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         free_percpu(svc->stats.cpustats);
 469         kfree(svc);
 470 }
 471
 472 static void ip_vs_service_rcu_free(struct rcu_head *head)
 473 {
 474         struct ip_vs_service *svc;
 475
 476         svc = container_of(head, struct ip_vs_service, rcu_head);
 477         ip_vs_service_free(svc);
 478 }
 479
 480 static void __ip_vs_svc_put(struct ip_vs_service *svc, bool do_delay)
 481 {
 482         if (atomic_dec_and_test(&svc->refcnt)) {
 483                 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
 484                               svc->fwmark,
 485                               IP_VS_DBG_ADDR(svc->af, &svc->addr),
 486                               ntohs(svc->port));
 487                 if (do_delay)
 488                         call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
 489                 else
 490                         ip_vs_service_free(svc);
 491         }
 492 }
 493
 494
 495 /*
 496  *      Returns hash value for real service
 497  */
 498 static inline unsigned int ip_vs_rs_hashkey(int af,
 499                                             const union nf_inet_addr *addr,
 500                                             __be16 port)
 501 {
 502         register unsigned int porth = ntohs(port);
 503         __be32 addr_fold = addr->ip;
 504
 505 #ifdef CONFIG_IP_VS_IPV6
 506         if (af == AF_INET6)
 507                 addr_fold = addr->ip6[0]^addr->ip6[1]^
 508                             addr->ip6[2]^addr->ip6[3];
 509 #endif
 510
 511         return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
 512                 & IP_VS_RTAB_MASK;
 513 }
 514
 515 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
 516 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
 517 {
 518         unsigned int hash;
 519
 520         if (dest->in_rs_table)
 521                 return;
 522
 523         /*
 524          *      Hash by proto,addr,port,
 525          *      which are the parameters of the real service.
 526          */
 527         hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
 528
 529         hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
 530         dest->in_rs_table = 1;
 531 }
 532
 533 /* Unhash ip_vs_dest from rs_table. */
 534 static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
 535 {
 536         /*
 537          * Remove it from the rs_table table.
 538          */
 539         if (dest->in_rs_table) {
 540                 hlist_del_rcu(&dest->d_list);
 541                 dest->in_rs_table = 0;
 542         }
 543 }
 544
 545 /* Check if real service by <proto,addr,port> is present */
 546 bool ip_vs_has_real_service(struct net *net, int af, __u16 protocol,
 547                             const union nf_inet_addr *daddr, __be16 dport)
 548 {
 549         struct netns_ipvs *ipvs = net_ipvs(net);
 550         unsigned int hash;
 551         struct ip_vs_dest *dest;
 552
 553         /* Check for "full" addressed entries */
 554         hash = ip_vs_rs_hashkey(af, daddr, dport);
 555
 556         rcu_read_lock();
 557         hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
 558                 if (dest->port == dport &&
 559                     dest->af == af &&
 560                     ip_vs_addr_equal(af, &dest->addr, daddr) &&
 561                     (dest->protocol == protocol || dest->vfwmark)) {
 562                         /* HIT */
 563                         rcu_read_unlock();
 564                         return true;
 565                 }
 566         }
 567         rcu_read_unlock();
 568
 569         return false;
 570 }
 571
 572 /* Lookup destination by {addr,port} in the given service
 573  * Called under RCU lock.
 574  */
 575 static struct ip_vs_dest *
 576 ip_vs_lookup_dest(struct ip_vs_service *svc, int dest_af,
 577                   const union nf_inet_addr *daddr, __be16 dport)
 578 {
 579         struct ip_vs_dest *dest;
 580
 581         /*
 582          * Find the destination for the given service
 583          */
 584         list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
 585                 if ((dest->af == dest_af) &&
 586                     ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
 587                     (dest->port == dport)) {
 588                         /* HIT */
 589                         return dest;
 590                 }
 591         }
 592
 593         return NULL;
 594 }
 595
 596 /*
 597  * Find destination by {daddr,dport,vaddr,protocol}
 598  * Created to be used in ip_vs_process_message() in
 599  * the backup synchronization daemon. It finds the
 600  * destination to be bound to the received connection
 601  * on the backup.
 602  * Called under RCU lock, no refcnt is returned.
 603  */
 604 struct ip_vs_dest *ip_vs_find_dest(struct net  *net, int svc_af, int dest_af,
 605                                    const union nf_inet_addr *daddr,
 606                                    __be16 dport,
 607                                    const union nf_inet_addr *vaddr,
 608                                    __be16 vport, __u16 protocol, __u32 fwmark,
 609                                    __u32 flags)
 610 {
 611         struct ip_vs_dest *dest;
 612         struct ip_vs_service *svc;
 613         __be16 port = dport;
 614
 615         svc = ip_vs_service_find(net, svc_af, fwmark, protocol, vaddr, vport);
 616         if (!svc)
 617                 return NULL;
 618         if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
 619                 port = 0;
 620         dest = ip_vs_lookup_dest(svc, dest_af, daddr, port);
 621         if (!dest)
 622                 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port ^ dport);
 623         return dest;
 624 }
 625
 626 void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
 627 {
 628         struct ip_vs_dest_dst *dest_dst = container_of(head,
 629                                                        struct ip_vs_dest_dst,
 630                                                        rcu_head);
 631
 632         dst_release(dest_dst->dst_cache);
 633         kfree(dest_dst);
 634 }
 635
 636 /* Release dest_dst and dst_cache for dest in user context */
 637 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
 638 {
 639         struct ip_vs_dest_dst *old;
 640
 641         old = rcu_dereference_protected(dest->dest_dst, 1);
 642         if (old) {
 643                 RCU_INIT_POINTER(dest->dest_dst, NULL);
 644                 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
 645         }
 646 }
 647
 648 /*
 649  *  Lookup dest by {svc,addr,port} in the destination trash.
 650  *  The destination trash is used to hold the destinations that are removed
 651  *  from the service table but are still referenced by some conn entries.
 652  *  The reason to add the destination trash is when the dest is temporary
 653  *  down (either by administrator or by monitor program), the dest can be
 654  *  picked back from the trash, the remaining connections to the dest can
 655  *  continue, and the counting information of the dest is also useful for
 656  *  scheduling.
 657  */
 658 static struct ip_vs_dest *
 659 ip_vs_trash_get_dest(struct ip_vs_service *svc, int dest_af,
 660                      const union nf_inet_addr *daddr, __be16 dport)
 661 {
 662         struct ip_vs_dest *dest;
 663         struct netns_ipvs *ipvs = net_ipvs(svc->net);
 664
 665         /*
 666          * Find the destination in trash
 667          */
 668         spin_lock_bh(&ipvs->dest_trash_lock);
 669         list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
 670                 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
 671                               "dest->refcnt=%d\n",
 672                               dest->vfwmark,
 673                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
 674                               ntohs(dest->port),
 675                               atomic_read(&dest->refcnt));
 676                 if (dest->af == dest_af &&
 677                     ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
 678                     dest->port == dport &&
 679                     dest->vfwmark == svc->fwmark &&
 680                     dest->protocol == svc->protocol &&
 681                     (svc->fwmark ||
 682                      (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
 683                       dest->vport == svc->port))) {
 684                         /* HIT */
 685                         list_del(&dest->t_list);
 686                         ip_vs_dest_hold(dest);
 687                         goto out;
 688                 }
 689         }
 690
 691         dest = NULL;
 692
 693 out:
 694         spin_unlock_bh(&ipvs->dest_trash_lock);
 695
 696         return dest;
 697 }
 698
 699 static void ip_vs_dest_free(struct ip_vs_dest *dest)
 700 {
 701         struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1);
 702
 703         __ip_vs_dst_cache_reset(dest);
 704         __ip_vs_svc_put(svc, false);
 705         free_percpu(dest->stats.cpustats);
 706         ip_vs_dest_put_and_free(dest);
 707 }
 708
 709 /*
 710  *  Clean up all the destinations in the trash
 711  *  Called by the ip_vs_control_cleanup()
 712  *
 713  *  When the ip_vs_control_clearup is activated by ipvs module exit,
 714  *  the service tables must have been flushed and all the connections
 715  *  are expired, and the refcnt of each destination in the trash must
 716  *  be 0, so we simply release them here.
 717  */
 718 static void ip_vs_trash_cleanup(struct net *net)
 719 {
 720         struct ip_vs_dest *dest, *nxt;
 721         struct netns_ipvs *ipvs = net_ipvs(net);
 722
 723         del_timer_sync(&ipvs->dest_trash_timer);
 724         /* No need to use dest_trash_lock */
 725         list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
 726                 list_del(&dest->t_list);
 727                 ip_vs_dest_free(dest);
 728         }
 729 }
 730
 731 static void
 732 ip_vs_copy_stats(struct ip_vs_kstats *dst, struct ip_vs_stats *src)
 733 {
 734 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c
 735
 736         spin_lock_bh(&src->lock);
 737
 738         IP_VS_SHOW_STATS_COUNTER(conns);
 739         IP_VS_SHOW_STATS_COUNTER(inpkts);
 740         IP_VS_SHOW_STATS_COUNTER(outpkts);
 741         IP_VS_SHOW_STATS_COUNTER(inbytes);
 742         IP_VS_SHOW_STATS_COUNTER(outbytes);
 743
 744         ip_vs_read_estimator(dst, src);
 745
 746         spin_unlock_bh(&src->lock);
 747 }
 748
 749 static void
 750 ip_vs_export_stats_user(struct ip_vs_stats_user *dst, struct ip_vs_kstats *src)
 751 {
 752         dst->conns = (u32)src->conns;
 753         dst->inpkts = (u32)src->inpkts;
 754         dst->outpkts = (u32)src->outpkts;
 755         dst->inbytes = src->inbytes;
 756         dst->outbytes = src->outbytes;
 757         dst->cps = (u32)src->cps;
 758         dst->inpps = (u32)src->inpps;
 759         dst->outpps = (u32)src->outpps;
 760         dst->inbps = (u32)src->inbps;
 761         dst->outbps = (u32)src->outbps;
 762 }
 763
 764 static void
 765 ip_vs_zero_stats(struct ip_vs_stats *stats)
 766 {
 767         spin_lock_bh(&stats->lock);
 768
 769         /* get current counters as zero point, rates are zeroed */
 770
 771 #define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c
 772
 773         IP_VS_ZERO_STATS_COUNTER(conns);
 774         IP_VS_ZERO_STATS_COUNTER(inpkts);
 775         IP_VS_ZERO_STATS_COUNTER(outpkts);
 776         IP_VS_ZERO_STATS_COUNTER(inbytes);
 777         IP_VS_ZERO_STATS_COUNTER(outbytes);
 778
 779         ip_vs_zero_estimator(stats);
 780
 781         spin_unlock_bh(&stats->lock);
 782 }
 783
 784 /*
 785  *      Update a destination in the given service
 786  */
 787 static void
 788 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
 789                     struct ip_vs_dest_user_kern *udest, int add)
 790 {
 791         struct netns_ipvs *ipvs = net_ipvs(svc->net);
 792         struct ip_vs_service *old_svc;
 793         struct ip_vs_scheduler *sched;
 794         int conn_flags;
 795
 796         /* We cannot modify an address and change the address family */
 797         BUG_ON(!add && udest->af != dest->af);
 798
 799         if (add && udest->af != svc->af)
 800                 ipvs->mixed_address_family_dests++;
 801
 802         /* set the weight and the flags */
 803         atomic_set(&dest->weight, udest->weight);
 804         conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
 805         conn_flags |= IP_VS_CONN_F_INACTIVE;
 806
 807         /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
 808         if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
 809                 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
 810         } else {
 811                 /*
 812                  *    Put the real service in rs_table if not present.
 813                  *    For now only for NAT!
 814                  */
 815                 ip_vs_rs_hash(ipvs, dest);
 816         }
 817         atomic_set(&dest->conn_flags, conn_flags);
 818
 819         /* bind the service */
 820         old_svc = rcu_dereference_protected(dest->svc, 1);
 821         if (!old_svc) {
 822                 __ip_vs_bind_svc(dest, svc);
 823         } else {
 824                 if (old_svc != svc) {
 825                         ip_vs_zero_stats(&dest->stats);
 826                         __ip_vs_bind_svc(dest, svc);
 827                         __ip_vs_svc_put(old_svc, true);
 828                 }
 829         }
 830
 831         /* set the dest status flags */
 832         dest->flags |= IP_VS_DEST_F_AVAILABLE;
 833
 834         if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
 835                 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
 836         dest->u_threshold = udest->u_threshold;
 837         dest->l_threshold = udest->l_threshold;
 838
 839         dest->af = udest->af;
 840
 841         spin_lock_bh(&dest->dst_lock);
 842         __ip_vs_dst_cache_reset(dest);
 843         spin_unlock_bh(&dest->dst_lock);
 844
 845         if (add) {
 846                 ip_vs_start_estimator(svc->net, &dest->stats);
 847                 list_add_rcu(&dest->n_list, &svc->destinations);
 848                 svc->num_dests++;
 849                 sched = rcu_dereference_protected(svc->scheduler, 1);
 850                 if (sched && sched->add_dest)
 851                         sched->add_dest(svc, dest);
 852         } else {
 853                 sched = rcu_dereference_protected(svc->scheduler, 1);
 854                 if (sched && sched->upd_dest)
 855                         sched->upd_dest(svc, dest);
 856         }
 857 }
 858
 859
 860 /*
 861  *      Create a destination for the given service
 862  */
 863 static int
 864 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
 865                struct ip_vs_dest **dest_p)
 866 {
 867         struct ip_vs_dest *dest;
 868         unsigned int atype, i;
 869
 870         EnterFunction(2);
 871
 872 #ifdef CONFIG_IP_VS_IPV6
 873         if (udest->af == AF_INET6) {
 874                 atype = ipv6_addr_type(&udest->addr.in6);
 875                 if ((!(atype & IPV6_ADDR_UNICAST) ||
 876                         atype & IPV6_ADDR_LINKLOCAL) &&
 877                         !__ip_vs_addr_is_local_v6(svc->net, &udest->addr.in6))
 878                         return -EINVAL;
 879         } else
 880 #endif
 881         {
 882                 atype = inet_addr_type(svc->net, udest->addr.ip);
 883                 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
 884                         return -EINVAL;
 885         }
 886
 887         dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
 888         if (dest == NULL)
 889                 return -ENOMEM;
 890
 891         dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
 892         if (!dest->stats.cpustats)
 893                 goto err_alloc;
 894
 895         for_each_possible_cpu(i) {
 896                 struct ip_vs_cpu_stats *ip_vs_dest_stats;
 897                 ip_vs_dest_stats = per_cpu_ptr(dest->stats.cpustats, i);
 898                 u64_stats_init(&ip_vs_dest_stats->syncp);
 899         }
 900
 901         dest->af = udest->af;
 902         dest->protocol = svc->protocol;
 903         dest->vaddr = svc->addr;
 904         dest->vport = svc->port;
 905         dest->vfwmark = svc->fwmark;
 906         ip_vs_addr_copy(udest->af, &dest->addr, &udest->addr);
 907         dest->port = udest->port;
 908
 909         atomic_set(&dest->activeconns, 0);
 910         atomic_set(&dest->inactconns, 0);
 911         atomic_set(&dest->persistconns, 0);
 912         atomic_set(&dest->refcnt, 1);
 913
 914         INIT_HLIST_NODE(&dest->d_list);
 915         spin_lock_init(&dest->dst_lock);
 916         spin_lock_init(&dest->stats.lock);
 917         __ip_vs_update_dest(svc, dest, udest, 1);
 918
 919         *dest_p = dest;
 920
 921         LeaveFunction(2);
 922         return 0;
 923
 924 err_alloc:
 925         kfree(dest);
 926         return -ENOMEM;
 927 }
 928
 929
 930 /*
 931  *      Add a destination into an existing service
 932  */
 933 static int
 934 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
 935 {
 936         struct ip_vs_dest *dest;
 937         union nf_inet_addr daddr;
 938         __be16 dport = udest->port;
 939         int ret;
 940
 941         EnterFunction(2);
 942
 943         if (udest->weight < 0) {
 944                 pr_err("%s(): server weight less than zero\n", __func__);
 945                 return -ERANGE;
 946         }
 947
 948         if (udest->l_threshold > udest->u_threshold) {
 949                 pr_err("%s(): lower threshold is higher than upper threshold\n",
 950                         __func__);
 951                 return -ERANGE;
 952         }
 953
 954         ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
 955
 956         /* We use function that requires RCU lock */
 957         rcu_read_lock();
 958         dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
 959         rcu_read_unlock();
 960
 961         if (dest != NULL) {
 962                 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
 963                 return -EEXIST;
 964         }
 965
 966         /*
 967          * Check if the dest already exists in the trash and
 968          * is from the same service
 969          */
 970         dest = ip_vs_trash_get_dest(svc, udest->af, &daddr, dport);
 971
 972         if (dest != NULL) {
 973                 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
 974                               "dest->refcnt=%d, service %u/%s:%u\n",
 975                               IP_VS_DBG_ADDR(udest->af, &daddr), ntohs(dport),
 976                               atomic_read(&dest->refcnt),
 977                               dest->vfwmark,
 978                               IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
 979                               ntohs(dest->vport));
 980
 981                 __ip_vs_update_dest(svc, dest, udest, 1);
 982                 ret = 0;
 983         } else {
 984                 /*
 985                  * Allocate and initialize the dest structure
 986                  */
 987                 ret = ip_vs_new_dest(svc, udest, &dest);
 988         }
 989         LeaveFunction(2);
 990
 991         return ret;
 992 }
 993
 994
 995 /*
 996  *      Edit a destination in the given service
 997  */
 998 static int
 999 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1000 {
1001         struct ip_vs_dest *dest;
1002         union nf_inet_addr daddr;
1003         __be16 dport = udest->port;
1004
1005         EnterFunction(2);
1006
1007         if (udest->weight < 0) {
1008                 pr_err("%s(): server weight less than zero\n", __func__);
1009                 return -ERANGE;
1010         }
1011
1012         if (udest->l_threshold > udest->u_threshold) {
1013                 pr_err("%s(): lower threshold is higher than upper threshold\n",
1014                         __func__);
1015                 return -ERANGE;
1016         }
1017
1018         ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1019
1020         /* We use function that requires RCU lock */
1021         rcu_read_lock();
1022         dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1023         rcu_read_unlock();
1024
1025         if (dest == NULL) {
1026                 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1027                 return -ENOENT;
1028         }
1029
1030         __ip_vs_update_dest(svc, dest, udest, 0);
1031         LeaveFunction(2);
1032
1033         return 0;
1034 }
1035
1036 /*
1037  *      Delete a destination (must be already unlinked from the service)
1038  */
1039 static void __ip_vs_del_dest(struct net *net, struct ip_vs_dest *dest,
1040                              bool cleanup)
1041 {
1042         struct netns_ipvs *ipvs = net_ipvs(net);
1043
1044         ip_vs_stop_estimator(net, &dest->stats);
1045
1046         /*
1047          *  Remove it from the d-linked list with the real services.
1048          */
1049         ip_vs_rs_unhash(dest);
1050
1051         spin_lock_bh(&ipvs->dest_trash_lock);
1052         IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1053                       IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1054                       atomic_read(&dest->refcnt));
1055         if (list_empty(&ipvs->dest_trash) && !cleanup)
1056                 mod_timer(&ipvs->dest_trash_timer,
1057                           jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1058         /* dest lives in trash without reference */
1059         list_add(&dest->t_list, &ipvs->dest_trash);
1060         dest->idle_start = 0;
1061         spin_unlock_bh(&ipvs->dest_trash_lock);
1062         ip_vs_dest_put(dest);
1063 }
1064
1065
1066 /*
1067  *      Unlink a destination from the given service
1068  */
1069 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1070                                 struct ip_vs_dest *dest,
1071                                 int svcupd)
1072 {
1073         dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1074
1075         /*
1076          *  Remove it from the d-linked destination list.
1077          */
1078         list_del_rcu(&dest->n_list);
1079         svc->num_dests--;
1080
1081         if (dest->af != svc->af)
1082                 net_ipvs(svc->net)->mixed_address_family_dests--;
1083
1084         if (svcupd) {
1085                 struct ip_vs_scheduler *sched;
1086
1087                 sched = rcu_dereference_protected(svc->scheduler, 1);
1088                 if (sched && sched->del_dest)
1089                         sched->del_dest(svc, dest);
1090         }
1091 }
1092
1093
1094 /*
1095  *      Delete a destination server in the given service
1096  */
1097 static int
1098 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1099 {
1100         struct ip_vs_dest *dest;
1101         __be16 dport = udest->port;
1102
1103         EnterFunction(2);
1104
1105         /* We use function that requires RCU lock */
1106         rcu_read_lock();
1107         dest = ip_vs_lookup_dest(svc, udest->af, &udest->addr, dport);
1108         rcu_read_unlock();
1109
1110         if (dest == NULL) {
1111                 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1112                 return -ENOENT;
1113         }
1114
1115         /*
1116          *      Unlink dest from the service
1117          */
1118         __ip_vs_unlink_dest(svc, dest, 1);
1119
1120         /*
1121          *      Delete the destination
1122          */
1123         __ip_vs_del_dest(svc->net, dest, false);
1124
1125         LeaveFunction(2);
1126
1127         return 0;
1128 }
1129
1130 static void ip_vs_dest_trash_expire(unsigned long data)
1131 {
1132         struct net *net = (struct net *) data;
1133         struct netns_ipvs *ipvs = net_ipvs(net);
1134         struct ip_vs_dest *dest, *next;
1135         unsigned long now = jiffies;
1136
1137         spin_lock(&ipvs->dest_trash_lock);
1138         list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1139                 if (atomic_read(&dest->refcnt) > 0)
1140                         continue;
1141                 if (dest->idle_start) {
1142                         if (time_before(now, dest->idle_start +
1143                                              IP_VS_DEST_TRASH_PERIOD))
1144                                 continue;
1145                 } else {
1146                         dest->idle_start = max(1UL, now);
1147                         continue;
1148                 }
1149                 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1150                               dest->vfwmark,
1151                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1152                               ntohs(dest->port));
1153                 list_del(&dest->t_list);
1154                 ip_vs_dest_free(dest);
1155         }
1156         if (!list_empty(&ipvs->dest_trash))
1157                 mod_timer(&ipvs->dest_trash_timer,
1158                           jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1159         spin_unlock(&ipvs->dest_trash_lock);
1160 }
1161
1162 /*
1163  *      Add a service into the service hash table
1164  */
1165 static int
1166 ip_vs_add_service(struct net *net, struct ip_vs_service_user_kern *u,
1167                   struct ip_vs_service **svc_p)
1168 {
1169         int ret = 0, i;
1170         struct ip_vs_scheduler *sched = NULL;
1171         struct ip_vs_pe *pe = NULL;
1172         struct ip_vs_service *svc = NULL;
1173         struct netns_ipvs *ipvs = net_ipvs(net);
1174
1175         /* increase the module use count */
1176         ip_vs_use_count_inc();
1177
1178         /* Lookup the scheduler by 'u->sched_name' */
1179         if (strcmp(u->sched_name, "none")) {
1180                 sched = ip_vs_scheduler_get(u->sched_name);
1181                 if (!sched) {
1182                         pr_info("Scheduler module ip_vs_%s not found\n",
1183                                 u->sched_name);
1184                         ret = -ENOENT;
1185                         goto out_err;
1186                 }
1187         }
1188
1189         if (u->pe_name && *u->pe_name) {
1190                 pe = ip_vs_pe_getbyname(u->pe_name);
1191                 if (pe == NULL) {
1192                         pr_info("persistence engine module ip_vs_pe_%s "
1193                                 "not found\n", u->pe_name);
1194                         ret = -ENOENT;
1195                         goto out_err;
1196                 }
1197         }
1198
1199 #ifdef CONFIG_IP_VS_IPV6
1200         if (u->af == AF_INET6) {
1201                 __u32 plen = (__force __u32) u->netmask;
1202
1203                 if (plen < 1 || plen > 128) {
1204                         ret = -EINVAL;
1205                         goto out_err;
1206                 }
1207         }
1208 #endif
1209
1210         svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1211         if (svc == NULL) {
1212                 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1213                 ret = -ENOMEM;
1214                 goto out_err;
1215         }
1216         svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1217         if (!svc->stats.cpustats) {
1218                 ret = -ENOMEM;
1219                 goto out_err;
1220         }
1221
1222         for_each_possible_cpu(i) {
1223                 struct ip_vs_cpu_stats *ip_vs_stats;
1224                 ip_vs_stats = per_cpu_ptr(svc->stats.cpustats, i);
1225                 u64_stats_init(&ip_vs_stats->syncp);
1226         }
1227
1228
1229         /* I'm the first user of the service */
1230         atomic_set(&svc->refcnt, 0);
1231
1232         svc->af = u->af;
1233         svc->protocol = u->protocol;
1234         ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1235         svc->port = u->port;
1236         svc->fwmark = u->fwmark;
1237         svc->flags = u->flags;
1238         svc->timeout = u->timeout * HZ;
1239         svc->netmask = u->netmask;
1240         svc->net = net;
1241
1242         INIT_LIST_HEAD(&svc->destinations);
1243         spin_lock_init(&svc->sched_lock);
1244         spin_lock_init(&svc->stats.lock);
1245
1246         /* Bind the scheduler */
1247         if (sched) {
1248                 ret = ip_vs_bind_scheduler(svc, sched);
1249                 if (ret)
1250                         goto out_err;
1251                 sched = NULL;
1252         }
1253
1254         /* Bind the ct retriever */
1255         RCU_INIT_POINTER(svc->pe, pe);
1256         pe = NULL;
1257
1258         /* Update the virtual service counters */
1259         if (svc->port == FTPPORT)
1260                 atomic_inc(&ipvs->ftpsvc_counter);
1261         else if (svc->port == 0)
1262                 atomic_inc(&ipvs->nullsvc_counter);
1263
1264         ip_vs_start_estimator(net, &svc->stats);
1265
1266         /* Count only IPv4 services for old get/setsockopt interface */
1267         if (svc->af == AF_INET)
1268                 ipvs->num_services++;
1269
1270         /* Hash the service into the service table */
1271         ip_vs_svc_hash(svc);
1272
1273         *svc_p = svc;
1274         /* Now there is a service - full throttle */
1275         ipvs->enable = 1;
1276         return 0;
1277
1278
1279  out_err:
1280         if (svc != NULL) {
1281                 ip_vs_unbind_scheduler(svc, sched);
1282                 ip_vs_service_free(svc);
1283         }
1284         ip_vs_scheduler_put(sched);
1285         ip_vs_pe_put(pe);
1286
1287         /* decrease the module use count */
1288         ip_vs_use_count_dec();
1289
1290         return ret;
1291 }
1292
1293
1294 /*
1295  *      Edit a service and bind it with a new scheduler
1296  */
1297 static int
1298 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1299 {
1300         struct ip_vs_scheduler *sched = NULL, *old_sched;
1301         struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1302         int ret = 0;
1303
1304         /*
1305          * Lookup the scheduler, by 'u->sched_name'
1306          */
1307         if (strcmp(u->sched_name, "none")) {
1308                 sched = ip_vs_scheduler_get(u->sched_name);
1309                 if (!sched) {
1310                         pr_info("Scheduler module ip_vs_%s not found\n",
1311                                 u->sched_name);
1312                         return -ENOENT;
1313                 }
1314         }
1315         old_sched = sched;
1316
1317         if (u->pe_name && *u->pe_name) {
1318                 pe = ip_vs_pe_getbyname(u->pe_name);
1319                 if (pe == NULL) {
1320                         pr_info("persistence engine module ip_vs_pe_%s "
1321                                 "not found\n", u->pe_name);
1322                         ret = -ENOENT;
1323                         goto out;
1324                 }
1325                 old_pe = pe;
1326         }
1327
1328 #ifdef CONFIG_IP_VS_IPV6
1329         if (u->af == AF_INET6) {
1330                 __u32 plen = (__force __u32) u->netmask;
1331
1332                 if (plen < 1 || plen > 128) {
1333                         ret = -EINVAL;
1334                         goto out;
1335                 }
1336         }
1337 #endif
1338
1339         old_sched = rcu_dereference_protected(svc->scheduler, 1);
1340         if (sched != old_sched) {
1341                 if (old_sched) {
1342                         ip_vs_unbind_scheduler(svc, old_sched);
1343                         RCU_INIT_POINTER(svc->scheduler, NULL);
1344                         /* Wait all svc->sched_data users */
1345                         synchronize_rcu();
1346                 }
1347                 /* Bind the new scheduler */
1348                 if (sched) {
1349                         ret = ip_vs_bind_scheduler(svc, sched);
1350                         if (ret) {
1351                                 ip_vs_scheduler_put(sched);
1352                                 goto out;
1353                         }
1354                 }
1355         }
1356
1357         /*
1358          * Set the flags and timeout value
1359          */
1360         svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1361         svc->timeout = u->timeout * HZ;
1362         svc->netmask = u->netmask;
1363
1364         old_pe = rcu_dereference_protected(svc->pe, 1);
1365         if (pe != old_pe)
1366                 rcu_assign_pointer(svc->pe, pe);
1367
1368 out:
1369         ip_vs_scheduler_put(old_sched);
1370         ip_vs_pe_put(old_pe);
1371         return ret;
1372 }
1373
1374 /*
1375  *      Delete a service from the service list
1376  *      - The service must be unlinked, unlocked and not referenced!
1377  *      - We are called under _bh lock
1378  */
1379 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1380 {
1381         struct ip_vs_dest *dest, *nxt;
1382         struct ip_vs_scheduler *old_sched;
1383         struct ip_vs_pe *old_pe;
1384         struct netns_ipvs *ipvs = net_ipvs(svc->net);
1385
1386         pr_info("%s: enter\n", __func__);
1387
1388         /* Count only IPv4 services for old get/setsockopt interface */
1389         if (svc->af == AF_INET)
1390                 ipvs->num_services--;
1391
1392         ip_vs_stop_estimator(svc->net, &svc->stats);
1393
1394         /* Unbind scheduler */
1395         old_sched = rcu_dereference_protected(svc->scheduler, 1);
1396         ip_vs_unbind_scheduler(svc, old_sched);
1397         ip_vs_scheduler_put(old_sched);
1398
1399         /* Unbind persistence engine, keep svc->pe */
1400         old_pe = rcu_dereference_protected(svc->pe, 1);
1401         ip_vs_pe_put(old_pe);
1402
1403         /*
1404          *    Unlink the whole destination list
1405          */
1406         list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1407                 __ip_vs_unlink_dest(svc, dest, 0);
1408                 __ip_vs_del_dest(svc->net, dest, cleanup);
1409         }
1410
1411         /*
1412          *    Update the virtual service counters
1413          */
1414         if (svc->port == FTPPORT)
1415                 atomic_dec(&ipvs->ftpsvc_counter);
1416         else if (svc->port == 0)
1417                 atomic_dec(&ipvs->nullsvc_counter);
1418
1419         /*
1420          *    Free the service if nobody refers to it
1421          */
1422         __ip_vs_svc_put(svc, true);
1423
1424         /* decrease the module use count */
1425         ip_vs_use_count_dec();
1426 }
1427
1428 /*
1429  * Unlink a service from list and try to delete it if its refcnt reached 0
1430  */
1431 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1432 {
1433         /* Hold svc to avoid double release from dest_trash */
1434         atomic_inc(&svc->refcnt);
1435         /*
1436          * Unhash it from the service table
1437          */
1438         ip_vs_svc_unhash(svc);
1439
1440         __ip_vs_del_service(svc, cleanup);
1441 }
1442
1443 /*
1444  *      Delete a service from the service list
1445  */
1446 static int ip_vs_del_service(struct ip_vs_service *svc)
1447 {
1448         if (svc == NULL)
1449                 return -EEXIST;
1450         ip_vs_unlink_service(svc, false);
1451
1452         return 0;
1453 }
1454
1455
1456 /*
1457  *      Flush all the virtual services
1458  */
1459 static int ip_vs_flush(struct net *net, bool cleanup)
1460 {
1461         int idx;
1462         struct ip_vs_service *svc;
1463         struct hlist_node *n;
1464
1465         /*
1466          * Flush the service table hashed by <netns,protocol,addr,port>
1467          */
1468         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1469                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1470                                           s_list) {
1471                         if (net_eq(svc->net, net))
1472                                 ip_vs_unlink_service(svc, cleanup);
1473                 }
1474         }
1475
1476         /*
1477          * Flush the service table hashed by fwmark
1478          */
1479         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1480                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1481                                           f_list) {
1482                         if (net_eq(svc->net, net))
1483                                 ip_vs_unlink_service(svc, cleanup);
1484                 }
1485         }
1486
1487         return 0;
1488 }
1489
1490 /*
1491  *      Delete service by {netns} in the service table.
1492  *      Called by __ip_vs_cleanup()
1493  */
1494 void ip_vs_service_net_cleanup(struct net *net)
1495 {
1496         EnterFunction(2);
1497         /* Check for "full" addressed entries */
1498         mutex_lock(&__ip_vs_mutex);
1499         ip_vs_flush(net, true);
1500         mutex_unlock(&__ip_vs_mutex);
1501         LeaveFunction(2);
1502 }
1503
1504 /* Put all references for device (dst_cache) */
1505 static inline void
1506 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1507 {
1508         struct ip_vs_dest_dst *dest_dst;
1509
1510         spin_lock_bh(&dest->dst_lock);
1511         dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1512         if (dest_dst && dest_dst->dst_cache->dev == dev) {
1513                 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1514                               dev->name,
1515                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1516                               ntohs(dest->port),
1517                               atomic_read(&dest->refcnt));
1518                 __ip_vs_dst_cache_reset(dest);
1519         }
1520         spin_unlock_bh(&dest->dst_lock);
1521
1522 }
1523 /* Netdev event receiver
1524  * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1525  */
1526 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1527                            void *ptr)
1528 {
1529         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1530         struct net *net = dev_net(dev);
1531         struct netns_ipvs *ipvs = net_ipvs(net);
1532         struct ip_vs_service *svc;
1533         struct ip_vs_dest *dest;
1534         unsigned int idx;
1535
1536         if (event != NETDEV_DOWN || !ipvs)
1537                 return NOTIFY_DONE;
1538         IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1539         EnterFunction(2);
1540         mutex_lock(&__ip_vs_mutex);
1541         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1542                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1543                         if (net_eq(svc->net, net)) {
1544                                 list_for_each_entry(dest, &svc->destinations,
1545                                                     n_list) {
1546                                         ip_vs_forget_dev(dest, dev);
1547                                 }
1548                         }
1549                 }
1550
1551                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1552                         if (net_eq(svc->net, net)) {
1553                                 list_for_each_entry(dest, &svc->destinations,
1554                                                     n_list) {
1555                                         ip_vs_forget_dev(dest, dev);
1556                                 }
1557                         }
1558
1559                 }
1560         }
1561
1562         spin_lock_bh(&ipvs->dest_trash_lock);
1563         list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1564                 ip_vs_forget_dev(dest, dev);
1565         }
1566         spin_unlock_bh(&ipvs->dest_trash_lock);
1567         mutex_unlock(&__ip_vs_mutex);
1568         LeaveFunction(2);
1569         return NOTIFY_DONE;
1570 }
1571
1572 /*
1573  *      Zero counters in a service or all services
1574  */
1575 static int ip_vs_zero_service(struct ip_vs_service *svc)
1576 {
1577         struct ip_vs_dest *dest;
1578
1579         list_for_each_entry(dest, &svc->destinations, n_list) {
1580                 ip_vs_zero_stats(&dest->stats);
1581         }
1582         ip_vs_zero_stats(&svc->stats);
1583         return 0;
1584 }
1585
1586 static int ip_vs_zero_all(struct net *net)
1587 {
1588         int idx;
1589         struct ip_vs_service *svc;
1590
1591         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1592                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1593                         if (net_eq(svc->net, net))
1594                                 ip_vs_zero_service(svc);
1595                 }
1596         }
1597
1598         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1599                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1600                         if (net_eq(svc->net, net))
1601                                 ip_vs_zero_service(svc);
1602                 }
1603         }
1604
1605         ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1606         return 0;
1607 }
1608
1609 #ifdef CONFIG_SYSCTL
1610
1611 static int zero;
1612 static int three = 3;
1613
1614 static int
1615 proc_do_defense_mode(struct ctl_table *table, int write,
1616                      void __user *buffer, size_t *lenp, loff_t *ppos)
1617 {
1618         struct net *net = current->nsproxy->net_ns;
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 > 3)) {
1626                         /* Restore the correct value */
1627                         *valp = val;
1628                 } else {
1629                         update_defense_level(net_ipvs(net));
1630                 }
1631         }
1632         return rc;
1633 }
1634
1635 static int
1636 proc_do_sync_threshold(struct ctl_table *table, int write,
1637                        void __user *buffer, size_t *lenp, loff_t *ppos)
1638 {
1639         int *valp = table->data;
1640         int val[2];
1641         int rc;
1642
1643         /* backup the value first */
1644         memcpy(val, valp, sizeof(val));
1645
1646         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1647         if (write && (valp[0] < 0 || valp[1] < 0 ||
1648             (valp[0] >= valp[1] && valp[1]))) {
1649                 /* Restore the correct value */
1650                 memcpy(valp, val, sizeof(val));
1651         }
1652         return rc;
1653 }
1654
1655 static int
1656 proc_do_sync_mode(struct ctl_table *table, int write,
1657                      void __user *buffer, size_t *lenp, loff_t *ppos)
1658 {
1659         int *valp = table->data;
1660         int val = *valp;
1661         int rc;
1662
1663         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1664         if (write && (*valp != val)) {
1665                 if ((*valp < 0) || (*valp > 1)) {
1666                         /* Restore the correct value */
1667                         *valp = val;
1668                 }
1669         }
1670         return rc;
1671 }
1672
1673 static int
1674 proc_do_sync_ports(struct ctl_table *table, int write,
1675                    void __user *buffer, size_t *lenp, loff_t *ppos)
1676 {
1677         int *valp = table->data;
1678         int val = *valp;
1679         int rc;
1680
1681         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1682         if (write && (*valp != val)) {
1683                 if (*valp < 1 || !is_power_of_2(*valp)) {
1684                         /* Restore the correct value */
1685                         *valp = val;
1686                 }
1687         }
1688         return rc;
1689 }
1690
1691 /*
1692  *      IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1693  *      Do not change order or insert new entries without
1694  *      align with netns init in ip_vs_control_net_init()
1695  */
1696
1697 static struct ctl_table vs_vars[] = {
1698         {
1699                 .procname       = "amemthresh",
1700                 .maxlen         = sizeof(int),
1701                 .mode           = 0644,
1702                 .proc_handler   = proc_dointvec,
1703         },
1704         {
1705                 .procname       = "am_droprate",
1706                 .maxlen         = sizeof(int),
1707                 .mode           = 0644,
1708                 .proc_handler   = proc_dointvec,
1709         },
1710         {
1711                 .procname       = "drop_entry",
1712                 .maxlen         = sizeof(int),
1713                 .mode           = 0644,
1714                 .proc_handler   = proc_do_defense_mode,
1715         },
1716         {
1717                 .procname       = "drop_packet",
1718                 .maxlen         = sizeof(int),
1719                 .mode           = 0644,
1720                 .proc_handler   = proc_do_defense_mode,
1721         },
1722 #ifdef CONFIG_IP_VS_NFCT
1723         {
1724                 .procname       = "conntrack",
1725                 .maxlen         = sizeof(int),
1726                 .mode           = 0644,
1727                 .proc_handler   = &proc_dointvec,
1728         },
1729 #endif
1730         {
1731                 .procname       = "secure_tcp",
1732                 .maxlen         = sizeof(int),
1733                 .mode           = 0644,
1734                 .proc_handler   = proc_do_defense_mode,
1735         },
1736         {
1737                 .procname       = "snat_reroute",
1738                 .maxlen         = sizeof(int),
1739                 .mode           = 0644,
1740                 .proc_handler   = &proc_dointvec,
1741         },
1742         {
1743                 .procname       = "sync_version",
1744                 .maxlen         = sizeof(int),
1745                 .mode           = 0644,
1746                 .proc_handler   = &proc_do_sync_mode,
1747         },
1748         {
1749                 .procname       = "sync_ports",
1750                 .maxlen         = sizeof(int),
1751                 .mode           = 0644,
1752                 .proc_handler   = &proc_do_sync_ports,
1753         },
1754         {
1755                 .procname       = "sync_persist_mode",
1756                 .maxlen         = sizeof(int),
1757                 .mode           = 0644,
1758                 .proc_handler   = proc_dointvec,
1759         },
1760         {
1761                 .procname       = "sync_qlen_max",
1762                 .maxlen         = sizeof(unsigned long),
1763                 .mode           = 0644,
1764                 .proc_handler   = proc_doulongvec_minmax,
1765         },
1766         {
1767                 .procname       = "sync_sock_size",
1768                 .maxlen         = sizeof(int),
1769                 .mode           = 0644,
1770                 .proc_handler   = proc_dointvec,
1771         },
1772         {
1773                 .procname       = "cache_bypass",
1774                 .maxlen         = sizeof(int),
1775                 .mode           = 0644,
1776                 .proc_handler   = proc_dointvec,
1777         },
1778         {
1779                 .procname       = "expire_nodest_conn",
1780                 .maxlen         = sizeof(int),
1781                 .mode           = 0644,
1782                 .proc_handler   = proc_dointvec,
1783         },
1784         {
1785                 .procname       = "sloppy_tcp",
1786                 .maxlen         = sizeof(int),
1787                 .mode           = 0644,
1788                 .proc_handler   = proc_dointvec,
1789         },
1790         {
1791                 .procname       = "sloppy_sctp",
1792                 .maxlen         = sizeof(int),
1793                 .mode           = 0644,
1794                 .proc_handler   = proc_dointvec,
1795         },
1796         {
1797                 .procname       = "expire_quiescent_template",
1798                 .maxlen         = sizeof(int),
1799                 .mode           = 0644,
1800                 .proc_handler   = proc_dointvec,
1801         },
1802         {
1803                 .procname       = "sync_threshold",
1804                 .maxlen         =
1805                         sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1806                 .mode           = 0644,
1807                 .proc_handler   = proc_do_sync_threshold,
1808         },
1809         {
1810                 .procname       = "sync_refresh_period",
1811                 .maxlen         = sizeof(int),
1812                 .mode           = 0644,
1813                 .proc_handler   = proc_dointvec_jiffies,
1814         },
1815         {
1816                 .procname       = "sync_retries",
1817                 .maxlen         = sizeof(int),
1818                 .mode           = 0644,
1819                 .proc_handler   = proc_dointvec_minmax,
1820                 .extra1         = &zero,
1821                 .extra2         = &three,
1822         },
1823         {
1824                 .procname       = "nat_icmp_send",
1825                 .maxlen         = sizeof(int),
1826                 .mode           = 0644,
1827                 .proc_handler   = proc_dointvec,
1828         },
1829         {
1830                 .procname       = "pmtu_disc",
1831                 .maxlen         = sizeof(int),
1832                 .mode           = 0644,
1833                 .proc_handler   = proc_dointvec,
1834         },
1835         {
1836                 .procname       = "backup_only",
1837                 .maxlen         = sizeof(int),
1838                 .mode           = 0644,
1839                 .proc_handler   = proc_dointvec,
1840         },
1841         {
1842                 .procname       = "conn_reuse_mode",
1843                 .maxlen         = sizeof(int),
1844                 .mode           = 0644,
1845                 .proc_handler   = proc_dointvec,
1846         },
1847 #ifdef CONFIG_IP_VS_DEBUG
1848         {
1849                 .procname       = "debug_level",
1850                 .data           = &sysctl_ip_vs_debug_level,
1851                 .maxlen         = sizeof(int),
1852                 .mode           = 0644,
1853                 .proc_handler   = proc_dointvec,
1854         },
1855 #endif
1856         { }
1857 };
1858
1859 #endif
1860
1861 #ifdef CONFIG_PROC_FS
1862
1863 struct ip_vs_iter {
1864         struct seq_net_private p;  /* Do not move this, netns depends upon it*/
1865         struct hlist_head *table;
1866         int bucket;
1867 };
1868
1869 /*
1870  *      Write the contents of the VS rule table to a PROCfs file.
1871  *      (It is kept just for backward compatibility)
1872  */
1873 static inline const char *ip_vs_fwd_name(unsigned int flags)
1874 {
1875         switch (flags & IP_VS_CONN_F_FWD_MASK) {
1876         case IP_VS_CONN_F_LOCALNODE:
1877                 return "Local";
1878         case IP_VS_CONN_F_TUNNEL:
1879                 return "Tunnel";
1880         case IP_VS_CONN_F_DROUTE:
1881                 return "Route";
1882         default:
1883                 return "Masq";
1884         }
1885 }
1886
1887
1888 /* Get the Nth entry in the two lists */
1889 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1890 {
1891         struct net *net = seq_file_net(seq);
1892         struct ip_vs_iter *iter = seq->private;
1893         int idx;
1894         struct ip_vs_service *svc;
1895
1896         /* look in hash by protocol */
1897         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1898                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1899                         if (net_eq(svc->net, net) && pos-- == 0) {
1900                                 iter->table = ip_vs_svc_table;
1901                                 iter->bucket = idx;
1902                                 return svc;
1903                         }
1904                 }
1905         }
1906
1907         /* keep looking in fwmark */
1908         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1909                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1910                                          f_list) {
1911                         if (net_eq(svc->net, net) && pos-- == 0) {
1912                                 iter->table = ip_vs_svc_fwm_table;
1913                                 iter->bucket = idx;
1914                                 return svc;
1915                         }
1916                 }
1917         }
1918
1919         return NULL;
1920 }
1921
1922 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1923         __acquires(RCU)
1924 {
1925         rcu_read_lock();
1926         return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1927 }
1928
1929
1930 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1931 {
1932         struct hlist_node *e;
1933         struct ip_vs_iter *iter;
1934         struct ip_vs_service *svc;
1935
1936         ++*pos;
1937         if (v == SEQ_START_TOKEN)
1938                 return ip_vs_info_array(seq,0);
1939
1940         svc = v;
1941         iter = seq->private;
1942
1943         if (iter->table == ip_vs_svc_table) {
1944                 /* next service in table hashed by protocol */
1945                 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
1946                 if (e)
1947                         return hlist_entry(e, struct ip_vs_service, s_list);
1948
1949                 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1950                         hlist_for_each_entry_rcu(svc,
1951                                                  &ip_vs_svc_table[iter->bucket],
1952                                                  s_list) {
1953                                 return svc;
1954                         }
1955                 }
1956
1957                 iter->table = ip_vs_svc_fwm_table;
1958                 iter->bucket = -1;
1959                 goto scan_fwmark;
1960         }
1961
1962         /* next service in hashed by fwmark */
1963         e = rcu_dereference(hlist_next_rcu(&svc->f_list));
1964         if (e)
1965                 return hlist_entry(e, struct ip_vs_service, f_list);
1966
1967  scan_fwmark:
1968         while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1969                 hlist_for_each_entry_rcu(svc,
1970                                          &ip_vs_svc_fwm_table[iter->bucket],
1971                                          f_list)
1972                         return svc;
1973         }
1974
1975         return NULL;
1976 }
1977
1978 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1979         __releases(RCU)
1980 {
1981         rcu_read_unlock();
1982 }
1983
1984
1985 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1986 {
1987         if (v == SEQ_START_TOKEN) {
1988                 seq_printf(seq,
1989                         "IP Virtual Server version %d.%d.%d (size=%d)\n",
1990                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1991                 seq_puts(seq,
1992                          "Prot LocalAddress:Port Scheduler Flags\n");
1993                 seq_puts(seq,
1994                          "  -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1995         } else {
1996                 const struct ip_vs_service *svc = v;
1997                 const struct ip_vs_iter *iter = seq->private;
1998                 const struct ip_vs_dest *dest;
1999                 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2000                 char *sched_name = sched ? sched->name : "none";
2001
2002                 if (iter->table == ip_vs_svc_table) {
2003 #ifdef CONFIG_IP_VS_IPV6
2004                         if (svc->af == AF_INET6)
2005                                 seq_printf(seq, "%s  [%pI6]:%04X %s ",
2006                                            ip_vs_proto_name(svc->protocol),
2007                                            &svc->addr.in6,
2008                                            ntohs(svc->port),
2009                                            sched_name);
2010                         else
2011 #endif
2012                                 seq_printf(seq, "%s  %08X:%04X %s %s ",
2013                                            ip_vs_proto_name(svc->protocol),
2014                                            ntohl(svc->addr.ip),
2015                                            ntohs(svc->port),
2016                                            sched_name,
2017                                            (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2018                 } else {
2019                         seq_printf(seq, "FWM  %08X %s %s",
2020                                    svc->fwmark, sched_name,
2021                                    (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2022                 }
2023
2024                 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2025                         seq_printf(seq, "persistent %d %08X\n",
2026                                 svc->timeout,
2027                                 ntohl(svc->netmask));
2028                 else
2029                         seq_putc(seq, '\n');
2030
2031                 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2032 #ifdef CONFIG_IP_VS_IPV6
2033                         if (dest->af == AF_INET6)
2034                                 seq_printf(seq,
2035                                            "  -> [%pI6]:%04X"
2036                                            "      %-7s %-6d %-10d %-10d\n",
2037                                            &dest->addr.in6,
2038                                            ntohs(dest->port),
2039                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2040                                            atomic_read(&dest->weight),
2041                                            atomic_read(&dest->activeconns),
2042                                            atomic_read(&dest->inactconns));
2043                         else
2044 #endif
2045                                 seq_printf(seq,
2046                                            "  -> %08X:%04X      "
2047                                            "%-7s %-6d %-10d %-10d\n",
2048                                            ntohl(dest->addr.ip),
2049                                            ntohs(dest->port),
2050                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2051                                            atomic_read(&dest->weight),
2052                                            atomic_read(&dest->activeconns),
2053                                            atomic_read(&dest->inactconns));
2054
2055                 }
2056         }
2057         return 0;
2058 }
2059
2060 static const struct seq_operations ip_vs_info_seq_ops = {
2061         .start = ip_vs_info_seq_start,
2062         .next  = ip_vs_info_seq_next,
2063         .stop  = ip_vs_info_seq_stop,
2064         .show  = ip_vs_info_seq_show,
2065 };
2066
2067 static int ip_vs_info_open(struct inode *inode, struct file *file)
2068 {
2069         return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2070                         sizeof(struct ip_vs_iter));
2071 }
2072
2073 static const struct file_operations ip_vs_info_fops = {
2074         .owner   = THIS_MODULE,
2075         .open    = ip_vs_info_open,
2076         .read    = seq_read,
2077         .llseek  = seq_lseek,
2078         .release = seq_release_net,
2079 };
2080
2081 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2082 {
2083         struct net *net = seq_file_single_net(seq);
2084         struct ip_vs_kstats show;
2085
2086 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2087         seq_puts(seq,
2088                  "   Total Incoming Outgoing         Incoming         Outgoing\n");
2089         seq_printf(seq,
2090                    "   Conns  Packets  Packets            Bytes            Bytes\n");
2091
2092         ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2093         seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n",
2094                    (unsigned long long)show.conns,
2095                    (unsigned long long)show.inpkts,
2096                    (unsigned long long)show.outpkts,
2097                    (unsigned long long)show.inbytes,
2098                    (unsigned long long)show.outbytes);
2099
2100 /*                01234567 01234567 01234567 0123456701234567 0123456701234567*/
2101         seq_puts(seq,
2102                  " Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2103         seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n",
2104                    (unsigned long long)show.cps,
2105                    (unsigned long long)show.inpps,
2106                    (unsigned long long)show.outpps,
2107                    (unsigned long long)show.inbps,
2108                    (unsigned long long)show.outbps);
2109
2110         return 0;
2111 }
2112
2113 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2114 {
2115         return single_open_net(inode, file, ip_vs_stats_show);
2116 }
2117
2118 static const struct file_operations ip_vs_stats_fops = {
2119         .owner = THIS_MODULE,
2120         .open = ip_vs_stats_seq_open,
2121         .read = seq_read,
2122         .llseek = seq_lseek,
2123         .release = single_release_net,
2124 };
2125
2126 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2127 {
2128         struct net *net = seq_file_single_net(seq);
2129         struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2130         struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2131         struct ip_vs_kstats kstats;
2132         int i;
2133
2134 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2135         seq_puts(seq,
2136                  "       Total Incoming Outgoing         Incoming         Outgoing\n");
2137         seq_printf(seq,
2138                    "CPU    Conns  Packets  Packets            Bytes            Bytes\n");
2139
2140         for_each_possible_cpu(i) {
2141                 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2142                 unsigned int start;
2143                 u64 conns, inpkts, outpkts, inbytes, outbytes;
2144
2145                 do {
2146                         start = u64_stats_fetch_begin_irq(&u->syncp);
2147                         conns = u->cnt.conns;
2148                         inpkts = u->cnt.inpkts;
2149                         outpkts = u->cnt.outpkts;
2150                         inbytes = u->cnt.inbytes;
2151                         outbytes = u->cnt.outbytes;
2152                 } while (u64_stats_fetch_retry_irq(&u->syncp, start));
2153
2154                 seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2155                            i, (u64)conns, (u64)inpkts,
2156                            (u64)outpkts, (u64)inbytes,
2157                            (u64)outbytes);
2158         }
2159
2160         ip_vs_copy_stats(&kstats, tot_stats);
2161
2162         seq_printf(seq, "  ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2163                    (unsigned long long)kstats.conns,
2164                    (unsigned long long)kstats.inpkts,
2165                    (unsigned long long)kstats.outpkts,
2166                    (unsigned long long)kstats.inbytes,
2167                    (unsigned long long)kstats.outbytes);
2168
2169 /*                ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2170         seq_puts(seq,
2171                  "     Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2172         seq_printf(seq, "    %8LX %8LX %8LX %16LX %16LX\n",
2173                    kstats.cps,
2174                    kstats.inpps,
2175                    kstats.outpps,
2176                    kstats.inbps,
2177                    kstats.outbps);
2178
2179         return 0;
2180 }
2181
2182 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2183 {
2184         return single_open_net(inode, file, ip_vs_stats_percpu_show);
2185 }
2186
2187 static const struct file_operations ip_vs_stats_percpu_fops = {
2188         .owner = THIS_MODULE,
2189         .open = ip_vs_stats_percpu_seq_open,
2190         .read = seq_read,
2191         .llseek = seq_lseek,
2192         .release = single_release_net,
2193 };
2194 #endif
2195
2196 /*
2197  *      Set timeout values for tcp tcpfin udp in the timeout_table.
2198  */
2199 static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
2200 {
2201 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2202         struct ip_vs_proto_data *pd;
2203 #endif
2204
2205         IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2206                   u->tcp_timeout,
2207                   u->tcp_fin_timeout,
2208                   u->udp_timeout);
2209
2210 #ifdef CONFIG_IP_VS_PROTO_TCP
2211         if (u->tcp_timeout) {
2212                 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2213                 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2214                         = u->tcp_timeout * HZ;
2215         }
2216
2217         if (u->tcp_fin_timeout) {
2218                 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2219                 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2220                         = u->tcp_fin_timeout * HZ;
2221         }
2222 #endif
2223
2224 #ifdef CONFIG_IP_VS_PROTO_UDP
2225         if (u->udp_timeout) {
2226                 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2227                 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2228                         = u->udp_timeout * HZ;
2229         }
2230 #endif
2231         return 0;
2232 }
2233
2234 #define CMDID(cmd)              (cmd - IP_VS_BASE_CTL)
2235
2236 struct ip_vs_svcdest_user {
2237         struct ip_vs_service_user       s;
2238         struct ip_vs_dest_user          d;
2239 };
2240
2241 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2242         [CMDID(IP_VS_SO_SET_ADD)]         = sizeof(struct ip_vs_service_user),
2243         [CMDID(IP_VS_SO_SET_EDIT)]        = sizeof(struct ip_vs_service_user),
2244         [CMDID(IP_VS_SO_SET_DEL)]         = sizeof(struct ip_vs_service_user),
2245         [CMDID(IP_VS_SO_SET_ADDDEST)]     = sizeof(struct ip_vs_svcdest_user),
2246         [CMDID(IP_VS_SO_SET_DELDEST)]     = sizeof(struct ip_vs_svcdest_user),
2247         [CMDID(IP_VS_SO_SET_EDITDEST)]    = sizeof(struct ip_vs_svcdest_user),
2248         [CMDID(IP_VS_SO_SET_TIMEOUT)]     = sizeof(struct ip_vs_timeout_user),
2249         [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2250         [CMDID(IP_VS_SO_SET_STOPDAEMON)]  = sizeof(struct ip_vs_daemon_user),
2251         [CMDID(IP_VS_SO_SET_ZERO)]        = sizeof(struct ip_vs_service_user),
2252 };
2253
2254 union ip_vs_set_arglen {
2255         struct ip_vs_service_user       field_IP_VS_SO_SET_ADD;
2256         struct ip_vs_service_user       field_IP_VS_SO_SET_EDIT;
2257         struct ip_vs_service_user       field_IP_VS_SO_SET_DEL;
2258         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_ADDDEST;
2259         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_DELDEST;
2260         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_EDITDEST;
2261         struct ip_vs_timeout_user       field_IP_VS_SO_SET_TIMEOUT;
2262         struct ip_vs_daemon_user        field_IP_VS_SO_SET_STARTDAEMON;
2263         struct ip_vs_daemon_user        field_IP_VS_SO_SET_STOPDAEMON;
2264         struct ip_vs_service_user       field_IP_VS_SO_SET_ZERO;
2265 };
2266
2267 #define MAX_SET_ARGLEN  sizeof(union ip_vs_set_arglen)
2268
2269 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2270                                   struct ip_vs_service_user *usvc_compat)
2271 {
2272         memset(usvc, 0, sizeof(*usvc));
2273
2274         usvc->af                = AF_INET;
2275         usvc->protocol          = usvc_compat->protocol;
2276         usvc->addr.ip           = usvc_compat->addr;
2277         usvc->port              = usvc_compat->port;
2278         usvc->fwmark            = usvc_compat->fwmark;
2279
2280         /* Deep copy of sched_name is not needed here */
2281         usvc->sched_name        = usvc_compat->sched_name;
2282
2283         usvc->flags             = usvc_compat->flags;
2284         usvc->timeout           = usvc_compat->timeout;
2285         usvc->netmask           = usvc_compat->netmask;
2286 }
2287
2288 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2289                                    struct ip_vs_dest_user *udest_compat)
2290 {
2291         memset(udest, 0, sizeof(*udest));
2292
2293         udest->addr.ip          = udest_compat->addr;
2294         udest->port             = udest_compat->port;
2295         udest->conn_flags       = udest_compat->conn_flags;
2296         udest->weight           = udest_compat->weight;
2297         udest->u_threshold      = udest_compat->u_threshold;
2298         udest->l_threshold      = udest_compat->l_threshold;
2299         udest->af               = AF_INET;
2300 }
2301
2302 static int
2303 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2304 {
2305         struct net *net = sock_net(sk);
2306         int ret;
2307         unsigned char arg[MAX_SET_ARGLEN];
2308         struct ip_vs_service_user *usvc_compat;
2309         struct ip_vs_service_user_kern usvc;
2310         struct ip_vs_service *svc;
2311         struct ip_vs_dest_user *udest_compat;
2312         struct ip_vs_dest_user_kern udest;
2313         struct netns_ipvs *ipvs = net_ipvs(net);
2314
2315         BUILD_BUG_ON(sizeof(arg) > 255);
2316         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2317                 return -EPERM;
2318
2319         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2320                 return -EINVAL;
2321         if (len != set_arglen[CMDID(cmd)]) {
2322                 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2323                           len, set_arglen[CMDID(cmd)]);
2324                 return -EINVAL;
2325         }
2326
2327         if (copy_from_user(arg, user, len) != 0)
2328                 return -EFAULT;
2329
2330         /* increase the module use count */
2331         ip_vs_use_count_inc();
2332
2333         /* Handle daemons since they have another lock */
2334         if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2335             cmd == IP_VS_SO_SET_STOPDAEMON) {
2336                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2337
2338                 mutex_lock(&ipvs->sync_mutex);
2339                 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2340                         ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2341                                                 dm->syncid);
2342                 else
2343                         ret = stop_sync_thread(net, dm->state);
2344                 mutex_unlock(&ipvs->sync_mutex);
2345                 goto out_dec;
2346         }
2347
2348         mutex_lock(&__ip_vs_mutex);
2349         if (cmd == IP_VS_SO_SET_FLUSH) {
2350                 /* Flush the virtual service */
2351                 ret = ip_vs_flush(net, false);
2352                 goto out_unlock;
2353         } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2354                 /* Set timeout values for (tcp tcpfin udp) */
2355                 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
2356                 goto out_unlock;
2357         }
2358
2359         usvc_compat = (struct ip_vs_service_user *)arg;
2360         udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2361
2362         /* We only use the new structs internally, so copy userspace compat
2363          * structs to extended internal versions */
2364         ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2365         ip_vs_copy_udest_compat(&udest, udest_compat);
2366
2367         if (cmd == IP_VS_SO_SET_ZERO) {
2368                 /* if no service address is set, zero counters in all */
2369                 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2370                         ret = ip_vs_zero_all(net);
2371                         goto out_unlock;
2372                 }
2373         }
2374
2375         /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2376         if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2377             usvc.protocol != IPPROTO_SCTP) {
2378                 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2379                        usvc.protocol, &usvc.addr.ip,
2380                        ntohs(usvc.port), usvc.sched_name);
2381                 ret = -EFAULT;
2382                 goto out_unlock;
2383         }
2384
2385         /* Lookup the exact service by <protocol, addr, port> or fwmark */
2386         rcu_read_lock();
2387         if (usvc.fwmark == 0)
2388                 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
2389                                            &usvc.addr, usvc.port);
2390         else
2391                 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
2392         rcu_read_unlock();
2393
2394         if (cmd != IP_VS_SO_SET_ADD
2395             && (svc == NULL || svc->protocol != usvc.protocol)) {
2396                 ret = -ESRCH;
2397                 goto out_unlock;
2398         }
2399
2400         switch (cmd) {
2401         case IP_VS_SO_SET_ADD:
2402                 if (svc != NULL)
2403                         ret = -EEXIST;
2404                 else
2405                         ret = ip_vs_add_service(net, &usvc, &svc);
2406                 break;
2407         case IP_VS_SO_SET_EDIT:
2408                 ret = ip_vs_edit_service(svc, &usvc);
2409                 break;
2410         case IP_VS_SO_SET_DEL:
2411                 ret = ip_vs_del_service(svc);
2412                 if (!ret)
2413                         goto out_unlock;
2414                 break;
2415         case IP_VS_SO_SET_ZERO:
2416                 ret = ip_vs_zero_service(svc);
2417                 break;
2418         case IP_VS_SO_SET_ADDDEST:
2419                 ret = ip_vs_add_dest(svc, &udest);
2420                 break;
2421         case IP_VS_SO_SET_EDITDEST:
2422                 ret = ip_vs_edit_dest(svc, &udest);
2423                 break;
2424         case IP_VS_SO_SET_DELDEST:
2425                 ret = ip_vs_del_dest(svc, &udest);
2426                 break;
2427         default:
2428                 ret = -EINVAL;
2429         }
2430
2431   out_unlock:
2432         mutex_unlock(&__ip_vs_mutex);
2433   out_dec:
2434         /* decrease the module use count */
2435         ip_vs_use_count_dec();
2436
2437         return ret;
2438 }
2439
2440
2441 static void
2442 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2443 {
2444         struct ip_vs_scheduler *sched;
2445         struct ip_vs_kstats kstats;
2446         char *sched_name;
2447
2448         sched = rcu_dereference_protected(src->scheduler, 1);
2449         sched_name = sched ? sched->name : "none";
2450         dst->protocol = src->protocol;
2451         dst->addr = src->addr.ip;
2452         dst->port = src->port;
2453         dst->fwmark = src->fwmark;
2454         strlcpy(dst->sched_name, sched_name, sizeof(dst->sched_name));
2455         dst->flags = src->flags;
2456         dst->timeout = src->timeout / HZ;
2457         dst->netmask = src->netmask;
2458         dst->num_dests = src->num_dests;
2459         ip_vs_copy_stats(&kstats, &src->stats);
2460         ip_vs_export_stats_user(&dst->stats, &kstats);
2461 }
2462
2463 static inline int
2464 __ip_vs_get_service_entries(struct net *net,
2465                             const struct ip_vs_get_services *get,
2466                             struct ip_vs_get_services __user *uptr)
2467 {
2468         int idx, count=0;
2469         struct ip_vs_service *svc;
2470         struct ip_vs_service_entry entry;
2471         int ret = 0;
2472
2473         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2474                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2475                         /* Only expose IPv4 entries to old interface */
2476                         if (svc->af != AF_INET || !net_eq(svc->net, net))
2477                                 continue;
2478
2479                         if (count >= get->num_services)
2480                                 goto out;
2481                         memset(&entry, 0, sizeof(entry));
2482                         ip_vs_copy_service(&entry, svc);
2483                         if (copy_to_user(&uptr->entrytable[count],
2484                                          &entry, sizeof(entry))) {
2485                                 ret = -EFAULT;
2486                                 goto out;
2487                         }
2488                         count++;
2489                 }
2490         }
2491
2492         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2493                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2494                         /* Only expose IPv4 entries to old interface */
2495                         if (svc->af != AF_INET || !net_eq(svc->net, net))
2496                                 continue;
2497
2498                         if (count >= get->num_services)
2499                                 goto out;
2500                         memset(&entry, 0, sizeof(entry));
2501                         ip_vs_copy_service(&entry, svc);
2502                         if (copy_to_user(&uptr->entrytable[count],
2503                                          &entry, sizeof(entry))) {
2504                                 ret = -EFAULT;
2505                                 goto out;
2506                         }
2507                         count++;
2508                 }
2509         }
2510 out:
2511         return ret;
2512 }
2513
2514 static inline int
2515 __ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
2516                          struct ip_vs_get_dests __user *uptr)
2517 {
2518         struct ip_vs_service *svc;
2519         union nf_inet_addr addr = { .ip = get->addr };
2520         int ret = 0;
2521
2522         rcu_read_lock();
2523         if (get->fwmark)
2524                 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
2525         else
2526                 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
2527                                            get->port);
2528         rcu_read_unlock();
2529
2530         if (svc) {
2531                 int count = 0;
2532                 struct ip_vs_dest *dest;
2533                 struct ip_vs_dest_entry entry;
2534                 struct ip_vs_kstats kstats;
2535
2536                 memset(&entry, 0, sizeof(entry));
2537                 list_for_each_entry(dest, &svc->destinations, n_list) {
2538                         if (count >= get->num_dests)
2539                                 break;
2540
2541                         /* Cannot expose heterogeneous members via sockopt
2542                          * interface
2543                          */
2544                         if (dest->af != svc->af)
2545                                 continue;
2546
2547                         entry.addr = dest->addr.ip;
2548                         entry.port = dest->port;
2549                         entry.conn_flags = atomic_read(&dest->conn_flags);
2550                         entry.weight = atomic_read(&dest->weight);
2551                         entry.u_threshold = dest->u_threshold;
2552                         entry.l_threshold = dest->l_threshold;
2553                         entry.activeconns = atomic_read(&dest->activeconns);
2554                         entry.inactconns = atomic_read(&dest->inactconns);
2555                         entry.persistconns = atomic_read(&dest->persistconns);
2556                         ip_vs_copy_stats(&kstats, &dest->stats);
2557                         ip_vs_export_stats_user(&entry.stats, &kstats);
2558                         if (copy_to_user(&uptr->entrytable[count],
2559                                          &entry, sizeof(entry))) {
2560                                 ret = -EFAULT;
2561                                 break;
2562                         }
2563                         count++;
2564                 }
2565         } else
2566                 ret = -ESRCH;
2567         return ret;
2568 }
2569
2570 static inline void
2571 __ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
2572 {
2573 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2574         struct ip_vs_proto_data *pd;
2575 #endif
2576
2577         memset(u, 0, sizeof (*u));
2578
2579 #ifdef CONFIG_IP_VS_PROTO_TCP
2580         pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2581         u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2582         u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2583 #endif
2584 #ifdef CONFIG_IP_VS_PROTO_UDP
2585         pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2586         u->udp_timeout =
2587                         pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2588 #endif
2589 }
2590
2591 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2592         [CMDID(IP_VS_SO_GET_VERSION)]  = 64,
2593         [CMDID(IP_VS_SO_GET_INFO)]     = sizeof(struct ip_vs_getinfo),
2594         [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2595         [CMDID(IP_VS_SO_GET_SERVICE)]  = sizeof(struct ip_vs_service_entry),
2596         [CMDID(IP_VS_SO_GET_DESTS)]    = sizeof(struct ip_vs_get_dests),
2597         [CMDID(IP_VS_SO_GET_TIMEOUT)]  = sizeof(struct ip_vs_timeout_user),
2598         [CMDID(IP_VS_SO_GET_DAEMON)]   = 2 * sizeof(struct ip_vs_daemon_user),
2599 };
2600
2601 union ip_vs_get_arglen {
2602         char                            field_IP_VS_SO_GET_VERSION[64];
2603         struct ip_vs_getinfo            field_IP_VS_SO_GET_INFO;
2604         struct ip_vs_get_services       field_IP_VS_SO_GET_SERVICES;
2605         struct ip_vs_service_entry      field_IP_VS_SO_GET_SERVICE;
2606         struct ip_vs_get_dests          field_IP_VS_SO_GET_DESTS;
2607         struct ip_vs_timeout_user       field_IP_VS_SO_GET_TIMEOUT;
2608         struct ip_vs_daemon_user        field_IP_VS_SO_GET_DAEMON[2];
2609 };
2610
2611 #define MAX_GET_ARGLEN  sizeof(union ip_vs_get_arglen)
2612
2613 static int
2614 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2615 {
2616         unsigned char arg[MAX_GET_ARGLEN];
2617         int ret = 0;
2618         unsigned int copylen;
2619         struct net *net = sock_net(sk);
2620         struct netns_ipvs *ipvs = net_ipvs(net);
2621
2622         BUG_ON(!net);
2623         BUILD_BUG_ON(sizeof(arg) > 255);
2624         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2625                 return -EPERM;
2626
2627         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2628                 return -EINVAL;
2629
2630         copylen = get_arglen[CMDID(cmd)];
2631         if (*len < (int) copylen) {
2632                 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
2633                 return -EINVAL;
2634         }
2635
2636         if (copy_from_user(arg, user, copylen) != 0)
2637                 return -EFAULT;
2638         /*
2639          * Handle daemons first since it has its own locking
2640          */
2641         if (cmd == IP_VS_SO_GET_DAEMON) {
2642                 struct ip_vs_daemon_user d[2];
2643
2644                 memset(&d, 0, sizeof(d));
2645                 mutex_lock(&ipvs->sync_mutex);
2646                 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2647                         d[0].state = IP_VS_STATE_MASTER;
2648                         strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2649                                 sizeof(d[0].mcast_ifn));
2650                         d[0].syncid = ipvs->master_syncid;
2651                 }
2652                 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2653                         d[1].state = IP_VS_STATE_BACKUP;
2654                         strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2655                                 sizeof(d[1].mcast_ifn));
2656                         d[1].syncid = ipvs->backup_syncid;
2657                 }
2658                 if (copy_to_user(user, &d, sizeof(d)) != 0)
2659                         ret = -EFAULT;
2660                 mutex_unlock(&ipvs->sync_mutex);
2661                 return ret;
2662         }
2663
2664         mutex_lock(&__ip_vs_mutex);
2665         switch (cmd) {
2666         case IP_VS_SO_GET_VERSION:
2667         {
2668                 char buf[64];
2669
2670                 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2671                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2672                 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2673                         ret = -EFAULT;
2674                         goto out;
2675                 }
2676                 *len = strlen(buf)+1;
2677         }
2678         break;
2679
2680         case IP_VS_SO_GET_INFO:
2681         {
2682                 struct ip_vs_getinfo info;
2683                 info.version = IP_VS_VERSION_CODE;
2684                 info.size = ip_vs_conn_tab_size;
2685                 info.num_services = ipvs->num_services;
2686                 if (copy_to_user(user, &info, sizeof(info)) != 0)
2687                         ret = -EFAULT;
2688         }
2689         break;
2690
2691         case IP_VS_SO_GET_SERVICES:
2692         {
2693                 struct ip_vs_get_services *get;
2694                 int size;
2695
2696                 get = (struct ip_vs_get_services *)arg;
2697                 size = sizeof(*get) +
2698                         sizeof(struct ip_vs_service_entry) * get->num_services;
2699                 if (*len != size) {
2700                         pr_err("length: %u != %u\n", *len, size);
2701                         ret = -EINVAL;
2702                         goto out;
2703                 }
2704                 ret = __ip_vs_get_service_entries(net, get, user);
2705         }
2706         break;
2707
2708         case IP_VS_SO_GET_SERVICE:
2709         {
2710                 struct ip_vs_service_entry *entry;
2711                 struct ip_vs_service *svc;
2712                 union nf_inet_addr addr;
2713
2714                 entry = (struct ip_vs_service_entry *)arg;
2715                 addr.ip = entry->addr;
2716                 rcu_read_lock();
2717                 if (entry->fwmark)
2718                         svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
2719                 else
2720                         svc = __ip_vs_service_find(net, AF_INET,
2721                                                    entry->protocol, &addr,
2722                                                    entry->port);
2723                 rcu_read_unlock();
2724                 if (svc) {
2725                         ip_vs_copy_service(entry, svc);
2726                         if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2727                                 ret = -EFAULT;
2728                 } else
2729                         ret = -ESRCH;
2730         }
2731         break;
2732
2733         case IP_VS_SO_GET_DESTS:
2734         {
2735                 struct ip_vs_get_dests *get;
2736                 int size;
2737
2738                 get = (struct ip_vs_get_dests *)arg;
2739                 size = sizeof(*get) +
2740                         sizeof(struct ip_vs_dest_entry) * get->num_dests;
2741                 if (*len != size) {
2742                         pr_err("length: %u != %u\n", *len, size);
2743                         ret = -EINVAL;
2744                         goto out;
2745                 }
2746                 ret = __ip_vs_get_dest_entries(net, get, user);
2747         }
2748         break;
2749
2750         case IP_VS_SO_GET_TIMEOUT:
2751         {
2752                 struct ip_vs_timeout_user t;
2753
2754                 __ip_vs_get_timeouts(net, &t);
2755                 if (copy_to_user(user, &t, sizeof(t)) != 0)
2756                         ret = -EFAULT;
2757         }
2758         break;
2759
2760         default:
2761                 ret = -EINVAL;
2762         }
2763
2764 out:
2765         mutex_unlock(&__ip_vs_mutex);
2766         return ret;
2767 }
2768
2769
2770 static struct nf_sockopt_ops ip_vs_sockopts = {
2771         .pf             = PF_INET,
2772         .set_optmin     = IP_VS_BASE_CTL,
2773         .set_optmax     = IP_VS_SO_SET_MAX+1,
2774         .set            = do_ip_vs_set_ctl,
2775         .get_optmin     = IP_VS_BASE_CTL,
2776         .get_optmax     = IP_VS_SO_GET_MAX+1,
2777         .get            = do_ip_vs_get_ctl,
2778         .owner          = THIS_MODULE,
2779 };
2780
2781 /*
2782  * Generic Netlink interface
2783  */
2784
2785 /* IPVS genetlink family */
2786 static struct genl_family ip_vs_genl_family = {
2787         .id             = GENL_ID_GENERATE,
2788         .hdrsize        = 0,
2789         .name           = IPVS_GENL_NAME,
2790         .version        = IPVS_GENL_VERSION,
2791         .maxattr        = IPVS_CMD_MAX,
2792         .netnsok        = true,         /* Make ipvsadm to work on netns */
2793 };
2794
2795 /* Policy used for first-level command attributes */
2796 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2797         [IPVS_CMD_ATTR_SERVICE]         = { .type = NLA_NESTED },
2798         [IPVS_CMD_ATTR_DEST]            = { .type = NLA_NESTED },
2799         [IPVS_CMD_ATTR_DAEMON]          = { .type = NLA_NESTED },
2800         [IPVS_CMD_ATTR_TIMEOUT_TCP]     = { .type = NLA_U32 },
2801         [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2802         [IPVS_CMD_ATTR_TIMEOUT_UDP]     = { .type = NLA_U32 },
2803 };
2804
2805 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2806 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2807         [IPVS_DAEMON_ATTR_STATE]        = { .type = NLA_U32 },
2808         [IPVS_DAEMON_ATTR_MCAST_IFN]    = { .type = NLA_NUL_STRING,
2809                                             .len = IP_VS_IFNAME_MAXLEN },
2810         [IPVS_DAEMON_ATTR_SYNC_ID]      = { .type = NLA_U32 },
2811 };
2812
2813 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2814 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2815         [IPVS_SVC_ATTR_AF]              = { .type = NLA_U16 },
2816         [IPVS_SVC_ATTR_PROTOCOL]        = { .type = NLA_U16 },
2817         [IPVS_SVC_ATTR_ADDR]            = { .type = NLA_BINARY,
2818                                             .len = sizeof(union nf_inet_addr) },
2819         [IPVS_SVC_ATTR_PORT]            = { .type = NLA_U16 },
2820         [IPVS_SVC_ATTR_FWMARK]          = { .type = NLA_U32 },
2821         [IPVS_SVC_ATTR_SCHED_NAME]      = { .type = NLA_NUL_STRING,
2822                                             .len = IP_VS_SCHEDNAME_MAXLEN },
2823         [IPVS_SVC_ATTR_PE_NAME]         = { .type = NLA_NUL_STRING,
2824                                             .len = IP_VS_PENAME_MAXLEN },
2825         [IPVS_SVC_ATTR_FLAGS]           = { .type = NLA_BINARY,
2826                                             .len = sizeof(struct ip_vs_flags) },
2827         [IPVS_SVC_ATTR_TIMEOUT]         = { .type = NLA_U32 },
2828         [IPVS_SVC_ATTR_NETMASK]         = { .type = NLA_U32 },
2829         [IPVS_SVC_ATTR_STATS]           = { .type = NLA_NESTED },
2830 };
2831
2832 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2833 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2834         [IPVS_DEST_ATTR_ADDR]           = { .type = NLA_BINARY,
2835                                             .len = sizeof(union nf_inet_addr) },
2836         [IPVS_DEST_ATTR_PORT]           = { .type = NLA_U16 },
2837         [IPVS_DEST_ATTR_FWD_METHOD]     = { .type = NLA_U32 },
2838         [IPVS_DEST_ATTR_WEIGHT]         = { .type = NLA_U32 },
2839         [IPVS_DEST_ATTR_U_THRESH]       = { .type = NLA_U32 },
2840         [IPVS_DEST_ATTR_L_THRESH]       = { .type = NLA_U32 },
2841         [IPVS_DEST_ATTR_ACTIVE_CONNS]   = { .type = NLA_U32 },
2842         [IPVS_DEST_ATTR_INACT_CONNS]    = { .type = NLA_U32 },
2843         [IPVS_DEST_ATTR_PERSIST_CONNS]  = { .type = NLA_U32 },
2844         [IPVS_DEST_ATTR_STATS]          = { .type = NLA_NESTED },
2845         [IPVS_DEST_ATTR_ADDR_FAMILY]    = { .type = NLA_U16 },
2846 };
2847
2848 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2849                                  struct ip_vs_kstats *kstats)
2850 {
2851         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2852
2853         if (!nl_stats)
2854                 return -EMSGSIZE;
2855
2856         if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) ||
2857             nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) ||
2858             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) ||
2859             nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes) ||
2860             nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes) ||
2861             nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) ||
2862             nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) ||
2863             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) ||
2864             nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) ||
2865             nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps))
2866                 goto nla_put_failure;
2867         nla_nest_end(skb, nl_stats);
2868
2869         return 0;
2870
2871 nla_put_failure:
2872         nla_nest_cancel(skb, nl_stats);
2873         return -EMSGSIZE;
2874 }
2875
2876 static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type,
2877                                    struct ip_vs_kstats *kstats)
2878 {
2879         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2880
2881         if (!nl_stats)
2882                 return -EMSGSIZE;
2883
2884         if (nla_put_u64(skb, IPVS_STATS_ATTR_CONNS, kstats->conns) ||
2885             nla_put_u64(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts) ||
2886             nla_put_u64(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts) ||
2887             nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes) ||
2888             nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes) ||
2889             nla_put_u64(skb, IPVS_STATS_ATTR_CPS, kstats->cps) ||
2890             nla_put_u64(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps) ||
2891             nla_put_u64(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps) ||
2892             nla_put_u64(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps) ||
2893             nla_put_u64(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps))
2894                 goto nla_put_failure;
2895         nla_nest_end(skb, nl_stats);
2896
2897         return 0;
2898
2899 nla_put_failure:
2900         nla_nest_cancel(skb, nl_stats);
2901         return -EMSGSIZE;
2902 }
2903
2904 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2905                                    struct ip_vs_service *svc)
2906 {
2907         struct ip_vs_scheduler *sched;
2908         struct ip_vs_pe *pe;
2909         struct nlattr *nl_service;
2910         struct ip_vs_flags flags = { .flags = svc->flags,
2911                                      .mask = ~0 };
2912         struct ip_vs_kstats kstats;
2913         char *sched_name;
2914
2915         nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2916         if (!nl_service)
2917                 return -EMSGSIZE;
2918
2919         if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2920                 goto nla_put_failure;
2921         if (svc->fwmark) {
2922                 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2923                         goto nla_put_failure;
2924         } else {
2925                 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2926                     nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2927                     nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2928                         goto nla_put_failure;
2929         }
2930
2931         sched = rcu_dereference_protected(svc->scheduler, 1);
2932         sched_name = sched ? sched->name : "none";
2933         pe = rcu_dereference_protected(svc->pe, 1);
2934         if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) ||
2935             (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
2936             nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2937             nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2938             nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2939                 goto nla_put_failure;
2940         ip_vs_copy_stats(&kstats, &svc->stats);
2941         if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats))
2942                 goto nla_put_failure;
2943         if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats))
2944                 goto nla_put_failure;
2945
2946         nla_nest_end(skb, nl_service);
2947
2948         return 0;
2949
2950 nla_put_failure:
2951         nla_nest_cancel(skb, nl_service);
2952         return -EMSGSIZE;
2953 }
2954
2955 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2956                                    struct ip_vs_service *svc,
2957                                    struct netlink_callback *cb)
2958 {
2959         void *hdr;
2960
2961         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2962                           &ip_vs_genl_family, NLM_F_MULTI,
2963                           IPVS_CMD_NEW_SERVICE);
2964         if (!hdr)
2965                 return -EMSGSIZE;
2966
2967         if (ip_vs_genl_fill_service(skb, svc) < 0)
2968                 goto nla_put_failure;
2969
2970         genlmsg_end(skb, hdr);
2971         return 0;
2972
2973 nla_put_failure:
2974         genlmsg_cancel(skb, hdr);
2975         return -EMSGSIZE;
2976 }
2977
2978 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2979                                     struct netlink_callback *cb)
2980 {
2981         int idx = 0, i;
2982         int start = cb->args[0];
2983         struct ip_vs_service *svc;
2984         struct net *net = skb_sknet(skb);
2985
2986         mutex_lock(&__ip_vs_mutex);
2987         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2988                 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2989                         if (++idx <= start || !net_eq(svc->net, net))
2990                                 continue;
2991                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2992                                 idx--;
2993                                 goto nla_put_failure;
2994                         }
2995                 }
2996         }
2997
2998         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2999                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3000                         if (++idx <= start || !net_eq(svc->net, net))
3001                                 continue;
3002                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3003                                 idx--;
3004                                 goto nla_put_failure;
3005                         }
3006                 }
3007         }
3008
3009 nla_put_failure:
3010         mutex_unlock(&__ip_vs_mutex);
3011         cb->args[0] = idx;
3012
3013         return skb->len;
3014 }
3015
3016 static int ip_vs_genl_parse_service(struct net *net,
3017                                     struct ip_vs_service_user_kern *usvc,
3018                                     struct nlattr *nla, int full_entry,
3019                                     struct ip_vs_service **ret_svc)
3020 {
3021         struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3022         struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3023         struct ip_vs_service *svc;
3024
3025         /* Parse mandatory identifying service fields first */
3026         if (nla == NULL ||
3027             nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
3028                 return -EINVAL;
3029
3030         nla_af          = attrs[IPVS_SVC_ATTR_AF];
3031         nla_protocol    = attrs[IPVS_SVC_ATTR_PROTOCOL];
3032         nla_addr        = attrs[IPVS_SVC_ATTR_ADDR];
3033         nla_port        = attrs[IPVS_SVC_ATTR_PORT];
3034         nla_fwmark      = attrs[IPVS_SVC_ATTR_FWMARK];
3035
3036         if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3037                 return -EINVAL;
3038
3039         memset(usvc, 0, sizeof(*usvc));
3040
3041         usvc->af = nla_get_u16(nla_af);
3042 #ifdef CONFIG_IP_VS_IPV6
3043         if (usvc->af != AF_INET && usvc->af != AF_INET6)
3044 #else
3045         if (usvc->af != AF_INET)
3046 #endif
3047                 return -EAFNOSUPPORT;
3048
3049         if (nla_fwmark) {
3050                 usvc->protocol = IPPROTO_TCP;
3051                 usvc->fwmark = nla_get_u32(nla_fwmark);
3052         } else {
3053                 usvc->protocol = nla_get_u16(nla_protocol);
3054                 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3055                 usvc->port = nla_get_be16(nla_port);
3056                 usvc->fwmark = 0;
3057         }
3058
3059         rcu_read_lock();
3060         if (usvc->fwmark)
3061                 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
3062         else
3063                 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
3064                                            &usvc->addr, usvc->port);
3065         rcu_read_unlock();
3066         *ret_svc = svc;
3067
3068         /* If a full entry was requested, check for the additional fields */
3069         if (full_entry) {
3070                 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3071                               *nla_netmask;
3072                 struct ip_vs_flags flags;
3073
3074                 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3075                 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3076                 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3077                 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3078                 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3079
3080                 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3081                         return -EINVAL;
3082
3083                 nla_memcpy(&flags, nla_flags, sizeof(flags));
3084
3085                 /* prefill flags from service if it already exists */
3086                 if (svc)
3087                         usvc->flags = svc->flags;
3088
3089                 /* set new flags from userland */
3090                 usvc->flags = (usvc->flags & ~flags.mask) |
3091                               (flags.flags & flags.mask);
3092                 usvc->sched_name = nla_data(nla_sched);
3093                 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3094                 usvc->timeout = nla_get_u32(nla_timeout);
3095                 usvc->netmask = nla_get_be32(nla_netmask);
3096         }
3097
3098         return 0;
3099 }
3100
3101 static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3102                                                      struct nlattr *nla)
3103 {
3104         struct ip_vs_service_user_kern usvc;
3105         struct ip_vs_service *svc;
3106         int ret;
3107
3108         ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
3109         return ret ? ERR_PTR(ret) : svc;
3110 }
3111
3112 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3113 {
3114         struct nlattr *nl_dest;
3115         struct ip_vs_kstats kstats;
3116
3117         nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3118         if (!nl_dest)
3119                 return -EMSGSIZE;
3120
3121         if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3122             nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3123             nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3124                         (atomic_read(&dest->conn_flags) &
3125                          IP_VS_CONN_F_FWD_MASK)) ||
3126             nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3127                         atomic_read(&dest->weight)) ||
3128             nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3129             nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3130             nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3131                         atomic_read(&dest->activeconns)) ||
3132             nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3133                         atomic_read(&dest->inactconns)) ||
3134             nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3135                         atomic_read(&dest->persistconns)) ||
3136             nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3137                 goto nla_put_failure;
3138         ip_vs_copy_stats(&kstats, &dest->stats);
3139         if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats))
3140                 goto nla_put_failure;
3141         if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats))
3142                 goto nla_put_failure;
3143
3144         nla_nest_end(skb, nl_dest);
3145
3146         return 0;
3147
3148 nla_put_failure:
3149         nla_nest_cancel(skb, nl_dest);
3150         return -EMSGSIZE;
3151 }
3152
3153 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3154                                 struct netlink_callback *cb)
3155 {
3156         void *hdr;
3157
3158         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3159                           &ip_vs_genl_family, NLM_F_MULTI,
3160                           IPVS_CMD_NEW_DEST);
3161         if (!hdr)
3162                 return -EMSGSIZE;
3163
3164         if (ip_vs_genl_fill_dest(skb, dest) < 0)
3165                 goto nla_put_failure;
3166
3167         genlmsg_end(skb, hdr);
3168         return 0;
3169
3170 nla_put_failure:
3171         genlmsg_cancel(skb, hdr);
3172         return -EMSGSIZE;
3173 }
3174
3175 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3176                                  struct netlink_callback *cb)
3177 {
3178         int idx = 0;
3179         int start = cb->args[0];
3180         struct ip_vs_service *svc;
3181         struct ip_vs_dest *dest;
3182         struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3183         struct net *net = skb_sknet(skb);
3184
3185         mutex_lock(&__ip_vs_mutex);
3186
3187         /* Try to find the service for which to dump destinations */
3188         if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3189                         IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3190                 goto out_err;
3191
3192
3193         svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
3194         if (IS_ERR(svc) || svc == NULL)
3195                 goto out_err;
3196
3197         /* Dump the destinations */
3198         list_for_each_entry(dest, &svc->destinations, n_list) {
3199                 if (++idx <= start)
3200                         continue;
3201                 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3202                         idx--;
3203                         goto nla_put_failure;
3204                 }
3205         }
3206
3207 nla_put_failure:
3208         cb->args[0] = idx;
3209
3210 out_err:
3211         mutex_unlock(&__ip_vs_mutex);
3212
3213         return skb->len;
3214 }
3215
3216 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3217                                  struct nlattr *nla, int full_entry)
3218 {
3219         struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3220         struct nlattr *nla_addr, *nla_port;
3221         struct nlattr *nla_addr_family;
3222
3223         /* Parse mandatory identifying destination fields first */
3224         if (nla == NULL ||
3225             nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3226                 return -EINVAL;
3227
3228         nla_addr        = attrs[IPVS_DEST_ATTR_ADDR];
3229         nla_port        = attrs[IPVS_DEST_ATTR_PORT];
3230         nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3231
3232         if (!(nla_addr && nla_port))
3233                 return -EINVAL;
3234
3235         memset(udest, 0, sizeof(*udest));
3236
3237         nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3238         udest->port = nla_get_be16(nla_port);
3239
3240         if (nla_addr_family)
3241                 udest->af = nla_get_u16(nla_addr_family);
3242         else
3243                 udest->af = 0;
3244
3245         /* If a full entry was requested, check for the additional fields */
3246         if (full_entry) {
3247                 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3248                               *nla_l_thresh;
3249
3250                 nla_fwd         = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3251                 nla_weight      = attrs[IPVS_DEST_ATTR_WEIGHT];
3252                 nla_u_thresh    = attrs[IPVS_DEST_ATTR_U_THRESH];
3253                 nla_l_thresh    = attrs[IPVS_DEST_ATTR_L_THRESH];
3254
3255                 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3256                         return -EINVAL;
3257
3258                 udest->conn_flags = nla_get_u32(nla_fwd)
3259                                     & IP_VS_CONN_F_FWD_MASK;
3260                 udest->weight = nla_get_u32(nla_weight);
3261                 udest->u_threshold = nla_get_u32(nla_u_thresh);
3262                 udest->l_threshold = nla_get_u32(nla_l_thresh);
3263         }
3264
3265         return 0;
3266 }
3267
3268 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3269                                   const char *mcast_ifn, __u32 syncid)
3270 {
3271         struct nlattr *nl_daemon;
3272
3273         nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3274         if (!nl_daemon)
3275                 return -EMSGSIZE;
3276
3277         if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3278             nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn) ||
3279             nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid))
3280                 goto nla_put_failure;
3281         nla_nest_end(skb, nl_daemon);
3282
3283         return 0;
3284
3285 nla_put_failure:
3286         nla_nest_cancel(skb, nl_daemon);
3287         return -EMSGSIZE;
3288 }
3289
3290 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3291                                   const char *mcast_ifn, __u32 syncid,
3292                                   struct netlink_callback *cb)
3293 {
3294         void *hdr;
3295         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3296                           &ip_vs_genl_family, NLM_F_MULTI,
3297                           IPVS_CMD_NEW_DAEMON);
3298         if (!hdr)
3299                 return -EMSGSIZE;
3300
3301         if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3302                 goto nla_put_failure;
3303
3304         genlmsg_end(skb, hdr);
3305         return 0;
3306
3307 nla_put_failure:
3308         genlmsg_cancel(skb, hdr);
3309         return -EMSGSIZE;
3310 }
3311
3312 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3313                                    struct netlink_callback *cb)
3314 {
3315         struct net *net = skb_sknet(skb);
3316         struct netns_ipvs *ipvs = net_ipvs(net);
3317
3318         mutex_lock(&ipvs->sync_mutex);
3319         if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3320                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3321                                            ipvs->master_mcast_ifn,
3322                                            ipvs->master_syncid, cb) < 0)
3323                         goto nla_put_failure;
3324
3325                 cb->args[0] = 1;
3326         }
3327
3328         if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3329                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3330                                            ipvs->backup_mcast_ifn,
3331                                            ipvs->backup_syncid, cb) < 0)
3332                         goto nla_put_failure;
3333
3334                 cb->args[1] = 1;
3335         }
3336
3337 nla_put_failure:
3338         mutex_unlock(&ipvs->sync_mutex);
3339
3340         return skb->len;
3341 }
3342
3343 static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
3344 {
3345         if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3346               attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3347               attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3348                 return -EINVAL;
3349
3350         /* The synchronization protocol is incompatible with mixed family
3351          * services
3352          */
3353         if (net_ipvs(net)->mixed_address_family_dests > 0)
3354                 return -EINVAL;
3355
3356         return start_sync_thread(net,
3357                                  nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3358                                  nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3359                                  nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3360 }
3361
3362 static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
3363 {
3364         if (!attrs[IPVS_DAEMON_ATTR_STATE])
3365                 return -EINVAL;
3366
3367         return stop_sync_thread(net,
3368                                 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3369 }
3370
3371 static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
3372 {
3373         struct ip_vs_timeout_user t;
3374
3375         __ip_vs_get_timeouts(net, &t);
3376
3377         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3378                 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3379
3380         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3381                 t.tcp_fin_timeout =
3382                         nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3383
3384         if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3385                 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3386
3387         return ip_vs_set_timeout(net, &t);
3388 }
3389
3390 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3391 {
3392         int ret = 0, cmd;
3393         struct net *net;
3394         struct netns_ipvs *ipvs;
3395
3396         net = skb_sknet(skb);
3397         ipvs = net_ipvs(net);
3398         cmd = info->genlhdr->cmd;
3399
3400         if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3401                 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3402
3403                 mutex_lock(&ipvs->sync_mutex);
3404                 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3405                     nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3406                                      info->attrs[IPVS_CMD_ATTR_DAEMON],
3407                                      ip_vs_daemon_policy)) {
3408                         ret = -EINVAL;
3409                         goto out;
3410                 }
3411
3412                 if (cmd == IPVS_CMD_NEW_DAEMON)
3413                         ret = ip_vs_genl_new_daemon(net, daemon_attrs);
3414                 else
3415                         ret = ip_vs_genl_del_daemon(net, daemon_attrs);
3416 out:
3417                 mutex_unlock(&ipvs->sync_mutex);
3418         }
3419         return ret;
3420 }
3421
3422 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3423 {
3424         struct ip_vs_service *svc = NULL;
3425         struct ip_vs_service_user_kern usvc;
3426         struct ip_vs_dest_user_kern udest;
3427         int ret = 0, cmd;
3428         int need_full_svc = 0, need_full_dest = 0;
3429         struct net *net;
3430
3431         net = skb_sknet(skb);
3432         cmd = info->genlhdr->cmd;
3433
3434         mutex_lock(&__ip_vs_mutex);
3435
3436         if (cmd == IPVS_CMD_FLUSH) {
3437                 ret = ip_vs_flush(net, false);
3438                 goto out;
3439         } else if (cmd == IPVS_CMD_SET_CONFIG) {
3440                 ret = ip_vs_genl_set_config(net, info->attrs);
3441                 goto out;
3442         } else if (cmd == IPVS_CMD_ZERO &&
3443                    !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3444                 ret = ip_vs_zero_all(net);
3445                 goto out;
3446         }
3447
3448         /* All following commands require a service argument, so check if we
3449          * received a valid one. We need a full service specification when
3450          * adding / editing a service. Only identifying members otherwise. */
3451         if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3452                 need_full_svc = 1;
3453
3454         ret = ip_vs_genl_parse_service(net, &usvc,
3455                                        info->attrs[IPVS_CMD_ATTR_SERVICE],
3456                                        need_full_svc, &svc);
3457         if (ret)
3458                 goto out;
3459
3460         /* Unless we're adding a new service, the service must already exist */
3461         if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3462                 ret = -ESRCH;
3463                 goto out;
3464         }
3465
3466         /* Destination commands require a valid destination argument. For
3467          * adding / editing a destination, we need a full destination
3468          * specification. */
3469         if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3470             cmd == IPVS_CMD_DEL_DEST) {
3471                 if (cmd != IPVS_CMD_DEL_DEST)
3472                         need_full_dest = 1;
3473
3474                 ret = ip_vs_genl_parse_dest(&udest,
3475                                             info->attrs[IPVS_CMD_ATTR_DEST],
3476                                             need_full_dest);
3477                 if (ret)
3478                         goto out;
3479
3480                 /* Old protocols did not allow the user to specify address
3481                  * family, so we set it to zero instead.  We also didn't
3482                  * allow heterogeneous pools in the old code, so it's safe
3483                  * to assume that this will have the same address family as
3484                  * the service.
3485                  */
3486                 if (udest.af == 0)
3487                         udest.af = svc->af;
3488
3489                 if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) {
3490                         /* The synchronization protocol is incompatible
3491                          * with mixed family services
3492                          */
3493                         if (net_ipvs(net)->sync_state) {
3494                                 ret = -EINVAL;
3495                                 goto out;
3496                         }
3497
3498                         /* Which connection types do we support? */
3499                         switch (udest.conn_flags) {
3500                         case IP_VS_CONN_F_TUNNEL:
3501                                 /* We are able to forward this */
3502                                 break;
3503                         default:
3504                                 ret = -EINVAL;
3505                                 goto out;
3506                         }
3507                 }
3508         }
3509
3510         switch (cmd) {
3511         case IPVS_CMD_NEW_SERVICE:
3512                 if (svc == NULL)
3513                         ret = ip_vs_add_service(net, &usvc, &svc);
3514                 else
3515                         ret = -EEXIST;
3516                 break;
3517         case IPVS_CMD_SET_SERVICE:
3518                 ret = ip_vs_edit_service(svc, &usvc);
3519                 break;
3520         case IPVS_CMD_DEL_SERVICE:
3521                 ret = ip_vs_del_service(svc);
3522                 /* do not use svc, it can be freed */
3523                 break;
3524         case IPVS_CMD_NEW_DEST:
3525                 ret = ip_vs_add_dest(svc, &udest);
3526                 break;
3527         case IPVS_CMD_SET_DEST:
3528                 ret = ip_vs_edit_dest(svc, &udest);
3529                 break;
3530         case IPVS_CMD_DEL_DEST:
3531                 ret = ip_vs_del_dest(svc, &udest);
3532                 break;
3533         case IPVS_CMD_ZERO:
3534                 ret = ip_vs_zero_service(svc);
3535                 break;
3536         default:
3537                 ret = -EINVAL;
3538         }
3539
3540 out:
3541         mutex_unlock(&__ip_vs_mutex);
3542
3543         return ret;
3544 }
3545
3546 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3547 {
3548         struct sk_buff *msg;
3549         void *reply;
3550         int ret, cmd, reply_cmd;
3551         struct net *net;
3552
3553         net = skb_sknet(skb);
3554         cmd = info->genlhdr->cmd;
3555
3556         if (cmd == IPVS_CMD_GET_SERVICE)
3557                 reply_cmd = IPVS_CMD_NEW_SERVICE;
3558         else if (cmd == IPVS_CMD_GET_INFO)
3559                 reply_cmd = IPVS_CMD_SET_INFO;
3560         else if (cmd == IPVS_CMD_GET_CONFIG)
3561                 reply_cmd = IPVS_CMD_SET_CONFIG;
3562         else {
3563                 pr_err("unknown Generic Netlink command\n");
3564                 return -EINVAL;
3565         }
3566
3567         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3568         if (!msg)
3569                 return -ENOMEM;
3570
3571         mutex_lock(&__ip_vs_mutex);
3572
3573         reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3574         if (reply == NULL)
3575                 goto nla_put_failure;
3576
3577         switch (cmd) {
3578         case IPVS_CMD_GET_SERVICE:
3579         {
3580                 struct ip_vs_service *svc;
3581
3582                 svc = ip_vs_genl_find_service(net,
3583                                               info->attrs[IPVS_CMD_ATTR_SERVICE]);
3584                 if (IS_ERR(svc)) {
3585                         ret = PTR_ERR(svc);
3586                         goto out_err;
3587                 } else if (svc) {
3588                         ret = ip_vs_genl_fill_service(msg, svc);
3589                         if (ret)
3590                                 goto nla_put_failure;
3591                 } else {
3592                         ret = -ESRCH;
3593                         goto out_err;
3594                 }
3595
3596                 break;
3597         }
3598
3599         case IPVS_CMD_GET_CONFIG:
3600         {
3601                 struct ip_vs_timeout_user t;
3602
3603                 __ip_vs_get_timeouts(net, &t);
3604 #ifdef CONFIG_IP_VS_PROTO_TCP
3605                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3606                                 t.tcp_timeout) ||
3607                     nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3608                                 t.tcp_fin_timeout))
3609                         goto nla_put_failure;
3610 #endif
3611 #ifdef CONFIG_IP_VS_PROTO_UDP
3612                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3613                         goto nla_put_failure;
3614 #endif
3615
3616                 break;
3617         }
3618
3619         case IPVS_CMD_GET_INFO:
3620                 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3621                                 IP_VS_VERSION_CODE) ||
3622                     nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3623                                 ip_vs_conn_tab_size))
3624                         goto nla_put_failure;
3625                 break;
3626         }
3627
3628         genlmsg_end(msg, reply);
3629         ret = genlmsg_reply(msg, info);
3630         goto out;
3631
3632 nla_put_failure:
3633         pr_err("not enough space in Netlink message\n");
3634         ret = -EMSGSIZE;
3635
3636 out_err:
3637         nlmsg_free(msg);
3638 out:
3639         mutex_unlock(&__ip_vs_mutex);
3640
3641         return ret;
3642 }
3643
3644
3645 static const struct genl_ops ip_vs_genl_ops[] = {
3646         {
3647                 .cmd    = IPVS_CMD_NEW_SERVICE,
3648                 .flags  = GENL_ADMIN_PERM,
3649                 .policy = ip_vs_cmd_policy,
3650                 .doit   = ip_vs_genl_set_cmd,
3651         },
3652         {
3653                 .cmd    = IPVS_CMD_SET_SERVICE,
3654                 .flags  = GENL_ADMIN_PERM,
3655                 .policy = ip_vs_cmd_policy,
3656                 .doit   = ip_vs_genl_set_cmd,
3657         },
3658         {
3659                 .cmd    = IPVS_CMD_DEL_SERVICE,
3660                 .flags  = GENL_ADMIN_PERM,
3661                 .policy = ip_vs_cmd_policy,
3662                 .doit   = ip_vs_genl_set_cmd,
3663         },
3664         {
3665                 .cmd    = IPVS_CMD_GET_SERVICE,
3666                 .flags  = GENL_ADMIN_PERM,
3667                 .doit   = ip_vs_genl_get_cmd,
3668                 .dumpit = ip_vs_genl_dump_services,
3669                 .policy = ip_vs_cmd_policy,
3670         },
3671         {
3672                 .cmd    = IPVS_CMD_NEW_DEST,
3673                 .flags  = GENL_ADMIN_PERM,
3674                 .policy = ip_vs_cmd_policy,
3675                 .doit   = ip_vs_genl_set_cmd,
3676         },
3677         {
3678                 .cmd    = IPVS_CMD_SET_DEST,
3679                 .flags  = GENL_ADMIN_PERM,
3680                 .policy = ip_vs_cmd_policy,
3681                 .doit   = ip_vs_genl_set_cmd,
3682         },
3683         {
3684                 .cmd    = IPVS_CMD_DEL_DEST,
3685                 .flags  = GENL_ADMIN_PERM,
3686                 .policy = ip_vs_cmd_policy,
3687                 .doit   = ip_vs_genl_set_cmd,
3688         },
3689         {
3690                 .cmd    = IPVS_CMD_GET_DEST,
3691                 .flags  = GENL_ADMIN_PERM,
3692                 .policy = ip_vs_cmd_policy,
3693                 .dumpit = ip_vs_genl_dump_dests,
3694         },
3695         {
3696                 .cmd    = IPVS_CMD_NEW_DAEMON,
3697                 .flags  = GENL_ADMIN_PERM,
3698                 .policy = ip_vs_cmd_policy,
3699                 .doit   = ip_vs_genl_set_daemon,
3700         },
3701         {
3702                 .cmd    = IPVS_CMD_DEL_DAEMON,
3703                 .flags  = GENL_ADMIN_PERM,
3704                 .policy = ip_vs_cmd_policy,
3705                 .doit   = ip_vs_genl_set_daemon,
3706         },
3707         {
3708                 .cmd    = IPVS_CMD_GET_DAEMON,
3709                 .flags  = GENL_ADMIN_PERM,
3710                 .dumpit = ip_vs_genl_dump_daemons,
3711         },
3712         {
3713                 .cmd    = IPVS_CMD_SET_CONFIG,
3714                 .flags  = GENL_ADMIN_PERM,
3715                 .policy = ip_vs_cmd_policy,
3716                 .doit   = ip_vs_genl_set_cmd,
3717         },
3718         {
3719                 .cmd    = IPVS_CMD_GET_CONFIG,
3720                 .flags  = GENL_ADMIN_PERM,
3721                 .doit   = ip_vs_genl_get_cmd,
3722         },
3723         {
3724                 .cmd    = IPVS_CMD_GET_INFO,
3725                 .flags  = GENL_ADMIN_PERM,
3726                 .doit   = ip_vs_genl_get_cmd,
3727         },
3728         {
3729                 .cmd    = IPVS_CMD_ZERO,
3730                 .flags  = GENL_ADMIN_PERM,
3731                 .policy = ip_vs_cmd_policy,
3732                 .doit   = ip_vs_genl_set_cmd,
3733         },
3734         {
3735                 .cmd    = IPVS_CMD_FLUSH,
3736                 .flags  = GENL_ADMIN_PERM,
3737                 .doit   = ip_vs_genl_set_cmd,
3738         },
3739 };
3740
3741 static int __init ip_vs_genl_register(void)
3742 {
3743         return genl_register_family_with_ops(&ip_vs_genl_family,
3744                                              ip_vs_genl_ops);
3745 }
3746
3747 static void ip_vs_genl_unregister(void)
3748 {
3749         genl_unregister_family(&ip_vs_genl_family);
3750 }
3751
3752 /* End of Generic Netlink interface definitions */
3753
3754 /*
3755  * per netns intit/exit func.
3756  */
3757 #ifdef CONFIG_SYSCTL
3758 static int __net_init ip_vs_control_net_init_sysctl(struct net *net)
3759 {
3760         int idx;
3761         struct netns_ipvs *ipvs = net_ipvs(net);
3762         struct ctl_table *tbl;
3763
3764         atomic_set(&ipvs->dropentry, 0);
3765         spin_lock_init(&ipvs->dropentry_lock);
3766         spin_lock_init(&ipvs->droppacket_lock);
3767         spin_lock_init(&ipvs->securetcp_lock);
3768
3769         if (!net_eq(net, &init_net)) {
3770                 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3771                 if (tbl == NULL)
3772                         return -ENOMEM;
3773
3774                 /* Don't export sysctls to unprivileged users */
3775                 if (net->user_ns != &init_user_ns)
3776                         tbl[0].procname = NULL;
3777         } else
3778                 tbl = vs_vars;
3779         /* Initialize sysctl defaults */
3780         idx = 0;
3781         ipvs->sysctl_amemthresh = 1024;
3782         tbl[idx++].data = &ipvs->sysctl_amemthresh;
3783         ipvs->sysctl_am_droprate = 10;
3784         tbl[idx++].data = &ipvs->sysctl_am_droprate;
3785         tbl[idx++].data = &ipvs->sysctl_drop_entry;
3786         tbl[idx++].data = &ipvs->sysctl_drop_packet;
3787 #ifdef CONFIG_IP_VS_NFCT
3788         tbl[idx++].data = &ipvs->sysctl_conntrack;
3789 #endif
3790         tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3791         ipvs->sysctl_snat_reroute = 1;
3792         tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3793         ipvs->sysctl_sync_ver = 1;
3794         tbl[idx++].data = &ipvs->sysctl_sync_ver;
3795         ipvs->sysctl_sync_ports = 1;
3796         tbl[idx++].data = &ipvs->sysctl_sync_ports;
3797         tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3798         ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3799         tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3800         ipvs->sysctl_sync_sock_size = 0;
3801         tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3802         tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3803         tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3804         tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3805         tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3806         tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3807         ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3808         ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3809         tbl[idx].data = &ipvs->sysctl_sync_threshold;
3810         tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3811         ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3812         tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3813         ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3814         tbl[idx++].data = &ipvs->sysctl_sync_retries;
3815         tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3816         ipvs->sysctl_pmtu_disc = 1;
3817         tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3818         tbl[idx++].data = &ipvs->sysctl_backup_only;
3819         ipvs->sysctl_conn_reuse_mode = 1;
3820         tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
3821
3822
3823         ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3824         if (ipvs->sysctl_hdr == NULL) {
3825                 if (!net_eq(net, &init_net))
3826                         kfree(tbl);
3827                 return -ENOMEM;
3828         }
3829         ip_vs_start_estimator(net, &ipvs->tot_stats);
3830         ipvs->sysctl_tbl = tbl;
3831         /* Schedule defense work */
3832         INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3833         schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3834
3835         return 0;
3836 }
3837
3838 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net)
3839 {
3840         struct netns_ipvs *ipvs = net_ipvs(net);
3841
3842         cancel_delayed_work_sync(&ipvs->defense_work);
3843         cancel_work_sync(&ipvs->defense_work.work);
3844         unregister_net_sysctl_table(ipvs->sysctl_hdr);
3845         ip_vs_stop_estimator(net, &ipvs->tot_stats);
3846
3847         if (!net_eq(net, &init_net))
3848                 kfree(ipvs->sysctl_tbl);
3849 }
3850
3851 #else
3852
3853 static int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
3854 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net) { }
3855
3856 #endif
3857
3858 static struct notifier_block ip_vs_dst_notifier = {
3859         .notifier_call = ip_vs_dst_event,
3860 };
3861
3862 int __net_init ip_vs_control_net_init(struct net *net)
3863 {
3864         int i, idx;
3865         struct netns_ipvs *ipvs = net_ipvs(net);
3866
3867         /* Initialize rs_table */
3868         for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3869                 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
3870
3871         INIT_LIST_HEAD(&ipvs->dest_trash);
3872         spin_lock_init(&ipvs->dest_trash_lock);
3873         setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
3874                     (unsigned long) net);
3875         atomic_set(&ipvs->ftpsvc_counter, 0);
3876         atomic_set(&ipvs->nullsvc_counter, 0);
3877
3878         /* procfs stats */
3879         ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3880         if (!ipvs->tot_stats.cpustats)
3881                 return -ENOMEM;
3882
3883         for_each_possible_cpu(i) {
3884                 struct ip_vs_cpu_stats *ipvs_tot_stats;
3885                 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
3886                 u64_stats_init(&ipvs_tot_stats->syncp);
3887         }
3888
3889         spin_lock_init(&ipvs->tot_stats.lock);
3890
3891         proc_create("ip_vs", 0, net->proc_net, &ip_vs_info_fops);
3892         proc_create("ip_vs_stats", 0, net->proc_net, &ip_vs_stats_fops);
3893         proc_create("ip_vs_stats_percpu", 0, net->proc_net,
3894                     &ip_vs_stats_percpu_fops);
3895
3896         if (ip_vs_control_net_init_sysctl(net))
3897                 goto err;
3898
3899         return 0;
3900
3901 err:
3902         free_percpu(ipvs->tot_stats.cpustats);
3903         return -ENOMEM;
3904 }
3905
3906 void __net_exit ip_vs_control_net_cleanup(struct net *net)
3907 {
3908         struct netns_ipvs *ipvs = net_ipvs(net);
3909
3910         ip_vs_trash_cleanup(net);
3911         ip_vs_control_net_cleanup_sysctl(net);
3912         remove_proc_entry("ip_vs_stats_percpu", net->proc_net);
3913         remove_proc_entry("ip_vs_stats", net->proc_net);
3914         remove_proc_entry("ip_vs", net->proc_net);
3915         free_percpu(ipvs->tot_stats.cpustats);
3916 }
3917
3918 int __init ip_vs_register_nl_ioctl(void)
3919 {
3920         int ret;
3921
3922         ret = nf_register_sockopt(&ip_vs_sockopts);
3923         if (ret) {
3924                 pr_err("cannot register sockopt.\n");
3925                 goto err_sock;
3926         }
3927
3928         ret = ip_vs_genl_register();
3929         if (ret) {
3930                 pr_err("cannot register Generic Netlink interface.\n");
3931                 goto err_genl;
3932         }
3933         return 0;
3934
3935 err_genl:
3936         nf_unregister_sockopt(&ip_vs_sockopts);
3937 err_sock:
3938         return ret;
3939 }
3940
3941 void ip_vs_unregister_nl_ioctl(void)
3942 {
3943         ip_vs_genl_unregister();
3944         nf_unregister_sockopt(&ip_vs_sockopts);
3945 }
3946
3947 int __init ip_vs_control_init(void)
3948 {
3949         int idx;
3950         int ret;
3951
3952         EnterFunction(2);
3953
3954         /* Initialize svc_table, ip_vs_svc_fwm_table */
3955         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3956                 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
3957                 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3958         }
3959
3960         smp_wmb();      /* Do we really need it now ? */
3961
3962         ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3963         if (ret < 0)
3964                 return ret;
3965
3966         LeaveFunction(2);
3967         return 0;
3968 }
3969
3970
3971 void ip_vs_control_cleanup(void)
3972 {
3973         EnterFunction(2);
3974         unregister_netdevice_notifier(&ip_vs_dst_notifier);
3975         LeaveFunction(2);
3976 }