net/netfilter/ipvs/ip_vs_ctl.c

   1 /*
   2  * IPVS         An implementation of the IP virtual server support for the
   3  *              LINUX operating system.  IPVS is now implemented as a module
   4  *              over the NetFilter framework. IPVS can be used to build a
   5  *              high-performance and highly available server based on a
   6  *              cluster of servers.
   7  *
   8  * Authors:     Wensong Zhang <wensong@linuxvirtualserver.org>
   9  *              Peter Kese <peter.kese@ijs.si>
  10  *              Julian Anastasov <ja@ssi.bg>
  11  *
  12  *              This program is free software; you can redistribute it and/or
  13  *              modify it under the terms of the GNU General Public License
  14  *              as published by the Free Software Foundation; either version
  15  *              2 of the License, or (at your option) any later version.
  16  *
  17  * Changes:
  18  *
  19  */
  20
  21 #define KMSG_COMPONENT "IPVS"
  22 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
  23
  24 #include <linux/module.h>
  25 #include <linux/init.h>
  26 #include <linux/types.h>
  27 #include <linux/capability.h>
  28 #include <linux/fs.h>
  29 #include <linux/sysctl.h>
  30 #include <linux/proc_fs.h>
  31 #include <linux/workqueue.h>
  32 #include <linux/swap.h>
  33 #include <linux/seq_file.h>
  34 #include <linux/slab.h>
  35
  36 #include <linux/netfilter.h>
  37 #include <linux/netfilter_ipv4.h>
  38 #include <linux/mutex.h>
  39
  40 #include <net/net_namespace.h>
  41 #include <net/ip.h>
  42 #ifdef CONFIG_IP_VS_IPV6
  43 #include <net/ipv6.h>
  44 #include <net/ip6_route.h>
  45 #endif
  46 #include <net/route.h>
  47 #include <net/sock.h>
  48 #include <net/genetlink.h>
  49
  50 #include <asm/uaccess.h>
  51
  52 #include <net/ip_vs.h>
  53
  54 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
  55 static DEFINE_MUTEX(__ip_vs_mutex);
  56
  57 /* lock for service table */
  58 static DEFINE_RWLOCK(__ip_vs_svc_lock);
  59
  60 /* lock for table with the real services */
  61 static DEFINE_RWLOCK(__ip_vs_rs_lock);
  62
  63 /* lock for state and timeout tables */
  64 static DEFINE_SPINLOCK(ip_vs_securetcp_lock);
  65
  66 /* lock for drop entry handling */
  67 static DEFINE_SPINLOCK(__ip_vs_dropentry_lock);
  68
  69 /* lock for drop packet handling */
  70 static DEFINE_SPINLOCK(__ip_vs_droppacket_lock);
  71
  72 /* 1/rate drop and drop-entry variables */
  73 int ip_vs_drop_rate = 0;
  74 int ip_vs_drop_counter = 0;
  75 static atomic_t ip_vs_dropentry = ATOMIC_INIT(0);
  76
  77 /* number of virtual services */
  78 static int ip_vs_num_services = 0;
  79
  80 /* sysctl variables */
  81 static int sysctl_ip_vs_drop_entry = 0;
  82 static int sysctl_ip_vs_drop_packet = 0;
  83 static int sysctl_ip_vs_secure_tcp = 0;
  84 static int sysctl_ip_vs_amemthresh = 1024;
  85 static int sysctl_ip_vs_am_droprate = 10;
  86 int sysctl_ip_vs_cache_bypass = 0;
  87 int sysctl_ip_vs_expire_nodest_conn = 0;
  88 int sysctl_ip_vs_expire_quiescent_template = 0;
  89 int sysctl_ip_vs_sync_threshold[2] = { 3, 50 };
  90 int sysctl_ip_vs_nat_icmp_send = 0;
  91 #ifdef CONFIG_IP_VS_NFCT
  92 int sysctl_ip_vs_conntrack;
  93 #endif
  94 int sysctl_ip_vs_snat_reroute = 1;
  95
  96
  97 #ifdef CONFIG_IP_VS_DEBUG
  98 static int sysctl_ip_vs_debug_level = 0;
  99
 100 int ip_vs_get_debug_level(void)
 101 {
 102         return sysctl_ip_vs_debug_level;
 103 }
 104 #endif
 105
 106 #ifdef CONFIG_IP_VS_IPV6
 107 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
 108 static int __ip_vs_addr_is_local_v6(const struct in6_addr *addr)
 109 {
 110         struct rt6_info *rt;
 111         struct flowi fl = {
 112                 .oif = 0,
 113                 .nl_u = {
 114                         .ip6_u = {
 115                                 .daddr = *addr,
 116                                 .saddr = { .s6_addr32 = {0, 0, 0, 0} }, } },
 117         };
 118
 119         rt = (struct rt6_info *)ip6_route_output(&init_net, NULL, &fl);
 120         if (rt && rt->rt6i_dev && (rt->rt6i_dev->flags & IFF_LOOPBACK))
 121                         return 1;
 122
 123         return 0;
 124 }
 125 #endif
 126 /*
 127  *      update_defense_level is called from keventd and from sysctl,
 128  *      so it needs to protect itself from softirqs
 129  */
 130 static void update_defense_level(void)
 131 {
 132         struct sysinfo i;
 133         static int old_secure_tcp = 0;
 134         int availmem;
 135         int nomem;
 136         int to_change = -1;
 137
 138         /* we only count free and buffered memory (in pages) */
 139         si_meminfo(&i);
 140         availmem = i.freeram + i.bufferram;
 141         /* however in linux 2.5 the i.bufferram is total page cache size,
 142            we need adjust it */
 143         /* si_swapinfo(&i); */
 144         /* availmem = availmem - (i.totalswap - i.freeswap); */
 145
 146         nomem = (availmem < sysctl_ip_vs_amemthresh);
 147
 148         local_bh_disable();
 149
 150         /* drop_entry */
 151         spin_lock(&__ip_vs_dropentry_lock);
 152         switch (sysctl_ip_vs_drop_entry) {
 153         case 0:
 154                 atomic_set(&ip_vs_dropentry, 0);
 155                 break;
 156         case 1:
 157                 if (nomem) {
 158                         atomic_set(&ip_vs_dropentry, 1);
 159                         sysctl_ip_vs_drop_entry = 2;
 160                 } else {
 161                         atomic_set(&ip_vs_dropentry, 0);
 162                 }
 163                 break;
 164         case 2:
 165                 if (nomem) {
 166                         atomic_set(&ip_vs_dropentry, 1);
 167                 } else {
 168                         atomic_set(&ip_vs_dropentry, 0);
 169                         sysctl_ip_vs_drop_entry = 1;
 170                 };
 171                 break;
 172         case 3:
 173                 atomic_set(&ip_vs_dropentry, 1);
 174                 break;
 175         }
 176         spin_unlock(&__ip_vs_dropentry_lock);
 177
 178         /* drop_packet */
 179         spin_lock(&__ip_vs_droppacket_lock);
 180         switch (sysctl_ip_vs_drop_packet) {
 181         case 0:
 182                 ip_vs_drop_rate = 0;
 183                 break;
 184         case 1:
 185                 if (nomem) {
 186                         ip_vs_drop_rate = ip_vs_drop_counter
 187                                 = sysctl_ip_vs_amemthresh /
 188                                 (sysctl_ip_vs_amemthresh-availmem);
 189                         sysctl_ip_vs_drop_packet = 2;
 190                 } else {
 191                         ip_vs_drop_rate = 0;
 192                 }
 193                 break;
 194         case 2:
 195                 if (nomem) {
 196                         ip_vs_drop_rate = ip_vs_drop_counter
 197                                 = sysctl_ip_vs_amemthresh /
 198                                 (sysctl_ip_vs_amemthresh-availmem);
 199                 } else {
 200                         ip_vs_drop_rate = 0;
 201                         sysctl_ip_vs_drop_packet = 1;
 202                 }
 203                 break;
 204         case 3:
 205                 ip_vs_drop_rate = sysctl_ip_vs_am_droprate;
 206                 break;
 207         }
 208         spin_unlock(&__ip_vs_droppacket_lock);
 209
 210         /* secure_tcp */
 211         spin_lock(&ip_vs_securetcp_lock);
 212         switch (sysctl_ip_vs_secure_tcp) {
 213         case 0:
 214                 if (old_secure_tcp >= 2)
 215                         to_change = 0;
 216                 break;
 217         case 1:
 218                 if (nomem) {
 219                         if (old_secure_tcp < 2)
 220                                 to_change = 1;
 221                         sysctl_ip_vs_secure_tcp = 2;
 222                 } else {
 223                         if (old_secure_tcp >= 2)
 224                                 to_change = 0;
 225                 }
 226                 break;
 227         case 2:
 228                 if (nomem) {
 229                         if (old_secure_tcp < 2)
 230                                 to_change = 1;
 231                 } else {
 232                         if (old_secure_tcp >= 2)
 233                                 to_change = 0;
 234                         sysctl_ip_vs_secure_tcp = 1;
 235                 }
 236                 break;
 237         case 3:
 238                 if (old_secure_tcp < 2)
 239                         to_change = 1;
 240                 break;
 241         }
 242         old_secure_tcp = sysctl_ip_vs_secure_tcp;
 243         if (to_change >= 0)
 244                 ip_vs_protocol_timeout_change(sysctl_ip_vs_secure_tcp>1);
 245         spin_unlock(&ip_vs_securetcp_lock);
 246
 247         local_bh_enable();
 248 }
 249
 250
 251 /*
 252  *      Timer for checking the defense
 253  */
 254 #define DEFENSE_TIMER_PERIOD    1*HZ
 255 static void defense_work_handler(struct work_struct *work);
 256 static DECLARE_DELAYED_WORK(defense_work, defense_work_handler);
 257
 258 static void defense_work_handler(struct work_struct *work)
 259 {
 260         update_defense_level();
 261         if (atomic_read(&ip_vs_dropentry))
 262                 ip_vs_random_dropentry();
 263
 264         schedule_delayed_work(&defense_work, DEFENSE_TIMER_PERIOD);
 265 }
 266
 267 int
 268 ip_vs_use_count_inc(void)
 269 {
 270         return try_module_get(THIS_MODULE);
 271 }
 272
 273 void
 274 ip_vs_use_count_dec(void)
 275 {
 276         module_put(THIS_MODULE);
 277 }
 278
 279
 280 /*
 281  *      Hash table: for virtual service lookups
 282  */
 283 #define IP_VS_SVC_TAB_BITS 8
 284 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
 285 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
 286
 287 /* the service table hashed by <protocol, addr, port> */
 288 static struct list_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
 289 /* the service table hashed by fwmark */
 290 static struct list_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
 291
 292 /*
 293  *      Hash table: for real service lookups
 294  */
 295 #define IP_VS_RTAB_BITS 4
 296 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
 297 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
 298
 299 static struct list_head ip_vs_rtable[IP_VS_RTAB_SIZE];
 300
 301 /*
 302  *      Trash for destinations
 303  */
 304 static LIST_HEAD(ip_vs_dest_trash);
 305
 306 /*
 307  *      FTP & NULL virtual service counters
 308  */
 309 static atomic_t ip_vs_ftpsvc_counter = ATOMIC_INIT(0);
 310 static atomic_t ip_vs_nullsvc_counter = ATOMIC_INIT(0);
 311
 312
 313 /*
 314  *      Returns hash value for virtual service
 315  */
 316 static __inline__ unsigned
 317 ip_vs_svc_hashkey(int af, unsigned proto, const union nf_inet_addr *addr,
 318                   __be16 port)
 319 {
 320         register unsigned porth = ntohs(port);
 321         __be32 addr_fold = addr->ip;
 322
 323 #ifdef CONFIG_IP_VS_IPV6
 324         if (af == AF_INET6)
 325                 addr_fold = addr->ip6[0]^addr->ip6[1]^
 326                             addr->ip6[2]^addr->ip6[3];
 327 #endif
 328
 329         return (proto^ntohl(addr_fold)^(porth>>IP_VS_SVC_TAB_BITS)^porth)
 330                 & IP_VS_SVC_TAB_MASK;
 331 }
 332
 333 /*
 334  *      Returns hash value of fwmark for virtual service lookup
 335  */
 336 static __inline__ unsigned ip_vs_svc_fwm_hashkey(__u32 fwmark)
 337 {
 338         return fwmark & IP_VS_SVC_TAB_MASK;
 339 }
 340
 341 /*
 342  *      Hashes a service in the ip_vs_svc_table by <proto,addr,port>
 343  *      or in the ip_vs_svc_fwm_table by fwmark.
 344  *      Should be called with locked tables.
 345  */
 346 static int ip_vs_svc_hash(struct ip_vs_service *svc)
 347 {
 348         unsigned hash;
 349
 350         if (svc->flags & IP_VS_SVC_F_HASHED) {
 351                 pr_err("%s(): request for already hashed, called from %pF\n",
 352                        __func__, __builtin_return_address(0));
 353                 return 0;
 354         }
 355
 356         if (svc->fwmark == 0) {
 357                 /*
 358                  *  Hash it by <protocol,addr,port> in ip_vs_svc_table
 359                  */
 360                 hash = ip_vs_svc_hashkey(svc->af, svc->protocol, &svc->addr,
 361                                          svc->port);
 362                 list_add(&svc->s_list, &ip_vs_svc_table[hash]);
 363         } else {
 364                 /*
 365                  *  Hash it by fwmark in ip_vs_svc_fwm_table
 366                  */
 367                 hash = ip_vs_svc_fwm_hashkey(svc->fwmark);
 368                 list_add(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
 369         }
 370
 371         svc->flags |= IP_VS_SVC_F_HASHED;
 372         /* increase its refcnt because it is referenced by the svc table */
 373         atomic_inc(&svc->refcnt);
 374         return 1;
 375 }
 376
 377
 378 /*
 379  *      Unhashes a service from ip_vs_svc_table/ip_vs_svc_fwm_table.
 380  *      Should be called with locked tables.
 381  */
 382 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
 383 {
 384         if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
 385                 pr_err("%s(): request for unhash flagged, called from %pF\n",
 386                        __func__, __builtin_return_address(0));
 387                 return 0;
 388         }
 389
 390         if (svc->fwmark == 0) {
 391                 /* Remove it from the ip_vs_svc_table table */
 392                 list_del(&svc->s_list);
 393         } else {
 394                 /* Remove it from the ip_vs_svc_fwm_table table */
 395                 list_del(&svc->f_list);
 396         }
 397
 398         svc->flags &= ~IP_VS_SVC_F_HASHED;
 399         atomic_dec(&svc->refcnt);
 400         return 1;
 401 }
 402
 403
 404 /*
 405  *      Get service by {proto,addr,port} in the service table.
 406  */
 407 static inline struct ip_vs_service *
 408 __ip_vs_service_find(int af, __u16 protocol, const union nf_inet_addr *vaddr,
 409                     __be16 vport)
 410 {
 411         unsigned hash;
 412         struct ip_vs_service *svc;
 413
 414         /* Check for "full" addressed entries */
 415         hash = ip_vs_svc_hashkey(af, protocol, vaddr, vport);
 416
 417         list_for_each_entry(svc, &ip_vs_svc_table[hash], s_list){
 418                 if ((svc->af == af)
 419                     && ip_vs_addr_equal(af, &svc->addr, vaddr)
 420                     && (svc->port == vport)
 421                     && (svc->protocol == protocol)) {
 422                         /* HIT */
 423                         return svc;
 424                 }
 425         }
 426
 427         return NULL;
 428 }
 429
 430
 431 /*
 432  *      Get service by {fwmark} in the service table.
 433  */
 434 static inline struct ip_vs_service *
 435 __ip_vs_svc_fwm_find(int af, __u32 fwmark)
 436 {
 437         unsigned hash;
 438         struct ip_vs_service *svc;
 439
 440         /* Check for fwmark addressed entries */
 441         hash = ip_vs_svc_fwm_hashkey(fwmark);
 442
 443         list_for_each_entry(svc, &ip_vs_svc_fwm_table[hash], f_list) {
 444                 if (svc->fwmark == fwmark && svc->af == af) {
 445                         /* HIT */
 446                         return svc;
 447                 }
 448         }
 449
 450         return NULL;
 451 }
 452
 453 struct ip_vs_service *
 454 ip_vs_service_get(int af, __u32 fwmark, __u16 protocol,
 455                   const union nf_inet_addr *vaddr, __be16 vport)
 456 {
 457         struct ip_vs_service *svc;
 458
 459         read_lock(&__ip_vs_svc_lock);
 460
 461         /*
 462          *      Check the table hashed by fwmark first
 463          */
 464         if (fwmark && (svc = __ip_vs_svc_fwm_find(af, fwmark)))
 465                 goto out;
 466
 467         /*
 468          *      Check the table hashed by <protocol,addr,port>
 469          *      for "full" addressed entries
 470          */
 471         svc = __ip_vs_service_find(af, protocol, vaddr, vport);
 472
 473         if (svc == NULL
 474             && protocol == IPPROTO_TCP
 475             && atomic_read(&ip_vs_ftpsvc_counter)
 476             && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
 477                 /*
 478                  * Check if ftp service entry exists, the packet
 479                  * might belong to FTP data connections.
 480                  */
 481                 svc = __ip_vs_service_find(af, protocol, vaddr, FTPPORT);
 482         }
 483
 484         if (svc == NULL
 485             && atomic_read(&ip_vs_nullsvc_counter)) {
 486                 /*
 487                  * Check if the catch-all port (port zero) exists
 488                  */
 489                 svc = __ip_vs_service_find(af, protocol, vaddr, 0);
 490         }
 491
 492   out:
 493         if (svc)
 494                 atomic_inc(&svc->usecnt);
 495         read_unlock(&__ip_vs_svc_lock);
 496
 497         IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
 498                       fwmark, ip_vs_proto_name(protocol),
 499                       IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
 500                       svc ? "hit" : "not hit");
 501
 502         return svc;
 503 }
 504
 505
 506 static inline void
 507 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
 508 {
 509         atomic_inc(&svc->refcnt);
 510         dest->svc = svc;
 511 }
 512
 513 static void
 514 __ip_vs_unbind_svc(struct ip_vs_dest *dest)
 515 {
 516         struct ip_vs_service *svc = dest->svc;
 517
 518         dest->svc = NULL;
 519         if (atomic_dec_and_test(&svc->refcnt)) {
 520                 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
 521                               svc->fwmark,
 522                               IP_VS_DBG_ADDR(svc->af, &svc->addr),
 523                               ntohs(svc->port), atomic_read(&svc->usecnt));
 524                 kfree(svc);
 525         }
 526 }
 527
 528
 529 /*
 530  *      Returns hash value for real service
 531  */
 532 static inline unsigned ip_vs_rs_hashkey(int af,
 533                                             const union nf_inet_addr *addr,
 534                                             __be16 port)
 535 {
 536         register unsigned porth = ntohs(port);
 537         __be32 addr_fold = addr->ip;
 538
 539 #ifdef CONFIG_IP_VS_IPV6
 540         if (af == AF_INET6)
 541                 addr_fold = addr->ip6[0]^addr->ip6[1]^
 542                             addr->ip6[2]^addr->ip6[3];
 543 #endif
 544
 545         return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
 546                 & IP_VS_RTAB_MASK;
 547 }
 548
 549 /*
 550  *      Hashes ip_vs_dest in ip_vs_rtable by <proto,addr,port>.
 551  *      should be called with locked tables.
 552  */
 553 static int ip_vs_rs_hash(struct ip_vs_dest *dest)
 554 {
 555         unsigned hash;
 556
 557         if (!list_empty(&dest->d_list)) {
 558                 return 0;
 559         }
 560
 561         /*
 562          *      Hash by proto,addr,port,
 563          *      which are the parameters of the real service.
 564          */
 565         hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
 566
 567         list_add(&dest->d_list, &ip_vs_rtable[hash]);
 568
 569         return 1;
 570 }
 571
 572 /*
 573  *      UNhashes ip_vs_dest from ip_vs_rtable.
 574  *      should be called with locked tables.
 575  */
 576 static int ip_vs_rs_unhash(struct ip_vs_dest *dest)
 577 {
 578         /*
 579          * Remove it from the ip_vs_rtable table.
 580          */
 581         if (!list_empty(&dest->d_list)) {
 582                 list_del(&dest->d_list);
 583                 INIT_LIST_HEAD(&dest->d_list);
 584         }
 585
 586         return 1;
 587 }
 588
 589 /*
 590  *      Lookup real service by <proto,addr,port> in the real service table.
 591  */
 592 struct ip_vs_dest *
 593 ip_vs_lookup_real_service(int af, __u16 protocol,
 594                           const union nf_inet_addr *daddr,
 595                           __be16 dport)
 596 {
 597         unsigned hash;
 598         struct ip_vs_dest *dest;
 599
 600         /*
 601          *      Check for "full" addressed entries
 602          *      Return the first found entry
 603          */
 604         hash = ip_vs_rs_hashkey(af, daddr, dport);
 605
 606         read_lock(&__ip_vs_rs_lock);
 607         list_for_each_entry(dest, &ip_vs_rtable[hash], d_list) {
 608                 if ((dest->af == af)
 609                     && ip_vs_addr_equal(af, &dest->addr, daddr)
 610                     && (dest->port == dport)
 611                     && ((dest->protocol == protocol) ||
 612                         dest->vfwmark)) {
 613                         /* HIT */
 614                         read_unlock(&__ip_vs_rs_lock);
 615                         return dest;
 616                 }
 617         }
 618         read_unlock(&__ip_vs_rs_lock);
 619
 620         return NULL;
 621 }
 622
 623 /*
 624  *      Lookup destination by {addr,port} in the given service
 625  */
 626 static struct ip_vs_dest *
 627 ip_vs_lookup_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
 628                   __be16 dport)
 629 {
 630         struct ip_vs_dest *dest;
 631
 632         /*
 633          * Find the destination for the given service
 634          */
 635         list_for_each_entry(dest, &svc->destinations, n_list) {
 636                 if ((dest->af == svc->af)
 637                     && ip_vs_addr_equal(svc->af, &dest->addr, daddr)
 638                     && (dest->port == dport)) {
 639                         /* HIT */
 640                         return dest;
 641                 }
 642         }
 643
 644         return NULL;
 645 }
 646
 647 /*
 648  * Find destination by {daddr,dport,vaddr,protocol}
 649  * Cretaed to be used in ip_vs_process_message() in
 650  * the backup synchronization daemon. It finds the
 651  * destination to be bound to the received connection
 652  * on the backup.
 653  *
 654  * ip_vs_lookup_real_service() looked promissing, but
 655  * seems not working as expected.
 656  */
 657 struct ip_vs_dest *ip_vs_find_dest(int af, const union nf_inet_addr *daddr,
 658                                    __be16 dport,
 659                                    const union nf_inet_addr *vaddr,
 660                                    __be16 vport, __u16 protocol)
 661 {
 662         struct ip_vs_dest *dest;
 663         struct ip_vs_service *svc;
 664
 665         svc = ip_vs_service_get(af, 0, protocol, vaddr, vport);
 666         if (!svc)
 667                 return NULL;
 668         dest = ip_vs_lookup_dest(svc, daddr, dport);
 669         if (dest)
 670                 atomic_inc(&dest->refcnt);
 671         ip_vs_service_put(svc);
 672         return dest;
 673 }
 674
 675 /*
 676  *  Lookup dest by {svc,addr,port} in the destination trash.
 677  *  The destination trash is used to hold the destinations that are removed
 678  *  from the service table but are still referenced by some conn entries.
 679  *  The reason to add the destination trash is when the dest is temporary
 680  *  down (either by administrator or by monitor program), the dest can be
 681  *  picked back from the trash, the remaining connections to the dest can
 682  *  continue, and the counting information of the dest is also useful for
 683  *  scheduling.
 684  */
 685 static struct ip_vs_dest *
 686 ip_vs_trash_get_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
 687                      __be16 dport)
 688 {
 689         struct ip_vs_dest *dest, *nxt;
 690
 691         /*
 692          * Find the destination in trash
 693          */
 694         list_for_each_entry_safe(dest, nxt, &ip_vs_dest_trash, n_list) {
 695                 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
 696                               "dest->refcnt=%d\n",
 697                               dest->vfwmark,
 698                               IP_VS_DBG_ADDR(svc->af, &dest->addr),
 699                               ntohs(dest->port),
 700                               atomic_read(&dest->refcnt));
 701                 if (dest->af == svc->af &&
 702                     ip_vs_addr_equal(svc->af, &dest->addr, daddr) &&
 703                     dest->port == dport &&
 704                     dest->vfwmark == svc->fwmark &&
 705                     dest->protocol == svc->protocol &&
 706                     (svc->fwmark ||
 707                      (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
 708                       dest->vport == svc->port))) {
 709                         /* HIT */
 710                         return dest;
 711                 }
 712
 713                 /*
 714                  * Try to purge the destination from trash if not referenced
 715                  */
 716                 if (atomic_read(&dest->refcnt) == 1) {
 717                         IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u "
 718                                       "from trash\n",
 719                                       dest->vfwmark,
 720                                       IP_VS_DBG_ADDR(svc->af, &dest->addr),
 721                                       ntohs(dest->port));
 722                         list_del(&dest->n_list);
 723                         ip_vs_dst_reset(dest);
 724                         __ip_vs_unbind_svc(dest);
 725                         kfree(dest);
 726                 }
 727         }
 728
 729         return NULL;
 730 }
 731
 732
 733 /*
 734  *  Clean up all the destinations in the trash
 735  *  Called by the ip_vs_control_cleanup()
 736  *
 737  *  When the ip_vs_control_clearup is activated by ipvs module exit,
 738  *  the service tables must have been flushed and all the connections
 739  *  are expired, and the refcnt of each destination in the trash must
 740  *  be 1, so we simply release them here.
 741  */
 742 static void ip_vs_trash_cleanup(void)
 743 {
 744         struct ip_vs_dest *dest, *nxt;
 745
 746         list_for_each_entry_safe(dest, nxt, &ip_vs_dest_trash, n_list) {
 747                 list_del(&dest->n_list);
 748                 ip_vs_dst_reset(dest);
 749                 __ip_vs_unbind_svc(dest);
 750                 kfree(dest);
 751         }
 752 }
 753
 754
 755 static void
 756 ip_vs_zero_stats(struct ip_vs_stats *stats)
 757 {
 758         spin_lock_bh(&stats->lock);
 759
 760         memset(&stats->ustats, 0, sizeof(stats->ustats));
 761         ip_vs_zero_estimator(stats);
 762
 763         spin_unlock_bh(&stats->lock);
 764 }
 765
 766 /*
 767  *      Update a destination in the given service
 768  */
 769 static void
 770 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
 771                     struct ip_vs_dest_user_kern *udest, int add)
 772 {
 773         int conn_flags;
 774
 775         /* set the weight and the flags */
 776         atomic_set(&dest->weight, udest->weight);
 777         conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
 778         conn_flags |= IP_VS_CONN_F_INACTIVE;
 779
 780         /* check if local node and update the flags */
 781 #ifdef CONFIG_IP_VS_IPV6
 782         if (svc->af == AF_INET6) {
 783                 if (__ip_vs_addr_is_local_v6(&udest->addr.in6)) {
 784                         conn_flags = (conn_flags & ~IP_VS_CONN_F_FWD_MASK)
 785                                 | IP_VS_CONN_F_LOCALNODE;
 786                 }
 787         } else
 788 #endif
 789                 if (inet_addr_type(&init_net, udest->addr.ip) == RTN_LOCAL) {
 790                         conn_flags = (conn_flags & ~IP_VS_CONN_F_FWD_MASK)
 791                                 | IP_VS_CONN_F_LOCALNODE;
 792                 }
 793
 794         /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
 795         if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
 796                 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
 797         } else {
 798                 /*
 799                  *    Put the real service in ip_vs_rtable if not present.
 800                  *    For now only for NAT!
 801                  */
 802                 write_lock_bh(&__ip_vs_rs_lock);
 803                 ip_vs_rs_hash(dest);
 804                 write_unlock_bh(&__ip_vs_rs_lock);
 805         }
 806         atomic_set(&dest->conn_flags, conn_flags);
 807
 808         /* bind the service */
 809         if (!dest->svc) {
 810                 __ip_vs_bind_svc(dest, svc);
 811         } else {
 812                 if (dest->svc != svc) {
 813                         __ip_vs_unbind_svc(dest);
 814                         ip_vs_zero_stats(&dest->stats);
 815                         __ip_vs_bind_svc(dest, svc);
 816                 }
 817         }
 818
 819         /* set the dest status flags */
 820         dest->flags |= IP_VS_DEST_F_AVAILABLE;
 821
 822         if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
 823                 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
 824         dest->u_threshold = udest->u_threshold;
 825         dest->l_threshold = udest->l_threshold;
 826
 827         if (add)
 828                 ip_vs_new_estimator(&dest->stats);
 829
 830         write_lock_bh(&__ip_vs_svc_lock);
 831
 832         /* Wait until all other svc users go away */
 833         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
 834
 835         if (add) {
 836                 list_add(&dest->n_list, &svc->destinations);
 837                 svc->num_dests++;
 838         }
 839
 840         /* call the update_service, because server weight may be changed */
 841         if (svc->scheduler->update_service)
 842                 svc->scheduler->update_service(svc);
 843
 844         write_unlock_bh(&__ip_vs_svc_lock);
 845 }
 846
 847
 848 /*
 849  *      Create a destination for the given service
 850  */
 851 static int
 852 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
 853                struct ip_vs_dest **dest_p)
 854 {
 855         struct ip_vs_dest *dest;
 856         unsigned atype;
 857
 858         EnterFunction(2);
 859
 860 #ifdef CONFIG_IP_VS_IPV6
 861         if (svc->af == AF_INET6) {
 862                 atype = ipv6_addr_type(&udest->addr.in6);
 863                 if ((!(atype & IPV6_ADDR_UNICAST) ||
 864                         atype & IPV6_ADDR_LINKLOCAL) &&
 865                         !__ip_vs_addr_is_local_v6(&udest->addr.in6))
 866                         return -EINVAL;
 867         } else
 868 #endif
 869         {
 870                 atype = inet_addr_type(&init_net, udest->addr.ip);
 871                 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
 872                         return -EINVAL;
 873         }
 874
 875         dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
 876         if (dest == NULL) {
 877                 pr_err("%s(): no memory.\n", __func__);
 878                 return -ENOMEM;
 879         }
 880
 881         dest->af = svc->af;
 882         dest->protocol = svc->protocol;
 883         dest->vaddr = svc->addr;
 884         dest->vport = svc->port;
 885         dest->vfwmark = svc->fwmark;
 886         ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
 887         dest->port = udest->port;
 888
 889         atomic_set(&dest->activeconns, 0);
 890         atomic_set(&dest->inactconns, 0);
 891         atomic_set(&dest->persistconns, 0);
 892         atomic_set(&dest->refcnt, 1);
 893
 894         INIT_LIST_HEAD(&dest->d_list);
 895         spin_lock_init(&dest->dst_lock);
 896         spin_lock_init(&dest->stats.lock);
 897         __ip_vs_update_dest(svc, dest, udest, 1);
 898
 899         *dest_p = dest;
 900
 901         LeaveFunction(2);
 902         return 0;
 903 }
 904
 905
 906 /*
 907  *      Add a destination into an existing service
 908  */
 909 static int
 910 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
 911 {
 912         struct ip_vs_dest *dest;
 913         union nf_inet_addr daddr;
 914         __be16 dport = udest->port;
 915         int ret;
 916
 917         EnterFunction(2);
 918
 919         if (udest->weight < 0) {
 920                 pr_err("%s(): server weight less than zero\n", __func__);
 921                 return -ERANGE;
 922         }
 923
 924         if (udest->l_threshold > udest->u_threshold) {
 925                 pr_err("%s(): lower threshold is higher than upper threshold\n",
 926                         __func__);
 927                 return -ERANGE;
 928         }
 929
 930         ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
 931
 932         /*
 933          * Check if the dest already exists in the list
 934          */
 935         dest = ip_vs_lookup_dest(svc, &daddr, dport);
 936
 937         if (dest != NULL) {
 938                 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
 939                 return -EEXIST;
 940         }
 941
 942         /*
 943          * Check if the dest already exists in the trash and
 944          * is from the same service
 945          */
 946         dest = ip_vs_trash_get_dest(svc, &daddr, dport);
 947
 948         if (dest != NULL) {
 949                 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
 950                               "dest->refcnt=%d, service %u/%s:%u\n",
 951                               IP_VS_DBG_ADDR(svc->af, &daddr), ntohs(dport),
 952                               atomic_read(&dest->refcnt),
 953                               dest->vfwmark,
 954                               IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
 955                               ntohs(dest->vport));
 956
 957                 /*
 958                  * Get the destination from the trash
 959                  */
 960                 list_del(&dest->n_list);
 961
 962                 __ip_vs_update_dest(svc, dest, udest, 1);
 963                 ret = 0;
 964         } else {
 965                 /*
 966                  * Allocate and initialize the dest structure
 967                  */
 968                 ret = ip_vs_new_dest(svc, udest, &dest);
 969         }
 970         LeaveFunction(2);
 971
 972         return ret;
 973 }
 974
 975
 976 /*
 977  *      Edit a destination in the given service
 978  */
 979 static int
 980 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
 981 {
 982         struct ip_vs_dest *dest;
 983         union nf_inet_addr daddr;
 984         __be16 dport = udest->port;
 985
 986         EnterFunction(2);
 987
 988         if (udest->weight < 0) {
 989                 pr_err("%s(): server weight less than zero\n", __func__);
 990                 return -ERANGE;
 991         }
 992
 993         if (udest->l_threshold > udest->u_threshold) {
 994                 pr_err("%s(): lower threshold is higher than upper threshold\n",
 995                         __func__);
 996                 return -ERANGE;
 997         }
 998
 999         ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
1000
1001         /*
1002          *  Lookup the destination list
1003          */
1004         dest = ip_vs_lookup_dest(svc, &daddr, dport);
1005
1006         if (dest == NULL) {
1007                 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1008                 return -ENOENT;
1009         }
1010
1011         __ip_vs_update_dest(svc, dest, udest, 0);
1012         LeaveFunction(2);
1013
1014         return 0;
1015 }
1016
1017
1018 /*
1019  *      Delete a destination (must be already unlinked from the service)
1020  */
1021 static void __ip_vs_del_dest(struct ip_vs_dest *dest)
1022 {
1023         ip_vs_kill_estimator(&dest->stats);
1024
1025         /*
1026          *  Remove it from the d-linked list with the real services.
1027          */
1028         write_lock_bh(&__ip_vs_rs_lock);
1029         ip_vs_rs_unhash(dest);
1030         write_unlock_bh(&__ip_vs_rs_lock);
1031
1032         /*
1033          *  Decrease the refcnt of the dest, and free the dest
1034          *  if nobody refers to it (refcnt=0). Otherwise, throw
1035          *  the destination into the trash.
1036          */
1037         if (atomic_dec_and_test(&dest->refcnt)) {
1038                 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u\n",
1039                               dest->vfwmark,
1040                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1041                               ntohs(dest->port));
1042                 ip_vs_dst_reset(dest);
1043                 /* simply decrease svc->refcnt here, let the caller check
1044                    and release the service if nobody refers to it.
1045                    Only user context can release destination and service,
1046                    and only one user context can update virtual service at a
1047                    time, so the operation here is OK */
1048                 atomic_dec(&dest->svc->refcnt);
1049                 kfree(dest);
1050         } else {
1051                 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, "
1052                               "dest->refcnt=%d\n",
1053                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1054                               ntohs(dest->port),
1055                               atomic_read(&dest->refcnt));
1056                 list_add(&dest->n_list, &ip_vs_dest_trash);
1057                 atomic_inc(&dest->refcnt);
1058         }
1059 }
1060
1061
1062 /*
1063  *      Unlink a destination from the given service
1064  */
1065 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1066                                 struct ip_vs_dest *dest,
1067                                 int svcupd)
1068 {
1069         dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1070
1071         /*
1072          *  Remove it from the d-linked destination list.
1073          */
1074         list_del(&dest->n_list);
1075         svc->num_dests--;
1076
1077         /*
1078          *  Call the update_service function of its scheduler
1079          */
1080         if (svcupd && svc->scheduler->update_service)
1081                         svc->scheduler->update_service(svc);
1082 }
1083
1084
1085 /*
1086  *      Delete a destination server in the given service
1087  */
1088 static int
1089 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1090 {
1091         struct ip_vs_dest *dest;
1092         __be16 dport = udest->port;
1093
1094         EnterFunction(2);
1095
1096         dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
1097
1098         if (dest == NULL) {
1099                 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1100                 return -ENOENT;
1101         }
1102
1103         write_lock_bh(&__ip_vs_svc_lock);
1104
1105         /*
1106          *      Wait until all other svc users go away.
1107          */
1108         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1109
1110         /*
1111          *      Unlink dest from the service
1112          */
1113         __ip_vs_unlink_dest(svc, dest, 1);
1114
1115         write_unlock_bh(&__ip_vs_svc_lock);
1116
1117         /*
1118          *      Delete the destination
1119          */
1120         __ip_vs_del_dest(dest);
1121
1122         LeaveFunction(2);
1123
1124         return 0;
1125 }
1126
1127
1128 /*
1129  *      Add a service into the service hash table
1130  */
1131 static int
1132 ip_vs_add_service(struct ip_vs_service_user_kern *u,
1133                   struct ip_vs_service **svc_p)
1134 {
1135         int ret = 0;
1136         struct ip_vs_scheduler *sched = NULL;
1137         struct ip_vs_service *svc = NULL;
1138
1139         /* increase the module use count */
1140         ip_vs_use_count_inc();
1141
1142         /* Lookup the scheduler by 'u->sched_name' */
1143         sched = ip_vs_scheduler_get(u->sched_name);
1144         if (sched == NULL) {
1145                 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1146                 ret = -ENOENT;
1147                 goto out_err;
1148         }
1149
1150 #ifdef CONFIG_IP_VS_IPV6
1151         if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1152                 ret = -EINVAL;
1153                 goto out_err;
1154         }
1155 #endif
1156
1157         svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1158         if (svc == NULL) {
1159                 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1160                 ret = -ENOMEM;
1161                 goto out_err;
1162         }
1163
1164         /* I'm the first user of the service */
1165         atomic_set(&svc->usecnt, 0);
1166         atomic_set(&svc->refcnt, 0);
1167
1168         svc->af = u->af;
1169         svc->protocol = u->protocol;
1170         ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1171         svc->port = u->port;
1172         svc->fwmark = u->fwmark;
1173         svc->flags = u->flags;
1174         svc->timeout = u->timeout * HZ;
1175         svc->netmask = u->netmask;
1176
1177         INIT_LIST_HEAD(&svc->destinations);
1178         rwlock_init(&svc->sched_lock);
1179         spin_lock_init(&svc->stats.lock);
1180
1181         /* Bind the scheduler */
1182         ret = ip_vs_bind_scheduler(svc, sched);
1183         if (ret)
1184                 goto out_err;
1185         sched = NULL;
1186
1187         /* Update the virtual service counters */
1188         if (svc->port == FTPPORT)
1189                 atomic_inc(&ip_vs_ftpsvc_counter);
1190         else if (svc->port == 0)
1191                 atomic_inc(&ip_vs_nullsvc_counter);
1192
1193         ip_vs_new_estimator(&svc->stats);
1194
1195         /* Count only IPv4 services for old get/setsockopt interface */
1196         if (svc->af == AF_INET)
1197                 ip_vs_num_services++;
1198
1199         /* Hash the service into the service table */
1200         write_lock_bh(&__ip_vs_svc_lock);
1201         ip_vs_svc_hash(svc);
1202         write_unlock_bh(&__ip_vs_svc_lock);
1203
1204         *svc_p = svc;
1205         return 0;
1206
1207  out_err:
1208         if (svc != NULL) {
1209                 ip_vs_unbind_scheduler(svc);
1210                 if (svc->inc) {
1211                         local_bh_disable();
1212                         ip_vs_app_inc_put(svc->inc);
1213                         local_bh_enable();
1214                 }
1215                 kfree(svc);
1216         }
1217         ip_vs_scheduler_put(sched);
1218
1219         /* decrease the module use count */
1220         ip_vs_use_count_dec();
1221
1222         return ret;
1223 }
1224
1225
1226 /*
1227  *      Edit a service and bind it with a new scheduler
1228  */
1229 static int
1230 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1231 {
1232         struct ip_vs_scheduler *sched, *old_sched;
1233         int ret = 0;
1234
1235         /*
1236          * Lookup the scheduler, by 'u->sched_name'
1237          */
1238         sched = ip_vs_scheduler_get(u->sched_name);
1239         if (sched == NULL) {
1240                 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1241                 return -ENOENT;
1242         }
1243         old_sched = sched;
1244
1245 #ifdef CONFIG_IP_VS_IPV6
1246         if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1247                 ret = -EINVAL;
1248                 goto out;
1249         }
1250 #endif
1251
1252         write_lock_bh(&__ip_vs_svc_lock);
1253
1254         /*
1255          * Wait until all other svc users go away.
1256          */
1257         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1258
1259         /*
1260          * Set the flags and timeout value
1261          */
1262         svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1263         svc->timeout = u->timeout * HZ;
1264         svc->netmask = u->netmask;
1265
1266         old_sched = svc->scheduler;
1267         if (sched != old_sched) {
1268                 /*
1269                  * Unbind the old scheduler
1270                  */
1271                 if ((ret = ip_vs_unbind_scheduler(svc))) {
1272                         old_sched = sched;
1273                         goto out_unlock;
1274                 }
1275
1276                 /*
1277                  * Bind the new scheduler
1278                  */
1279                 if ((ret = ip_vs_bind_scheduler(svc, sched))) {
1280                         /*
1281                          * If ip_vs_bind_scheduler fails, restore the old
1282                          * scheduler.
1283                          * The main reason of failure is out of memory.
1284                          *
1285                          * The question is if the old scheduler can be
1286                          * restored all the time. TODO: if it cannot be
1287                          * restored some time, we must delete the service,
1288                          * otherwise the system may crash.
1289                          */
1290                         ip_vs_bind_scheduler(svc, old_sched);
1291                         old_sched = sched;
1292                         goto out_unlock;
1293                 }
1294         }
1295
1296   out_unlock:
1297         write_unlock_bh(&__ip_vs_svc_lock);
1298 #ifdef CONFIG_IP_VS_IPV6
1299   out:
1300 #endif
1301         ip_vs_scheduler_put(old_sched);
1302         return ret;
1303 }
1304
1305
1306 /*
1307  *      Delete a service from the service list
1308  *      - The service must be unlinked, unlocked and not referenced!
1309  *      - We are called under _bh lock
1310  */
1311 static void __ip_vs_del_service(struct ip_vs_service *svc)
1312 {
1313         struct ip_vs_dest *dest, *nxt;
1314         struct ip_vs_scheduler *old_sched;
1315
1316         /* Count only IPv4 services for old get/setsockopt interface */
1317         if (svc->af == AF_INET)
1318                 ip_vs_num_services--;
1319
1320         ip_vs_kill_estimator(&svc->stats);
1321
1322         /* Unbind scheduler */
1323         old_sched = svc->scheduler;
1324         ip_vs_unbind_scheduler(svc);
1325         ip_vs_scheduler_put(old_sched);
1326
1327         /* Unbind app inc */
1328         if (svc->inc) {
1329                 ip_vs_app_inc_put(svc->inc);
1330                 svc->inc = NULL;
1331         }
1332
1333         /*
1334          *    Unlink the whole destination list
1335          */
1336         list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1337                 __ip_vs_unlink_dest(svc, dest, 0);
1338                 __ip_vs_del_dest(dest);
1339         }
1340
1341         /*
1342          *    Update the virtual service counters
1343          */
1344         if (svc->port == FTPPORT)
1345                 atomic_dec(&ip_vs_ftpsvc_counter);
1346         else if (svc->port == 0)
1347                 atomic_dec(&ip_vs_nullsvc_counter);
1348
1349         /*
1350          *    Free the service if nobody refers to it
1351          */
1352         if (atomic_read(&svc->refcnt) == 0) {
1353                 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
1354                               svc->fwmark,
1355                               IP_VS_DBG_ADDR(svc->af, &svc->addr),
1356                               ntohs(svc->port), atomic_read(&svc->usecnt));
1357                 kfree(svc);
1358         }
1359
1360         /* decrease the module use count */
1361         ip_vs_use_count_dec();
1362 }
1363
1364 /*
1365  * Unlink a service from list and try to delete it if its refcnt reached 0
1366  */
1367 static void ip_vs_unlink_service(struct ip_vs_service *svc)
1368 {
1369         /*
1370          * Unhash it from the service table
1371          */
1372         write_lock_bh(&__ip_vs_svc_lock);
1373
1374         ip_vs_svc_unhash(svc);
1375
1376         /*
1377          * Wait until all the svc users go away.
1378          */
1379         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1380
1381         __ip_vs_del_service(svc);
1382
1383         write_unlock_bh(&__ip_vs_svc_lock);
1384 }
1385
1386 /*
1387  *      Delete a service from the service list
1388  */
1389 static int ip_vs_del_service(struct ip_vs_service *svc)
1390 {
1391         if (svc == NULL)
1392                 return -EEXIST;
1393         ip_vs_unlink_service(svc);
1394
1395         return 0;
1396 }
1397
1398
1399 /*
1400  *      Flush all the virtual services
1401  */
1402 static int ip_vs_flush(void)
1403 {
1404         int idx;
1405         struct ip_vs_service *svc, *nxt;
1406
1407         /*
1408          * Flush the service table hashed by <protocol,addr,port>
1409          */
1410         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1411                 list_for_each_entry_safe(svc, nxt, &ip_vs_svc_table[idx], s_list) {
1412                         ip_vs_unlink_service(svc);
1413                 }
1414         }
1415
1416         /*
1417          * Flush the service table hashed by fwmark
1418          */
1419         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1420                 list_for_each_entry_safe(svc, nxt,
1421                                          &ip_vs_svc_fwm_table[idx], f_list) {
1422                         ip_vs_unlink_service(svc);
1423                 }
1424         }
1425
1426         return 0;
1427 }
1428
1429
1430 /*
1431  *      Zero counters in a service or all services
1432  */
1433 static int ip_vs_zero_service(struct ip_vs_service *svc)
1434 {
1435         struct ip_vs_dest *dest;
1436
1437         write_lock_bh(&__ip_vs_svc_lock);
1438         list_for_each_entry(dest, &svc->destinations, n_list) {
1439                 ip_vs_zero_stats(&dest->stats);
1440         }
1441         ip_vs_zero_stats(&svc->stats);
1442         write_unlock_bh(&__ip_vs_svc_lock);
1443         return 0;
1444 }
1445
1446 static int ip_vs_zero_all(void)
1447 {
1448         int idx;
1449         struct ip_vs_service *svc;
1450
1451         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1452                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1453                         ip_vs_zero_service(svc);
1454                 }
1455         }
1456
1457         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1458                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1459                         ip_vs_zero_service(svc);
1460                 }
1461         }
1462
1463         ip_vs_zero_stats(&ip_vs_stats);
1464         return 0;
1465 }
1466
1467
1468 static int
1469 proc_do_defense_mode(ctl_table *table, int write,
1470                      void __user *buffer, size_t *lenp, loff_t *ppos)
1471 {
1472         int *valp = table->data;
1473         int val = *valp;
1474         int rc;
1475
1476         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1477         if (write && (*valp != val)) {
1478                 if ((*valp < 0) || (*valp > 3)) {
1479                         /* Restore the correct value */
1480                         *valp = val;
1481                 } else {
1482                         update_defense_level();
1483                 }
1484         }
1485         return rc;
1486 }
1487
1488
1489 static int
1490 proc_do_sync_threshold(ctl_table *table, int write,
1491                        void __user *buffer, size_t *lenp, loff_t *ppos)
1492 {
1493         int *valp = table->data;
1494         int val[2];
1495         int rc;
1496
1497         /* backup the value first */
1498         memcpy(val, valp, sizeof(val));
1499
1500         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1501         if (write && (valp[0] < 0 || valp[1] < 0 || valp[0] >= valp[1])) {
1502                 /* Restore the correct value */
1503                 memcpy(valp, val, sizeof(val));
1504         }
1505         return rc;
1506 }
1507
1508
1509 /*
1510  *      IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1511  */
1512
1513 static struct ctl_table vs_vars[] = {
1514         {
1515                 .procname       = "amemthresh",
1516                 .data           = &sysctl_ip_vs_amemthresh,
1517                 .maxlen         = sizeof(int),
1518                 .mode           = 0644,
1519                 .proc_handler   = proc_dointvec,
1520         },
1521 #ifdef CONFIG_IP_VS_DEBUG
1522         {
1523                 .procname       = "debug_level",
1524                 .data           = &sysctl_ip_vs_debug_level,
1525                 .maxlen         = sizeof(int),
1526                 .mode           = 0644,
1527                 .proc_handler   = proc_dointvec,
1528         },
1529 #endif
1530         {
1531                 .procname       = "am_droprate",
1532                 .data           = &sysctl_ip_vs_am_droprate,
1533                 .maxlen         = sizeof(int),
1534                 .mode           = 0644,
1535                 .proc_handler   = proc_dointvec,
1536         },
1537         {
1538                 .procname       = "drop_entry",
1539                 .data           = &sysctl_ip_vs_drop_entry,
1540                 .maxlen         = sizeof(int),
1541                 .mode           = 0644,
1542                 .proc_handler   = proc_do_defense_mode,
1543         },
1544         {
1545                 .procname       = "drop_packet",
1546                 .data           = &sysctl_ip_vs_drop_packet,
1547                 .maxlen         = sizeof(int),
1548                 .mode           = 0644,
1549                 .proc_handler   = proc_do_defense_mode,
1550         },
1551 #ifdef CONFIG_IP_VS_NFCT
1552         {
1553                 .procname       = "conntrack",
1554                 .data           = &sysctl_ip_vs_conntrack,
1555                 .maxlen         = sizeof(int),
1556                 .mode           = 0644,
1557                 .proc_handler   = &proc_dointvec,
1558         },
1559 #endif
1560         {
1561                 .procname       = "secure_tcp",
1562                 .data           = &sysctl_ip_vs_secure_tcp,
1563                 .maxlen         = sizeof(int),
1564                 .mode           = 0644,
1565                 .proc_handler   = proc_do_defense_mode,
1566         },
1567         {
1568                 .procname       = "snat_reroute",
1569                 .data           = &sysctl_ip_vs_snat_reroute,
1570                 .maxlen         = sizeof(int),
1571                 .mode           = 0644,
1572                 .proc_handler   = &proc_dointvec,
1573         },
1574 #if 0
1575         {
1576                 .procname       = "timeout_established",
1577                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1578                 .maxlen         = sizeof(int),
1579                 .mode           = 0644,
1580                 .proc_handler   = proc_dointvec_jiffies,
1581         },
1582         {
1583                 .procname       = "timeout_synsent",
1584                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1585                 .maxlen         = sizeof(int),
1586                 .mode           = 0644,
1587                 .proc_handler   = proc_dointvec_jiffies,
1588         },
1589         {
1590                 .procname       = "timeout_synrecv",
1591                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1592                 .maxlen         = sizeof(int),
1593                 .mode           = 0644,
1594                 .proc_handler   = proc_dointvec_jiffies,
1595         },
1596         {
1597                 .procname       = "timeout_finwait",
1598                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1599                 .maxlen         = sizeof(int),
1600                 .mode           = 0644,
1601                 .proc_handler   = proc_dointvec_jiffies,
1602         },
1603         {
1604                 .procname       = "timeout_timewait",
1605                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1606                 .maxlen         = sizeof(int),
1607                 .mode           = 0644,
1608                 .proc_handler   = proc_dointvec_jiffies,
1609         },
1610         {
1611                 .procname       = "timeout_close",
1612                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1613                 .maxlen         = sizeof(int),
1614                 .mode           = 0644,
1615                 .proc_handler   = proc_dointvec_jiffies,
1616         },
1617         {
1618                 .procname       = "timeout_closewait",
1619                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1620                 .maxlen         = sizeof(int),
1621                 .mode           = 0644,
1622                 .proc_handler   = proc_dointvec_jiffies,
1623         },
1624         {
1625                 .procname       = "timeout_lastack",
1626                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1627                 .maxlen         = sizeof(int),
1628                 .mode           = 0644,
1629                 .proc_handler   = proc_dointvec_jiffies,
1630         },
1631         {
1632                 .procname       = "timeout_listen",
1633                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1634                 .maxlen         = sizeof(int),
1635                 .mode           = 0644,
1636                 .proc_handler   = proc_dointvec_jiffies,
1637         },
1638         {
1639                 .procname       = "timeout_synack",
1640                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1641                 .maxlen         = sizeof(int),
1642                 .mode           = 0644,
1643                 .proc_handler   = proc_dointvec_jiffies,
1644         },
1645         {
1646                 .procname       = "timeout_udp",
1647                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1648                 .maxlen         = sizeof(int),
1649                 .mode           = 0644,
1650                 .proc_handler   = proc_dointvec_jiffies,
1651         },
1652         {
1653                 .procname       = "timeout_icmp",
1654                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1655                 .maxlen         = sizeof(int),
1656                 .mode           = 0644,
1657                 .proc_handler   = proc_dointvec_jiffies,
1658         },
1659 #endif
1660         {
1661                 .procname       = "cache_bypass",
1662                 .data           = &sysctl_ip_vs_cache_bypass,
1663                 .maxlen         = sizeof(int),
1664                 .mode           = 0644,
1665                 .proc_handler   = proc_dointvec,
1666         },
1667         {
1668                 .procname       = "expire_nodest_conn",
1669                 .data           = &sysctl_ip_vs_expire_nodest_conn,
1670                 .maxlen         = sizeof(int),
1671                 .mode           = 0644,
1672                 .proc_handler   = proc_dointvec,
1673         },
1674         {
1675                 .procname       = "expire_quiescent_template",
1676                 .data           = &sysctl_ip_vs_expire_quiescent_template,
1677                 .maxlen         = sizeof(int),
1678                 .mode           = 0644,
1679                 .proc_handler   = proc_dointvec,
1680         },
1681         {
1682                 .procname       = "sync_threshold",
1683                 .data           = &sysctl_ip_vs_sync_threshold,
1684                 .maxlen         = sizeof(sysctl_ip_vs_sync_threshold),
1685                 .mode           = 0644,
1686                 .proc_handler   = proc_do_sync_threshold,
1687         },
1688         {
1689                 .procname       = "nat_icmp_send",
1690                 .data           = &sysctl_ip_vs_nat_icmp_send,
1691                 .maxlen         = sizeof(int),
1692                 .mode           = 0644,
1693                 .proc_handler   = proc_dointvec,
1694         },
1695         { }
1696 };
1697
1698 const struct ctl_path net_vs_ctl_path[] = {
1699         { .procname = "net", },
1700         { .procname = "ipv4", },
1701         { .procname = "vs", },
1702         { }
1703 };
1704 EXPORT_SYMBOL_GPL(net_vs_ctl_path);
1705
1706 static struct ctl_table_header * sysctl_header;
1707
1708 #ifdef CONFIG_PROC_FS
1709
1710 struct ip_vs_iter {
1711         struct list_head *table;
1712         int bucket;
1713 };
1714
1715 /*
1716  *      Write the contents of the VS rule table to a PROCfs file.
1717  *      (It is kept just for backward compatibility)
1718  */
1719 static inline const char *ip_vs_fwd_name(unsigned flags)
1720 {
1721         switch (flags & IP_VS_CONN_F_FWD_MASK) {
1722         case IP_VS_CONN_F_LOCALNODE:
1723                 return "Local";
1724         case IP_VS_CONN_F_TUNNEL:
1725                 return "Tunnel";
1726         case IP_VS_CONN_F_DROUTE:
1727                 return "Route";
1728         default:
1729                 return "Masq";
1730         }
1731 }
1732
1733
1734 /* Get the Nth entry in the two lists */
1735 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1736 {
1737         struct ip_vs_iter *iter = seq->private;
1738         int idx;
1739         struct ip_vs_service *svc;
1740
1741         /* look in hash by protocol */
1742         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1743                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1744                         if (pos-- == 0){
1745                                 iter->table = ip_vs_svc_table;
1746                                 iter->bucket = idx;
1747                                 return svc;
1748                         }
1749                 }
1750         }
1751
1752         /* keep looking in fwmark */
1753         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1754                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1755                         if (pos-- == 0) {
1756                                 iter->table = ip_vs_svc_fwm_table;
1757                                 iter->bucket = idx;
1758                                 return svc;
1759                         }
1760                 }
1761         }
1762
1763         return NULL;
1764 }
1765
1766 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1767 __acquires(__ip_vs_svc_lock)
1768 {
1769
1770         read_lock_bh(&__ip_vs_svc_lock);
1771         return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1772 }
1773
1774
1775 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1776 {
1777         struct list_head *e;
1778         struct ip_vs_iter *iter;
1779         struct ip_vs_service *svc;
1780
1781         ++*pos;
1782         if (v == SEQ_START_TOKEN)
1783                 return ip_vs_info_array(seq,0);
1784
1785         svc = v;
1786         iter = seq->private;
1787
1788         if (iter->table == ip_vs_svc_table) {
1789                 /* next service in table hashed by protocol */
1790                 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
1791                         return list_entry(e, struct ip_vs_service, s_list);
1792
1793
1794                 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1795                         list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
1796                                             s_list) {
1797                                 return svc;
1798                         }
1799                 }
1800
1801                 iter->table = ip_vs_svc_fwm_table;
1802                 iter->bucket = -1;
1803                 goto scan_fwmark;
1804         }
1805
1806         /* next service in hashed by fwmark */
1807         if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
1808                 return list_entry(e, struct ip_vs_service, f_list);
1809
1810  scan_fwmark:
1811         while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1812                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
1813                                     f_list)
1814                         return svc;
1815         }
1816
1817         return NULL;
1818 }
1819
1820 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1821 __releases(__ip_vs_svc_lock)
1822 {
1823         read_unlock_bh(&__ip_vs_svc_lock);
1824 }
1825
1826
1827 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1828 {
1829         if (v == SEQ_START_TOKEN) {
1830                 seq_printf(seq,
1831                         "IP Virtual Server version %d.%d.%d (size=%d)\n",
1832                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1833                 seq_puts(seq,
1834                          "Prot LocalAddress:Port Scheduler Flags\n");
1835                 seq_puts(seq,
1836                          "  -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1837         } else {
1838                 const struct ip_vs_service *svc = v;
1839                 const struct ip_vs_iter *iter = seq->private;
1840                 const struct ip_vs_dest *dest;
1841
1842                 if (iter->table == ip_vs_svc_table) {
1843 #ifdef CONFIG_IP_VS_IPV6
1844                         if (svc->af == AF_INET6)
1845                                 seq_printf(seq, "%s  [%pI6]:%04X %s ",
1846                                            ip_vs_proto_name(svc->protocol),
1847                                            &svc->addr.in6,
1848                                            ntohs(svc->port),
1849                                            svc->scheduler->name);
1850                         else
1851 #endif
1852                                 seq_printf(seq, "%s  %08X:%04X %s %s ",
1853                                            ip_vs_proto_name(svc->protocol),
1854                                            ntohl(svc->addr.ip),
1855                                            ntohs(svc->port),
1856                                            svc->scheduler->name,
1857                                            (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
1858                 } else {
1859                         seq_printf(seq, "FWM  %08X %s %s",
1860                                    svc->fwmark, svc->scheduler->name,
1861                                    (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
1862                 }
1863
1864                 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
1865                         seq_printf(seq, "persistent %d %08X\n",
1866                                 svc->timeout,
1867                                 ntohl(svc->netmask));
1868                 else
1869                         seq_putc(seq, '\n');
1870
1871                 list_for_each_entry(dest, &svc->destinations, n_list) {
1872 #ifdef CONFIG_IP_VS_IPV6
1873                         if (dest->af == AF_INET6)
1874                                 seq_printf(seq,
1875                                            "  -> [%pI6]:%04X"
1876                                            "      %-7s %-6d %-10d %-10d\n",
1877                                            &dest->addr.in6,
1878                                            ntohs(dest->port),
1879                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1880                                            atomic_read(&dest->weight),
1881                                            atomic_read(&dest->activeconns),
1882                                            atomic_read(&dest->inactconns));
1883                         else
1884 #endif
1885                                 seq_printf(seq,
1886                                            "  -> %08X:%04X      "
1887                                            "%-7s %-6d %-10d %-10d\n",
1888                                            ntohl(dest->addr.ip),
1889                                            ntohs(dest->port),
1890                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1891                                            atomic_read(&dest->weight),
1892                                            atomic_read(&dest->activeconns),
1893                                            atomic_read(&dest->inactconns));
1894
1895                 }
1896         }
1897         return 0;
1898 }
1899
1900 static const struct seq_operations ip_vs_info_seq_ops = {
1901         .start = ip_vs_info_seq_start,
1902         .next  = ip_vs_info_seq_next,
1903         .stop  = ip_vs_info_seq_stop,
1904         .show  = ip_vs_info_seq_show,
1905 };
1906
1907 static int ip_vs_info_open(struct inode *inode, struct file *file)
1908 {
1909         return seq_open_private(file, &ip_vs_info_seq_ops,
1910                         sizeof(struct ip_vs_iter));
1911 }
1912
1913 static const struct file_operations ip_vs_info_fops = {
1914         .owner   = THIS_MODULE,
1915         .open    = ip_vs_info_open,
1916         .read    = seq_read,
1917         .llseek  = seq_lseek,
1918         .release = seq_release_private,
1919 };
1920
1921 #endif
1922
1923 struct ip_vs_stats ip_vs_stats = {
1924         .lock = __SPIN_LOCK_UNLOCKED(ip_vs_stats.lock),
1925 };
1926
1927 #ifdef CONFIG_PROC_FS
1928 static int ip_vs_stats_show(struct seq_file *seq, void *v)
1929 {
1930
1931 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
1932         seq_puts(seq,
1933                  "   Total Incoming Outgoing         Incoming         Outgoing\n");
1934         seq_printf(seq,
1935                    "   Conns  Packets  Packets            Bytes            Bytes\n");
1936
1937         spin_lock_bh(&ip_vs_stats.lock);
1938         seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", ip_vs_stats.ustats.conns,
1939                    ip_vs_stats.ustats.inpkts, ip_vs_stats.ustats.outpkts,
1940                    (unsigned long long) ip_vs_stats.ustats.inbytes,
1941                    (unsigned long long) ip_vs_stats.ustats.outbytes);
1942
1943 /*                 01234567 01234567 01234567 0123456701234567 0123456701234567 */
1944         seq_puts(seq,
1945                    " Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
1946         seq_printf(seq,"%8X %8X %8X %16X %16X\n",
1947                         ip_vs_stats.ustats.cps,
1948                         ip_vs_stats.ustats.inpps,
1949                         ip_vs_stats.ustats.outpps,
1950                         ip_vs_stats.ustats.inbps,
1951                         ip_vs_stats.ustats.outbps);
1952         spin_unlock_bh(&ip_vs_stats.lock);
1953
1954         return 0;
1955 }
1956
1957 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
1958 {
1959         return single_open(file, ip_vs_stats_show, NULL);
1960 }
1961
1962 static const struct file_operations ip_vs_stats_fops = {
1963         .owner = THIS_MODULE,
1964         .open = ip_vs_stats_seq_open,
1965         .read = seq_read,
1966         .llseek = seq_lseek,
1967         .release = single_release,
1968 };
1969
1970 #endif
1971
1972 /*
1973  *      Set timeout values for tcp tcpfin udp in the timeout_table.
1974  */
1975 static int ip_vs_set_timeout(struct ip_vs_timeout_user *u)
1976 {
1977         IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
1978                   u->tcp_timeout,
1979                   u->tcp_fin_timeout,
1980                   u->udp_timeout);
1981
1982 #ifdef CONFIG_IP_VS_PROTO_TCP
1983         if (u->tcp_timeout) {
1984                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED]
1985                         = u->tcp_timeout * HZ;
1986         }
1987
1988         if (u->tcp_fin_timeout) {
1989                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT]
1990                         = u->tcp_fin_timeout * HZ;
1991         }
1992 #endif
1993
1994 #ifdef CONFIG_IP_VS_PROTO_UDP
1995         if (u->udp_timeout) {
1996                 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL]
1997                         = u->udp_timeout * HZ;
1998         }
1999 #endif
2000         return 0;
2001 }
2002
2003
2004 #define SET_CMDID(cmd)          (cmd - IP_VS_BASE_CTL)
2005 #define SERVICE_ARG_LEN         (sizeof(struct ip_vs_service_user))
2006 #define SVCDEST_ARG_LEN         (sizeof(struct ip_vs_service_user) +    \
2007                                  sizeof(struct ip_vs_dest_user))
2008 #define TIMEOUT_ARG_LEN         (sizeof(struct ip_vs_timeout_user))
2009 #define DAEMON_ARG_LEN          (sizeof(struct ip_vs_daemon_user))
2010 #define MAX_ARG_LEN             SVCDEST_ARG_LEN
2011
2012 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2013         [SET_CMDID(IP_VS_SO_SET_ADD)]           = SERVICE_ARG_LEN,
2014         [SET_CMDID(IP_VS_SO_SET_EDIT)]          = SERVICE_ARG_LEN,
2015         [SET_CMDID(IP_VS_SO_SET_DEL)]           = SERVICE_ARG_LEN,
2016         [SET_CMDID(IP_VS_SO_SET_FLUSH)]         = 0,
2017         [SET_CMDID(IP_VS_SO_SET_ADDDEST)]       = SVCDEST_ARG_LEN,
2018         [SET_CMDID(IP_VS_SO_SET_DELDEST)]       = SVCDEST_ARG_LEN,
2019         [SET_CMDID(IP_VS_SO_SET_EDITDEST)]      = SVCDEST_ARG_LEN,
2020         [SET_CMDID(IP_VS_SO_SET_TIMEOUT)]       = TIMEOUT_ARG_LEN,
2021         [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)]   = DAEMON_ARG_LEN,
2022         [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)]    = DAEMON_ARG_LEN,
2023         [SET_CMDID(IP_VS_SO_SET_ZERO)]          = SERVICE_ARG_LEN,
2024 };
2025
2026 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2027                                   struct ip_vs_service_user *usvc_compat)
2028 {
2029         usvc->af                = AF_INET;
2030         usvc->protocol          = usvc_compat->protocol;
2031         usvc->addr.ip           = usvc_compat->addr;
2032         usvc->port              = usvc_compat->port;
2033         usvc->fwmark            = usvc_compat->fwmark;
2034
2035         /* Deep copy of sched_name is not needed here */
2036         usvc->sched_name        = usvc_compat->sched_name;
2037
2038         usvc->flags             = usvc_compat->flags;
2039         usvc->timeout           = usvc_compat->timeout;
2040         usvc->netmask           = usvc_compat->netmask;
2041 }
2042
2043 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2044                                    struct ip_vs_dest_user *udest_compat)
2045 {
2046         udest->addr.ip          = udest_compat->addr;
2047         udest->port             = udest_compat->port;
2048         udest->conn_flags       = udest_compat->conn_flags;
2049         udest->weight           = udest_compat->weight;
2050         udest->u_threshold      = udest_compat->u_threshold;
2051         udest->l_threshold      = udest_compat->l_threshold;
2052 }
2053
2054 static int
2055 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2056 {
2057         int ret;
2058         unsigned char arg[MAX_ARG_LEN];
2059         struct ip_vs_service_user *usvc_compat;
2060         struct ip_vs_service_user_kern usvc;
2061         struct ip_vs_service *svc;
2062         struct ip_vs_dest_user *udest_compat;
2063         struct ip_vs_dest_user_kern udest;
2064
2065         if (!capable(CAP_NET_ADMIN))
2066                 return -EPERM;
2067
2068         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2069                 return -EINVAL;
2070         if (len < 0 || len >  MAX_ARG_LEN)
2071                 return -EINVAL;
2072         if (len != set_arglen[SET_CMDID(cmd)]) {
2073                 pr_err("set_ctl: len %u != %u\n",
2074                        len, set_arglen[SET_CMDID(cmd)]);
2075                 return -EINVAL;
2076         }
2077
2078         if (copy_from_user(arg, user, len) != 0)
2079                 return -EFAULT;
2080
2081         /* increase the module use count */
2082         ip_vs_use_count_inc();
2083
2084         if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2085                 ret = -ERESTARTSYS;
2086                 goto out_dec;
2087         }
2088
2089         if (cmd == IP_VS_SO_SET_FLUSH) {
2090                 /* Flush the virtual service */
2091                 ret = ip_vs_flush();
2092                 goto out_unlock;
2093         } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2094                 /* Set timeout values for (tcp tcpfin udp) */
2095                 ret = ip_vs_set_timeout((struct ip_vs_timeout_user *)arg);
2096                 goto out_unlock;
2097         } else if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2098                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2099                 ret = start_sync_thread(dm->state, dm->mcast_ifn, dm->syncid);
2100                 goto out_unlock;
2101         } else if (cmd == IP_VS_SO_SET_STOPDAEMON) {
2102                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2103                 ret = stop_sync_thread(dm->state);
2104                 goto out_unlock;
2105         }
2106
2107         usvc_compat = (struct ip_vs_service_user *)arg;
2108         udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2109
2110         /* We only use the new structs internally, so copy userspace compat
2111          * structs to extended internal versions */
2112         ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2113         ip_vs_copy_udest_compat(&udest, udest_compat);
2114
2115         if (cmd == IP_VS_SO_SET_ZERO) {
2116                 /* if no service address is set, zero counters in all */
2117                 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2118                         ret = ip_vs_zero_all();
2119                         goto out_unlock;
2120                 }
2121         }
2122
2123         /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2124         if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2125             usvc.protocol != IPPROTO_SCTP) {
2126                 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2127                        usvc.protocol, &usvc.addr.ip,
2128                        ntohs(usvc.port), usvc.sched_name);
2129                 ret = -EFAULT;
2130                 goto out_unlock;
2131         }
2132
2133         /* Lookup the exact service by <protocol, addr, port> or fwmark */
2134         if (usvc.fwmark == 0)
2135                 svc = __ip_vs_service_find(usvc.af, usvc.protocol,
2136                                            &usvc.addr, usvc.port);
2137         else
2138                 svc = __ip_vs_svc_fwm_find(usvc.af, usvc.fwmark);
2139
2140         if (cmd != IP_VS_SO_SET_ADD
2141             && (svc == NULL || svc->protocol != usvc.protocol)) {
2142                 ret = -ESRCH;
2143                 goto out_unlock;
2144         }
2145
2146         switch (cmd) {
2147         case IP_VS_SO_SET_ADD:
2148                 if (svc != NULL)
2149                         ret = -EEXIST;
2150                 else
2151                         ret = ip_vs_add_service(&usvc, &svc);
2152                 break;
2153         case IP_VS_SO_SET_EDIT:
2154                 ret = ip_vs_edit_service(svc, &usvc);
2155                 break;
2156         case IP_VS_SO_SET_DEL:
2157                 ret = ip_vs_del_service(svc);
2158                 if (!ret)
2159                         goto out_unlock;
2160                 break;
2161         case IP_VS_SO_SET_ZERO:
2162                 ret = ip_vs_zero_service(svc);
2163                 break;
2164         case IP_VS_SO_SET_ADDDEST:
2165                 ret = ip_vs_add_dest(svc, &udest);
2166                 break;
2167         case IP_VS_SO_SET_EDITDEST:
2168                 ret = ip_vs_edit_dest(svc, &udest);
2169                 break;
2170         case IP_VS_SO_SET_DELDEST:
2171                 ret = ip_vs_del_dest(svc, &udest);
2172                 break;
2173         default:
2174                 ret = -EINVAL;
2175         }
2176
2177   out_unlock:
2178         mutex_unlock(&__ip_vs_mutex);
2179   out_dec:
2180         /* decrease the module use count */
2181         ip_vs_use_count_dec();
2182
2183         return ret;
2184 }
2185
2186
2187 static void
2188 ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
2189 {
2190         spin_lock_bh(&src->lock);
2191         memcpy(dst, &src->ustats, sizeof(*dst));
2192         spin_unlock_bh(&src->lock);
2193 }
2194
2195 static void
2196 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2197 {
2198         dst->protocol = src->protocol;
2199         dst->addr = src->addr.ip;
2200         dst->port = src->port;
2201         dst->fwmark = src->fwmark;
2202         strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
2203         dst->flags = src->flags;
2204         dst->timeout = src->timeout / HZ;
2205         dst->netmask = src->netmask;
2206         dst->num_dests = src->num_dests;
2207         ip_vs_copy_stats(&dst->stats, &src->stats);
2208 }
2209
2210 static inline int
2211 __ip_vs_get_service_entries(const struct ip_vs_get_services *get,
2212                             struct ip_vs_get_services __user *uptr)
2213 {
2214         int idx, count=0;
2215         struct ip_vs_service *svc;
2216         struct ip_vs_service_entry entry;
2217         int ret = 0;
2218
2219         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2220                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2221                         /* Only expose IPv4 entries to old interface */
2222                         if (svc->af != AF_INET)
2223                                 continue;
2224
2225                         if (count >= get->num_services)
2226                                 goto out;
2227                         memset(&entry, 0, sizeof(entry));
2228                         ip_vs_copy_service(&entry, svc);
2229                         if (copy_to_user(&uptr->entrytable[count],
2230                                          &entry, sizeof(entry))) {
2231                                 ret = -EFAULT;
2232                                 goto out;
2233                         }
2234                         count++;
2235                 }
2236         }
2237
2238         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2239                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2240                         /* Only expose IPv4 entries to old interface */
2241                         if (svc->af != AF_INET)
2242                                 continue;
2243
2244                         if (count >= get->num_services)
2245                                 goto out;
2246                         memset(&entry, 0, sizeof(entry));
2247                         ip_vs_copy_service(&entry, svc);
2248                         if (copy_to_user(&uptr->entrytable[count],
2249                                          &entry, sizeof(entry))) {
2250                                 ret = -EFAULT;
2251                                 goto out;
2252                         }
2253                         count++;
2254                 }
2255         }
2256   out:
2257         return ret;
2258 }
2259
2260 static inline int
2261 __ip_vs_get_dest_entries(const struct ip_vs_get_dests *get,
2262                          struct ip_vs_get_dests __user *uptr)
2263 {
2264         struct ip_vs_service *svc;
2265         union nf_inet_addr addr = { .ip = get->addr };
2266         int ret = 0;
2267
2268         if (get->fwmark)
2269                 svc = __ip_vs_svc_fwm_find(AF_INET, get->fwmark);
2270         else
2271                 svc = __ip_vs_service_find(AF_INET, get->protocol, &addr,
2272                                            get->port);
2273
2274         if (svc) {
2275                 int count = 0;
2276                 struct ip_vs_dest *dest;
2277                 struct ip_vs_dest_entry entry;
2278
2279                 list_for_each_entry(dest, &svc->destinations, n_list) {
2280                         if (count >= get->num_dests)
2281                                 break;
2282
2283                         entry.addr = dest->addr.ip;
2284                         entry.port = dest->port;
2285                         entry.conn_flags = atomic_read(&dest->conn_flags);
2286                         entry.weight = atomic_read(&dest->weight);
2287                         entry.u_threshold = dest->u_threshold;
2288                         entry.l_threshold = dest->l_threshold;
2289                         entry.activeconns = atomic_read(&dest->activeconns);
2290                         entry.inactconns = atomic_read(&dest->inactconns);
2291                         entry.persistconns = atomic_read(&dest->persistconns);
2292                         ip_vs_copy_stats(&entry.stats, &dest->stats);
2293                         if (copy_to_user(&uptr->entrytable[count],
2294                                          &entry, sizeof(entry))) {
2295                                 ret = -EFAULT;
2296                                 break;
2297                         }
2298                         count++;
2299                 }
2300         } else
2301                 ret = -ESRCH;
2302         return ret;
2303 }
2304
2305 static inline void
2306 __ip_vs_get_timeouts(struct ip_vs_timeout_user *u)
2307 {
2308 #ifdef CONFIG_IP_VS_PROTO_TCP
2309         u->tcp_timeout =
2310                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2311         u->tcp_fin_timeout =
2312                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2313 #endif
2314 #ifdef CONFIG_IP_VS_PROTO_UDP
2315         u->udp_timeout =
2316                 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2317 #endif
2318 }
2319
2320
2321 #define GET_CMDID(cmd)          (cmd - IP_VS_BASE_CTL)
2322 #define GET_INFO_ARG_LEN        (sizeof(struct ip_vs_getinfo))
2323 #define GET_SERVICES_ARG_LEN    (sizeof(struct ip_vs_get_services))
2324 #define GET_SERVICE_ARG_LEN     (sizeof(struct ip_vs_service_entry))
2325 #define GET_DESTS_ARG_LEN       (sizeof(struct ip_vs_get_dests))
2326 #define GET_TIMEOUT_ARG_LEN     (sizeof(struct ip_vs_timeout_user))
2327 #define GET_DAEMON_ARG_LEN      (sizeof(struct ip_vs_daemon_user) * 2)
2328
2329 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2330         [GET_CMDID(IP_VS_SO_GET_VERSION)]       = 64,
2331         [GET_CMDID(IP_VS_SO_GET_INFO)]          = GET_INFO_ARG_LEN,
2332         [GET_CMDID(IP_VS_SO_GET_SERVICES)]      = GET_SERVICES_ARG_LEN,
2333         [GET_CMDID(IP_VS_SO_GET_SERVICE)]       = GET_SERVICE_ARG_LEN,
2334         [GET_CMDID(IP_VS_SO_GET_DESTS)]         = GET_DESTS_ARG_LEN,
2335         [GET_CMDID(IP_VS_SO_GET_TIMEOUT)]       = GET_TIMEOUT_ARG_LEN,
2336         [GET_CMDID(IP_VS_SO_GET_DAEMON)]        = GET_DAEMON_ARG_LEN,
2337 };
2338
2339 static int
2340 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2341 {
2342         unsigned char arg[128];
2343         int ret = 0;
2344         unsigned int copylen;
2345
2346         if (!capable(CAP_NET_ADMIN))
2347                 return -EPERM;
2348
2349         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2350                 return -EINVAL;
2351
2352         if (*len < get_arglen[GET_CMDID(cmd)]) {
2353                 pr_err("get_ctl: len %u < %u\n",
2354                        *len, get_arglen[GET_CMDID(cmd)]);
2355                 return -EINVAL;
2356         }
2357
2358         copylen = get_arglen[GET_CMDID(cmd)];
2359         if (copylen > 128)
2360                 return -EINVAL;
2361
2362         if (copy_from_user(arg, user, copylen) != 0)
2363                 return -EFAULT;
2364
2365         if (mutex_lock_interruptible(&__ip_vs_mutex))
2366                 return -ERESTARTSYS;
2367
2368         switch (cmd) {
2369         case IP_VS_SO_GET_VERSION:
2370         {
2371                 char buf[64];
2372
2373                 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2374                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2375                 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2376                         ret = -EFAULT;
2377                         goto out;
2378                 }
2379                 *len = strlen(buf)+1;
2380         }
2381         break;
2382
2383         case IP_VS_SO_GET_INFO:
2384         {
2385                 struct ip_vs_getinfo info;
2386                 info.version = IP_VS_VERSION_CODE;
2387                 info.size = ip_vs_conn_tab_size;
2388                 info.num_services = ip_vs_num_services;
2389                 if (copy_to_user(user, &info, sizeof(info)) != 0)
2390                         ret = -EFAULT;
2391         }
2392         break;
2393
2394         case IP_VS_SO_GET_SERVICES:
2395         {
2396                 struct ip_vs_get_services *get;
2397                 int size;
2398
2399                 get = (struct ip_vs_get_services *)arg;
2400                 size = sizeof(*get) +
2401                         sizeof(struct ip_vs_service_entry) * get->num_services;
2402                 if (*len != size) {
2403                         pr_err("length: %u != %u\n", *len, size);
2404                         ret = -EINVAL;
2405                         goto out;
2406                 }
2407                 ret = __ip_vs_get_service_entries(get, user);
2408         }
2409         break;
2410
2411         case IP_VS_SO_GET_SERVICE:
2412         {
2413                 struct ip_vs_service_entry *entry;
2414                 struct ip_vs_service *svc;
2415                 union nf_inet_addr addr;
2416
2417                 entry = (struct ip_vs_service_entry *)arg;
2418                 addr.ip = entry->addr;
2419                 if (entry->fwmark)
2420                         svc = __ip_vs_svc_fwm_find(AF_INET, entry->fwmark);
2421                 else
2422                         svc = __ip_vs_service_find(AF_INET, entry->protocol,
2423                                                    &addr, entry->port);
2424                 if (svc) {
2425                         ip_vs_copy_service(entry, svc);
2426                         if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2427                                 ret = -EFAULT;
2428                 } else
2429                         ret = -ESRCH;
2430         }
2431         break;
2432
2433         case IP_VS_SO_GET_DESTS:
2434         {
2435                 struct ip_vs_get_dests *get;
2436                 int size;
2437
2438                 get = (struct ip_vs_get_dests *)arg;
2439                 size = sizeof(*get) +
2440                         sizeof(struct ip_vs_dest_entry) * get->num_dests;
2441                 if (*len != size) {
2442                         pr_err("length: %u != %u\n", *len, size);
2443                         ret = -EINVAL;
2444                         goto out;
2445                 }
2446                 ret = __ip_vs_get_dest_entries(get, user);
2447         }
2448         break;
2449
2450         case IP_VS_SO_GET_TIMEOUT:
2451         {
2452                 struct ip_vs_timeout_user t;
2453
2454                 __ip_vs_get_timeouts(&t);
2455                 if (copy_to_user(user, &t, sizeof(t)) != 0)
2456                         ret = -EFAULT;
2457         }
2458         break;
2459
2460         case IP_VS_SO_GET_DAEMON:
2461         {
2462                 struct ip_vs_daemon_user d[2];
2463
2464                 memset(&d, 0, sizeof(d));
2465                 if (ip_vs_sync_state & IP_VS_STATE_MASTER) {
2466                         d[0].state = IP_VS_STATE_MASTER;
2467                         strlcpy(d[0].mcast_ifn, ip_vs_master_mcast_ifn, sizeof(d[0].mcast_ifn));
2468                         d[0].syncid = ip_vs_master_syncid;
2469                 }
2470                 if (ip_vs_sync_state & IP_VS_STATE_BACKUP) {
2471                         d[1].state = IP_VS_STATE_BACKUP;
2472                         strlcpy(d[1].mcast_ifn, ip_vs_backup_mcast_ifn, sizeof(d[1].mcast_ifn));
2473                         d[1].syncid = ip_vs_backup_syncid;
2474                 }
2475                 if (copy_to_user(user, &d, sizeof(d)) != 0)
2476                         ret = -EFAULT;
2477         }
2478         break;
2479
2480         default:
2481                 ret = -EINVAL;
2482         }
2483
2484   out:
2485         mutex_unlock(&__ip_vs_mutex);
2486         return ret;
2487 }
2488
2489
2490 static struct nf_sockopt_ops ip_vs_sockopts = {
2491         .pf             = PF_INET,
2492         .set_optmin     = IP_VS_BASE_CTL,
2493         .set_optmax     = IP_VS_SO_SET_MAX+1,
2494         .set            = do_ip_vs_set_ctl,
2495         .get_optmin     = IP_VS_BASE_CTL,
2496         .get_optmax     = IP_VS_SO_GET_MAX+1,
2497         .get            = do_ip_vs_get_ctl,
2498         .owner          = THIS_MODULE,
2499 };
2500
2501 /*
2502  * Generic Netlink interface
2503  */
2504
2505 /* IPVS genetlink family */
2506 static struct genl_family ip_vs_genl_family = {
2507         .id             = GENL_ID_GENERATE,
2508         .hdrsize        = 0,
2509         .name           = IPVS_GENL_NAME,
2510         .version        = IPVS_GENL_VERSION,
2511         .maxattr        = IPVS_CMD_MAX,
2512 };
2513
2514 /* Policy used for first-level command attributes */
2515 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2516         [IPVS_CMD_ATTR_SERVICE]         = { .type = NLA_NESTED },
2517         [IPVS_CMD_ATTR_DEST]            = { .type = NLA_NESTED },
2518         [IPVS_CMD_ATTR_DAEMON]          = { .type = NLA_NESTED },
2519         [IPVS_CMD_ATTR_TIMEOUT_TCP]     = { .type = NLA_U32 },
2520         [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2521         [IPVS_CMD_ATTR_TIMEOUT_UDP]     = { .type = NLA_U32 },
2522 };
2523
2524 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2525 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2526         [IPVS_DAEMON_ATTR_STATE]        = { .type = NLA_U32 },
2527         [IPVS_DAEMON_ATTR_MCAST_IFN]    = { .type = NLA_NUL_STRING,
2528                                             .len = IP_VS_IFNAME_MAXLEN },
2529         [IPVS_DAEMON_ATTR_SYNC_ID]      = { .type = NLA_U32 },
2530 };
2531
2532 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2533 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2534         [IPVS_SVC_ATTR_AF]              = { .type = NLA_U16 },
2535         [IPVS_SVC_ATTR_PROTOCOL]        = { .type = NLA_U16 },
2536         [IPVS_SVC_ATTR_ADDR]            = { .type = NLA_BINARY,
2537                                             .len = sizeof(union nf_inet_addr) },
2538         [IPVS_SVC_ATTR_PORT]            = { .type = NLA_U16 },
2539         [IPVS_SVC_ATTR_FWMARK]          = { .type = NLA_U32 },
2540         [IPVS_SVC_ATTR_SCHED_NAME]      = { .type = NLA_NUL_STRING,
2541                                             .len = IP_VS_SCHEDNAME_MAXLEN },
2542         [IPVS_SVC_ATTR_FLAGS]           = { .type = NLA_BINARY,
2543                                             .len = sizeof(struct ip_vs_flags) },
2544         [IPVS_SVC_ATTR_TIMEOUT]         = { .type = NLA_U32 },
2545         [IPVS_SVC_ATTR_NETMASK]         = { .type = NLA_U32 },
2546         [IPVS_SVC_ATTR_STATS]           = { .type = NLA_NESTED },
2547 };
2548
2549 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2550 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2551         [IPVS_DEST_ATTR_ADDR]           = { .type = NLA_BINARY,
2552                                             .len = sizeof(union nf_inet_addr) },
2553         [IPVS_DEST_ATTR_PORT]           = { .type = NLA_U16 },
2554         [IPVS_DEST_ATTR_FWD_METHOD]     = { .type = NLA_U32 },
2555         [IPVS_DEST_ATTR_WEIGHT]         = { .type = NLA_U32 },
2556         [IPVS_DEST_ATTR_U_THRESH]       = { .type = NLA_U32 },
2557         [IPVS_DEST_ATTR_L_THRESH]       = { .type = NLA_U32 },
2558         [IPVS_DEST_ATTR_ACTIVE_CONNS]   = { .type = NLA_U32 },
2559         [IPVS_DEST_ATTR_INACT_CONNS]    = { .type = NLA_U32 },
2560         [IPVS_DEST_ATTR_PERSIST_CONNS]  = { .type = NLA_U32 },
2561         [IPVS_DEST_ATTR_STATS]          = { .type = NLA_NESTED },
2562 };
2563
2564 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2565                                  struct ip_vs_stats *stats)
2566 {
2567         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2568         if (!nl_stats)
2569                 return -EMSGSIZE;
2570
2571         spin_lock_bh(&stats->lock);
2572
2573         NLA_PUT_U32(skb, IPVS_STATS_ATTR_CONNS, stats->ustats.conns);
2574         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPKTS, stats->ustats.inpkts);
2575         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPKTS, stats->ustats.outpkts);
2576         NLA_PUT_U64(skb, IPVS_STATS_ATTR_INBYTES, stats->ustats.inbytes);
2577         NLA_PUT_U64(skb, IPVS_STATS_ATTR_OUTBYTES, stats->ustats.outbytes);
2578         NLA_PUT_U32(skb, IPVS_STATS_ATTR_CPS, stats->ustats.cps);
2579         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPPS, stats->ustats.inpps);
2580         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPPS, stats->ustats.outpps);
2581         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INBPS, stats->ustats.inbps);
2582         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTBPS, stats->ustats.outbps);
2583
2584         spin_unlock_bh(&stats->lock);
2585
2586         nla_nest_end(skb, nl_stats);
2587
2588         return 0;
2589
2590 nla_put_failure:
2591         spin_unlock_bh(&stats->lock);
2592         nla_nest_cancel(skb, nl_stats);
2593         return -EMSGSIZE;
2594 }
2595
2596 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2597                                    struct ip_vs_service *svc)
2598 {
2599         struct nlattr *nl_service;
2600         struct ip_vs_flags flags = { .flags = svc->flags,
2601                                      .mask = ~0 };
2602
2603         nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2604         if (!nl_service)
2605                 return -EMSGSIZE;
2606
2607         NLA_PUT_U16(skb, IPVS_SVC_ATTR_AF, svc->af);
2608
2609         if (svc->fwmark) {
2610                 NLA_PUT_U32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark);
2611         } else {
2612                 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol);
2613                 NLA_PUT(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr);
2614                 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PORT, svc->port);
2615         }
2616
2617         NLA_PUT_STRING(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name);
2618         NLA_PUT(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags);
2619         NLA_PUT_U32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ);
2620         NLA_PUT_U32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask);
2621
2622         if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2623                 goto nla_put_failure;
2624
2625         nla_nest_end(skb, nl_service);
2626
2627         return 0;
2628
2629 nla_put_failure:
2630         nla_nest_cancel(skb, nl_service);
2631         return -EMSGSIZE;
2632 }
2633
2634 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2635                                    struct ip_vs_service *svc,
2636                                    struct netlink_callback *cb)
2637 {
2638         void *hdr;
2639
2640         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2641                           &ip_vs_genl_family, NLM_F_MULTI,
2642                           IPVS_CMD_NEW_SERVICE);
2643         if (!hdr)
2644                 return -EMSGSIZE;
2645
2646         if (ip_vs_genl_fill_service(skb, svc) < 0)
2647                 goto nla_put_failure;
2648
2649         return genlmsg_end(skb, hdr);
2650
2651 nla_put_failure:
2652         genlmsg_cancel(skb, hdr);
2653         return -EMSGSIZE;
2654 }
2655
2656 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2657                                     struct netlink_callback *cb)
2658 {
2659         int idx = 0, i;
2660         int start = cb->args[0];
2661         struct ip_vs_service *svc;
2662
2663         mutex_lock(&__ip_vs_mutex);
2664         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2665                 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2666                         if (++idx <= start)
2667                                 continue;
2668                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2669                                 idx--;
2670                                 goto nla_put_failure;
2671                         }
2672                 }
2673         }
2674
2675         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2676                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2677                         if (++idx <= start)
2678                                 continue;
2679                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2680                                 idx--;
2681                                 goto nla_put_failure;
2682                         }
2683                 }
2684         }
2685
2686 nla_put_failure:
2687         mutex_unlock(&__ip_vs_mutex);
2688         cb->args[0] = idx;
2689
2690         return skb->len;
2691 }
2692
2693 static int ip_vs_genl_parse_service(struct ip_vs_service_user_kern *usvc,
2694                                     struct nlattr *nla, int full_entry,
2695                                     struct ip_vs_service **ret_svc)
2696 {
2697         struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2698         struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2699         struct ip_vs_service *svc;
2700
2701         /* Parse mandatory identifying service fields first */
2702         if (nla == NULL ||
2703             nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2704                 return -EINVAL;
2705
2706         nla_af          = attrs[IPVS_SVC_ATTR_AF];
2707         nla_protocol    = attrs[IPVS_SVC_ATTR_PROTOCOL];
2708         nla_addr        = attrs[IPVS_SVC_ATTR_ADDR];
2709         nla_port        = attrs[IPVS_SVC_ATTR_PORT];
2710         nla_fwmark      = attrs[IPVS_SVC_ATTR_FWMARK];
2711
2712         if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2713                 return -EINVAL;
2714
2715         memset(usvc, 0, sizeof(*usvc));
2716
2717         usvc->af = nla_get_u16(nla_af);
2718 #ifdef CONFIG_IP_VS_IPV6
2719         if (usvc->af != AF_INET && usvc->af != AF_INET6)
2720 #else
2721         if (usvc->af != AF_INET)
2722 #endif
2723                 return -EAFNOSUPPORT;
2724
2725         if (nla_fwmark) {
2726                 usvc->protocol = IPPROTO_TCP;
2727                 usvc->fwmark = nla_get_u32(nla_fwmark);
2728         } else {
2729                 usvc->protocol = nla_get_u16(nla_protocol);
2730                 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
2731                 usvc->port = nla_get_u16(nla_port);
2732                 usvc->fwmark = 0;
2733         }
2734
2735         if (usvc->fwmark)
2736                 svc = __ip_vs_svc_fwm_find(usvc->af, usvc->fwmark);
2737         else
2738                 svc = __ip_vs_service_find(usvc->af, usvc->protocol,
2739                                            &usvc->addr, usvc->port);
2740         *ret_svc = svc;
2741
2742         /* If a full entry was requested, check for the additional fields */
2743         if (full_entry) {
2744                 struct nlattr *nla_sched, *nla_flags, *nla_timeout,
2745                               *nla_netmask;
2746                 struct ip_vs_flags flags;
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 (svc)
2760                         usvc->flags = svc->flags;
2761
2762                 /* set new flags from userland */
2763                 usvc->flags = (usvc->flags & ~flags.mask) |
2764                               (flags.flags & flags.mask);
2765                 usvc->sched_name = nla_data(nla_sched);
2766                 usvc->timeout = nla_get_u32(nla_timeout);
2767                 usvc->netmask = nla_get_u32(nla_netmask);
2768         }
2769
2770         return 0;
2771 }
2772
2773 static struct ip_vs_service *ip_vs_genl_find_service(struct nlattr *nla)
2774 {
2775         struct ip_vs_service_user_kern usvc;
2776         struct ip_vs_service *svc;
2777         int ret;
2778
2779         ret = ip_vs_genl_parse_service(&usvc, nla, 0, &svc);
2780         return ret ? ERR_PTR(ret) : svc;
2781 }
2782
2783 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
2784 {
2785         struct nlattr *nl_dest;
2786
2787         nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
2788         if (!nl_dest)
2789                 return -EMSGSIZE;
2790
2791         NLA_PUT(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr);
2792         NLA_PUT_U16(skb, IPVS_DEST_ATTR_PORT, dest->port);
2793
2794         NLA_PUT_U32(skb, IPVS_DEST_ATTR_FWD_METHOD,
2795                     atomic_read(&dest->conn_flags) & IP_VS_CONN_F_FWD_MASK);
2796         NLA_PUT_U32(skb, IPVS_DEST_ATTR_WEIGHT, atomic_read(&dest->weight));
2797         NLA_PUT_U32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold);
2798         NLA_PUT_U32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold);
2799         NLA_PUT_U32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
2800                     atomic_read(&dest->activeconns));
2801         NLA_PUT_U32(skb, IPVS_DEST_ATTR_INACT_CONNS,
2802                     atomic_read(&dest->inactconns));
2803         NLA_PUT_U32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
2804                     atomic_read(&dest->persistconns));
2805
2806         if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
2807                 goto nla_put_failure;
2808
2809         nla_nest_end(skb, nl_dest);
2810
2811         return 0;
2812
2813 nla_put_failure:
2814         nla_nest_cancel(skb, nl_dest);
2815         return -EMSGSIZE;
2816 }
2817
2818 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
2819                                 struct netlink_callback *cb)
2820 {
2821         void *hdr;
2822
2823         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2824                           &ip_vs_genl_family, NLM_F_MULTI,
2825                           IPVS_CMD_NEW_DEST);
2826         if (!hdr)
2827                 return -EMSGSIZE;
2828
2829         if (ip_vs_genl_fill_dest(skb, dest) < 0)
2830                 goto nla_put_failure;
2831
2832         return genlmsg_end(skb, hdr);
2833
2834 nla_put_failure:
2835         genlmsg_cancel(skb, hdr);
2836         return -EMSGSIZE;
2837 }
2838
2839 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
2840                                  struct netlink_callback *cb)
2841 {
2842         int idx = 0;
2843         int start = cb->args[0];
2844         struct ip_vs_service *svc;
2845         struct ip_vs_dest *dest;
2846         struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
2847
2848         mutex_lock(&__ip_vs_mutex);
2849
2850         /* Try to find the service for which to dump destinations */
2851         if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
2852                         IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
2853                 goto out_err;
2854
2855         svc = ip_vs_genl_find_service(attrs[IPVS_CMD_ATTR_SERVICE]);
2856         if (IS_ERR(svc) || svc == NULL)
2857                 goto out_err;
2858
2859         /* Dump the destinations */
2860         list_for_each_entry(dest, &svc->destinations, n_list) {
2861                 if (++idx <= start)
2862                         continue;
2863                 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
2864                         idx--;
2865                         goto nla_put_failure;
2866                 }
2867         }
2868
2869 nla_put_failure:
2870         cb->args[0] = idx;
2871
2872 out_err:
2873         mutex_unlock(&__ip_vs_mutex);
2874
2875         return skb->len;
2876 }
2877
2878 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
2879                                  struct nlattr *nla, int full_entry)
2880 {
2881         struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
2882         struct nlattr *nla_addr, *nla_port;
2883
2884         /* Parse mandatory identifying destination fields first */
2885         if (nla == NULL ||
2886             nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
2887                 return -EINVAL;
2888
2889         nla_addr        = attrs[IPVS_DEST_ATTR_ADDR];
2890         nla_port        = attrs[IPVS_DEST_ATTR_PORT];
2891
2892         if (!(nla_addr && nla_port))
2893                 return -EINVAL;
2894
2895         memset(udest, 0, sizeof(*udest));
2896
2897         nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
2898         udest->port = nla_get_u16(nla_port);
2899
2900         /* If a full entry was requested, check for the additional fields */
2901         if (full_entry) {
2902                 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
2903                               *nla_l_thresh;
2904
2905                 nla_fwd         = attrs[IPVS_DEST_ATTR_FWD_METHOD];
2906                 nla_weight      = attrs[IPVS_DEST_ATTR_WEIGHT];
2907                 nla_u_thresh    = attrs[IPVS_DEST_ATTR_U_THRESH];
2908                 nla_l_thresh    = attrs[IPVS_DEST_ATTR_L_THRESH];
2909
2910                 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
2911                         return -EINVAL;
2912
2913                 udest->conn_flags = nla_get_u32(nla_fwd)
2914                                     & IP_VS_CONN_F_FWD_MASK;
2915                 udest->weight = nla_get_u32(nla_weight);
2916                 udest->u_threshold = nla_get_u32(nla_u_thresh);
2917                 udest->l_threshold = nla_get_u32(nla_l_thresh);
2918         }
2919
2920         return 0;
2921 }
2922
2923 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
2924                                   const char *mcast_ifn, __be32 syncid)
2925 {
2926         struct nlattr *nl_daemon;
2927
2928         nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
2929         if (!nl_daemon)
2930                 return -EMSGSIZE;
2931
2932         NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_STATE, state);
2933         NLA_PUT_STRING(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn);
2934         NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid);
2935
2936         nla_nest_end(skb, nl_daemon);
2937
2938         return 0;
2939
2940 nla_put_failure:
2941         nla_nest_cancel(skb, nl_daemon);
2942         return -EMSGSIZE;
2943 }
2944
2945 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
2946                                   const char *mcast_ifn, __be32 syncid,
2947                                   struct netlink_callback *cb)
2948 {
2949         void *hdr;
2950         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2951                           &ip_vs_genl_family, NLM_F_MULTI,
2952                           IPVS_CMD_NEW_DAEMON);
2953         if (!hdr)
2954                 return -EMSGSIZE;
2955
2956         if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
2957                 goto nla_put_failure;
2958
2959         return genlmsg_end(skb, hdr);
2960
2961 nla_put_failure:
2962         genlmsg_cancel(skb, hdr);
2963         return -EMSGSIZE;
2964 }
2965
2966 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
2967                                    struct netlink_callback *cb)
2968 {
2969         mutex_lock(&__ip_vs_mutex);
2970         if ((ip_vs_sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
2971                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
2972                                            ip_vs_master_mcast_ifn,
2973                                            ip_vs_master_syncid, cb) < 0)
2974                         goto nla_put_failure;
2975
2976                 cb->args[0] = 1;
2977         }
2978
2979         if ((ip_vs_sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
2980                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
2981                                            ip_vs_backup_mcast_ifn,
2982                                            ip_vs_backup_syncid, cb) < 0)
2983                         goto nla_put_failure;
2984
2985                 cb->args[1] = 1;
2986         }
2987
2988 nla_put_failure:
2989         mutex_unlock(&__ip_vs_mutex);
2990
2991         return skb->len;
2992 }
2993
2994 static int ip_vs_genl_new_daemon(struct nlattr **attrs)
2995 {
2996         if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
2997               attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
2998               attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
2999                 return -EINVAL;
3000
3001         return start_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3002                                  nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3003                                  nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3004 }
3005
3006 static int ip_vs_genl_del_daemon(struct nlattr **attrs)
3007 {
3008         if (!attrs[IPVS_DAEMON_ATTR_STATE])
3009                 return -EINVAL;
3010
3011         return stop_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3012 }
3013
3014 static int ip_vs_genl_set_config(struct nlattr **attrs)
3015 {
3016         struct ip_vs_timeout_user t;
3017
3018         __ip_vs_get_timeouts(&t);
3019
3020         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3021                 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3022
3023         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3024                 t.tcp_fin_timeout =
3025                         nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3026
3027         if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3028                 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3029
3030         return ip_vs_set_timeout(&t);
3031 }
3032
3033 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3034 {
3035         struct ip_vs_service *svc = NULL;
3036         struct ip_vs_service_user_kern usvc;
3037         struct ip_vs_dest_user_kern udest;
3038         int ret = 0, cmd;
3039         int need_full_svc = 0, need_full_dest = 0;
3040
3041         cmd = info->genlhdr->cmd;
3042
3043         mutex_lock(&__ip_vs_mutex);
3044
3045         if (cmd == IPVS_CMD_FLUSH) {
3046                 ret = ip_vs_flush();
3047                 goto out;
3048         } else if (cmd == IPVS_CMD_SET_CONFIG) {
3049                 ret = ip_vs_genl_set_config(info->attrs);
3050                 goto out;
3051         } else if (cmd == IPVS_CMD_NEW_DAEMON ||
3052                    cmd == IPVS_CMD_DEL_DAEMON) {
3053
3054                 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3055
3056                 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3057                     nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3058                                      info->attrs[IPVS_CMD_ATTR_DAEMON],
3059                                      ip_vs_daemon_policy)) {
3060                         ret = -EINVAL;
3061                         goto out;
3062                 }
3063
3064                 if (cmd == IPVS_CMD_NEW_DAEMON)
3065                         ret = ip_vs_genl_new_daemon(daemon_attrs);
3066                 else
3067                         ret = ip_vs_genl_del_daemon(daemon_attrs);
3068                 goto out;
3069         } else if (cmd == IPVS_CMD_ZERO &&
3070                    !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3071                 ret = ip_vs_zero_all();
3072                 goto out;
3073         }
3074
3075         /* All following commands require a service argument, so check if we
3076          * received a valid one. We need a full service specification when
3077          * adding / editing a service. Only identifying members otherwise. */
3078         if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3079                 need_full_svc = 1;
3080
3081         ret = ip_vs_genl_parse_service(&usvc,
3082                                        info->attrs[IPVS_CMD_ATTR_SERVICE],
3083                                        need_full_svc, &svc);
3084         if (ret)
3085                 goto out;
3086
3087         /* Unless we're adding a new service, the service must already exist */
3088         if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3089                 ret = -ESRCH;
3090                 goto out;
3091         }
3092
3093         /* Destination commands require a valid destination argument. For
3094          * adding / editing a destination, we need a full destination
3095          * specification. */
3096         if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3097             cmd == IPVS_CMD_DEL_DEST) {
3098                 if (cmd != IPVS_CMD_DEL_DEST)
3099                         need_full_dest = 1;
3100
3101                 ret = ip_vs_genl_parse_dest(&udest,
3102                                             info->attrs[IPVS_CMD_ATTR_DEST],
3103                                             need_full_dest);
3104                 if (ret)
3105                         goto out;
3106         }
3107
3108         switch (cmd) {
3109         case IPVS_CMD_NEW_SERVICE:
3110                 if (svc == NULL)
3111                         ret = ip_vs_add_service(&usvc, &svc);
3112                 else
3113                         ret = -EEXIST;
3114                 break;
3115         case IPVS_CMD_SET_SERVICE:
3116                 ret = ip_vs_edit_service(svc, &usvc);
3117                 break;
3118         case IPVS_CMD_DEL_SERVICE:
3119                 ret = ip_vs_del_service(svc);
3120                 /* do not use svc, it can be freed */
3121                 break;
3122         case IPVS_CMD_NEW_DEST:
3123                 ret = ip_vs_add_dest(svc, &udest);
3124                 break;
3125         case IPVS_CMD_SET_DEST:
3126                 ret = ip_vs_edit_dest(svc, &udest);
3127                 break;
3128         case IPVS_CMD_DEL_DEST:
3129                 ret = ip_vs_del_dest(svc, &udest);
3130                 break;
3131         case IPVS_CMD_ZERO:
3132                 ret = ip_vs_zero_service(svc);
3133                 break;
3134         default:
3135                 ret = -EINVAL;
3136         }
3137
3138 out:
3139         mutex_unlock(&__ip_vs_mutex);
3140
3141         return ret;
3142 }
3143
3144 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3145 {
3146         struct sk_buff *msg;
3147         void *reply;
3148         int ret, cmd, reply_cmd;
3149
3150         cmd = info->genlhdr->cmd;
3151
3152         if (cmd == IPVS_CMD_GET_SERVICE)
3153                 reply_cmd = IPVS_CMD_NEW_SERVICE;
3154         else if (cmd == IPVS_CMD_GET_INFO)
3155                 reply_cmd = IPVS_CMD_SET_INFO;
3156         else if (cmd == IPVS_CMD_GET_CONFIG)
3157                 reply_cmd = IPVS_CMD_SET_CONFIG;
3158         else {
3159                 pr_err("unknown Generic Netlink command\n");
3160                 return -EINVAL;
3161         }
3162
3163         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3164         if (!msg)
3165                 return -ENOMEM;
3166
3167         mutex_lock(&__ip_vs_mutex);
3168
3169         reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3170         if (reply == NULL)
3171                 goto nla_put_failure;
3172
3173         switch (cmd) {
3174         case IPVS_CMD_GET_SERVICE:
3175         {
3176                 struct ip_vs_service *svc;
3177
3178                 svc = ip_vs_genl_find_service(info->attrs[IPVS_CMD_ATTR_SERVICE]);
3179                 if (IS_ERR(svc)) {
3180                         ret = PTR_ERR(svc);
3181                         goto out_err;
3182                 } else if (svc) {
3183                         ret = ip_vs_genl_fill_service(msg, svc);
3184                         if (ret)
3185                                 goto nla_put_failure;
3186                 } else {
3187                         ret = -ESRCH;
3188                         goto out_err;
3189                 }
3190
3191                 break;
3192         }
3193
3194         case IPVS_CMD_GET_CONFIG:
3195         {
3196                 struct ip_vs_timeout_user t;
3197
3198                 __ip_vs_get_timeouts(&t);
3199 #ifdef CONFIG_IP_VS_PROTO_TCP
3200                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP, t.tcp_timeout);
3201                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3202                             t.tcp_fin_timeout);
3203 #endif
3204 #ifdef CONFIG_IP_VS_PROTO_UDP
3205                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout);
3206 #endif
3207
3208                 break;
3209         }
3210
3211         case IPVS_CMD_GET_INFO:
3212                 NLA_PUT_U32(msg, IPVS_INFO_ATTR_VERSION, IP_VS_VERSION_CODE);
3213                 NLA_PUT_U32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3214                             ip_vs_conn_tab_size);
3215                 break;
3216         }
3217
3218         genlmsg_end(msg, reply);
3219         ret = genlmsg_reply(msg, info);
3220         goto out;
3221
3222 nla_put_failure:
3223         pr_err("not enough space in Netlink message\n");
3224         ret = -EMSGSIZE;
3225
3226 out_err:
3227         nlmsg_free(msg);
3228 out:
3229         mutex_unlock(&__ip_vs_mutex);
3230
3231         return ret;
3232 }
3233
3234
3235 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3236         {
3237                 .cmd    = IPVS_CMD_NEW_SERVICE,
3238                 .flags  = GENL_ADMIN_PERM,
3239                 .policy = ip_vs_cmd_policy,
3240                 .doit   = ip_vs_genl_set_cmd,
3241         },
3242         {
3243                 .cmd    = IPVS_CMD_SET_SERVICE,
3244                 .flags  = GENL_ADMIN_PERM,
3245                 .policy = ip_vs_cmd_policy,
3246                 .doit   = ip_vs_genl_set_cmd,
3247         },
3248         {
3249                 .cmd    = IPVS_CMD_DEL_SERVICE,
3250                 .flags  = GENL_ADMIN_PERM,
3251                 .policy = ip_vs_cmd_policy,
3252                 .doit   = ip_vs_genl_set_cmd,
3253         },
3254         {
3255                 .cmd    = IPVS_CMD_GET_SERVICE,
3256                 .flags  = GENL_ADMIN_PERM,
3257                 .doit   = ip_vs_genl_get_cmd,
3258                 .dumpit = ip_vs_genl_dump_services,
3259                 .policy = ip_vs_cmd_policy,
3260         },
3261         {
3262                 .cmd    = IPVS_CMD_NEW_DEST,
3263                 .flags  = GENL_ADMIN_PERM,
3264                 .policy = ip_vs_cmd_policy,
3265                 .doit   = ip_vs_genl_set_cmd,
3266         },
3267         {
3268                 .cmd    = IPVS_CMD_SET_DEST,
3269                 .flags  = GENL_ADMIN_PERM,
3270                 .policy = ip_vs_cmd_policy,
3271                 .doit   = ip_vs_genl_set_cmd,
3272         },
3273         {
3274                 .cmd    = IPVS_CMD_DEL_DEST,
3275                 .flags  = GENL_ADMIN_PERM,
3276                 .policy = ip_vs_cmd_policy,
3277                 .doit   = ip_vs_genl_set_cmd,
3278         },
3279         {
3280                 .cmd    = IPVS_CMD_GET_DEST,
3281                 .flags  = GENL_ADMIN_PERM,
3282                 .policy = ip_vs_cmd_policy,
3283                 .dumpit = ip_vs_genl_dump_dests,
3284         },
3285         {
3286                 .cmd    = IPVS_CMD_NEW_DAEMON,
3287                 .flags  = GENL_ADMIN_PERM,
3288                 .policy = ip_vs_cmd_policy,
3289                 .doit   = ip_vs_genl_set_cmd,
3290         },
3291         {
3292                 .cmd    = IPVS_CMD_DEL_DAEMON,
3293                 .flags  = GENL_ADMIN_PERM,
3294                 .policy = ip_vs_cmd_policy,
3295                 .doit   = ip_vs_genl_set_cmd,
3296         },
3297         {
3298                 .cmd    = IPVS_CMD_GET_DAEMON,
3299                 .flags  = GENL_ADMIN_PERM,
3300                 .dumpit = ip_vs_genl_dump_daemons,
3301         },
3302         {
3303                 .cmd    = IPVS_CMD_SET_CONFIG,
3304                 .flags  = GENL_ADMIN_PERM,
3305                 .policy = ip_vs_cmd_policy,
3306                 .doit   = ip_vs_genl_set_cmd,
3307         },
3308         {
3309                 .cmd    = IPVS_CMD_GET_CONFIG,
3310                 .flags  = GENL_ADMIN_PERM,
3311                 .doit   = ip_vs_genl_get_cmd,
3312         },
3313         {
3314                 .cmd    = IPVS_CMD_GET_INFO,
3315                 .flags  = GENL_ADMIN_PERM,
3316                 .doit   = ip_vs_genl_get_cmd,
3317         },
3318         {
3319                 .cmd    = IPVS_CMD_ZERO,
3320                 .flags  = GENL_ADMIN_PERM,
3321                 .policy = ip_vs_cmd_policy,
3322                 .doit   = ip_vs_genl_set_cmd,
3323         },
3324         {
3325                 .cmd    = IPVS_CMD_FLUSH,
3326                 .flags  = GENL_ADMIN_PERM,
3327                 .doit   = ip_vs_genl_set_cmd,
3328         },
3329 };
3330
3331 static int __init ip_vs_genl_register(void)
3332 {
3333         return genl_register_family_with_ops(&ip_vs_genl_family,
3334                 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
3335 }
3336
3337 static void ip_vs_genl_unregister(void)
3338 {
3339         genl_unregister_family(&ip_vs_genl_family);
3340 }
3341
3342 /* End of Generic Netlink interface definitions */
3343
3344
3345 int __init ip_vs_control_init(void)
3346 {
3347         int ret;
3348         int idx;
3349
3350         EnterFunction(2);
3351
3352         ret = nf_register_sockopt(&ip_vs_sockopts);
3353         if (ret) {
3354                 pr_err("cannot register sockopt.\n");
3355                 return ret;
3356         }
3357
3358         ret = ip_vs_genl_register();
3359         if (ret) {
3360                 pr_err("cannot register Generic Netlink interface.\n");
3361                 nf_unregister_sockopt(&ip_vs_sockopts);
3362                 return ret;
3363         }
3364
3365         proc_net_fops_create(&init_net, "ip_vs", 0, &ip_vs_info_fops);
3366         proc_net_fops_create(&init_net, "ip_vs_stats",0, &ip_vs_stats_fops);
3367
3368         sysctl_header = register_sysctl_paths(net_vs_ctl_path, vs_vars);
3369
3370         /* Initialize ip_vs_svc_table, ip_vs_svc_fwm_table, ip_vs_rtable */
3371         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++)  {
3372                 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3373                 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3374         }
3375         for(idx = 0; idx < IP_VS_RTAB_SIZE; idx++)  {
3376                 INIT_LIST_HEAD(&ip_vs_rtable[idx]);
3377         }
3378
3379         ip_vs_new_estimator(&ip_vs_stats);
3380
3381         /* Hook the defense timer */
3382         schedule_delayed_work(&defense_work, DEFENSE_TIMER_PERIOD);
3383
3384         LeaveFunction(2);
3385         return 0;
3386 }
3387
3388
3389 void ip_vs_control_cleanup(void)
3390 {
3391         EnterFunction(2);
3392         ip_vs_trash_cleanup();
3393         cancel_rearming_delayed_work(&defense_work);
3394         cancel_work_sync(&defense_work.work);
3395         ip_vs_kill_estimator(&ip_vs_stats);
3396         unregister_sysctl_table(sysctl_header);
3397         proc_net_remove(&init_net, "ip_vs_stats");
3398         proc_net_remove(&init_net, "ip_vs");
3399         ip_vs_genl_unregister();
3400         nf_unregister_sockopt(&ip_vs_sockopts);
3401         LeaveFunction(2);
3402 }