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 #endif
  47 #include <net/route.h>
  48 #include <net/sock.h>
  49 #include <net/genetlink.h>
  50
  51 #include <asm/uaccess.h>
  52
  53 #include <net/ip_vs.h>
  54
  55 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
  56 static DEFINE_MUTEX(__ip_vs_mutex);
  57
  58 /* sysctl variables */
  59
  60 #ifdef CONFIG_IP_VS_DEBUG
  61 static int sysctl_ip_vs_debug_level = 0;
  62
  63 int ip_vs_get_debug_level(void)
  64 {
  65         return sysctl_ip_vs_debug_level;
  66 }
  67 #endif
  68
  69
  70 /*  Protos */
  71 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup);
  72
  73
  74 #ifdef CONFIG_IP_VS_IPV6
  75 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
  76 static bool __ip_vs_addr_is_local_v6(struct net *net,
  77                                      const struct in6_addr *addr)
  78 {
  79         struct flowi6 fl6 = {
  80                 .daddr = *addr,
  81         };
  82         struct dst_entry *dst = ip6_route_output(net, NULL, &fl6);
  83         bool is_local;
  84
  85         is_local = !dst->error && dst->dev && (dst->dev->flags & IFF_LOOPBACK);
  86
  87         dst_release(dst);
  88         return is_local;
  89 }
  90 #endif
  91
  92 #ifdef CONFIG_SYSCTL
  93 /*
  94  *      update_defense_level is called from keventd and from sysctl,
  95  *      so it needs to protect itself from softirqs
  96  */
  97 static void update_defense_level(struct netns_ipvs *ipvs)
  98 {
  99         struct sysinfo i;
 100         static int old_secure_tcp = 0;
 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 (old_secure_tcp >= 2)
 182                         to_change = 0;
 183                 break;
 184         case 1:
 185                 if (nomem) {
 186                         if (old_secure_tcp < 2)
 187                                 to_change = 1;
 188                         ipvs->sysctl_secure_tcp = 2;
 189                 } else {
 190                         if (old_secure_tcp >= 2)
 191                                 to_change = 0;
 192                 }
 193                 break;
 194         case 2:
 195                 if (nomem) {
 196                         if (old_secure_tcp < 2)
 197                                 to_change = 1;
 198                 } else {
 199                         if (old_secure_tcp >= 2)
 200                                 to_change = 0;
 201                         ipvs->sysctl_secure_tcp = 1;
 202                 }
 203                 break;
 204         case 3:
 205                 if (old_secure_tcp < 2)
 206                         to_change = 1;
 207                 break;
 208         }
 209         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->net);
 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 net *net, 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) net >> 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 net *net, __u32 fwmark)
 289 {
 290         return (((size_t)net>>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->net, 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->net, 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 net *net, 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(net, 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                     && net_eq(svc->net, net)) {
 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 net *net, 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(net, 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                     && net_eq(svc->net, net)) {
 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 net *net, int af, __u32 fwmark, __u16 protocol,
 410                    const union nf_inet_addr *vaddr, __be16 vport)
 411 {
 412         struct ip_vs_service *svc;
 413         struct netns_ipvs *ipvs = net_ipvs(net);
 414
 415         /*
 416          *      Check the table hashed by fwmark first
 417          */
 418         if (fwmark) {
 419                 svc = __ip_vs_svc_fwm_find(net, af, fwmark);
 420                 if (svc)
 421                         goto out;
 422         }
 423
 424         /*
 425          *      Check the table hashed by <protocol,addr,port>
 426          *      for "full" addressed entries
 427          */
 428         svc = __ip_vs_service_find(net, af, protocol, vaddr, vport);
 429
 430         if (svc == NULL
 431             && protocol == IPPROTO_TCP
 432             && atomic_read(&ipvs->ftpsvc_counter)
 433             && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
 434                 /*
 435                  * Check if ftp service entry exists, the packet
 436                  * might belong to FTP data connections.
 437                  */
 438                 svc = __ip_vs_service_find(net, af, protocol, vaddr, FTPPORT);
 439         }
 440
 441         if (svc == NULL
 442             && atomic_read(&ipvs->nullsvc_counter)) {
 443                 /*
 444                  * Check if the catch-all port (port zero) exists
 445                  */
 446                 svc = __ip_vs_service_find(net, af, protocol, vaddr, 0);
 447         }
 448
 449   out:
 450         IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
 451                       fwmark, ip_vs_proto_name(protocol),
 452                       IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
 453                       svc ? "hit" : "not hit");
 454
 455         return svc;
 456 }
 457
 458
 459 static inline void
 460 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
 461 {
 462         atomic_inc(&svc->refcnt);
 463         rcu_assign_pointer(dest->svc, svc);
 464 }
 465
 466 static void ip_vs_service_free(struct ip_vs_service *svc)
 467 {
 468         if (svc->stats.cpustats)
 469                 free_percpu(svc->stats.cpustats);
 470         kfree(svc);
 471 }
 472
 473 static void ip_vs_service_rcu_free(struct rcu_head *head)
 474 {
 475         struct ip_vs_service *svc;
 476
 477         svc = container_of(head, struct ip_vs_service, rcu_head);
 478         ip_vs_service_free(svc);
 479 }
 480
 481 static void __ip_vs_svc_put(struct ip_vs_service *svc, bool do_delay)
 482 {
 483         if (atomic_dec_and_test(&svc->refcnt)) {
 484                 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
 485                               svc->fwmark,
 486                               IP_VS_DBG_ADDR(svc->af, &svc->addr),
 487                               ntohs(svc->port));
 488                 if (do_delay)
 489                         call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
 490                 else
 491                         ip_vs_service_free(svc);
 492         }
 493 }
 494
 495
 496 /*
 497  *      Returns hash value for real service
 498  */
 499 static inline unsigned int ip_vs_rs_hashkey(int af,
 500                                             const union nf_inet_addr *addr,
 501                                             __be16 port)
 502 {
 503         register unsigned int porth = ntohs(port);
 504         __be32 addr_fold = addr->ip;
 505
 506 #ifdef CONFIG_IP_VS_IPV6
 507         if (af == AF_INET6)
 508                 addr_fold = addr->ip6[0]^addr->ip6[1]^
 509                             addr->ip6[2]^addr->ip6[3];
 510 #endif
 511
 512         return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
 513                 & IP_VS_RTAB_MASK;
 514 }
 515
 516 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
 517 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
 518 {
 519         unsigned int hash;
 520
 521         if (dest->in_rs_table)
 522                 return;
 523
 524         /*
 525          *      Hash by proto,addr,port,
 526          *      which are the parameters of the real service.
 527          */
 528         hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
 529
 530         hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
 531         dest->in_rs_table = 1;
 532 }
 533
 534 /* Unhash ip_vs_dest from rs_table. */
 535 static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
 536 {
 537         /*
 538          * Remove it from the rs_table table.
 539          */
 540         if (dest->in_rs_table) {
 541                 hlist_del_rcu(&dest->d_list);
 542                 dest->in_rs_table = 0;
 543         }
 544 }
 545
 546 /* Check if real service by <proto,addr,port> is present */
 547 bool ip_vs_has_real_service(struct net *net, int af, __u16 protocol,
 548                             const union nf_inet_addr *daddr, __be16 dport)
 549 {
 550         struct netns_ipvs *ipvs = net_ipvs(net);
 551         unsigned int hash;
 552         struct ip_vs_dest *dest;
 553
 554         /* Check for "full" addressed entries */
 555         hash = ip_vs_rs_hashkey(af, daddr, dport);
 556
 557         rcu_read_lock();
 558         hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
 559                 if (dest->port == dport &&
 560                     dest->af == af &&
 561                     ip_vs_addr_equal(af, &dest->addr, daddr) &&
 562                     (dest->protocol == protocol || dest->vfwmark)) {
 563                         /* HIT */
 564                         rcu_read_unlock();
 565                         return true;
 566                 }
 567         }
 568         rcu_read_unlock();
 569
 570         return false;
 571 }
 572
 573 /* Lookup destination by {addr,port} in the given service
 574  * Called under RCU lock.
 575  */
 576 static struct ip_vs_dest *
 577 ip_vs_lookup_dest(struct ip_vs_service *svc, int dest_af,
 578                   const union nf_inet_addr *daddr, __be16 dport)
 579 {
 580         struct ip_vs_dest *dest;
 581
 582         /*
 583          * Find the destination for the given service
 584          */
 585         list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
 586                 if ((dest->af == dest_af) &&
 587                     ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
 588                     (dest->port == dport)) {
 589                         /* HIT */
 590                         return dest;
 591                 }
 592         }
 593
 594         return NULL;
 595 }
 596
 597 /*
 598  * Find destination by {daddr,dport,vaddr,protocol}
 599  * Created to be used in ip_vs_process_message() in
 600  * the backup synchronization daemon. It finds the
 601  * destination to be bound to the received connection
 602  * on the backup.
 603  * Called under RCU lock, no refcnt is returned.
 604  */
 605 struct ip_vs_dest *ip_vs_find_dest(struct net  *net, int svc_af, int dest_af,
 606                                    const union nf_inet_addr *daddr,
 607                                    __be16 dport,
 608                                    const union nf_inet_addr *vaddr,
 609                                    __be16 vport, __u16 protocol, __u32 fwmark,
 610                                    __u32 flags)
 611 {
 612         struct ip_vs_dest *dest;
 613         struct ip_vs_service *svc;
 614         __be16 port = dport;
 615
 616         svc = ip_vs_service_find(net, svc_af, fwmark, protocol, vaddr, vport);
 617         if (!svc)
 618                 return NULL;
 619         if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
 620                 port = 0;
 621         dest = ip_vs_lookup_dest(svc, dest_af, daddr, port);
 622         if (!dest)
 623                 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port ^ dport);
 624         return dest;
 625 }
 626
 627 void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
 628 {
 629         struct ip_vs_dest_dst *dest_dst = container_of(head,
 630                                                        struct ip_vs_dest_dst,
 631                                                        rcu_head);
 632
 633         dst_release(dest_dst->dst_cache);
 634         kfree(dest_dst);
 635 }
 636
 637 /* Release dest_dst and dst_cache for dest in user context */
 638 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
 639 {
 640         struct ip_vs_dest_dst *old;
 641
 642         old = rcu_dereference_protected(dest->dest_dst, 1);
 643         if (old) {
 644                 RCU_INIT_POINTER(dest->dest_dst, NULL);
 645                 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
 646         }
 647 }
 648
 649 /*
 650  *  Lookup dest by {svc,addr,port} in the destination trash.
 651  *  The destination trash is used to hold the destinations that are removed
 652  *  from the service table but are still referenced by some conn entries.
 653  *  The reason to add the destination trash is when the dest is temporary
 654  *  down (either by administrator or by monitor program), the dest can be
 655  *  picked back from the trash, the remaining connections to the dest can
 656  *  continue, and the counting information of the dest is also useful for
 657  *  scheduling.
 658  */
 659 static struct ip_vs_dest *
 660 ip_vs_trash_get_dest(struct ip_vs_service *svc, int dest_af,
 661                      const union nf_inet_addr *daddr, __be16 dport)
 662 {
 663         struct ip_vs_dest *dest;
 664         struct netns_ipvs *ipvs = net_ipvs(svc->net);
 665
 666         /*
 667          * Find the destination in trash
 668          */
 669         spin_lock_bh(&ipvs->dest_trash_lock);
 670         list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
 671                 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
 672                               "dest->refcnt=%d\n",
 673                               dest->vfwmark,
 674                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
 675                               ntohs(dest->port),
 676                               atomic_read(&dest->refcnt));
 677                 if (dest->af == dest_af &&
 678                     ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
 679                     dest->port == dport &&
 680                     dest->vfwmark == svc->fwmark &&
 681                     dest->protocol == svc->protocol &&
 682                     (svc->fwmark ||
 683                      (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
 684                       dest->vport == svc->port))) {
 685                         /* HIT */
 686                         list_del(&dest->t_list);
 687                         ip_vs_dest_hold(dest);
 688                         goto out;
 689                 }
 690         }
 691
 692         dest = NULL;
 693
 694 out:
 695         spin_unlock_bh(&ipvs->dest_trash_lock);
 696
 697         return dest;
 698 }
 699
 700 static void ip_vs_dest_free(struct ip_vs_dest *dest)
 701 {
 702         struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1);
 703
 704         __ip_vs_dst_cache_reset(dest);
 705         __ip_vs_svc_put(svc, false);
 706         free_percpu(dest->stats.cpustats);
 707         ip_vs_dest_put_and_free(dest);
 708 }
 709
 710 /*
 711  *  Clean up all the destinations in the trash
 712  *  Called by the ip_vs_control_cleanup()
 713  *
 714  *  When the ip_vs_control_clearup is activated by ipvs module exit,
 715  *  the service tables must have been flushed and all the connections
 716  *  are expired, and the refcnt of each destination in the trash must
 717  *  be 0, so we simply release them here.
 718  */
 719 static void ip_vs_trash_cleanup(struct net *net)
 720 {
 721         struct ip_vs_dest *dest, *nxt;
 722         struct netns_ipvs *ipvs = net_ipvs(net);
 723
 724         del_timer_sync(&ipvs->dest_trash_timer);
 725         /* No need to use dest_trash_lock */
 726         list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
 727                 list_del(&dest->t_list);
 728                 ip_vs_dest_free(dest);
 729         }
 730 }
 731
 732 static void
 733 ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
 734 {
 735 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->ustats.c - src->ustats0.c
 736
 737         spin_lock_bh(&src->lock);
 738
 739         IP_VS_SHOW_STATS_COUNTER(conns);
 740         IP_VS_SHOW_STATS_COUNTER(inpkts);
 741         IP_VS_SHOW_STATS_COUNTER(outpkts);
 742         IP_VS_SHOW_STATS_COUNTER(inbytes);
 743         IP_VS_SHOW_STATS_COUNTER(outbytes);
 744
 745         ip_vs_read_estimator(dst, src);
 746
 747         spin_unlock_bh(&src->lock);
 748 }
 749
 750 static void
 751 ip_vs_zero_stats(struct ip_vs_stats *stats)
 752 {
 753         spin_lock_bh(&stats->lock);
 754
 755         /* get current counters as zero point, rates are zeroed */
 756
 757 #define IP_VS_ZERO_STATS_COUNTER(c) stats->ustats0.c = stats->ustats.c
 758
 759         IP_VS_ZERO_STATS_COUNTER(conns);
 760         IP_VS_ZERO_STATS_COUNTER(inpkts);
 761         IP_VS_ZERO_STATS_COUNTER(outpkts);
 762         IP_VS_ZERO_STATS_COUNTER(inbytes);
 763         IP_VS_ZERO_STATS_COUNTER(outbytes);
 764
 765         ip_vs_zero_estimator(stats);
 766
 767         spin_unlock_bh(&stats->lock);
 768 }
 769
 770 /*
 771  *      Update a destination in the given service
 772  */
 773 static void
 774 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
 775                     struct ip_vs_dest_user_kern *udest, int add)
 776 {
 777         struct netns_ipvs *ipvs = net_ipvs(svc->net);
 778         struct ip_vs_service *old_svc;
 779         struct ip_vs_scheduler *sched;
 780         int conn_flags;
 781
 782         /* We cannot modify an address and change the address family */
 783         BUG_ON(!add && udest->af != dest->af);
 784
 785         if (add && udest->af != svc->af)
 786                 ipvs->mixed_address_family_dests++;
 787
 788         /* set the weight and the flags */
 789         atomic_set(&dest->weight, udest->weight);
 790         conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
 791         conn_flags |= IP_VS_CONN_F_INACTIVE;
 792
 793         /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
 794         if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
 795                 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
 796         } else {
 797                 /*
 798                  *    Put the real service in rs_table if not present.
 799                  *    For now only for NAT!
 800                  */
 801                 ip_vs_rs_hash(ipvs, dest);
 802         }
 803         atomic_set(&dest->conn_flags, conn_flags);
 804
 805         /* bind the service */
 806         old_svc = rcu_dereference_protected(dest->svc, 1);
 807         if (!old_svc) {
 808                 __ip_vs_bind_svc(dest, svc);
 809         } else {
 810                 if (old_svc != svc) {
 811                         ip_vs_zero_stats(&dest->stats);
 812                         __ip_vs_bind_svc(dest, svc);
 813                         __ip_vs_svc_put(old_svc, true);
 814                 }
 815         }
 816
 817         /* set the dest status flags */
 818         dest->flags |= IP_VS_DEST_F_AVAILABLE;
 819
 820         if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
 821                 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
 822         dest->u_threshold = udest->u_threshold;
 823         dest->l_threshold = udest->l_threshold;
 824
 825         dest->af = udest->af;
 826
 827         spin_lock_bh(&dest->dst_lock);
 828         __ip_vs_dst_cache_reset(dest);
 829         spin_unlock_bh(&dest->dst_lock);
 830
 831         sched = rcu_dereference_protected(svc->scheduler, 1);
 832         if (add) {
 833                 ip_vs_start_estimator(svc->net, &dest->stats);
 834                 list_add_rcu(&dest->n_list, &svc->destinations);
 835                 svc->num_dests++;
 836                 if (sched->add_dest)
 837                         sched->add_dest(svc, dest);
 838         } else {
 839                 if (sched->upd_dest)
 840                         sched->upd_dest(svc, dest);
 841         }
 842 }
 843
 844
 845 /*
 846  *      Create a destination for the given service
 847  */
 848 static int
 849 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
 850                struct ip_vs_dest **dest_p)
 851 {
 852         struct ip_vs_dest *dest;
 853         unsigned int atype, i;
 854
 855         EnterFunction(2);
 856
 857 #ifdef CONFIG_IP_VS_IPV6
 858         if (udest->af == AF_INET6) {
 859                 atype = ipv6_addr_type(&udest->addr.in6);
 860                 if ((!(atype & IPV6_ADDR_UNICAST) ||
 861                         atype & IPV6_ADDR_LINKLOCAL) &&
 862                         !__ip_vs_addr_is_local_v6(svc->net, &udest->addr.in6))
 863                         return -EINVAL;
 864         } else
 865 #endif
 866         {
 867                 atype = inet_addr_type(svc->net, udest->addr.ip);
 868                 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
 869                         return -EINVAL;
 870         }
 871
 872         dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
 873         if (dest == NULL)
 874                 return -ENOMEM;
 875
 876         dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
 877         if (!dest->stats.cpustats)
 878                 goto err_alloc;
 879
 880         for_each_possible_cpu(i) {
 881                 struct ip_vs_cpu_stats *ip_vs_dest_stats;
 882                 ip_vs_dest_stats = per_cpu_ptr(dest->stats.cpustats, i);
 883                 u64_stats_init(&ip_vs_dest_stats->syncp);
 884         }
 885
 886         dest->af = udest->af;
 887         dest->protocol = svc->protocol;
 888         dest->vaddr = svc->addr;
 889         dest->vport = svc->port;
 890         dest->vfwmark = svc->fwmark;
 891         ip_vs_addr_copy(udest->af, &dest->addr, &udest->addr);
 892         dest->port = udest->port;
 893
 894         atomic_set(&dest->activeconns, 0);
 895         atomic_set(&dest->inactconns, 0);
 896         atomic_set(&dest->persistconns, 0);
 897         atomic_set(&dest->refcnt, 1);
 898
 899         INIT_HLIST_NODE(&dest->d_list);
 900         spin_lock_init(&dest->dst_lock);
 901         spin_lock_init(&dest->stats.lock);
 902         __ip_vs_update_dest(svc, dest, udest, 1);
 903
 904         *dest_p = dest;
 905
 906         LeaveFunction(2);
 907         return 0;
 908
 909 err_alloc:
 910         kfree(dest);
 911         return -ENOMEM;
 912 }
 913
 914
 915 /*
 916  *      Add a destination into an existing service
 917  */
 918 static int
 919 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
 920 {
 921         struct ip_vs_dest *dest;
 922         union nf_inet_addr daddr;
 923         __be16 dport = udest->port;
 924         int ret;
 925
 926         EnterFunction(2);
 927
 928         if (udest->weight < 0) {
 929                 pr_err("%s(): server weight less than zero\n", __func__);
 930                 return -ERANGE;
 931         }
 932
 933         if (udest->l_threshold > udest->u_threshold) {
 934                 pr_err("%s(): lower threshold is higher than upper threshold\n",
 935                         __func__);
 936                 return -ERANGE;
 937         }
 938
 939         ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
 940
 941         /* We use function that requires RCU lock */
 942         rcu_read_lock();
 943         dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
 944         rcu_read_unlock();
 945
 946         if (dest != NULL) {
 947                 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
 948                 return -EEXIST;
 949         }
 950
 951         /*
 952          * Check if the dest already exists in the trash and
 953          * is from the same service
 954          */
 955         dest = ip_vs_trash_get_dest(svc, udest->af, &daddr, dport);
 956
 957         if (dest != NULL) {
 958                 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
 959                               "dest->refcnt=%d, service %u/%s:%u\n",
 960                               IP_VS_DBG_ADDR(udest->af, &daddr), ntohs(dport),
 961                               atomic_read(&dest->refcnt),
 962                               dest->vfwmark,
 963                               IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
 964                               ntohs(dest->vport));
 965
 966                 __ip_vs_update_dest(svc, dest, udest, 1);
 967                 ret = 0;
 968         } else {
 969                 /*
 970                  * Allocate and initialize the dest structure
 971                  */
 972                 ret = ip_vs_new_dest(svc, udest, &dest);
 973         }
 974         LeaveFunction(2);
 975
 976         return ret;
 977 }
 978
 979
 980 /*
 981  *      Edit a destination in the given service
 982  */
 983 static int
 984 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
 985 {
 986         struct ip_vs_dest *dest;
 987         union nf_inet_addr daddr;
 988         __be16 dport = udest->port;
 989
 990         EnterFunction(2);
 991
 992         if (udest->weight < 0) {
 993                 pr_err("%s(): server weight less than zero\n", __func__);
 994                 return -ERANGE;
 995         }
 996
 997         if (udest->l_threshold > udest->u_threshold) {
 998                 pr_err("%s(): lower threshold is higher than upper threshold\n",
 999                         __func__);
1000                 return -ERANGE;
1001         }
1002
1003         ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1004
1005         /* We use function that requires RCU lock */
1006         rcu_read_lock();
1007         dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1008         rcu_read_unlock();
1009
1010         if (dest == NULL) {
1011                 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1012                 return -ENOENT;
1013         }
1014
1015         __ip_vs_update_dest(svc, dest, udest, 0);
1016         LeaveFunction(2);
1017
1018         return 0;
1019 }
1020
1021 /*
1022  *      Delete a destination (must be already unlinked from the service)
1023  */
1024 static void __ip_vs_del_dest(struct net *net, struct ip_vs_dest *dest,
1025                              bool cleanup)
1026 {
1027         struct netns_ipvs *ipvs = net_ipvs(net);
1028
1029         ip_vs_stop_estimator(net, &dest->stats);
1030
1031         /*
1032          *  Remove it from the d-linked list with the real services.
1033          */
1034         ip_vs_rs_unhash(dest);
1035
1036         spin_lock_bh(&ipvs->dest_trash_lock);
1037         IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1038                       IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1039                       atomic_read(&dest->refcnt));
1040         if (list_empty(&ipvs->dest_trash) && !cleanup)
1041                 mod_timer(&ipvs->dest_trash_timer,
1042                           jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1043         /* dest lives in trash without reference */
1044         list_add(&dest->t_list, &ipvs->dest_trash);
1045         dest->idle_start = 0;
1046         spin_unlock_bh(&ipvs->dest_trash_lock);
1047         ip_vs_dest_put(dest);
1048 }
1049
1050
1051 /*
1052  *      Unlink a destination from the given service
1053  */
1054 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1055                                 struct ip_vs_dest *dest,
1056                                 int svcupd)
1057 {
1058         dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1059
1060         /*
1061          *  Remove it from the d-linked destination list.
1062          */
1063         list_del_rcu(&dest->n_list);
1064         svc->num_dests--;
1065
1066         if (dest->af != svc->af)
1067                 net_ipvs(svc->net)->mixed_address_family_dests--;
1068
1069         if (svcupd) {
1070                 struct ip_vs_scheduler *sched;
1071
1072                 sched = rcu_dereference_protected(svc->scheduler, 1);
1073                 if (sched->del_dest)
1074                         sched->del_dest(svc, dest);
1075         }
1076 }
1077
1078
1079 /*
1080  *      Delete a destination server in the given service
1081  */
1082 static int
1083 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1084 {
1085         struct ip_vs_dest *dest;
1086         __be16 dport = udest->port;
1087
1088         EnterFunction(2);
1089
1090         /* We use function that requires RCU lock */
1091         rcu_read_lock();
1092         dest = ip_vs_lookup_dest(svc, udest->af, &udest->addr, dport);
1093         rcu_read_unlock();
1094
1095         if (dest == NULL) {
1096                 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1097                 return -ENOENT;
1098         }
1099
1100         /*
1101          *      Unlink dest from the service
1102          */
1103         __ip_vs_unlink_dest(svc, dest, 1);
1104
1105         /*
1106          *      Delete the destination
1107          */
1108         __ip_vs_del_dest(svc->net, dest, false);
1109
1110         LeaveFunction(2);
1111
1112         return 0;
1113 }
1114
1115 static void ip_vs_dest_trash_expire(unsigned long data)
1116 {
1117         struct net *net = (struct net *) data;
1118         struct netns_ipvs *ipvs = net_ipvs(net);
1119         struct ip_vs_dest *dest, *next;
1120         unsigned long now = jiffies;
1121
1122         spin_lock(&ipvs->dest_trash_lock);
1123         list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1124                 if (atomic_read(&dest->refcnt) > 0)
1125                         continue;
1126                 if (dest->idle_start) {
1127                         if (time_before(now, dest->idle_start +
1128                                              IP_VS_DEST_TRASH_PERIOD))
1129                                 continue;
1130                 } else {
1131                         dest->idle_start = max(1UL, now);
1132                         continue;
1133                 }
1134                 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1135                               dest->vfwmark,
1136                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1137                               ntohs(dest->port));
1138                 list_del(&dest->t_list);
1139                 ip_vs_dest_free(dest);
1140         }
1141         if (!list_empty(&ipvs->dest_trash))
1142                 mod_timer(&ipvs->dest_trash_timer,
1143                           jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1144         spin_unlock(&ipvs->dest_trash_lock);
1145 }
1146
1147 /*
1148  *      Add a service into the service hash table
1149  */
1150 static int
1151 ip_vs_add_service(struct net *net, struct ip_vs_service_user_kern *u,
1152                   struct ip_vs_service **svc_p)
1153 {
1154         int ret = 0, i;
1155         struct ip_vs_scheduler *sched = NULL;
1156         struct ip_vs_pe *pe = NULL;
1157         struct ip_vs_service *svc = NULL;
1158         struct netns_ipvs *ipvs = net_ipvs(net);
1159
1160         /* increase the module use count */
1161         ip_vs_use_count_inc();
1162
1163         /* Lookup the scheduler by 'u->sched_name' */
1164         sched = ip_vs_scheduler_get(u->sched_name);
1165         if (sched == NULL) {
1166                 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1167                 ret = -ENOENT;
1168                 goto out_err;
1169         }
1170
1171         if (u->pe_name && *u->pe_name) {
1172                 pe = ip_vs_pe_getbyname(u->pe_name);
1173                 if (pe == NULL) {
1174                         pr_info("persistence engine module ip_vs_pe_%s "
1175                                 "not found\n", u->pe_name);
1176                         ret = -ENOENT;
1177                         goto out_err;
1178                 }
1179         }
1180
1181 #ifdef CONFIG_IP_VS_IPV6
1182         if (u->af == AF_INET6) {
1183                 __u32 plen = (__force __u32) u->netmask;
1184
1185                 if (plen < 1 || plen > 128) {
1186                         ret = -EINVAL;
1187                         goto out_err;
1188                 }
1189         }
1190 #endif
1191
1192         svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1193         if (svc == NULL) {
1194                 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1195                 ret = -ENOMEM;
1196                 goto out_err;
1197         }
1198         svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1199         if (!svc->stats.cpustats) {
1200                 ret = -ENOMEM;
1201                 goto out_err;
1202         }
1203
1204         for_each_possible_cpu(i) {
1205                 struct ip_vs_cpu_stats *ip_vs_stats;
1206                 ip_vs_stats = per_cpu_ptr(svc->stats.cpustats, i);
1207                 u64_stats_init(&ip_vs_stats->syncp);
1208         }
1209
1210
1211         /* I'm the first user of the service */
1212         atomic_set(&svc->refcnt, 0);
1213
1214         svc->af = u->af;
1215         svc->protocol = u->protocol;
1216         ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1217         svc->port = u->port;
1218         svc->fwmark = u->fwmark;
1219         svc->flags = u->flags;
1220         svc->timeout = u->timeout * HZ;
1221         svc->netmask = u->netmask;
1222         svc->net = net;
1223
1224         INIT_LIST_HEAD(&svc->destinations);
1225         spin_lock_init(&svc->sched_lock);
1226         spin_lock_init(&svc->stats.lock);
1227
1228         /* Bind the scheduler */
1229         ret = ip_vs_bind_scheduler(svc, sched);
1230         if (ret)
1231                 goto out_err;
1232         sched = NULL;
1233
1234         /* Bind the ct retriever */
1235         RCU_INIT_POINTER(svc->pe, pe);
1236         pe = NULL;
1237
1238         /* Update the virtual service counters */
1239         if (svc->port == FTPPORT)
1240                 atomic_inc(&ipvs->ftpsvc_counter);
1241         else if (svc->port == 0)
1242                 atomic_inc(&ipvs->nullsvc_counter);
1243
1244         ip_vs_start_estimator(net, &svc->stats);
1245
1246         /* Count only IPv4 services for old get/setsockopt interface */
1247         if (svc->af == AF_INET)
1248                 ipvs->num_services++;
1249
1250         /* Hash the service into the service table */
1251         ip_vs_svc_hash(svc);
1252
1253         *svc_p = svc;
1254         /* Now there is a service - full throttle */
1255         ipvs->enable = 1;
1256         return 0;
1257
1258
1259  out_err:
1260         if (svc != NULL) {
1261                 ip_vs_unbind_scheduler(svc, sched);
1262                 ip_vs_service_free(svc);
1263         }
1264         ip_vs_scheduler_put(sched);
1265         ip_vs_pe_put(pe);
1266
1267         /* decrease the module use count */
1268         ip_vs_use_count_dec();
1269
1270         return ret;
1271 }
1272
1273
1274 /*
1275  *      Edit a service and bind it with a new scheduler
1276  */
1277 static int
1278 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1279 {
1280         struct ip_vs_scheduler *sched, *old_sched;
1281         struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1282         int ret = 0;
1283
1284         /*
1285          * Lookup the scheduler, by 'u->sched_name'
1286          */
1287         sched = ip_vs_scheduler_get(u->sched_name);
1288         if (sched == NULL) {
1289                 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1290                 return -ENOENT;
1291         }
1292         old_sched = sched;
1293
1294         if (u->pe_name && *u->pe_name) {
1295                 pe = ip_vs_pe_getbyname(u->pe_name);
1296                 if (pe == NULL) {
1297                         pr_info("persistence engine module ip_vs_pe_%s "
1298                                 "not found\n", u->pe_name);
1299                         ret = -ENOENT;
1300                         goto out;
1301                 }
1302                 old_pe = pe;
1303         }
1304
1305 #ifdef CONFIG_IP_VS_IPV6
1306         if (u->af == AF_INET6) {
1307                 __u32 plen = (__force __u32) u->netmask;
1308
1309                 if (plen < 1 || plen > 128) {
1310                         ret = -EINVAL;
1311                         goto out;
1312                 }
1313         }
1314 #endif
1315
1316         old_sched = rcu_dereference_protected(svc->scheduler, 1);
1317         if (sched != old_sched) {
1318                 /* Bind the new scheduler */
1319                 ret = ip_vs_bind_scheduler(svc, sched);
1320                 if (ret) {
1321                         old_sched = sched;
1322                         goto out;
1323                 }
1324                 /* Unbind the old scheduler on success */
1325                 ip_vs_unbind_scheduler(svc, old_sched);
1326         }
1327
1328         /*
1329          * Set the flags and timeout value
1330          */
1331         svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1332         svc->timeout = u->timeout * HZ;
1333         svc->netmask = u->netmask;
1334
1335         old_pe = rcu_dereference_protected(svc->pe, 1);
1336         if (pe != old_pe)
1337                 rcu_assign_pointer(svc->pe, pe);
1338
1339 out:
1340         ip_vs_scheduler_put(old_sched);
1341         ip_vs_pe_put(old_pe);
1342         return ret;
1343 }
1344
1345 /*
1346  *      Delete a service from the service list
1347  *      - The service must be unlinked, unlocked and not referenced!
1348  *      - We are called under _bh lock
1349  */
1350 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1351 {
1352         struct ip_vs_dest *dest, *nxt;
1353         struct ip_vs_scheduler *old_sched;
1354         struct ip_vs_pe *old_pe;
1355         struct netns_ipvs *ipvs = net_ipvs(svc->net);
1356
1357         pr_info("%s: enter\n", __func__);
1358
1359         /* Count only IPv4 services for old get/setsockopt interface */
1360         if (svc->af == AF_INET)
1361                 ipvs->num_services--;
1362
1363         ip_vs_stop_estimator(svc->net, &svc->stats);
1364
1365         /* Unbind scheduler */
1366         old_sched = rcu_dereference_protected(svc->scheduler, 1);
1367         ip_vs_unbind_scheduler(svc, old_sched);
1368         ip_vs_scheduler_put(old_sched);
1369
1370         /* Unbind persistence engine, keep svc->pe */
1371         old_pe = rcu_dereference_protected(svc->pe, 1);
1372         ip_vs_pe_put(old_pe);
1373
1374         /*
1375          *    Unlink the whole destination list
1376          */
1377         list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1378                 __ip_vs_unlink_dest(svc, dest, 0);
1379                 __ip_vs_del_dest(svc->net, dest, cleanup);
1380         }
1381
1382         /*
1383          *    Update the virtual service counters
1384          */
1385         if (svc->port == FTPPORT)
1386                 atomic_dec(&ipvs->ftpsvc_counter);
1387         else if (svc->port == 0)
1388                 atomic_dec(&ipvs->nullsvc_counter);
1389
1390         /*
1391          *    Free the service if nobody refers to it
1392          */
1393         __ip_vs_svc_put(svc, true);
1394
1395         /* decrease the module use count */
1396         ip_vs_use_count_dec();
1397 }
1398
1399 /*
1400  * Unlink a service from list and try to delete it if its refcnt reached 0
1401  */
1402 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1403 {
1404         /* Hold svc to avoid double release from dest_trash */
1405         atomic_inc(&svc->refcnt);
1406         /*
1407          * Unhash it from the service table
1408          */
1409         ip_vs_svc_unhash(svc);
1410
1411         __ip_vs_del_service(svc, cleanup);
1412 }
1413
1414 /*
1415  *      Delete a service from the service list
1416  */
1417 static int ip_vs_del_service(struct ip_vs_service *svc)
1418 {
1419         if (svc == NULL)
1420                 return -EEXIST;
1421         ip_vs_unlink_service(svc, false);
1422
1423         return 0;
1424 }
1425
1426
1427 /*
1428  *      Flush all the virtual services
1429  */
1430 static int ip_vs_flush(struct net *net, bool cleanup)
1431 {
1432         int idx;
1433         struct ip_vs_service *svc;
1434         struct hlist_node *n;
1435
1436         /*
1437          * Flush the service table hashed by <netns,protocol,addr,port>
1438          */
1439         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1440                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1441                                           s_list) {
1442                         if (net_eq(svc->net, net))
1443                                 ip_vs_unlink_service(svc, cleanup);
1444                 }
1445         }
1446
1447         /*
1448          * Flush the service table hashed by fwmark
1449          */
1450         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1451                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1452                                           f_list) {
1453                         if (net_eq(svc->net, net))
1454                                 ip_vs_unlink_service(svc, cleanup);
1455                 }
1456         }
1457
1458         return 0;
1459 }
1460
1461 /*
1462  *      Delete service by {netns} in the service table.
1463  *      Called by __ip_vs_cleanup()
1464  */
1465 void ip_vs_service_net_cleanup(struct net *net)
1466 {
1467         EnterFunction(2);
1468         /* Check for "full" addressed entries */
1469         mutex_lock(&__ip_vs_mutex);
1470         ip_vs_flush(net, true);
1471         mutex_unlock(&__ip_vs_mutex);
1472         LeaveFunction(2);
1473 }
1474
1475 /* Put all references for device (dst_cache) */
1476 static inline void
1477 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1478 {
1479         struct ip_vs_dest_dst *dest_dst;
1480
1481         spin_lock_bh(&dest->dst_lock);
1482         dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1483         if (dest_dst && dest_dst->dst_cache->dev == dev) {
1484                 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1485                               dev->name,
1486                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1487                               ntohs(dest->port),
1488                               atomic_read(&dest->refcnt));
1489                 __ip_vs_dst_cache_reset(dest);
1490         }
1491         spin_unlock_bh(&dest->dst_lock);
1492
1493 }
1494 /* Netdev event receiver
1495  * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1496  */
1497 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1498                            void *ptr)
1499 {
1500         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1501         struct net *net = dev_net(dev);
1502         struct netns_ipvs *ipvs = net_ipvs(net);
1503         struct ip_vs_service *svc;
1504         struct ip_vs_dest *dest;
1505         unsigned int idx;
1506
1507         if (event != NETDEV_DOWN || !ipvs)
1508                 return NOTIFY_DONE;
1509         IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1510         EnterFunction(2);
1511         mutex_lock(&__ip_vs_mutex);
1512         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1513                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1514                         if (net_eq(svc->net, net)) {
1515                                 list_for_each_entry(dest, &svc->destinations,
1516                                                     n_list) {
1517                                         ip_vs_forget_dev(dest, dev);
1518                                 }
1519                         }
1520                 }
1521
1522                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1523                         if (net_eq(svc->net, net)) {
1524                                 list_for_each_entry(dest, &svc->destinations,
1525                                                     n_list) {
1526                                         ip_vs_forget_dev(dest, dev);
1527                                 }
1528                         }
1529
1530                 }
1531         }
1532
1533         spin_lock_bh(&ipvs->dest_trash_lock);
1534         list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1535                 ip_vs_forget_dev(dest, dev);
1536         }
1537         spin_unlock_bh(&ipvs->dest_trash_lock);
1538         mutex_unlock(&__ip_vs_mutex);
1539         LeaveFunction(2);
1540         return NOTIFY_DONE;
1541 }
1542
1543 /*
1544  *      Zero counters in a service or all services
1545  */
1546 static int ip_vs_zero_service(struct ip_vs_service *svc)
1547 {
1548         struct ip_vs_dest *dest;
1549
1550         list_for_each_entry(dest, &svc->destinations, n_list) {
1551                 ip_vs_zero_stats(&dest->stats);
1552         }
1553         ip_vs_zero_stats(&svc->stats);
1554         return 0;
1555 }
1556
1557 static int ip_vs_zero_all(struct net *net)
1558 {
1559         int idx;
1560         struct ip_vs_service *svc;
1561
1562         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1563                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1564                         if (net_eq(svc->net, net))
1565                                 ip_vs_zero_service(svc);
1566                 }
1567         }
1568
1569         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1570                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1571                         if (net_eq(svc->net, net))
1572                                 ip_vs_zero_service(svc);
1573                 }
1574         }
1575
1576         ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1577         return 0;
1578 }
1579
1580 #ifdef CONFIG_SYSCTL
1581
1582 static int zero;
1583 static int three = 3;
1584
1585 static int
1586 proc_do_defense_mode(struct ctl_table *table, int write,
1587                      void __user *buffer, size_t *lenp, loff_t *ppos)
1588 {
1589         struct net *net = current->nsproxy->net_ns;
1590         int *valp = table->data;
1591         int val = *valp;
1592         int rc;
1593
1594         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1595         if (write && (*valp != val)) {
1596                 if ((*valp < 0) || (*valp > 3)) {
1597                         /* Restore the correct value */
1598                         *valp = val;
1599                 } else {
1600                         update_defense_level(net_ipvs(net));
1601                 }
1602         }
1603         return rc;
1604 }
1605
1606 static int
1607 proc_do_sync_threshold(struct ctl_table *table, int write,
1608                        void __user *buffer, size_t *lenp, loff_t *ppos)
1609 {
1610         int *valp = table->data;
1611         int val[2];
1612         int rc;
1613
1614         /* backup the value first */
1615         memcpy(val, valp, sizeof(val));
1616
1617         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1618         if (write && (valp[0] < 0 || valp[1] < 0 ||
1619             (valp[0] >= valp[1] && valp[1]))) {
1620                 /* Restore the correct value */
1621                 memcpy(valp, val, sizeof(val));
1622         }
1623         return rc;
1624 }
1625
1626 static int
1627 proc_do_sync_mode(struct ctl_table *table, int write,
1628                      void __user *buffer, size_t *lenp, loff_t *ppos)
1629 {
1630         int *valp = table->data;
1631         int val = *valp;
1632         int rc;
1633
1634         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1635         if (write && (*valp != val)) {
1636                 if ((*valp < 0) || (*valp > 1)) {
1637                         /* Restore the correct value */
1638                         *valp = val;
1639                 }
1640         }
1641         return rc;
1642 }
1643
1644 static int
1645 proc_do_sync_ports(struct ctl_table *table, int write,
1646                    void __user *buffer, size_t *lenp, loff_t *ppos)
1647 {
1648         int *valp = table->data;
1649         int val = *valp;
1650         int rc;
1651
1652         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1653         if (write && (*valp != val)) {
1654                 if (*valp < 1 || !is_power_of_2(*valp)) {
1655                         /* Restore the correct value */
1656                         *valp = val;
1657                 }
1658         }
1659         return rc;
1660 }
1661
1662 /*
1663  *      IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1664  *      Do not change order or insert new entries without
1665  *      align with netns init in ip_vs_control_net_init()
1666  */
1667
1668 static struct ctl_table vs_vars[] = {
1669         {
1670                 .procname       = "amemthresh",
1671                 .maxlen         = sizeof(int),
1672                 .mode           = 0644,
1673                 .proc_handler   = proc_dointvec,
1674         },
1675         {
1676                 .procname       = "am_droprate",
1677                 .maxlen         = sizeof(int),
1678                 .mode           = 0644,
1679                 .proc_handler   = proc_dointvec,
1680         },
1681         {
1682                 .procname       = "drop_entry",
1683                 .maxlen         = sizeof(int),
1684                 .mode           = 0644,
1685                 .proc_handler   = proc_do_defense_mode,
1686         },
1687         {
1688                 .procname       = "drop_packet",
1689                 .maxlen         = sizeof(int),
1690                 .mode           = 0644,
1691                 .proc_handler   = proc_do_defense_mode,
1692         },
1693 #ifdef CONFIG_IP_VS_NFCT
1694         {
1695                 .procname       = "conntrack",
1696                 .maxlen         = sizeof(int),
1697                 .mode           = 0644,
1698                 .proc_handler   = &proc_dointvec,
1699         },
1700 #endif
1701         {
1702                 .procname       = "secure_tcp",
1703                 .maxlen         = sizeof(int),
1704                 .mode           = 0644,
1705                 .proc_handler   = proc_do_defense_mode,
1706         },
1707         {
1708                 .procname       = "snat_reroute",
1709                 .maxlen         = sizeof(int),
1710                 .mode           = 0644,
1711                 .proc_handler   = &proc_dointvec,
1712         },
1713         {
1714                 .procname       = "sync_version",
1715                 .maxlen         = sizeof(int),
1716                 .mode           = 0644,
1717                 .proc_handler   = &proc_do_sync_mode,
1718         },
1719         {
1720                 .procname       = "sync_ports",
1721                 .maxlen         = sizeof(int),
1722                 .mode           = 0644,
1723                 .proc_handler   = &proc_do_sync_ports,
1724         },
1725         {
1726                 .procname       = "sync_persist_mode",
1727                 .maxlen         = sizeof(int),
1728                 .mode           = 0644,
1729                 .proc_handler   = proc_dointvec,
1730         },
1731         {
1732                 .procname       = "sync_qlen_max",
1733                 .maxlen         = sizeof(unsigned long),
1734                 .mode           = 0644,
1735                 .proc_handler   = proc_doulongvec_minmax,
1736         },
1737         {
1738                 .procname       = "sync_sock_size",
1739                 .maxlen         = sizeof(int),
1740                 .mode           = 0644,
1741                 .proc_handler   = proc_dointvec,
1742         },
1743         {
1744                 .procname       = "cache_bypass",
1745                 .maxlen         = sizeof(int),
1746                 .mode           = 0644,
1747                 .proc_handler   = proc_dointvec,
1748         },
1749         {
1750                 .procname       = "expire_nodest_conn",
1751                 .maxlen         = sizeof(int),
1752                 .mode           = 0644,
1753                 .proc_handler   = proc_dointvec,
1754         },
1755         {
1756                 .procname       = "sloppy_tcp",
1757                 .maxlen         = sizeof(int),
1758                 .mode           = 0644,
1759                 .proc_handler   = proc_dointvec,
1760         },
1761         {
1762                 .procname       = "sloppy_sctp",
1763                 .maxlen         = sizeof(int),
1764                 .mode           = 0644,
1765                 .proc_handler   = proc_dointvec,
1766         },
1767         {
1768                 .procname       = "expire_quiescent_template",
1769                 .maxlen         = sizeof(int),
1770                 .mode           = 0644,
1771                 .proc_handler   = proc_dointvec,
1772         },
1773         {
1774                 .procname       = "sync_threshold",
1775                 .maxlen         =
1776                         sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1777                 .mode           = 0644,
1778                 .proc_handler   = proc_do_sync_threshold,
1779         },
1780         {
1781                 .procname       = "sync_refresh_period",
1782                 .maxlen         = sizeof(int),
1783                 .mode           = 0644,
1784                 .proc_handler   = proc_dointvec_jiffies,
1785         },
1786         {
1787                 .procname       = "sync_retries",
1788                 .maxlen         = sizeof(int),
1789                 .mode           = 0644,
1790                 .proc_handler   = proc_dointvec_minmax,
1791                 .extra1         = &zero,
1792                 .extra2         = &three,
1793         },
1794         {
1795                 .procname       = "nat_icmp_send",
1796                 .maxlen         = sizeof(int),
1797                 .mode           = 0644,
1798                 .proc_handler   = proc_dointvec,
1799         },
1800         {
1801                 .procname       = "pmtu_disc",
1802                 .maxlen         = sizeof(int),
1803                 .mode           = 0644,
1804                 .proc_handler   = proc_dointvec,
1805         },
1806         {
1807                 .procname       = "backup_only",
1808                 .maxlen         = sizeof(int),
1809                 .mode           = 0644,
1810                 .proc_handler   = proc_dointvec,
1811         },
1812 #ifdef CONFIG_IP_VS_DEBUG
1813         {
1814                 .procname       = "debug_level",
1815                 .data           = &sysctl_ip_vs_debug_level,
1816                 .maxlen         = sizeof(int),
1817                 .mode           = 0644,
1818                 .proc_handler   = proc_dointvec,
1819         },
1820 #endif
1821         { }
1822 };
1823
1824 #endif
1825
1826 #ifdef CONFIG_PROC_FS
1827
1828 struct ip_vs_iter {
1829         struct seq_net_private p;  /* Do not move this, netns depends upon it*/
1830         struct hlist_head *table;
1831         int bucket;
1832 };
1833
1834 /*
1835  *      Write the contents of the VS rule table to a PROCfs file.
1836  *      (It is kept just for backward compatibility)
1837  */
1838 static inline const char *ip_vs_fwd_name(unsigned int flags)
1839 {
1840         switch (flags & IP_VS_CONN_F_FWD_MASK) {
1841         case IP_VS_CONN_F_LOCALNODE:
1842                 return "Local";
1843         case IP_VS_CONN_F_TUNNEL:
1844                 return "Tunnel";
1845         case IP_VS_CONN_F_DROUTE:
1846                 return "Route";
1847         default:
1848                 return "Masq";
1849         }
1850 }
1851
1852
1853 /* Get the Nth entry in the two lists */
1854 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1855 {
1856         struct net *net = seq_file_net(seq);
1857         struct ip_vs_iter *iter = seq->private;
1858         int idx;
1859         struct ip_vs_service *svc;
1860
1861         /* look in hash by protocol */
1862         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1863                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1864                         if (net_eq(svc->net, net) && pos-- == 0) {
1865                                 iter->table = ip_vs_svc_table;
1866                                 iter->bucket = idx;
1867                                 return svc;
1868                         }
1869                 }
1870         }
1871
1872         /* keep looking in fwmark */
1873         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1874                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1875                                          f_list) {
1876                         if (net_eq(svc->net, net) && pos-- == 0) {
1877                                 iter->table = ip_vs_svc_fwm_table;
1878                                 iter->bucket = idx;
1879                                 return svc;
1880                         }
1881                 }
1882         }
1883
1884         return NULL;
1885 }
1886
1887 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1888         __acquires(RCU)
1889 {
1890         rcu_read_lock();
1891         return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1892 }
1893
1894
1895 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1896 {
1897         struct hlist_node *e;
1898         struct ip_vs_iter *iter;
1899         struct ip_vs_service *svc;
1900
1901         ++*pos;
1902         if (v == SEQ_START_TOKEN)
1903                 return ip_vs_info_array(seq,0);
1904
1905         svc = v;
1906         iter = seq->private;
1907
1908         if (iter->table == ip_vs_svc_table) {
1909                 /* next service in table hashed by protocol */
1910                 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
1911                 if (e)
1912                         return hlist_entry(e, struct ip_vs_service, s_list);
1913
1914                 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1915                         hlist_for_each_entry_rcu(svc,
1916                                                  &ip_vs_svc_table[iter->bucket],
1917                                                  s_list) {
1918                                 return svc;
1919                         }
1920                 }
1921
1922                 iter->table = ip_vs_svc_fwm_table;
1923                 iter->bucket = -1;
1924                 goto scan_fwmark;
1925         }
1926
1927         /* next service in hashed by fwmark */
1928         e = rcu_dereference(hlist_next_rcu(&svc->f_list));
1929         if (e)
1930                 return hlist_entry(e, struct ip_vs_service, f_list);
1931
1932  scan_fwmark:
1933         while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1934                 hlist_for_each_entry_rcu(svc,
1935                                          &ip_vs_svc_fwm_table[iter->bucket],
1936                                          f_list)
1937                         return svc;
1938         }
1939
1940         return NULL;
1941 }
1942
1943 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1944         __releases(RCU)
1945 {
1946         rcu_read_unlock();
1947 }
1948
1949
1950 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1951 {
1952         if (v == SEQ_START_TOKEN) {
1953                 seq_printf(seq,
1954                         "IP Virtual Server version %d.%d.%d (size=%d)\n",
1955                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1956                 seq_puts(seq,
1957                          "Prot LocalAddress:Port Scheduler Flags\n");
1958                 seq_puts(seq,
1959                          "  -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1960         } else {
1961                 const struct ip_vs_service *svc = v;
1962                 const struct ip_vs_iter *iter = seq->private;
1963                 const struct ip_vs_dest *dest;
1964                 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
1965
1966                 if (iter->table == ip_vs_svc_table) {
1967 #ifdef CONFIG_IP_VS_IPV6
1968                         if (svc->af == AF_INET6)
1969                                 seq_printf(seq, "%s  [%pI6]:%04X %s ",
1970                                            ip_vs_proto_name(svc->protocol),
1971                                            &svc->addr.in6,
1972                                            ntohs(svc->port),
1973                                            sched->name);
1974                         else
1975 #endif
1976                                 seq_printf(seq, "%s  %08X:%04X %s %s ",
1977                                            ip_vs_proto_name(svc->protocol),
1978                                            ntohl(svc->addr.ip),
1979                                            ntohs(svc->port),
1980                                            sched->name,
1981                                            (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
1982                 } else {
1983                         seq_printf(seq, "FWM  %08X %s %s",
1984                                    svc->fwmark, sched->name,
1985                                    (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
1986                 }
1987
1988                 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
1989                         seq_printf(seq, "persistent %d %08X\n",
1990                                 svc->timeout,
1991                                 ntohl(svc->netmask));
1992                 else
1993                         seq_putc(seq, '\n');
1994
1995                 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
1996 #ifdef CONFIG_IP_VS_IPV6
1997                         if (dest->af == AF_INET6)
1998                                 seq_printf(seq,
1999                                            "  -> [%pI6]:%04X"
2000                                            "      %-7s %-6d %-10d %-10d\n",
2001                                            &dest->addr.in6,
2002                                            ntohs(dest->port),
2003                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2004                                            atomic_read(&dest->weight),
2005                                            atomic_read(&dest->activeconns),
2006                                            atomic_read(&dest->inactconns));
2007                         else
2008 #endif
2009                                 seq_printf(seq,
2010                                            "  -> %08X:%04X      "
2011                                            "%-7s %-6d %-10d %-10d\n",
2012                                            ntohl(dest->addr.ip),
2013                                            ntohs(dest->port),
2014                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2015                                            atomic_read(&dest->weight),
2016                                            atomic_read(&dest->activeconns),
2017                                            atomic_read(&dest->inactconns));
2018
2019                 }
2020         }
2021         return 0;
2022 }
2023
2024 static const struct seq_operations ip_vs_info_seq_ops = {
2025         .start = ip_vs_info_seq_start,
2026         .next  = ip_vs_info_seq_next,
2027         .stop  = ip_vs_info_seq_stop,
2028         .show  = ip_vs_info_seq_show,
2029 };
2030
2031 static int ip_vs_info_open(struct inode *inode, struct file *file)
2032 {
2033         return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2034                         sizeof(struct ip_vs_iter));
2035 }
2036
2037 static const struct file_operations ip_vs_info_fops = {
2038         .owner   = THIS_MODULE,
2039         .open    = ip_vs_info_open,
2040         .read    = seq_read,
2041         .llseek  = seq_lseek,
2042         .release = seq_release_net,
2043 };
2044
2045 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2046 {
2047         struct net *net = seq_file_single_net(seq);
2048         struct ip_vs_stats_user show;
2049
2050 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2051         seq_puts(seq,
2052                  "   Total Incoming Outgoing         Incoming         Outgoing\n");
2053         seq_printf(seq,
2054                    "   Conns  Packets  Packets            Bytes            Bytes\n");
2055
2056         ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2057         seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", show.conns,
2058                    show.inpkts, show.outpkts,
2059                    (unsigned long long) show.inbytes,
2060                    (unsigned long long) show.outbytes);
2061
2062 /*                 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2063         seq_puts(seq,
2064                    " Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2065         seq_printf(seq, "%8X %8X %8X %16X %16X\n",
2066                         show.cps, show.inpps, show.outpps,
2067                         show.inbps, show.outbps);
2068
2069         return 0;
2070 }
2071
2072 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2073 {
2074         return single_open_net(inode, file, ip_vs_stats_show);
2075 }
2076
2077 static const struct file_operations ip_vs_stats_fops = {
2078         .owner = THIS_MODULE,
2079         .open = ip_vs_stats_seq_open,
2080         .read = seq_read,
2081         .llseek = seq_lseek,
2082         .release = single_release_net,
2083 };
2084
2085 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2086 {
2087         struct net *net = seq_file_single_net(seq);
2088         struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2089         struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2090         struct ip_vs_stats_user rates;
2091         int i;
2092
2093 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2094         seq_puts(seq,
2095                  "       Total Incoming Outgoing         Incoming         Outgoing\n");
2096         seq_printf(seq,
2097                    "CPU    Conns  Packets  Packets            Bytes            Bytes\n");
2098
2099         for_each_possible_cpu(i) {
2100                 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2101                 unsigned int start;
2102                 __u64 inbytes, outbytes;
2103
2104                 do {
2105                         start = u64_stats_fetch_begin_irq(&u->syncp);
2106                         inbytes = u->ustats.inbytes;
2107                         outbytes = u->ustats.outbytes;
2108                 } while (u64_stats_fetch_retry_irq(&u->syncp, start));
2109
2110                 seq_printf(seq, "%3X %8X %8X %8X %16LX %16LX\n",
2111                            i, u->ustats.conns, u->ustats.inpkts,
2112                            u->ustats.outpkts, (__u64)inbytes,
2113                            (__u64)outbytes);
2114         }
2115
2116         spin_lock_bh(&tot_stats->lock);
2117
2118         seq_printf(seq, "  ~ %8X %8X %8X %16LX %16LX\n\n",
2119                    tot_stats->ustats.conns, tot_stats->ustats.inpkts,
2120                    tot_stats->ustats.outpkts,
2121                    (unsigned long long) tot_stats->ustats.inbytes,
2122                    (unsigned long long) tot_stats->ustats.outbytes);
2123
2124         ip_vs_read_estimator(&rates, tot_stats);
2125
2126         spin_unlock_bh(&tot_stats->lock);
2127
2128 /*                 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2129         seq_puts(seq,
2130                    "     Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2131         seq_printf(seq, "    %8X %8X %8X %16X %16X\n",
2132                         rates.cps,
2133                         rates.inpps,
2134                         rates.outpps,
2135                         rates.inbps,
2136                         rates.outbps);
2137
2138         return 0;
2139 }
2140
2141 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2142 {
2143         return single_open_net(inode, file, ip_vs_stats_percpu_show);
2144 }
2145
2146 static const struct file_operations ip_vs_stats_percpu_fops = {
2147         .owner = THIS_MODULE,
2148         .open = ip_vs_stats_percpu_seq_open,
2149         .read = seq_read,
2150         .llseek = seq_lseek,
2151         .release = single_release_net,
2152 };
2153 #endif
2154
2155 /*
2156  *      Set timeout values for tcp tcpfin udp in the timeout_table.
2157  */
2158 static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
2159 {
2160 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2161         struct ip_vs_proto_data *pd;
2162 #endif
2163
2164         IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2165                   u->tcp_timeout,
2166                   u->tcp_fin_timeout,
2167                   u->udp_timeout);
2168
2169 #ifdef CONFIG_IP_VS_PROTO_TCP
2170         if (u->tcp_timeout) {
2171                 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2172                 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2173                         = u->tcp_timeout * HZ;
2174         }
2175
2176         if (u->tcp_fin_timeout) {
2177                 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2178                 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2179                         = u->tcp_fin_timeout * HZ;
2180         }
2181 #endif
2182
2183 #ifdef CONFIG_IP_VS_PROTO_UDP
2184         if (u->udp_timeout) {
2185                 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2186                 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2187                         = u->udp_timeout * HZ;
2188         }
2189 #endif
2190         return 0;
2191 }
2192
2193 #define CMDID(cmd)              (cmd - IP_VS_BASE_CTL)
2194
2195 struct ip_vs_svcdest_user {
2196         struct ip_vs_service_user       s;
2197         struct ip_vs_dest_user          d;
2198 };
2199
2200 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2201         [CMDID(IP_VS_SO_SET_ADD)]         = sizeof(struct ip_vs_service_user),
2202         [CMDID(IP_VS_SO_SET_EDIT)]        = sizeof(struct ip_vs_service_user),
2203         [CMDID(IP_VS_SO_SET_DEL)]         = sizeof(struct ip_vs_service_user),
2204         [CMDID(IP_VS_SO_SET_ADDDEST)]     = sizeof(struct ip_vs_svcdest_user),
2205         [CMDID(IP_VS_SO_SET_DELDEST)]     = sizeof(struct ip_vs_svcdest_user),
2206         [CMDID(IP_VS_SO_SET_EDITDEST)]    = sizeof(struct ip_vs_svcdest_user),
2207         [CMDID(IP_VS_SO_SET_TIMEOUT)]     = sizeof(struct ip_vs_timeout_user),
2208         [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2209         [CMDID(IP_VS_SO_SET_STOPDAEMON)]  = sizeof(struct ip_vs_daemon_user),
2210         [CMDID(IP_VS_SO_SET_ZERO)]        = sizeof(struct ip_vs_service_user),
2211 };
2212
2213 union ip_vs_set_arglen {
2214         struct ip_vs_service_user       field_IP_VS_SO_SET_ADD;
2215         struct ip_vs_service_user       field_IP_VS_SO_SET_EDIT;
2216         struct ip_vs_service_user       field_IP_VS_SO_SET_DEL;
2217         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_ADDDEST;
2218         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_DELDEST;
2219         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_EDITDEST;
2220         struct ip_vs_timeout_user       field_IP_VS_SO_SET_TIMEOUT;
2221         struct ip_vs_daemon_user        field_IP_VS_SO_SET_STARTDAEMON;
2222         struct ip_vs_daemon_user        field_IP_VS_SO_SET_STOPDAEMON;
2223         struct ip_vs_service_user       field_IP_VS_SO_SET_ZERO;
2224 };
2225
2226 #define MAX_SET_ARGLEN  sizeof(union ip_vs_set_arglen)
2227
2228 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2229                                   struct ip_vs_service_user *usvc_compat)
2230 {
2231         memset(usvc, 0, sizeof(*usvc));
2232
2233         usvc->af                = AF_INET;
2234         usvc->protocol          = usvc_compat->protocol;
2235         usvc->addr.ip           = usvc_compat->addr;
2236         usvc->port              = usvc_compat->port;
2237         usvc->fwmark            = usvc_compat->fwmark;
2238
2239         /* Deep copy of sched_name is not needed here */
2240         usvc->sched_name        = usvc_compat->sched_name;
2241
2242         usvc->flags             = usvc_compat->flags;
2243         usvc->timeout           = usvc_compat->timeout;
2244         usvc->netmask           = usvc_compat->netmask;
2245 }
2246
2247 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2248                                    struct ip_vs_dest_user *udest_compat)
2249 {
2250         memset(udest, 0, sizeof(*udest));
2251
2252         udest->addr.ip          = udest_compat->addr;
2253         udest->port             = udest_compat->port;
2254         udest->conn_flags       = udest_compat->conn_flags;
2255         udest->weight           = udest_compat->weight;
2256         udest->u_threshold      = udest_compat->u_threshold;
2257         udest->l_threshold      = udest_compat->l_threshold;
2258         udest->af               = AF_INET;
2259 }
2260
2261 static int
2262 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2263 {
2264         struct net *net = sock_net(sk);
2265         int ret;
2266         unsigned char arg[MAX_SET_ARGLEN];
2267         struct ip_vs_service_user *usvc_compat;
2268         struct ip_vs_service_user_kern usvc;
2269         struct ip_vs_service *svc;
2270         struct ip_vs_dest_user *udest_compat;
2271         struct ip_vs_dest_user_kern udest;
2272         struct netns_ipvs *ipvs = net_ipvs(net);
2273
2274         BUILD_BUG_ON(sizeof(arg) > 255);
2275         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2276                 return -EPERM;
2277
2278         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2279                 return -EINVAL;
2280         if (len != set_arglen[CMDID(cmd)]) {
2281                 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2282                           len, set_arglen[CMDID(cmd)]);
2283                 return -EINVAL;
2284         }
2285
2286         if (copy_from_user(arg, user, len) != 0)
2287                 return -EFAULT;
2288
2289         /* increase the module use count */
2290         ip_vs_use_count_inc();
2291
2292         /* Handle daemons since they have another lock */
2293         if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2294             cmd == IP_VS_SO_SET_STOPDAEMON) {
2295                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2296
2297                 mutex_lock(&ipvs->sync_mutex);
2298                 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2299                         ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2300                                                 dm->syncid);
2301                 else
2302                         ret = stop_sync_thread(net, dm->state);
2303                 mutex_unlock(&ipvs->sync_mutex);
2304                 goto out_dec;
2305         }
2306
2307         mutex_lock(&__ip_vs_mutex);
2308         if (cmd == IP_VS_SO_SET_FLUSH) {
2309                 /* Flush the virtual service */
2310                 ret = ip_vs_flush(net, false);
2311                 goto out_unlock;
2312         } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2313                 /* Set timeout values for (tcp tcpfin udp) */
2314                 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
2315                 goto out_unlock;
2316         }
2317
2318         usvc_compat = (struct ip_vs_service_user *)arg;
2319         udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2320
2321         /* We only use the new structs internally, so copy userspace compat
2322          * structs to extended internal versions */
2323         ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2324         ip_vs_copy_udest_compat(&udest, udest_compat);
2325
2326         if (cmd == IP_VS_SO_SET_ZERO) {
2327                 /* if no service address is set, zero counters in all */
2328                 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2329                         ret = ip_vs_zero_all(net);
2330                         goto out_unlock;
2331                 }
2332         }
2333
2334         /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2335         if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2336             usvc.protocol != IPPROTO_SCTP) {
2337                 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2338                        usvc.protocol, &usvc.addr.ip,
2339                        ntohs(usvc.port), usvc.sched_name);
2340                 ret = -EFAULT;
2341                 goto out_unlock;
2342         }
2343
2344         /* Lookup the exact service by <protocol, addr, port> or fwmark */
2345         rcu_read_lock();
2346         if (usvc.fwmark == 0)
2347                 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
2348                                            &usvc.addr, usvc.port);
2349         else
2350                 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
2351         rcu_read_unlock();
2352
2353         if (cmd != IP_VS_SO_SET_ADD
2354             && (svc == NULL || svc->protocol != usvc.protocol)) {
2355                 ret = -ESRCH;
2356                 goto out_unlock;
2357         }
2358
2359         switch (cmd) {
2360         case IP_VS_SO_SET_ADD:
2361                 if (svc != NULL)
2362                         ret = -EEXIST;
2363                 else
2364                         ret = ip_vs_add_service(net, &usvc, &svc);
2365                 break;
2366         case IP_VS_SO_SET_EDIT:
2367                 ret = ip_vs_edit_service(svc, &usvc);
2368                 break;
2369         case IP_VS_SO_SET_DEL:
2370                 ret = ip_vs_del_service(svc);
2371                 if (!ret)
2372                         goto out_unlock;
2373                 break;
2374         case IP_VS_SO_SET_ZERO:
2375                 ret = ip_vs_zero_service(svc);
2376                 break;
2377         case IP_VS_SO_SET_ADDDEST:
2378                 ret = ip_vs_add_dest(svc, &udest);
2379                 break;
2380         case IP_VS_SO_SET_EDITDEST:
2381                 ret = ip_vs_edit_dest(svc, &udest);
2382                 break;
2383         case IP_VS_SO_SET_DELDEST:
2384                 ret = ip_vs_del_dest(svc, &udest);
2385                 break;
2386         default:
2387                 ret = -EINVAL;
2388         }
2389
2390   out_unlock:
2391         mutex_unlock(&__ip_vs_mutex);
2392   out_dec:
2393         /* decrease the module use count */
2394         ip_vs_use_count_dec();
2395
2396         return ret;
2397 }
2398
2399
2400 static void
2401 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2402 {
2403         struct ip_vs_scheduler *sched;
2404
2405         sched = rcu_dereference_protected(src->scheduler, 1);
2406         dst->protocol = src->protocol;
2407         dst->addr = src->addr.ip;
2408         dst->port = src->port;
2409         dst->fwmark = src->fwmark;
2410         strlcpy(dst->sched_name, sched->name, sizeof(dst->sched_name));
2411         dst->flags = src->flags;
2412         dst->timeout = src->timeout / HZ;
2413         dst->netmask = src->netmask;
2414         dst->num_dests = src->num_dests;
2415         ip_vs_copy_stats(&dst->stats, &src->stats);
2416 }
2417
2418 static inline int
2419 __ip_vs_get_service_entries(struct net *net,
2420                             const struct ip_vs_get_services *get,
2421                             struct ip_vs_get_services __user *uptr)
2422 {
2423         int idx, count=0;
2424         struct ip_vs_service *svc;
2425         struct ip_vs_service_entry entry;
2426         int ret = 0;
2427
2428         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2429                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2430                         /* Only expose IPv4 entries to old interface */
2431                         if (svc->af != AF_INET || !net_eq(svc->net, net))
2432                                 continue;
2433
2434                         if (count >= get->num_services)
2435                                 goto out;
2436                         memset(&entry, 0, sizeof(entry));
2437                         ip_vs_copy_service(&entry, svc);
2438                         if (copy_to_user(&uptr->entrytable[count],
2439                                          &entry, sizeof(entry))) {
2440                                 ret = -EFAULT;
2441                                 goto out;
2442                         }
2443                         count++;
2444                 }
2445         }
2446
2447         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2448                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2449                         /* Only expose IPv4 entries to old interface */
2450                         if (svc->af != AF_INET || !net_eq(svc->net, net))
2451                                 continue;
2452
2453                         if (count >= get->num_services)
2454                                 goto out;
2455                         memset(&entry, 0, sizeof(entry));
2456                         ip_vs_copy_service(&entry, svc);
2457                         if (copy_to_user(&uptr->entrytable[count],
2458                                          &entry, sizeof(entry))) {
2459                                 ret = -EFAULT;
2460                                 goto out;
2461                         }
2462                         count++;
2463                 }
2464         }
2465 out:
2466         return ret;
2467 }
2468
2469 static inline int
2470 __ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
2471                          struct ip_vs_get_dests __user *uptr)
2472 {
2473         struct ip_vs_service *svc;
2474         union nf_inet_addr addr = { .ip = get->addr };
2475         int ret = 0;
2476
2477         rcu_read_lock();
2478         if (get->fwmark)
2479                 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
2480         else
2481                 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
2482                                            get->port);
2483         rcu_read_unlock();
2484
2485         if (svc) {
2486                 int count = 0;
2487                 struct ip_vs_dest *dest;
2488                 struct ip_vs_dest_entry entry;
2489
2490                 memset(&entry, 0, sizeof(entry));
2491                 list_for_each_entry(dest, &svc->destinations, n_list) {
2492                         if (count >= get->num_dests)
2493                                 break;
2494
2495                         /* Cannot expose heterogeneous members via sockopt
2496                          * interface
2497                          */
2498                         if (dest->af != svc->af)
2499                                 continue;
2500
2501                         entry.addr = dest->addr.ip;
2502                         entry.port = dest->port;
2503                         entry.conn_flags = atomic_read(&dest->conn_flags);
2504                         entry.weight = atomic_read(&dest->weight);
2505                         entry.u_threshold = dest->u_threshold;
2506                         entry.l_threshold = dest->l_threshold;
2507                         entry.activeconns = atomic_read(&dest->activeconns);
2508                         entry.inactconns = atomic_read(&dest->inactconns);
2509                         entry.persistconns = atomic_read(&dest->persistconns);
2510                         ip_vs_copy_stats(&entry.stats, &dest->stats);
2511                         if (copy_to_user(&uptr->entrytable[count],
2512                                          &entry, sizeof(entry))) {
2513                                 ret = -EFAULT;
2514                                 break;
2515                         }
2516                         count++;
2517                 }
2518         } else
2519                 ret = -ESRCH;
2520         return ret;
2521 }
2522
2523 static inline void
2524 __ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
2525 {
2526 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2527         struct ip_vs_proto_data *pd;
2528 #endif
2529
2530         memset(u, 0, sizeof (*u));
2531
2532 #ifdef CONFIG_IP_VS_PROTO_TCP
2533         pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2534         u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2535         u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2536 #endif
2537 #ifdef CONFIG_IP_VS_PROTO_UDP
2538         pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2539         u->udp_timeout =
2540                         pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2541 #endif
2542 }
2543
2544 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2545         [CMDID(IP_VS_SO_GET_VERSION)]  = 64,
2546         [CMDID(IP_VS_SO_GET_INFO)]     = sizeof(struct ip_vs_getinfo),
2547         [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2548         [CMDID(IP_VS_SO_GET_SERVICE)]  = sizeof(struct ip_vs_service_entry),
2549         [CMDID(IP_VS_SO_GET_DESTS)]    = sizeof(struct ip_vs_get_dests),
2550         [CMDID(IP_VS_SO_GET_TIMEOUT)]  = sizeof(struct ip_vs_timeout_user),
2551         [CMDID(IP_VS_SO_GET_DAEMON)]   = 2 * sizeof(struct ip_vs_daemon_user),
2552 };
2553
2554 union ip_vs_get_arglen {
2555         char                            field_IP_VS_SO_GET_VERSION[64];
2556         struct ip_vs_getinfo            field_IP_VS_SO_GET_INFO;
2557         struct ip_vs_get_services       field_IP_VS_SO_GET_SERVICES;
2558         struct ip_vs_service_entry      field_IP_VS_SO_GET_SERVICE;
2559         struct ip_vs_get_dests          field_IP_VS_SO_GET_DESTS;
2560         struct ip_vs_timeout_user       field_IP_VS_SO_GET_TIMEOUT;
2561         struct ip_vs_daemon_user        field_IP_VS_SO_GET_DAEMON[2];
2562 };
2563
2564 #define MAX_GET_ARGLEN  sizeof(union ip_vs_get_arglen)
2565
2566 static int
2567 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2568 {
2569         unsigned char arg[MAX_GET_ARGLEN];
2570         int ret = 0;
2571         unsigned int copylen;
2572         struct net *net = sock_net(sk);
2573         struct netns_ipvs *ipvs = net_ipvs(net);
2574
2575         BUG_ON(!net);
2576         BUILD_BUG_ON(sizeof(arg) > 255);
2577         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2578                 return -EPERM;
2579
2580         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2581                 return -EINVAL;
2582
2583         copylen = get_arglen[CMDID(cmd)];
2584         if (*len < (int) copylen) {
2585                 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
2586                 return -EINVAL;
2587         }
2588
2589         if (copy_from_user(arg, user, copylen) != 0)
2590                 return -EFAULT;
2591         /*
2592          * Handle daemons first since it has its own locking
2593          */
2594         if (cmd == IP_VS_SO_GET_DAEMON) {
2595                 struct ip_vs_daemon_user d[2];
2596
2597                 memset(&d, 0, sizeof(d));
2598                 mutex_lock(&ipvs->sync_mutex);
2599                 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2600                         d[0].state = IP_VS_STATE_MASTER;
2601                         strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2602                                 sizeof(d[0].mcast_ifn));
2603                         d[0].syncid = ipvs->master_syncid;
2604                 }
2605                 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2606                         d[1].state = IP_VS_STATE_BACKUP;
2607                         strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2608                                 sizeof(d[1].mcast_ifn));
2609                         d[1].syncid = ipvs->backup_syncid;
2610                 }
2611                 if (copy_to_user(user, &d, sizeof(d)) != 0)
2612                         ret = -EFAULT;
2613                 mutex_unlock(&ipvs->sync_mutex);
2614                 return ret;
2615         }
2616
2617         mutex_lock(&__ip_vs_mutex);
2618         switch (cmd) {
2619         case IP_VS_SO_GET_VERSION:
2620         {
2621                 char buf[64];
2622
2623                 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2624                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2625                 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2626                         ret = -EFAULT;
2627                         goto out;
2628                 }
2629                 *len = strlen(buf)+1;
2630         }
2631         break;
2632
2633         case IP_VS_SO_GET_INFO:
2634         {
2635                 struct ip_vs_getinfo info;
2636                 info.version = IP_VS_VERSION_CODE;
2637                 info.size = ip_vs_conn_tab_size;
2638                 info.num_services = ipvs->num_services;
2639                 if (copy_to_user(user, &info, sizeof(info)) != 0)
2640                         ret = -EFAULT;
2641         }
2642         break;
2643
2644         case IP_VS_SO_GET_SERVICES:
2645         {
2646                 struct ip_vs_get_services *get;
2647                 int size;
2648
2649                 get = (struct ip_vs_get_services *)arg;
2650                 size = sizeof(*get) +
2651                         sizeof(struct ip_vs_service_entry) * get->num_services;
2652                 if (*len != size) {
2653                         pr_err("length: %u != %u\n", *len, size);
2654                         ret = -EINVAL;
2655                         goto out;
2656                 }
2657                 ret = __ip_vs_get_service_entries(net, get, user);
2658         }
2659         break;
2660
2661         case IP_VS_SO_GET_SERVICE:
2662         {
2663                 struct ip_vs_service_entry *entry;
2664                 struct ip_vs_service *svc;
2665                 union nf_inet_addr addr;
2666
2667                 entry = (struct ip_vs_service_entry *)arg;
2668                 addr.ip = entry->addr;
2669                 rcu_read_lock();
2670                 if (entry->fwmark)
2671                         svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
2672                 else
2673                         svc = __ip_vs_service_find(net, AF_INET,
2674                                                    entry->protocol, &addr,
2675                                                    entry->port);
2676                 rcu_read_unlock();
2677                 if (svc) {
2678                         ip_vs_copy_service(entry, svc);
2679                         if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2680                                 ret = -EFAULT;
2681                 } else
2682                         ret = -ESRCH;
2683         }
2684         break;
2685
2686         case IP_VS_SO_GET_DESTS:
2687         {
2688                 struct ip_vs_get_dests *get;
2689                 int size;
2690
2691                 get = (struct ip_vs_get_dests *)arg;
2692                 size = sizeof(*get) +
2693                         sizeof(struct ip_vs_dest_entry) * get->num_dests;
2694                 if (*len != size) {
2695                         pr_err("length: %u != %u\n", *len, size);
2696                         ret = -EINVAL;
2697                         goto out;
2698                 }
2699                 ret = __ip_vs_get_dest_entries(net, get, user);
2700         }
2701         break;
2702
2703         case IP_VS_SO_GET_TIMEOUT:
2704         {
2705                 struct ip_vs_timeout_user t;
2706
2707                 __ip_vs_get_timeouts(net, &t);
2708                 if (copy_to_user(user, &t, sizeof(t)) != 0)
2709                         ret = -EFAULT;
2710         }
2711         break;
2712
2713         default:
2714                 ret = -EINVAL;
2715         }
2716
2717 out:
2718         mutex_unlock(&__ip_vs_mutex);
2719         return ret;
2720 }
2721
2722
2723 static struct nf_sockopt_ops ip_vs_sockopts = {
2724         .pf             = PF_INET,
2725         .set_optmin     = IP_VS_BASE_CTL,
2726         .set_optmax     = IP_VS_SO_SET_MAX+1,
2727         .set            = do_ip_vs_set_ctl,
2728         .get_optmin     = IP_VS_BASE_CTL,
2729         .get_optmax     = IP_VS_SO_GET_MAX+1,
2730         .get            = do_ip_vs_get_ctl,
2731         .owner          = THIS_MODULE,
2732 };
2733
2734 /*
2735  * Generic Netlink interface
2736  */
2737
2738 /* IPVS genetlink family */
2739 static struct genl_family ip_vs_genl_family = {
2740         .id             = GENL_ID_GENERATE,
2741         .hdrsize        = 0,
2742         .name           = IPVS_GENL_NAME,
2743         .version        = IPVS_GENL_VERSION,
2744         .maxattr        = IPVS_CMD_MAX,
2745         .netnsok        = true,         /* Make ipvsadm to work on netns */
2746 };
2747
2748 /* Policy used for first-level command attributes */
2749 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2750         [IPVS_CMD_ATTR_SERVICE]         = { .type = NLA_NESTED },
2751         [IPVS_CMD_ATTR_DEST]            = { .type = NLA_NESTED },
2752         [IPVS_CMD_ATTR_DAEMON]          = { .type = NLA_NESTED },
2753         [IPVS_CMD_ATTR_TIMEOUT_TCP]     = { .type = NLA_U32 },
2754         [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2755         [IPVS_CMD_ATTR_TIMEOUT_UDP]     = { .type = NLA_U32 },
2756 };
2757
2758 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2759 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2760         [IPVS_DAEMON_ATTR_STATE]        = { .type = NLA_U32 },
2761         [IPVS_DAEMON_ATTR_MCAST_IFN]    = { .type = NLA_NUL_STRING,
2762                                             .len = IP_VS_IFNAME_MAXLEN },
2763         [IPVS_DAEMON_ATTR_SYNC_ID]      = { .type = NLA_U32 },
2764 };
2765
2766 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2767 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2768         [IPVS_SVC_ATTR_AF]              = { .type = NLA_U16 },
2769         [IPVS_SVC_ATTR_PROTOCOL]        = { .type = NLA_U16 },
2770         [IPVS_SVC_ATTR_ADDR]            = { .type = NLA_BINARY,
2771                                             .len = sizeof(union nf_inet_addr) },
2772         [IPVS_SVC_ATTR_PORT]            = { .type = NLA_U16 },
2773         [IPVS_SVC_ATTR_FWMARK]          = { .type = NLA_U32 },
2774         [IPVS_SVC_ATTR_SCHED_NAME]      = { .type = NLA_NUL_STRING,
2775                                             .len = IP_VS_SCHEDNAME_MAXLEN },
2776         [IPVS_SVC_ATTR_PE_NAME]         = { .type = NLA_NUL_STRING,
2777                                             .len = IP_VS_PENAME_MAXLEN },
2778         [IPVS_SVC_ATTR_FLAGS]           = { .type = NLA_BINARY,
2779                                             .len = sizeof(struct ip_vs_flags) },
2780         [IPVS_SVC_ATTR_TIMEOUT]         = { .type = NLA_U32 },
2781         [IPVS_SVC_ATTR_NETMASK]         = { .type = NLA_U32 },
2782         [IPVS_SVC_ATTR_STATS]           = { .type = NLA_NESTED },
2783 };
2784
2785 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2786 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2787         [IPVS_DEST_ATTR_ADDR]           = { .type = NLA_BINARY,
2788                                             .len = sizeof(union nf_inet_addr) },
2789         [IPVS_DEST_ATTR_PORT]           = { .type = NLA_U16 },
2790         [IPVS_DEST_ATTR_FWD_METHOD]     = { .type = NLA_U32 },
2791         [IPVS_DEST_ATTR_WEIGHT]         = { .type = NLA_U32 },
2792         [IPVS_DEST_ATTR_U_THRESH]       = { .type = NLA_U32 },
2793         [IPVS_DEST_ATTR_L_THRESH]       = { .type = NLA_U32 },
2794         [IPVS_DEST_ATTR_ACTIVE_CONNS]   = { .type = NLA_U32 },
2795         [IPVS_DEST_ATTR_INACT_CONNS]    = { .type = NLA_U32 },
2796         [IPVS_DEST_ATTR_PERSIST_CONNS]  = { .type = NLA_U32 },
2797         [IPVS_DEST_ATTR_STATS]          = { .type = NLA_NESTED },
2798         [IPVS_DEST_ATTR_ADDR_FAMILY]    = { .type = NLA_U16 },
2799 };
2800
2801 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2802                                  struct ip_vs_stats *stats)
2803 {
2804         struct ip_vs_stats_user ustats;
2805         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2806         if (!nl_stats)
2807                 return -EMSGSIZE;
2808
2809         ip_vs_copy_stats(&ustats, stats);
2810
2811         if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns) ||
2812             nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts) ||
2813             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts) ||
2814             nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes) ||
2815             nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes) ||
2816             nla_put_u32(skb, IPVS_STATS_ATTR_CPS, ustats.cps) ||
2817             nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps) ||
2818             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps) ||
2819             nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps) ||
2820             nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps))
2821                 goto nla_put_failure;
2822         nla_nest_end(skb, nl_stats);
2823
2824         return 0;
2825
2826 nla_put_failure:
2827         nla_nest_cancel(skb, nl_stats);
2828         return -EMSGSIZE;
2829 }
2830
2831 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2832                                    struct ip_vs_service *svc)
2833 {
2834         struct ip_vs_scheduler *sched;
2835         struct ip_vs_pe *pe;
2836         struct nlattr *nl_service;
2837         struct ip_vs_flags flags = { .flags = svc->flags,
2838                                      .mask = ~0 };
2839
2840         nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2841         if (!nl_service)
2842                 return -EMSGSIZE;
2843
2844         if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2845                 goto nla_put_failure;
2846         if (svc->fwmark) {
2847                 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2848                         goto nla_put_failure;
2849         } else {
2850                 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2851                     nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2852                     nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2853                         goto nla_put_failure;
2854         }
2855
2856         sched = rcu_dereference_protected(svc->scheduler, 1);
2857         pe = rcu_dereference_protected(svc->pe, 1);
2858         if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched->name) ||
2859             (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
2860             nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2861             nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2862             nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2863                 goto nla_put_failure;
2864         if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2865                 goto nla_put_failure;
2866
2867         nla_nest_end(skb, nl_service);
2868
2869         return 0;
2870
2871 nla_put_failure:
2872         nla_nest_cancel(skb, nl_service);
2873         return -EMSGSIZE;
2874 }
2875
2876 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2877                                    struct ip_vs_service *svc,
2878                                    struct netlink_callback *cb)
2879 {
2880         void *hdr;
2881
2882         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2883                           &ip_vs_genl_family, NLM_F_MULTI,
2884                           IPVS_CMD_NEW_SERVICE);
2885         if (!hdr)
2886                 return -EMSGSIZE;
2887
2888         if (ip_vs_genl_fill_service(skb, svc) < 0)
2889                 goto nla_put_failure;
2890
2891         return genlmsg_end(skb, hdr);
2892
2893 nla_put_failure:
2894         genlmsg_cancel(skb, hdr);
2895         return -EMSGSIZE;
2896 }
2897
2898 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2899                                     struct netlink_callback *cb)
2900 {
2901         int idx = 0, i;
2902         int start = cb->args[0];
2903         struct ip_vs_service *svc;
2904         struct net *net = skb_sknet(skb);
2905
2906         mutex_lock(&__ip_vs_mutex);
2907         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2908                 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2909                         if (++idx <= start || !net_eq(svc->net, net))
2910                                 continue;
2911                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2912                                 idx--;
2913                                 goto nla_put_failure;
2914                         }
2915                 }
2916         }
2917
2918         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2919                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2920                         if (++idx <= start || !net_eq(svc->net, net))
2921                                 continue;
2922                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2923                                 idx--;
2924                                 goto nla_put_failure;
2925                         }
2926                 }
2927         }
2928
2929 nla_put_failure:
2930         mutex_unlock(&__ip_vs_mutex);
2931         cb->args[0] = idx;
2932
2933         return skb->len;
2934 }
2935
2936 static int ip_vs_genl_parse_service(struct net *net,
2937                                     struct ip_vs_service_user_kern *usvc,
2938                                     struct nlattr *nla, int full_entry,
2939                                     struct ip_vs_service **ret_svc)
2940 {
2941         struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2942         struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2943         struct ip_vs_service *svc;
2944
2945         /* Parse mandatory identifying service fields first */
2946         if (nla == NULL ||
2947             nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2948                 return -EINVAL;
2949
2950         nla_af          = attrs[IPVS_SVC_ATTR_AF];
2951         nla_protocol    = attrs[IPVS_SVC_ATTR_PROTOCOL];
2952         nla_addr        = attrs[IPVS_SVC_ATTR_ADDR];
2953         nla_port        = attrs[IPVS_SVC_ATTR_PORT];
2954         nla_fwmark      = attrs[IPVS_SVC_ATTR_FWMARK];
2955
2956         if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2957                 return -EINVAL;
2958
2959         memset(usvc, 0, sizeof(*usvc));
2960
2961         usvc->af = nla_get_u16(nla_af);
2962 #ifdef CONFIG_IP_VS_IPV6
2963         if (usvc->af != AF_INET && usvc->af != AF_INET6)
2964 #else
2965         if (usvc->af != AF_INET)
2966 #endif
2967                 return -EAFNOSUPPORT;
2968
2969         if (nla_fwmark) {
2970                 usvc->protocol = IPPROTO_TCP;
2971                 usvc->fwmark = nla_get_u32(nla_fwmark);
2972         } else {
2973                 usvc->protocol = nla_get_u16(nla_protocol);
2974                 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
2975                 usvc->port = nla_get_be16(nla_port);
2976                 usvc->fwmark = 0;
2977         }
2978
2979         rcu_read_lock();
2980         if (usvc->fwmark)
2981                 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
2982         else
2983                 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
2984                                            &usvc->addr, usvc->port);
2985         rcu_read_unlock();
2986         *ret_svc = svc;
2987
2988         /* If a full entry was requested, check for the additional fields */
2989         if (full_entry) {
2990                 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
2991                               *nla_netmask;
2992                 struct ip_vs_flags flags;
2993
2994                 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
2995                 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
2996                 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
2997                 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
2998                 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
2999
3000                 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3001                         return -EINVAL;
3002
3003                 nla_memcpy(&flags, nla_flags, sizeof(flags));
3004
3005                 /* prefill flags from service if it already exists */
3006                 if (svc)
3007                         usvc->flags = svc->flags;
3008
3009                 /* set new flags from userland */
3010                 usvc->flags = (usvc->flags & ~flags.mask) |
3011                               (flags.flags & flags.mask);
3012                 usvc->sched_name = nla_data(nla_sched);
3013                 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3014                 usvc->timeout = nla_get_u32(nla_timeout);
3015                 usvc->netmask = nla_get_be32(nla_netmask);
3016         }
3017
3018         return 0;
3019 }
3020
3021 static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3022                                                      struct nlattr *nla)
3023 {
3024         struct ip_vs_service_user_kern usvc;
3025         struct ip_vs_service *svc;
3026         int ret;
3027
3028         ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
3029         return ret ? ERR_PTR(ret) : svc;
3030 }
3031
3032 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3033 {
3034         struct nlattr *nl_dest;
3035
3036         nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3037         if (!nl_dest)
3038                 return -EMSGSIZE;
3039
3040         if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3041             nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3042             nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3043                         (atomic_read(&dest->conn_flags) &
3044                          IP_VS_CONN_F_FWD_MASK)) ||
3045             nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3046                         atomic_read(&dest->weight)) ||
3047             nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3048             nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3049             nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3050                         atomic_read(&dest->activeconns)) ||
3051             nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3052                         atomic_read(&dest->inactconns)) ||
3053             nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3054                         atomic_read(&dest->persistconns)) ||
3055             nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3056                 goto nla_put_failure;
3057         if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3058                 goto nla_put_failure;
3059
3060         nla_nest_end(skb, nl_dest);
3061
3062         return 0;
3063
3064 nla_put_failure:
3065         nla_nest_cancel(skb, nl_dest);
3066         return -EMSGSIZE;
3067 }
3068
3069 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3070                                 struct netlink_callback *cb)
3071 {
3072         void *hdr;
3073
3074         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3075                           &ip_vs_genl_family, NLM_F_MULTI,
3076                           IPVS_CMD_NEW_DEST);
3077         if (!hdr)
3078                 return -EMSGSIZE;
3079
3080         if (ip_vs_genl_fill_dest(skb, dest) < 0)
3081                 goto nla_put_failure;
3082
3083         return genlmsg_end(skb, hdr);
3084
3085 nla_put_failure:
3086         genlmsg_cancel(skb, hdr);
3087         return -EMSGSIZE;
3088 }
3089
3090 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3091                                  struct netlink_callback *cb)
3092 {
3093         int idx = 0;
3094         int start = cb->args[0];
3095         struct ip_vs_service *svc;
3096         struct ip_vs_dest *dest;
3097         struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3098         struct net *net = skb_sknet(skb);
3099
3100         mutex_lock(&__ip_vs_mutex);
3101
3102         /* Try to find the service for which to dump destinations */
3103         if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3104                         IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3105                 goto out_err;
3106
3107
3108         svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
3109         if (IS_ERR(svc) || svc == NULL)
3110                 goto out_err;
3111
3112         /* Dump the destinations */
3113         list_for_each_entry(dest, &svc->destinations, n_list) {
3114                 if (++idx <= start)
3115                         continue;
3116                 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3117                         idx--;
3118                         goto nla_put_failure;
3119                 }
3120         }
3121
3122 nla_put_failure:
3123         cb->args[0] = idx;
3124
3125 out_err:
3126         mutex_unlock(&__ip_vs_mutex);
3127
3128         return skb->len;
3129 }
3130
3131 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3132                                  struct nlattr *nla, int full_entry)
3133 {
3134         struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3135         struct nlattr *nla_addr, *nla_port;
3136         struct nlattr *nla_addr_family;
3137
3138         /* Parse mandatory identifying destination fields first */
3139         if (nla == NULL ||
3140             nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3141                 return -EINVAL;
3142
3143         nla_addr        = attrs[IPVS_DEST_ATTR_ADDR];
3144         nla_port        = attrs[IPVS_DEST_ATTR_PORT];
3145         nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3146
3147         if (!(nla_addr && nla_port))
3148                 return -EINVAL;
3149
3150         memset(udest, 0, sizeof(*udest));
3151
3152         nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3153         udest->port = nla_get_be16(nla_port);
3154
3155         if (nla_addr_family)
3156                 udest->af = nla_get_u16(nla_addr_family);
3157         else
3158                 udest->af = 0;
3159
3160         /* If a full entry was requested, check for the additional fields */
3161         if (full_entry) {
3162                 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3163                               *nla_l_thresh;
3164
3165                 nla_fwd         = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3166                 nla_weight      = attrs[IPVS_DEST_ATTR_WEIGHT];
3167                 nla_u_thresh    = attrs[IPVS_DEST_ATTR_U_THRESH];
3168                 nla_l_thresh    = attrs[IPVS_DEST_ATTR_L_THRESH];
3169
3170                 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3171                         return -EINVAL;
3172
3173                 udest->conn_flags = nla_get_u32(nla_fwd)
3174                                     & IP_VS_CONN_F_FWD_MASK;
3175                 udest->weight = nla_get_u32(nla_weight);
3176                 udest->u_threshold = nla_get_u32(nla_u_thresh);
3177                 udest->l_threshold = nla_get_u32(nla_l_thresh);
3178         }
3179
3180         return 0;
3181 }
3182
3183 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3184                                   const char *mcast_ifn, __u32 syncid)
3185 {
3186         struct nlattr *nl_daemon;
3187
3188         nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3189         if (!nl_daemon)
3190                 return -EMSGSIZE;
3191
3192         if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3193             nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn) ||
3194             nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid))
3195                 goto nla_put_failure;
3196         nla_nest_end(skb, nl_daemon);
3197
3198         return 0;
3199
3200 nla_put_failure:
3201         nla_nest_cancel(skb, nl_daemon);
3202         return -EMSGSIZE;
3203 }
3204
3205 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3206                                   const char *mcast_ifn, __u32 syncid,
3207                                   struct netlink_callback *cb)
3208 {
3209         void *hdr;
3210         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3211                           &ip_vs_genl_family, NLM_F_MULTI,
3212                           IPVS_CMD_NEW_DAEMON);
3213         if (!hdr)
3214                 return -EMSGSIZE;
3215
3216         if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3217                 goto nla_put_failure;
3218
3219         return genlmsg_end(skb, hdr);
3220
3221 nla_put_failure:
3222         genlmsg_cancel(skb, hdr);
3223         return -EMSGSIZE;
3224 }
3225
3226 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3227                                    struct netlink_callback *cb)
3228 {
3229         struct net *net = skb_sknet(skb);
3230         struct netns_ipvs *ipvs = net_ipvs(net);
3231
3232         mutex_lock(&ipvs->sync_mutex);
3233         if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3234                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3235                                            ipvs->master_mcast_ifn,
3236                                            ipvs->master_syncid, cb) < 0)
3237                         goto nla_put_failure;
3238
3239                 cb->args[0] = 1;
3240         }
3241
3242         if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3243                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3244                                            ipvs->backup_mcast_ifn,
3245                                            ipvs->backup_syncid, cb) < 0)
3246                         goto nla_put_failure;
3247
3248                 cb->args[1] = 1;
3249         }
3250
3251 nla_put_failure:
3252         mutex_unlock(&ipvs->sync_mutex);
3253
3254         return skb->len;
3255 }
3256
3257 static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
3258 {
3259         if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3260               attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3261               attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3262                 return -EINVAL;
3263
3264         /* The synchronization protocol is incompatible with mixed family
3265          * services
3266          */
3267         if (net_ipvs(net)->mixed_address_family_dests > 0)
3268                 return -EINVAL;
3269
3270         return start_sync_thread(net,
3271                                  nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3272                                  nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3273                                  nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3274 }
3275
3276 static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
3277 {
3278         if (!attrs[IPVS_DAEMON_ATTR_STATE])
3279                 return -EINVAL;
3280
3281         return stop_sync_thread(net,
3282                                 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3283 }
3284
3285 static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
3286 {
3287         struct ip_vs_timeout_user t;
3288
3289         __ip_vs_get_timeouts(net, &t);
3290
3291         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3292                 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3293
3294         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3295                 t.tcp_fin_timeout =
3296                         nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3297
3298         if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3299                 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3300
3301         return ip_vs_set_timeout(net, &t);
3302 }
3303
3304 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3305 {
3306         int ret = 0, cmd;
3307         struct net *net;
3308         struct netns_ipvs *ipvs;
3309
3310         net = skb_sknet(skb);
3311         ipvs = net_ipvs(net);
3312         cmd = info->genlhdr->cmd;
3313
3314         if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3315                 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3316
3317                 mutex_lock(&ipvs->sync_mutex);
3318                 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3319                     nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3320                                      info->attrs[IPVS_CMD_ATTR_DAEMON],
3321                                      ip_vs_daemon_policy)) {
3322                         ret = -EINVAL;
3323                         goto out;
3324                 }
3325
3326                 if (cmd == IPVS_CMD_NEW_DAEMON)
3327                         ret = ip_vs_genl_new_daemon(net, daemon_attrs);
3328                 else
3329                         ret = ip_vs_genl_del_daemon(net, daemon_attrs);
3330 out:
3331                 mutex_unlock(&ipvs->sync_mutex);
3332         }
3333         return ret;
3334 }
3335
3336 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3337 {
3338         struct ip_vs_service *svc = NULL;
3339         struct ip_vs_service_user_kern usvc;
3340         struct ip_vs_dest_user_kern udest;
3341         int ret = 0, cmd;
3342         int need_full_svc = 0, need_full_dest = 0;
3343         struct net *net;
3344
3345         net = skb_sknet(skb);
3346         cmd = info->genlhdr->cmd;
3347
3348         mutex_lock(&__ip_vs_mutex);
3349
3350         if (cmd == IPVS_CMD_FLUSH) {
3351                 ret = ip_vs_flush(net, false);
3352                 goto out;
3353         } else if (cmd == IPVS_CMD_SET_CONFIG) {
3354                 ret = ip_vs_genl_set_config(net, info->attrs);
3355                 goto out;
3356         } else if (cmd == IPVS_CMD_ZERO &&
3357                    !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3358                 ret = ip_vs_zero_all(net);
3359                 goto out;
3360         }
3361
3362         /* All following commands require a service argument, so check if we
3363          * received a valid one. We need a full service specification when
3364          * adding / editing a service. Only identifying members otherwise. */
3365         if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3366                 need_full_svc = 1;
3367
3368         ret = ip_vs_genl_parse_service(net, &usvc,
3369                                        info->attrs[IPVS_CMD_ATTR_SERVICE],
3370                                        need_full_svc, &svc);
3371         if (ret)
3372                 goto out;
3373
3374         /* Unless we're adding a new service, the service must already exist */
3375         if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3376                 ret = -ESRCH;
3377                 goto out;
3378         }
3379
3380         /* Destination commands require a valid destination argument. For
3381          * adding / editing a destination, we need a full destination
3382          * specification. */
3383         if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3384             cmd == IPVS_CMD_DEL_DEST) {
3385                 if (cmd != IPVS_CMD_DEL_DEST)
3386                         need_full_dest = 1;
3387
3388                 ret = ip_vs_genl_parse_dest(&udest,
3389                                             info->attrs[IPVS_CMD_ATTR_DEST],
3390                                             need_full_dest);
3391                 if (ret)
3392                         goto out;
3393
3394                 /* Old protocols did not allow the user to specify address
3395                  * family, so we set it to zero instead.  We also didn't
3396                  * allow heterogeneous pools in the old code, so it's safe
3397                  * to assume that this will have the same address family as
3398                  * the service.
3399                  */
3400                 if (udest.af == 0)
3401                         udest.af = svc->af;
3402
3403                 if (udest.af != svc->af) {
3404                         /* The synchronization protocol is incompatible
3405                          * with mixed family services
3406                          */
3407                         if (net_ipvs(net)->sync_state) {
3408                                 ret = -EINVAL;
3409                                 goto out;
3410                         }
3411
3412                         /* Which connection types do we support? */
3413                         switch (udest.conn_flags) {
3414                         case IP_VS_CONN_F_TUNNEL:
3415                                 /* We are able to forward this */
3416                                 break;
3417                         default:
3418                                 ret = -EINVAL;
3419                                 goto out;
3420                         }
3421                 }
3422         }
3423
3424         switch (cmd) {
3425         case IPVS_CMD_NEW_SERVICE:
3426                 if (svc == NULL)
3427                         ret = ip_vs_add_service(net, &usvc, &svc);
3428                 else
3429                         ret = -EEXIST;
3430                 break;
3431         case IPVS_CMD_SET_SERVICE:
3432                 ret = ip_vs_edit_service(svc, &usvc);
3433                 break;
3434         case IPVS_CMD_DEL_SERVICE:
3435                 ret = ip_vs_del_service(svc);
3436                 /* do not use svc, it can be freed */
3437                 break;
3438         case IPVS_CMD_NEW_DEST:
3439                 ret = ip_vs_add_dest(svc, &udest);
3440                 break;
3441         case IPVS_CMD_SET_DEST:
3442                 ret = ip_vs_edit_dest(svc, &udest);
3443                 break;
3444         case IPVS_CMD_DEL_DEST:
3445                 ret = ip_vs_del_dest(svc, &udest);
3446                 break;
3447         case IPVS_CMD_ZERO:
3448                 ret = ip_vs_zero_service(svc);
3449                 break;
3450         default:
3451                 ret = -EINVAL;
3452         }
3453
3454 out:
3455         mutex_unlock(&__ip_vs_mutex);
3456
3457         return ret;
3458 }
3459
3460 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3461 {
3462         struct sk_buff *msg;
3463         void *reply;
3464         int ret, cmd, reply_cmd;
3465         struct net *net;
3466
3467         net = skb_sknet(skb);
3468         cmd = info->genlhdr->cmd;
3469
3470         if (cmd == IPVS_CMD_GET_SERVICE)
3471                 reply_cmd = IPVS_CMD_NEW_SERVICE;
3472         else if (cmd == IPVS_CMD_GET_INFO)
3473                 reply_cmd = IPVS_CMD_SET_INFO;
3474         else if (cmd == IPVS_CMD_GET_CONFIG)
3475                 reply_cmd = IPVS_CMD_SET_CONFIG;
3476         else {
3477                 pr_err("unknown Generic Netlink command\n");
3478                 return -EINVAL;
3479         }
3480
3481         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3482         if (!msg)
3483                 return -ENOMEM;
3484
3485         mutex_lock(&__ip_vs_mutex);
3486
3487         reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3488         if (reply == NULL)
3489                 goto nla_put_failure;
3490
3491         switch (cmd) {
3492         case IPVS_CMD_GET_SERVICE:
3493         {
3494                 struct ip_vs_service *svc;
3495
3496                 svc = ip_vs_genl_find_service(net,
3497                                               info->attrs[IPVS_CMD_ATTR_SERVICE]);
3498                 if (IS_ERR(svc)) {
3499                         ret = PTR_ERR(svc);
3500                         goto out_err;
3501                 } else if (svc) {
3502                         ret = ip_vs_genl_fill_service(msg, svc);
3503                         if (ret)
3504                                 goto nla_put_failure;
3505                 } else {
3506                         ret = -ESRCH;
3507                         goto out_err;
3508                 }
3509
3510                 break;
3511         }
3512
3513         case IPVS_CMD_GET_CONFIG:
3514         {
3515                 struct ip_vs_timeout_user t;
3516
3517                 __ip_vs_get_timeouts(net, &t);
3518 #ifdef CONFIG_IP_VS_PROTO_TCP
3519                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3520                                 t.tcp_timeout) ||
3521                     nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3522                                 t.tcp_fin_timeout))
3523                         goto nla_put_failure;
3524 #endif
3525 #ifdef CONFIG_IP_VS_PROTO_UDP
3526                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3527                         goto nla_put_failure;
3528 #endif
3529
3530                 break;
3531         }
3532
3533         case IPVS_CMD_GET_INFO:
3534                 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3535                                 IP_VS_VERSION_CODE) ||
3536                     nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3537                                 ip_vs_conn_tab_size))
3538                         goto nla_put_failure;
3539                 break;
3540         }
3541
3542         genlmsg_end(msg, reply);
3543         ret = genlmsg_reply(msg, info);
3544         goto out;
3545
3546 nla_put_failure:
3547         pr_err("not enough space in Netlink message\n");
3548         ret = -EMSGSIZE;
3549
3550 out_err:
3551         nlmsg_free(msg);
3552 out:
3553         mutex_unlock(&__ip_vs_mutex);
3554
3555         return ret;
3556 }
3557
3558
3559 static const struct genl_ops ip_vs_genl_ops[] = {
3560         {
3561                 .cmd    = IPVS_CMD_NEW_SERVICE,
3562                 .flags  = GENL_ADMIN_PERM,
3563                 .policy = ip_vs_cmd_policy,
3564                 .doit   = ip_vs_genl_set_cmd,
3565         },
3566         {
3567                 .cmd    = IPVS_CMD_SET_SERVICE,
3568                 .flags  = GENL_ADMIN_PERM,
3569                 .policy = ip_vs_cmd_policy,
3570                 .doit   = ip_vs_genl_set_cmd,
3571         },
3572         {
3573                 .cmd    = IPVS_CMD_DEL_SERVICE,
3574                 .flags  = GENL_ADMIN_PERM,
3575                 .policy = ip_vs_cmd_policy,
3576                 .doit   = ip_vs_genl_set_cmd,
3577         },
3578         {
3579                 .cmd    = IPVS_CMD_GET_SERVICE,
3580                 .flags  = GENL_ADMIN_PERM,
3581                 .doit   = ip_vs_genl_get_cmd,
3582                 .dumpit = ip_vs_genl_dump_services,
3583                 .policy = ip_vs_cmd_policy,
3584         },
3585         {
3586                 .cmd    = IPVS_CMD_NEW_DEST,
3587                 .flags  = GENL_ADMIN_PERM,
3588                 .policy = ip_vs_cmd_policy,
3589                 .doit   = ip_vs_genl_set_cmd,
3590         },
3591         {
3592                 .cmd    = IPVS_CMD_SET_DEST,
3593                 .flags  = GENL_ADMIN_PERM,
3594                 .policy = ip_vs_cmd_policy,
3595                 .doit   = ip_vs_genl_set_cmd,
3596         },
3597         {
3598                 .cmd    = IPVS_CMD_DEL_DEST,
3599                 .flags  = GENL_ADMIN_PERM,
3600                 .policy = ip_vs_cmd_policy,
3601                 .doit   = ip_vs_genl_set_cmd,
3602         },
3603         {
3604                 .cmd    = IPVS_CMD_GET_DEST,
3605                 .flags  = GENL_ADMIN_PERM,
3606                 .policy = ip_vs_cmd_policy,
3607                 .dumpit = ip_vs_genl_dump_dests,
3608         },
3609         {
3610                 .cmd    = IPVS_CMD_NEW_DAEMON,
3611                 .flags  = GENL_ADMIN_PERM,
3612                 .policy = ip_vs_cmd_policy,
3613                 .doit   = ip_vs_genl_set_daemon,
3614         },
3615         {
3616                 .cmd    = IPVS_CMD_DEL_DAEMON,
3617                 .flags  = GENL_ADMIN_PERM,
3618                 .policy = ip_vs_cmd_policy,
3619                 .doit   = ip_vs_genl_set_daemon,
3620         },
3621         {
3622                 .cmd    = IPVS_CMD_GET_DAEMON,
3623                 .flags  = GENL_ADMIN_PERM,
3624                 .dumpit = ip_vs_genl_dump_daemons,
3625         },
3626         {
3627                 .cmd    = IPVS_CMD_SET_CONFIG,
3628                 .flags  = GENL_ADMIN_PERM,
3629                 .policy = ip_vs_cmd_policy,
3630                 .doit   = ip_vs_genl_set_cmd,
3631         },
3632         {
3633                 .cmd    = IPVS_CMD_GET_CONFIG,
3634                 .flags  = GENL_ADMIN_PERM,
3635                 .doit   = ip_vs_genl_get_cmd,
3636         },
3637         {
3638                 .cmd    = IPVS_CMD_GET_INFO,
3639                 .flags  = GENL_ADMIN_PERM,
3640                 .doit   = ip_vs_genl_get_cmd,
3641         },
3642         {
3643                 .cmd    = IPVS_CMD_ZERO,
3644                 .flags  = GENL_ADMIN_PERM,
3645                 .policy = ip_vs_cmd_policy,
3646                 .doit   = ip_vs_genl_set_cmd,
3647         },
3648         {
3649                 .cmd    = IPVS_CMD_FLUSH,
3650                 .flags  = GENL_ADMIN_PERM,
3651                 .doit   = ip_vs_genl_set_cmd,
3652         },
3653 };
3654
3655 static int __init ip_vs_genl_register(void)
3656 {
3657         return genl_register_family_with_ops(&ip_vs_genl_family,
3658                                              ip_vs_genl_ops);
3659 }
3660
3661 static void ip_vs_genl_unregister(void)
3662 {
3663         genl_unregister_family(&ip_vs_genl_family);
3664 }
3665
3666 /* End of Generic Netlink interface definitions */
3667
3668 /*
3669  * per netns intit/exit func.
3670  */
3671 #ifdef CONFIG_SYSCTL
3672 static int __net_init ip_vs_control_net_init_sysctl(struct net *net)
3673 {
3674         int idx;
3675         struct netns_ipvs *ipvs = net_ipvs(net);
3676         struct ctl_table *tbl;
3677
3678         atomic_set(&ipvs->dropentry, 0);
3679         spin_lock_init(&ipvs->dropentry_lock);
3680         spin_lock_init(&ipvs->droppacket_lock);
3681         spin_lock_init(&ipvs->securetcp_lock);
3682
3683         if (!net_eq(net, &init_net)) {
3684                 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3685                 if (tbl == NULL)
3686                         return -ENOMEM;
3687
3688                 /* Don't export sysctls to unprivileged users */
3689                 if (net->user_ns != &init_user_ns)
3690                         tbl[0].procname = NULL;
3691         } else
3692                 tbl = vs_vars;
3693         /* Initialize sysctl defaults */
3694         idx = 0;
3695         ipvs->sysctl_amemthresh = 1024;
3696         tbl[idx++].data = &ipvs->sysctl_amemthresh;
3697         ipvs->sysctl_am_droprate = 10;
3698         tbl[idx++].data = &ipvs->sysctl_am_droprate;
3699         tbl[idx++].data = &ipvs->sysctl_drop_entry;
3700         tbl[idx++].data = &ipvs->sysctl_drop_packet;
3701 #ifdef CONFIG_IP_VS_NFCT
3702         tbl[idx++].data = &ipvs->sysctl_conntrack;
3703 #endif
3704         tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3705         ipvs->sysctl_snat_reroute = 1;
3706         tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3707         ipvs->sysctl_sync_ver = 1;
3708         tbl[idx++].data = &ipvs->sysctl_sync_ver;
3709         ipvs->sysctl_sync_ports = 1;
3710         tbl[idx++].data = &ipvs->sysctl_sync_ports;
3711         tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3712         ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3713         tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3714         ipvs->sysctl_sync_sock_size = 0;
3715         tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3716         tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3717         tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3718         tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3719         tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3720         tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3721         ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3722         ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3723         tbl[idx].data = &ipvs->sysctl_sync_threshold;
3724         tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3725         ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3726         tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3727         ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3728         tbl[idx++].data = &ipvs->sysctl_sync_retries;
3729         tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3730         ipvs->sysctl_pmtu_disc = 1;
3731         tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3732         tbl[idx++].data = &ipvs->sysctl_backup_only;
3733
3734
3735         ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3736         if (ipvs->sysctl_hdr == NULL) {
3737                 if (!net_eq(net, &init_net))
3738                         kfree(tbl);
3739                 return -ENOMEM;
3740         }
3741         ip_vs_start_estimator(net, &ipvs->tot_stats);
3742         ipvs->sysctl_tbl = tbl;
3743         /* Schedule defense work */
3744         INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3745         schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3746
3747         return 0;
3748 }
3749
3750 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net)
3751 {
3752         struct netns_ipvs *ipvs = net_ipvs(net);
3753
3754         cancel_delayed_work_sync(&ipvs->defense_work);
3755         cancel_work_sync(&ipvs->defense_work.work);
3756         unregister_net_sysctl_table(ipvs->sysctl_hdr);
3757         ip_vs_stop_estimator(net, &ipvs->tot_stats);
3758 }
3759
3760 #else
3761
3762 static int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
3763 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net) { }
3764
3765 #endif
3766
3767 static struct notifier_block ip_vs_dst_notifier = {
3768         .notifier_call = ip_vs_dst_event,
3769 };
3770
3771 int __net_init ip_vs_control_net_init(struct net *net)
3772 {
3773         int i, idx;
3774         struct netns_ipvs *ipvs = net_ipvs(net);
3775
3776         /* Initialize rs_table */
3777         for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3778                 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
3779
3780         INIT_LIST_HEAD(&ipvs->dest_trash);
3781         spin_lock_init(&ipvs->dest_trash_lock);
3782         setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
3783                     (unsigned long) net);
3784         atomic_set(&ipvs->ftpsvc_counter, 0);
3785         atomic_set(&ipvs->nullsvc_counter, 0);
3786
3787         /* procfs stats */
3788         ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3789         if (!ipvs->tot_stats.cpustats)
3790                 return -ENOMEM;
3791
3792         for_each_possible_cpu(i) {
3793                 struct ip_vs_cpu_stats *ipvs_tot_stats;
3794                 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
3795                 u64_stats_init(&ipvs_tot_stats->syncp);
3796         }
3797
3798         spin_lock_init(&ipvs->tot_stats.lock);
3799
3800         proc_create("ip_vs", 0, net->proc_net, &ip_vs_info_fops);
3801         proc_create("ip_vs_stats", 0, net->proc_net, &ip_vs_stats_fops);
3802         proc_create("ip_vs_stats_percpu", 0, net->proc_net,
3803                     &ip_vs_stats_percpu_fops);
3804
3805         if (ip_vs_control_net_init_sysctl(net))
3806                 goto err;
3807
3808         return 0;
3809
3810 err:
3811         free_percpu(ipvs->tot_stats.cpustats);
3812         return -ENOMEM;
3813 }
3814
3815 void __net_exit ip_vs_control_net_cleanup(struct net *net)
3816 {
3817         struct netns_ipvs *ipvs = net_ipvs(net);
3818
3819         ip_vs_trash_cleanup(net);
3820         ip_vs_control_net_cleanup_sysctl(net);
3821         remove_proc_entry("ip_vs_stats_percpu", net->proc_net);
3822         remove_proc_entry("ip_vs_stats", net->proc_net);
3823         remove_proc_entry("ip_vs", net->proc_net);
3824         free_percpu(ipvs->tot_stats.cpustats);
3825 }
3826
3827 int __init ip_vs_register_nl_ioctl(void)
3828 {
3829         int ret;
3830
3831         ret = nf_register_sockopt(&ip_vs_sockopts);
3832         if (ret) {
3833                 pr_err("cannot register sockopt.\n");
3834                 goto err_sock;
3835         }
3836
3837         ret = ip_vs_genl_register();
3838         if (ret) {
3839                 pr_err("cannot register Generic Netlink interface.\n");
3840                 goto err_genl;
3841         }
3842         return 0;
3843
3844 err_genl:
3845         nf_unregister_sockopt(&ip_vs_sockopts);
3846 err_sock:
3847         return ret;
3848 }
3849
3850 void ip_vs_unregister_nl_ioctl(void)
3851 {
3852         ip_vs_genl_unregister();
3853         nf_unregister_sockopt(&ip_vs_sockopts);
3854 }
3855
3856 int __init ip_vs_control_init(void)
3857 {
3858         int idx;
3859         int ret;
3860
3861         EnterFunction(2);
3862
3863         /* Initialize svc_table, ip_vs_svc_fwm_table */
3864         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3865                 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
3866                 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3867         }
3868
3869         smp_wmb();      /* Do we really need it now ? */
3870
3871         ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3872         if (ret < 0)
3873                 return ret;
3874
3875         LeaveFunction(2);
3876         return 0;
3877 }
3878
3879
3880 void ip_vs_control_cleanup(void)
3881 {
3882         EnterFunction(2);
3883         unregister_netdevice_notifier(&ip_vs_dst_notifier);
3884         LeaveFunction(2);
3885 }