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, const union nf_inet_addr *daddr,
 578                   __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 == svc->af)
 587                     && ip_vs_addr_equal(svc->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 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, 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, daddr, port);
 622         if (!dest)
 623                 dest = ip_vs_lookup_dest(svc, 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, const union nf_inet_addr *daddr,
 661                      __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(svc->af, &dest->addr),
 675                               ntohs(dest->port),
 676                               atomic_read(&dest->refcnt));
 677                 if (dest->af == svc->af &&
 678                     ip_vs_addr_equal(svc->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         /* set the weight and the flags */
 783         atomic_set(&dest->weight, udest->weight);
 784         conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
 785         conn_flags |= IP_VS_CONN_F_INACTIVE;
 786
 787         /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
 788         if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
 789                 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
 790         } else {
 791                 /*
 792                  *    Put the real service in rs_table if not present.
 793                  *    For now only for NAT!
 794                  */
 795                 ip_vs_rs_hash(ipvs, dest);
 796         }
 797         atomic_set(&dest->conn_flags, conn_flags);
 798
 799         /* bind the service */
 800         old_svc = rcu_dereference_protected(dest->svc, 1);
 801         if (!old_svc) {
 802                 __ip_vs_bind_svc(dest, svc);
 803         } else {
 804                 if (old_svc != svc) {
 805                         ip_vs_zero_stats(&dest->stats);
 806                         __ip_vs_bind_svc(dest, svc);
 807                         __ip_vs_svc_put(old_svc, true);
 808                 }
 809         }
 810
 811         /* set the dest status flags */
 812         dest->flags |= IP_VS_DEST_F_AVAILABLE;
 813
 814         if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
 815                 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
 816         dest->u_threshold = udest->u_threshold;
 817         dest->l_threshold = udest->l_threshold;
 818
 819         spin_lock_bh(&dest->dst_lock);
 820         __ip_vs_dst_cache_reset(dest);
 821         spin_unlock_bh(&dest->dst_lock);
 822
 823         sched = rcu_dereference_protected(svc->scheduler, 1);
 824         if (add) {
 825                 ip_vs_start_estimator(svc->net, &dest->stats);
 826                 list_add_rcu(&dest->n_list, &svc->destinations);
 827                 svc->num_dests++;
 828                 if (sched->add_dest)
 829                         sched->add_dest(svc, dest);
 830         } else {
 831                 if (sched->upd_dest)
 832                         sched->upd_dest(svc, dest);
 833         }
 834 }
 835
 836
 837 /*
 838  *      Create a destination for the given service
 839  */
 840 static int
 841 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
 842                struct ip_vs_dest **dest_p)
 843 {
 844         struct ip_vs_dest *dest;
 845         unsigned int atype, i;
 846
 847         EnterFunction(2);
 848
 849 #ifdef CONFIG_IP_VS_IPV6
 850         if (svc->af == AF_INET6) {
 851                 atype = ipv6_addr_type(&udest->addr.in6);
 852                 if ((!(atype & IPV6_ADDR_UNICAST) ||
 853                         atype & IPV6_ADDR_LINKLOCAL) &&
 854                         !__ip_vs_addr_is_local_v6(svc->net, &udest->addr.in6))
 855                         return -EINVAL;
 856         } else
 857 #endif
 858         {
 859                 atype = inet_addr_type(svc->net, udest->addr.ip);
 860                 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
 861                         return -EINVAL;
 862         }
 863
 864         dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
 865         if (dest == NULL)
 866                 return -ENOMEM;
 867
 868         dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
 869         if (!dest->stats.cpustats)
 870                 goto err_alloc;
 871
 872         for_each_possible_cpu(i) {
 873                 struct ip_vs_cpu_stats *ip_vs_dest_stats;
 874                 ip_vs_dest_stats = per_cpu_ptr(dest->stats.cpustats, i);
 875                 u64_stats_init(&ip_vs_dest_stats->syncp);
 876         }
 877
 878         dest->af = svc->af;
 879         dest->protocol = svc->protocol;
 880         dest->vaddr = svc->addr;
 881         dest->vport = svc->port;
 882         dest->vfwmark = svc->fwmark;
 883         ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
 884         dest->port = udest->port;
 885
 886         atomic_set(&dest->activeconns, 0);
 887         atomic_set(&dest->inactconns, 0);
 888         atomic_set(&dest->persistconns, 0);
 889         atomic_set(&dest->refcnt, 1);
 890
 891         INIT_HLIST_NODE(&dest->d_list);
 892         spin_lock_init(&dest->dst_lock);
 893         spin_lock_init(&dest->stats.lock);
 894         __ip_vs_update_dest(svc, dest, udest, 1);
 895
 896         *dest_p = dest;
 897
 898         LeaveFunction(2);
 899         return 0;
 900
 901 err_alloc:
 902         kfree(dest);
 903         return -ENOMEM;
 904 }
 905
 906
 907 /*
 908  *      Add a destination into an existing service
 909  */
 910 static int
 911 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
 912 {
 913         struct ip_vs_dest *dest;
 914         union nf_inet_addr daddr;
 915         __be16 dport = udest->port;
 916         int ret;
 917
 918         EnterFunction(2);
 919
 920         if (udest->weight < 0) {
 921                 pr_err("%s(): server weight less than zero\n", __func__);
 922                 return -ERANGE;
 923         }
 924
 925         if (udest->l_threshold > udest->u_threshold) {
 926                 pr_err("%s(): lower threshold is higher than upper threshold\n",
 927                         __func__);
 928                 return -ERANGE;
 929         }
 930
 931         ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
 932
 933         /* We use function that requires RCU lock */
 934         rcu_read_lock();
 935         dest = ip_vs_lookup_dest(svc, &daddr, dport);
 936         rcu_read_unlock();
 937
 938         if (dest != NULL) {
 939                 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
 940                 return -EEXIST;
 941         }
 942
 943         /*
 944          * Check if the dest already exists in the trash and
 945          * is from the same service
 946          */
 947         dest = ip_vs_trash_get_dest(svc, &daddr, dport);
 948
 949         if (dest != NULL) {
 950                 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
 951                               "dest->refcnt=%d, service %u/%s:%u\n",
 952                               IP_VS_DBG_ADDR(svc->af, &daddr), ntohs(dport),
 953                               atomic_read(&dest->refcnt),
 954                               dest->vfwmark,
 955                               IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
 956                               ntohs(dest->vport));
 957
 958                 __ip_vs_update_dest(svc, dest, udest, 1);
 959                 ret = 0;
 960         } else {
 961                 /*
 962                  * Allocate and initialize the dest structure
 963                  */
 964                 ret = ip_vs_new_dest(svc, udest, &dest);
 965         }
 966         LeaveFunction(2);
 967
 968         return ret;
 969 }
 970
 971
 972 /*
 973  *      Edit a destination in the given service
 974  */
 975 static int
 976 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
 977 {
 978         struct ip_vs_dest *dest;
 979         union nf_inet_addr daddr;
 980         __be16 dport = udest->port;
 981
 982         EnterFunction(2);
 983
 984         if (udest->weight < 0) {
 985                 pr_err("%s(): server weight less than zero\n", __func__);
 986                 return -ERANGE;
 987         }
 988
 989         if (udest->l_threshold > udest->u_threshold) {
 990                 pr_err("%s(): lower threshold is higher than upper threshold\n",
 991                         __func__);
 992                 return -ERANGE;
 993         }
 994
 995         ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
 996
 997         /* We use function that requires RCU lock */
 998         rcu_read_lock();
 999         dest = ip_vs_lookup_dest(svc, &daddr, dport);
1000         rcu_read_unlock();
1001
1002         if (dest == NULL) {
1003                 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1004                 return -ENOENT;
1005         }
1006
1007         __ip_vs_update_dest(svc, dest, udest, 0);
1008         LeaveFunction(2);
1009
1010         return 0;
1011 }
1012
1013 /*
1014  *      Delete a destination (must be already unlinked from the service)
1015  */
1016 static void __ip_vs_del_dest(struct net *net, struct ip_vs_dest *dest,
1017                              bool cleanup)
1018 {
1019         struct netns_ipvs *ipvs = net_ipvs(net);
1020
1021         ip_vs_stop_estimator(net, &dest->stats);
1022
1023         /*
1024          *  Remove it from the d-linked list with the real services.
1025          */
1026         ip_vs_rs_unhash(dest);
1027
1028         spin_lock_bh(&ipvs->dest_trash_lock);
1029         IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1030                       IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1031                       atomic_read(&dest->refcnt));
1032         if (list_empty(&ipvs->dest_trash) && !cleanup)
1033                 mod_timer(&ipvs->dest_trash_timer,
1034                           jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1035         /* dest lives in trash without reference */
1036         list_add(&dest->t_list, &ipvs->dest_trash);
1037         dest->idle_start = 0;
1038         spin_unlock_bh(&ipvs->dest_trash_lock);
1039         ip_vs_dest_put(dest);
1040 }
1041
1042
1043 /*
1044  *      Unlink a destination from the given service
1045  */
1046 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1047                                 struct ip_vs_dest *dest,
1048                                 int svcupd)
1049 {
1050         dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1051
1052         /*
1053          *  Remove it from the d-linked destination list.
1054          */
1055         list_del_rcu(&dest->n_list);
1056         svc->num_dests--;
1057
1058         if (svcupd) {
1059                 struct ip_vs_scheduler *sched;
1060
1061                 sched = rcu_dereference_protected(svc->scheduler, 1);
1062                 if (sched->del_dest)
1063                         sched->del_dest(svc, dest);
1064         }
1065 }
1066
1067
1068 /*
1069  *      Delete a destination server in the given service
1070  */
1071 static int
1072 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1073 {
1074         struct ip_vs_dest *dest;
1075         __be16 dport = udest->port;
1076
1077         EnterFunction(2);
1078
1079         /* We use function that requires RCU lock */
1080         rcu_read_lock();
1081         dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
1082         rcu_read_unlock();
1083
1084         if (dest == NULL) {
1085                 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1086                 return -ENOENT;
1087         }
1088
1089         /*
1090          *      Unlink dest from the service
1091          */
1092         __ip_vs_unlink_dest(svc, dest, 1);
1093
1094         /*
1095          *      Delete the destination
1096          */
1097         __ip_vs_del_dest(svc->net, dest, false);
1098
1099         LeaveFunction(2);
1100
1101         return 0;
1102 }
1103
1104 static void ip_vs_dest_trash_expire(unsigned long data)
1105 {
1106         struct net *net = (struct net *) data;
1107         struct netns_ipvs *ipvs = net_ipvs(net);
1108         struct ip_vs_dest *dest, *next;
1109         unsigned long now = jiffies;
1110
1111         spin_lock(&ipvs->dest_trash_lock);
1112         list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1113                 if (atomic_read(&dest->refcnt) > 0)
1114                         continue;
1115                 if (dest->idle_start) {
1116                         if (time_before(now, dest->idle_start +
1117                                              IP_VS_DEST_TRASH_PERIOD))
1118                                 continue;
1119                 } else {
1120                         dest->idle_start = max(1UL, now);
1121                         continue;
1122                 }
1123                 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1124                               dest->vfwmark,
1125                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1126                               ntohs(dest->port));
1127                 list_del(&dest->t_list);
1128                 ip_vs_dest_free(dest);
1129         }
1130         if (!list_empty(&ipvs->dest_trash))
1131                 mod_timer(&ipvs->dest_trash_timer,
1132                           jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1133         spin_unlock(&ipvs->dest_trash_lock);
1134 }
1135
1136 /*
1137  *      Add a service into the service hash table
1138  */
1139 static int
1140 ip_vs_add_service(struct net *net, struct ip_vs_service_user_kern *u,
1141                   struct ip_vs_service **svc_p)
1142 {
1143         int ret = 0, i;
1144         struct ip_vs_scheduler *sched = NULL;
1145         struct ip_vs_pe *pe = NULL;
1146         struct ip_vs_service *svc = NULL;
1147         struct netns_ipvs *ipvs = net_ipvs(net);
1148
1149         /* increase the module use count */
1150         ip_vs_use_count_inc();
1151
1152         /* Lookup the scheduler by 'u->sched_name' */
1153         sched = ip_vs_scheduler_get(u->sched_name);
1154         if (sched == NULL) {
1155                 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1156                 ret = -ENOENT;
1157                 goto out_err;
1158         }
1159
1160         if (u->pe_name && *u->pe_name) {
1161                 pe = ip_vs_pe_getbyname(u->pe_name);
1162                 if (pe == NULL) {
1163                         pr_info("persistence engine module ip_vs_pe_%s "
1164                                 "not found\n", u->pe_name);
1165                         ret = -ENOENT;
1166                         goto out_err;
1167                 }
1168         }
1169
1170 #ifdef CONFIG_IP_VS_IPV6
1171         if (u->af == AF_INET6) {
1172                 __u32 plen = (__force __u32) u->netmask;
1173
1174                 if (plen < 1 || plen > 128) {
1175                         ret = -EINVAL;
1176                         goto out_err;
1177                 }
1178         }
1179 #endif
1180
1181         svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1182         if (svc == NULL) {
1183                 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1184                 ret = -ENOMEM;
1185                 goto out_err;
1186         }
1187         svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1188         if (!svc->stats.cpustats) {
1189                 ret = -ENOMEM;
1190                 goto out_err;
1191         }
1192
1193         for_each_possible_cpu(i) {
1194                 struct ip_vs_cpu_stats *ip_vs_stats;
1195                 ip_vs_stats = per_cpu_ptr(svc->stats.cpustats, i);
1196                 u64_stats_init(&ip_vs_stats->syncp);
1197         }
1198
1199
1200         /* I'm the first user of the service */
1201         atomic_set(&svc->refcnt, 0);
1202
1203         svc->af = u->af;
1204         svc->protocol = u->protocol;
1205         ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1206         svc->port = u->port;
1207         svc->fwmark = u->fwmark;
1208         svc->flags = u->flags;
1209         svc->timeout = u->timeout * HZ;
1210         svc->netmask = u->netmask;
1211         svc->net = net;
1212
1213         INIT_LIST_HEAD(&svc->destinations);
1214         spin_lock_init(&svc->sched_lock);
1215         spin_lock_init(&svc->stats.lock);
1216
1217         /* Bind the scheduler */
1218         ret = ip_vs_bind_scheduler(svc, sched);
1219         if (ret)
1220                 goto out_err;
1221         sched = NULL;
1222
1223         /* Bind the ct retriever */
1224         RCU_INIT_POINTER(svc->pe, pe);
1225         pe = NULL;
1226
1227         /* Update the virtual service counters */
1228         if (svc->port == FTPPORT)
1229                 atomic_inc(&ipvs->ftpsvc_counter);
1230         else if (svc->port == 0)
1231                 atomic_inc(&ipvs->nullsvc_counter);
1232
1233         ip_vs_start_estimator(net, &svc->stats);
1234
1235         /* Count only IPv4 services for old get/setsockopt interface */
1236         if (svc->af == AF_INET)
1237                 ipvs->num_services++;
1238
1239         /* Hash the service into the service table */
1240         ip_vs_svc_hash(svc);
1241
1242         *svc_p = svc;
1243         /* Now there is a service - full throttle */
1244         ipvs->enable = 1;
1245         return 0;
1246
1247
1248  out_err:
1249         if (svc != NULL) {
1250                 ip_vs_unbind_scheduler(svc, sched);
1251                 ip_vs_service_free(svc);
1252         }
1253         ip_vs_scheduler_put(sched);
1254         ip_vs_pe_put(pe);
1255
1256         /* decrease the module use count */
1257         ip_vs_use_count_dec();
1258
1259         return ret;
1260 }
1261
1262
1263 /*
1264  *      Edit a service and bind it with a new scheduler
1265  */
1266 static int
1267 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1268 {
1269         struct ip_vs_scheduler *sched, *old_sched;
1270         struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1271         int ret = 0;
1272
1273         /*
1274          * Lookup the scheduler, by 'u->sched_name'
1275          */
1276         sched = ip_vs_scheduler_get(u->sched_name);
1277         if (sched == NULL) {
1278                 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1279                 return -ENOENT;
1280         }
1281         old_sched = sched;
1282
1283         if (u->pe_name && *u->pe_name) {
1284                 pe = ip_vs_pe_getbyname(u->pe_name);
1285                 if (pe == NULL) {
1286                         pr_info("persistence engine module ip_vs_pe_%s "
1287                                 "not found\n", u->pe_name);
1288                         ret = -ENOENT;
1289                         goto out;
1290                 }
1291                 old_pe = pe;
1292         }
1293
1294 #ifdef CONFIG_IP_VS_IPV6
1295         if (u->af == AF_INET6) {
1296                 __u32 plen = (__force __u32) u->netmask;
1297
1298                 if (plen < 1 || plen > 128) {
1299                         ret = -EINVAL;
1300                         goto out;
1301                 }
1302         }
1303 #endif
1304
1305         old_sched = rcu_dereference_protected(svc->scheduler, 1);
1306         if (sched != old_sched) {
1307                 /* Bind the new scheduler */
1308                 ret = ip_vs_bind_scheduler(svc, sched);
1309                 if (ret) {
1310                         old_sched = sched;
1311                         goto out;
1312                 }
1313                 /* Unbind the old scheduler on success */
1314                 ip_vs_unbind_scheduler(svc, old_sched);
1315         }
1316
1317         /*
1318          * Set the flags and timeout value
1319          */
1320         svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1321         svc->timeout = u->timeout * HZ;
1322         svc->netmask = u->netmask;
1323
1324         old_pe = rcu_dereference_protected(svc->pe, 1);
1325         if (pe != old_pe)
1326                 rcu_assign_pointer(svc->pe, pe);
1327
1328 out:
1329         ip_vs_scheduler_put(old_sched);
1330         ip_vs_pe_put(old_pe);
1331         return ret;
1332 }
1333
1334 /*
1335  *      Delete a service from the service list
1336  *      - The service must be unlinked, unlocked and not referenced!
1337  *      - We are called under _bh lock
1338  */
1339 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1340 {
1341         struct ip_vs_dest *dest, *nxt;
1342         struct ip_vs_scheduler *old_sched;
1343         struct ip_vs_pe *old_pe;
1344         struct netns_ipvs *ipvs = net_ipvs(svc->net);
1345
1346         pr_info("%s: enter\n", __func__);
1347
1348         /* Count only IPv4 services for old get/setsockopt interface */
1349         if (svc->af == AF_INET)
1350                 ipvs->num_services--;
1351
1352         ip_vs_stop_estimator(svc->net, &svc->stats);
1353
1354         /* Unbind scheduler */
1355         old_sched = rcu_dereference_protected(svc->scheduler, 1);
1356         ip_vs_unbind_scheduler(svc, old_sched);
1357         ip_vs_scheduler_put(old_sched);
1358
1359         /* Unbind persistence engine, keep svc->pe */
1360         old_pe = rcu_dereference_protected(svc->pe, 1);
1361         ip_vs_pe_put(old_pe);
1362
1363         /*
1364          *    Unlink the whole destination list
1365          */
1366         list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1367                 __ip_vs_unlink_dest(svc, dest, 0);
1368                 __ip_vs_del_dest(svc->net, dest, cleanup);
1369         }
1370
1371         /*
1372          *    Update the virtual service counters
1373          */
1374         if (svc->port == FTPPORT)
1375                 atomic_dec(&ipvs->ftpsvc_counter);
1376         else if (svc->port == 0)
1377                 atomic_dec(&ipvs->nullsvc_counter);
1378
1379         /*
1380          *    Free the service if nobody refers to it
1381          */
1382         __ip_vs_svc_put(svc, true);
1383
1384         /* decrease the module use count */
1385         ip_vs_use_count_dec();
1386 }
1387
1388 /*
1389  * Unlink a service from list and try to delete it if its refcnt reached 0
1390  */
1391 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1392 {
1393         /* Hold svc to avoid double release from dest_trash */
1394         atomic_inc(&svc->refcnt);
1395         /*
1396          * Unhash it from the service table
1397          */
1398         ip_vs_svc_unhash(svc);
1399
1400         __ip_vs_del_service(svc, cleanup);
1401 }
1402
1403 /*
1404  *      Delete a service from the service list
1405  */
1406 static int ip_vs_del_service(struct ip_vs_service *svc)
1407 {
1408         if (svc == NULL)
1409                 return -EEXIST;
1410         ip_vs_unlink_service(svc, false);
1411
1412         return 0;
1413 }
1414
1415
1416 /*
1417  *      Flush all the virtual services
1418  */
1419 static int ip_vs_flush(struct net *net, bool cleanup)
1420 {
1421         int idx;
1422         struct ip_vs_service *svc;
1423         struct hlist_node *n;
1424
1425         /*
1426          * Flush the service table hashed by <netns,protocol,addr,port>
1427          */
1428         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1429                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1430                                           s_list) {
1431                         if (net_eq(svc->net, net))
1432                                 ip_vs_unlink_service(svc, cleanup);
1433                 }
1434         }
1435
1436         /*
1437          * Flush the service table hashed by fwmark
1438          */
1439         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1440                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1441                                           f_list) {
1442                         if (net_eq(svc->net, net))
1443                                 ip_vs_unlink_service(svc, cleanup);
1444                 }
1445         }
1446
1447         return 0;
1448 }
1449
1450 /*
1451  *      Delete service by {netns} in the service table.
1452  *      Called by __ip_vs_cleanup()
1453  */
1454 void ip_vs_service_net_cleanup(struct net *net)
1455 {
1456         EnterFunction(2);
1457         /* Check for "full" addressed entries */
1458         mutex_lock(&__ip_vs_mutex);
1459         ip_vs_flush(net, true);
1460         mutex_unlock(&__ip_vs_mutex);
1461         LeaveFunction(2);
1462 }
1463
1464 /* Put all references for device (dst_cache) */
1465 static inline void
1466 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1467 {
1468         struct ip_vs_dest_dst *dest_dst;
1469
1470         spin_lock_bh(&dest->dst_lock);
1471         dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1472         if (dest_dst && dest_dst->dst_cache->dev == dev) {
1473                 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1474                               dev->name,
1475                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1476                               ntohs(dest->port),
1477                               atomic_read(&dest->refcnt));
1478                 __ip_vs_dst_cache_reset(dest);
1479         }
1480         spin_unlock_bh(&dest->dst_lock);
1481
1482 }
1483 /* Netdev event receiver
1484  * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1485  */
1486 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1487                            void *ptr)
1488 {
1489         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1490         struct net *net = dev_net(dev);
1491         struct netns_ipvs *ipvs = net_ipvs(net);
1492         struct ip_vs_service *svc;
1493         struct ip_vs_dest *dest;
1494         unsigned int idx;
1495
1496         if (event != NETDEV_DOWN || !ipvs)
1497                 return NOTIFY_DONE;
1498         IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1499         EnterFunction(2);
1500         mutex_lock(&__ip_vs_mutex);
1501         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1502                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1503                         if (net_eq(svc->net, net)) {
1504                                 list_for_each_entry(dest, &svc->destinations,
1505                                                     n_list) {
1506                                         ip_vs_forget_dev(dest, dev);
1507                                 }
1508                         }
1509                 }
1510
1511                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1512                         if (net_eq(svc->net, net)) {
1513                                 list_for_each_entry(dest, &svc->destinations,
1514                                                     n_list) {
1515                                         ip_vs_forget_dev(dest, dev);
1516                                 }
1517                         }
1518
1519                 }
1520         }
1521
1522         spin_lock_bh(&ipvs->dest_trash_lock);
1523         list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1524                 ip_vs_forget_dev(dest, dev);
1525         }
1526         spin_unlock_bh(&ipvs->dest_trash_lock);
1527         mutex_unlock(&__ip_vs_mutex);
1528         LeaveFunction(2);
1529         return NOTIFY_DONE;
1530 }
1531
1532 /*
1533  *      Zero counters in a service or all services
1534  */
1535 static int ip_vs_zero_service(struct ip_vs_service *svc)
1536 {
1537         struct ip_vs_dest *dest;
1538
1539         list_for_each_entry(dest, &svc->destinations, n_list) {
1540                 ip_vs_zero_stats(&dest->stats);
1541         }
1542         ip_vs_zero_stats(&svc->stats);
1543         return 0;
1544 }
1545
1546 static int ip_vs_zero_all(struct net *net)
1547 {
1548         int idx;
1549         struct ip_vs_service *svc;
1550
1551         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1552                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1553                         if (net_eq(svc->net, net))
1554                                 ip_vs_zero_service(svc);
1555                 }
1556         }
1557
1558         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1559                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1560                         if (net_eq(svc->net, net))
1561                                 ip_vs_zero_service(svc);
1562                 }
1563         }
1564
1565         ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1566         return 0;
1567 }
1568
1569 #ifdef CONFIG_SYSCTL
1570
1571 static int zero;
1572 static int three = 3;
1573
1574 static int
1575 proc_do_defense_mode(struct ctl_table *table, int write,
1576                      void __user *buffer, size_t *lenp, loff_t *ppos)
1577 {
1578         struct net *net = current->nsproxy->net_ns;
1579         int *valp = table->data;
1580         int val = *valp;
1581         int rc;
1582
1583         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1584         if (write && (*valp != val)) {
1585                 if ((*valp < 0) || (*valp > 3)) {
1586                         /* Restore the correct value */
1587                         *valp = val;
1588                 } else {
1589                         update_defense_level(net_ipvs(net));
1590                 }
1591         }
1592         return rc;
1593 }
1594
1595 static int
1596 proc_do_sync_threshold(struct ctl_table *table, int write,
1597                        void __user *buffer, size_t *lenp, loff_t *ppos)
1598 {
1599         int *valp = table->data;
1600         int val[2];
1601         int rc;
1602
1603         /* backup the value first */
1604         memcpy(val, valp, sizeof(val));
1605
1606         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1607         if (write && (valp[0] < 0 || valp[1] < 0 ||
1608             (valp[0] >= valp[1] && valp[1]))) {
1609                 /* Restore the correct value */
1610                 memcpy(valp, val, sizeof(val));
1611         }
1612         return rc;
1613 }
1614
1615 static int
1616 proc_do_sync_mode(struct ctl_table *table, int write,
1617                      void __user *buffer, size_t *lenp, loff_t *ppos)
1618 {
1619         int *valp = table->data;
1620         int val = *valp;
1621         int rc;
1622
1623         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1624         if (write && (*valp != val)) {
1625                 if ((*valp < 0) || (*valp > 1)) {
1626                         /* Restore the correct value */
1627                         *valp = val;
1628                 }
1629         }
1630         return rc;
1631 }
1632
1633 static int
1634 proc_do_sync_ports(struct ctl_table *table, int write,
1635                    void __user *buffer, size_t *lenp, loff_t *ppos)
1636 {
1637         int *valp = table->data;
1638         int val = *valp;
1639         int rc;
1640
1641         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1642         if (write && (*valp != val)) {
1643                 if (*valp < 1 || !is_power_of_2(*valp)) {
1644                         /* Restore the correct value */
1645                         *valp = val;
1646                 }
1647         }
1648         return rc;
1649 }
1650
1651 /*
1652  *      IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1653  *      Do not change order or insert new entries without
1654  *      align with netns init in ip_vs_control_net_init()
1655  */
1656
1657 static struct ctl_table vs_vars[] = {
1658         {
1659                 .procname       = "amemthresh",
1660                 .maxlen         = sizeof(int),
1661                 .mode           = 0644,
1662                 .proc_handler   = proc_dointvec,
1663         },
1664         {
1665                 .procname       = "am_droprate",
1666                 .maxlen         = sizeof(int),
1667                 .mode           = 0644,
1668                 .proc_handler   = proc_dointvec,
1669         },
1670         {
1671                 .procname       = "drop_entry",
1672                 .maxlen         = sizeof(int),
1673                 .mode           = 0644,
1674                 .proc_handler   = proc_do_defense_mode,
1675         },
1676         {
1677                 .procname       = "drop_packet",
1678                 .maxlen         = sizeof(int),
1679                 .mode           = 0644,
1680                 .proc_handler   = proc_do_defense_mode,
1681         },
1682 #ifdef CONFIG_IP_VS_NFCT
1683         {
1684                 .procname       = "conntrack",
1685                 .maxlen         = sizeof(int),
1686                 .mode           = 0644,
1687                 .proc_handler   = &proc_dointvec,
1688         },
1689 #endif
1690         {
1691                 .procname       = "secure_tcp",
1692                 .maxlen         = sizeof(int),
1693                 .mode           = 0644,
1694                 .proc_handler   = proc_do_defense_mode,
1695         },
1696         {
1697                 .procname       = "snat_reroute",
1698                 .maxlen         = sizeof(int),
1699                 .mode           = 0644,
1700                 .proc_handler   = &proc_dointvec,
1701         },
1702         {
1703                 .procname       = "sync_version",
1704                 .maxlen         = sizeof(int),
1705                 .mode           = 0644,
1706                 .proc_handler   = &proc_do_sync_mode,
1707         },
1708         {
1709                 .procname       = "sync_ports",
1710                 .maxlen         = sizeof(int),
1711                 .mode           = 0644,
1712                 .proc_handler   = &proc_do_sync_ports,
1713         },
1714         {
1715                 .procname       = "sync_persist_mode",
1716                 .maxlen         = sizeof(int),
1717                 .mode           = 0644,
1718                 .proc_handler   = proc_dointvec,
1719         },
1720         {
1721                 .procname       = "sync_qlen_max",
1722                 .maxlen         = sizeof(unsigned long),
1723                 .mode           = 0644,
1724                 .proc_handler   = proc_doulongvec_minmax,
1725         },
1726         {
1727                 .procname       = "sync_sock_size",
1728                 .maxlen         = sizeof(int),
1729                 .mode           = 0644,
1730                 .proc_handler   = proc_dointvec,
1731         },
1732         {
1733                 .procname       = "cache_bypass",
1734                 .maxlen         = sizeof(int),
1735                 .mode           = 0644,
1736                 .proc_handler   = proc_dointvec,
1737         },
1738         {
1739                 .procname       = "expire_nodest_conn",
1740                 .maxlen         = sizeof(int),
1741                 .mode           = 0644,
1742                 .proc_handler   = proc_dointvec,
1743         },
1744         {
1745                 .procname       = "sloppy_tcp",
1746                 .maxlen         = sizeof(int),
1747                 .mode           = 0644,
1748                 .proc_handler   = proc_dointvec,
1749         },
1750         {
1751                 .procname       = "sloppy_sctp",
1752                 .maxlen         = sizeof(int),
1753                 .mode           = 0644,
1754                 .proc_handler   = proc_dointvec,
1755         },
1756         {
1757                 .procname       = "expire_quiescent_template",
1758                 .maxlen         = sizeof(int),
1759                 .mode           = 0644,
1760                 .proc_handler   = proc_dointvec,
1761         },
1762         {
1763                 .procname       = "sync_threshold",
1764                 .maxlen         =
1765                         sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1766                 .mode           = 0644,
1767                 .proc_handler   = proc_do_sync_threshold,
1768         },
1769         {
1770                 .procname       = "sync_refresh_period",
1771                 .maxlen         = sizeof(int),
1772                 .mode           = 0644,
1773                 .proc_handler   = proc_dointvec_jiffies,
1774         },
1775         {
1776                 .procname       = "sync_retries",
1777                 .maxlen         = sizeof(int),
1778                 .mode           = 0644,
1779                 .proc_handler   = proc_dointvec_minmax,
1780                 .extra1         = &zero,
1781                 .extra2         = &three,
1782         },
1783         {
1784                 .procname       = "nat_icmp_send",
1785                 .maxlen         = sizeof(int),
1786                 .mode           = 0644,
1787                 .proc_handler   = proc_dointvec,
1788         },
1789         {
1790                 .procname       = "pmtu_disc",
1791                 .maxlen         = sizeof(int),
1792                 .mode           = 0644,
1793                 .proc_handler   = proc_dointvec,
1794         },
1795         {
1796                 .procname       = "backup_only",
1797                 .maxlen         = sizeof(int),
1798                 .mode           = 0644,
1799                 .proc_handler   = proc_dointvec,
1800         },
1801 #ifdef CONFIG_IP_VS_DEBUG
1802         {
1803                 .procname       = "debug_level",
1804                 .data           = &sysctl_ip_vs_debug_level,
1805                 .maxlen         = sizeof(int),
1806                 .mode           = 0644,
1807                 .proc_handler   = proc_dointvec,
1808         },
1809 #endif
1810 #if 0
1811         {
1812                 .procname       = "timeout_established",
1813                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1814                 .maxlen         = sizeof(int),
1815                 .mode           = 0644,
1816                 .proc_handler   = proc_dointvec_jiffies,
1817         },
1818         {
1819                 .procname       = "timeout_synsent",
1820                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1821                 .maxlen         = sizeof(int),
1822                 .mode           = 0644,
1823                 .proc_handler   = proc_dointvec_jiffies,
1824         },
1825         {
1826                 .procname       = "timeout_synrecv",
1827                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1828                 .maxlen         = sizeof(int),
1829                 .mode           = 0644,
1830                 .proc_handler   = proc_dointvec_jiffies,
1831         },
1832         {
1833                 .procname       = "timeout_finwait",
1834                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1835                 .maxlen         = sizeof(int),
1836                 .mode           = 0644,
1837                 .proc_handler   = proc_dointvec_jiffies,
1838         },
1839         {
1840                 .procname       = "timeout_timewait",
1841                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1842                 .maxlen         = sizeof(int),
1843                 .mode           = 0644,
1844                 .proc_handler   = proc_dointvec_jiffies,
1845         },
1846         {
1847                 .procname       = "timeout_close",
1848                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1849                 .maxlen         = sizeof(int),
1850                 .mode           = 0644,
1851                 .proc_handler   = proc_dointvec_jiffies,
1852         },
1853         {
1854                 .procname       = "timeout_closewait",
1855                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1856                 .maxlen         = sizeof(int),
1857                 .mode           = 0644,
1858                 .proc_handler   = proc_dointvec_jiffies,
1859         },
1860         {
1861                 .procname       = "timeout_lastack",
1862                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1863                 .maxlen         = sizeof(int),
1864                 .mode           = 0644,
1865                 .proc_handler   = proc_dointvec_jiffies,
1866         },
1867         {
1868                 .procname       = "timeout_listen",
1869                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1870                 .maxlen         = sizeof(int),
1871                 .mode           = 0644,
1872                 .proc_handler   = proc_dointvec_jiffies,
1873         },
1874         {
1875                 .procname       = "timeout_synack",
1876                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1877                 .maxlen         = sizeof(int),
1878                 .mode           = 0644,
1879                 .proc_handler   = proc_dointvec_jiffies,
1880         },
1881         {
1882                 .procname       = "timeout_udp",
1883                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1884                 .maxlen         = sizeof(int),
1885                 .mode           = 0644,
1886                 .proc_handler   = proc_dointvec_jiffies,
1887         },
1888         {
1889                 .procname       = "timeout_icmp",
1890                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1891                 .maxlen         = sizeof(int),
1892                 .mode           = 0644,
1893                 .proc_handler   = proc_dointvec_jiffies,
1894         },
1895 #endif
1896         { }
1897 };
1898
1899 #endif
1900
1901 #ifdef CONFIG_PROC_FS
1902
1903 struct ip_vs_iter {
1904         struct seq_net_private p;  /* Do not move this, netns depends upon it*/
1905         struct hlist_head *table;
1906         int bucket;
1907 };
1908
1909 /*
1910  *      Write the contents of the VS rule table to a PROCfs file.
1911  *      (It is kept just for backward compatibility)
1912  */
1913 static inline const char *ip_vs_fwd_name(unsigned int flags)
1914 {
1915         switch (flags & IP_VS_CONN_F_FWD_MASK) {
1916         case IP_VS_CONN_F_LOCALNODE:
1917                 return "Local";
1918         case IP_VS_CONN_F_TUNNEL:
1919                 return "Tunnel";
1920         case IP_VS_CONN_F_DROUTE:
1921                 return "Route";
1922         default:
1923                 return "Masq";
1924         }
1925 }
1926
1927
1928 /* Get the Nth entry in the two lists */
1929 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1930 {
1931         struct net *net = seq_file_net(seq);
1932         struct ip_vs_iter *iter = seq->private;
1933         int idx;
1934         struct ip_vs_service *svc;
1935
1936         /* look in hash by protocol */
1937         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1938                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1939                         if (net_eq(svc->net, net) && pos-- == 0) {
1940                                 iter->table = ip_vs_svc_table;
1941                                 iter->bucket = idx;
1942                                 return svc;
1943                         }
1944                 }
1945         }
1946
1947         /* keep looking in fwmark */
1948         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1949                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1950                                          f_list) {
1951                         if (net_eq(svc->net, net) && pos-- == 0) {
1952                                 iter->table = ip_vs_svc_fwm_table;
1953                                 iter->bucket = idx;
1954                                 return svc;
1955                         }
1956                 }
1957         }
1958
1959         return NULL;
1960 }
1961
1962 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1963         __acquires(RCU)
1964 {
1965         rcu_read_lock();
1966         return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1967 }
1968
1969
1970 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1971 {
1972         struct hlist_node *e;
1973         struct ip_vs_iter *iter;
1974         struct ip_vs_service *svc;
1975
1976         ++*pos;
1977         if (v == SEQ_START_TOKEN)
1978                 return ip_vs_info_array(seq,0);
1979
1980         svc = v;
1981         iter = seq->private;
1982
1983         if (iter->table == ip_vs_svc_table) {
1984                 /* next service in table hashed by protocol */
1985                 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
1986                 if (e)
1987                         return hlist_entry(e, struct ip_vs_service, s_list);
1988
1989                 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1990                         hlist_for_each_entry_rcu(svc,
1991                                                  &ip_vs_svc_table[iter->bucket],
1992                                                  s_list) {
1993                                 return svc;
1994                         }
1995                 }
1996
1997                 iter->table = ip_vs_svc_fwm_table;
1998                 iter->bucket = -1;
1999                 goto scan_fwmark;
2000         }
2001
2002         /* next service in hashed by fwmark */
2003         e = rcu_dereference(hlist_next_rcu(&svc->f_list));
2004         if (e)
2005                 return hlist_entry(e, struct ip_vs_service, f_list);
2006
2007  scan_fwmark:
2008         while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2009                 hlist_for_each_entry_rcu(svc,
2010                                          &ip_vs_svc_fwm_table[iter->bucket],
2011                                          f_list)
2012                         return svc;
2013         }
2014
2015         return NULL;
2016 }
2017
2018 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2019         __releases(RCU)
2020 {
2021         rcu_read_unlock();
2022 }
2023
2024
2025 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2026 {
2027         if (v == SEQ_START_TOKEN) {
2028                 seq_printf(seq,
2029                         "IP Virtual Server version %d.%d.%d (size=%d)\n",
2030                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2031                 seq_puts(seq,
2032                          "Prot LocalAddress:Port Scheduler Flags\n");
2033                 seq_puts(seq,
2034                          "  -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2035         } else {
2036                 const struct ip_vs_service *svc = v;
2037                 const struct ip_vs_iter *iter = seq->private;
2038                 const struct ip_vs_dest *dest;
2039                 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2040
2041                 if (iter->table == ip_vs_svc_table) {
2042 #ifdef CONFIG_IP_VS_IPV6
2043                         if (svc->af == AF_INET6)
2044                                 seq_printf(seq, "%s  [%pI6]:%04X %s ",
2045                                            ip_vs_proto_name(svc->protocol),
2046                                            &svc->addr.in6,
2047                                            ntohs(svc->port),
2048                                            sched->name);
2049                         else
2050 #endif
2051                                 seq_printf(seq, "%s  %08X:%04X %s %s ",
2052                                            ip_vs_proto_name(svc->protocol),
2053                                            ntohl(svc->addr.ip),
2054                                            ntohs(svc->port),
2055                                            sched->name,
2056                                            (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2057                 } else {
2058                         seq_printf(seq, "FWM  %08X %s %s",
2059                                    svc->fwmark, sched->name,
2060                                    (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2061                 }
2062
2063                 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2064                         seq_printf(seq, "persistent %d %08X\n",
2065                                 svc->timeout,
2066                                 ntohl(svc->netmask));
2067                 else
2068                         seq_putc(seq, '\n');
2069
2070                 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2071 #ifdef CONFIG_IP_VS_IPV6
2072                         if (dest->af == AF_INET6)
2073                                 seq_printf(seq,
2074                                            "  -> [%pI6]:%04X"
2075                                            "      %-7s %-6d %-10d %-10d\n",
2076                                            &dest->addr.in6,
2077                                            ntohs(dest->port),
2078                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2079                                            atomic_read(&dest->weight),
2080                                            atomic_read(&dest->activeconns),
2081                                            atomic_read(&dest->inactconns));
2082                         else
2083 #endif
2084                                 seq_printf(seq,
2085                                            "  -> %08X:%04X      "
2086                                            "%-7s %-6d %-10d %-10d\n",
2087                                            ntohl(dest->addr.ip),
2088                                            ntohs(dest->port),
2089                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2090                                            atomic_read(&dest->weight),
2091                                            atomic_read(&dest->activeconns),
2092                                            atomic_read(&dest->inactconns));
2093
2094                 }
2095         }
2096         return 0;
2097 }
2098
2099 static const struct seq_operations ip_vs_info_seq_ops = {
2100         .start = ip_vs_info_seq_start,
2101         .next  = ip_vs_info_seq_next,
2102         .stop  = ip_vs_info_seq_stop,
2103         .show  = ip_vs_info_seq_show,
2104 };
2105
2106 static int ip_vs_info_open(struct inode *inode, struct file *file)
2107 {
2108         return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2109                         sizeof(struct ip_vs_iter));
2110 }
2111
2112 static const struct file_operations ip_vs_info_fops = {
2113         .owner   = THIS_MODULE,
2114         .open    = ip_vs_info_open,
2115         .read    = seq_read,
2116         .llseek  = seq_lseek,
2117         .release = seq_release_net,
2118 };
2119
2120 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2121 {
2122         struct net *net = seq_file_single_net(seq);
2123         struct ip_vs_stats_user show;
2124
2125 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2126         seq_puts(seq,
2127                  "   Total Incoming Outgoing         Incoming         Outgoing\n");
2128         seq_printf(seq,
2129                    "   Conns  Packets  Packets            Bytes            Bytes\n");
2130
2131         ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2132         seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", show.conns,
2133                    show.inpkts, show.outpkts,
2134                    (unsigned long long) show.inbytes,
2135                    (unsigned long long) show.outbytes);
2136
2137 /*                 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2138         seq_puts(seq,
2139                    " Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2140         seq_printf(seq, "%8X %8X %8X %16X %16X\n",
2141                         show.cps, show.inpps, show.outpps,
2142                         show.inbps, show.outbps);
2143
2144         return 0;
2145 }
2146
2147 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2148 {
2149         return single_open_net(inode, file, ip_vs_stats_show);
2150 }
2151
2152 static const struct file_operations ip_vs_stats_fops = {
2153         .owner = THIS_MODULE,
2154         .open = ip_vs_stats_seq_open,
2155         .read = seq_read,
2156         .llseek = seq_lseek,
2157         .release = single_release_net,
2158 };
2159
2160 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2161 {
2162         struct net *net = seq_file_single_net(seq);
2163         struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2164         struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2165         struct ip_vs_stats_user rates;
2166         int i;
2167
2168 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2169         seq_puts(seq,
2170                  "       Total Incoming Outgoing         Incoming         Outgoing\n");
2171         seq_printf(seq,
2172                    "CPU    Conns  Packets  Packets            Bytes            Bytes\n");
2173
2174         for_each_possible_cpu(i) {
2175                 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2176                 unsigned int start;
2177                 __u64 inbytes, outbytes;
2178
2179                 do {
2180                         start = u64_stats_fetch_begin_bh(&u->syncp);
2181                         inbytes = u->ustats.inbytes;
2182                         outbytes = u->ustats.outbytes;
2183                 } while (u64_stats_fetch_retry_bh(&u->syncp, start));
2184
2185                 seq_printf(seq, "%3X %8X %8X %8X %16LX %16LX\n",
2186                            i, u->ustats.conns, u->ustats.inpkts,
2187                            u->ustats.outpkts, (__u64)inbytes,
2188                            (__u64)outbytes);
2189         }
2190
2191         spin_lock_bh(&tot_stats->lock);
2192
2193         seq_printf(seq, "  ~ %8X %8X %8X %16LX %16LX\n\n",
2194                    tot_stats->ustats.conns, tot_stats->ustats.inpkts,
2195                    tot_stats->ustats.outpkts,
2196                    (unsigned long long) tot_stats->ustats.inbytes,
2197                    (unsigned long long) tot_stats->ustats.outbytes);
2198
2199         ip_vs_read_estimator(&rates, tot_stats);
2200
2201         spin_unlock_bh(&tot_stats->lock);
2202
2203 /*                 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2204         seq_puts(seq,
2205                    "     Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2206         seq_printf(seq, "    %8X %8X %8X %16X %16X\n",
2207                         rates.cps,
2208                         rates.inpps,
2209                         rates.outpps,
2210                         rates.inbps,
2211                         rates.outbps);
2212
2213         return 0;
2214 }
2215
2216 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2217 {
2218         return single_open_net(inode, file, ip_vs_stats_percpu_show);
2219 }
2220
2221 static const struct file_operations ip_vs_stats_percpu_fops = {
2222         .owner = THIS_MODULE,
2223         .open = ip_vs_stats_percpu_seq_open,
2224         .read = seq_read,
2225         .llseek = seq_lseek,
2226         .release = single_release_net,
2227 };
2228 #endif
2229
2230 /*
2231  *      Set timeout values for tcp tcpfin udp in the timeout_table.
2232  */
2233 static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
2234 {
2235 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2236         struct ip_vs_proto_data *pd;
2237 #endif
2238
2239         IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2240                   u->tcp_timeout,
2241                   u->tcp_fin_timeout,
2242                   u->udp_timeout);
2243
2244 #ifdef CONFIG_IP_VS_PROTO_TCP
2245         if (u->tcp_timeout) {
2246                 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2247                 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2248                         = u->tcp_timeout * HZ;
2249         }
2250
2251         if (u->tcp_fin_timeout) {
2252                 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2253                 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2254                         = u->tcp_fin_timeout * HZ;
2255         }
2256 #endif
2257
2258 #ifdef CONFIG_IP_VS_PROTO_UDP
2259         if (u->udp_timeout) {
2260                 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2261                 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2262                         = u->udp_timeout * HZ;
2263         }
2264 #endif
2265         return 0;
2266 }
2267
2268
2269 #define SET_CMDID(cmd)          (cmd - IP_VS_BASE_CTL)
2270 #define SERVICE_ARG_LEN         (sizeof(struct ip_vs_service_user))
2271 #define SVCDEST_ARG_LEN         (sizeof(struct ip_vs_service_user) +    \
2272                                  sizeof(struct ip_vs_dest_user))
2273 #define TIMEOUT_ARG_LEN         (sizeof(struct ip_vs_timeout_user))
2274 #define DAEMON_ARG_LEN          (sizeof(struct ip_vs_daemon_user))
2275 #define MAX_ARG_LEN             SVCDEST_ARG_LEN
2276
2277 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2278         [SET_CMDID(IP_VS_SO_SET_ADD)]           = SERVICE_ARG_LEN,
2279         [SET_CMDID(IP_VS_SO_SET_EDIT)]          = SERVICE_ARG_LEN,
2280         [SET_CMDID(IP_VS_SO_SET_DEL)]           = SERVICE_ARG_LEN,
2281         [SET_CMDID(IP_VS_SO_SET_FLUSH)]         = 0,
2282         [SET_CMDID(IP_VS_SO_SET_ADDDEST)]       = SVCDEST_ARG_LEN,
2283         [SET_CMDID(IP_VS_SO_SET_DELDEST)]       = SVCDEST_ARG_LEN,
2284         [SET_CMDID(IP_VS_SO_SET_EDITDEST)]      = SVCDEST_ARG_LEN,
2285         [SET_CMDID(IP_VS_SO_SET_TIMEOUT)]       = TIMEOUT_ARG_LEN,
2286         [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)]   = DAEMON_ARG_LEN,
2287         [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)]    = DAEMON_ARG_LEN,
2288         [SET_CMDID(IP_VS_SO_SET_ZERO)]          = SERVICE_ARG_LEN,
2289 };
2290
2291 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2292                                   struct ip_vs_service_user *usvc_compat)
2293 {
2294         memset(usvc, 0, sizeof(*usvc));
2295
2296         usvc->af                = AF_INET;
2297         usvc->protocol          = usvc_compat->protocol;
2298         usvc->addr.ip           = usvc_compat->addr;
2299         usvc->port              = usvc_compat->port;
2300         usvc->fwmark            = usvc_compat->fwmark;
2301
2302         /* Deep copy of sched_name is not needed here */
2303         usvc->sched_name        = usvc_compat->sched_name;
2304
2305         usvc->flags             = usvc_compat->flags;
2306         usvc->timeout           = usvc_compat->timeout;
2307         usvc->netmask           = usvc_compat->netmask;
2308 }
2309
2310 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2311                                    struct ip_vs_dest_user *udest_compat)
2312 {
2313         memset(udest, 0, sizeof(*udest));
2314
2315         udest->addr.ip          = udest_compat->addr;
2316         udest->port             = udest_compat->port;
2317         udest->conn_flags       = udest_compat->conn_flags;
2318         udest->weight           = udest_compat->weight;
2319         udest->u_threshold      = udest_compat->u_threshold;
2320         udest->l_threshold      = udest_compat->l_threshold;
2321 }
2322
2323 static int
2324 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2325 {
2326         struct net *net = sock_net(sk);
2327         int ret;
2328         unsigned char arg[MAX_ARG_LEN];
2329         struct ip_vs_service_user *usvc_compat;
2330         struct ip_vs_service_user_kern usvc;
2331         struct ip_vs_service *svc;
2332         struct ip_vs_dest_user *udest_compat;
2333         struct ip_vs_dest_user_kern udest;
2334         struct netns_ipvs *ipvs = net_ipvs(net);
2335
2336         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2337                 return -EPERM;
2338
2339         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2340                 return -EINVAL;
2341         if (len < 0 || len >  MAX_ARG_LEN)
2342                 return -EINVAL;
2343         if (len != set_arglen[SET_CMDID(cmd)]) {
2344                 pr_err("set_ctl: len %u != %u\n",
2345                        len, set_arglen[SET_CMDID(cmd)]);
2346                 return -EINVAL;
2347         }
2348
2349         if (copy_from_user(arg, user, len) != 0)
2350                 return -EFAULT;
2351
2352         /* increase the module use count */
2353         ip_vs_use_count_inc();
2354
2355         /* Handle daemons since they have another lock */
2356         if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2357             cmd == IP_VS_SO_SET_STOPDAEMON) {
2358                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2359
2360                 if (mutex_lock_interruptible(&ipvs->sync_mutex)) {
2361                         ret = -ERESTARTSYS;
2362                         goto out_dec;
2363                 }
2364                 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2365                         ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2366                                                 dm->syncid);
2367                 else
2368                         ret = stop_sync_thread(net, dm->state);
2369                 mutex_unlock(&ipvs->sync_mutex);
2370                 goto out_dec;
2371         }
2372
2373         if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2374                 ret = -ERESTARTSYS;
2375                 goto out_dec;
2376         }
2377
2378         if (cmd == IP_VS_SO_SET_FLUSH) {
2379                 /* Flush the virtual service */
2380                 ret = ip_vs_flush(net, false);
2381                 goto out_unlock;
2382         } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2383                 /* Set timeout values for (tcp tcpfin udp) */
2384                 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
2385                 goto out_unlock;
2386         }
2387
2388         usvc_compat = (struct ip_vs_service_user *)arg;
2389         udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2390
2391         /* We only use the new structs internally, so copy userspace compat
2392          * structs to extended internal versions */
2393         ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2394         ip_vs_copy_udest_compat(&udest, udest_compat);
2395
2396         if (cmd == IP_VS_SO_SET_ZERO) {
2397                 /* if no service address is set, zero counters in all */
2398                 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2399                         ret = ip_vs_zero_all(net);
2400                         goto out_unlock;
2401                 }
2402         }
2403
2404         /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2405         if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2406             usvc.protocol != IPPROTO_SCTP) {
2407                 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2408                        usvc.protocol, &usvc.addr.ip,
2409                        ntohs(usvc.port), usvc.sched_name);
2410                 ret = -EFAULT;
2411                 goto out_unlock;
2412         }
2413
2414         /* Lookup the exact service by <protocol, addr, port> or fwmark */
2415         rcu_read_lock();
2416         if (usvc.fwmark == 0)
2417                 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
2418                                            &usvc.addr, usvc.port);
2419         else
2420                 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
2421         rcu_read_unlock();
2422
2423         if (cmd != IP_VS_SO_SET_ADD
2424             && (svc == NULL || svc->protocol != usvc.protocol)) {
2425                 ret = -ESRCH;
2426                 goto out_unlock;
2427         }
2428
2429         switch (cmd) {
2430         case IP_VS_SO_SET_ADD:
2431                 if (svc != NULL)
2432                         ret = -EEXIST;
2433                 else
2434                         ret = ip_vs_add_service(net, &usvc, &svc);
2435                 break;
2436         case IP_VS_SO_SET_EDIT:
2437                 ret = ip_vs_edit_service(svc, &usvc);
2438                 break;
2439         case IP_VS_SO_SET_DEL:
2440                 ret = ip_vs_del_service(svc);
2441                 if (!ret)
2442                         goto out_unlock;
2443                 break;
2444         case IP_VS_SO_SET_ZERO:
2445                 ret = ip_vs_zero_service(svc);
2446                 break;
2447         case IP_VS_SO_SET_ADDDEST:
2448                 ret = ip_vs_add_dest(svc, &udest);
2449                 break;
2450         case IP_VS_SO_SET_EDITDEST:
2451                 ret = ip_vs_edit_dest(svc, &udest);
2452                 break;
2453         case IP_VS_SO_SET_DELDEST:
2454                 ret = ip_vs_del_dest(svc, &udest);
2455                 break;
2456         default:
2457                 ret = -EINVAL;
2458         }
2459
2460   out_unlock:
2461         mutex_unlock(&__ip_vs_mutex);
2462   out_dec:
2463         /* decrease the module use count */
2464         ip_vs_use_count_dec();
2465
2466         return ret;
2467 }
2468
2469
2470 static void
2471 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2472 {
2473         struct ip_vs_scheduler *sched;
2474
2475         sched = rcu_dereference_protected(src->scheduler, 1);
2476         dst->protocol = src->protocol;
2477         dst->addr = src->addr.ip;
2478         dst->port = src->port;
2479         dst->fwmark = src->fwmark;
2480         strlcpy(dst->sched_name, sched->name, sizeof(dst->sched_name));
2481         dst->flags = src->flags;
2482         dst->timeout = src->timeout / HZ;
2483         dst->netmask = src->netmask;
2484         dst->num_dests = src->num_dests;
2485         ip_vs_copy_stats(&dst->stats, &src->stats);
2486 }
2487
2488 static inline int
2489 __ip_vs_get_service_entries(struct net *net,
2490                             const struct ip_vs_get_services *get,
2491                             struct ip_vs_get_services __user *uptr)
2492 {
2493         int idx, count=0;
2494         struct ip_vs_service *svc;
2495         struct ip_vs_service_entry entry;
2496         int ret = 0;
2497
2498         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2499                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2500                         /* Only expose IPv4 entries to old interface */
2501                         if (svc->af != AF_INET || !net_eq(svc->net, net))
2502                                 continue;
2503
2504                         if (count >= get->num_services)
2505                                 goto out;
2506                         memset(&entry, 0, sizeof(entry));
2507                         ip_vs_copy_service(&entry, svc);
2508                         if (copy_to_user(&uptr->entrytable[count],
2509                                          &entry, sizeof(entry))) {
2510                                 ret = -EFAULT;
2511                                 goto out;
2512                         }
2513                         count++;
2514                 }
2515         }
2516
2517         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2518                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2519                         /* Only expose IPv4 entries to old interface */
2520                         if (svc->af != AF_INET || !net_eq(svc->net, net))
2521                                 continue;
2522
2523                         if (count >= get->num_services)
2524                                 goto out;
2525                         memset(&entry, 0, sizeof(entry));
2526                         ip_vs_copy_service(&entry, svc);
2527                         if (copy_to_user(&uptr->entrytable[count],
2528                                          &entry, sizeof(entry))) {
2529                                 ret = -EFAULT;
2530                                 goto out;
2531                         }
2532                         count++;
2533                 }
2534         }
2535 out:
2536         return ret;
2537 }
2538
2539 static inline int
2540 __ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
2541                          struct ip_vs_get_dests __user *uptr)
2542 {
2543         struct ip_vs_service *svc;
2544         union nf_inet_addr addr = { .ip = get->addr };
2545         int ret = 0;
2546
2547         rcu_read_lock();
2548         if (get->fwmark)
2549                 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
2550         else
2551                 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
2552                                            get->port);
2553         rcu_read_unlock();
2554
2555         if (svc) {
2556                 int count = 0;
2557                 struct ip_vs_dest *dest;
2558                 struct ip_vs_dest_entry entry;
2559
2560                 memset(&entry, 0, sizeof(entry));
2561                 list_for_each_entry(dest, &svc->destinations, n_list) {
2562                         if (count >= get->num_dests)
2563                                 break;
2564
2565                         entry.addr = dest->addr.ip;
2566                         entry.port = dest->port;
2567                         entry.conn_flags = atomic_read(&dest->conn_flags);
2568                         entry.weight = atomic_read(&dest->weight);
2569                         entry.u_threshold = dest->u_threshold;
2570                         entry.l_threshold = dest->l_threshold;
2571                         entry.activeconns = atomic_read(&dest->activeconns);
2572                         entry.inactconns = atomic_read(&dest->inactconns);
2573                         entry.persistconns = atomic_read(&dest->persistconns);
2574                         ip_vs_copy_stats(&entry.stats, &dest->stats);
2575                         if (copy_to_user(&uptr->entrytable[count],
2576                                          &entry, sizeof(entry))) {
2577                                 ret = -EFAULT;
2578                                 break;
2579                         }
2580                         count++;
2581                 }
2582         } else
2583                 ret = -ESRCH;
2584         return ret;
2585 }
2586
2587 static inline void
2588 __ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
2589 {
2590 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2591         struct ip_vs_proto_data *pd;
2592 #endif
2593
2594         memset(u, 0, sizeof (*u));
2595
2596 #ifdef CONFIG_IP_VS_PROTO_TCP
2597         pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2598         u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2599         u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2600 #endif
2601 #ifdef CONFIG_IP_VS_PROTO_UDP
2602         pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2603         u->udp_timeout =
2604                         pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2605 #endif
2606 }
2607
2608
2609 #define GET_CMDID(cmd)          (cmd - IP_VS_BASE_CTL)
2610 #define GET_INFO_ARG_LEN        (sizeof(struct ip_vs_getinfo))
2611 #define GET_SERVICES_ARG_LEN    (sizeof(struct ip_vs_get_services))
2612 #define GET_SERVICE_ARG_LEN     (sizeof(struct ip_vs_service_entry))
2613 #define GET_DESTS_ARG_LEN       (sizeof(struct ip_vs_get_dests))
2614 #define GET_TIMEOUT_ARG_LEN     (sizeof(struct ip_vs_timeout_user))
2615 #define GET_DAEMON_ARG_LEN      (sizeof(struct ip_vs_daemon_user) * 2)
2616
2617 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2618         [GET_CMDID(IP_VS_SO_GET_VERSION)]       = 64,
2619         [GET_CMDID(IP_VS_SO_GET_INFO)]          = GET_INFO_ARG_LEN,
2620         [GET_CMDID(IP_VS_SO_GET_SERVICES)]      = GET_SERVICES_ARG_LEN,
2621         [GET_CMDID(IP_VS_SO_GET_SERVICE)]       = GET_SERVICE_ARG_LEN,
2622         [GET_CMDID(IP_VS_SO_GET_DESTS)]         = GET_DESTS_ARG_LEN,
2623         [GET_CMDID(IP_VS_SO_GET_TIMEOUT)]       = GET_TIMEOUT_ARG_LEN,
2624         [GET_CMDID(IP_VS_SO_GET_DAEMON)]        = GET_DAEMON_ARG_LEN,
2625 };
2626
2627 static int
2628 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2629 {
2630         unsigned char arg[128];
2631         int ret = 0;
2632         unsigned int copylen;
2633         struct net *net = sock_net(sk);
2634         struct netns_ipvs *ipvs = net_ipvs(net);
2635
2636         BUG_ON(!net);
2637         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2638                 return -EPERM;
2639
2640         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2641                 return -EINVAL;
2642
2643         if (*len < get_arglen[GET_CMDID(cmd)]) {
2644                 pr_err("get_ctl: len %u < %u\n",
2645                        *len, get_arglen[GET_CMDID(cmd)]);
2646                 return -EINVAL;
2647         }
2648
2649         copylen = get_arglen[GET_CMDID(cmd)];
2650         if (copylen > 128)
2651                 return -EINVAL;
2652
2653         if (copy_from_user(arg, user, copylen) != 0)
2654                 return -EFAULT;
2655         /*
2656          * Handle daemons first since it has its own locking
2657          */
2658         if (cmd == IP_VS_SO_GET_DAEMON) {
2659                 struct ip_vs_daemon_user d[2];
2660
2661                 memset(&d, 0, sizeof(d));
2662                 if (mutex_lock_interruptible(&ipvs->sync_mutex))
2663                         return -ERESTARTSYS;
2664
2665                 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2666                         d[0].state = IP_VS_STATE_MASTER;
2667                         strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2668                                 sizeof(d[0].mcast_ifn));
2669                         d[0].syncid = ipvs->master_syncid;
2670                 }
2671                 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2672                         d[1].state = IP_VS_STATE_BACKUP;
2673                         strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2674                                 sizeof(d[1].mcast_ifn));
2675                         d[1].syncid = ipvs->backup_syncid;
2676                 }
2677                 if (copy_to_user(user, &d, sizeof(d)) != 0)
2678                         ret = -EFAULT;
2679                 mutex_unlock(&ipvs->sync_mutex);
2680                 return ret;
2681         }
2682
2683         if (mutex_lock_interruptible(&__ip_vs_mutex))
2684                 return -ERESTARTSYS;
2685
2686         switch (cmd) {
2687         case IP_VS_SO_GET_VERSION:
2688         {
2689                 char buf[64];
2690
2691                 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2692                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2693                 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2694                         ret = -EFAULT;
2695                         goto out;
2696                 }
2697                 *len = strlen(buf)+1;
2698         }
2699         break;
2700
2701         case IP_VS_SO_GET_INFO:
2702         {
2703                 struct ip_vs_getinfo info;
2704                 info.version = IP_VS_VERSION_CODE;
2705                 info.size = ip_vs_conn_tab_size;
2706                 info.num_services = ipvs->num_services;
2707                 if (copy_to_user(user, &info, sizeof(info)) != 0)
2708                         ret = -EFAULT;
2709         }
2710         break;
2711
2712         case IP_VS_SO_GET_SERVICES:
2713         {
2714                 struct ip_vs_get_services *get;
2715                 int size;
2716
2717                 get = (struct ip_vs_get_services *)arg;
2718                 size = sizeof(*get) +
2719                         sizeof(struct ip_vs_service_entry) * get->num_services;
2720                 if (*len != size) {
2721                         pr_err("length: %u != %u\n", *len, size);
2722                         ret = -EINVAL;
2723                         goto out;
2724                 }
2725                 ret = __ip_vs_get_service_entries(net, get, user);
2726         }
2727         break;
2728
2729         case IP_VS_SO_GET_SERVICE:
2730         {
2731                 struct ip_vs_service_entry *entry;
2732                 struct ip_vs_service *svc;
2733                 union nf_inet_addr addr;
2734
2735                 entry = (struct ip_vs_service_entry *)arg;
2736                 addr.ip = entry->addr;
2737                 rcu_read_lock();
2738                 if (entry->fwmark)
2739                         svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
2740                 else
2741                         svc = __ip_vs_service_find(net, AF_INET,
2742                                                    entry->protocol, &addr,
2743                                                    entry->port);
2744                 rcu_read_unlock();
2745                 if (svc) {
2746                         ip_vs_copy_service(entry, svc);
2747                         if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2748                                 ret = -EFAULT;
2749                 } else
2750                         ret = -ESRCH;
2751         }
2752         break;
2753
2754         case IP_VS_SO_GET_DESTS:
2755         {
2756                 struct ip_vs_get_dests *get;
2757                 int size;
2758
2759                 get = (struct ip_vs_get_dests *)arg;
2760                 size = sizeof(*get) +
2761                         sizeof(struct ip_vs_dest_entry) * get->num_dests;
2762                 if (*len != size) {
2763                         pr_err("length: %u != %u\n", *len, size);
2764                         ret = -EINVAL;
2765                         goto out;
2766                 }
2767                 ret = __ip_vs_get_dest_entries(net, get, user);
2768         }
2769         break;
2770
2771         case IP_VS_SO_GET_TIMEOUT:
2772         {
2773                 struct ip_vs_timeout_user t;
2774
2775                 __ip_vs_get_timeouts(net, &t);
2776                 if (copy_to_user(user, &t, sizeof(t)) != 0)
2777                         ret = -EFAULT;
2778         }
2779         break;
2780
2781         default:
2782                 ret = -EINVAL;
2783         }
2784
2785 out:
2786         mutex_unlock(&__ip_vs_mutex);
2787         return ret;
2788 }
2789
2790
2791 static struct nf_sockopt_ops ip_vs_sockopts = {
2792         .pf             = PF_INET,
2793         .set_optmin     = IP_VS_BASE_CTL,
2794         .set_optmax     = IP_VS_SO_SET_MAX+1,
2795         .set            = do_ip_vs_set_ctl,
2796         .get_optmin     = IP_VS_BASE_CTL,
2797         .get_optmax     = IP_VS_SO_GET_MAX+1,
2798         .get            = do_ip_vs_get_ctl,
2799         .owner          = THIS_MODULE,
2800 };
2801
2802 /*
2803  * Generic Netlink interface
2804  */
2805
2806 /* IPVS genetlink family */
2807 static struct genl_family ip_vs_genl_family = {
2808         .id             = GENL_ID_GENERATE,
2809         .hdrsize        = 0,
2810         .name           = IPVS_GENL_NAME,
2811         .version        = IPVS_GENL_VERSION,
2812         .maxattr        = IPVS_CMD_MAX,
2813         .netnsok        = true,         /* Make ipvsadm to work on netns */
2814 };
2815
2816 /* Policy used for first-level command attributes */
2817 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2818         [IPVS_CMD_ATTR_SERVICE]         = { .type = NLA_NESTED },
2819         [IPVS_CMD_ATTR_DEST]            = { .type = NLA_NESTED },
2820         [IPVS_CMD_ATTR_DAEMON]          = { .type = NLA_NESTED },
2821         [IPVS_CMD_ATTR_TIMEOUT_TCP]     = { .type = NLA_U32 },
2822         [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2823         [IPVS_CMD_ATTR_TIMEOUT_UDP]     = { .type = NLA_U32 },
2824 };
2825
2826 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2827 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2828         [IPVS_DAEMON_ATTR_STATE]        = { .type = NLA_U32 },
2829         [IPVS_DAEMON_ATTR_MCAST_IFN]    = { .type = NLA_NUL_STRING,
2830                                             .len = IP_VS_IFNAME_MAXLEN },
2831         [IPVS_DAEMON_ATTR_SYNC_ID]      = { .type = NLA_U32 },
2832 };
2833
2834 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2835 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2836         [IPVS_SVC_ATTR_AF]              = { .type = NLA_U16 },
2837         [IPVS_SVC_ATTR_PROTOCOL]        = { .type = NLA_U16 },
2838         [IPVS_SVC_ATTR_ADDR]            = { .type = NLA_BINARY,
2839                                             .len = sizeof(union nf_inet_addr) },
2840         [IPVS_SVC_ATTR_PORT]            = { .type = NLA_U16 },
2841         [IPVS_SVC_ATTR_FWMARK]          = { .type = NLA_U32 },
2842         [IPVS_SVC_ATTR_SCHED_NAME]      = { .type = NLA_NUL_STRING,
2843                                             .len = IP_VS_SCHEDNAME_MAXLEN },
2844         [IPVS_SVC_ATTR_PE_NAME]         = { .type = NLA_NUL_STRING,
2845                                             .len = IP_VS_PENAME_MAXLEN },
2846         [IPVS_SVC_ATTR_FLAGS]           = { .type = NLA_BINARY,
2847                                             .len = sizeof(struct ip_vs_flags) },
2848         [IPVS_SVC_ATTR_TIMEOUT]         = { .type = NLA_U32 },
2849         [IPVS_SVC_ATTR_NETMASK]         = { .type = NLA_U32 },
2850         [IPVS_SVC_ATTR_STATS]           = { .type = NLA_NESTED },
2851 };
2852
2853 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2854 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2855         [IPVS_DEST_ATTR_ADDR]           = { .type = NLA_BINARY,
2856                                             .len = sizeof(union nf_inet_addr) },
2857         [IPVS_DEST_ATTR_PORT]           = { .type = NLA_U16 },
2858         [IPVS_DEST_ATTR_FWD_METHOD]     = { .type = NLA_U32 },
2859         [IPVS_DEST_ATTR_WEIGHT]         = { .type = NLA_U32 },
2860         [IPVS_DEST_ATTR_U_THRESH]       = { .type = NLA_U32 },
2861         [IPVS_DEST_ATTR_L_THRESH]       = { .type = NLA_U32 },
2862         [IPVS_DEST_ATTR_ACTIVE_CONNS]   = { .type = NLA_U32 },
2863         [IPVS_DEST_ATTR_INACT_CONNS]    = { .type = NLA_U32 },
2864         [IPVS_DEST_ATTR_PERSIST_CONNS]  = { .type = NLA_U32 },
2865         [IPVS_DEST_ATTR_STATS]          = { .type = NLA_NESTED },
2866 };
2867
2868 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2869                                  struct ip_vs_stats *stats)
2870 {
2871         struct ip_vs_stats_user ustats;
2872         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2873         if (!nl_stats)
2874                 return -EMSGSIZE;
2875
2876         ip_vs_copy_stats(&ustats, stats);
2877
2878         if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns) ||
2879             nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts) ||
2880             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts) ||
2881             nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes) ||
2882             nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes) ||
2883             nla_put_u32(skb, IPVS_STATS_ATTR_CPS, ustats.cps) ||
2884             nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps) ||
2885             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps) ||
2886             nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps) ||
2887             nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps))
2888                 goto nla_put_failure;
2889         nla_nest_end(skb, nl_stats);
2890
2891         return 0;
2892
2893 nla_put_failure:
2894         nla_nest_cancel(skb, nl_stats);
2895         return -EMSGSIZE;
2896 }
2897
2898 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2899                                    struct ip_vs_service *svc)
2900 {
2901         struct ip_vs_scheduler *sched;
2902         struct ip_vs_pe *pe;
2903         struct nlattr *nl_service;
2904         struct ip_vs_flags flags = { .flags = svc->flags,
2905                                      .mask = ~0 };
2906
2907         nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2908         if (!nl_service)
2909                 return -EMSGSIZE;
2910
2911         if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2912                 goto nla_put_failure;
2913         if (svc->fwmark) {
2914                 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2915                         goto nla_put_failure;
2916         } else {
2917                 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2918                     nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2919                     nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2920                         goto nla_put_failure;
2921         }
2922
2923         sched = rcu_dereference_protected(svc->scheduler, 1);
2924         pe = rcu_dereference_protected(svc->pe, 1);
2925         if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched->name) ||
2926             (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
2927             nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2928             nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2929             nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2930                 goto nla_put_failure;
2931         if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2932                 goto nla_put_failure;
2933
2934         nla_nest_end(skb, nl_service);
2935
2936         return 0;
2937
2938 nla_put_failure:
2939         nla_nest_cancel(skb, nl_service);
2940         return -EMSGSIZE;
2941 }
2942
2943 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2944                                    struct ip_vs_service *svc,
2945                                    struct netlink_callback *cb)
2946 {
2947         void *hdr;
2948
2949         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2950                           &ip_vs_genl_family, NLM_F_MULTI,
2951                           IPVS_CMD_NEW_SERVICE);
2952         if (!hdr)
2953                 return -EMSGSIZE;
2954
2955         if (ip_vs_genl_fill_service(skb, svc) < 0)
2956                 goto nla_put_failure;
2957
2958         return genlmsg_end(skb, hdr);
2959
2960 nla_put_failure:
2961         genlmsg_cancel(skb, hdr);
2962         return -EMSGSIZE;
2963 }
2964
2965 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2966                                     struct netlink_callback *cb)
2967 {
2968         int idx = 0, i;
2969         int start = cb->args[0];
2970         struct ip_vs_service *svc;
2971         struct net *net = skb_sknet(skb);
2972
2973         mutex_lock(&__ip_vs_mutex);
2974         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2975                 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2976                         if (++idx <= start || !net_eq(svc->net, net))
2977                                 continue;
2978                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2979                                 idx--;
2980                                 goto nla_put_failure;
2981                         }
2982                 }
2983         }
2984
2985         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2986                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2987                         if (++idx <= start || !net_eq(svc->net, net))
2988                                 continue;
2989                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2990                                 idx--;
2991                                 goto nla_put_failure;
2992                         }
2993                 }
2994         }
2995
2996 nla_put_failure:
2997         mutex_unlock(&__ip_vs_mutex);
2998         cb->args[0] = idx;
2999
3000         return skb->len;
3001 }
3002
3003 static int ip_vs_genl_parse_service(struct net *net,
3004                                     struct ip_vs_service_user_kern *usvc,
3005                                     struct nlattr *nla, int full_entry,
3006                                     struct ip_vs_service **ret_svc)
3007 {
3008         struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3009         struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3010         struct ip_vs_service *svc;
3011
3012         /* Parse mandatory identifying service fields first */
3013         if (nla == NULL ||
3014             nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
3015                 return -EINVAL;
3016
3017         nla_af          = attrs[IPVS_SVC_ATTR_AF];
3018         nla_protocol    = attrs[IPVS_SVC_ATTR_PROTOCOL];
3019         nla_addr        = attrs[IPVS_SVC_ATTR_ADDR];
3020         nla_port        = attrs[IPVS_SVC_ATTR_PORT];
3021         nla_fwmark      = attrs[IPVS_SVC_ATTR_FWMARK];
3022
3023         if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3024                 return -EINVAL;
3025
3026         memset(usvc, 0, sizeof(*usvc));
3027
3028         usvc->af = nla_get_u16(nla_af);
3029 #ifdef CONFIG_IP_VS_IPV6
3030         if (usvc->af != AF_INET && usvc->af != AF_INET6)
3031 #else
3032         if (usvc->af != AF_INET)
3033 #endif
3034                 return -EAFNOSUPPORT;
3035
3036         if (nla_fwmark) {
3037                 usvc->protocol = IPPROTO_TCP;
3038                 usvc->fwmark = nla_get_u32(nla_fwmark);
3039         } else {
3040                 usvc->protocol = nla_get_u16(nla_protocol);
3041                 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3042                 usvc->port = nla_get_be16(nla_port);
3043                 usvc->fwmark = 0;
3044         }
3045
3046         rcu_read_lock();
3047         if (usvc->fwmark)
3048                 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
3049         else
3050                 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
3051                                            &usvc->addr, usvc->port);
3052         rcu_read_unlock();
3053         *ret_svc = svc;
3054
3055         /* If a full entry was requested, check for the additional fields */
3056         if (full_entry) {
3057                 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3058                               *nla_netmask;
3059                 struct ip_vs_flags flags;
3060
3061                 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3062                 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3063                 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3064                 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3065                 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3066
3067                 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3068                         return -EINVAL;
3069
3070                 nla_memcpy(&flags, nla_flags, sizeof(flags));
3071
3072                 /* prefill flags from service if it already exists */
3073                 if (svc)
3074                         usvc->flags = svc->flags;
3075
3076                 /* set new flags from userland */
3077                 usvc->flags = (usvc->flags & ~flags.mask) |
3078                               (flags.flags & flags.mask);
3079                 usvc->sched_name = nla_data(nla_sched);
3080                 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3081                 usvc->timeout = nla_get_u32(nla_timeout);
3082                 usvc->netmask = nla_get_be32(nla_netmask);
3083         }
3084
3085         return 0;
3086 }
3087
3088 static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3089                                                      struct nlattr *nla)
3090 {
3091         struct ip_vs_service_user_kern usvc;
3092         struct ip_vs_service *svc;
3093         int ret;
3094
3095         ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
3096         return ret ? ERR_PTR(ret) : svc;
3097 }
3098
3099 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3100 {
3101         struct nlattr *nl_dest;
3102
3103         nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3104         if (!nl_dest)
3105                 return -EMSGSIZE;
3106
3107         if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3108             nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3109             nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3110                         (atomic_read(&dest->conn_flags) &
3111                          IP_VS_CONN_F_FWD_MASK)) ||
3112             nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3113                         atomic_read(&dest->weight)) ||
3114             nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3115             nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3116             nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3117                         atomic_read(&dest->activeconns)) ||
3118             nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3119                         atomic_read(&dest->inactconns)) ||
3120             nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3121                         atomic_read(&dest->persistconns)))
3122                 goto nla_put_failure;
3123         if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3124                 goto nla_put_failure;
3125
3126         nla_nest_end(skb, nl_dest);
3127
3128         return 0;
3129
3130 nla_put_failure:
3131         nla_nest_cancel(skb, nl_dest);
3132         return -EMSGSIZE;
3133 }
3134
3135 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3136                                 struct netlink_callback *cb)
3137 {
3138         void *hdr;
3139
3140         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3141                           &ip_vs_genl_family, NLM_F_MULTI,
3142                           IPVS_CMD_NEW_DEST);
3143         if (!hdr)
3144                 return -EMSGSIZE;
3145
3146         if (ip_vs_genl_fill_dest(skb, dest) < 0)
3147                 goto nla_put_failure;
3148
3149         return genlmsg_end(skb, hdr);
3150
3151 nla_put_failure:
3152         genlmsg_cancel(skb, hdr);
3153         return -EMSGSIZE;
3154 }
3155
3156 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3157                                  struct netlink_callback *cb)
3158 {
3159         int idx = 0;
3160         int start = cb->args[0];
3161         struct ip_vs_service *svc;
3162         struct ip_vs_dest *dest;
3163         struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3164         struct net *net = skb_sknet(skb);
3165
3166         mutex_lock(&__ip_vs_mutex);
3167
3168         /* Try to find the service for which to dump destinations */
3169         if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3170                         IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3171                 goto out_err;
3172
3173
3174         svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
3175         if (IS_ERR(svc) || svc == NULL)
3176                 goto out_err;
3177
3178         /* Dump the destinations */
3179         list_for_each_entry(dest, &svc->destinations, n_list) {
3180                 if (++idx <= start)
3181                         continue;
3182                 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3183                         idx--;
3184                         goto nla_put_failure;
3185                 }
3186         }
3187
3188 nla_put_failure:
3189         cb->args[0] = idx;
3190
3191 out_err:
3192         mutex_unlock(&__ip_vs_mutex);
3193
3194         return skb->len;
3195 }
3196
3197 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3198                                  struct nlattr *nla, int full_entry)
3199 {
3200         struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3201         struct nlattr *nla_addr, *nla_port;
3202
3203         /* Parse mandatory identifying destination fields first */
3204         if (nla == NULL ||
3205             nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3206                 return -EINVAL;
3207
3208         nla_addr        = attrs[IPVS_DEST_ATTR_ADDR];
3209         nla_port        = attrs[IPVS_DEST_ATTR_PORT];
3210
3211         if (!(nla_addr && nla_port))
3212                 return -EINVAL;
3213
3214         memset(udest, 0, sizeof(*udest));
3215
3216         nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3217         udest->port = nla_get_be16(nla_port);
3218
3219         /* If a full entry was requested, check for the additional fields */
3220         if (full_entry) {
3221                 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3222                               *nla_l_thresh;
3223
3224                 nla_fwd         = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3225                 nla_weight      = attrs[IPVS_DEST_ATTR_WEIGHT];
3226                 nla_u_thresh    = attrs[IPVS_DEST_ATTR_U_THRESH];
3227                 nla_l_thresh    = attrs[IPVS_DEST_ATTR_L_THRESH];
3228
3229                 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3230                         return -EINVAL;
3231
3232                 udest->conn_flags = nla_get_u32(nla_fwd)
3233                                     & IP_VS_CONN_F_FWD_MASK;
3234                 udest->weight = nla_get_u32(nla_weight);
3235                 udest->u_threshold = nla_get_u32(nla_u_thresh);
3236                 udest->l_threshold = nla_get_u32(nla_l_thresh);
3237         }
3238
3239         return 0;
3240 }
3241
3242 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3243                                   const char *mcast_ifn, __u32 syncid)
3244 {
3245         struct nlattr *nl_daemon;
3246
3247         nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3248         if (!nl_daemon)
3249                 return -EMSGSIZE;
3250
3251         if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3252             nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn) ||
3253             nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid))
3254                 goto nla_put_failure;
3255         nla_nest_end(skb, nl_daemon);
3256
3257         return 0;
3258
3259 nla_put_failure:
3260         nla_nest_cancel(skb, nl_daemon);
3261         return -EMSGSIZE;
3262 }
3263
3264 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3265                                   const char *mcast_ifn, __u32 syncid,
3266                                   struct netlink_callback *cb)
3267 {
3268         void *hdr;
3269         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3270                           &ip_vs_genl_family, NLM_F_MULTI,
3271                           IPVS_CMD_NEW_DAEMON);
3272         if (!hdr)
3273                 return -EMSGSIZE;
3274
3275         if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3276                 goto nla_put_failure;
3277
3278         return genlmsg_end(skb, hdr);
3279
3280 nla_put_failure:
3281         genlmsg_cancel(skb, hdr);
3282         return -EMSGSIZE;
3283 }
3284
3285 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3286                                    struct netlink_callback *cb)
3287 {
3288         struct net *net = skb_sknet(skb);
3289         struct netns_ipvs *ipvs = net_ipvs(net);
3290
3291         mutex_lock(&ipvs->sync_mutex);
3292         if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3293                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3294                                            ipvs->master_mcast_ifn,
3295                                            ipvs->master_syncid, cb) < 0)
3296                         goto nla_put_failure;
3297
3298                 cb->args[0] = 1;
3299         }
3300
3301         if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3302                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3303                                            ipvs->backup_mcast_ifn,
3304                                            ipvs->backup_syncid, cb) < 0)
3305                         goto nla_put_failure;
3306
3307                 cb->args[1] = 1;
3308         }
3309
3310 nla_put_failure:
3311         mutex_unlock(&ipvs->sync_mutex);
3312
3313         return skb->len;
3314 }
3315
3316 static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
3317 {
3318         if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3319               attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3320               attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3321                 return -EINVAL;
3322
3323         return start_sync_thread(net,
3324                                  nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3325                                  nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3326                                  nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3327 }
3328
3329 static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
3330 {
3331         if (!attrs[IPVS_DAEMON_ATTR_STATE])
3332                 return -EINVAL;
3333
3334         return stop_sync_thread(net,
3335                                 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3336 }
3337
3338 static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
3339 {
3340         struct ip_vs_timeout_user t;
3341
3342         __ip_vs_get_timeouts(net, &t);
3343
3344         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3345                 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3346
3347         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3348                 t.tcp_fin_timeout =
3349                         nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3350
3351         if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3352                 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3353
3354         return ip_vs_set_timeout(net, &t);
3355 }
3356
3357 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3358 {
3359         int ret = 0, cmd;
3360         struct net *net;
3361         struct netns_ipvs *ipvs;
3362
3363         net = skb_sknet(skb);
3364         ipvs = net_ipvs(net);
3365         cmd = info->genlhdr->cmd;
3366
3367         if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3368                 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3369
3370                 mutex_lock(&ipvs->sync_mutex);
3371                 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3372                     nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3373                                      info->attrs[IPVS_CMD_ATTR_DAEMON],
3374                                      ip_vs_daemon_policy)) {
3375                         ret = -EINVAL;
3376                         goto out;
3377                 }
3378
3379                 if (cmd == IPVS_CMD_NEW_DAEMON)
3380                         ret = ip_vs_genl_new_daemon(net, daemon_attrs);
3381                 else
3382                         ret = ip_vs_genl_del_daemon(net, daemon_attrs);
3383 out:
3384                 mutex_unlock(&ipvs->sync_mutex);
3385         }
3386         return ret;
3387 }
3388
3389 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3390 {
3391         struct ip_vs_service *svc = NULL;
3392         struct ip_vs_service_user_kern usvc;
3393         struct ip_vs_dest_user_kern udest;
3394         int ret = 0, cmd;
3395         int need_full_svc = 0, need_full_dest = 0;
3396         struct net *net;
3397
3398         net = skb_sknet(skb);
3399         cmd = info->genlhdr->cmd;
3400
3401         mutex_lock(&__ip_vs_mutex);
3402
3403         if (cmd == IPVS_CMD_FLUSH) {
3404                 ret = ip_vs_flush(net, false);
3405                 goto out;
3406         } else if (cmd == IPVS_CMD_SET_CONFIG) {
3407                 ret = ip_vs_genl_set_config(net, info->attrs);
3408                 goto out;
3409         } else if (cmd == IPVS_CMD_ZERO &&
3410                    !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3411                 ret = ip_vs_zero_all(net);
3412                 goto out;
3413         }
3414
3415         /* All following commands require a service argument, so check if we
3416          * received a valid one. We need a full service specification when
3417          * adding / editing a service. Only identifying members otherwise. */
3418         if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3419                 need_full_svc = 1;
3420
3421         ret = ip_vs_genl_parse_service(net, &usvc,
3422                                        info->attrs[IPVS_CMD_ATTR_SERVICE],
3423                                        need_full_svc, &svc);
3424         if (ret)
3425                 goto out;
3426
3427         /* Unless we're adding a new service, the service must already exist */
3428         if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3429                 ret = -ESRCH;
3430                 goto out;
3431         }
3432
3433         /* Destination commands require a valid destination argument. For
3434          * adding / editing a destination, we need a full destination
3435          * specification. */
3436         if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3437             cmd == IPVS_CMD_DEL_DEST) {
3438                 if (cmd != IPVS_CMD_DEL_DEST)
3439                         need_full_dest = 1;
3440
3441                 ret = ip_vs_genl_parse_dest(&udest,
3442                                             info->attrs[IPVS_CMD_ATTR_DEST],
3443                                             need_full_dest);
3444                 if (ret)
3445                         goto out;
3446         }
3447
3448         switch (cmd) {
3449         case IPVS_CMD_NEW_SERVICE:
3450                 if (svc == NULL)
3451                         ret = ip_vs_add_service(net, &usvc, &svc);
3452                 else
3453                         ret = -EEXIST;
3454                 break;
3455         case IPVS_CMD_SET_SERVICE:
3456                 ret = ip_vs_edit_service(svc, &usvc);
3457                 break;
3458         case IPVS_CMD_DEL_SERVICE:
3459                 ret = ip_vs_del_service(svc);
3460                 /* do not use svc, it can be freed */
3461                 break;
3462         case IPVS_CMD_NEW_DEST:
3463                 ret = ip_vs_add_dest(svc, &udest);
3464                 break;
3465         case IPVS_CMD_SET_DEST:
3466                 ret = ip_vs_edit_dest(svc, &udest);
3467                 break;
3468         case IPVS_CMD_DEL_DEST:
3469                 ret = ip_vs_del_dest(svc, &udest);
3470                 break;
3471         case IPVS_CMD_ZERO:
3472                 ret = ip_vs_zero_service(svc);
3473                 break;
3474         default:
3475                 ret = -EINVAL;
3476         }
3477
3478 out:
3479         mutex_unlock(&__ip_vs_mutex);
3480
3481         return ret;
3482 }
3483
3484 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3485 {
3486         struct sk_buff *msg;
3487         void *reply;
3488         int ret, cmd, reply_cmd;
3489         struct net *net;
3490
3491         net = skb_sknet(skb);
3492         cmd = info->genlhdr->cmd;
3493
3494         if (cmd == IPVS_CMD_GET_SERVICE)
3495                 reply_cmd = IPVS_CMD_NEW_SERVICE;
3496         else if (cmd == IPVS_CMD_GET_INFO)
3497                 reply_cmd = IPVS_CMD_SET_INFO;
3498         else if (cmd == IPVS_CMD_GET_CONFIG)
3499                 reply_cmd = IPVS_CMD_SET_CONFIG;
3500         else {
3501                 pr_err("unknown Generic Netlink command\n");
3502                 return -EINVAL;
3503         }
3504
3505         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3506         if (!msg)
3507                 return -ENOMEM;
3508
3509         mutex_lock(&__ip_vs_mutex);
3510
3511         reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3512         if (reply == NULL)
3513                 goto nla_put_failure;
3514
3515         switch (cmd) {
3516         case IPVS_CMD_GET_SERVICE:
3517         {
3518                 struct ip_vs_service *svc;
3519
3520                 svc = ip_vs_genl_find_service(net,
3521                                               info->attrs[IPVS_CMD_ATTR_SERVICE]);
3522                 if (IS_ERR(svc)) {
3523                         ret = PTR_ERR(svc);
3524                         goto out_err;
3525                 } else if (svc) {
3526                         ret = ip_vs_genl_fill_service(msg, svc);
3527                         if (ret)
3528                                 goto nla_put_failure;
3529                 } else {
3530                         ret = -ESRCH;
3531                         goto out_err;
3532                 }
3533
3534                 break;
3535         }
3536
3537         case IPVS_CMD_GET_CONFIG:
3538         {
3539                 struct ip_vs_timeout_user t;
3540
3541                 __ip_vs_get_timeouts(net, &t);
3542 #ifdef CONFIG_IP_VS_PROTO_TCP
3543                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3544                                 t.tcp_timeout) ||
3545                     nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3546                                 t.tcp_fin_timeout))
3547                         goto nla_put_failure;
3548 #endif
3549 #ifdef CONFIG_IP_VS_PROTO_UDP
3550                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3551                         goto nla_put_failure;
3552 #endif
3553
3554                 break;
3555         }
3556
3557         case IPVS_CMD_GET_INFO:
3558                 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3559                                 IP_VS_VERSION_CODE) ||
3560                     nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3561                                 ip_vs_conn_tab_size))
3562                         goto nla_put_failure;
3563                 break;
3564         }
3565
3566         genlmsg_end(msg, reply);
3567         ret = genlmsg_reply(msg, info);
3568         goto out;
3569
3570 nla_put_failure:
3571         pr_err("not enough space in Netlink message\n");
3572         ret = -EMSGSIZE;
3573
3574 out_err:
3575         nlmsg_free(msg);
3576 out:
3577         mutex_unlock(&__ip_vs_mutex);
3578
3579         return ret;
3580 }
3581
3582
3583 static const struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3584         {
3585                 .cmd    = IPVS_CMD_NEW_SERVICE,
3586                 .flags  = GENL_ADMIN_PERM,
3587                 .policy = ip_vs_cmd_policy,
3588                 .doit   = ip_vs_genl_set_cmd,
3589         },
3590         {
3591                 .cmd    = IPVS_CMD_SET_SERVICE,
3592                 .flags  = GENL_ADMIN_PERM,
3593                 .policy = ip_vs_cmd_policy,
3594                 .doit   = ip_vs_genl_set_cmd,
3595         },
3596         {
3597                 .cmd    = IPVS_CMD_DEL_SERVICE,
3598                 .flags  = GENL_ADMIN_PERM,
3599                 .policy = ip_vs_cmd_policy,
3600                 .doit   = ip_vs_genl_set_cmd,
3601         },
3602         {
3603                 .cmd    = IPVS_CMD_GET_SERVICE,
3604                 .flags  = GENL_ADMIN_PERM,
3605                 .doit   = ip_vs_genl_get_cmd,
3606                 .dumpit = ip_vs_genl_dump_services,
3607                 .policy = ip_vs_cmd_policy,
3608         },
3609         {
3610                 .cmd    = IPVS_CMD_NEW_DEST,
3611                 .flags  = GENL_ADMIN_PERM,
3612                 .policy = ip_vs_cmd_policy,
3613                 .doit   = ip_vs_genl_set_cmd,
3614         },
3615         {
3616                 .cmd    = IPVS_CMD_SET_DEST,
3617                 .flags  = GENL_ADMIN_PERM,
3618                 .policy = ip_vs_cmd_policy,
3619                 .doit   = ip_vs_genl_set_cmd,
3620         },
3621         {
3622                 .cmd    = IPVS_CMD_DEL_DEST,
3623                 .flags  = GENL_ADMIN_PERM,
3624                 .policy = ip_vs_cmd_policy,
3625                 .doit   = ip_vs_genl_set_cmd,
3626         },
3627         {
3628                 .cmd    = IPVS_CMD_GET_DEST,
3629                 .flags  = GENL_ADMIN_PERM,
3630                 .policy = ip_vs_cmd_policy,
3631                 .dumpit = ip_vs_genl_dump_dests,
3632         },
3633         {
3634                 .cmd    = IPVS_CMD_NEW_DAEMON,
3635                 .flags  = GENL_ADMIN_PERM,
3636                 .policy = ip_vs_cmd_policy,
3637                 .doit   = ip_vs_genl_set_daemon,
3638         },
3639         {
3640                 .cmd    = IPVS_CMD_DEL_DAEMON,
3641                 .flags  = GENL_ADMIN_PERM,
3642                 .policy = ip_vs_cmd_policy,
3643                 .doit   = ip_vs_genl_set_daemon,
3644         },
3645         {
3646                 .cmd    = IPVS_CMD_GET_DAEMON,
3647                 .flags  = GENL_ADMIN_PERM,
3648                 .dumpit = ip_vs_genl_dump_daemons,
3649         },
3650         {
3651                 .cmd    = IPVS_CMD_SET_CONFIG,
3652                 .flags  = GENL_ADMIN_PERM,
3653                 .policy = ip_vs_cmd_policy,
3654                 .doit   = ip_vs_genl_set_cmd,
3655         },
3656         {
3657                 .cmd    = IPVS_CMD_GET_CONFIG,
3658                 .flags  = GENL_ADMIN_PERM,
3659                 .doit   = ip_vs_genl_get_cmd,
3660         },
3661         {
3662                 .cmd    = IPVS_CMD_GET_INFO,
3663                 .flags  = GENL_ADMIN_PERM,
3664                 .doit   = ip_vs_genl_get_cmd,
3665         },
3666         {
3667                 .cmd    = IPVS_CMD_ZERO,
3668                 .flags  = GENL_ADMIN_PERM,
3669                 .policy = ip_vs_cmd_policy,
3670                 .doit   = ip_vs_genl_set_cmd,
3671         },
3672         {
3673                 .cmd    = IPVS_CMD_FLUSH,
3674                 .flags  = GENL_ADMIN_PERM,
3675                 .doit   = ip_vs_genl_set_cmd,
3676         },
3677 };
3678
3679 static int __init ip_vs_genl_register(void)
3680 {
3681         return genl_register_family_with_ops(&ip_vs_genl_family,
3682                                              ip_vs_genl_ops);
3683 }
3684
3685 static void ip_vs_genl_unregister(void)
3686 {
3687         genl_unregister_family(&ip_vs_genl_family);
3688 }
3689
3690 /* End of Generic Netlink interface definitions */
3691
3692 /*
3693  * per netns intit/exit func.
3694  */
3695 #ifdef CONFIG_SYSCTL
3696 static int __net_init ip_vs_control_net_init_sysctl(struct net *net)
3697 {
3698         int idx;
3699         struct netns_ipvs *ipvs = net_ipvs(net);
3700         struct ctl_table *tbl;
3701
3702         atomic_set(&ipvs->dropentry, 0);
3703         spin_lock_init(&ipvs->dropentry_lock);
3704         spin_lock_init(&ipvs->droppacket_lock);
3705         spin_lock_init(&ipvs->securetcp_lock);
3706
3707         if (!net_eq(net, &init_net)) {
3708                 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3709                 if (tbl == NULL)
3710                         return -ENOMEM;
3711
3712                 /* Don't export sysctls to unprivileged users */
3713                 if (net->user_ns != &init_user_ns)
3714                         tbl[0].procname = NULL;
3715         } else
3716                 tbl = vs_vars;
3717         /* Initialize sysctl defaults */
3718         idx = 0;
3719         ipvs->sysctl_amemthresh = 1024;
3720         tbl[idx++].data = &ipvs->sysctl_amemthresh;
3721         ipvs->sysctl_am_droprate = 10;
3722         tbl[idx++].data = &ipvs->sysctl_am_droprate;
3723         tbl[idx++].data = &ipvs->sysctl_drop_entry;
3724         tbl[idx++].data = &ipvs->sysctl_drop_packet;
3725 #ifdef CONFIG_IP_VS_NFCT
3726         tbl[idx++].data = &ipvs->sysctl_conntrack;
3727 #endif
3728         tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3729         ipvs->sysctl_snat_reroute = 1;
3730         tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3731         ipvs->sysctl_sync_ver = 1;
3732         tbl[idx++].data = &ipvs->sysctl_sync_ver;
3733         ipvs->sysctl_sync_ports = 1;
3734         tbl[idx++].data = &ipvs->sysctl_sync_ports;
3735         tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3736         ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3737         tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3738         ipvs->sysctl_sync_sock_size = 0;
3739         tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3740         tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3741         tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3742         tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3743         tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3744         tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3745         ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3746         ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3747         tbl[idx].data = &ipvs->sysctl_sync_threshold;
3748         tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3749         ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3750         tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3751         ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3752         tbl[idx++].data = &ipvs->sysctl_sync_retries;
3753         tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3754         ipvs->sysctl_pmtu_disc = 1;
3755         tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3756         tbl[idx++].data = &ipvs->sysctl_backup_only;
3757
3758
3759         ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3760         if (ipvs->sysctl_hdr == NULL) {
3761                 if (!net_eq(net, &init_net))
3762                         kfree(tbl);
3763                 return -ENOMEM;
3764         }
3765         ip_vs_start_estimator(net, &ipvs->tot_stats);
3766         ipvs->sysctl_tbl = tbl;
3767         /* Schedule defense work */
3768         INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3769         schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3770
3771         return 0;
3772 }
3773
3774 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net)
3775 {
3776         struct netns_ipvs *ipvs = net_ipvs(net);
3777
3778         cancel_delayed_work_sync(&ipvs->defense_work);
3779         cancel_work_sync(&ipvs->defense_work.work);
3780         unregister_net_sysctl_table(ipvs->sysctl_hdr);
3781 }
3782
3783 #else
3784
3785 static int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
3786 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net) { }
3787
3788 #endif
3789
3790 static struct notifier_block ip_vs_dst_notifier = {
3791         .notifier_call = ip_vs_dst_event,
3792 };
3793
3794 int __net_init ip_vs_control_net_init(struct net *net)
3795 {
3796         int i, idx;
3797         struct netns_ipvs *ipvs = net_ipvs(net);
3798
3799         /* Initialize rs_table */
3800         for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3801                 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
3802
3803         INIT_LIST_HEAD(&ipvs->dest_trash);
3804         spin_lock_init(&ipvs->dest_trash_lock);
3805         setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
3806                     (unsigned long) net);
3807         atomic_set(&ipvs->ftpsvc_counter, 0);
3808         atomic_set(&ipvs->nullsvc_counter, 0);
3809
3810         /* procfs stats */
3811         ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3812         if (!ipvs->tot_stats.cpustats)
3813                 return -ENOMEM;
3814
3815         for_each_possible_cpu(i) {
3816                 struct ip_vs_cpu_stats *ipvs_tot_stats;
3817                 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
3818                 u64_stats_init(&ipvs_tot_stats->syncp);
3819         }
3820
3821         spin_lock_init(&ipvs->tot_stats.lock);
3822
3823         proc_create("ip_vs", 0, net->proc_net, &ip_vs_info_fops);
3824         proc_create("ip_vs_stats", 0, net->proc_net, &ip_vs_stats_fops);
3825         proc_create("ip_vs_stats_percpu", 0, net->proc_net,
3826                     &ip_vs_stats_percpu_fops);
3827
3828         if (ip_vs_control_net_init_sysctl(net))
3829                 goto err;
3830
3831         return 0;
3832
3833 err:
3834         free_percpu(ipvs->tot_stats.cpustats);
3835         return -ENOMEM;
3836 }
3837
3838 void __net_exit ip_vs_control_net_cleanup(struct net *net)
3839 {
3840         struct netns_ipvs *ipvs = net_ipvs(net);
3841
3842         ip_vs_trash_cleanup(net);
3843         ip_vs_stop_estimator(net, &ipvs->tot_stats);
3844         ip_vs_control_net_cleanup_sysctl(net);
3845         remove_proc_entry("ip_vs_stats_percpu", net->proc_net);
3846         remove_proc_entry("ip_vs_stats", net->proc_net);
3847         remove_proc_entry("ip_vs", net->proc_net);
3848         free_percpu(ipvs->tot_stats.cpustats);
3849 }
3850
3851 int __init ip_vs_register_nl_ioctl(void)
3852 {
3853         int ret;
3854
3855         ret = nf_register_sockopt(&ip_vs_sockopts);
3856         if (ret) {
3857                 pr_err("cannot register sockopt.\n");
3858                 goto err_sock;
3859         }
3860
3861         ret = ip_vs_genl_register();
3862         if (ret) {
3863                 pr_err("cannot register Generic Netlink interface.\n");
3864                 goto err_genl;
3865         }
3866         return 0;
3867
3868 err_genl:
3869         nf_unregister_sockopt(&ip_vs_sockopts);
3870 err_sock:
3871         return ret;
3872 }
3873
3874 void ip_vs_unregister_nl_ioctl(void)
3875 {
3876         ip_vs_genl_unregister();
3877         nf_unregister_sockopt(&ip_vs_sockopts);
3878 }
3879
3880 int __init ip_vs_control_init(void)
3881 {
3882         int idx;
3883         int ret;
3884
3885         EnterFunction(2);
3886
3887         /* Initialize svc_table, ip_vs_svc_fwm_table */
3888         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3889                 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
3890                 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3891         }
3892
3893         smp_wmb();      /* Do we really need it now ? */
3894
3895         ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3896         if (ret < 0)
3897                 return ret;
3898
3899         LeaveFunction(2);
3900         return 0;
3901 }
3902
3903
3904 void ip_vs_control_cleanup(void)
3905 {
3906         EnterFunction(2);
3907         unregister_netdevice_notifier(&ip_vs_dst_notifier);
3908         LeaveFunction(2);
3909 }