_array(seq,0);
1797
1798         svc = v;
1799         iter = seq->private;
1800
1801         if (iter->table == ip_vs_svc_table) {
1802                 /* next service in table hashed by protocol */
1803                 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
1804                         return list_entry(e, struct ip_vs_service, s_list);
1805
1806
1807                 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1808                         list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
1809                                             s_list) {
1810                                 return svc;
1811                         }
1812                 }
1813
1814                 iter->table = ip_vs_svc_fwm_table;
1815                 iter->bucket = -1;
1816                 goto scan_fwmark;
1817         }
1818
1819         /* next service in hashed by fwmark */
1820         if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
1821                 return list_entry(e, struct ip_vs_service, f_list);
1822
1823  scan_fwmark:
1824         while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1825                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
1826                                     f_list)
1827                         return svc;
1828         }
1829
1830         return NULL;
1831 }
1832
1833 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1834 __releases(__ip_vs_svc_lock)
1835 {
1836         read_unlock_bh(&__ip_vs_svc_lock);
1837 }
1838
1839
1840 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1841 {
1842         if (v == SEQ_START_TOKEN) {
1843                 seq_printf(seq,
1844                         "IP Virtual Server version %d.%d.%d (size=%d)\n",
1845                         NVERSION(IP_VS_VERSION_CODE), IP_VS_CONN_TAB_SIZE);
1846                 seq_puts(seq,
1847                          "Prot LocalAddress:Port Scheduler Flags\n");
1848                 seq_puts(seq,
1849                          "  -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1850         } else {
1851                 const struct ip_vs_service *svc = v;
1852                 const struct ip_vs_iter *iter = seq->private;
1853                 const struct ip_vs_dest *dest;
1854
1855                 if (iter->table == ip_vs_svc_table) {
1856 #ifdef CONFIG_IP_VS_IPV6
1857                         if (svc->af == AF_INET6)
1858                                 seq_printf(seq, "%s  [%pI6]:%04X %s ",
1859                                            ip_vs_proto_name(svc->protocol),
1860                                            &svc->addr.in6,
1861                                            ntohs(svc->port),
1862                                            svc->scheduler->name);
1863                         else
1864 #endif
1865                                 seq_printf(seq, "%s  %08X:%04X %s ",
1866                                            ip_vs_proto_name(svc->protocol),
1867                                            ntohl(svc->addr.ip),
1868                                            ntohs(svc->port),
1869                                            svc->scheduler->name);
1870                 } else {
1871                         seq_printf(seq, "FWM  %08X %s ",
1872                                    svc->fwmark, svc->scheduler->name);
1873                 }
1874
1875                 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
1876                         seq_printf(seq, "persistent %d %08X\n",
1877                                 svc->timeout,
1878                                 ntohl(svc->netmask));
1879                 else
1880                         seq_putc(seq, '\n');
1881
1882                 list_for_each_entry(dest, &svc->destinations, n_list) {
1883 #ifdef CONFIG_IP_VS_IPV6
1884                         if (dest->af == AF_INET6)
1885                                 seq_printf(seq,
1886                                            "  -> [%pI6]:%04X"
1887                                            "      %-7s %-6d %-10d %-10d\n",
1888                                            &dest->addr.in6,
1889                                            ntohs(dest->port),
1890                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1891                                            atomic_read(&dest->weight),
1892                                            atomic_read(&dest->activeconns),
1893                                            atomic_read(&dest->inactconns));
1894                         else
1895 #endif
1896                                 seq_printf(seq,
1897                                            "  -> %08X:%04X      "
1898                                            "%-7s %-6d %-10d %-10d\n",
1899                                            ntohl(dest->addr.ip),
1900                                            ntohs(dest->port),
1901                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1902                                            atomic_read(&dest->weight),
1903                                            atomic_read(&dest->activeconns),
1904                                            atomic_read(&dest->inactconns));
1905
1906                 }
1907         }
1908         return 0;
1909 }
1910
1911 static const struct seq_operations ip_vs_info_seq_ops = {
1912         .start = ip_vs_info_seq_start,
1913         .next  = ip_vs_info_seq_next,
1914         .stop  = ip_vs_info_seq_stop,
1915         .show  = ip_vs_info_seq_show,
1916 };
1917
1918 static int ip_vs_info_open(struct inode *inode, struct file *file)
1919 {
1920         return seq_open_private(file, &ip_vs_info_seq_ops,
1921                         sizeof(struct ip_vs_iter));
1922 }
1923
1924 static const struct file_operations ip_vs_info_fops = {
1925         .owner   = THIS_MODULE,
1926         .open    = ip_vs_info_open,
1927         .read    = seq_read,
1928         .llseek  = seq_lseek,
1929         .release = seq_release_private,
1930 };
1931
1932 #endif
1933
1934 struct ip_vs_stats ip_vs_stats = {
1935         .lock = __SPIN_LOCK_UNLOCKED(ip_vs_stats.lock),
1936 };
1937
1938 #ifdef CONFIG_PROC_FS
1939 static int ip_vs_stats_show(struct seq_file *seq, void *v)
1940 {
1941
1942 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
1943         seq_puts(seq,
1944                  "   Total Incoming Outgoing         Incoming         Outgoing\n");
1945         seq_printf(seq,
1946                    "   Conns  Packets  Packets            Bytes            Bytes\n");
1947
1948         spin_lock_bh(&ip_vs_stats.lock);
1949         seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", ip_vs_stats.ustats.conns,
1950                    ip_vs_stats.ustats.inpkts, ip_vs_stats.ustats.outpkts,
1951                    (unsigned long long) ip_vs_stats.ustats.inbytes,
1952                    (unsigned long long) ip_vs_stats.ustats.outbytes);
1953
1954 /*                 01234567 01234567 01234567 0123456701234567 0123456701234567 */
1955         seq_puts(seq,
1956                    " Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
1957         seq_printf(seq,"%8X %8X %8X %16X %16X\n",
1958                         ip_vs_stats.ustats.cps,
1959                         ip_vs_stats.ustats.inpps,
1960                         ip_vs_stats.ustats.outpps,
1961                         ip_vs_stats.ustats.inbps,
1962                         ip_vs_stats.ustats.outbps);
1963         spin_unlock_bh(&ip_vs_stats.lock);
1964
1965         return 0;
1966 }
1967
1968 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
1969 {
1970         return single_open(file, ip_vs_stats_show, NULL);
1971 }
1972
1973 static const struct file_operations ip_vs_stats_fops = {
1974         .owner = THIS_MODULE,
1975         .open = ip_vs_stats_seq_open,
1976         .read = seq_read,
1977         .llseek = seq_lseek,
1978         .release = single_release,
1979 };
1980
1981 #endif
1982
1983 /*
1984  *      Set timeout values for tcp tcpfin udp in the timeout_table.
1985  */
1986 static int ip_vs_set_timeout(struct ip_vs_timeout_user *u)
1987 {
1988         IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
1989                   u->tcp_timeout,
1990                   u->tcp_fin_timeout,
1991                   u->udp_timeout);
1992
1993 #ifdef CONFIG_IP_VS_PROTO_TCP
1994         if (u->tcp_timeout) {
1995                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED]
1996                         = u->tcp_timeout * HZ;
1997         }
1998
1999         if (u->tcp_fin_timeout) {
2000                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT]
2001                         = u->tcp_fin_timeout * HZ;
2002         }
2003 #endif
2004
2005 #ifdef CONFIG_IP_VS_PROTO_UDP
2006         if (u->udp_timeout) {
2007                 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL]
2008                         = u->udp_timeout * HZ;
2009         }
2010 #endif
2011         return 0;
2012 }
2013
2014
2015 #define SET_CMDID(cmd)          (cmd - IP_VS_BASE_CTL)
2016 #define SERVICE_ARG_LEN         (sizeof(struct ip_vs_service_user))
2017 #define SVCDEST_ARG_LEN         (sizeof(struct ip_vs_service_user) +    \
2018                                  sizeof(struct ip_vs_dest_user))
2019 #define TIMEOUT_ARG_LEN         (sizeof(struct ip_vs_timeout_user))
2020 #define DAEMON_ARG_LEN          (sizeof(struct ip_vs_daemon_user))
2021 #define MAX_ARG_LEN             SVCDEST_ARG_LEN
2022
2023 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2024         [SET_CMDID(IP_VS_SO_SET_ADD)]           = SERVICE_ARG_LEN,
2025         [SET_CMDID(IP_VS_SO_SET_EDIT)]          = SERVICE_ARG_LEN,
2026         [SET_CMDID(IP_VS_SO_SET_DEL)]           = SERVICE_ARG_LEN,
2027         [SET_CMDID(IP_VS_SO_SET_FLUSH)]         = 0,
2028         [SET_CMDID(IP_VS_SO_SET_ADDDEST)]       = SVCDEST_ARG_LEN,
2029         [SET_CMDID(IP_VS_SO_SET_DELDEST)]       = SVCDEST_ARG_LEN,
2030         [SET_CMDID(IP_VS_SO_SET_EDITDEST)]      = SVCDEST_ARG_LEN,
2031         [SET_CMDID(IP_VS_SO_SET_TIMEOUT)]       = TIMEOUT_ARG_LEN,
2032         [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)]   = DAEMON_ARG_LEN,
2033         [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)]    = DAEMON_ARG_LEN,
2034         [SET_CMDID(IP_VS_SO_SET_ZERO)]          = SERVICE_ARG_LEN,
2035 };
2036
2037 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2038                                   struct ip_vs_service_user *usvc_compat)
2039 {
2040         usvc->af                = AF_INET;
2041         usvc->protocol          = usvc_compat->protocol;
2042         usvc->addr.ip           = usvc_compat->addr;
2043         usvc->port              = usvc_compat->port;
2044         usvc->fwmark            = usvc_compat->fwmark;
2045
2046         /* Deep copy of sched_name is not needed here */
2047         usvc->sched_name        = usvc_compat->sched_name;
2048
2049         usvc->flags             = usvc_compat->flags;
2050         usvc->timeout           = usvc_compat->timeout;
2051         usvc->netmask           = usvc_compat->netmask;
2052 }
2053
2054 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2055                                    struct ip_vs_dest_user *udest_compat)
2056 {
2057         udest->addr.ip          = udest_compat->addr;
2058         udest->port             = udest_compat->port;
2059         udest->conn_flags       = udest_compat->conn_flags;
2060         udest->weight           = udest_compat->weight;
2061         udest->u_threshold      = udest_compat->u_threshold;
2062         udest->l_threshold      = udest_compat->l_threshold;
2063 }
2064
2065 static int
2066 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2067 {
2068         int ret;
2069         unsigned char arg[MAX_ARG_LEN];
2070         struct ip_vs_service_user *usvc_compat;
2071         struct ip_vs_service_user_kern usvc;
2072         struct ip_vs_service *svc;
2073         struct ip_vs_dest_user *udest_compat;
2074         struct ip_vs_dest_user_kern udest;
2075
2076         if (!capable(CAP_NET_ADMIN))
2077                 return -EPERM;
2078
2079         if (len != set_arglen[SET_CMDID(cmd)]) {
2080                 IP_VS_ERR("set_ctl: len %u != %u\n",
2081                           len, set_arglen[SET_CMDID(cmd)]);
2082                 return -EINVAL;
2083         }
2084
2085         if (copy_from_user(arg, user, len) != 0)
2086                 return -EFAULT;
2087
2088         /* increase the module use count */
2089         ip_vs_use_count_inc();
2090
2091         if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2092                 ret = -ERESTARTSYS;
2093                 goto out_dec;
2094         }
2095
2096         if (cmd == IP_VS_SO_SET_FLUSH) {
2097                 /* Flush the virtual service */
2098                 ret = ip_vs_flush();
2099                 goto out_unlock;
2100         } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2101                 /* Set timeout values for (tcp tcpfin udp) */
2102                 ret = ip_vs_set_timeout((struct ip_vs_timeout_user *)arg);
2103                 goto out_unlock;
2104         } else if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2105                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2106                 ret = start_sync_thread(dm->state, dm->mcast_ifn, dm->syncid);
2107                 goto out_unlock;
2108         } else if (cmd == IP_VS_SO_SET_STOPDAEMON) {
2109                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2110                 ret = stop_sync_thread(dm->state);
2111                 goto out_unlock;
2112         }
2113
2114         usvc_compat = (struct ip_vs_service_user *)arg;
2115         udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2116
2117         /* We only use the new structs internally, so copy userspace compat
2118          * structs to extended internal versions */
2119         ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2120         ip_vs_copy_udest_compat(&udest, udest_compat);
2121
2122         if (cmd == IP_VS_SO_SET_ZERO) {
2123                 /* if no service address is set, zero counters in all */
2124                 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2125                         ret = ip_vs_zero_all();
2126                         goto out_unlock;
2127                 }
2128         }
2129
2130         /* Check for valid protocol: TCP or UDP, even for fwmark!=0 */
2131         if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP) {
2132                 IP_VS_ERR("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2133                           usvc.protocol, &usvc.addr.ip,
2134                           ntohs(usvc.port), usvc.sched_name);
2135                 ret = -EFAULT;
2136                 goto out_unlock;
2137         }
2138
2139         /* Lookup the exact service by <protocol, addr, port> or fwmark */
2140         if (usvc.fwmark == 0)
2141                 svc = __ip_vs_service_get(usvc.af, usvc.protocol,
2142                                           &usvc.addr, usvc.port);
2143         else
2144                 svc = __ip_vs_svc_fwm_get(usvc.af, usvc.fwmark);
2145
2146         if (cmd != IP_VS_SO_SET_ADD
2147             && (svc == NULL || svc->protocol != usvc.protocol)) {
2148                 ret = -ESRCH;
2149                 goto out_unlock;
2150         }
2151
2152         switch (cmd) {
2153         case IP_VS_SO_SET_ADD:
2154                 if (svc != NULL)
2155                         ret = -EEXIST;
2156                 else
2157                         ret = ip_vs_add_service(&usvc, &svc);
2158                 break;
2159         case IP_VS_SO_SET_EDIT:
2160                 ret = ip_vs_edit_service(svc, &usvc);
2161                 break;
2162         case IP_VS_SO_SET_DEL:
2163                 ret = ip_vs_del_service(svc);
2164                 if (!ret)
2165                         goto out_unlock;
2166                 break;
2167         case IP_VS_SO_SET_ZERO:
2168                 ret = ip_vs_zero_service(svc);
2169                 break;
2170         case IP_VS_SO_SET_ADDDEST:
2171                 ret = ip_vs_add_dest(svc, &udest);
2172                 break;
2173         case IP_VS_SO_SET_EDITDEST:
2174                 ret = ip_vs_edit_dest(svc, &udest);
2175                 break;
2176         case IP_VS_SO_SET_DELDEST:
2177                 ret = ip_vs_del_dest(svc, &udest);
2178                 break;
2179         default:
2180                 ret = -EINVAL;
2181         }
2182
2183         if (svc)
2184                 ip_vs_service_put(svc);
2185
2186   out_unlock:
2187         mutex_unlock(&__ip_vs_mutex);
2188   out_dec:
2189         /* decrease the module use count */
2190         ip_vs_use_count_dec();
2191
2192         return ret;
2193 }
2194
2195
2196 static void
2197 ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
2198 {
2199         spin_lock_bh(&src->lock);
2200         memcpy(dst, &src->ustats, sizeof(*dst));
2201         spin_unlock_bh(&src->lock);
2202 }
2203
2204 static void
2205 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2206 {
2207         dst->protocol = src->protocol;
2208         dst->addr = src->addr.ip;
2209         dst->port = src->port;
2210         dst->fwmark = src->fwmark;
2211         strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
2212         dst->flags = src->flags;
2213         dst->timeout = src->timeout / HZ;
2214         dst->netmask = src->netmask;
2215         dst->num_dests = src->num_dests;
2216         ip_vs_copy_stats(&dst->stats, &src->stats);
2217 }
2218
2219 static inline int
2220 __ip_vs_get_service_entries(const struct ip_vs_get_services *get,
2221                             struct ip_vs_get_services __user *uptr)
2222 {
2223         int idx, count=0;
2224         struct ip_vs_service *svc;
2225         struct ip_vs_service_entry entry;
2226         int ret = 0;
2227
2228         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2229                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2230                         /* Only expose IPv4 entries to old interface */
2231                         if (svc->af != AF_INET)
2232                                 continue;
2233
2234                         if (count >= get->num_services)
2235                                 goto out;
2236                         memset(&entry, 0, sizeof(entry));
2237                         ip_vs_copy_service(&entry, svc);
2238                         if (copy_to_user(&uptr->entrytable[count],
2239                                          &entry, sizeof(entry))) {
2240                                 ret = -EFAULT;
2241                                 goto out;
2242                         }
2243                         count++;
2244                 }
2245         }
2246
2247         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2248                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2249                         /* Only expose IPv4 entries to old interface */
2250                         if (svc->af != AF_INET)
2251                                 continue;
2252
2253                         if (count >= get->num_services)
2254                                 goto out;
2255                         memset(&entry, 0, sizeof(entry));
2256                         ip_vs_copy_service(&entry, svc);
2257                         if (copy_to_user(&uptr->entrytable[count],
2258                                          &entry, sizeof(entry))) {
2259                                 ret = -EFAULT;
2260                                 goto out;
2261                         }
2262                         count++;
2263                 }
2264         }
2265   out:
2266         return ret;
2267 }
2268
2269 static inline int
2270 __ip_vs_get_dest_entries(const struct ip_vs_get_dests *get,
2271                          struct ip_vs_get_dests __user *uptr)
2272 {
2273         struct ip_vs_service *svc;
2274         union nf_inet_addr addr = { .ip = get->addr };
2275         int ret = 0;
2276
2277         if (get->fwmark)
2278                 svc = __ip_vs_svc_fwm_get(AF_INET, get->fwmark);
2279         else
2280                 svc = __ip_vs_service_get(AF_INET, get->protocol, &addr,
2281                                           get->port);
2282
2283         if (svc) {
2284                 int count = 0;
2285                 struct ip_vs_dest *dest;
2286                 struct ip_vs_dest_entry entry;
2287
2288                 list_for_each_entry(dest, &svc->destinations, n_list) {
2289                         if (count >= get->num_dests)
2290                                 break;
2291
2292                         entry.addr = dest->addr.ip;
2293                         entry.port = dest->port;
2294                         entry.conn_flags = atomic_read(&dest->conn_flags);
2295                         entry.weight = atomic_read(&dest->weight);
2296                         entry.u_threshold = dest->u_threshold;
2297                         entry.l_threshold = dest->l_threshold;
2298                         entry.activeconns = atomic_read(&dest->activeconns);
2299                         entry.inactconns = atomic_read(&dest->inactconns);
2300                         entry.persistconns = atomic_read(&dest->persistconns);
2301                         ip_vs_copy_stats(&entry.stats, &dest->stats);
2302                         if (copy_to_user(&uptr->entrytable[count],
2303                                          &entry, sizeof(entry))) {
2304                                 ret = -EFAULT;
2305                                 break;
2306                         }
2307                         count++;
2308                 }
2309                 ip_vs_service_put(svc);
2310         } else
2311                 ret = -ESRCH;
2312         return ret;
2313 }
2314
2315 static inline void
2316 __ip_vs_get_timeouts(struct ip_vs_timeout_user *u)
2317 {
2318 #ifdef CONFIG_IP_VS_PROTO_TCP
2319         u->tcp_timeout =
2320                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2321         u->tcp_fin_timeout =
2322                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2323 #endif
2324 #ifdef CONFIG_IP_VS_PROTO_UDP
2325         u->udp_timeout =
2326                 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2327 #endif
2328 }
2329
2330
2331 #define GET_CMDID(cmd)          (cmd - IP_VS_BASE_CTL)
2332 #define GET_INFO_ARG_LEN        (sizeof(struct ip_vs_getinfo))
2333 #define GET_SERVICES_ARG_LEN    (sizeof(struct ip_vs_get_services))
2334 #define GET_SERVICE_ARG_LEN     (sizeof(struct ip_vs_service_entry))
2335 #define GET_DESTS_ARG_LEN       (sizeof(struct ip_vs_get_dests))
2336 #define GET_TIMEOUT_ARG_LEN     (sizeof(struct ip_vs_timeout_user))
2337 #define GET_DAEMON_ARG_LEN      (sizeof(struct ip_vs_daemon_user) * 2)
2338
2339 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2340         [GET_CMDID(IP_VS_SO_GET_VERSION)]       = 64,
2341         [GET_CMDID(IP_VS_SO_GET_INFO)]          = GET_INFO_ARG_LEN,
2342         [GET_CMDID(IP_VS_SO_GET_SERVICES)]      = GET_SERVICES_ARG_LEN,
2343         [GET_CMDID(IP_VS_SO_GET_SERVICE)]       = GET_SERVICE_ARG_LEN,
2344         [GET_CMDID(IP_VS_SO_GET_DESTS)]         = GET_DESTS_ARG_LEN,
2345         [GET_CMDID(IP_VS_SO_GET_TIMEOUT)]       = GET_TIMEOUT_ARG_LEN,
2346         [GET_CMDID(IP_VS_SO_GET_DAEMON)]        = GET_DAEMON_ARG_LEN,
2347 };
2348
2349 static int
2350 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2351 {
2352         unsigned char arg[128];
2353         int ret = 0;
2354
2355         if (!capable(CAP_NET_ADMIN))
2356                 return -EPERM;
2357
2358         if (*len < get_arglen[GET_CMDID(cmd)]) {
2359                 IP_VS_ERR("get_ctl: len %u < %u\n",
2360                           *len, get_arglen[GET_CMDID(cmd)]);
2361                 return -EINVAL;
2362         }
2363
2364         if (copy_from_user(arg, user, get_arglen[GET_CMDID(cmd)]) != 0)
2365                 return -EFAULT;
2366
2367         if (mutex_lock_interruptible(&__ip_vs_mutex))
2368                 return -ERESTARTSYS;
2369
2370         switch (cmd) {
2371         case IP_VS_SO_GET_VERSION:
2372         {
2373                 char buf[64];
2374
2375                 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2376                         NVERSION(IP_VS_VERSION_CODE), IP_VS_CONN_TAB_SIZE);
2377                 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2378                         ret = -EFAULT;
2379                         goto out;
2380                 }
2381                 *len = strlen(buf)+1;
2382         }
2383         break;
2384
2385         case IP_VS_SO_GET_INFO:
2386         {
2387                 struct ip_vs_getinfo info;
2388                 info.version = IP_VS_VERSION_CODE;
2389                 info.size = IP_VS_CONN_TAB_SIZE;
2390                 info.num_services = ip_vs_num_services;
2391                 if (copy_to_user(user, &info, sizeof(info)) != 0)
2392                         ret = -EFAULT;
2393         }
2394         break;
2395
2396         case IP_VS_SO_GET_SERVICES:
2397         {
2398                 struct ip_vs_get_services *get;
2399                 int size;
2400
2401                 get = (struct ip_vs_get_services *)arg;
2402                 size = sizeof(*get) +
2403                         sizeof(struct ip_vs_service_entry) * get->num_services;
2404                 if (*len != size) {
2405                         IP_VS_ERR("length: %u != %u\n", *len, size);
2406                         ret = -EINVAL;
2407                         goto out;
2408                 }
2409                 ret = __ip_vs_get_service_entries(get, user);
2410         }
2411         break;
2412
2413         case IP_VS_SO_GET_SERVICE:
2414         {
2415                 struct ip_vs_service_entry *entry;
2416                 struct ip_vs_service *svc;
2417                 union nf_inet_addr addr;
2418
2419                 entry = (struct ip_vs_service_entry *)arg;
2420                 addr.ip = entry->addr;
2421                 if (entry->fwmark)
2422                         svc = __ip_vs_svc_fwm_get(AF_INET, entry->fwmark);
2423                 else
2424                         svc = __ip_vs_service_get(AF_INET, entry->protocol,
2425                                                   &addr, entry->port);
2426                 if (svc) {
2427                         ip_vs_copy_service(entry, svc);
2428                         if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2429                                 ret = -EFAULT;
2430                         ip_vs_service_put(svc);
2431                 } else
2432                         ret = -ESRCH;
2433         }
2434         break;
2435
2436         case IP_VS_SO_GET_DESTS:
2437         {
2438                 struct ip_vs_get_dests *get;
2439                 int size;
2440
2441                 get = (struct ip_vs_get_dests *)arg;
2442                 size = sizeof(*get) +
2443                         sizeof(struct ip_vs_dest_entry) * get->num_dests;
2444                 if (*len != size) {
2445                         IP_VS_ERR("length: %u != %u\n", *len, size);
2446                         ret = -EINVAL;
2447                         goto out;
2448                 }
2449                 ret = __ip_vs_get_dest_entries(get, user);
2450         }
2451         break;
2452
2453         case IP_VS_SO_GET_TIMEOUT:
2454         {
2455                 struct ip_vs_timeout_user t;
2456
2457                 __ip_vs_get_timeouts(&t);
2458                 if (copy_to_user(user, &t, sizeof(t)) != 0)
2459                         ret = -EFAULT;
2460         }
2461         break;
2462
2463         case IP_VS_SO_GET_DAEMON:
2464         {
2465                 struct ip_vs_daemon_user d[2];
2466
2467                 memset(&d, 0, sizeof(d));
2468                 if (ip_vs_sync_state & IP_VS_STATE_MASTER) {
2469                         d[0].state = IP_VS_STATE_MASTER;
2470                         strlcpy(d[0].mcast_ifn, ip_vs_master_mcast_ifn, sizeof(d[0].mcast_ifn));
2471                         d[0].syncid = ip_vs_master_syncid;
2472                 }
2473                 if (ip_vs_sync_state & IP_VS_STATE_BACKUP) {
2474                         d[1].state = IP_VS_STATE_BACKUP;
2475                         strlcpy(d[1].mcast_ifn, ip_vs_backup_mcast_ifn, sizeof(d[1].mcast_ifn));
2476                         d[1].syncid = ip_vs_backup_syncid;
2477                 }
2478                 if (copy_to_user(user, &d, sizeof(d)) != 0)
2479                         ret = -EFAULT;
2480         }
2481         break;
2482
2483         default:
2484                 ret = -EINVAL;
2485         }
2486
2487   out:
2488         mutex_unlock(&__ip_vs_mutex);
2489         return ret;
2490 }
2491
2492
2493 static struct nf_sockopt_ops ip_vs_sockopts = {
2494         .pf             = PF_INET,
2495         .set_optmin     = IP_VS_BASE_CTL,
2496         .set_optmax     = IP_VS_SO_SET_MAX+1,
2497         .set            = do_ip_vs_set_ctl,
2498         .get_optmin     = IP_VS_BASE_CTL,
2499         .get_optmax     = IP_VS_SO_GET_MAX+1,
2500         .get            = do_ip_vs_get_ctl,
2501         .owner          = THIS_MODULE,
2502 };
2503
2504 /*
2505  * Generic Netlink interface
2506  */
2507
2508 /* IPVS genetlink family */
2509 static struct genl_family ip_vs_genl_family = {
2510         .id             = GENL_ID_GENERATE,
2511         .hdrsize        = 0,
2512         .name           = IPVS_GENL_NAME,
2513         .version        = IPVS_GENL_VERSION,
2514         .maxattr        = IPVS_CMD_MAX,
2515 };
2516
2517 /* Policy used for first-level command attributes */
2518 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2519         [IPVS_CMD_ATTR_SERVICE]         = { .type = NLA_NESTED },
2520         [IPVS_CMD_ATTR_DEST]            = { .type = NLA_NESTED },
2521         [IPVS_CMD_ATTR_DAEMON]          = { .type = NLA_NESTED },
2522         [IPVS_CMD_ATTR_TIMEOUT_TCP]     = { .type = NLA_U32 },
2523         [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2524         [IPVS_CMD_ATTR_TIMEOUT_UDP]     = { .type = NLA_U32 },
2525 };
2526
2527 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2528 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2529         [IPVS_DAEMON_ATTR_STATE]        = { .type = NLA_U32 },
2530         [IPVS_DAEMON_ATTR_MCAST_IFN]    = { .type = NLA_NUL_STRING,
2531                                             .len = IP_VS_IFNAME_MAXLEN },
2532         [IPVS_DAEMON_ATTR_SYNC_ID]      = { .type = NLA_U32 },
2533 };
2534
2535 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2536 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2537         [IPVS_SVC_ATTR_AF]              = { .type = NLA_U16 },
2538         [IPVS_SVC_ATTR_PROTOCOL]        = { .type = NLA_U16 },
2539         [IPVS_SVC_ATTR_ADDR]            = { .type = NLA_BINARY,
2540                                             .len = sizeof(union nf_inet_addr) },
2541         [IPVS_SVC_ATTR_PORT]            = { .type = NLA_U16 },
2542         [IPVS_SVC_ATTR_FWMARK]          = { .type = NLA_U32 },
2543         [IPVS_SVC_ATTR_SCHED_NAME]      = { .type = NLA_NUL_STRING,
2544                                             .len = IP_VS_SCHEDNAME_MAXLEN },
2545         [IPVS_SVC_ATTR_FLAGS]           = { .type = NLA_BINARY,
2546                                             .len = sizeof(struct ip_vs_flags) },
2547         [IPVS_SVC_ATTR_TIMEOUT]         = { .type = NLA_U32 },
2548         [IPVS_SVC_ATTR_NETMASK]         = { .type = NLA_U32 },
2549         [IPVS_SVC_ATTR_STATS]           = { .type = NLA_NESTED },
2550 };
2551
2552 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2553 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2554         [IPVS_DEST_ATTR_ADDR]           = { .type = NLA_BINARY,
2555                                             .len = sizeof(union nf_inet_addr) },
2556         [IPVS_DEST_ATTR_PORT]           = { .type = NLA_U16 },
2557         [IPVS_DEST_ATTR_FWD_METHOD]     = { .type = NLA_U32 },
2558         [IPVS_DEST_ATTR_WEIGHT]         = { .type = NLA_U32 },
2559         [IPVS_DEST_ATTR_U_THRESH]       = { .type = NLA_U32 },
2560         [IPVS_DEST_ATTR_L_THRESH]       = { .type = NLA_U32 },
2561         [IPVS_DEST_ATTR_ACTIVE_CONNS]   = { .type = NLA_U32 },
2562         [IPVS_DEST_ATTR_INACT_CONNS]    = { .type = NLA_U32 },
2563         [IPVS_DEST_ATTR_PERSIST_CONNS]  = { .type = NLA_U32 },
2564         [IPVS_DEST_ATTR_STATS]          = { .type = NLA_NESTED },
2565 };
2566
2567 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2568                                  struct ip_vs_stats *stats)
2569 {
2570         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2571         if (!nl_stats)
2572                 return -EMSGSIZE;
2573
2574         spin_lock_bh(&stats->lock);
2575
2576         NLA_PUT_U32(skb, IPVS_STATS_ATTR_CONNS, stats->ustats.conns);
2577         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPKTS, stats->ustats.inpkts);
2578         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPKTS, stats->ustats.outpkts);
2579         NLA_PUT_U64(skb, IPVS_STATS_ATTR_INBYTES, stats->ustats.inbytes);
2580         NLA_PUT_U64(skb, IPVS_STATS_ATTR_OUTBYTES, stats->ustats.outbytes);
2581         NLA_PUT_U32(skb, IPVS_STATS_ATTR_CPS, stats->ustats.cps);
2582         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPPS, stats->ustats.inpps);
2583         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPPS, stats->ustats.outpps);
2584         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INBPS, stats->ustats.inbps);
2585         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTBPS, stats->ustats.outbps);
2586
2587         spin_unlock_bh(&stats->lock);
2588
2589         nla_nest_end(skb, nl_stats);
2590
2591         return 0;
2592
2593 nla_put_failure:
2594         spin_unlock_bh(&stats->lock);
2595         nla_nest_cancel(skb, nl_stats);
2596         return -EMSGSIZE;
2597 }
2598
2599 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2600                                    struct ip_vs_service *svc)
2601 {
2602         struct nlattr *nl_service;
2603         struct ip_vs_flags flags = { .flags = svc->flags,
2604                                      .mask = ~0 };
2605
2606         nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2607         if (!nl_service)
2608                 return -EMSGSIZE;
2609
2610         NLA_PUT_U16(skb, IPVS_SVC_ATTR_AF, svc->af);
2611
2612         if (svc->fwmark) {
2613                 NLA_PUT_U32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark);
2614         } else {
2615                 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol);
2616                 NLA_PUT(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr);
2617                 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PORT, svc->port);
2618         }
2619
2620         NLA_PUT_STRING(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name);
2621         NLA_PUT(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags);
2622         NLA_PUT_U32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ);
2623         NLA_PUT_U32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask);
2624
2625         if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2626                 goto nla_put_failure;
2627
2628         nla_nest_end(skb, nl_service);
2629
2630         return 0;
2631
2632 nla_put_failure:
2633         nla_nest_cancel(skb, nl_service);
2634         return -EMSGSIZE;
2635 }
2636
2637 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2638                                    struct ip_vs_service *svc,
2639                                    struct netlink_callback *cb)
2640 {
2641         void *hdr;
2642
2643         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2644                           &ip_vs_genl_family, NLM_F_MULTI,
2645                           IPVS_CMD_NEW_SERVICE);
2646         if (!hdr)
2647                 return -EMSGSIZE;
2648
2649         if (ip_vs_genl_fill_service(skb, svc) < 0)
2650                 goto nla_put_failure;
2651
2652         return genlmsg_end(skb, hdr);
2653
2654 nla_put_failure:
2655         genlmsg_cancel(skb, hdr);
2656         return -EMSGSIZE;
2657 }
2658
2659 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2660                                     struct netlink_callback *cb)
2661 {
2662         int idx = 0, i;
2663         int start = cb->args[0];
2664         struct ip_vs_service *svc;
2665
2666         mutex_lock(&__ip_vs_mutex);
2667         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2668                 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2669                         if (++idx <= start)
2670                                 continue;
2671                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2672                                 idx--;
2673                                 goto nla_put_failure;
2674                         }
2675                 }
2676         }
2677
2678         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2679                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2680                         if (++idx <= start)
2681                                 continue;
2682                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2683                                 idx--;
2684                                 goto nla_put_failure;
2685                         }
2686                 }
2687         }
2688
2689 nla_put_failure:
2690         mutex_unlock(&__ip_vs_mutex);
2691         cb->args[0] = idx;
2692
2693         return skb->len;
2694 }
2695
2696 static int ip_vs_genl_parse_service(struct ip_vs_service_user_kern *usvc,
2697                                     struct nlattr *nla, int full_entry)
2698 {
2699         struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2700         struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2701
2702         /* Parse mandatory identifying service fields first */
2703         if (nla == NULL ||
2704             nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2705                 return -EINVAL;
2706
2707         nla_af          = attrs[IPVS_SVC_ATTR_AF];
2708         nla_protocol    = attrs[IPVS_SVC_ATTR_PROTOCOL];
2709         nla_addr        = attrs[IPVS_SVC_ATTR_ADDR];
2710         nla_port        = attrs[IPVS_SVC_ATTR_PORT];
2711         nla_fwmark      = attrs[IPVS_SVC_ATTR_FWMARK];
2712
2713         if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2714                 return -EINVAL;
2715
2716         usvc->af = nla_get_u16(nla_af);
2717 #ifdef CONFIG_IP_VS_IPV6
2718         if (usvc->af != AF_INET && usvc->af != AF_INET6)
2719 #else
2720         if (usvc->af != AF_INET)
2721 #endif
2722                 return -EAFNOSUPPORT;
2723
2724         if (nla_fwmark) {
2725                 usvc->protocol = IPPROTO_TCP;
2726                 usvc->fwmark = nla_get_u32(nla_fwmark);
2727         } else {
2728                 usvc->protocol = nla_get_u16(nla_protocol);
2729                 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
2730                 usvc->port = nla_get_u16(nla_port);
2731                 usvc->fwmark = 0;
2732         }
2733
2734         /* If a full entry was requested, check for the additional fields */
2735         if (full_entry) {
2736                 struct nlattr *nla_sched, *nla_flags, *nla_timeout,
2737                               *nla_netmask;
2738                 struct ip_vs_flags flags;
2739                 struct ip_vs_service *svc;
2740
2741                 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
2742                 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
2743                 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
2744                 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
2745
2746                 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
2747                         return -EINVAL;
2748
2749                 nla_memcpy(&flags, nla_flags, sizeof(flags));
2750
2751                 /* prefill flags from service if it already exists */
2752                 if (usvc->fwmark)
2753                         svc = __ip_vs_svc_fwm_get(usvc->af, usvc->fwmark);
2754                 else
2755                         svc = __ip_vs_service_get(usvc->af, usvc->protocol,
2756                                                   &usvc->addr, usvc->port);
2757                 if (svc) {
2758                         usvc->flags = svc->flags;
2759                         ip_vs_service_put(svc);
2760                 } else
2761                         usvc->flags = 0;
2762
2763                 /* set new flags from userland */
2764                 usvc->flags = (usvc->flags & ~flags.mask) |
2765                               (flags.flags & flags.mask);
2766                 usvc->sched_name = nla_data(nla_sched);
2767                 usvc->timeout = nla_get_u32(nla_timeout);
2768                 usvc->netmask = nla_get_u32(nla_netmask);
2769         }
2770
2771         return 0;
2772 }
2773
2774 static struct ip_vs_service *ip_vs_genl_find_service(struct nlattr *nla)
2775 {
2776         struct ip_vs_service_user_kern usvc;
2777         int ret;
2778
2779         ret = ip_vs_genl_parse_service(&usvc, nla, 0);
2780         if (ret)
2781                 return ERR_PTR(ret);
2782
2783         if (usvc.fwmark)
2784                 return __ip_vs_svc_fwm_get(usvc.af, usvc.fwmark);
2785         else
2786                 return __ip_vs_service_get(usvc.af, usvc.protocol,
2787                                            &usvc.addr, usvc.port);
2788 }
2789
2790 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
2791 {
2792         struct nlattr *nl_dest;
2793
2794         nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
2795         if (!nl_dest)
2796                 return -EMSGSIZE;
2797
2798         NLA_PUT(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr);
2799         NLA_PUT_U16(skb, IPVS_DEST_ATTR_PORT, dest->port);
2800
2801         NLA_PUT_U32(skb, IPVS_DEST_ATTR_FWD_METHOD,
2802                     atomic_read(&dest->conn_flags) & IP_VS_CONN_F_FWD_MASK);
2803         NLA_PUT_U32(skb, IPVS_DEST_ATTR_WEIGHT, atomic_read(&dest->weight));
2804         NLA_PUT_U32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold);
2805         NLA_PUT_U32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold);
2806         NLA_PUT_U32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
2807                     atomic_read(&dest->activeconns));
2808         NLA_PUT_U32(skb, IPVS_DEST_ATTR_INACT_CONNS,
2809                     atomic_read(&dest->inactconns));
2810         NLA_PUT_U32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
2811                     atomic_read(&dest->persistconns));
2812
2813         if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
2814                 goto nla_put_failure;
2815
2816         nla_nest_end(skb, nl_dest);
2817
2818         return 0;
2819
2820 nla_put_failure:
2821         nla_nest_cancel(skb, nl_dest);
2822         return -EMSGSIZE;
2823 }
2824
2825 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
2826                                 struct netlink_callback *cb)
2827 {
2828         void *hdr;
2829
2830         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2831                           &ip_vs_genl_family, NLM_F_MULTI,
2832                           IPVS_CMD_NEW_DEST);
2833         if (!hdr)
2834                 return -EMSGSIZE;
2835
2836         if (ip_vs_genl_fill_dest(skb, dest) < 0)
2837                 goto nla_put_failure;
2838
2839         return genlmsg_end(skb, hdr);
2840
2841 nla_put_failure:
2842         genlmsg_cancel(skb, hdr);
2843         return -EMSGSIZE;
2844 }
2845
2846 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
2847                                  struct netlink_callback *cb)
2848 {
2849         int idx = 0;
2850         int start = cb->args[0];
2851         struct ip_vs_service *svc;
2852         struct ip_vs_dest *dest;
2853         struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
2854
2855         mutex_lock(&__ip_vs_mutex);
2856
2857         /* Try to find the service for which to dump destinations */
2858         if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
2859                         IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
2860                 goto out_err;
2861
2862         svc = ip_vs_genl_find_service(attrs[IPVS_CMD_ATTR_SERVICE]);
2863         if (IS_ERR(svc) || svc == NULL)
2864                 goto out_err;
2865
2866         /* Dump the destinations */
2867         list_for_each_entry(dest, &svc->destinations, n_list) {
2868                 if (++idx <= start)
2869                         continue;
2870                 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
2871                         idx--;
2872                         goto nla_put_failure;
2873                 }
2874         }
2875
2876 nla_put_failure:
2877         cb->args[0] = idx;
2878         ip_vs_service_put(svc);
2879
2880 out_err:
2881         mutex_unlock(&__ip_vs_mutex);
2882
2883         return skb->len;
2884 }
2885
2886 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
2887                                  struct nlattr *nla, int full_entry)
2888 {
2889         struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
2890         struct nlattr *nla_addr, *nla_port;
2891
2892         /* Parse mandatory identifying destination fields first */
2893         if (nla == NULL ||
2894             nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
2895                 return -EINVAL;
2896
2897         nla_addr        = attrs[IPVS_DEST_ATTR_ADDR];
2898         nla_port        = attrs[IPVS_DEST_ATTR_PORT];
2899
2900         if (!(nla_addr && nla_port))
2901                 return -EINVAL;
2902
2903         nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
2904         udest->port = nla_get_u16(nla_port);
2905
2906         /* If a full entry was requested, check for the additional fields */
2907         if (full_entry) {
2908                 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
2909                               *nla_l_thresh;
2910
2911                 nla_fwd         = attrs[IPVS_DEST_ATTR_FWD_METHOD];
2912                 nla_weight      = attrs[IPVS_DEST_ATTR_WEIGHT];
2913                 nla_u_thresh    = attrs[IPVS_DEST_ATTR_U_THRESH];
2914                 nla_l_thresh    = attrs[IPVS_DEST_ATTR_L_THRESH];
2915
2916                 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
2917                         return -EINVAL;
2918
2919                 udest->conn_flags = nla_get_u32(nla_fwd)
2920                                     & IP_VS_CONN_F_FWD_MASK;
2921                 udest->weight = nla_get_u32(nla_weight);
2922                 udest->u_threshold = nla_get_u32(nla_u_thresh);
2923                 udest->l_threshold = nla_get_u32(nla_l_thresh);
2924         }
2925
2926         return 0;
2927 }
2928
2929 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
2930                                   const char *mcast_ifn, __be32 syncid)
2931 {
2932         struct nlattr *nl_daemon;
2933
2934         nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
2935         if (!nl_daemon)
2936                 return -EMSGSIZE;
2937
2938         NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_STATE, state);
2939         NLA_PUT_STRING(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn);
2940         NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid);
2941
2942         nla_nest_end(skb, nl_daemon);
2943
2944         return 0;
2945
2946 nla_put_failure:
2947         nla_nest_cancel(skb, nl_daemon);
2948         return -EMSGSIZE;
2949 }
2950
2951 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
2952                                   const char *mcast_ifn, __be32 syncid,
2953                                   struct netlink_callback *cb)
2954 {
2955         void *hdr;
2956         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2957                           &ip_vs_genl_family, NLM_F_MULTI,
2958                           IPVS_CMD_NEW_DAEMON);
2959         if (!hdr)
2960                 return -EMSGSIZE;
2961
2962         if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
2963                 goto nla_put_failure;
2964
2965         return genlmsg_end(skb, hdr);
2966
2967 nla_put_failure:
2968         genlmsg_cancel(skb, hdr);
2969         return -EMSGSIZE;
2970 }
2971
2972 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
2973                                    struct netlink_callback *cb)
2974 {
2975         mutex_lock(&__ip_vs_mutex);
2976         if ((ip_vs_sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
2977                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
2978                                            ip_vs_master_mcast_ifn,
2979                                            ip_vs_master_syncid, cb) < 0)
2980                         goto nla_put_failure;
2981
2982                 cb->args[0] = 1;
2983         }
2984
2985         if ((ip_vs_sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
2986                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
2987                                            ip_vs_backup_mcast_ifn,
2988                                            ip_vs_backup_syncid, cb) < 0)
2989                         goto nla_put_failure;
2990
2991                 cb->args[1] = 1;
2992         }
2993
2994 nla_put_failure:
2995         mutex_unlock(&__ip_vs_mutex);
2996
2997         return skb->len;
2998 }
2999
3000 static int ip_vs_genl_new_daemon(struct nlattr **attrs)
3001 {
3002         if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3003               attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3004               attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3005                 return -EINVAL;
3006
3007         return start_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3008                                  nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3009                                  nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3010 }
3011
3012 static int ip_vs_genl_del_daemon(struct nlattr **attrs)
3013 {
3014         if (!attrs[IPVS_DAEMON_ATTR_STATE])
3015                 return -EINVAL;
3016
3017         return stop_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3018 }
3019
3020 static int ip_vs_genl_set_config(struct nlattr **attrs)
3021 {
3022         struct ip_vs_timeout_user t;
3023
3024         __ip_vs_get_timeouts(&t);
3025
3026         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3027                 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3028
3029         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3030                 t.tcp_fin_timeout =
3031                         nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3032
3033         if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3034                 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3035
3036         return ip_vs_set_timeout(&t);
3037 }
3038
3039 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3040 {
3041         struct ip_vs_service *svc = NULL;
3042         struct ip_vs_service_user_kern usvc;
3043         struct ip_vs_dest_user_kern udest;
3044         int ret = 0, cmd;
3045         int need_full_svc = 0, need_full_dest = 0;
3046
3047         cmd = info->genlhdr->cmd;
3048
3049         mutex_lock(&__ip_vs_mutex);
3050
3051         if (cmd == IPVS_CMD_FLUSH) {
3052                 ret = ip_vs_flush();
3053                 goto out;
3054         } else if (cmd == IPVS_CMD_SET_CONFIG) {
3055                 ret = ip_vs_genl_set_config(info->attrs);
3056                 goto out;
3057         } else if (cmd == IPVS_CMD_NEW_DAEMON ||
3058                    cmd == IPVS_CMD_DEL_DAEMON) {
3059
3060                 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3061
3062                 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3063                     nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3064                                      info->attrs[IPVS_CMD_ATTR_DAEMON],
3065                                      ip_vs_daemon_policy)) {
3066                         ret = -EINVAL;
3067                         goto out;
3068                 }
3069
3070                 if (cmd == IPVS_CMD_NEW_DAEMON)
3071                         ret = ip_vs_genl_new_daemon(daemon_attrs);
3072                 else
3073                         ret = ip_vs_genl_del_daemon(daemon_attrs);
3074                 goto out;
3075         } else if (cmd == IPVS_CMD_ZERO &&
3076                    !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3077                 ret = ip_vs_zero_all();
3078                 goto out;
3079         }
3080
3081         /* All following commands require a service argument, so check if we
3082          * received a valid one. We need a full service specification when
3083          * adding / editing a service. Only identifying members otherwise. */
3084         if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3085                 need_full_svc = 1;
3086
3087         ret = ip_vs_genl_parse_service(&usvc,
3088                                        info->attrs[IPVS_CMD_ATTR_SERVICE],
3089                                        need_full_svc);
3090         if (ret)
3091                 goto out;
3092
3093         /* Lookup the exact service by <protocol, addr, port> or fwmark */
3094         if (usvc.fwmark == 0)
3095                 svc = __ip_vs_service_get(usvc.af, usvc.protocol,
3096                                           &usvc.addr, usvc.port);
3097         else
3098                 svc = __ip_vs_svc_fwm_get(usvc.af, usvc.fwmark);
3099
3100         /* Unless we're adding a new service, the service must already exist */
3101         if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3102                 ret = -ESRCH;
3103                 goto out;
3104         }
3105
3106         /* Destination commands require a valid destination argument. For
3107          * adding / editing a destination, we need a full destination
3108          * specification. */
3109         if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3110             cmd == IPVS_CMD_DEL_DEST) {
3111                 if (cmd != IPVS_CMD_DEL_DEST)
3112                         need_full_dest = 1;
3113
3114                 ret = ip_vs_genl_parse_dest(&udest,
3115                                             info->attrs[IPVS_CMD_ATTR_DEST],
3116                                             need_full_dest);
3117                 if (ret)
3118                         goto out;
3119         }
3120
3121         switch (cmd) {
3122         case IPVS_CMD_NEW_SERVICE:
3123                 if (svc == NULL)
3124                         ret = ip_vs_add_service(&usvc, &svc);
3125                 else
3126                         ret = -EEXIST;
3127                 break;
3128         case IPVS_CMD_SET_SERVICE:
3129                 ret = ip_vs_edit_service(svc, &usvc);
3130                 break;
3131         case IPVS_CMD_DEL_SERVICE:
3132                 ret = ip_vs_del_service(svc);
3133                 break;
3134         case IPVS_CMD_NEW_DEST:
3135                 ret = ip_vs_add_dest(svc, &udest);
3136                 break;
3137         case IPVS_CMD_SET_DEST:
3138                 ret = ip_vs_edit_dest(svc, &udest);
3139                 break;
3140         case IPVS_CMD_DEL_DEST:
3141                 ret = ip_vs_del_dest(svc, &udest);
3142                 break;
3143         case IPVS_CMD_ZERO:
3144                 ret = ip_vs_zero_service(svc);
3145                 break;
3146         default:
3147                 ret = -EINVAL;
3148         }
3149
3150 out:
3151         if (svc)
3152                 ip_vs_service_put(svc);
3153         mutex_unlock(&__ip_vs_mutex);
3154
3155         return ret;
3156 }
3157
3158 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3159 {
3160         struct sk_buff *msg;
3161         void *reply;
3162         int ret, cmd, reply_cmd;
3163
3164         cmd = info->genlhdr->cmd;
3165
3166         if (cmd == IPVS_CMD_GET_SERVICE)
3167                 reply_cmd = IPVS_CMD_NEW_SERVICE;
3168         else if (cmd == IPVS_CMD_GET_INFO)
3169                 reply_cmd = IPVS_CMD_SET_INFO;
3170         else if (cmd == IPVS_CMD_GET_CONFIG)
3171                 reply_cmd = IPVS_CMD_SET_CONFIG;
3172         else {
3173                 IP_VS_ERR("unknown Generic Netlink command\n");
3174                 return -EINVAL;
3175         }
3176
3177         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3178         if (!msg)
3179                 return -ENOMEM;
3180
3181         mutex_lock(&__ip_vs_mutex);
3182
3183         reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3184         if (reply == NULL)
3185                 goto nla_put_failure;
3186
3187         switch (cmd) {
3188         case IPVS_CMD_GET_SERVICE:
3189         {
3190                 struct ip_vs_service *svc;
3191
3192                 svc = ip_vs_genl_find_service(info->attrs[IPVS_CMD_ATTR_SERVICE]);
3193                 if (IS_ERR(svc)) {
3194                         ret = PTR_ERR(svc);
3195                         goto out_err;
3196                 } else if (svc) {
3197                         ret = ip_vs_genl_fill_service(msg, svc);
3198                         ip_vs_service_put(svc);
3199                         if (ret)
3200                                 goto nla_put_failure;
3201                 } else {
3202                         ret = -ESRCH;
3203                         goto out_err;
3204                 }
3205
3206                 break;
3207         }
3208
3209         case IPVS_CMD_GET_CONFIG:
3210         {
3211                 struct ip_vs_timeout_user t;
3212
3213                 __ip_vs_get_timeouts(&t);
3214 #ifdef CONFIG_IP_VS_PROTO_TCP
3215                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP, t.tcp_timeout);
3216                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3217                             t.tcp_fin_timeout);
3218 #endif
3219 #ifdef CONFIG_IP_VS_PROTO_UDP
3220                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout);
3221 #endif
3222
3223                 break;
3224         }
3225
3226         case IPVS_CMD_GET_INFO:
3227                 NLA_PUT_U32(msg, IPVS_INFO_ATTR_VERSION, IP_VS_VERSION_CODE);
3228                 NLA_PUT_U32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3229                             IP_VS_CONN_TAB_SIZE);
3230                 break;
3231         }
3232
3233         genlmsg_end(msg, reply);
3234         ret = genlmsg_unicast(msg, info->snd_pid);
3235         goto out;
3236
3237 nla_put_failure:
3238         IP_VS_ERR("not enough space in Netlink message\n");
3239         ret = -EMSGSIZE;
3240
3241 out_err:
3242         nlmsg_free(msg);
3243 out:
3244         mutex_unlock(&__ip_vs_mutex);
3245
3246         return ret;
3247 }
3248
3249
3250 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3251         {
3252                 .cmd    = IPVS_CMD_NEW_SERVICE,
3253                 .flags  = GENL_ADMIN_PERM,
3254                 .policy = ip_vs_cmd_policy,
3255                 .doit   = ip_vs_genl_set_cmd,
3256         },
3257         {
3258                 .cmd    = IPVS_CMD_SET_SERVICE,
3259                 .flags  = GENL_ADMIN_PERM,
3260                 .policy = ip_vs_cmd_policy,
3261                 .doit   = ip_vs_genl_set_cmd,
3262         },
3263         {
3264                 .cmd    = IPVS_CMD_DEL_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_GET_SERVICE,
3271                 .flags  = GENL_ADMIN_PERM,
3272                 .doit   = ip_vs_genl_get_cmd,
3273                 .dumpit = ip_vs_genl_dump_services,
3274                 .policy = ip_vs_cmd_policy,
3275         },
3276         {
3277                 .cmd    = IPVS_CMD_NEW_DEST,
3278                 .flags  = GENL_ADMIN_PERM,
3279                 .policy = ip_vs_cmd_policy,
3280                 .doit   = ip_vs_genl_set_cmd,
3281         },
3282         {
3283                 .cmd    = IPVS_CMD_SET_DEST,
3284                 .flags  = GENL_ADMIN_PERM,
3285                 .policy = ip_vs_cmd_policy,
3286                 .doit   = ip_vs_genl_set_cmd,
3287         },
3288         {
3289                 .cmd    = IPVS_CMD_DEL_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_GET_DEST,
3296                 .flags  = GENL_ADMIN_PERM,
3297                 .policy = ip_vs_cmd_policy,
3298                 .dumpit = ip_vs_genl_dump_dests,
3299         },
3300         {
3301                 .cmd    = IPVS_CMD_NEW_DAEMON,
3302                 .flags  = GENL_ADMIN_PERM,
3303                 .policy = ip_vs_cmd_policy,
3304                 .doit   = ip_vs_genl_set_cmd,
3305         },
3306         {
3307                 .cmd    = IPVS_CMD_DEL_DAEMON,
3308                 .flags  = GENL_ADMIN_PERM,
3309                 .policy = ip_vs_cmd_policy,
3310                 .doit   = ip_vs_genl_set_cmd,
3311         },
3312         {
3313                 .cmd    = IPVS_CMD_GET_DAEMON,
3314                 .flags  = GENL_ADMIN_PERM,
3315                 .dumpit = ip_vs_genl_dump_daemons,
3316         },
3317         {
3318                 .cmd    = IPVS_CMD_SET_CONFIG,
3319                 .flags  = GENL_ADMIN_PERM,
3320                 .policy = ip_vs_cmd_policy,
3321                 .doit   = ip_vs_genl_set_cmd,
3322         },
3323         {
3324                 .cmd    = IPVS_CMD_GET_CONFIG,
3325                 .flags  = GENL_ADMIN_PERM,
3326                 .doit   = ip_vs_genl_get_cmd,
3327         },
3328         {
3329                 .cmd    = IPVS_CMD_GET_INFO,
3330                 .flags  = GENL_ADMIN_PERM,
3331                 .doit   = ip_vs_genl_get_cmd,
3332         },
3333         {
3334                 .cmd    = IPVS_CMD_ZERO,
3335                 .flags  = GENL_ADMIN_PERM,
3336                 .policy = ip_vs_cmd_policy,
3337                 .doit   = ip_vs_genl_set_cmd,
3338         },
3339         {
3340                 .cmd    = IPVS_CMD_FLUSH,
3341                 .flags  = GENL_ADMIN_PERM,
3342                 .doit   = ip_vs_genl_set_cmd,
3343         },
3344 };
3345
3346 static int __init ip_vs_genl_register(void)
3347 {
3348         int ret, i;
3349
3350         ret = genl_register_family(&ip_vs_genl_family);
3351         if (ret)
3352                 return ret;
3353
3354         for (i = 0; i < ARRAY_SIZE(ip_vs_genl_ops); i++) {
3355                 ret = genl_register_ops(&ip_vs_genl_family, &ip_vs_genl_ops[i]);
3356                 if (ret)
3357                         goto err_out;
3358         }
3359         return 0;
3360
3361 err_out:
3362         genl_unregister_family(&ip_vs_genl_family);
3363         return ret;
3364 }
3365
3366 static void ip_vs_genl_unregister(void)
3367 {
3368         genl_unregister_family(&ip_vs_genl_family);
3369 }
3370
3371 /* End of Generic Netlink interface definitions */
3372
3373
3374 int __init ip_vs_control_init(void)
3375 {
3376         int ret;
3377         int idx;
3378
3379         EnterFunction(2);
3380
3381         ret = nf_register_sockopt(&ip_vs_sockopts);
3382         if (ret) {
3383                 IP_VS_ERR("cannot register sockopt.\n");
3384                 return ret;
3385         }
3386
3387         ret = ip_vs_genl_register();
3388         if (ret) {
3389                 IP_VS_ERR("cannot register Generic Netlink interface.\n");
3390                 nf_unregister_sockopt(&ip_vs_sockopts);
3391                 return ret;
3392         }
3393
3394         proc_net_fops_create(&init_net, "ip_vs", 0, &ip_vs_info_fops);
3395         proc_net_fops_create(&init_net, "ip_vs_stats",0, &ip_vs_stats_fops);
3396
3397         sysctl_header = register_sysctl_paths(net_vs_ctl_path, vs_vars);
3398
3399         /* Initialize ip_vs_svc_table, ip_vs_svc_fwm_table, ip_vs_rtable */
3400         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++)  {
3401                 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3402                 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3403         }
3404         for(idx = 0; idx < IP_VS_RTAB_SIZE; idx++)  {
3405                 INIT_LIST_HEAD(&ip_vs_rtable[idx]);
3406         }
3407
3408         ip_vs_new_estimator(&ip_vs_stats);
3409
3410         /* Hook the defense timer */
3411         schedule_delayed_work(&defense_work, DEFENSE_TIMER_PERIOD);
3412
3413         LeaveFunction(2);
3414         return 0;
3415 }
3416
3417
3418 void ip_vs_control_cleanup(void)
3419 {
3420         EnterFunction(2);
3421         ip_vs_trash_cleanup();
3422         cancel_rearming_delayed_work(&defense_work);
3423         cancel_work_sync(&defense_work.work);
3424         ip_vs_kill_estimator(&ip_vs_stats);
3425         unregister_sysctl_table(sysctl_header);
3426         proc_net_remove(&init_net, "ip_vs_stats");
3427         proc_net_remove(&init_net, "ip_vs");
3428         ip_vs_genl_unregister();
3429         nf_unregister_sockopt(&ip_vs_sockopts);
3430         LeaveFunction(2);
3431 }