net/ipv4/ipvs/ip_vs_ctl.c

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