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 <linux/nsproxy.h>
  42 #include <net/ip.h>
  43 #ifdef CONFIG_IP_VS_IPV6
  44 #include <net/ipv6.h>
  45 #include <net/ip6_route.h>
  46 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
  47 #endif
  48 #include <net/route.h>
  49 #include <net/sock.h>
  50 #include <net/genetlink.h>
  51
  52 #include <linux/uaccess.h>
  53
  54 #include <net/ip_vs.h>
  55
  56 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
  57 static DEFINE_MUTEX(__ip_vs_mutex);
  58
  59 /* sysctl variables */
  60
  61 #ifdef CONFIG_IP_VS_DEBUG
  62 static int sysctl_ip_vs_debug_level = 0;
  63
  64 int ip_vs_get_debug_level(void)
  65 {
  66         return sysctl_ip_vs_debug_level;
  67 }
  68 #endif
  69
  70
  71 /*  Protos */
  72 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup);
  73
  74
  75 #ifdef CONFIG_IP_VS_IPV6
  76 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
  77 static bool __ip_vs_addr_is_local_v6(struct net *net,
  78                                      const struct in6_addr *addr)
  79 {
  80         struct flowi6 fl6 = {
  81                 .daddr = *addr,
  82         };
  83         struct dst_entry *dst = ip6_route_output(net, NULL, &fl6);
  84         bool is_local;
  85
  86         is_local = !dst->error && dst->dev && (dst->dev->flags & IFF_LOOPBACK);
  87
  88         dst_release(dst);
  89         return is_local;
  90 }
  91 #endif
  92
  93 #ifdef CONFIG_SYSCTL
  94 /*
  95  *      update_defense_level is called from keventd and from sysctl,
  96  *      so it needs to protect itself from softirqs
  97  */
  98 static void update_defense_level(struct netns_ipvs *ipvs)
  99 {
 100         struct sysinfo i;
 101         int availmem;
 102         int nomem;
 103         int to_change = -1;
 104
 105         /* we only count free and buffered memory (in pages) */
 106         si_meminfo(&i);
 107         availmem = i.freeram + i.bufferram;
 108         /* however in linux 2.5 the i.bufferram is total page cache size,
 109            we need adjust it */
 110         /* si_swapinfo(&i); */
 111         /* availmem = availmem - (i.totalswap - i.freeswap); */
 112
 113         nomem = (availmem < ipvs->sysctl_amemthresh);
 114
 115         local_bh_disable();
 116
 117         /* drop_entry */
 118         spin_lock(&ipvs->dropentry_lock);
 119         switch (ipvs->sysctl_drop_entry) {
 120         case 0:
 121                 atomic_set(&ipvs->dropentry, 0);
 122                 break;
 123         case 1:
 124                 if (nomem) {
 125                         atomic_set(&ipvs->dropentry, 1);
 126                         ipvs->sysctl_drop_entry = 2;
 127                 } else {
 128                         atomic_set(&ipvs->dropentry, 0);
 129                 }
 130                 break;
 131         case 2:
 132                 if (nomem) {
 133                         atomic_set(&ipvs->dropentry, 1);
 134                 } else {
 135                         atomic_set(&ipvs->dropentry, 0);
 136                         ipvs->sysctl_drop_entry = 1;
 137                 };
 138                 break;
 139         case 3:
 140                 atomic_set(&ipvs->dropentry, 1);
 141                 break;
 142         }
 143         spin_unlock(&ipvs->dropentry_lock);
 144
 145         /* drop_packet */
 146         spin_lock(&ipvs->droppacket_lock);
 147         switch (ipvs->sysctl_drop_packet) {
 148         case 0:
 149                 ipvs->drop_rate = 0;
 150                 break;
 151         case 1:
 152                 if (nomem) {
 153                         ipvs->drop_rate = ipvs->drop_counter
 154                                 = ipvs->sysctl_amemthresh /
 155                                 (ipvs->sysctl_amemthresh-availmem);
 156                         ipvs->sysctl_drop_packet = 2;
 157                 } else {
 158                         ipvs->drop_rate = 0;
 159                 }
 160                 break;
 161         case 2:
 162                 if (nomem) {
 163                         ipvs->drop_rate = ipvs->drop_counter
 164                                 = ipvs->sysctl_amemthresh /
 165                                 (ipvs->sysctl_amemthresh-availmem);
 166                 } else {
 167                         ipvs->drop_rate = 0;
 168                         ipvs->sysctl_drop_packet = 1;
 169                 }
 170                 break;
 171         case 3:
 172                 ipvs->drop_rate = ipvs->sysctl_am_droprate;
 173                 break;
 174         }
 175         spin_unlock(&ipvs->droppacket_lock);
 176
 177         /* secure_tcp */
 178         spin_lock(&ipvs->securetcp_lock);
 179         switch (ipvs->sysctl_secure_tcp) {
 180         case 0:
 181                 if (ipvs->old_secure_tcp >= 2)
 182                         to_change = 0;
 183                 break;
 184         case 1:
 185                 if (nomem) {
 186                         if (ipvs->old_secure_tcp < 2)
 187                                 to_change = 1;
 188                         ipvs->sysctl_secure_tcp = 2;
 189                 } else {
 190                         if (ipvs->old_secure_tcp >= 2)
 191                                 to_change = 0;
 192                 }
 193                 break;
 194         case 2:
 195                 if (nomem) {
 196                         if (ipvs->old_secure_tcp < 2)
 197                                 to_change = 1;
 198                 } else {
 199                         if (ipvs->old_secure_tcp >= 2)
 200                                 to_change = 0;
 201                         ipvs->sysctl_secure_tcp = 1;
 202                 }
 203                 break;
 204         case 3:
 205                 if (ipvs->old_secure_tcp < 2)
 206                         to_change = 1;
 207                 break;
 208         }
 209         ipvs->old_secure_tcp = ipvs->sysctl_secure_tcp;
 210         if (to_change >= 0)
 211                 ip_vs_protocol_timeout_change(ipvs,
 212                                               ipvs->sysctl_secure_tcp > 1);
 213         spin_unlock(&ipvs->securetcp_lock);
 214
 215         local_bh_enable();
 216 }
 217
 218
 219 /*
 220  *      Timer for checking the defense
 221  */
 222 #define DEFENSE_TIMER_PERIOD    1*HZ
 223
 224 static void defense_work_handler(struct work_struct *work)
 225 {
 226         struct netns_ipvs *ipvs =
 227                 container_of(work, struct netns_ipvs, defense_work.work);
 228
 229         update_defense_level(ipvs);
 230         if (atomic_read(&ipvs->dropentry))
 231                 ip_vs_random_dropentry(ipvs);
 232         schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
 233 }
 234 #endif
 235
 236 int
 237 ip_vs_use_count_inc(void)
 238 {
 239         return try_module_get(THIS_MODULE);
 240 }
 241
 242 void
 243 ip_vs_use_count_dec(void)
 244 {
 245         module_put(THIS_MODULE);
 246 }
 247
 248
 249 /*
 250  *      Hash table: for virtual service lookups
 251  */
 252 #define IP_VS_SVC_TAB_BITS 8
 253 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
 254 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
 255
 256 /* the service table hashed by <protocol, addr, port> */
 257 static struct hlist_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
 258 /* the service table hashed by fwmark */
 259 static struct hlist_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
 260
 261
 262 /*
 263  *      Returns hash value for virtual service
 264  */
 265 static inline unsigned int
 266 ip_vs_svc_hashkey(struct netns_ipvs *ipvs, int af, unsigned int proto,
 267                   const union nf_inet_addr *addr, __be16 port)
 268 {
 269         register unsigned int porth = ntohs(port);
 270         __be32 addr_fold = addr->ip;
 271         __u32 ahash;
 272
 273 #ifdef CONFIG_IP_VS_IPV6
 274         if (af == AF_INET6)
 275                 addr_fold = addr->ip6[0]^addr->ip6[1]^
 276                             addr->ip6[2]^addr->ip6[3];
 277 #endif
 278         ahash = ntohl(addr_fold);
 279         ahash ^= ((size_t) ipvs >> 8);
 280
 281         return (proto ^ ahash ^ (porth >> IP_VS_SVC_TAB_BITS) ^ porth) &
 282                IP_VS_SVC_TAB_MASK;
 283 }
 284
 285 /*
 286  *      Returns hash value of fwmark for virtual service lookup
 287  */
 288 static inline unsigned int ip_vs_svc_fwm_hashkey(struct netns_ipvs *ipvs, __u32 fwmark)
 289 {
 290         return (((size_t)ipvs>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK;
 291 }
 292
 293 /*
 294  *      Hashes a service in the ip_vs_svc_table by <netns,proto,addr,port>
 295  *      or in the ip_vs_svc_fwm_table by fwmark.
 296  *      Should be called with locked tables.
 297  */
 298 static int ip_vs_svc_hash(struct ip_vs_service *svc)
 299 {
 300         unsigned int hash;
 301
 302         if (svc->flags & IP_VS_SVC_F_HASHED) {
 303                 pr_err("%s(): request for already hashed, called from %pF\n",
 304                        __func__, __builtin_return_address(0));
 305                 return 0;
 306         }
 307
 308         if (svc->fwmark == 0) {
 309                 /*
 310                  *  Hash it by <netns,protocol,addr,port> in ip_vs_svc_table
 311                  */
 312                 hash = ip_vs_svc_hashkey(svc->ipvs, svc->af, svc->protocol,
 313                                          &svc->addr, svc->port);
 314                 hlist_add_head_rcu(&svc->s_list, &ip_vs_svc_table[hash]);
 315         } else {
 316                 /*
 317                  *  Hash it by fwmark in svc_fwm_table
 318                  */
 319                 hash = ip_vs_svc_fwm_hashkey(svc->ipvs, svc->fwmark);
 320                 hlist_add_head_rcu(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
 321         }
 322
 323         svc->flags |= IP_VS_SVC_F_HASHED;
 324         /* increase its refcnt because it is referenced by the svc table */
 325         atomic_inc(&svc->refcnt);
 326         return 1;
 327 }
 328
 329
 330 /*
 331  *      Unhashes a service from svc_table / svc_fwm_table.
 332  *      Should be called with locked tables.
 333  */
 334 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
 335 {
 336         if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
 337                 pr_err("%s(): request for unhash flagged, called from %pF\n",
 338                        __func__, __builtin_return_address(0));
 339                 return 0;
 340         }
 341
 342         if (svc->fwmark == 0) {
 343                 /* Remove it from the svc_table table */
 344                 hlist_del_rcu(&svc->s_list);
 345         } else {
 346                 /* Remove it from the svc_fwm_table table */
 347                 hlist_del_rcu(&svc->f_list);
 348         }
 349
 350         svc->flags &= ~IP_VS_SVC_F_HASHED;
 351         atomic_dec(&svc->refcnt);
 352         return 1;
 353 }
 354
 355
 356 /*
 357  *      Get service by {netns, proto,addr,port} in the service table.
 358  */
 359 static inline struct ip_vs_service *
 360 __ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u16 protocol,
 361                      const union nf_inet_addr *vaddr, __be16 vport)
 362 {
 363         unsigned int hash;
 364         struct ip_vs_service *svc;
 365
 366         /* Check for "full" addressed entries */
 367         hash = ip_vs_svc_hashkey(ipvs, af, protocol, vaddr, vport);
 368
 369         hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[hash], s_list) {
 370                 if ((svc->af == af)
 371                     && ip_vs_addr_equal(af, &svc->addr, vaddr)
 372                     && (svc->port == vport)
 373                     && (svc->protocol == protocol)
 374                     && (svc->ipvs == ipvs)) {
 375                         /* HIT */
 376                         return svc;
 377                 }
 378         }
 379
 380         return NULL;
 381 }
 382
 383
 384 /*
 385  *      Get service by {fwmark} in the service table.
 386  */
 387 static inline struct ip_vs_service *
 388 __ip_vs_svc_fwm_find(struct netns_ipvs *ipvs, int af, __u32 fwmark)
 389 {
 390         unsigned int hash;
 391         struct ip_vs_service *svc;
 392
 393         /* Check for fwmark addressed entries */
 394         hash = ip_vs_svc_fwm_hashkey(ipvs, fwmark);
 395
 396         hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[hash], f_list) {
 397                 if (svc->fwmark == fwmark && svc->af == af
 398                     && (svc->ipvs == ipvs)) {
 399                         /* HIT */
 400                         return svc;
 401                 }
 402         }
 403
 404         return NULL;
 405 }
 406
 407 /* Find service, called under RCU lock */
 408 struct ip_vs_service *
 409 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
 410                    const union nf_inet_addr *vaddr, __be16 vport)
 411 {
 412         struct ip_vs_service *svc;
 413
 414         /*
 415          *      Check the table hashed by fwmark first
 416          */
 417         if (fwmark) {
 418                 svc = __ip_vs_svc_fwm_find(ipvs, af, fwmark);
 419                 if (svc)
 420                         goto out;
 421         }
 422
 423         /*
 424          *      Check the table hashed by <protocol,addr,port>
 425          *      for "full" addressed entries
 426          */
 427         svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, vport);
 428
 429         if (!svc && protocol == IPPROTO_TCP &&
 430             atomic_read(&ipvs->ftpsvc_counter) &&
 431             (vport == FTPDATA || ntohs(vport) >= inet_prot_sock(ipvs->net))) {
 432                 /*
 433                  * Check if ftp service entry exists, the packet
 434                  * might belong to FTP data connections.
 435                  */
 436                 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, FTPPORT);
 437         }
 438
 439         if (svc == NULL
 440             && atomic_read(&ipvs->nullsvc_counter)) {
 441                 /*
 442                  * Check if the catch-all port (port zero) exists
 443                  */
 444                 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, 0);
 445         }
 446
 447   out:
 448         IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
 449                       fwmark, ip_vs_proto_name(protocol),
 450                       IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
 451                       svc ? "hit" : "not hit");
 452
 453         return svc;
 454 }
 455
 456
 457 static inline void
 458 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
 459 {
 460         atomic_inc(&svc->refcnt);
 461         rcu_assign_pointer(dest->svc, svc);
 462 }
 463
 464 static void ip_vs_service_free(struct ip_vs_service *svc)
 465 {
 466         free_percpu(svc->stats.cpustats);
 467         kfree(svc);
 468 }
 469
 470 static void ip_vs_service_rcu_free(struct rcu_head *head)
 471 {
 472         struct ip_vs_service *svc;
 473
 474         svc = container_of(head, struct ip_vs_service, rcu_head);
 475         ip_vs_service_free(svc);
 476 }
 477
 478 static void __ip_vs_svc_put(struct ip_vs_service *svc, bool do_delay)
 479 {
 480         if (atomic_dec_and_test(&svc->refcnt)) {
 481                 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
 482                               svc->fwmark,
 483                               IP_VS_DBG_ADDR(svc->af, &svc->addr),
 484                               ntohs(svc->port));
 485                 if (do_delay)
 486                         call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
 487                 else
 488                         ip_vs_service_free(svc);
 489         }
 490 }
 491
 492
 493 /*
 494  *      Returns hash value for real service
 495  */
 496 static inline unsigned int ip_vs_rs_hashkey(int af,
 497                                             const union nf_inet_addr *addr,
 498                                             __be16 port)
 499 {
 500         register unsigned int porth = ntohs(port);
 501         __be32 addr_fold = addr->ip;
 502
 503 #ifdef CONFIG_IP_VS_IPV6
 504         if (af == AF_INET6)
 505                 addr_fold = addr->ip6[0]^addr->ip6[1]^
 506                             addr->ip6[2]^addr->ip6[3];
 507 #endif
 508
 509         return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
 510                 & IP_VS_RTAB_MASK;
 511 }
 512
 513 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
 514 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
 515 {
 516         unsigned int hash;
 517
 518         if (dest->in_rs_table)
 519                 return;
 520
 521         /*
 522          *      Hash by proto,addr,port,
 523          *      which are the parameters of the real service.
 524          */
 525         hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
 526
 527         hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
 528         dest->in_rs_table = 1;
 529 }
 530
 531 /* Unhash ip_vs_dest from rs_table. */
 532 static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
 533 {
 534         /*
 535          * Remove it from the rs_table table.
 536          */
 537         if (dest->in_rs_table) {
 538                 hlist_del_rcu(&dest->d_list);
 539                 dest->in_rs_table = 0;
 540         }
 541 }
 542
 543 /* Check if real service by <proto,addr,port> is present */
 544 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
 545                             const union nf_inet_addr *daddr, __be16 dport)
 546 {
 547         unsigned int hash;
 548         struct ip_vs_dest *dest;
 549
 550         /* Check for "full" addressed entries */
 551         hash = ip_vs_rs_hashkey(af, daddr, dport);
 552
 553         hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
 554                 if (dest->port == dport &&
 555                     dest->af == af &&
 556                     ip_vs_addr_equal(af, &dest->addr, daddr) &&
 557                     (dest->protocol == protocol || dest->vfwmark)) {
 558                         /* HIT */
 559                         return true;
 560                 }
 561         }
 562
 563         return false;
 564 }
 565
 566 /* Find real service record by <proto,addr,port>.
 567  * In case of multiple records with the same <proto,addr,port>, only
 568  * the first found record is returned.
 569  *
 570  * To be called under RCU lock.
 571  */
 572 struct ip_vs_dest *ip_vs_find_real_service(struct netns_ipvs *ipvs, int af,
 573                                            __u16 protocol,
 574                                            const union nf_inet_addr *daddr,
 575                                            __be16 dport)
 576 {
 577         unsigned int hash;
 578         struct ip_vs_dest *dest;
 579
 580         /* Check for "full" addressed entries */
 581         hash = ip_vs_rs_hashkey(af, daddr, dport);
 582
 583         hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
 584                 if (dest->port == dport &&
 585                     dest->af == af &&
 586                     ip_vs_addr_equal(af, &dest->addr, daddr) &&
 587                         (dest->protocol == protocol || dest->vfwmark)) {
 588                         /* HIT */
 589                         return dest;
 590                 }
 591         }
 592
 593         return NULL;
 594 }
 595
 596 /* Lookup destination by {addr,port} in the given service
 597  * Called under RCU lock.
 598  */
 599 static struct ip_vs_dest *
 600 ip_vs_lookup_dest(struct ip_vs_service *svc, int dest_af,
 601                   const union nf_inet_addr *daddr, __be16 dport)
 602 {
 603         struct ip_vs_dest *dest;
 604
 605         /*
 606          * Find the destination for the given service
 607          */
 608         list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
 609                 if ((dest->af == dest_af) &&
 610                     ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
 611                     (dest->port == dport)) {
 612                         /* HIT */
 613                         return dest;
 614                 }
 615         }
 616
 617         return NULL;
 618 }
 619
 620 /*
 621  * Find destination by {daddr,dport,vaddr,protocol}
 622  * Created to be used in ip_vs_process_message() in
 623  * the backup synchronization daemon. It finds the
 624  * destination to be bound to the received connection
 625  * on the backup.
 626  * Called under RCU lock, no refcnt is returned.
 627  */
 628 struct ip_vs_dest *ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
 629                                    const union nf_inet_addr *daddr,
 630                                    __be16 dport,
 631                                    const union nf_inet_addr *vaddr,
 632                                    __be16 vport, __u16 protocol, __u32 fwmark,
 633                                    __u32 flags)
 634 {
 635         struct ip_vs_dest *dest;
 636         struct ip_vs_service *svc;
 637         __be16 port = dport;
 638
 639         svc = ip_vs_service_find(ipvs, svc_af, fwmark, protocol, vaddr, vport);
 640         if (!svc)
 641                 return NULL;
 642         if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
 643                 port = 0;
 644         dest = ip_vs_lookup_dest(svc, dest_af, daddr, port);
 645         if (!dest)
 646                 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port ^ dport);
 647         return dest;
 648 }
 649
 650 void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
 651 {
 652         struct ip_vs_dest_dst *dest_dst = container_of(head,
 653                                                        struct ip_vs_dest_dst,
 654                                                        rcu_head);
 655
 656         dst_release(dest_dst->dst_cache);
 657         kfree(dest_dst);
 658 }
 659
 660 /* Release dest_dst and dst_cache for dest in user context */
 661 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
 662 {
 663         struct ip_vs_dest_dst *old;
 664
 665         old = rcu_dereference_protected(dest->dest_dst, 1);
 666         if (old) {
 667                 RCU_INIT_POINTER(dest->dest_dst, NULL);
 668                 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
 669         }
 670 }
 671
 672 /*
 673  *  Lookup dest by {svc,addr,port} in the destination trash.
 674  *  The destination trash is used to hold the destinations that are removed
 675  *  from the service table but are still referenced by some conn entries.
 676  *  The reason to add the destination trash is when the dest is temporary
 677  *  down (either by administrator or by monitor program), the dest can be
 678  *  picked back from the trash, the remaining connections to the dest can
 679  *  continue, and the counting information of the dest is also useful for
 680  *  scheduling.
 681  */
 682 static struct ip_vs_dest *
 683 ip_vs_trash_get_dest(struct ip_vs_service *svc, int dest_af,
 684                      const union nf_inet_addr *daddr, __be16 dport)
 685 {
 686         struct ip_vs_dest *dest;
 687         struct netns_ipvs *ipvs = svc->ipvs;
 688
 689         /*
 690          * Find the destination in trash
 691          */
 692         spin_lock_bh(&ipvs->dest_trash_lock);
 693         list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
 694                 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
 695                               "dest->refcnt=%d\n",
 696                               dest->vfwmark,
 697                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
 698                               ntohs(dest->port),
 699                               refcount_read(&dest->refcnt));
 700                 if (dest->af == dest_af &&
 701                     ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
 702                     dest->port == dport &&
 703                     dest->vfwmark == svc->fwmark &&
 704                     dest->protocol == svc->protocol &&
 705                     (svc->fwmark ||
 706                      (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
 707                       dest->vport == svc->port))) {
 708                         /* HIT */
 709                         list_del(&dest->t_list);
 710                         goto out;
 711                 }
 712         }
 713
 714         dest = NULL;
 715
 716 out:
 717         spin_unlock_bh(&ipvs->dest_trash_lock);
 718
 719         return dest;
 720 }
 721
 722 static void ip_vs_dest_free(struct ip_vs_dest *dest)
 723 {
 724         struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1);
 725
 726         __ip_vs_dst_cache_reset(dest);
 727         __ip_vs_svc_put(svc, false);
 728         free_percpu(dest->stats.cpustats);
 729         ip_vs_dest_put_and_free(dest);
 730 }
 731
 732 /*
 733  *  Clean up all the destinations in the trash
 734  *  Called by the ip_vs_control_cleanup()
 735  *
 736  *  When the ip_vs_control_clearup is activated by ipvs module exit,
 737  *  the service tables must have been flushed and all the connections
 738  *  are expired, and the refcnt of each destination in the trash must
 739  *  be 1, so we simply release them here.
 740  */
 741 static void ip_vs_trash_cleanup(struct netns_ipvs *ipvs)
 742 {
 743         struct ip_vs_dest *dest, *nxt;
 744
 745         del_timer_sync(&ipvs->dest_trash_timer);
 746         /* No need to use dest_trash_lock */
 747         list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
 748                 list_del(&dest->t_list);
 749                 ip_vs_dest_free(dest);
 750         }
 751 }
 752
 753 static void
 754 ip_vs_copy_stats(struct ip_vs_kstats *dst, struct ip_vs_stats *src)
 755 {
 756 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c
 757
 758         spin_lock_bh(&src->lock);
 759
 760         IP_VS_SHOW_STATS_COUNTER(conns);
 761         IP_VS_SHOW_STATS_COUNTER(inpkts);
 762         IP_VS_SHOW_STATS_COUNTER(outpkts);
 763         IP_VS_SHOW_STATS_COUNTER(inbytes);
 764         IP_VS_SHOW_STATS_COUNTER(outbytes);
 765
 766         ip_vs_read_estimator(dst, src);
 767
 768         spin_unlock_bh(&src->lock);
 769 }
 770
 771 static void
 772 ip_vs_export_stats_user(struct ip_vs_stats_user *dst, struct ip_vs_kstats *src)
 773 {
 774         dst->conns = (u32)src->conns;
 775         dst->inpkts = (u32)src->inpkts;
 776         dst->outpkts = (u32)src->outpkts;
 777         dst->inbytes = src->inbytes;
 778         dst->outbytes = src->outbytes;
 779         dst->cps = (u32)src->cps;
 780         dst->inpps = (u32)src->inpps;
 781         dst->outpps = (u32)src->outpps;
 782         dst->inbps = (u32)src->inbps;
 783         dst->outbps = (u32)src->outbps;
 784 }
 785
 786 static void
 787 ip_vs_zero_stats(struct ip_vs_stats *stats)
 788 {
 789         spin_lock_bh(&stats->lock);
 790
 791         /* get current counters as zero point, rates are zeroed */
 792
 793 #define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c
 794
 795         IP_VS_ZERO_STATS_COUNTER(conns);
 796         IP_VS_ZERO_STATS_COUNTER(inpkts);
 797         IP_VS_ZERO_STATS_COUNTER(outpkts);
 798         IP_VS_ZERO_STATS_COUNTER(inbytes);
 799         IP_VS_ZERO_STATS_COUNTER(outbytes);
 800
 801         ip_vs_zero_estimator(stats);
 802
 803         spin_unlock_bh(&stats->lock);
 804 }
 805
 806 /*
 807  *      Update a destination in the given service
 808  */
 809 static void
 810 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
 811                     struct ip_vs_dest_user_kern *udest, int add)
 812 {
 813         struct netns_ipvs *ipvs = svc->ipvs;
 814         struct ip_vs_service *old_svc;
 815         struct ip_vs_scheduler *sched;
 816         int conn_flags;
 817
 818         /* We cannot modify an address and change the address family */
 819         BUG_ON(!add && udest->af != dest->af);
 820
 821         if (add && udest->af != svc->af)
 822                 ipvs->mixed_address_family_dests++;
 823
 824         /* set the weight and the flags */
 825         atomic_set(&dest->weight, udest->weight);
 826         conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
 827         conn_flags |= IP_VS_CONN_F_INACTIVE;
 828
 829         /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
 830         if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
 831                 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
 832         } else {
 833                 /*
 834                  *    Put the real service in rs_table if not present.
 835                  *    For now only for NAT!
 836                  */
 837                 ip_vs_rs_hash(ipvs, dest);
 838         }
 839         atomic_set(&dest->conn_flags, conn_flags);
 840
 841         /* bind the service */
 842         old_svc = rcu_dereference_protected(dest->svc, 1);
 843         if (!old_svc) {
 844                 __ip_vs_bind_svc(dest, svc);
 845         } else {
 846                 if (old_svc != svc) {
 847                         ip_vs_zero_stats(&dest->stats);
 848                         __ip_vs_bind_svc(dest, svc);
 849                         __ip_vs_svc_put(old_svc, true);
 850                 }
 851         }
 852
 853         /* set the dest status flags */
 854         dest->flags |= IP_VS_DEST_F_AVAILABLE;
 855
 856         if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
 857                 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
 858         dest->u_threshold = udest->u_threshold;
 859         dest->l_threshold = udest->l_threshold;
 860
 861         dest->af = udest->af;
 862
 863         spin_lock_bh(&dest->dst_lock);
 864         __ip_vs_dst_cache_reset(dest);
 865         spin_unlock_bh(&dest->dst_lock);
 866
 867         if (add) {
 868                 ip_vs_start_estimator(svc->ipvs, &dest->stats);
 869                 list_add_rcu(&dest->n_list, &svc->destinations);
 870                 svc->num_dests++;
 871                 sched = rcu_dereference_protected(svc->scheduler, 1);
 872                 if (sched && sched->add_dest)
 873                         sched->add_dest(svc, dest);
 874         } else {
 875                 sched = rcu_dereference_protected(svc->scheduler, 1);
 876                 if (sched && sched->upd_dest)
 877                         sched->upd_dest(svc, dest);
 878         }
 879 }
 880
 881
 882 /*
 883  *      Create a destination for the given service
 884  */
 885 static int
 886 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
 887                struct ip_vs_dest **dest_p)
 888 {
 889         struct ip_vs_dest *dest;
 890         unsigned int atype, i;
 891
 892         EnterFunction(2);
 893
 894 #ifdef CONFIG_IP_VS_IPV6
 895         if (udest->af == AF_INET6) {
 896                 int ret;
 897
 898                 atype = ipv6_addr_type(&udest->addr.in6);
 899                 if ((!(atype & IPV6_ADDR_UNICAST) ||
 900                         atype & IPV6_ADDR_LINKLOCAL) &&
 901                         !__ip_vs_addr_is_local_v6(svc->ipvs->net, &udest->addr.in6))
 902                         return -EINVAL;
 903
 904                 ret = nf_defrag_ipv6_enable(svc->ipvs->net);
 905                 if (ret)
 906                         return ret;
 907         } else
 908 #endif
 909         {
 910                 atype = inet_addr_type(svc->ipvs->net, udest->addr.ip);
 911                 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
 912                         return -EINVAL;
 913         }
 914
 915         dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
 916         if (dest == NULL)
 917                 return -ENOMEM;
 918
 919         dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
 920         if (!dest->stats.cpustats)
 921                 goto err_alloc;
 922
 923         for_each_possible_cpu(i) {
 924                 struct ip_vs_cpu_stats *ip_vs_dest_stats;
 925                 ip_vs_dest_stats = per_cpu_ptr(dest->stats.cpustats, i);
 926                 u64_stats_init(&ip_vs_dest_stats->syncp);
 927         }
 928
 929         dest->af = udest->af;
 930         dest->protocol = svc->protocol;
 931         dest->vaddr = svc->addr;
 932         dest->vport = svc->port;
 933         dest->vfwmark = svc->fwmark;
 934         ip_vs_addr_copy(udest->af, &dest->addr, &udest->addr);
 935         dest->port = udest->port;
 936
 937         atomic_set(&dest->activeconns, 0);
 938         atomic_set(&dest->inactconns, 0);
 939         atomic_set(&dest->persistconns, 0);
 940         refcount_set(&dest->refcnt, 1);
 941
 942         INIT_HLIST_NODE(&dest->d_list);
 943         spin_lock_init(&dest->dst_lock);
 944         spin_lock_init(&dest->stats.lock);
 945         __ip_vs_update_dest(svc, dest, udest, 1);
 946
 947         *dest_p = dest;
 948
 949         LeaveFunction(2);
 950         return 0;
 951
 952 err_alloc:
 953         kfree(dest);
 954         return -ENOMEM;
 955 }
 956
 957
 958 /*
 959  *      Add a destination into an existing service
 960  */
 961 static int
 962 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
 963 {
 964         struct ip_vs_dest *dest;
 965         union nf_inet_addr daddr;
 966         __be16 dport = udest->port;
 967         int ret;
 968
 969         EnterFunction(2);
 970
 971         if (udest->weight < 0) {
 972                 pr_err("%s(): server weight less than zero\n", __func__);
 973                 return -ERANGE;
 974         }
 975
 976         if (udest->l_threshold > udest->u_threshold) {
 977                 pr_err("%s(): lower threshold is higher than upper threshold\n",
 978                         __func__);
 979                 return -ERANGE;
 980         }
 981
 982         ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
 983
 984         /* We use function that requires RCU lock */
 985         rcu_read_lock();
 986         dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
 987         rcu_read_unlock();
 988
 989         if (dest != NULL) {
 990                 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
 991                 return -EEXIST;
 992         }
 993
 994         /*
 995          * Check if the dest already exists in the trash and
 996          * is from the same service
 997          */
 998         dest = ip_vs_trash_get_dest(svc, udest->af, &daddr, dport);
 999
1000         if (dest != NULL) {
1001                 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
1002                               "dest->refcnt=%d, service %u/%s:%u\n",
1003                               IP_VS_DBG_ADDR(udest->af, &daddr), ntohs(dport),
1004                               refcount_read(&dest->refcnt),
1005                               dest->vfwmark,
1006                               IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
1007                               ntohs(dest->vport));
1008
1009                 __ip_vs_update_dest(svc, dest, udest, 1);
1010                 ret = 0;
1011         } else {
1012                 /*
1013                  * Allocate and initialize the dest structure
1014                  */
1015                 ret = ip_vs_new_dest(svc, udest, &dest);
1016         }
1017         LeaveFunction(2);
1018
1019         return ret;
1020 }
1021
1022
1023 /*
1024  *      Edit a destination in the given service
1025  */
1026 static int
1027 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1028 {
1029         struct ip_vs_dest *dest;
1030         union nf_inet_addr daddr;
1031         __be16 dport = udest->port;
1032
1033         EnterFunction(2);
1034
1035         if (udest->weight < 0) {
1036                 pr_err("%s(): server weight less than zero\n", __func__);
1037                 return -ERANGE;
1038         }
1039
1040         if (udest->l_threshold > udest->u_threshold) {
1041                 pr_err("%s(): lower threshold is higher than upper threshold\n",
1042                         __func__);
1043                 return -ERANGE;
1044         }
1045
1046         ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1047
1048         /* We use function that requires RCU lock */
1049         rcu_read_lock();
1050         dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1051         rcu_read_unlock();
1052
1053         if (dest == NULL) {
1054                 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1055                 return -ENOENT;
1056         }
1057
1058         __ip_vs_update_dest(svc, dest, udest, 0);
1059         LeaveFunction(2);
1060
1061         return 0;
1062 }
1063
1064 /*
1065  *      Delete a destination (must be already unlinked from the service)
1066  */
1067 static void __ip_vs_del_dest(struct netns_ipvs *ipvs, struct ip_vs_dest *dest,
1068                              bool cleanup)
1069 {
1070         ip_vs_stop_estimator(ipvs, &dest->stats);
1071
1072         /*
1073          *  Remove it from the d-linked list with the real services.
1074          */
1075         ip_vs_rs_unhash(dest);
1076
1077         spin_lock_bh(&ipvs->dest_trash_lock);
1078         IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1079                       IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1080                       refcount_read(&dest->refcnt));
1081         if (list_empty(&ipvs->dest_trash) && !cleanup)
1082                 mod_timer(&ipvs->dest_trash_timer,
1083                           jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1084         /* dest lives in trash with reference */
1085         list_add(&dest->t_list, &ipvs->dest_trash);
1086         dest->idle_start = 0;
1087         spin_unlock_bh(&ipvs->dest_trash_lock);
1088 }
1089
1090
1091 /*
1092  *      Unlink a destination from the given service
1093  */
1094 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1095                                 struct ip_vs_dest *dest,
1096                                 int svcupd)
1097 {
1098         dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1099
1100         /*
1101          *  Remove it from the d-linked destination list.
1102          */
1103         list_del_rcu(&dest->n_list);
1104         svc->num_dests--;
1105
1106         if (dest->af != svc->af)
1107                 svc->ipvs->mixed_address_family_dests--;
1108
1109         if (svcupd) {
1110                 struct ip_vs_scheduler *sched;
1111
1112                 sched = rcu_dereference_protected(svc->scheduler, 1);
1113                 if (sched && sched->del_dest)
1114                         sched->del_dest(svc, dest);
1115         }
1116 }
1117
1118
1119 /*
1120  *      Delete a destination server in the given service
1121  */
1122 static int
1123 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1124 {
1125         struct ip_vs_dest *dest;
1126         __be16 dport = udest->port;
1127
1128         EnterFunction(2);
1129
1130         /* We use function that requires RCU lock */
1131         rcu_read_lock();
1132         dest = ip_vs_lookup_dest(svc, udest->af, &udest->addr, dport);
1133         rcu_read_unlock();
1134
1135         if (dest == NULL) {
1136                 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1137                 return -ENOENT;
1138         }
1139
1140         /*
1141          *      Unlink dest from the service
1142          */
1143         __ip_vs_unlink_dest(svc, dest, 1);
1144
1145         /*
1146          *      Delete the destination
1147          */
1148         __ip_vs_del_dest(svc->ipvs, dest, false);
1149
1150         LeaveFunction(2);
1151
1152         return 0;
1153 }
1154
1155 static void ip_vs_dest_trash_expire(unsigned long data)
1156 {
1157         struct netns_ipvs *ipvs = (struct netns_ipvs *)data;
1158         struct ip_vs_dest *dest, *next;
1159         unsigned long now = jiffies;
1160
1161         spin_lock(&ipvs->dest_trash_lock);
1162         list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1163                 if (refcount_read(&dest->refcnt) > 1)
1164                         continue;
1165                 if (dest->idle_start) {
1166                         if (time_before(now, dest->idle_start +
1167                                              IP_VS_DEST_TRASH_PERIOD))
1168                                 continue;
1169                 } else {
1170                         dest->idle_start = max(1UL, now);
1171                         continue;
1172                 }
1173                 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1174                               dest->vfwmark,
1175                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1176                               ntohs(dest->port));
1177                 list_del(&dest->t_list);
1178                 ip_vs_dest_free(dest);
1179         }
1180         if (!list_empty(&ipvs->dest_trash))
1181                 mod_timer(&ipvs->dest_trash_timer,
1182                           jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1183         spin_unlock(&ipvs->dest_trash_lock);
1184 }
1185
1186 /*
1187  *      Add a service into the service hash table
1188  */
1189 static int
1190 ip_vs_add_service(struct netns_ipvs *ipvs, struct ip_vs_service_user_kern *u,
1191                   struct ip_vs_service **svc_p)
1192 {
1193         int ret = 0, i;
1194         struct ip_vs_scheduler *sched = NULL;
1195         struct ip_vs_pe *pe = NULL;
1196         struct ip_vs_service *svc = NULL;
1197
1198         /* increase the module use count */
1199         if (!ip_vs_use_count_inc())
1200                 return -ENOPROTOOPT;
1201
1202         /* Lookup the scheduler by 'u->sched_name' */
1203         if (strcmp(u->sched_name, "none")) {
1204                 sched = ip_vs_scheduler_get(u->sched_name);
1205                 if (!sched) {
1206                         pr_info("Scheduler module ip_vs_%s not found\n",
1207                                 u->sched_name);
1208                         ret = -ENOENT;
1209                         goto out_err;
1210                 }
1211         }
1212
1213         if (u->pe_name && *u->pe_name) {
1214                 pe = ip_vs_pe_getbyname(u->pe_name);
1215                 if (pe == NULL) {
1216                         pr_info("persistence engine module ip_vs_pe_%s "
1217                                 "not found\n", u->pe_name);
1218                         ret = -ENOENT;
1219                         goto out_err;
1220                 }
1221         }
1222
1223 #ifdef CONFIG_IP_VS_IPV6
1224         if (u->af == AF_INET6) {
1225                 __u32 plen = (__force __u32) u->netmask;
1226
1227                 if (plen < 1 || plen > 128) {
1228                         ret = -EINVAL;
1229                         goto out_err;
1230                 }
1231
1232                 ret = nf_defrag_ipv6_enable(ipvs->net);
1233                 if (ret)
1234                         goto out_err;
1235         }
1236 #endif
1237
1238         svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1239         if (svc == NULL) {
1240                 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1241                 ret = -ENOMEM;
1242                 goto out_err;
1243         }
1244         svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1245         if (!svc->stats.cpustats) {
1246                 ret = -ENOMEM;
1247                 goto out_err;
1248         }
1249
1250         for_each_possible_cpu(i) {
1251                 struct ip_vs_cpu_stats *ip_vs_stats;
1252                 ip_vs_stats = per_cpu_ptr(svc->stats.cpustats, i);
1253                 u64_stats_init(&ip_vs_stats->syncp);
1254         }
1255
1256
1257         /* I'm the first user of the service */
1258         atomic_set(&svc->refcnt, 0);
1259
1260         svc->af = u->af;
1261         svc->protocol = u->protocol;
1262         ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1263         svc->port = u->port;
1264         svc->fwmark = u->fwmark;
1265         svc->flags = u->flags & ~IP_VS_SVC_F_HASHED;
1266         svc->timeout = u->timeout * HZ;
1267         svc->netmask = u->netmask;
1268         svc->ipvs = ipvs;
1269
1270         INIT_LIST_HEAD(&svc->destinations);
1271         spin_lock_init(&svc->sched_lock);
1272         spin_lock_init(&svc->stats.lock);
1273
1274         /* Bind the scheduler */
1275         if (sched) {
1276                 ret = ip_vs_bind_scheduler(svc, sched);
1277                 if (ret)
1278                         goto out_err;
1279                 sched = NULL;
1280         }
1281
1282         /* Bind the ct retriever */
1283         RCU_INIT_POINTER(svc->pe, pe);
1284         pe = NULL;
1285
1286         /* Update the virtual service counters */
1287         if (svc->port == FTPPORT)
1288                 atomic_inc(&ipvs->ftpsvc_counter);
1289         else if (svc->port == 0)
1290                 atomic_inc(&ipvs->nullsvc_counter);
1291         if (svc->pe && svc->pe->conn_out)
1292                 atomic_inc(&ipvs->conn_out_counter);
1293
1294         ip_vs_start_estimator(ipvs, &svc->stats);
1295
1296         /* Count only IPv4 services for old get/setsockopt interface */
1297         if (svc->af == AF_INET)
1298                 ipvs->num_services++;
1299
1300         /* Hash the service into the service table */
1301         ip_vs_svc_hash(svc);
1302
1303         *svc_p = svc;
1304         /* Now there is a service - full throttle */
1305         ipvs->enable = 1;
1306         return 0;
1307
1308
1309  out_err:
1310         if (svc != NULL) {
1311                 ip_vs_unbind_scheduler(svc, sched);
1312                 ip_vs_service_free(svc);
1313         }
1314         ip_vs_scheduler_put(sched);
1315         ip_vs_pe_put(pe);
1316
1317         /* decrease the module use count */
1318         ip_vs_use_count_dec();
1319
1320         return ret;
1321 }
1322
1323
1324 /*
1325  *      Edit a service and bind it with a new scheduler
1326  */
1327 static int
1328 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1329 {
1330         struct ip_vs_scheduler *sched = NULL, *old_sched;
1331         struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1332         int ret = 0;
1333         bool new_pe_conn_out, old_pe_conn_out;
1334
1335         /*
1336          * Lookup the scheduler, by 'u->sched_name'
1337          */
1338         if (strcmp(u->sched_name, "none")) {
1339                 sched = ip_vs_scheduler_get(u->sched_name);
1340                 if (!sched) {
1341                         pr_info("Scheduler module ip_vs_%s not found\n",
1342                                 u->sched_name);
1343                         return -ENOENT;
1344                 }
1345         }
1346         old_sched = sched;
1347
1348         if (u->pe_name && *u->pe_name) {
1349                 pe = ip_vs_pe_getbyname(u->pe_name);
1350                 if (pe == NULL) {
1351                         pr_info("persistence engine module ip_vs_pe_%s "
1352                                 "not found\n", u->pe_name);
1353                         ret = -ENOENT;
1354                         goto out;
1355                 }
1356                 old_pe = pe;
1357         }
1358
1359 #ifdef CONFIG_IP_VS_IPV6
1360         if (u->af == AF_INET6) {
1361                 __u32 plen = (__force __u32) u->netmask;
1362
1363                 if (plen < 1 || plen > 128) {
1364                         ret = -EINVAL;
1365                         goto out;
1366                 }
1367         }
1368 #endif
1369
1370         old_sched = rcu_dereference_protected(svc->scheduler, 1);
1371         if (sched != old_sched) {
1372                 if (old_sched) {
1373                         ip_vs_unbind_scheduler(svc, old_sched);
1374                         RCU_INIT_POINTER(svc->scheduler, NULL);
1375                         /* Wait all svc->sched_data users */
1376                         synchronize_rcu();
1377                 }
1378                 /* Bind the new scheduler */
1379                 if (sched) {
1380                         ret = ip_vs_bind_scheduler(svc, sched);
1381                         if (ret) {
1382                                 ip_vs_scheduler_put(sched);
1383                                 goto out;
1384                         }
1385                 }
1386         }
1387
1388         /*
1389          * Set the flags and timeout value
1390          */
1391         svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1392         svc->timeout = u->timeout * HZ;
1393         svc->netmask = u->netmask;
1394
1395         old_pe = rcu_dereference_protected(svc->pe, 1);
1396         if (pe != old_pe) {
1397                 rcu_assign_pointer(svc->pe, pe);
1398                 /* check for optional methods in new pe */
1399                 new_pe_conn_out = (pe && pe->conn_out) ? true : false;
1400                 old_pe_conn_out = (old_pe && old_pe->conn_out) ? true : false;
1401                 if (new_pe_conn_out && !old_pe_conn_out)
1402                         atomic_inc(&svc->ipvs->conn_out_counter);
1403                 if (old_pe_conn_out && !new_pe_conn_out)
1404                         atomic_dec(&svc->ipvs->conn_out_counter);
1405         }
1406
1407 out:
1408         ip_vs_scheduler_put(old_sched);
1409         ip_vs_pe_put(old_pe);
1410         return ret;
1411 }
1412
1413 /*
1414  *      Delete a service from the service list
1415  *      - The service must be unlinked, unlocked and not referenced!
1416  *      - We are called under _bh lock
1417  */
1418 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1419 {
1420         struct ip_vs_dest *dest, *nxt;
1421         struct ip_vs_scheduler *old_sched;
1422         struct ip_vs_pe *old_pe;
1423         struct netns_ipvs *ipvs = svc->ipvs;
1424
1425         /* Count only IPv4 services for old get/setsockopt interface */
1426         if (svc->af == AF_INET)
1427                 ipvs->num_services--;
1428
1429         ip_vs_stop_estimator(svc->ipvs, &svc->stats);
1430
1431         /* Unbind scheduler */
1432         old_sched = rcu_dereference_protected(svc->scheduler, 1);
1433         ip_vs_unbind_scheduler(svc, old_sched);
1434         ip_vs_scheduler_put(old_sched);
1435
1436         /* Unbind persistence engine, keep svc->pe */
1437         old_pe = rcu_dereference_protected(svc->pe, 1);
1438         if (old_pe && old_pe->conn_out)
1439                 atomic_dec(&ipvs->conn_out_counter);
1440         ip_vs_pe_put(old_pe);
1441
1442         /*
1443          *    Unlink the whole destination list
1444          */
1445         list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1446                 __ip_vs_unlink_dest(svc, dest, 0);
1447                 __ip_vs_del_dest(svc->ipvs, dest, cleanup);
1448         }
1449
1450         /*
1451          *    Update the virtual service counters
1452          */
1453         if (svc->port == FTPPORT)
1454                 atomic_dec(&ipvs->ftpsvc_counter);
1455         else if (svc->port == 0)
1456                 atomic_dec(&ipvs->nullsvc_counter);
1457
1458         /*
1459          *    Free the service if nobody refers to it
1460          */
1461         __ip_vs_svc_put(svc, true);
1462
1463         /* decrease the module use count */
1464         ip_vs_use_count_dec();
1465 }
1466
1467 /*
1468  * Unlink a service from list and try to delete it if its refcnt reached 0
1469  */
1470 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1471 {
1472         /* Hold svc to avoid double release from dest_trash */
1473         atomic_inc(&svc->refcnt);
1474         /*
1475          * Unhash it from the service table
1476          */
1477         ip_vs_svc_unhash(svc);
1478
1479         __ip_vs_del_service(svc, cleanup);
1480 }
1481
1482 /*
1483  *      Delete a service from the service list
1484  */
1485 static int ip_vs_del_service(struct ip_vs_service *svc)
1486 {
1487         if (svc == NULL)
1488                 return -EEXIST;
1489         ip_vs_unlink_service(svc, false);
1490
1491         return 0;
1492 }
1493
1494
1495 /*
1496  *      Flush all the virtual services
1497  */
1498 static int ip_vs_flush(struct netns_ipvs *ipvs, bool cleanup)
1499 {
1500         int idx;
1501         struct ip_vs_service *svc;
1502         struct hlist_node *n;
1503
1504         /*
1505          * Flush the service table hashed by <netns,protocol,addr,port>
1506          */
1507         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1508                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1509                                           s_list) {
1510                         if (svc->ipvs == ipvs)
1511                                 ip_vs_unlink_service(svc, cleanup);
1512                 }
1513         }
1514
1515         /*
1516          * Flush the service table hashed by fwmark
1517          */
1518         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1519                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1520                                           f_list) {
1521                         if (svc->ipvs == ipvs)
1522                                 ip_vs_unlink_service(svc, cleanup);
1523                 }
1524         }
1525
1526         return 0;
1527 }
1528
1529 /*
1530  *      Delete service by {netns} in the service table.
1531  *      Called by __ip_vs_cleanup()
1532  */
1533 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs)
1534 {
1535         EnterFunction(2);
1536         /* Check for "full" addressed entries */
1537         mutex_lock(&__ip_vs_mutex);
1538         ip_vs_flush(ipvs, true);
1539         mutex_unlock(&__ip_vs_mutex);
1540         LeaveFunction(2);
1541 }
1542
1543 /* Put all references for device (dst_cache) */
1544 static inline void
1545 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1546 {
1547         struct ip_vs_dest_dst *dest_dst;
1548
1549         spin_lock_bh(&dest->dst_lock);
1550         dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1551         if (dest_dst && dest_dst->dst_cache->dev == dev) {
1552                 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1553                               dev->name,
1554                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1555                               ntohs(dest->port),
1556                               refcount_read(&dest->refcnt));
1557                 __ip_vs_dst_cache_reset(dest);
1558         }
1559         spin_unlock_bh(&dest->dst_lock);
1560
1561 }
1562 /* Netdev event receiver
1563  * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1564  */
1565 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1566                            void *ptr)
1567 {
1568         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1569         struct net *net = dev_net(dev);
1570         struct netns_ipvs *ipvs = net_ipvs(net);
1571         struct ip_vs_service *svc;
1572         struct ip_vs_dest *dest;
1573         unsigned int idx;
1574
1575         if (event != NETDEV_DOWN || !ipvs)
1576                 return NOTIFY_DONE;
1577         IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1578         EnterFunction(2);
1579         mutex_lock(&__ip_vs_mutex);
1580         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1581                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1582                         if (svc->ipvs == ipvs) {
1583                                 list_for_each_entry(dest, &svc->destinations,
1584                                                     n_list) {
1585                                         ip_vs_forget_dev(dest, dev);
1586                                 }
1587                         }
1588                 }
1589
1590                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1591                         if (svc->ipvs == ipvs) {
1592                                 list_for_each_entry(dest, &svc->destinations,
1593                                                     n_list) {
1594                                         ip_vs_forget_dev(dest, dev);
1595                                 }
1596                         }
1597
1598                 }
1599         }
1600
1601         spin_lock_bh(&ipvs->dest_trash_lock);
1602         list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1603                 ip_vs_forget_dev(dest, dev);
1604         }
1605         spin_unlock_bh(&ipvs->dest_trash_lock);
1606         mutex_unlock(&__ip_vs_mutex);
1607         LeaveFunction(2);
1608         return NOTIFY_DONE;
1609 }
1610
1611 /*
1612  *      Zero counters in a service or all services
1613  */
1614 static int ip_vs_zero_service(struct ip_vs_service *svc)
1615 {
1616         struct ip_vs_dest *dest;
1617
1618         list_for_each_entry(dest, &svc->destinations, n_list) {
1619                 ip_vs_zero_stats(&dest->stats);
1620         }
1621         ip_vs_zero_stats(&svc->stats);
1622         return 0;
1623 }
1624
1625 static int ip_vs_zero_all(struct netns_ipvs *ipvs)
1626 {
1627         int idx;
1628         struct ip_vs_service *svc;
1629
1630         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1631                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1632                         if (svc->ipvs == ipvs)
1633                                 ip_vs_zero_service(svc);
1634                 }
1635         }
1636
1637         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1638                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1639                         if (svc->ipvs == ipvs)
1640                                 ip_vs_zero_service(svc);
1641                 }
1642         }
1643
1644         ip_vs_zero_stats(&ipvs->tot_stats);
1645         return 0;
1646 }
1647
1648 #ifdef CONFIG_SYSCTL
1649
1650 static int zero;
1651 static int three = 3;
1652
1653 static int
1654 proc_do_defense_mode(struct ctl_table *table, int write,
1655                      void __user *buffer, size_t *lenp, loff_t *ppos)
1656 {
1657         struct netns_ipvs *ipvs = table->extra2;
1658         int *valp = table->data;
1659         int val = *valp;
1660         int rc;
1661
1662         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1663         if (write && (*valp != val)) {
1664                 if ((*valp < 0) || (*valp > 3)) {
1665                         /* Restore the correct value */
1666                         *valp = val;
1667                 } else {
1668                         update_defense_level(ipvs);
1669                 }
1670         }
1671         return rc;
1672 }
1673
1674 static int
1675 proc_do_sync_threshold(struct ctl_table *table, int write,
1676                        void __user *buffer, size_t *lenp, loff_t *ppos)
1677 {
1678         int *valp = table->data;
1679         int val[2];
1680         int rc;
1681
1682         /* backup the value first */
1683         memcpy(val, valp, sizeof(val));
1684
1685         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1686         if (write && (valp[0] < 0 || valp[1] < 0 ||
1687             (valp[0] >= valp[1] && valp[1]))) {
1688                 /* Restore the correct value */
1689                 memcpy(valp, val, sizeof(val));
1690         }
1691         return rc;
1692 }
1693
1694 static int
1695 proc_do_sync_mode(struct ctl_table *table, int write,
1696                      void __user *buffer, size_t *lenp, loff_t *ppos)
1697 {
1698         int *valp = table->data;
1699         int val = *valp;
1700         int rc;
1701
1702         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1703         if (write && (*valp != val)) {
1704                 if ((*valp < 0) || (*valp > 1)) {
1705                         /* Restore the correct value */
1706                         *valp = val;
1707                 }
1708         }
1709         return rc;
1710 }
1711
1712 static int
1713 proc_do_sync_ports(struct ctl_table *table, int write,
1714                    void __user *buffer, size_t *lenp, loff_t *ppos)
1715 {
1716         int *valp = table->data;
1717         int val = *valp;
1718         int rc;
1719
1720         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1721         if (write && (*valp != val)) {
1722                 if (*valp < 1 || !is_power_of_2(*valp)) {
1723                         /* Restore the correct value */
1724                         *valp = val;
1725                 }
1726         }
1727         return rc;
1728 }
1729
1730 /*
1731  *      IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1732  *      Do not change order or insert new entries without
1733  *      align with netns init in ip_vs_control_net_init()
1734  */
1735
1736 static struct ctl_table vs_vars[] = {
1737         {
1738                 .procname       = "amemthresh",
1739                 .maxlen         = sizeof(int),
1740                 .mode           = 0644,
1741                 .proc_handler   = proc_dointvec,
1742         },
1743         {
1744                 .procname       = "am_droprate",
1745                 .maxlen         = sizeof(int),
1746                 .mode           = 0644,
1747                 .proc_handler   = proc_dointvec,
1748         },
1749         {
1750                 .procname       = "drop_entry",
1751                 .maxlen         = sizeof(int),
1752                 .mode           = 0644,
1753                 .proc_handler   = proc_do_defense_mode,
1754         },
1755         {
1756                 .procname       = "drop_packet",
1757                 .maxlen         = sizeof(int),
1758                 .mode           = 0644,
1759                 .proc_handler   = proc_do_defense_mode,
1760         },
1761 #ifdef CONFIG_IP_VS_NFCT
1762         {
1763                 .procname       = "conntrack",
1764                 .maxlen         = sizeof(int),
1765                 .mode           = 0644,
1766                 .proc_handler   = &proc_dointvec,
1767         },
1768 #endif
1769         {
1770                 .procname       = "secure_tcp",
1771                 .maxlen         = sizeof(int),
1772                 .mode           = 0644,
1773                 .proc_handler   = proc_do_defense_mode,
1774         },
1775         {
1776                 .procname       = "snat_reroute",
1777                 .maxlen         = sizeof(int),
1778                 .mode           = 0644,
1779                 .proc_handler   = &proc_dointvec,
1780         },
1781         {
1782                 .procname       = "sync_version",
1783                 .maxlen         = sizeof(int),
1784                 .mode           = 0644,
1785                 .proc_handler   = proc_do_sync_mode,
1786         },
1787         {
1788                 .procname       = "sync_ports",
1789                 .maxlen         = sizeof(int),
1790                 .mode           = 0644,
1791                 .proc_handler   = proc_do_sync_ports,
1792         },
1793         {
1794                 .procname       = "sync_persist_mode",
1795                 .maxlen         = sizeof(int),
1796                 .mode           = 0644,
1797                 .proc_handler   = proc_dointvec,
1798         },
1799         {
1800                 .procname       = "sync_qlen_max",
1801                 .maxlen         = sizeof(unsigned long),
1802                 .mode           = 0644,
1803                 .proc_handler   = proc_doulongvec_minmax,
1804         },
1805         {
1806                 .procname       = "sync_sock_size",
1807                 .maxlen         = sizeof(int),
1808                 .mode           = 0644,
1809                 .proc_handler   = proc_dointvec,
1810         },
1811         {
1812                 .procname       = "cache_bypass",
1813                 .maxlen         = sizeof(int),
1814                 .mode           = 0644,
1815                 .proc_handler   = proc_dointvec,
1816         },
1817         {
1818                 .procname       = "expire_nodest_conn",
1819                 .maxlen         = sizeof(int),
1820                 .mode           = 0644,
1821                 .proc_handler   = proc_dointvec,
1822         },
1823         {
1824                 .procname       = "sloppy_tcp",
1825                 .maxlen         = sizeof(int),
1826                 .mode           = 0644,
1827                 .proc_handler   = proc_dointvec,
1828         },
1829         {
1830                 .procname       = "sloppy_sctp",
1831                 .maxlen         = sizeof(int),
1832                 .mode           = 0644,
1833                 .proc_handler   = proc_dointvec,
1834         },
1835         {
1836                 .procname       = "expire_quiescent_template",
1837                 .maxlen         = sizeof(int),
1838                 .mode           = 0644,
1839                 .proc_handler   = proc_dointvec,
1840         },
1841         {
1842                 .procname       = "sync_threshold",
1843                 .maxlen         =
1844                         sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1845                 .mode           = 0644,
1846                 .proc_handler   = proc_do_sync_threshold,
1847         },
1848         {
1849                 .procname       = "sync_refresh_period",
1850                 .maxlen         = sizeof(int),
1851                 .mode           = 0644,
1852                 .proc_handler   = proc_dointvec_jiffies,
1853         },
1854         {
1855                 .procname       = "sync_retries",
1856                 .maxlen         = sizeof(int),
1857                 .mode           = 0644,
1858                 .proc_handler   = proc_dointvec_minmax,
1859                 .extra1         = &zero,
1860                 .extra2         = &three,
1861         },
1862         {
1863                 .procname       = "nat_icmp_send",
1864                 .maxlen         = sizeof(int),
1865                 .mode           = 0644,
1866                 .proc_handler   = proc_dointvec,
1867         },
1868         {
1869                 .procname       = "pmtu_disc",
1870                 .maxlen         = sizeof(int),
1871                 .mode           = 0644,
1872                 .proc_handler   = proc_dointvec,
1873         },
1874         {
1875                 .procname       = "backup_only",
1876                 .maxlen         = sizeof(int),
1877                 .mode           = 0644,
1878                 .proc_handler   = proc_dointvec,
1879         },
1880         {
1881                 .procname       = "conn_reuse_mode",
1882                 .maxlen         = sizeof(int),
1883                 .mode           = 0644,
1884                 .proc_handler   = proc_dointvec,
1885         },
1886         {
1887                 .procname       = "schedule_icmp",
1888                 .maxlen         = sizeof(int),
1889                 .mode           = 0644,
1890                 .proc_handler   = proc_dointvec,
1891         },
1892         {
1893                 .procname       = "ignore_tunneled",
1894                 .maxlen         = sizeof(int),
1895                 .mode           = 0644,
1896                 .proc_handler   = proc_dointvec,
1897         },
1898 #ifdef CONFIG_IP_VS_DEBUG
1899         {
1900                 .procname       = "debug_level",
1901                 .data           = &sysctl_ip_vs_debug_level,
1902                 .maxlen         = sizeof(int),
1903                 .mode           = 0644,
1904                 .proc_handler   = proc_dointvec,
1905         },
1906 #endif
1907         { }
1908 };
1909
1910 #endif
1911
1912 #ifdef CONFIG_PROC_FS
1913
1914 struct ip_vs_iter {
1915         struct seq_net_private p;  /* Do not move this, netns depends upon it*/
1916         struct hlist_head *table;
1917         int bucket;
1918 };
1919
1920 /*
1921  *      Write the contents of the VS rule table to a PROCfs file.
1922  *      (It is kept just for backward compatibility)
1923  */
1924 static inline const char *ip_vs_fwd_name(unsigned int flags)
1925 {
1926         switch (flags & IP_VS_CONN_F_FWD_MASK) {
1927         case IP_VS_CONN_F_LOCALNODE:
1928                 return "Local";
1929         case IP_VS_CONN_F_TUNNEL:
1930                 return "Tunnel";
1931         case IP_VS_CONN_F_DROUTE:
1932                 return "Route";
1933         default:
1934                 return "Masq";
1935         }
1936 }
1937
1938
1939 /* Get the Nth entry in the two lists */
1940 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1941 {
1942         struct net *net = seq_file_net(seq);
1943         struct netns_ipvs *ipvs = net_ipvs(net);
1944         struct ip_vs_iter *iter = seq->private;
1945         int idx;
1946         struct ip_vs_service *svc;
1947
1948         /* look in hash by protocol */
1949         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1950                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1951                         if ((svc->ipvs == ipvs) && pos-- == 0) {
1952                                 iter->table = ip_vs_svc_table;
1953                                 iter->bucket = idx;
1954                                 return svc;
1955                         }
1956                 }
1957         }
1958
1959         /* keep looking in fwmark */
1960         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1961                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1962                                          f_list) {
1963                         if ((svc->ipvs == ipvs) && pos-- == 0) {
1964                                 iter->table = ip_vs_svc_fwm_table;
1965                                 iter->bucket = idx;
1966                                 return svc;
1967                         }
1968                 }
1969         }
1970
1971         return NULL;
1972 }
1973
1974 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1975         __acquires(RCU)
1976 {
1977         rcu_read_lock();
1978         return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1979 }
1980
1981
1982 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1983 {
1984         struct hlist_node *e;
1985         struct ip_vs_iter *iter;
1986         struct ip_vs_service *svc;
1987
1988         ++*pos;
1989         if (v == SEQ_START_TOKEN)
1990                 return ip_vs_info_array(seq,0);
1991
1992         svc = v;
1993         iter = seq->private;
1994
1995         if (iter->table == ip_vs_svc_table) {
1996                 /* next service in table hashed by protocol */
1997                 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
1998                 if (e)
1999                         return hlist_entry(e, struct ip_vs_service, s_list);
2000
2001                 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2002                         hlist_for_each_entry_rcu(svc,
2003                                                  &ip_vs_svc_table[iter->bucket],
2004                                                  s_list) {
2005                                 return svc;
2006                         }
2007                 }
2008
2009                 iter->table = ip_vs_svc_fwm_table;
2010                 iter->bucket = -1;
2011                 goto scan_fwmark;
2012         }
2013
2014         /* next service in hashed by fwmark */
2015         e = rcu_dereference(hlist_next_rcu(&svc->f_list));
2016         if (e)
2017                 return hlist_entry(e, struct ip_vs_service, f_list);
2018
2019  scan_fwmark:
2020         while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2021                 hlist_for_each_entry_rcu(svc,
2022                                          &ip_vs_svc_fwm_table[iter->bucket],
2023                                          f_list)
2024                         return svc;
2025         }
2026
2027         return NULL;
2028 }
2029
2030 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2031         __releases(RCU)
2032 {
2033         rcu_read_unlock();
2034 }
2035
2036
2037 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2038 {
2039         if (v == SEQ_START_TOKEN) {
2040                 seq_printf(seq,
2041                         "IP Virtual Server version %d.%d.%d (size=%d)\n",
2042                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2043                 seq_puts(seq,
2044                          "Prot LocalAddress:Port Scheduler Flags\n");
2045                 seq_puts(seq,
2046                          "  -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2047         } else {
2048                 struct net *net = seq_file_net(seq);
2049                 struct netns_ipvs *ipvs = net_ipvs(net);
2050                 const struct ip_vs_service *svc = v;
2051                 const struct ip_vs_iter *iter = seq->private;
2052                 const struct ip_vs_dest *dest;
2053                 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2054                 char *sched_name = sched ? sched->name : "none";
2055
2056                 if (svc->ipvs != ipvs)
2057                         return 0;
2058                 if (iter->table == ip_vs_svc_table) {
2059 #ifdef CONFIG_IP_VS_IPV6
2060                         if (svc->af == AF_INET6)
2061                                 seq_printf(seq, "%s  [%pI6]:%04X %s ",
2062                                            ip_vs_proto_name(svc->protocol),
2063                                            &svc->addr.in6,
2064                                            ntohs(svc->port),
2065                                            sched_name);
2066                         else
2067 #endif
2068                                 seq_printf(seq, "%s  %08X:%04X %s %s ",
2069                                            ip_vs_proto_name(svc->protocol),
2070                                            ntohl(svc->addr.ip),
2071                                            ntohs(svc->port),
2072                                            sched_name,
2073                                            (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2074                 } else {
2075                         seq_printf(seq, "FWM  %08X %s %s",
2076                                    svc->fwmark, sched_name,
2077                                    (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2078                 }
2079
2080                 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2081                         seq_printf(seq, "persistent %d %08X\n",
2082                                 svc->timeout,
2083                                 ntohl(svc->netmask));
2084                 else
2085                         seq_putc(seq, '\n');
2086
2087                 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2088 #ifdef CONFIG_IP_VS_IPV6
2089                         if (dest->af == AF_INET6)
2090                                 seq_printf(seq,
2091                                            "  -> [%pI6]:%04X"
2092                                            "      %-7s %-6d %-10d %-10d\n",
2093                                            &dest->addr.in6,
2094                                            ntohs(dest->port),
2095                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2096                                            atomic_read(&dest->weight),
2097                                            atomic_read(&dest->activeconns),
2098                                            atomic_read(&dest->inactconns));
2099                         else
2100 #endif
2101                                 seq_printf(seq,
2102                                            "  -> %08X:%04X      "
2103                                            "%-7s %-6d %-10d %-10d\n",
2104                                            ntohl(dest->addr.ip),
2105                                            ntohs(dest->port),
2106                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2107                                            atomic_read(&dest->weight),
2108                                            atomic_read(&dest->activeconns),
2109                                            atomic_read(&dest->inactconns));
2110
2111                 }
2112         }
2113         return 0;
2114 }
2115
2116 static const struct seq_operations ip_vs_info_seq_ops = {
2117         .start = ip_vs_info_seq_start,
2118         .next  = ip_vs_info_seq_next,
2119         .stop  = ip_vs_info_seq_stop,
2120         .show  = ip_vs_info_seq_show,
2121 };
2122
2123 static int ip_vs_info_open(struct inode *inode, struct file *file)
2124 {
2125         return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2126                         sizeof(struct ip_vs_iter));
2127 }
2128
2129 static const struct file_operations ip_vs_info_fops = {
2130         .owner   = THIS_MODULE,
2131         .open    = ip_vs_info_open,
2132         .read    = seq_read,
2133         .llseek  = seq_lseek,
2134         .release = seq_release_net,
2135 };
2136
2137 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2138 {
2139         struct net *net = seq_file_single_net(seq);
2140         struct ip_vs_kstats show;
2141
2142 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2143         seq_puts(seq,
2144                  "   Total Incoming Outgoing         Incoming         Outgoing\n");
2145         seq_puts(seq,
2146                  "   Conns  Packets  Packets            Bytes            Bytes\n");
2147
2148         ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2149         seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n",
2150                    (unsigned long long)show.conns,
2151                    (unsigned long long)show.inpkts,
2152                    (unsigned long long)show.outpkts,
2153                    (unsigned long long)show.inbytes,
2154                    (unsigned long long)show.outbytes);
2155
2156 /*                01234567 01234567 01234567 0123456701234567 0123456701234567*/
2157         seq_puts(seq,
2158                  " Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2159         seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n",
2160                    (unsigned long long)show.cps,
2161                    (unsigned long long)show.inpps,
2162                    (unsigned long long)show.outpps,
2163                    (unsigned long long)show.inbps,
2164                    (unsigned long long)show.outbps);
2165
2166         return 0;
2167 }
2168
2169 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2170 {
2171         return single_open_net(inode, file, ip_vs_stats_show);
2172 }
2173
2174 static const struct file_operations ip_vs_stats_fops = {
2175         .owner = THIS_MODULE,
2176         .open = ip_vs_stats_seq_open,
2177         .read = seq_read,
2178         .llseek = seq_lseek,
2179         .release = single_release_net,
2180 };
2181
2182 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2183 {
2184         struct net *net = seq_file_single_net(seq);
2185         struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2186         struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2187         struct ip_vs_kstats kstats;
2188         int i;
2189
2190 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2191         seq_puts(seq,
2192                  "       Total Incoming Outgoing         Incoming         Outgoing\n");
2193         seq_puts(seq,
2194                  "CPU    Conns  Packets  Packets            Bytes            Bytes\n");
2195
2196         for_each_possible_cpu(i) {
2197                 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2198                 unsigned int start;
2199                 u64 conns, inpkts, outpkts, inbytes, outbytes;
2200
2201                 do {
2202                         start = u64_stats_fetch_begin_irq(&u->syncp);
2203                         conns = u->cnt.conns;
2204                         inpkts = u->cnt.inpkts;
2205                         outpkts = u->cnt.outpkts;
2206                         inbytes = u->cnt.inbytes;
2207                         outbytes = u->cnt.outbytes;
2208                 } while (u64_stats_fetch_retry_irq(&u->syncp, start));
2209
2210                 seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2211                            i, (u64)conns, (u64)inpkts,
2212                            (u64)outpkts, (u64)inbytes,
2213                            (u64)outbytes);
2214         }
2215
2216         ip_vs_copy_stats(&kstats, tot_stats);
2217
2218         seq_printf(seq, "  ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2219                    (unsigned long long)kstats.conns,
2220                    (unsigned long long)kstats.inpkts,
2221                    (unsigned long long)kstats.outpkts,
2222                    (unsigned long long)kstats.inbytes,
2223                    (unsigned long long)kstats.outbytes);
2224
2225 /*                ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2226         seq_puts(seq,
2227                  "     Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2228         seq_printf(seq, "    %8LX %8LX %8LX %16LX %16LX\n",
2229                    kstats.cps,
2230                    kstats.inpps,
2231                    kstats.outpps,
2232                    kstats.inbps,
2233                    kstats.outbps);
2234
2235         return 0;
2236 }
2237
2238 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2239 {
2240         return single_open_net(inode, file, ip_vs_stats_percpu_show);
2241 }
2242
2243 static const struct file_operations ip_vs_stats_percpu_fops = {
2244         .owner = THIS_MODULE,
2245         .open = ip_vs_stats_percpu_seq_open,
2246         .read = seq_read,
2247         .llseek = seq_lseek,
2248         .release = single_release_net,
2249 };
2250 #endif
2251
2252 /*
2253  *      Set timeout values for tcp tcpfin udp in the timeout_table.
2254  */
2255 static int ip_vs_set_timeout(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2256 {
2257 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2258         struct ip_vs_proto_data *pd;
2259 #endif
2260
2261         IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2262                   u->tcp_timeout,
2263                   u->tcp_fin_timeout,
2264                   u->udp_timeout);
2265
2266 #ifdef CONFIG_IP_VS_PROTO_TCP
2267         if (u->tcp_timeout < 0 || u->tcp_timeout > (INT_MAX / HZ) ||
2268             u->tcp_fin_timeout < 0 || u->tcp_fin_timeout > (INT_MAX / HZ)) {
2269                 return -EINVAL;
2270         }
2271 #endif
2272
2273 #ifdef CONFIG_IP_VS_PROTO_UDP
2274         if (u->udp_timeout < 0 || u->udp_timeout > (INT_MAX / HZ))
2275                 return -EINVAL;
2276 #endif
2277
2278 #ifdef CONFIG_IP_VS_PROTO_TCP
2279         if (u->tcp_timeout) {
2280                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2281                 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2282                         = u->tcp_timeout * HZ;
2283         }
2284
2285         if (u->tcp_fin_timeout) {
2286                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2287                 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2288                         = u->tcp_fin_timeout * HZ;
2289         }
2290 #endif
2291
2292 #ifdef CONFIG_IP_VS_PROTO_UDP
2293         if (u->udp_timeout) {
2294                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2295                 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2296                         = u->udp_timeout * HZ;
2297         }
2298 #endif
2299         return 0;
2300 }
2301
2302 #define CMDID(cmd)              (cmd - IP_VS_BASE_CTL)
2303
2304 struct ip_vs_svcdest_user {
2305         struct ip_vs_service_user       s;
2306         struct ip_vs_dest_user          d;
2307 };
2308
2309 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2310         [CMDID(IP_VS_SO_SET_ADD)]         = sizeof(struct ip_vs_service_user),
2311         [CMDID(IP_VS_SO_SET_EDIT)]        = sizeof(struct ip_vs_service_user),
2312         [CMDID(IP_VS_SO_SET_DEL)]         = sizeof(struct ip_vs_service_user),
2313         [CMDID(IP_VS_SO_SET_ADDDEST)]     = sizeof(struct ip_vs_svcdest_user),
2314         [CMDID(IP_VS_SO_SET_DELDEST)]     = sizeof(struct ip_vs_svcdest_user),
2315         [CMDID(IP_VS_SO_SET_EDITDEST)]    = sizeof(struct ip_vs_svcdest_user),
2316         [CMDID(IP_VS_SO_SET_TIMEOUT)]     = sizeof(struct ip_vs_timeout_user),
2317         [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2318         [CMDID(IP_VS_SO_SET_STOPDAEMON)]  = sizeof(struct ip_vs_daemon_user),
2319         [CMDID(IP_VS_SO_SET_ZERO)]        = sizeof(struct ip_vs_service_user),
2320 };
2321
2322 union ip_vs_set_arglen {
2323         struct ip_vs_service_user       field_IP_VS_SO_SET_ADD;
2324         struct ip_vs_service_user       field_IP_VS_SO_SET_EDIT;
2325         struct ip_vs_service_user       field_IP_VS_SO_SET_DEL;
2326         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_ADDDEST;
2327         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_DELDEST;
2328         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_EDITDEST;
2329         struct ip_vs_timeout_user       field_IP_VS_SO_SET_TIMEOUT;
2330         struct ip_vs_daemon_user        field_IP_VS_SO_SET_STARTDAEMON;
2331         struct ip_vs_daemon_user        field_IP_VS_SO_SET_STOPDAEMON;
2332         struct ip_vs_service_user       field_IP_VS_SO_SET_ZERO;
2333 };
2334
2335 #define MAX_SET_ARGLEN  sizeof(union ip_vs_set_arglen)
2336
2337 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2338                                   struct ip_vs_service_user *usvc_compat)
2339 {
2340         memset(usvc, 0, sizeof(*usvc));
2341
2342         usvc->af                = AF_INET;
2343         usvc->protocol          = usvc_compat->protocol;
2344         usvc->addr.ip           = usvc_compat->addr;
2345         usvc->port              = usvc_compat->port;
2346         usvc->fwmark            = usvc_compat->fwmark;
2347
2348         /* Deep copy of sched_name is not needed here */
2349         usvc->sched_name        = usvc_compat->sched_name;
2350
2351         usvc->flags             = usvc_compat->flags;
2352         usvc->timeout           = usvc_compat->timeout;
2353         usvc->netmask           = usvc_compat->netmask;
2354 }
2355
2356 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2357                                    struct ip_vs_dest_user *udest_compat)
2358 {
2359         memset(udest, 0, sizeof(*udest));
2360
2361         udest->addr.ip          = udest_compat->addr;
2362         udest->port             = udest_compat->port;
2363         udest->conn_flags       = udest_compat->conn_flags;
2364         udest->weight           = udest_compat->weight;
2365         udest->u_threshold      = udest_compat->u_threshold;
2366         udest->l_threshold      = udest_compat->l_threshold;
2367         udest->af               = AF_INET;
2368 }
2369
2370 static int
2371 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2372 {
2373         struct net *net = sock_net(sk);
2374         int ret;
2375         unsigned char arg[MAX_SET_ARGLEN];
2376         struct ip_vs_service_user *usvc_compat;
2377         struct ip_vs_service_user_kern usvc;
2378         struct ip_vs_service *svc;
2379         struct ip_vs_dest_user *udest_compat;
2380         struct ip_vs_dest_user_kern udest;
2381         struct netns_ipvs *ipvs = net_ipvs(net);
2382
2383         BUILD_BUG_ON(sizeof(arg) > 255);
2384         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2385                 return -EPERM;
2386
2387         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2388                 return -EINVAL;
2389         if (len != set_arglen[CMDID(cmd)]) {
2390                 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2391                           len, set_arglen[CMDID(cmd)]);
2392                 return -EINVAL;
2393         }
2394
2395         if (copy_from_user(arg, user, len) != 0)
2396                 return -EFAULT;
2397
2398         /* Handle daemons since they have another lock */
2399         if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2400             cmd == IP_VS_SO_SET_STOPDAEMON) {
2401                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2402
2403                 if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2404                         struct ipvs_sync_daemon_cfg cfg;
2405
2406                         memset(&cfg, 0, sizeof(cfg));
2407                         ret = -EINVAL;
2408                         if (strscpy(cfg.mcast_ifn, dm->mcast_ifn,
2409                                     sizeof(cfg.mcast_ifn)) <= 0)
2410                                 return ret;
2411                         cfg.syncid = dm->syncid;
2412                         ret = start_sync_thread(ipvs, &cfg, dm->state);
2413                 } else {
2414                         ret = stop_sync_thread(ipvs, dm->state);
2415                 }
2416                 return ret;
2417         }
2418
2419         mutex_lock(&__ip_vs_mutex);
2420         if (cmd == IP_VS_SO_SET_FLUSH) {
2421                 /* Flush the virtual service */
2422                 ret = ip_vs_flush(ipvs, false);
2423                 goto out_unlock;
2424         } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2425                 /* Set timeout values for (tcp tcpfin udp) */
2426                 ret = ip_vs_set_timeout(ipvs, (struct ip_vs_timeout_user *)arg);
2427                 goto out_unlock;
2428         } else if (!len) {
2429                 /* No more commands with len == 0 below */
2430                 ret = -EINVAL;
2431                 goto out_unlock;
2432         }
2433
2434         usvc_compat = (struct ip_vs_service_user *)arg;
2435         udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2436
2437         /* We only use the new structs internally, so copy userspace compat
2438          * structs to extended internal versions */
2439         ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2440         ip_vs_copy_udest_compat(&udest, udest_compat);
2441
2442         if (cmd == IP_VS_SO_SET_ZERO) {
2443                 /* if no service address is set, zero counters in all */
2444                 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2445                         ret = ip_vs_zero_all(ipvs);
2446                         goto out_unlock;
2447                 }
2448         }
2449
2450         if ((cmd == IP_VS_SO_SET_ADD || cmd == IP_VS_SO_SET_EDIT) &&
2451             strnlen(usvc.sched_name, IP_VS_SCHEDNAME_MAXLEN) ==
2452             IP_VS_SCHEDNAME_MAXLEN) {
2453                 ret = -EINVAL;
2454                 goto out_unlock;
2455         }
2456
2457         /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2458         if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2459             usvc.protocol != IPPROTO_SCTP) {
2460                 pr_err("set_ctl: invalid protocol: %d %pI4:%d\n",
2461                        usvc.protocol, &usvc.addr.ip,
2462                        ntohs(usvc.port));
2463                 ret = -EFAULT;
2464                 goto out_unlock;
2465         }
2466
2467         /* Lookup the exact service by <protocol, addr, port> or fwmark */
2468         rcu_read_lock();
2469         if (usvc.fwmark == 0)
2470                 svc = __ip_vs_service_find(ipvs, usvc.af, usvc.protocol,
2471                                            &usvc.addr, usvc.port);
2472         else
2473                 svc = __ip_vs_svc_fwm_find(ipvs, usvc.af, usvc.fwmark);
2474         rcu_read_unlock();
2475
2476         if (cmd != IP_VS_SO_SET_ADD
2477             && (svc == NULL || svc->protocol != usvc.protocol)) {
2478                 ret = -ESRCH;
2479                 goto out_unlock;
2480         }
2481
2482         switch (cmd) {
2483         case IP_VS_SO_SET_ADD:
2484                 if (svc != NULL)
2485                         ret = -EEXIST;
2486                 else
2487                         ret = ip_vs_add_service(ipvs, &usvc, &svc);
2488                 break;
2489         case IP_VS_SO_SET_EDIT:
2490                 ret = ip_vs_edit_service(svc, &usvc);
2491                 break;
2492         case IP_VS_SO_SET_DEL:
2493                 ret = ip_vs_del_service(svc);
2494                 if (!ret)
2495                         goto out_unlock;
2496                 break;
2497         case IP_VS_SO_SET_ZERO:
2498                 ret = ip_vs_zero_service(svc);
2499                 break;
2500         case IP_VS_SO_SET_ADDDEST:
2501                 ret = ip_vs_add_dest(svc, &udest);
2502                 break;
2503         case IP_VS_SO_SET_EDITDEST:
2504                 ret = ip_vs_edit_dest(svc, &udest);
2505                 break;
2506         case IP_VS_SO_SET_DELDEST:
2507                 ret = ip_vs_del_dest(svc, &udest);
2508         }
2509
2510   out_unlock:
2511         mutex_unlock(&__ip_vs_mutex);
2512         return ret;
2513 }
2514
2515
2516 static void
2517 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2518 {
2519         struct ip_vs_scheduler *sched;
2520         struct ip_vs_kstats kstats;
2521         char *sched_name;
2522
2523         sched = rcu_dereference_protected(src->scheduler, 1);
2524         sched_name = sched ? sched->name : "none";
2525         dst->protocol = src->protocol;
2526         dst->addr = src->addr.ip;
2527         dst->port = src->port;
2528         dst->fwmark = src->fwmark;
2529         strlcpy(dst->sched_name, sched_name, sizeof(dst->sched_name));
2530         dst->flags = src->flags;
2531         dst->timeout = src->timeout / HZ;
2532         dst->netmask = src->netmask;
2533         dst->num_dests = src->num_dests;
2534         ip_vs_copy_stats(&kstats, &src->stats);
2535         ip_vs_export_stats_user(&dst->stats, &kstats);
2536 }
2537
2538 static inline int
2539 __ip_vs_get_service_entries(struct netns_ipvs *ipvs,
2540                             const struct ip_vs_get_services *get,
2541                             struct ip_vs_get_services __user *uptr)
2542 {
2543         int idx, count=0;
2544         struct ip_vs_service *svc;
2545         struct ip_vs_service_entry entry;
2546         int ret = 0;
2547
2548         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2549                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2550                         /* Only expose IPv4 entries to old interface */
2551                         if (svc->af != AF_INET || (svc->ipvs != ipvs))
2552                                 continue;
2553
2554                         if (count >= get->num_services)
2555                                 goto out;
2556                         memset(&entry, 0, sizeof(entry));
2557                         ip_vs_copy_service(&entry, svc);
2558                         if (copy_to_user(&uptr->entrytable[count],
2559                                          &entry, sizeof(entry))) {
2560                                 ret = -EFAULT;
2561                                 goto out;
2562                         }
2563                         count++;
2564                 }
2565         }
2566
2567         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2568                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2569                         /* Only expose IPv4 entries to old interface */
2570                         if (svc->af != AF_INET || (svc->ipvs != ipvs))
2571                                 continue;
2572
2573                         if (count >= get->num_services)
2574                                 goto out;
2575                         memset(&entry, 0, sizeof(entry));
2576                         ip_vs_copy_service(&entry, svc);
2577                         if (copy_to_user(&uptr->entrytable[count],
2578                                          &entry, sizeof(entry))) {
2579                                 ret = -EFAULT;
2580                                 goto out;
2581                         }
2582                         count++;
2583                 }
2584         }
2585 out:
2586         return ret;
2587 }
2588
2589 static inline int
2590 __ip_vs_get_dest_entries(struct netns_ipvs *ipvs, const struct ip_vs_get_dests *get,
2591                          struct ip_vs_get_dests __user *uptr)
2592 {
2593         struct ip_vs_service *svc;
2594         union nf_inet_addr addr = { .ip = get->addr };
2595         int ret = 0;
2596
2597         rcu_read_lock();
2598         if (get->fwmark)
2599                 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, get->fwmark);
2600         else
2601                 svc = __ip_vs_service_find(ipvs, AF_INET, get->protocol, &addr,
2602                                            get->port);
2603         rcu_read_unlock();
2604
2605         if (svc) {
2606                 int count = 0;
2607                 struct ip_vs_dest *dest;
2608                 struct ip_vs_dest_entry entry;
2609                 struct ip_vs_kstats kstats;
2610
2611                 memset(&entry, 0, sizeof(entry));
2612                 list_for_each_entry(dest, &svc->destinations, n_list) {
2613                         if (count >= get->num_dests)
2614                                 break;
2615
2616                         /* Cannot expose heterogeneous members via sockopt
2617                          * interface
2618                          */
2619                         if (dest->af != svc->af)
2620                                 continue;
2621
2622                         entry.addr = dest->addr.ip;
2623                         entry.port = dest->port;
2624                         entry.conn_flags = atomic_read(&dest->conn_flags);
2625                         entry.weight = atomic_read(&dest->weight);
2626                         entry.u_threshold = dest->u_threshold;
2627                         entry.l_threshold = dest->l_threshold;
2628                         entry.activeconns = atomic_read(&dest->activeconns);
2629                         entry.inactconns = atomic_read(&dest->inactconns);
2630                         entry.persistconns = atomic_read(&dest->persistconns);
2631                         ip_vs_copy_stats(&kstats, &dest->stats);
2632                         ip_vs_export_stats_user(&entry.stats, &kstats);
2633                         if (copy_to_user(&uptr->entrytable[count],
2634                                          &entry, sizeof(entry))) {
2635                                 ret = -EFAULT;
2636                                 break;
2637                         }
2638                         count++;
2639                 }
2640         } else
2641                 ret = -ESRCH;
2642         return ret;
2643 }
2644
2645 static inline void
2646 __ip_vs_get_timeouts(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2647 {
2648 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2649         struct ip_vs_proto_data *pd;
2650 #endif
2651
2652         memset(u, 0, sizeof (*u));
2653
2654 #ifdef CONFIG_IP_VS_PROTO_TCP
2655         pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2656         u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2657         u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2658 #endif
2659 #ifdef CONFIG_IP_VS_PROTO_UDP
2660         pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2661         u->udp_timeout =
2662                         pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2663 #endif
2664 }
2665
2666 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2667         [CMDID(IP_VS_SO_GET_VERSION)]  = 64,
2668         [CMDID(IP_VS_SO_GET_INFO)]     = sizeof(struct ip_vs_getinfo),
2669         [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2670         [CMDID(IP_VS_SO_GET_SERVICE)]  = sizeof(struct ip_vs_service_entry),
2671         [CMDID(IP_VS_SO_GET_DESTS)]    = sizeof(struct ip_vs_get_dests),
2672         [CMDID(IP_VS_SO_GET_TIMEOUT)]  = sizeof(struct ip_vs_timeout_user),
2673         [CMDID(IP_VS_SO_GET_DAEMON)]   = 2 * sizeof(struct ip_vs_daemon_user),
2674 };
2675
2676 union ip_vs_get_arglen {
2677         char                            field_IP_VS_SO_GET_VERSION[64];
2678         struct ip_vs_getinfo            field_IP_VS_SO_GET_INFO;
2679         struct ip_vs_get_services       field_IP_VS_SO_GET_SERVICES;
2680         struct ip_vs_service_entry      field_IP_VS_SO_GET_SERVICE;
2681         struct ip_vs_get_dests          field_IP_VS_SO_GET_DESTS;
2682         struct ip_vs_timeout_user       field_IP_VS_SO_GET_TIMEOUT;
2683         struct ip_vs_daemon_user        field_IP_VS_SO_GET_DAEMON[2];
2684 };
2685
2686 #define MAX_GET_ARGLEN  sizeof(union ip_vs_get_arglen)
2687
2688 static int
2689 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2690 {
2691         unsigned char arg[MAX_GET_ARGLEN];
2692         int ret = 0;
2693         unsigned int copylen;
2694         struct net *net = sock_net(sk);
2695         struct netns_ipvs *ipvs = net_ipvs(net);
2696
2697         BUG_ON(!net);
2698         BUILD_BUG_ON(sizeof(arg) > 255);
2699         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2700                 return -EPERM;
2701
2702         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2703                 return -EINVAL;
2704
2705         copylen = get_arglen[CMDID(cmd)];
2706         if (*len < (int) copylen) {
2707                 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
2708                 return -EINVAL;
2709         }
2710
2711         if (copy_from_user(arg, user, copylen) != 0)
2712                 return -EFAULT;
2713         /*
2714          * Handle daemons first since it has its own locking
2715          */
2716         if (cmd == IP_VS_SO_GET_DAEMON) {
2717                 struct ip_vs_daemon_user d[2];
2718
2719                 memset(&d, 0, sizeof(d));
2720                 mutex_lock(&ipvs->sync_mutex);
2721                 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2722                         d[0].state = IP_VS_STATE_MASTER;
2723                         strlcpy(d[0].mcast_ifn, ipvs->mcfg.mcast_ifn,
2724                                 sizeof(d[0].mcast_ifn));
2725                         d[0].syncid = ipvs->mcfg.syncid;
2726                 }
2727                 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2728                         d[1].state = IP_VS_STATE_BACKUP;
2729                         strlcpy(d[1].mcast_ifn, ipvs->bcfg.mcast_ifn,
2730                                 sizeof(d[1].mcast_ifn));
2731                         d[1].syncid = ipvs->bcfg.syncid;
2732                 }
2733                 if (copy_to_user(user, &d, sizeof(d)) != 0)
2734                         ret = -EFAULT;
2735                 mutex_unlock(&ipvs->sync_mutex);
2736                 return ret;
2737         }
2738
2739         mutex_lock(&__ip_vs_mutex);
2740         switch (cmd) {
2741         case IP_VS_SO_GET_VERSION:
2742         {
2743                 char buf[64];
2744
2745                 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2746                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2747                 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2748                         ret = -EFAULT;
2749                         goto out;
2750                 }
2751                 *len = strlen(buf)+1;
2752         }
2753         break;
2754
2755         case IP_VS_SO_GET_INFO:
2756         {
2757                 struct ip_vs_getinfo info;
2758                 info.version = IP_VS_VERSION_CODE;
2759                 info.size = ip_vs_conn_tab_size;
2760                 info.num_services = ipvs->num_services;
2761                 if (copy_to_user(user, &info, sizeof(info)) != 0)
2762                         ret = -EFAULT;
2763         }
2764         break;
2765
2766         case IP_VS_SO_GET_SERVICES:
2767         {
2768                 struct ip_vs_get_services *get;
2769                 int size;
2770
2771                 get = (struct ip_vs_get_services *)arg;
2772                 size = sizeof(*get) +
2773                         sizeof(struct ip_vs_service_entry) * get->num_services;
2774                 if (*len != size) {
2775                         pr_err("length: %u != %u\n", *len, size);
2776                         ret = -EINVAL;
2777                         goto out;
2778                 }
2779                 ret = __ip_vs_get_service_entries(ipvs, get, user);
2780         }
2781         break;
2782
2783         case IP_VS_SO_GET_SERVICE:
2784         {
2785                 struct ip_vs_service_entry *entry;
2786                 struct ip_vs_service *svc;
2787                 union nf_inet_addr addr;
2788
2789                 entry = (struct ip_vs_service_entry *)arg;
2790                 addr.ip = entry->addr;
2791                 rcu_read_lock();
2792                 if (entry->fwmark)
2793                         svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, entry->fwmark);
2794                 else
2795                         svc = __ip_vs_service_find(ipvs, AF_INET,
2796                                                    entry->protocol, &addr,
2797                                                    entry->port);
2798                 rcu_read_unlock();
2799                 if (svc) {
2800                         ip_vs_copy_service(entry, svc);
2801                         if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2802                                 ret = -EFAULT;
2803                 } else
2804                         ret = -ESRCH;
2805         }
2806         break;
2807
2808         case IP_VS_SO_GET_DESTS:
2809         {
2810                 struct ip_vs_get_dests *get;
2811                 int size;
2812
2813                 get = (struct ip_vs_get_dests *)arg;
2814                 size = sizeof(*get) +
2815                         sizeof(struct ip_vs_dest_entry) * get->num_dests;
2816                 if (*len != size) {
2817                         pr_err("length: %u != %u\n", *len, size);
2818                         ret = -EINVAL;
2819                         goto out;
2820                 }
2821                 ret = __ip_vs_get_dest_entries(ipvs, get, user);
2822         }
2823         break;
2824
2825         case IP_VS_SO_GET_TIMEOUT:
2826         {
2827                 struct ip_vs_timeout_user t;
2828
2829                 __ip_vs_get_timeouts(ipvs, &t);
2830                 if (copy_to_user(user, &t, sizeof(t)) != 0)
2831                         ret = -EFAULT;
2832         }
2833         break;
2834
2835         default:
2836                 ret = -EINVAL;
2837         }
2838
2839 out:
2840         mutex_unlock(&__ip_vs_mutex);
2841         return ret;
2842 }
2843
2844
2845 static struct nf_sockopt_ops ip_vs_sockopts = {
2846         .pf             = PF_INET,
2847         .set_optmin     = IP_VS_BASE_CTL,
2848         .set_optmax     = IP_VS_SO_SET_MAX+1,
2849         .set            = do_ip_vs_set_ctl,
2850         .get_optmin     = IP_VS_BASE_CTL,
2851         .get_optmax     = IP_VS_SO_GET_MAX+1,
2852         .get            = do_ip_vs_get_ctl,
2853         .owner          = THIS_MODULE,
2854 };
2855
2856 /*
2857  * Generic Netlink interface
2858  */
2859
2860 /* IPVS genetlink family */
2861 static struct genl_family ip_vs_genl_family;
2862
2863 /* Policy used for first-level command attributes */
2864 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2865         [IPVS_CMD_ATTR_SERVICE]         = { .type = NLA_NESTED },
2866         [IPVS_CMD_ATTR_DEST]            = { .type = NLA_NESTED },
2867         [IPVS_CMD_ATTR_DAEMON]          = { .type = NLA_NESTED },
2868         [IPVS_CMD_ATTR_TIMEOUT_TCP]     = { .type = NLA_U32 },
2869         [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2870         [IPVS_CMD_ATTR_TIMEOUT_UDP]     = { .type = NLA_U32 },
2871 };
2872
2873 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2874 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2875         [IPVS_DAEMON_ATTR_STATE]        = { .type = NLA_U32 },
2876         [IPVS_DAEMON_ATTR_MCAST_IFN]    = { .type = NLA_NUL_STRING,
2877                                             .len = IP_VS_IFNAME_MAXLEN - 1 },
2878         [IPVS_DAEMON_ATTR_SYNC_ID]      = { .type = NLA_U32 },
2879         [IPVS_DAEMON_ATTR_SYNC_MAXLEN]  = { .type = NLA_U16 },
2880         [IPVS_DAEMON_ATTR_MCAST_GROUP]  = { .type = NLA_U32 },
2881         [IPVS_DAEMON_ATTR_MCAST_GROUP6] = { .len = sizeof(struct in6_addr) },
2882         [IPVS_DAEMON_ATTR_MCAST_PORT]   = { .type = NLA_U16 },
2883         [IPVS_DAEMON_ATTR_MCAST_TTL]    = { .type = NLA_U8 },
2884 };
2885
2886 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2887 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2888         [IPVS_SVC_ATTR_AF]              = { .type = NLA_U16 },
2889         [IPVS_SVC_ATTR_PROTOCOL]        = { .type = NLA_U16 },
2890         [IPVS_SVC_ATTR_ADDR]            = { .type = NLA_BINARY,
2891                                             .len = sizeof(union nf_inet_addr) },
2892         [IPVS_SVC_ATTR_PORT]            = { .type = NLA_U16 },
2893         [IPVS_SVC_ATTR_FWMARK]          = { .type = NLA_U32 },
2894         [IPVS_SVC_ATTR_SCHED_NAME]      = { .type = NLA_NUL_STRING,
2895                                             .len = IP_VS_SCHEDNAME_MAXLEN - 1 },
2896         [IPVS_SVC_ATTR_PE_NAME]         = { .type = NLA_NUL_STRING,
2897                                             .len = IP_VS_PENAME_MAXLEN },
2898         [IPVS_SVC_ATTR_FLAGS]           = { .type = NLA_BINARY,
2899                                             .len = sizeof(struct ip_vs_flags) },
2900         [IPVS_SVC_ATTR_TIMEOUT]         = { .type = NLA_U32 },
2901         [IPVS_SVC_ATTR_NETMASK]         = { .type = NLA_U32 },
2902         [IPVS_SVC_ATTR_STATS]           = { .type = NLA_NESTED },
2903 };
2904
2905 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2906 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2907         [IPVS_DEST_ATTR_ADDR]           = { .type = NLA_BINARY,
2908                                             .len = sizeof(union nf_inet_addr) },
2909         [IPVS_DEST_ATTR_PORT]           = { .type = NLA_U16 },
2910         [IPVS_DEST_ATTR_FWD_METHOD]     = { .type = NLA_U32 },
2911         [IPVS_DEST_ATTR_WEIGHT]         = { .type = NLA_U32 },
2912         [IPVS_DEST_ATTR_U_THRESH]       = { .type = NLA_U32 },
2913         [IPVS_DEST_ATTR_L_THRESH]       = { .type = NLA_U32 },
2914         [IPVS_DEST_ATTR_ACTIVE_CONNS]   = { .type = NLA_U32 },
2915         [IPVS_DEST_ATTR_INACT_CONNS]    = { .type = NLA_U32 },
2916         [IPVS_DEST_ATTR_PERSIST_CONNS]  = { .type = NLA_U32 },
2917         [IPVS_DEST_ATTR_STATS]          = { .type = NLA_NESTED },
2918         [IPVS_DEST_ATTR_ADDR_FAMILY]    = { .type = NLA_U16 },
2919 };
2920
2921 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2922                                  struct ip_vs_kstats *kstats)
2923 {
2924         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2925
2926         if (!nl_stats)
2927                 return -EMSGSIZE;
2928
2929         if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) ||
2930             nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) ||
2931             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) ||
2932             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
2933                               IPVS_STATS_ATTR_PAD) ||
2934             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
2935                               IPVS_STATS_ATTR_PAD) ||
2936             nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) ||
2937             nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) ||
2938             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) ||
2939             nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) ||
2940             nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps))
2941                 goto nla_put_failure;
2942         nla_nest_end(skb, nl_stats);
2943
2944         return 0;
2945
2946 nla_put_failure:
2947         nla_nest_cancel(skb, nl_stats);
2948         return -EMSGSIZE;
2949 }
2950
2951 static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type,
2952                                    struct ip_vs_kstats *kstats)
2953 {
2954         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2955
2956         if (!nl_stats)
2957                 return -EMSGSIZE;
2958
2959         if (nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CONNS, kstats->conns,
2960                               IPVS_STATS_ATTR_PAD) ||
2961             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts,
2962                               IPVS_STATS_ATTR_PAD) ||
2963             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts,
2964                               IPVS_STATS_ATTR_PAD) ||
2965             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
2966                               IPVS_STATS_ATTR_PAD) ||
2967             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
2968                               IPVS_STATS_ATTR_PAD) ||
2969             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CPS, kstats->cps,
2970                               IPVS_STATS_ATTR_PAD) ||
2971             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps,
2972                               IPVS_STATS_ATTR_PAD) ||
2973             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps,
2974                               IPVS_STATS_ATTR_PAD) ||
2975             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps,
2976                               IPVS_STATS_ATTR_PAD) ||
2977             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps,
2978                               IPVS_STATS_ATTR_PAD))
2979                 goto nla_put_failure;
2980         nla_nest_end(skb, nl_stats);
2981
2982         return 0;
2983
2984 nla_put_failure:
2985         nla_nest_cancel(skb, nl_stats);
2986         return -EMSGSIZE;
2987 }
2988
2989 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2990                                    struct ip_vs_service *svc)
2991 {
2992         struct ip_vs_scheduler *sched;
2993         struct ip_vs_pe *pe;
2994         struct nlattr *nl_service;
2995         struct ip_vs_flags flags = { .flags = svc->flags,
2996                                      .mask = ~0 };
2997         struct ip_vs_kstats kstats;
2998         char *sched_name;
2999
3000         nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
3001         if (!nl_service)
3002                 return -EMSGSIZE;
3003
3004         if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
3005                 goto nla_put_failure;
3006         if (svc->fwmark) {
3007                 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
3008                         goto nla_put_failure;
3009         } else {
3010                 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
3011                     nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
3012                     nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
3013                         goto nla_put_failure;
3014         }
3015
3016         sched = rcu_dereference_protected(svc->scheduler, 1);
3017         sched_name = sched ? sched->name : "none";
3018         pe = rcu_dereference_protected(svc->pe, 1);
3019         if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) ||
3020             (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
3021             nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
3022             nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
3023             nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
3024                 goto nla_put_failure;
3025         ip_vs_copy_stats(&kstats, &svc->stats);
3026         if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats))
3027                 goto nla_put_failure;
3028         if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats))
3029                 goto nla_put_failure;
3030
3031         nla_nest_end(skb, nl_service);
3032
3033         return 0;
3034
3035 nla_put_failure:
3036         nla_nest_cancel(skb, nl_service);
3037         return -EMSGSIZE;
3038 }
3039
3040 static int ip_vs_genl_dump_service(struct sk_buff *skb,
3041                                    struct ip_vs_service *svc,
3042                                    struct netlink_callback *cb)
3043 {
3044         void *hdr;
3045
3046         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3047                           &ip_vs_genl_family, NLM_F_MULTI,
3048                           IPVS_CMD_NEW_SERVICE);
3049         if (!hdr)
3050                 return -EMSGSIZE;
3051
3052         if (ip_vs_genl_fill_service(skb, svc) < 0)
3053                 goto nla_put_failure;
3054
3055         genlmsg_end(skb, hdr);
3056         return 0;
3057
3058 nla_put_failure:
3059         genlmsg_cancel(skb, hdr);
3060         return -EMSGSIZE;
3061 }
3062
3063 static int ip_vs_genl_dump_services(struct sk_buff *skb,
3064                                     struct netlink_callback *cb)
3065 {
3066         int idx = 0, i;
3067         int start = cb->args[0];
3068         struct ip_vs_service *svc;
3069         struct net *net = sock_net(skb->sk);
3070         struct netns_ipvs *ipvs = net_ipvs(net);
3071
3072         mutex_lock(&__ip_vs_mutex);
3073         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3074                 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
3075                         if (++idx <= start || (svc->ipvs != ipvs))
3076                                 continue;
3077                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3078                                 idx--;
3079                                 goto nla_put_failure;
3080                         }
3081                 }
3082         }
3083
3084         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3085                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3086                         if (++idx <= start || (svc->ipvs != ipvs))
3087                                 continue;
3088                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3089                                 idx--;
3090                                 goto nla_put_failure;
3091                         }
3092                 }
3093         }
3094
3095 nla_put_failure:
3096         mutex_unlock(&__ip_vs_mutex);
3097         cb->args[0] = idx;
3098
3099         return skb->len;
3100 }
3101
3102 static bool ip_vs_is_af_valid(int af)
3103 {
3104         if (af == AF_INET)
3105                 return true;
3106 #ifdef CONFIG_IP_VS_IPV6
3107         if (af == AF_INET6 && ipv6_mod_enabled())
3108                 return true;
3109 #endif
3110         return false;
3111 }
3112
3113 static int ip_vs_genl_parse_service(struct netns_ipvs *ipvs,
3114                                     struct ip_vs_service_user_kern *usvc,
3115                                     struct nlattr *nla, int full_entry,
3116                                     struct ip_vs_service **ret_svc)
3117 {
3118         struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3119         struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3120         struct ip_vs_service *svc;
3121
3122         /* Parse mandatory identifying service fields first */
3123         if (nla == NULL ||
3124             nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla,
3125                              ip_vs_svc_policy, NULL))
3126                 return -EINVAL;
3127
3128         nla_af          = attrs[IPVS_SVC_ATTR_AF];
3129         nla_protocol    = attrs[IPVS_SVC_ATTR_PROTOCOL];
3130         nla_addr        = attrs[IPVS_SVC_ATTR_ADDR];
3131         nla_port        = attrs[IPVS_SVC_ATTR_PORT];
3132         nla_fwmark      = attrs[IPVS_SVC_ATTR_FWMARK];
3133
3134         if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3135                 return -EINVAL;
3136
3137         memset(usvc, 0, sizeof(*usvc));
3138
3139         usvc->af = nla_get_u16(nla_af);
3140         if (!ip_vs_is_af_valid(usvc->af))
3141                 return -EAFNOSUPPORT;
3142
3143         if (nla_fwmark) {
3144                 usvc->protocol = IPPROTO_TCP;
3145                 usvc->fwmark = nla_get_u32(nla_fwmark);
3146         } else {
3147                 usvc->protocol = nla_get_u16(nla_protocol);
3148                 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3149                 usvc->port = nla_get_be16(nla_port);
3150                 usvc->fwmark = 0;
3151         }
3152
3153         rcu_read_lock();
3154         if (usvc->fwmark)
3155                 svc = __ip_vs_svc_fwm_find(ipvs, usvc->af, usvc->fwmark);
3156         else
3157                 svc = __ip_vs_service_find(ipvs, usvc->af, usvc->protocol,
3158                                            &usvc->addr, usvc->port);
3159         rcu_read_unlock();
3160         *ret_svc = svc;
3161
3162         /* If a full entry was requested, check for the additional fields */
3163         if (full_entry) {
3164                 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3165                               *nla_netmask;
3166                 struct ip_vs_flags flags;
3167
3168                 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3169                 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3170                 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3171                 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3172                 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3173
3174                 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3175                         return -EINVAL;
3176
3177                 nla_memcpy(&flags, nla_flags, sizeof(flags));
3178
3179                 /* prefill flags from service if it already exists */
3180                 if (svc)
3181                         usvc->flags = svc->flags;
3182
3183                 /* set new flags from userland */
3184                 usvc->flags = (usvc->flags & ~flags.mask) |
3185                               (flags.flags & flags.mask);
3186                 usvc->sched_name = nla_data(nla_sched);
3187                 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3188                 usvc->timeout = nla_get_u32(nla_timeout);
3189                 usvc->netmask = nla_get_be32(nla_netmask);
3190         }
3191
3192         return 0;
3193 }
3194
3195 static struct ip_vs_service *ip_vs_genl_find_service(struct netns_ipvs *ipvs,
3196                                                      struct nlattr *nla)
3197 {
3198         struct ip_vs_service_user_kern usvc;
3199         struct ip_vs_service *svc;
3200         int ret;
3201
3202         ret = ip_vs_genl_parse_service(ipvs, &usvc, nla, 0, &svc);
3203         return ret ? ERR_PTR(ret) : svc;
3204 }
3205
3206 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3207 {
3208         struct nlattr *nl_dest;
3209         struct ip_vs_kstats kstats;
3210
3211         nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3212         if (!nl_dest)
3213                 return -EMSGSIZE;
3214
3215         if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3216             nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3217             nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3218                         (atomic_read(&dest->conn_flags) &
3219                          IP_VS_CONN_F_FWD_MASK)) ||
3220             nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3221                         atomic_read(&dest->weight)) ||
3222             nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3223             nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3224             nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3225                         atomic_read(&dest->activeconns)) ||
3226             nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3227                         atomic_read(&dest->inactconns)) ||
3228             nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3229                         atomic_read(&dest->persistconns)) ||
3230             nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3231                 goto nla_put_failure;
3232         ip_vs_copy_stats(&kstats, &dest->stats);
3233         if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats))
3234                 goto nla_put_failure;
3235         if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats))
3236                 goto nla_put_failure;
3237
3238         nla_nest_end(skb, nl_dest);
3239
3240         return 0;
3241
3242 nla_put_failure:
3243         nla_nest_cancel(skb, nl_dest);
3244         return -EMSGSIZE;
3245 }
3246
3247 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3248                                 struct netlink_callback *cb)
3249 {
3250         void *hdr;
3251
3252         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3253                           &ip_vs_genl_family, NLM_F_MULTI,
3254                           IPVS_CMD_NEW_DEST);
3255         if (!hdr)
3256                 return -EMSGSIZE;
3257
3258         if (ip_vs_genl_fill_dest(skb, dest) < 0)
3259                 goto nla_put_failure;
3260
3261         genlmsg_end(skb, hdr);
3262         return 0;
3263
3264 nla_put_failure:
3265         genlmsg_cancel(skb, hdr);
3266         return -EMSGSIZE;
3267 }
3268
3269 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3270                                  struct netlink_callback *cb)
3271 {
3272         int idx = 0;
3273         int start = cb->args[0];
3274         struct ip_vs_service *svc;
3275         struct ip_vs_dest *dest;
3276         struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3277         struct net *net = sock_net(skb->sk);
3278         struct netns_ipvs *ipvs = net_ipvs(net);
3279
3280         mutex_lock(&__ip_vs_mutex);
3281
3282         /* Try to find the service for which to dump destinations */
3283         if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs, IPVS_CMD_ATTR_MAX,
3284                         ip_vs_cmd_policy, NULL))
3285                 goto out_err;
3286
3287
3288         svc = ip_vs_genl_find_service(ipvs, attrs[IPVS_CMD_ATTR_SERVICE]);
3289         if (IS_ERR_OR_NULL(svc))
3290                 goto out_err;
3291
3292         /* Dump the destinations */
3293         list_for_each_entry(dest, &svc->destinations, n_list) {
3294                 if (++idx <= start)
3295                         continue;
3296                 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3297                         idx--;
3298                         goto nla_put_failure;
3299                 }
3300         }
3301
3302 nla_put_failure:
3303         cb->args[0] = idx;
3304
3305 out_err:
3306         mutex_unlock(&__ip_vs_mutex);
3307
3308         return skb->len;
3309 }
3310
3311 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3312                                  struct nlattr *nla, int full_entry)
3313 {
3314         struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3315         struct nlattr *nla_addr, *nla_port;
3316         struct nlattr *nla_addr_family;
3317
3318         /* Parse mandatory identifying destination fields first */
3319         if (nla == NULL ||
3320             nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla,
3321                              ip_vs_dest_policy, NULL))
3322                 return -EINVAL;
3323
3324         nla_addr        = attrs[IPVS_DEST_ATTR_ADDR];
3325         nla_port        = attrs[IPVS_DEST_ATTR_PORT];
3326         nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3327
3328         if (!(nla_addr && nla_port))
3329                 return -EINVAL;
3330
3331         memset(udest, 0, sizeof(*udest));
3332
3333         nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3334         udest->port = nla_get_be16(nla_port);
3335
3336         if (nla_addr_family)
3337                 udest->af = nla_get_u16(nla_addr_family);
3338         else
3339                 udest->af = 0;
3340
3341         /* If a full entry was requested, check for the additional fields */
3342         if (full_entry) {
3343                 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3344                               *nla_l_thresh;
3345
3346                 nla_fwd         = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3347                 nla_weight      = attrs[IPVS_DEST_ATTR_WEIGHT];
3348                 nla_u_thresh    = attrs[IPVS_DEST_ATTR_U_THRESH];
3349                 nla_l_thresh    = attrs[IPVS_DEST_ATTR_L_THRESH];
3350
3351                 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3352                         return -EINVAL;
3353
3354                 udest->conn_flags = nla_get_u32(nla_fwd)
3355                                     & IP_VS_CONN_F_FWD_MASK;
3356                 udest->weight = nla_get_u32(nla_weight);
3357                 udest->u_threshold = nla_get_u32(nla_u_thresh);
3358                 udest->l_threshold = nla_get_u32(nla_l_thresh);
3359         }
3360
3361         return 0;
3362 }
3363
3364 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3365                                   struct ipvs_sync_daemon_cfg *c)
3366 {
3367         struct nlattr *nl_daemon;
3368
3369         nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3370         if (!nl_daemon)
3371                 return -EMSGSIZE;
3372
3373         if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3374             nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, c->mcast_ifn) ||
3375             nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, c->syncid) ||
3376             nla_put_u16(skb, IPVS_DAEMON_ATTR_SYNC_MAXLEN, c->sync_maxlen) ||
3377             nla_put_u16(skb, IPVS_DAEMON_ATTR_MCAST_PORT, c->mcast_port) ||
3378             nla_put_u8(skb, IPVS_DAEMON_ATTR_MCAST_TTL, c->mcast_ttl))
3379                 goto nla_put_failure;
3380 #ifdef CONFIG_IP_VS_IPV6
3381         if (c->mcast_af == AF_INET6) {
3382                 if (nla_put_in6_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP6,
3383                                      &c->mcast_group.in6))
3384                         goto nla_put_failure;
3385         } else
3386 #endif
3387                 if (c->mcast_af == AF_INET &&
3388                     nla_put_in_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP,
3389                                     c->mcast_group.ip))
3390                         goto nla_put_failure;
3391         nla_nest_end(skb, nl_daemon);
3392
3393         return 0;
3394
3395 nla_put_failure:
3396         nla_nest_cancel(skb, nl_daemon);
3397         return -EMSGSIZE;
3398 }
3399
3400 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3401                                   struct ipvs_sync_daemon_cfg *c,
3402                                   struct netlink_callback *cb)
3403 {
3404         void *hdr;
3405         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3406                           &ip_vs_genl_family, NLM_F_MULTI,
3407                           IPVS_CMD_NEW_DAEMON);
3408         if (!hdr)
3409                 return -EMSGSIZE;
3410
3411         if (ip_vs_genl_fill_daemon(skb, state, c))
3412                 goto nla_put_failure;
3413
3414         genlmsg_end(skb, hdr);
3415         return 0;
3416
3417 nla_put_failure:
3418         genlmsg_cancel(skb, hdr);
3419         return -EMSGSIZE;
3420 }
3421
3422 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3423                                    struct netlink_callback *cb)
3424 {
3425         struct net *net = sock_net(skb->sk);
3426         struct netns_ipvs *ipvs = net_ipvs(net);
3427
3428         mutex_lock(&ipvs->sync_mutex);
3429         if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3430                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3431                                            &ipvs->mcfg, cb) < 0)
3432                         goto nla_put_failure;
3433
3434                 cb->args[0] = 1;
3435         }
3436
3437         if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3438                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3439                                            &ipvs->bcfg, cb) < 0)
3440                         goto nla_put_failure;
3441
3442                 cb->args[1] = 1;
3443         }
3444
3445 nla_put_failure:
3446         mutex_unlock(&ipvs->sync_mutex);
3447
3448         return skb->len;
3449 }
3450
3451 static int ip_vs_genl_new_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3452 {
3453         struct ipvs_sync_daemon_cfg c;
3454         struct nlattr *a;
3455         int ret;
3456
3457         memset(&c, 0, sizeof(c));
3458         if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3459               attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3460               attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3461                 return -EINVAL;
3462         strlcpy(c.mcast_ifn, nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3463                 sizeof(c.mcast_ifn));
3464         c.syncid = nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]);
3465
3466         a = attrs[IPVS_DAEMON_ATTR_SYNC_MAXLEN];
3467         if (a)
3468                 c.sync_maxlen = nla_get_u16(a);
3469
3470         a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP];
3471         if (a) {
3472                 c.mcast_af = AF_INET;
3473                 c.mcast_group.ip = nla_get_in_addr(a);
3474                 if (!ipv4_is_multicast(c.mcast_group.ip))
3475                         return -EINVAL;
3476         } else {
3477                 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP6];
3478                 if (a) {
3479 #ifdef CONFIG_IP_VS_IPV6
3480                         int addr_type;
3481
3482                         c.mcast_af = AF_INET6;
3483                         c.mcast_group.in6 = nla_get_in6_addr(a);
3484                         addr_type = ipv6_addr_type(&c.mcast_group.in6);
3485                         if (!(addr_type & IPV6_ADDR_MULTICAST))
3486                                 return -EINVAL;
3487 #else
3488                         return -EAFNOSUPPORT;
3489 #endif
3490                 }
3491         }
3492
3493         a = attrs[IPVS_DAEMON_ATTR_MCAST_PORT];
3494         if (a)
3495                 c.mcast_port = nla_get_u16(a);
3496
3497         a = attrs[IPVS_DAEMON_ATTR_MCAST_TTL];
3498         if (a)
3499                 c.mcast_ttl = nla_get_u8(a);
3500
3501         /* The synchronization protocol is incompatible with mixed family
3502          * services
3503          */
3504         if (ipvs->mixed_address_family_dests > 0)
3505                 return -EINVAL;
3506
3507         ret = start_sync_thread(ipvs, &c,
3508                                 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3509         return ret;
3510 }
3511
3512 static int ip_vs_genl_del_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3513 {
3514         int ret;
3515
3516         if (!attrs[IPVS_DAEMON_ATTR_STATE])
3517                 return -EINVAL;
3518
3519         ret = stop_sync_thread(ipvs,
3520                                nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3521         return ret;
3522 }
3523
3524 static int ip_vs_genl_set_config(struct netns_ipvs *ipvs, struct nlattr **attrs)
3525 {
3526         struct ip_vs_timeout_user t;
3527
3528         __ip_vs_get_timeouts(ipvs, &t);
3529
3530         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3531                 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3532
3533         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3534                 t.tcp_fin_timeout =
3535                         nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3536
3537         if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3538                 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3539
3540         return ip_vs_set_timeout(ipvs, &t);
3541 }
3542
3543 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3544 {
3545         int ret = -EINVAL, cmd;
3546         struct net *net = sock_net(skb->sk);
3547         struct netns_ipvs *ipvs = net_ipvs(net);
3548
3549         cmd = info->genlhdr->cmd;
3550
3551         if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3552                 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3553
3554                 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3555                     nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3556                                      info->attrs[IPVS_CMD_ATTR_DAEMON],
3557                                      ip_vs_daemon_policy, info->extack))
3558                         goto out;
3559
3560                 if (cmd == IPVS_CMD_NEW_DAEMON)
3561                         ret = ip_vs_genl_new_daemon(ipvs, daemon_attrs);
3562                 else
3563                         ret = ip_vs_genl_del_daemon(ipvs, daemon_attrs);
3564         }
3565
3566 out:
3567         return ret;
3568 }
3569
3570 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3571 {
3572         struct ip_vs_service *svc = NULL;
3573         struct ip_vs_service_user_kern usvc;
3574         struct ip_vs_dest_user_kern udest;
3575         int ret = 0, cmd;
3576         int need_full_svc = 0, need_full_dest = 0;
3577         struct net *net = sock_net(skb->sk);
3578         struct netns_ipvs *ipvs = net_ipvs(net);
3579
3580         cmd = info->genlhdr->cmd;
3581
3582         mutex_lock(&__ip_vs_mutex);
3583
3584         if (cmd == IPVS_CMD_FLUSH) {
3585                 ret = ip_vs_flush(ipvs, false);
3586                 goto out;
3587         } else if (cmd == IPVS_CMD_SET_CONFIG) {
3588                 ret = ip_vs_genl_set_config(ipvs, info->attrs);
3589                 goto out;
3590         } else if (cmd == IPVS_CMD_ZERO &&
3591                    !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3592                 ret = ip_vs_zero_all(ipvs);
3593                 goto out;
3594         }
3595
3596         /* All following commands require a service argument, so check if we
3597          * received a valid one. We need a full service specification when
3598          * adding / editing a service. Only identifying members otherwise. */
3599         if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3600                 need_full_svc = 1;
3601
3602         ret = ip_vs_genl_parse_service(ipvs, &usvc,
3603                                        info->attrs[IPVS_CMD_ATTR_SERVICE],
3604                                        need_full_svc, &svc);
3605         if (ret)
3606                 goto out;
3607
3608         /* Unless we're adding a new service, the service must already exist */
3609         if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3610                 ret = -ESRCH;
3611                 goto out;
3612         }
3613
3614         /* Destination commands require a valid destination argument. For
3615          * adding / editing a destination, we need a full destination
3616          * specification. */
3617         if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3618             cmd == IPVS_CMD_DEL_DEST) {
3619                 if (cmd != IPVS_CMD_DEL_DEST)
3620                         need_full_dest = 1;
3621
3622                 ret = ip_vs_genl_parse_dest(&udest,
3623                                             info->attrs[IPVS_CMD_ATTR_DEST],
3624                                             need_full_dest);
3625                 if (ret)
3626                         goto out;
3627
3628                 /* Old protocols did not allow the user to specify address
3629                  * family, so we set it to zero instead.  We also didn't
3630                  * allow heterogeneous pools in the old code, so it's safe
3631                  * to assume that this will have the same address family as
3632                  * the service.
3633                  */
3634                 if (udest.af == 0)
3635                         udest.af = svc->af;
3636
3637                 if (!ip_vs_is_af_valid(udest.af)) {
3638                         ret = -EAFNOSUPPORT;
3639                         goto out;
3640                 }
3641
3642                 if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) {
3643                         /* The synchronization protocol is incompatible
3644                          * with mixed family services
3645                          */
3646                         if (ipvs->sync_state) {
3647                                 ret = -EINVAL;
3648                                 goto out;
3649                         }
3650
3651                         /* Which connection types do we support? */
3652                         switch (udest.conn_flags) {
3653                         case IP_VS_CONN_F_TUNNEL:
3654                                 /* We are able to forward this */
3655                                 break;
3656                         default:
3657                                 ret = -EINVAL;
3658                                 goto out;
3659                         }
3660                 }
3661         }
3662
3663         switch (cmd) {
3664         case IPVS_CMD_NEW_SERVICE:
3665                 if (svc == NULL)
3666                         ret = ip_vs_add_service(ipvs, &usvc, &svc);
3667                 else
3668                         ret = -EEXIST;
3669                 break;
3670         case IPVS_CMD_SET_SERVICE:
3671                 ret = ip_vs_edit_service(svc, &usvc);
3672                 break;
3673         case IPVS_CMD_DEL_SERVICE:
3674                 ret = ip_vs_del_service(svc);
3675                 /* do not use svc, it can be freed */
3676                 break;
3677         case IPVS_CMD_NEW_DEST:
3678                 ret = ip_vs_add_dest(svc, &udest);
3679                 break;
3680         case IPVS_CMD_SET_DEST:
3681                 ret = ip_vs_edit_dest(svc, &udest);
3682                 break;
3683         case IPVS_CMD_DEL_DEST:
3684                 ret = ip_vs_del_dest(svc, &udest);
3685                 break;
3686         case IPVS_CMD_ZERO:
3687                 ret = ip_vs_zero_service(svc);
3688                 break;
3689         default:
3690                 ret = -EINVAL;
3691         }
3692
3693 out:
3694         mutex_unlock(&__ip_vs_mutex);
3695
3696         return ret;
3697 }
3698
3699 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3700 {
3701         struct sk_buff *msg;
3702         void *reply;
3703         int ret, cmd, reply_cmd;
3704         struct net *net = sock_net(skb->sk);
3705         struct netns_ipvs *ipvs = net_ipvs(net);
3706
3707         cmd = info->genlhdr->cmd;
3708
3709         if (cmd == IPVS_CMD_GET_SERVICE)
3710                 reply_cmd = IPVS_CMD_NEW_SERVICE;
3711         else if (cmd == IPVS_CMD_GET_INFO)
3712                 reply_cmd = IPVS_CMD_SET_INFO;
3713         else if (cmd == IPVS_CMD_GET_CONFIG)
3714                 reply_cmd = IPVS_CMD_SET_CONFIG;
3715         else {
3716                 pr_err("unknown Generic Netlink command\n");
3717                 return -EINVAL;
3718         }
3719
3720         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3721         if (!msg)
3722                 return -ENOMEM;
3723
3724         mutex_lock(&__ip_vs_mutex);
3725
3726         reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3727         if (reply == NULL)
3728                 goto nla_put_failure;
3729
3730         switch (cmd) {
3731         case IPVS_CMD_GET_SERVICE:
3732         {
3733                 struct ip_vs_service *svc;
3734
3735                 svc = ip_vs_genl_find_service(ipvs,
3736                                               info->attrs[IPVS_CMD_ATTR_SERVICE]);
3737                 if (IS_ERR(svc)) {
3738                         ret = PTR_ERR(svc);
3739                         goto out_err;
3740                 } else if (svc) {
3741                         ret = ip_vs_genl_fill_service(msg, svc);
3742                         if (ret)
3743                                 goto nla_put_failure;
3744                 } else {
3745                         ret = -ESRCH;
3746                         goto out_err;
3747                 }
3748
3749                 break;
3750         }
3751
3752         case IPVS_CMD_GET_CONFIG:
3753         {
3754                 struct ip_vs_timeout_user t;
3755
3756                 __ip_vs_get_timeouts(ipvs, &t);
3757 #ifdef CONFIG_IP_VS_PROTO_TCP
3758                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3759                                 t.tcp_timeout) ||
3760                     nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3761                                 t.tcp_fin_timeout))
3762                         goto nla_put_failure;
3763 #endif
3764 #ifdef CONFIG_IP_VS_PROTO_UDP
3765                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3766                         goto nla_put_failure;
3767 #endif
3768
3769                 break;
3770         }
3771
3772         case IPVS_CMD_GET_INFO:
3773                 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3774                                 IP_VS_VERSION_CODE) ||
3775                     nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3776                                 ip_vs_conn_tab_size))
3777                         goto nla_put_failure;
3778                 break;
3779         }
3780
3781         genlmsg_end(msg, reply);
3782         ret = genlmsg_reply(msg, info);
3783         goto out;
3784
3785 nla_put_failure:
3786         pr_err("not enough space in Netlink message\n");
3787         ret = -EMSGSIZE;
3788
3789 out_err:
3790         nlmsg_free(msg);
3791 out:
3792         mutex_unlock(&__ip_vs_mutex);
3793
3794         return ret;
3795 }
3796
3797
3798 static const struct genl_ops ip_vs_genl_ops[] = {
3799         {
3800                 .cmd    = IPVS_CMD_NEW_SERVICE,
3801                 .flags  = GENL_ADMIN_PERM,
3802                 .policy = ip_vs_cmd_policy,
3803                 .doit   = ip_vs_genl_set_cmd,
3804         },
3805         {
3806                 .cmd    = IPVS_CMD_SET_SERVICE,
3807                 .flags  = GENL_ADMIN_PERM,
3808                 .policy = ip_vs_cmd_policy,
3809                 .doit   = ip_vs_genl_set_cmd,
3810         },
3811         {
3812                 .cmd    = IPVS_CMD_DEL_SERVICE,
3813                 .flags  = GENL_ADMIN_PERM,
3814                 .policy = ip_vs_cmd_policy,
3815                 .doit   = ip_vs_genl_set_cmd,
3816         },
3817         {
3818                 .cmd    = IPVS_CMD_GET_SERVICE,
3819                 .flags  = GENL_ADMIN_PERM,
3820                 .doit   = ip_vs_genl_get_cmd,
3821                 .dumpit = ip_vs_genl_dump_services,
3822                 .policy = ip_vs_cmd_policy,
3823         },
3824         {
3825                 .cmd    = IPVS_CMD_NEW_DEST,
3826                 .flags  = GENL_ADMIN_PERM,
3827                 .policy = ip_vs_cmd_policy,
3828                 .doit   = ip_vs_genl_set_cmd,
3829         },
3830         {
3831                 .cmd    = IPVS_CMD_SET_DEST,
3832                 .flags  = GENL_ADMIN_PERM,
3833                 .policy = ip_vs_cmd_policy,
3834                 .doit   = ip_vs_genl_set_cmd,
3835         },
3836         {
3837                 .cmd    = IPVS_CMD_DEL_DEST,
3838                 .flags  = GENL_ADMIN_PERM,
3839                 .policy = ip_vs_cmd_policy,
3840                 .doit   = ip_vs_genl_set_cmd,
3841         },
3842         {
3843                 .cmd    = IPVS_CMD_GET_DEST,
3844                 .flags  = GENL_ADMIN_PERM,
3845                 .policy = ip_vs_cmd_policy,
3846                 .dumpit = ip_vs_genl_dump_dests,
3847         },
3848         {
3849                 .cmd    = IPVS_CMD_NEW_DAEMON,
3850                 .flags  = GENL_ADMIN_PERM,
3851                 .policy = ip_vs_cmd_policy,
3852                 .doit   = ip_vs_genl_set_daemon,
3853         },
3854         {
3855                 .cmd    = IPVS_CMD_DEL_DAEMON,
3856                 .flags  = GENL_ADMIN_PERM,
3857                 .policy = ip_vs_cmd_policy,
3858                 .doit   = ip_vs_genl_set_daemon,
3859         },
3860         {
3861                 .cmd    = IPVS_CMD_GET_DAEMON,
3862                 .flags  = GENL_ADMIN_PERM,
3863                 .dumpit = ip_vs_genl_dump_daemons,
3864         },
3865         {
3866                 .cmd    = IPVS_CMD_SET_CONFIG,
3867                 .flags  = GENL_ADMIN_PERM,
3868                 .policy = ip_vs_cmd_policy,
3869                 .doit   = ip_vs_genl_set_cmd,
3870         },
3871         {
3872                 .cmd    = IPVS_CMD_GET_CONFIG,
3873                 .flags  = GENL_ADMIN_PERM,
3874                 .doit   = ip_vs_genl_get_cmd,
3875         },
3876         {
3877                 .cmd    = IPVS_CMD_GET_INFO,
3878                 .flags  = GENL_ADMIN_PERM,
3879                 .doit   = ip_vs_genl_get_cmd,
3880         },
3881         {
3882                 .cmd    = IPVS_CMD_ZERO,
3883                 .flags  = GENL_ADMIN_PERM,
3884                 .policy = ip_vs_cmd_policy,
3885                 .doit   = ip_vs_genl_set_cmd,
3886         },
3887         {
3888                 .cmd    = IPVS_CMD_FLUSH,
3889                 .flags  = GENL_ADMIN_PERM,
3890                 .doit   = ip_vs_genl_set_cmd,
3891         },
3892 };
3893
3894 static struct genl_family ip_vs_genl_family __ro_after_init = {
3895         .hdrsize        = 0,
3896         .name           = IPVS_GENL_NAME,
3897         .version        = IPVS_GENL_VERSION,
3898         .maxattr        = IPVS_CMD_ATTR_MAX,
3899         .netnsok        = true,         /* Make ipvsadm to work on netns */
3900         .module         = THIS_MODULE,
3901         .ops            = ip_vs_genl_ops,
3902         .n_ops          = ARRAY_SIZE(ip_vs_genl_ops),
3903 };
3904
3905 static int __init ip_vs_genl_register(void)
3906 {
3907         return genl_register_family(&ip_vs_genl_family);
3908 }
3909
3910 static void ip_vs_genl_unregister(void)
3911 {
3912         genl_unregister_family(&ip_vs_genl_family);
3913 }
3914
3915 /* End of Generic Netlink interface definitions */
3916
3917 /*
3918  * per netns intit/exit func.
3919  */
3920 #ifdef CONFIG_SYSCTL
3921 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs)
3922 {
3923         struct net *net = ipvs->net;
3924         int idx;
3925         struct ctl_table *tbl;
3926
3927         atomic_set(&ipvs->dropentry, 0);
3928         spin_lock_init(&ipvs->dropentry_lock);
3929         spin_lock_init(&ipvs->droppacket_lock);
3930         spin_lock_init(&ipvs->securetcp_lock);
3931
3932         if (!net_eq(net, &init_net)) {
3933                 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3934                 if (tbl == NULL)
3935                         return -ENOMEM;
3936
3937                 /* Don't export sysctls to unprivileged users */
3938                 if (net->user_ns != &init_user_ns)
3939                         tbl[0].procname = NULL;
3940         } else
3941                 tbl = vs_vars;
3942         /* Initialize sysctl defaults */
3943         for (idx = 0; idx < ARRAY_SIZE(vs_vars); idx++) {
3944                 if (tbl[idx].proc_handler == proc_do_defense_mode)
3945                         tbl[idx].extra2 = ipvs;
3946         }
3947         idx = 0;
3948         ipvs->sysctl_amemthresh = 1024;
3949         tbl[idx++].data = &ipvs->sysctl_amemthresh;
3950         ipvs->sysctl_am_droprate = 10;
3951         tbl[idx++].data = &ipvs->sysctl_am_droprate;
3952         tbl[idx++].data = &ipvs->sysctl_drop_entry;
3953         tbl[idx++].data = &ipvs->sysctl_drop_packet;
3954 #ifdef CONFIG_IP_VS_NFCT
3955         tbl[idx++].data = &ipvs->sysctl_conntrack;
3956 #endif
3957         tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3958         ipvs->sysctl_snat_reroute = 1;
3959         tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3960         ipvs->sysctl_sync_ver = 1;
3961         tbl[idx++].data = &ipvs->sysctl_sync_ver;
3962         ipvs->sysctl_sync_ports = 1;
3963         tbl[idx++].data = &ipvs->sysctl_sync_ports;
3964         tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3965         ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3966         tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3967         ipvs->sysctl_sync_sock_size = 0;
3968         tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3969         tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3970         tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3971         tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3972         tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3973         tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3974         ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3975         ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3976         tbl[idx].data = &ipvs->sysctl_sync_threshold;
3977         tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3978         ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3979         tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3980         ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3981         tbl[idx++].data = &ipvs->sysctl_sync_retries;
3982         tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3983         ipvs->sysctl_pmtu_disc = 1;
3984         tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3985         tbl[idx++].data = &ipvs->sysctl_backup_only;
3986         ipvs->sysctl_conn_reuse_mode = 1;
3987         tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
3988         tbl[idx++].data = &ipvs->sysctl_schedule_icmp;
3989         tbl[idx++].data = &ipvs->sysctl_ignore_tunneled;
3990 #ifdef CONFIG_IP_VS_DEBUG
3991         /* Global sysctls must be ro in non-init netns */
3992         if (!net_eq(net, &init_net))
3993                 tbl[idx++].mode = 0444;
3994 #endif
3995
3996         ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3997         if (ipvs->sysctl_hdr == NULL) {
3998                 if (!net_eq(net, &init_net))
3999                         kfree(tbl);
4000                 return -ENOMEM;
4001         }
4002         ip_vs_start_estimator(ipvs, &ipvs->tot_stats);
4003         ipvs->sysctl_tbl = tbl;
4004         /* Schedule defense work */
4005         INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
4006         schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
4007
4008         return 0;
4009 }
4010
4011 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs)
4012 {
4013         struct net *net = ipvs->net;
4014
4015         cancel_delayed_work_sync(&ipvs->defense_work);
4016         cancel_work_sync(&ipvs->defense_work.work);
4017         unregister_net_sysctl_table(ipvs->sysctl_hdr);
4018         ip_vs_stop_estimator(ipvs, &ipvs->tot_stats);
4019
4020         if (!net_eq(net, &init_net))
4021                 kfree(ipvs->sysctl_tbl);
4022 }
4023
4024 #else
4025
4026 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) { return 0; }
4027 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs) { }
4028
4029 #endif
4030
4031 static struct notifier_block ip_vs_dst_notifier = {
4032         .notifier_call = ip_vs_dst_event,
4033 #ifdef CONFIG_IP_VS_IPV6
4034         .priority = ADDRCONF_NOTIFY_PRIORITY + 5,
4035 #endif
4036 };
4037
4038 int __net_init ip_vs_control_net_init(struct netns_ipvs *ipvs)
4039 {
4040         int i, idx;
4041
4042         /* Initialize rs_table */
4043         for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
4044                 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
4045
4046         INIT_LIST_HEAD(&ipvs->dest_trash);
4047         spin_lock_init(&ipvs->dest_trash_lock);
4048         setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
4049                     (unsigned long) ipvs);
4050         atomic_set(&ipvs->ftpsvc_counter, 0);
4051         atomic_set(&ipvs->nullsvc_counter, 0);
4052         atomic_set(&ipvs->conn_out_counter, 0);
4053
4054         /* procfs stats */
4055         ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
4056         if (!ipvs->tot_stats.cpustats)
4057                 return -ENOMEM;
4058
4059         for_each_possible_cpu(i) {
4060                 struct ip_vs_cpu_stats *ipvs_tot_stats;
4061                 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
4062                 u64_stats_init(&ipvs_tot_stats->syncp);
4063         }
4064
4065         spin_lock_init(&ipvs->tot_stats.lock);
4066
4067         proc_create("ip_vs", 0, ipvs->net->proc_net, &ip_vs_info_fops);
4068         proc_create("ip_vs_stats", 0, ipvs->net->proc_net, &ip_vs_stats_fops);
4069         proc_create("ip_vs_stats_percpu", 0, ipvs->net->proc_net,
4070                     &ip_vs_stats_percpu_fops);
4071
4072         if (ip_vs_control_net_init_sysctl(ipvs))
4073                 goto err;
4074
4075         return 0;
4076
4077 err:
4078         free_percpu(ipvs->tot_stats.cpustats);
4079         return -ENOMEM;
4080 }
4081
4082 void __net_exit ip_vs_control_net_cleanup(struct netns_ipvs *ipvs)
4083 {
4084         ip_vs_trash_cleanup(ipvs);
4085         ip_vs_control_net_cleanup_sysctl(ipvs);
4086         remove_proc_entry("ip_vs_stats_percpu", ipvs->net->proc_net);
4087         remove_proc_entry("ip_vs_stats", ipvs->net->proc_net);
4088         remove_proc_entry("ip_vs", ipvs->net->proc_net);
4089         free_percpu(ipvs->tot_stats.cpustats);
4090 }
4091
4092 int __init ip_vs_register_nl_ioctl(void)
4093 {
4094         int ret;
4095
4096         ret = nf_register_sockopt(&ip_vs_sockopts);
4097         if (ret) {
4098                 pr_err("cannot register sockopt.\n");
4099                 goto err_sock;
4100         }
4101
4102         ret = ip_vs_genl_register();
4103         if (ret) {
4104                 pr_err("cannot register Generic Netlink interface.\n");
4105                 goto err_genl;
4106         }
4107         return 0;
4108
4109 err_genl:
4110         nf_unregister_sockopt(&ip_vs_sockopts);
4111 err_sock:
4112         return ret;
4113 }
4114
4115 void ip_vs_unregister_nl_ioctl(void)
4116 {
4117         ip_vs_genl_unregister();
4118         nf_unregister_sockopt(&ip_vs_sockopts);
4119 }
4120
4121 int __init ip_vs_control_init(void)
4122 {
4123         int idx;
4124         int ret;
4125
4126         EnterFunction(2);
4127
4128         /* Initialize svc_table, ip_vs_svc_fwm_table */
4129         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
4130                 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
4131                 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
4132         }
4133
4134         smp_wmb();      /* Do we really need it now ? */
4135
4136         ret = register_netdevice_notifier(&ip_vs_dst_notifier);
4137         if (ret < 0)
4138                 return ret;
4139
4140         LeaveFunction(2);
4141         return 0;
4142 }
4143
4144
4145 void ip_vs_control_cleanup(void)
4146 {
4147         EnterFunction(2);
4148         unregister_netdevice_notifier(&ip_vs_dst_notifier);
4149         LeaveFunction(2);
4150 }