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);
 232         schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
 233 }
 234 #endif
 235
 236 int
 237 ip_vs_use_count_inc(void)
 238 {
 239         return try_module_get(THIS_MODULE);
 240 }
 241
 242 void
 243 ip_vs_use_count_dec(void)
 244 {
 245         module_put(THIS_MODULE);
 246 }
 247
 248
 249 /*
 250  *      Hash table: for virtual service lookups
 251  */
 252 #define IP_VS_SVC_TAB_BITS 8
 253 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
 254 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
 255
 256 /* the service table hashed by <protocol, addr, port> */
 257 static struct hlist_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
 258 /* the service table hashed by fwmark */
 259 static struct hlist_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
 260
 261
 262 /*
 263  *      Returns hash value for virtual service
 264  */
 265 static inline unsigned int
 266 ip_vs_svc_hashkey(struct netns_ipvs *ipvs, int af, unsigned int proto,
 267                   const union nf_inet_addr *addr, __be16 port)
 268 {
 269         register unsigned int porth = ntohs(port);
 270         __be32 addr_fold = addr->ip;
 271         __u32 ahash;
 272
 273 #ifdef CONFIG_IP_VS_IPV6
 274         if (af == AF_INET6)
 275                 addr_fold = addr->ip6[0]^addr->ip6[1]^
 276                             addr->ip6[2]^addr->ip6[3];
 277 #endif
 278         ahash = ntohl(addr_fold);
 279         ahash ^= ((size_t) ipvs >> 8);
 280
 281         return (proto ^ ahash ^ (porth >> IP_VS_SVC_TAB_BITS) ^ porth) &
 282                IP_VS_SVC_TAB_MASK;
 283 }
 284
 285 /*
 286  *      Returns hash value of fwmark for virtual service lookup
 287  */
 288 static inline unsigned int ip_vs_svc_fwm_hashkey(struct netns_ipvs *ipvs, __u32 fwmark)
 289 {
 290         return (((size_t)ipvs>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK;
 291 }
 292
 293 /*
 294  *      Hashes a service in the ip_vs_svc_table by <netns,proto,addr,port>
 295  *      or in the ip_vs_svc_fwm_table by fwmark.
 296  *      Should be called with locked tables.
 297  */
 298 static int ip_vs_svc_hash(struct ip_vs_service *svc)
 299 {
 300         unsigned int hash;
 301
 302         if (svc->flags & IP_VS_SVC_F_HASHED) {
 303                 pr_err("%s(): request for already hashed, called from %pF\n",
 304                        __func__, __builtin_return_address(0));
 305                 return 0;
 306         }
 307
 308         if (svc->fwmark == 0) {
 309                 /*
 310                  *  Hash it by <netns,protocol,addr,port> in ip_vs_svc_table
 311                  */
 312                 hash = ip_vs_svc_hashkey(svc->ipvs, svc->af, svc->protocol,
 313                                          &svc->addr, svc->port);
 314                 hlist_add_head_rcu(&svc->s_list, &ip_vs_svc_table[hash]);
 315         } else {
 316                 /*
 317                  *  Hash it by fwmark in svc_fwm_table
 318                  */
 319                 hash = ip_vs_svc_fwm_hashkey(svc->ipvs, svc->fwmark);
 320                 hlist_add_head_rcu(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
 321         }
 322
 323         svc->flags |= IP_VS_SVC_F_HASHED;
 324         /* increase its refcnt because it is referenced by the svc table */
 325         atomic_inc(&svc->refcnt);
 326         return 1;
 327 }
 328
 329
 330 /*
 331  *      Unhashes a service from svc_table / svc_fwm_table.
 332  *      Should be called with locked tables.
 333  */
 334 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
 335 {
 336         if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
 337                 pr_err("%s(): request for unhash flagged, called from %pF\n",
 338                        __func__, __builtin_return_address(0));
 339                 return 0;
 340         }
 341
 342         if (svc->fwmark == 0) {
 343                 /* Remove it from the svc_table table */
 344                 hlist_del_rcu(&svc->s_list);
 345         } else {
 346                 /* Remove it from the svc_fwm_table table */
 347                 hlist_del_rcu(&svc->f_list);
 348         }
 349
 350         svc->flags &= ~IP_VS_SVC_F_HASHED;
 351         atomic_dec(&svc->refcnt);
 352         return 1;
 353 }
 354
 355
 356 /*
 357  *      Get service by {netns, proto,addr,port} in the service table.
 358  */
 359 static inline struct ip_vs_service *
 360 __ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u16 protocol,
 361                      const union nf_inet_addr *vaddr, __be16 vport)
 362 {
 363         unsigned int hash;
 364         struct ip_vs_service *svc;
 365
 366         /* Check for "full" addressed entries */
 367         hash = ip_vs_svc_hashkey(ipvs, af, protocol, vaddr, vport);
 368
 369         hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[hash], s_list) {
 370                 if ((svc->af == af)
 371                     && ip_vs_addr_equal(af, &svc->addr, vaddr)
 372                     && (svc->port == vport)
 373                     && (svc->protocol == protocol)
 374                     && (svc->ipvs == ipvs)) {
 375                         /* HIT */
 376                         return svc;
 377                 }
 378         }
 379
 380         return NULL;
 381 }
 382
 383
 384 /*
 385  *      Get service by {fwmark} in the service table.
 386  */
 387 static inline struct ip_vs_service *
 388 __ip_vs_svc_fwm_find(struct netns_ipvs *ipvs, int af, __u32 fwmark)
 389 {
 390         unsigned int hash;
 391         struct ip_vs_service *svc;
 392
 393         /* Check for fwmark addressed entries */
 394         hash = ip_vs_svc_fwm_hashkey(ipvs, fwmark);
 395
 396         hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[hash], f_list) {
 397                 if (svc->fwmark == fwmark && svc->af == af
 398                     && (svc->ipvs == ipvs)) {
 399                         /* HIT */
 400                         return svc;
 401                 }
 402         }
 403
 404         return NULL;
 405 }
 406
 407 /* Find service, called under RCU lock */
 408 struct ip_vs_service *
 409 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
 410                    const union nf_inet_addr *vaddr, __be16 vport)
 411 {
 412         struct ip_vs_service *svc;
 413
 414         /*
 415          *      Check the table hashed by fwmark first
 416          */
 417         if (fwmark) {
 418                 svc = __ip_vs_svc_fwm_find(ipvs, af, fwmark);
 419                 if (svc)
 420                         goto out;
 421         }
 422
 423         /*
 424          *      Check the table hashed by <protocol,addr,port>
 425          *      for "full" addressed entries
 426          */
 427         svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, vport);
 428
 429         if (svc == NULL
 430             && protocol == IPPROTO_TCP
 431             && atomic_read(&ipvs->ftpsvc_counter)
 432             && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
 433                 /*
 434                  * Check if ftp service entry exists, the packet
 435                  * might belong to FTP data connections.
 436                  */
 437                 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, FTPPORT);
 438         }
 439
 440         if (svc == NULL
 441             && atomic_read(&ipvs->nullsvc_counter)) {
 442                 /*
 443                  * Check if the catch-all port (port zero) exists
 444                  */
 445                 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, 0);
 446         }
 447
 448   out:
 449         IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
 450                       fwmark, ip_vs_proto_name(protocol),
 451                       IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
 452                       svc ? "hit" : "not hit");
 453
 454         return svc;
 455 }
 456
 457
 458 static inline void
 459 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
 460 {
 461         atomic_inc(&svc->refcnt);
 462         rcu_assign_pointer(dest->svc, svc);
 463 }
 464
 465 static void ip_vs_service_free(struct ip_vs_service *svc)
 466 {
 467         free_percpu(svc->stats.cpustats);
 468         kfree(svc);
 469 }
 470
 471 static void ip_vs_service_rcu_free(struct rcu_head *head)
 472 {
 473         struct ip_vs_service *svc;
 474
 475         svc = container_of(head, struct ip_vs_service, rcu_head);
 476         ip_vs_service_free(svc);
 477 }
 478
 479 static void __ip_vs_svc_put(struct ip_vs_service *svc, bool do_delay)
 480 {
 481         if (atomic_dec_and_test(&svc->refcnt)) {
 482                 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
 483                               svc->fwmark,
 484                               IP_VS_DBG_ADDR(svc->af, &svc->addr),
 485                               ntohs(svc->port));
 486                 if (do_delay)
 487                         call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
 488                 else
 489                         ip_vs_service_free(svc);
 490         }
 491 }
 492
 493
 494 /*
 495  *      Returns hash value for real service
 496  */
 497 static inline unsigned int ip_vs_rs_hashkey(int af,
 498                                             const union nf_inet_addr *addr,
 499                                             __be16 port)
 500 {
 501         register unsigned int porth = ntohs(port);
 502         __be32 addr_fold = addr->ip;
 503
 504 #ifdef CONFIG_IP_VS_IPV6
 505         if (af == AF_INET6)
 506                 addr_fold = addr->ip6[0]^addr->ip6[1]^
 507                             addr->ip6[2]^addr->ip6[3];
 508 #endif
 509
 510         return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
 511                 & IP_VS_RTAB_MASK;
 512 }
 513
 514 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
 515 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
 516 {
 517         unsigned int hash;
 518
 519         if (dest->in_rs_table)
 520                 return;
 521
 522         /*
 523          *      Hash by proto,addr,port,
 524          *      which are the parameters of the real service.
 525          */
 526         hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
 527
 528         hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
 529         dest->in_rs_table = 1;
 530 }
 531
 532 /* Unhash ip_vs_dest from rs_table. */
 533 static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
 534 {
 535         /*
 536          * Remove it from the rs_table table.
 537          */
 538         if (dest->in_rs_table) {
 539                 hlist_del_rcu(&dest->d_list);
 540                 dest->in_rs_table = 0;
 541         }
 542 }
 543
 544 /* Check if real service by <proto,addr,port> is present */
 545 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
 546                             const union nf_inet_addr *daddr, __be16 dport)
 547 {
 548         unsigned int hash;
 549         struct ip_vs_dest *dest;
 550
 551         /* Check for "full" addressed entries */
 552         hash = ip_vs_rs_hashkey(af, daddr, dport);
 553
 554         rcu_read_lock();
 555         hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
 556                 if (dest->port == dport &&
 557                     dest->af == af &&
 558                     ip_vs_addr_equal(af, &dest->addr, daddr) &&
 559                     (dest->protocol == protocol || dest->vfwmark)) {
 560                         /* HIT */
 561                         rcu_read_unlock();
 562                         return true;
 563                 }
 564         }
 565         rcu_read_unlock();
 566
 567         return false;
 568 }
 569
 570 /* Lookup destination by {addr,port} in the given service
 571  * Called under RCU lock.
 572  */
 573 static struct ip_vs_dest *
 574 ip_vs_lookup_dest(struct ip_vs_service *svc, int dest_af,
 575                   const union nf_inet_addr *daddr, __be16 dport)
 576 {
 577         struct ip_vs_dest *dest;
 578
 579         /*
 580          * Find the destination for the given service
 581          */
 582         list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
 583                 if ((dest->af == dest_af) &&
 584                     ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
 585                     (dest->port == dport)) {
 586                         /* HIT */
 587                         return dest;
 588                 }
 589         }
 590
 591         return NULL;
 592 }
 593
 594 /*
 595  * Find destination by {daddr,dport,vaddr,protocol}
 596  * Created to be used in ip_vs_process_message() in
 597  * the backup synchronization daemon. It finds the
 598  * destination to be bound to the received connection
 599  * on the backup.
 600  * Called under RCU lock, no refcnt is returned.
 601  */
 602 struct ip_vs_dest *ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
 603                                    const union nf_inet_addr *daddr,
 604                                    __be16 dport,
 605                                    const union nf_inet_addr *vaddr,
 606                                    __be16 vport, __u16 protocol, __u32 fwmark,
 607                                    __u32 flags)
 608 {
 609         struct ip_vs_dest *dest;
 610         struct ip_vs_service *svc;
 611         __be16 port = dport;
 612
 613         svc = ip_vs_service_find(ipvs, svc_af, fwmark, protocol, vaddr, vport);
 614         if (!svc)
 615                 return NULL;
 616         if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
 617                 port = 0;
 618         dest = ip_vs_lookup_dest(svc, dest_af, daddr, port);
 619         if (!dest)
 620                 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port ^ dport);
 621         return dest;
 622 }
 623
 624 void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
 625 {
 626         struct ip_vs_dest_dst *dest_dst = container_of(head,
 627                                                        struct ip_vs_dest_dst,
 628                                                        rcu_head);
 629
 630         dst_release(dest_dst->dst_cache);
 631         kfree(dest_dst);
 632 }
 633
 634 /* Release dest_dst and dst_cache for dest in user context */
 635 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
 636 {
 637         struct ip_vs_dest_dst *old;
 638
 639         old = rcu_dereference_protected(dest->dest_dst, 1);
 640         if (old) {
 641                 RCU_INIT_POINTER(dest->dest_dst, NULL);
 642                 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
 643         }
 644 }
 645
 646 /*
 647  *  Lookup dest by {svc,addr,port} in the destination trash.
 648  *  The destination trash is used to hold the destinations that are removed
 649  *  from the service table but are still referenced by some conn entries.
 650  *  The reason to add the destination trash is when the dest is temporary
 651  *  down (either by administrator or by monitor program), the dest can be
 652  *  picked back from the trash, the remaining connections to the dest can
 653  *  continue, and the counting information of the dest is also useful for
 654  *  scheduling.
 655  */
 656 static struct ip_vs_dest *
 657 ip_vs_trash_get_dest(struct ip_vs_service *svc, int dest_af,
 658                      const union nf_inet_addr *daddr, __be16 dport)
 659 {
 660         struct ip_vs_dest *dest;
 661         struct netns_ipvs *ipvs = svc->ipvs;
 662
 663         /*
 664          * Find the destination in trash
 665          */
 666         spin_lock_bh(&ipvs->dest_trash_lock);
 667         list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
 668                 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
 669                               "dest->refcnt=%d\n",
 670                               dest->vfwmark,
 671                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
 672                               ntohs(dest->port),
 673                               atomic_read(&dest->refcnt));
 674                 if (dest->af == dest_af &&
 675                     ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
 676                     dest->port == dport &&
 677                     dest->vfwmark == svc->fwmark &&
 678                     dest->protocol == svc->protocol &&
 679                     (svc->fwmark ||
 680                      (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
 681                       dest->vport == svc->port))) {
 682                         /* HIT */
 683                         list_del(&dest->t_list);
 684                         ip_vs_dest_hold(dest);
 685                         goto out;
 686                 }
 687         }
 688
 689         dest = NULL;
 690
 691 out:
 692         spin_unlock_bh(&ipvs->dest_trash_lock);
 693
 694         return dest;
 695 }
 696
 697 static void ip_vs_dest_free(struct ip_vs_dest *dest)
 698 {
 699         struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1);
 700
 701         __ip_vs_dst_cache_reset(dest);
 702         __ip_vs_svc_put(svc, false);
 703         free_percpu(dest->stats.cpustats);
 704         ip_vs_dest_put_and_free(dest);
 705 }
 706
 707 /*
 708  *  Clean up all the destinations in the trash
 709  *  Called by the ip_vs_control_cleanup()
 710  *
 711  *  When the ip_vs_control_clearup is activated by ipvs module exit,
 712  *  the service tables must have been flushed and all the connections
 713  *  are expired, and the refcnt of each destination in the trash must
 714  *  be 0, so we simply release them here.
 715  */
 716 static void ip_vs_trash_cleanup(struct netns_ipvs *ipvs)
 717 {
 718         struct ip_vs_dest *dest, *nxt;
 719
 720         del_timer_sync(&ipvs->dest_trash_timer);
 721         /* No need to use dest_trash_lock */
 722         list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
 723                 list_del(&dest->t_list);
 724                 ip_vs_dest_free(dest);
 725         }
 726 }
 727
 728 static void
 729 ip_vs_copy_stats(struct ip_vs_kstats *dst, struct ip_vs_stats *src)
 730 {
 731 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c
 732
 733         spin_lock_bh(&src->lock);
 734
 735         IP_VS_SHOW_STATS_COUNTER(conns);
 736         IP_VS_SHOW_STATS_COUNTER(inpkts);
 737         IP_VS_SHOW_STATS_COUNTER(outpkts);
 738         IP_VS_SHOW_STATS_COUNTER(inbytes);
 739         IP_VS_SHOW_STATS_COUNTER(outbytes);
 740
 741         ip_vs_read_estimator(dst, src);
 742
 743         spin_unlock_bh(&src->lock);
 744 }
 745
 746 static void
 747 ip_vs_export_stats_user(struct ip_vs_stats_user *dst, struct ip_vs_kstats *src)
 748 {
 749         dst->conns = (u32)src->conns;
 750         dst->inpkts = (u32)src->inpkts;
 751         dst->outpkts = (u32)src->outpkts;
 752         dst->inbytes = src->inbytes;
 753         dst->outbytes = src->outbytes;
 754         dst->cps = (u32)src->cps;
 755         dst->inpps = (u32)src->inpps;
 756         dst->outpps = (u32)src->outpps;
 757         dst->inbps = (u32)src->inbps;
 758         dst->outbps = (u32)src->outbps;
 759 }
 760
 761 static void
 762 ip_vs_zero_stats(struct ip_vs_stats *stats)
 763 {
 764         spin_lock_bh(&stats->lock);
 765
 766         /* get current counters as zero point, rates are zeroed */
 767
 768 #define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c
 769
 770         IP_VS_ZERO_STATS_COUNTER(conns);
 771         IP_VS_ZERO_STATS_COUNTER(inpkts);
 772         IP_VS_ZERO_STATS_COUNTER(outpkts);
 773         IP_VS_ZERO_STATS_COUNTER(inbytes);
 774         IP_VS_ZERO_STATS_COUNTER(outbytes);
 775
 776         ip_vs_zero_estimator(stats);
 777
 778         spin_unlock_bh(&stats->lock);
 779 }
 780
 781 /*
 782  *      Update a destination in the given service
 783  */
 784 static void
 785 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
 786                     struct ip_vs_dest_user_kern *udest, int add)
 787 {
 788         struct netns_ipvs *ipvs = svc->ipvs;
 789         struct ip_vs_service *old_svc;
 790         struct ip_vs_scheduler *sched;
 791         int conn_flags;
 792
 793         /* We cannot modify an address and change the address family */
 794         BUG_ON(!add && udest->af != dest->af);
 795
 796         if (add && udest->af != svc->af)
 797                 ipvs->mixed_address_family_dests++;
 798
 799         /* set the weight and the flags */
 800         atomic_set(&dest->weight, udest->weight);
 801         conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
 802         conn_flags |= IP_VS_CONN_F_INACTIVE;
 803
 804         /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
 805         if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
 806                 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
 807         } else {
 808                 /*
 809                  *    Put the real service in rs_table if not present.
 810                  *    For now only for NAT!
 811                  */
 812                 ip_vs_rs_hash(ipvs, dest);
 813         }
 814         atomic_set(&dest->conn_flags, conn_flags);
 815
 816         /* bind the service */
 817         old_svc = rcu_dereference_protected(dest->svc, 1);
 818         if (!old_svc) {
 819                 __ip_vs_bind_svc(dest, svc);
 820         } else {
 821                 if (old_svc != svc) {
 822                         ip_vs_zero_stats(&dest->stats);
 823                         __ip_vs_bind_svc(dest, svc);
 824                         __ip_vs_svc_put(old_svc, true);
 825                 }
 826         }
 827
 828         /* set the dest status flags */
 829         dest->flags |= IP_VS_DEST_F_AVAILABLE;
 830
 831         if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
 832                 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
 833         dest->u_threshold = udest->u_threshold;
 834         dest->l_threshold = udest->l_threshold;
 835
 836         dest->af = udest->af;
 837
 838         spin_lock_bh(&dest->dst_lock);
 839         __ip_vs_dst_cache_reset(dest);
 840         spin_unlock_bh(&dest->dst_lock);
 841
 842         if (add) {
 843                 ip_vs_start_estimator(svc->ipvs, &dest->stats);
 844                 list_add_rcu(&dest->n_list, &svc->destinations);
 845                 svc->num_dests++;
 846                 sched = rcu_dereference_protected(svc->scheduler, 1);
 847                 if (sched && sched->add_dest)
 848                         sched->add_dest(svc, dest);
 849         } else {
 850                 sched = rcu_dereference_protected(svc->scheduler, 1);
 851                 if (sched && sched->upd_dest)
 852                         sched->upd_dest(svc, dest);
 853         }
 854 }
 855
 856
 857 /*
 858  *      Create a destination for the given service
 859  */
 860 static int
 861 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
 862                struct ip_vs_dest **dest_p)
 863 {
 864         struct ip_vs_dest *dest;
 865         unsigned int atype, i;
 866
 867         EnterFunction(2);
 868
 869 #ifdef CONFIG_IP_VS_IPV6
 870         if (udest->af == AF_INET6) {
 871                 atype = ipv6_addr_type(&udest->addr.in6);
 872                 if ((!(atype & IPV6_ADDR_UNICAST) ||
 873                         atype & IPV6_ADDR_LINKLOCAL) &&
 874                         !__ip_vs_addr_is_local_v6(svc->ipvs->net, &udest->addr.in6))
 875                         return -EINVAL;
 876         } else
 877 #endif
 878         {
 879                 atype = inet_addr_type(svc->ipvs->net, udest->addr.ip);
 880                 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
 881                         return -EINVAL;
 882         }
 883
 884         dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
 885         if (dest == NULL)
 886                 return -ENOMEM;
 887
 888         dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
 889         if (!dest->stats.cpustats)
 890                 goto err_alloc;
 891
 892         for_each_possible_cpu(i) {
 893                 struct ip_vs_cpu_stats *ip_vs_dest_stats;
 894                 ip_vs_dest_stats = per_cpu_ptr(dest->stats.cpustats, i);
 895                 u64_stats_init(&ip_vs_dest_stats->syncp);
 896         }
 897
 898         dest->af = udest->af;
 899         dest->protocol = svc->protocol;
 900         dest->vaddr = svc->addr;
 901         dest->vport = svc->port;
 902         dest->vfwmark = svc->fwmark;
 903         ip_vs_addr_copy(udest->af, &dest->addr, &udest->addr);
 904         dest->port = udest->port;
 905
 906         atomic_set(&dest->activeconns, 0);
 907         atomic_set(&dest->inactconns, 0);
 908         atomic_set(&dest->persistconns, 0);
 909         atomic_set(&dest->refcnt, 1);
 910
 911         INIT_HLIST_NODE(&dest->d_list);
 912         spin_lock_init(&dest->dst_lock);
 913         spin_lock_init(&dest->stats.lock);
 914         __ip_vs_update_dest(svc, dest, udest, 1);
 915
 916         *dest_p = dest;
 917
 918         LeaveFunction(2);
 919         return 0;
 920
 921 err_alloc:
 922         kfree(dest);
 923         return -ENOMEM;
 924 }
 925
 926
 927 /*
 928  *      Add a destination into an existing service
 929  */
 930 static int
 931 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
 932 {
 933         struct ip_vs_dest *dest;
 934         union nf_inet_addr daddr;
 935         __be16 dport = udest->port;
 936         int ret;
 937
 938         EnterFunction(2);
 939
 940         if (udest->weight < 0) {
 941                 pr_err("%s(): server weight less than zero\n", __func__);
 942                 return -ERANGE;
 943         }
 944
 945         if (udest->l_threshold > udest->u_threshold) {
 946                 pr_err("%s(): lower threshold is higher than upper threshold\n",
 947                         __func__);
 948                 return -ERANGE;
 949         }
 950
 951         ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
 952
 953         /* We use function that requires RCU lock */
 954         rcu_read_lock();
 955         dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
 956         rcu_read_unlock();
 957
 958         if (dest != NULL) {
 959                 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
 960                 return -EEXIST;
 961         }
 962
 963         /*
 964          * Check if the dest already exists in the trash and
 965          * is from the same service
 966          */
 967         dest = ip_vs_trash_get_dest(svc, udest->af, &daddr, dport);
 968
 969         if (dest != NULL) {
 970                 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
 971                               "dest->refcnt=%d, service %u/%s:%u\n",
 972                               IP_VS_DBG_ADDR(udest->af, &daddr), ntohs(dport),
 973                               atomic_read(&dest->refcnt),
 974                               dest->vfwmark,
 975                               IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
 976                               ntohs(dest->vport));
 977
 978                 __ip_vs_update_dest(svc, dest, udest, 1);
 979                 ret = 0;
 980         } else {
 981                 /*
 982                  * Allocate and initialize the dest structure
 983                  */
 984                 ret = ip_vs_new_dest(svc, udest, &dest);
 985         }
 986         LeaveFunction(2);
 987
 988         return ret;
 989 }
 990
 991
 992 /*
 993  *      Edit a destination in the given service
 994  */
 995 static int
 996 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
 997 {
 998         struct ip_vs_dest *dest;
 999         union nf_inet_addr daddr;
1000         __be16 dport = udest->port;
1001
1002         EnterFunction(2);
1003
1004         if (udest->weight < 0) {
1005                 pr_err("%s(): server weight less than zero\n", __func__);
1006                 return -ERANGE;
1007         }
1008
1009         if (udest->l_threshold > udest->u_threshold) {
1010                 pr_err("%s(): lower threshold is higher than upper threshold\n",
1011                         __func__);
1012                 return -ERANGE;
1013         }
1014
1015         ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1016
1017         /* We use function that requires RCU lock */
1018         rcu_read_lock();
1019         dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1020         rcu_read_unlock();
1021
1022         if (dest == NULL) {
1023                 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1024                 return -ENOENT;
1025         }
1026
1027         __ip_vs_update_dest(svc, dest, udest, 0);
1028         LeaveFunction(2);
1029
1030         return 0;
1031 }
1032
1033 /*
1034  *      Delete a destination (must be already unlinked from the service)
1035  */
1036 static void __ip_vs_del_dest(struct netns_ipvs *ipvs, struct ip_vs_dest *dest,
1037                              bool cleanup)
1038 {
1039         ip_vs_stop_estimator(ipvs, &dest->stats);
1040
1041         /*
1042          *  Remove it from the d-linked list with the real services.
1043          */
1044         ip_vs_rs_unhash(dest);
1045
1046         spin_lock_bh(&ipvs->dest_trash_lock);
1047         IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1048                       IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1049                       atomic_read(&dest->refcnt));
1050         if (list_empty(&ipvs->dest_trash) && !cleanup)
1051                 mod_timer(&ipvs->dest_trash_timer,
1052                           jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1053         /* dest lives in trash without reference */
1054         list_add(&dest->t_list, &ipvs->dest_trash);
1055         dest->idle_start = 0;
1056         spin_unlock_bh(&ipvs->dest_trash_lock);
1057         ip_vs_dest_put(dest);
1058 }
1059
1060
1061 /*
1062  *      Unlink a destination from the given service
1063  */
1064 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1065                                 struct ip_vs_dest *dest,
1066                                 int svcupd)
1067 {
1068         dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1069
1070         /*
1071          *  Remove it from the d-linked destination list.
1072          */
1073         list_del_rcu(&dest->n_list);
1074         svc->num_dests--;
1075
1076         if (dest->af != svc->af)
1077                 svc->ipvs->mixed_address_family_dests--;
1078
1079         if (svcupd) {
1080                 struct ip_vs_scheduler *sched;
1081
1082                 sched = rcu_dereference_protected(svc->scheduler, 1);
1083                 if (sched && sched->del_dest)
1084                         sched->del_dest(svc, dest);
1085         }
1086 }
1087
1088
1089 /*
1090  *      Delete a destination server in the given service
1091  */
1092 static int
1093 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1094 {
1095         struct ip_vs_dest *dest;
1096         __be16 dport = udest->port;
1097
1098         EnterFunction(2);
1099
1100         /* We use function that requires RCU lock */
1101         rcu_read_lock();
1102         dest = ip_vs_lookup_dest(svc, udest->af, &udest->addr, dport);
1103         rcu_read_unlock();
1104
1105         if (dest == NULL) {
1106                 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1107                 return -ENOENT;
1108         }
1109
1110         /*
1111          *      Unlink dest from the service
1112          */
1113         __ip_vs_unlink_dest(svc, dest, 1);
1114
1115         /*
1116          *      Delete the destination
1117          */
1118         __ip_vs_del_dest(svc->ipvs, dest, false);
1119
1120         LeaveFunction(2);
1121
1122         return 0;
1123 }
1124
1125 static void ip_vs_dest_trash_expire(unsigned long data)
1126 {
1127         struct netns_ipvs *ipvs = (struct netns_ipvs *)data;
1128         struct ip_vs_dest *dest, *next;
1129         unsigned long now = jiffies;
1130
1131         spin_lock(&ipvs->dest_trash_lock);
1132         list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1133                 if (atomic_read(&dest->refcnt) > 0)
1134                         continue;
1135                 if (dest->idle_start) {
1136                         if (time_before(now, dest->idle_start +
1137                                              IP_VS_DEST_TRASH_PERIOD))
1138                                 continue;
1139                 } else {
1140                         dest->idle_start = max(1UL, now);
1141                         continue;
1142                 }
1143                 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1144                               dest->vfwmark,
1145                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1146                               ntohs(dest->port));
1147                 list_del(&dest->t_list);
1148                 ip_vs_dest_free(dest);
1149         }
1150         if (!list_empty(&ipvs->dest_trash))
1151                 mod_timer(&ipvs->dest_trash_timer,
1152                           jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1153         spin_unlock(&ipvs->dest_trash_lock);
1154 }
1155
1156 /*
1157  *      Add a service into the service hash table
1158  */
1159 static int
1160 ip_vs_add_service(struct netns_ipvs *ipvs, struct ip_vs_service_user_kern *u,
1161                   struct ip_vs_service **svc_p)
1162 {
1163         int ret = 0, i;
1164         struct ip_vs_scheduler *sched = NULL;
1165         struct ip_vs_pe *pe = NULL;
1166         struct ip_vs_service *svc = NULL;
1167
1168         /* increase the module use count */
1169         ip_vs_use_count_inc();
1170
1171         /* Lookup the scheduler by 'u->sched_name' */
1172         if (strcmp(u->sched_name, "none")) {
1173                 sched = ip_vs_scheduler_get(u->sched_name);
1174                 if (!sched) {
1175                         pr_info("Scheduler module ip_vs_%s not found\n",
1176                                 u->sched_name);
1177                         ret = -ENOENT;
1178                         goto out_err;
1179                 }
1180         }
1181
1182         if (u->pe_name && *u->pe_name) {
1183                 pe = ip_vs_pe_getbyname(u->pe_name);
1184                 if (pe == NULL) {
1185                         pr_info("persistence engine module ip_vs_pe_%s "
1186                                 "not found\n", u->pe_name);
1187                         ret = -ENOENT;
1188                         goto out_err;
1189                 }
1190         }
1191
1192 #ifdef CONFIG_IP_VS_IPV6
1193         if (u->af == AF_INET6) {
1194                 __u32 plen = (__force __u32) u->netmask;
1195
1196                 if (plen < 1 || plen > 128) {
1197                         ret = -EINVAL;
1198                         goto out_err;
1199                 }
1200         }
1201 #endif
1202
1203         svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1204         if (svc == NULL) {
1205                 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1206                 ret = -ENOMEM;
1207                 goto out_err;
1208         }
1209         svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1210         if (!svc->stats.cpustats) {
1211                 ret = -ENOMEM;
1212                 goto out_err;
1213         }
1214
1215         for_each_possible_cpu(i) {
1216                 struct ip_vs_cpu_stats *ip_vs_stats;
1217                 ip_vs_stats = per_cpu_ptr(svc->stats.cpustats, i);
1218                 u64_stats_init(&ip_vs_stats->syncp);
1219         }
1220
1221
1222         /* I'm the first user of the service */
1223         atomic_set(&svc->refcnt, 0);
1224
1225         svc->af = u->af;
1226         svc->protocol = u->protocol;
1227         ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1228         svc->port = u->port;
1229         svc->fwmark = u->fwmark;
1230         svc->flags = u->flags;
1231         svc->timeout = u->timeout * HZ;
1232         svc->netmask = u->netmask;
1233         svc->ipvs = ipvs;
1234
1235         INIT_LIST_HEAD(&svc->destinations);
1236         spin_lock_init(&svc->sched_lock);
1237         spin_lock_init(&svc->stats.lock);
1238
1239         /* Bind the scheduler */
1240         if (sched) {
1241                 ret = ip_vs_bind_scheduler(svc, sched);
1242                 if (ret)
1243                         goto out_err;
1244                 sched = NULL;
1245         }
1246
1247         /* Bind the ct retriever */
1248         RCU_INIT_POINTER(svc->pe, pe);
1249         pe = NULL;
1250
1251         /* Update the virtual service counters */
1252         if (svc->port == FTPPORT)
1253                 atomic_inc(&ipvs->ftpsvc_counter);
1254         else if (svc->port == 0)
1255                 atomic_inc(&ipvs->nullsvc_counter);
1256
1257         ip_vs_start_estimator(ipvs, &svc->stats);
1258
1259         /* Count only IPv4 services for old get/setsockopt interface */
1260         if (svc->af == AF_INET)
1261                 ipvs->num_services++;
1262
1263         /* Hash the service into the service table */
1264         ip_vs_svc_hash(svc);
1265
1266         *svc_p = svc;
1267         /* Now there is a service - full throttle */
1268         ipvs->enable = 1;
1269         return 0;
1270
1271
1272  out_err:
1273         if (svc != NULL) {
1274                 ip_vs_unbind_scheduler(svc, sched);
1275                 ip_vs_service_free(svc);
1276         }
1277         ip_vs_scheduler_put(sched);
1278         ip_vs_pe_put(pe);
1279
1280         /* decrease the module use count */
1281         ip_vs_use_count_dec();
1282
1283         return ret;
1284 }
1285
1286
1287 /*
1288  *      Edit a service and bind it with a new scheduler
1289  */
1290 static int
1291 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1292 {
1293         struct ip_vs_scheduler *sched = NULL, *old_sched;
1294         struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1295         int ret = 0;
1296
1297         /*
1298          * Lookup the scheduler, by 'u->sched_name'
1299          */
1300         if (strcmp(u->sched_name, "none")) {
1301                 sched = ip_vs_scheduler_get(u->sched_name);
1302                 if (!sched) {
1303                         pr_info("Scheduler module ip_vs_%s not found\n",
1304                                 u->sched_name);
1305                         return -ENOENT;
1306                 }
1307         }
1308         old_sched = sched;
1309
1310         if (u->pe_name && *u->pe_name) {
1311                 pe = ip_vs_pe_getbyname(u->pe_name);
1312                 if (pe == NULL) {
1313                         pr_info("persistence engine module ip_vs_pe_%s "
1314                                 "not found\n", u->pe_name);
1315                         ret = -ENOENT;
1316                         goto out;
1317                 }
1318                 old_pe = pe;
1319         }
1320
1321 #ifdef CONFIG_IP_VS_IPV6
1322         if (u->af == AF_INET6) {
1323                 __u32 plen = (__force __u32) u->netmask;
1324
1325                 if (plen < 1 || plen > 128) {
1326                         ret = -EINVAL;
1327                         goto out;
1328                 }
1329         }
1330 #endif
1331
1332         old_sched = rcu_dereference_protected(svc->scheduler, 1);
1333         if (sched != old_sched) {
1334                 if (old_sched) {
1335                         ip_vs_unbind_scheduler(svc, old_sched);
1336                         RCU_INIT_POINTER(svc->scheduler, NULL);
1337                         /* Wait all svc->sched_data users */
1338                         synchronize_rcu();
1339                 }
1340                 /* Bind the new scheduler */
1341                 if (sched) {
1342                         ret = ip_vs_bind_scheduler(svc, sched);
1343                         if (ret) {
1344                                 ip_vs_scheduler_put(sched);
1345                                 goto out;
1346                         }
1347                 }
1348         }
1349
1350         /*
1351          * Set the flags and timeout value
1352          */
1353         svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1354         svc->timeout = u->timeout * HZ;
1355         svc->netmask = u->netmask;
1356
1357         old_pe = rcu_dereference_protected(svc->pe, 1);
1358         if (pe != old_pe)
1359                 rcu_assign_pointer(svc->pe, pe);
1360
1361 out:
1362         ip_vs_scheduler_put(old_sched);
1363         ip_vs_pe_put(old_pe);
1364         return ret;
1365 }
1366
1367 /*
1368  *      Delete a service from the service list
1369  *      - The service must be unlinked, unlocked and not referenced!
1370  *      - We are called under _bh lock
1371  */
1372 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1373 {
1374         struct ip_vs_dest *dest, *nxt;
1375         struct ip_vs_scheduler *old_sched;
1376         struct ip_vs_pe *old_pe;
1377         struct netns_ipvs *ipvs = svc->ipvs;
1378
1379         /* Count only IPv4 services for old get/setsockopt interface */
1380         if (svc->af == AF_INET)
1381                 ipvs->num_services--;
1382
1383         ip_vs_stop_estimator(svc->ipvs, &svc->stats);
1384
1385         /* Unbind scheduler */
1386         old_sched = rcu_dereference_protected(svc->scheduler, 1);
1387         ip_vs_unbind_scheduler(svc, old_sched);
1388         ip_vs_scheduler_put(old_sched);
1389
1390         /* Unbind persistence engine, keep svc->pe */
1391         old_pe = rcu_dereference_protected(svc->pe, 1);
1392         ip_vs_pe_put(old_pe);
1393
1394         /*
1395          *    Unlink the whole destination list
1396          */
1397         list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1398                 __ip_vs_unlink_dest(svc, dest, 0);
1399                 __ip_vs_del_dest(svc->ipvs, dest, cleanup);
1400         }
1401
1402         /*
1403          *    Update the virtual service counters
1404          */
1405         if (svc->port == FTPPORT)
1406                 atomic_dec(&ipvs->ftpsvc_counter);
1407         else if (svc->port == 0)
1408                 atomic_dec(&ipvs->nullsvc_counter);
1409
1410         /*
1411          *    Free the service if nobody refers to it
1412          */
1413         __ip_vs_svc_put(svc, true);
1414
1415         /* decrease the module use count */
1416         ip_vs_use_count_dec();
1417 }
1418
1419 /*
1420  * Unlink a service from list and try to delete it if its refcnt reached 0
1421  */
1422 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1423 {
1424         /* Hold svc to avoid double release from dest_trash */
1425         atomic_inc(&svc->refcnt);
1426         /*
1427          * Unhash it from the service table
1428          */
1429         ip_vs_svc_unhash(svc);
1430
1431         __ip_vs_del_service(svc, cleanup);
1432 }
1433
1434 /*
1435  *      Delete a service from the service list
1436  */
1437 static int ip_vs_del_service(struct ip_vs_service *svc)
1438 {
1439         if (svc == NULL)
1440                 return -EEXIST;
1441         ip_vs_unlink_service(svc, false);
1442
1443         return 0;
1444 }
1445
1446
1447 /*
1448  *      Flush all the virtual services
1449  */
1450 static int ip_vs_flush(struct netns_ipvs *ipvs, bool cleanup)
1451 {
1452         int idx;
1453         struct ip_vs_service *svc;
1454         struct hlist_node *n;
1455
1456         /*
1457          * Flush the service table hashed by <netns,protocol,addr,port>
1458          */
1459         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1460                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1461                                           s_list) {
1462                         if (svc->ipvs == ipvs)
1463                                 ip_vs_unlink_service(svc, cleanup);
1464                 }
1465         }
1466
1467         /*
1468          * Flush the service table hashed by fwmark
1469          */
1470         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1471                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1472                                           f_list) {
1473                         if (svc->ipvs == ipvs)
1474                                 ip_vs_unlink_service(svc, cleanup);
1475                 }
1476         }
1477
1478         return 0;
1479 }
1480
1481 /*
1482  *      Delete service by {netns} in the service table.
1483  *      Called by __ip_vs_cleanup()
1484  */
1485 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs)
1486 {
1487         EnterFunction(2);
1488         /* Check for "full" addressed entries */
1489         mutex_lock(&__ip_vs_mutex);
1490         ip_vs_flush(ipvs, true);
1491         mutex_unlock(&__ip_vs_mutex);
1492         LeaveFunction(2);
1493 }
1494
1495 /* Put all references for device (dst_cache) */
1496 static inline void
1497 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1498 {
1499         struct ip_vs_dest_dst *dest_dst;
1500
1501         spin_lock_bh(&dest->dst_lock);
1502         dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1503         if (dest_dst && dest_dst->dst_cache->dev == dev) {
1504                 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1505                               dev->name,
1506                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1507                               ntohs(dest->port),
1508                               atomic_read(&dest->refcnt));
1509                 __ip_vs_dst_cache_reset(dest);
1510         }
1511         spin_unlock_bh(&dest->dst_lock);
1512
1513 }
1514 /* Netdev event receiver
1515  * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1516  */
1517 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1518                            void *ptr)
1519 {
1520         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1521         struct net *net = dev_net(dev);
1522         struct netns_ipvs *ipvs = net_ipvs(net);
1523         struct ip_vs_service *svc;
1524         struct ip_vs_dest *dest;
1525         unsigned int idx;
1526
1527         if (event != NETDEV_DOWN || !ipvs)
1528                 return NOTIFY_DONE;
1529         IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1530         EnterFunction(2);
1531         mutex_lock(&__ip_vs_mutex);
1532         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1533                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1534                         if (svc->ipvs == ipvs) {
1535                                 list_for_each_entry(dest, &svc->destinations,
1536                                                     n_list) {
1537                                         ip_vs_forget_dev(dest, dev);
1538                                 }
1539                         }
1540                 }
1541
1542                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1543                         if (svc->ipvs == ipvs) {
1544                                 list_for_each_entry(dest, &svc->destinations,
1545                                                     n_list) {
1546                                         ip_vs_forget_dev(dest, dev);
1547                                 }
1548                         }
1549
1550                 }
1551         }
1552
1553         spin_lock_bh(&ipvs->dest_trash_lock);
1554         list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1555                 ip_vs_forget_dev(dest, dev);
1556         }
1557         spin_unlock_bh(&ipvs->dest_trash_lock);
1558         mutex_unlock(&__ip_vs_mutex);
1559         LeaveFunction(2);
1560         return NOTIFY_DONE;
1561 }
1562
1563 /*
1564  *      Zero counters in a service or all services
1565  */
1566 static int ip_vs_zero_service(struct ip_vs_service *svc)
1567 {
1568         struct ip_vs_dest *dest;
1569
1570         list_for_each_entry(dest, &svc->destinations, n_list) {
1571                 ip_vs_zero_stats(&dest->stats);
1572         }
1573         ip_vs_zero_stats(&svc->stats);
1574         return 0;
1575 }
1576
1577 static int ip_vs_zero_all(struct netns_ipvs *ipvs)
1578 {
1579         int idx;
1580         struct ip_vs_service *svc;
1581
1582         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1583                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1584                         if (svc->ipvs == ipvs)
1585                                 ip_vs_zero_service(svc);
1586                 }
1587         }
1588
1589         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1590                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1591                         if (svc->ipvs == ipvs)
1592                                 ip_vs_zero_service(svc);
1593                 }
1594         }
1595
1596         ip_vs_zero_stats(&ipvs->tot_stats);
1597         return 0;
1598 }
1599
1600 #ifdef CONFIG_SYSCTL
1601
1602 static int zero;
1603 static int three = 3;
1604
1605 static int
1606 proc_do_defense_mode(struct ctl_table *table, int write,
1607                      void __user *buffer, size_t *lenp, loff_t *ppos)
1608 {
1609         struct netns_ipvs *ipvs = table->extra2;
1610         int *valp = table->data;
1611         int val = *valp;
1612         int rc;
1613
1614         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1615         if (write && (*valp != val)) {
1616                 if ((*valp < 0) || (*valp > 3)) {
1617                         /* Restore the correct value */
1618                         *valp = val;
1619                 } else {
1620                         update_defense_level(ipvs);
1621                 }
1622         }
1623         return rc;
1624 }
1625
1626 static int
1627 proc_do_sync_threshold(struct ctl_table *table, int write,
1628                        void __user *buffer, size_t *lenp, loff_t *ppos)
1629 {
1630         int *valp = table->data;
1631         int val[2];
1632         int rc;
1633
1634         /* backup the value first */
1635         memcpy(val, valp, sizeof(val));
1636
1637         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1638         if (write && (valp[0] < 0 || valp[1] < 0 ||
1639             (valp[0] >= valp[1] && valp[1]))) {
1640                 /* Restore the correct value */
1641                 memcpy(valp, val, sizeof(val));
1642         }
1643         return rc;
1644 }
1645
1646 static int
1647 proc_do_sync_mode(struct ctl_table *table, int write,
1648                      void __user *buffer, size_t *lenp, loff_t *ppos)
1649 {
1650         int *valp = table->data;
1651         int val = *valp;
1652         int rc;
1653
1654         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1655         if (write && (*valp != val)) {
1656                 if ((*valp < 0) || (*valp > 1)) {
1657                         /* Restore the correct value */
1658                         *valp = val;
1659                 }
1660         }
1661         return rc;
1662 }
1663
1664 static int
1665 proc_do_sync_ports(struct ctl_table *table, int write,
1666                    void __user *buffer, size_t *lenp, loff_t *ppos)
1667 {
1668         int *valp = table->data;
1669         int val = *valp;
1670         int rc;
1671
1672         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1673         if (write && (*valp != val)) {
1674                 if (*valp < 1 || !is_power_of_2(*valp)) {
1675                         /* Restore the correct value */
1676                         *valp = val;
1677                 }
1678         }
1679         return rc;
1680 }
1681
1682 /*
1683  *      IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1684  *      Do not change order or insert new entries without
1685  *      align with netns init in ip_vs_control_net_init()
1686  */
1687
1688 static struct ctl_table vs_vars[] = {
1689         {
1690                 .procname       = "amemthresh",
1691                 .maxlen         = sizeof(int),
1692                 .mode           = 0644,
1693                 .proc_handler   = proc_dointvec,
1694         },
1695         {
1696                 .procname       = "am_droprate",
1697                 .maxlen         = sizeof(int),
1698                 .mode           = 0644,
1699                 .proc_handler   = proc_dointvec,
1700         },
1701         {
1702                 .procname       = "drop_entry",
1703                 .maxlen         = sizeof(int),
1704                 .mode           = 0644,
1705                 .proc_handler   = proc_do_defense_mode,
1706         },
1707         {
1708                 .procname       = "drop_packet",
1709                 .maxlen         = sizeof(int),
1710                 .mode           = 0644,
1711                 .proc_handler   = proc_do_defense_mode,
1712         },
1713 #ifdef CONFIG_IP_VS_NFCT
1714         {
1715                 .procname       = "conntrack",
1716                 .maxlen         = sizeof(int),
1717                 .mode           = 0644,
1718                 .proc_handler   = &proc_dointvec,
1719         },
1720 #endif
1721         {
1722                 .procname       = "secure_tcp",
1723                 .maxlen         = sizeof(int),
1724                 .mode           = 0644,
1725                 .proc_handler   = proc_do_defense_mode,
1726         },
1727         {
1728                 .procname       = "snat_reroute",
1729                 .maxlen         = sizeof(int),
1730                 .mode           = 0644,
1731                 .proc_handler   = &proc_dointvec,
1732         },
1733         {
1734                 .procname       = "sync_version",
1735                 .maxlen         = sizeof(int),
1736                 .mode           = 0644,
1737                 .proc_handler   = &proc_do_sync_mode,
1738         },
1739         {
1740                 .procname       = "sync_ports",
1741                 .maxlen         = sizeof(int),
1742                 .mode           = 0644,
1743                 .proc_handler   = &proc_do_sync_ports,
1744         },
1745         {
1746                 .procname       = "sync_persist_mode",
1747                 .maxlen         = sizeof(int),
1748                 .mode           = 0644,
1749                 .proc_handler   = proc_dointvec,
1750         },
1751         {
1752                 .procname       = "sync_qlen_max",
1753                 .maxlen         = sizeof(unsigned long),
1754                 .mode           = 0644,
1755                 .proc_handler   = proc_doulongvec_minmax,
1756         },
1757         {
1758                 .procname       = "sync_sock_size",
1759                 .maxlen         = sizeof(int),
1760                 .mode           = 0644,
1761                 .proc_handler   = proc_dointvec,
1762         },
1763         {
1764                 .procname       = "cache_bypass",
1765                 .maxlen         = sizeof(int),
1766                 .mode           = 0644,
1767                 .proc_handler   = proc_dointvec,
1768         },
1769         {
1770                 .procname       = "expire_nodest_conn",
1771                 .maxlen         = sizeof(int),
1772                 .mode           = 0644,
1773                 .proc_handler   = proc_dointvec,
1774         },
1775         {
1776                 .procname       = "sloppy_tcp",
1777                 .maxlen         = sizeof(int),
1778                 .mode           = 0644,
1779                 .proc_handler   = proc_dointvec,
1780         },
1781         {
1782                 .procname       = "sloppy_sctp",
1783                 .maxlen         = sizeof(int),
1784                 .mode           = 0644,
1785                 .proc_handler   = proc_dointvec,
1786         },
1787         {
1788                 .procname       = "expire_quiescent_template",
1789                 .maxlen         = sizeof(int),
1790                 .mode           = 0644,
1791                 .proc_handler   = proc_dointvec,
1792         },
1793         {
1794                 .procname       = "sync_threshold",
1795                 .maxlen         =
1796                         sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1797                 .mode           = 0644,
1798                 .proc_handler   = proc_do_sync_threshold,
1799         },
1800         {
1801                 .procname       = "sync_refresh_period",
1802                 .maxlen         = sizeof(int),
1803                 .mode           = 0644,
1804                 .proc_handler   = proc_dointvec_jiffies,
1805         },
1806         {
1807                 .procname       = "sync_retries",
1808                 .maxlen         = sizeof(int),
1809                 .mode           = 0644,
1810                 .proc_handler   = proc_dointvec_minmax,
1811                 .extra1         = &zero,
1812                 .extra2         = &three,
1813         },
1814         {
1815                 .procname       = "nat_icmp_send",
1816                 .maxlen         = sizeof(int),
1817                 .mode           = 0644,
1818                 .proc_handler   = proc_dointvec,
1819         },
1820         {
1821                 .procname       = "pmtu_disc",
1822                 .maxlen         = sizeof(int),
1823                 .mode           = 0644,
1824                 .proc_handler   = proc_dointvec,
1825         },
1826         {
1827                 .procname       = "backup_only",
1828                 .maxlen         = sizeof(int),
1829                 .mode           = 0644,
1830                 .proc_handler   = proc_dointvec,
1831         },
1832         {
1833                 .procname       = "conn_reuse_mode",
1834                 .maxlen         = sizeof(int),
1835                 .mode           = 0644,
1836                 .proc_handler   = proc_dointvec,
1837         },
1838         {
1839                 .procname       = "schedule_icmp",
1840                 .maxlen         = sizeof(int),
1841                 .mode           = 0644,
1842                 .proc_handler   = proc_dointvec,
1843         },
1844         {
1845                 .procname       = "ignore_tunneled",
1846                 .maxlen         = sizeof(int),
1847                 .mode           = 0644,
1848                 .proc_handler   = proc_dointvec,
1849         },
1850 #ifdef CONFIG_IP_VS_DEBUG
1851         {
1852                 .procname       = "debug_level",
1853                 .data           = &sysctl_ip_vs_debug_level,
1854                 .maxlen         = sizeof(int),
1855                 .mode           = 0644,
1856                 .proc_handler   = proc_dointvec,
1857         },
1858 #endif
1859         { }
1860 };
1861
1862 #endif
1863
1864 #ifdef CONFIG_PROC_FS
1865
1866 struct ip_vs_iter {
1867         struct seq_net_private p;  /* Do not move this, netns depends upon it*/
1868         struct hlist_head *table;
1869         int bucket;
1870 };
1871
1872 /*
1873  *      Write the contents of the VS rule table to a PROCfs file.
1874  *      (It is kept just for backward compatibility)
1875  */
1876 static inline const char *ip_vs_fwd_name(unsigned int flags)
1877 {
1878         switch (flags & IP_VS_CONN_F_FWD_MASK) {
1879         case IP_VS_CONN_F_LOCALNODE:
1880                 return "Local";
1881         case IP_VS_CONN_F_TUNNEL:
1882                 return "Tunnel";
1883         case IP_VS_CONN_F_DROUTE:
1884                 return "Route";
1885         default:
1886                 return "Masq";
1887         }
1888 }
1889
1890
1891 /* Get the Nth entry in the two lists */
1892 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1893 {
1894         struct net *net = seq_file_net(seq);
1895         struct netns_ipvs *ipvs = net_ipvs(net);
1896         struct ip_vs_iter *iter = seq->private;
1897         int idx;
1898         struct ip_vs_service *svc;
1899
1900         /* look in hash by protocol */
1901         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1902                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1903                         if ((svc->ipvs == ipvs) && pos-- == 0) {
1904                                 iter->table = ip_vs_svc_table;
1905                                 iter->bucket = idx;
1906                                 return svc;
1907                         }
1908                 }
1909         }
1910
1911         /* keep looking in fwmark */
1912         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1913                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1914                                          f_list) {
1915                         if ((svc->ipvs == ipvs) && pos-- == 0) {
1916                                 iter->table = ip_vs_svc_fwm_table;
1917                                 iter->bucket = idx;
1918                                 return svc;
1919                         }
1920                 }
1921         }
1922
1923         return NULL;
1924 }
1925
1926 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1927         __acquires(RCU)
1928 {
1929         rcu_read_lock();
1930         return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1931 }
1932
1933
1934 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1935 {
1936         struct hlist_node *e;
1937         struct ip_vs_iter *iter;
1938         struct ip_vs_service *svc;
1939
1940         ++*pos;
1941         if (v == SEQ_START_TOKEN)
1942                 return ip_vs_info_array(seq,0);
1943
1944         svc = v;
1945         iter = seq->private;
1946
1947         if (iter->table == ip_vs_svc_table) {
1948                 /* next service in table hashed by protocol */
1949                 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
1950                 if (e)
1951                         return hlist_entry(e, struct ip_vs_service, s_list);
1952
1953                 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1954                         hlist_for_each_entry_rcu(svc,
1955                                                  &ip_vs_svc_table[iter->bucket],
1956                                                  s_list) {
1957                                 return svc;
1958                         }
1959                 }
1960
1961                 iter->table = ip_vs_svc_fwm_table;
1962                 iter->bucket = -1;
1963                 goto scan_fwmark;
1964         }
1965
1966         /* next service in hashed by fwmark */
1967         e = rcu_dereference(hlist_next_rcu(&svc->f_list));
1968         if (e)
1969                 return hlist_entry(e, struct ip_vs_service, f_list);
1970
1971  scan_fwmark:
1972         while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1973                 hlist_for_each_entry_rcu(svc,
1974                                          &ip_vs_svc_fwm_table[iter->bucket],
1975                                          f_list)
1976                         return svc;
1977         }
1978
1979         return NULL;
1980 }
1981
1982 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1983         __releases(RCU)
1984 {
1985         rcu_read_unlock();
1986 }
1987
1988
1989 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1990 {
1991         if (v == SEQ_START_TOKEN) {
1992                 seq_printf(seq,
1993                         "IP Virtual Server version %d.%d.%d (size=%d)\n",
1994                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1995                 seq_puts(seq,
1996                          "Prot LocalAddress:Port Scheduler Flags\n");
1997                 seq_puts(seq,
1998                          "  -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1999         } else {
2000                 const struct ip_vs_service *svc = v;
2001                 const struct ip_vs_iter *iter = seq->private;
2002                 const struct ip_vs_dest *dest;
2003                 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2004                 char *sched_name = sched ? sched->name : "none";
2005
2006                 if (iter->table == ip_vs_svc_table) {
2007 #ifdef CONFIG_IP_VS_IPV6
2008                         if (svc->af == AF_INET6)
2009                                 seq_printf(seq, "%s  [%pI6]:%04X %s ",
2010                                            ip_vs_proto_name(svc->protocol),
2011                                            &svc->addr.in6,
2012                                            ntohs(svc->port),
2013                                            sched_name);
2014                         else
2015 #endif
2016                                 seq_printf(seq, "%s  %08X:%04X %s %s ",
2017                                            ip_vs_proto_name(svc->protocol),
2018                                            ntohl(svc->addr.ip),
2019                                            ntohs(svc->port),
2020                                            sched_name,
2021                                            (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2022                 } else {
2023                         seq_printf(seq, "FWM  %08X %s %s",
2024                                    svc->fwmark, sched_name,
2025                                    (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2026                 }
2027
2028                 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2029                         seq_printf(seq, "persistent %d %08X\n",
2030                                 svc->timeout,
2031                                 ntohl(svc->netmask));
2032                 else
2033                         seq_putc(seq, '\n');
2034
2035                 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2036 #ifdef CONFIG_IP_VS_IPV6
2037                         if (dest->af == AF_INET6)
2038                                 seq_printf(seq,
2039                                            "  -> [%pI6]:%04X"
2040                                            "      %-7s %-6d %-10d %-10d\n",
2041                                            &dest->addr.in6,
2042                                            ntohs(dest->port),
2043                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2044                                            atomic_read(&dest->weight),
2045                                            atomic_read(&dest->activeconns),
2046                                            atomic_read(&dest->inactconns));
2047                         else
2048 #endif
2049                                 seq_printf(seq,
2050                                            "  -> %08X:%04X      "
2051                                            "%-7s %-6d %-10d %-10d\n",
2052                                            ntohl(dest->addr.ip),
2053                                            ntohs(dest->port),
2054                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2055                                            atomic_read(&dest->weight),
2056                                            atomic_read(&dest->activeconns),
2057                                            atomic_read(&dest->inactconns));
2058
2059                 }
2060         }
2061         return 0;
2062 }
2063
2064 static const struct seq_operations ip_vs_info_seq_ops = {
2065         .start = ip_vs_info_seq_start,
2066         .next  = ip_vs_info_seq_next,
2067         .stop  = ip_vs_info_seq_stop,
2068         .show  = ip_vs_info_seq_show,
2069 };
2070
2071 static int ip_vs_info_open(struct inode *inode, struct file *file)
2072 {
2073         return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2074                         sizeof(struct ip_vs_iter));
2075 }
2076
2077 static const struct file_operations ip_vs_info_fops = {
2078         .owner   = THIS_MODULE,
2079         .open    = ip_vs_info_open,
2080         .read    = seq_read,
2081         .llseek  = seq_lseek,
2082         .release = seq_release_net,
2083 };
2084
2085 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2086 {
2087         struct net *net = seq_file_single_net(seq);
2088         struct ip_vs_kstats show;
2089
2090 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2091         seq_puts(seq,
2092                  "   Total Incoming Outgoing         Incoming         Outgoing\n");
2093         seq_printf(seq,
2094                    "   Conns  Packets  Packets            Bytes            Bytes\n");
2095
2096         ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2097         seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n",
2098                    (unsigned long long)show.conns,
2099                    (unsigned long long)show.inpkts,
2100                    (unsigned long long)show.outpkts,
2101                    (unsigned long long)show.inbytes,
2102                    (unsigned long long)show.outbytes);
2103
2104 /*                01234567 01234567 01234567 0123456701234567 0123456701234567*/
2105         seq_puts(seq,
2106                  " Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2107         seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n",
2108                    (unsigned long long)show.cps,
2109                    (unsigned long long)show.inpps,
2110                    (unsigned long long)show.outpps,
2111                    (unsigned long long)show.inbps,
2112                    (unsigned long long)show.outbps);
2113
2114         return 0;
2115 }
2116
2117 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2118 {
2119         return single_open_net(inode, file, ip_vs_stats_show);
2120 }
2121
2122 static const struct file_operations ip_vs_stats_fops = {
2123         .owner = THIS_MODULE,
2124         .open = ip_vs_stats_seq_open,
2125         .read = seq_read,
2126         .llseek = seq_lseek,
2127         .release = single_release_net,
2128 };
2129
2130 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2131 {
2132         struct net *net = seq_file_single_net(seq);
2133         struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2134         struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2135         struct ip_vs_kstats kstats;
2136         int i;
2137
2138 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2139         seq_puts(seq,
2140                  "       Total Incoming Outgoing         Incoming         Outgoing\n");
2141         seq_printf(seq,
2142                    "CPU    Conns  Packets  Packets            Bytes            Bytes\n");
2143
2144         for_each_possible_cpu(i) {
2145                 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2146                 unsigned int start;
2147                 u64 conns, inpkts, outpkts, inbytes, outbytes;
2148
2149                 do {
2150                         start = u64_stats_fetch_begin_irq(&u->syncp);
2151                         conns = u->cnt.conns;
2152                         inpkts = u->cnt.inpkts;
2153                         outpkts = u->cnt.outpkts;
2154                         inbytes = u->cnt.inbytes;
2155                         outbytes = u->cnt.outbytes;
2156                 } while (u64_stats_fetch_retry_irq(&u->syncp, start));
2157
2158                 seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2159                            i, (u64)conns, (u64)inpkts,
2160                            (u64)outpkts, (u64)inbytes,
2161                            (u64)outbytes);
2162         }
2163
2164         ip_vs_copy_stats(&kstats, tot_stats);
2165
2166         seq_printf(seq, "  ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2167                    (unsigned long long)kstats.conns,
2168                    (unsigned long long)kstats.inpkts,
2169                    (unsigned long long)kstats.outpkts,
2170                    (unsigned long long)kstats.inbytes,
2171                    (unsigned long long)kstats.outbytes);
2172
2173 /*                ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2174         seq_puts(seq,
2175                  "     Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2176         seq_printf(seq, "    %8LX %8LX %8LX %16LX %16LX\n",
2177                    kstats.cps,
2178                    kstats.inpps,
2179                    kstats.outpps,
2180                    kstats.inbps,
2181                    kstats.outbps);
2182
2183         return 0;
2184 }
2185
2186 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2187 {
2188         return single_open_net(inode, file, ip_vs_stats_percpu_show);
2189 }
2190
2191 static const struct file_operations ip_vs_stats_percpu_fops = {
2192         .owner = THIS_MODULE,
2193         .open = ip_vs_stats_percpu_seq_open,
2194         .read = seq_read,
2195         .llseek = seq_lseek,
2196         .release = single_release_net,
2197 };
2198 #endif
2199
2200 /*
2201  *      Set timeout values for tcp tcpfin udp in the timeout_table.
2202  */
2203 static int ip_vs_set_timeout(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2204 {
2205 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2206         struct ip_vs_proto_data *pd;
2207 #endif
2208
2209         IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2210                   u->tcp_timeout,
2211                   u->tcp_fin_timeout,
2212                   u->udp_timeout);
2213
2214 #ifdef CONFIG_IP_VS_PROTO_TCP
2215         if (u->tcp_timeout) {
2216                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2217                 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2218                         = u->tcp_timeout * HZ;
2219         }
2220
2221         if (u->tcp_fin_timeout) {
2222                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2223                 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2224                         = u->tcp_fin_timeout * HZ;
2225         }
2226 #endif
2227
2228 #ifdef CONFIG_IP_VS_PROTO_UDP
2229         if (u->udp_timeout) {
2230                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2231                 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2232                         = u->udp_timeout * HZ;
2233         }
2234 #endif
2235         return 0;
2236 }
2237
2238 #define CMDID(cmd)              (cmd - IP_VS_BASE_CTL)
2239
2240 struct ip_vs_svcdest_user {
2241         struct ip_vs_service_user       s;
2242         struct ip_vs_dest_user          d;
2243 };
2244
2245 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2246         [CMDID(IP_VS_SO_SET_ADD)]         = sizeof(struct ip_vs_service_user),
2247         [CMDID(IP_VS_SO_SET_EDIT)]        = sizeof(struct ip_vs_service_user),
2248         [CMDID(IP_VS_SO_SET_DEL)]         = sizeof(struct ip_vs_service_user),
2249         [CMDID(IP_VS_SO_SET_ADDDEST)]     = sizeof(struct ip_vs_svcdest_user),
2250         [CMDID(IP_VS_SO_SET_DELDEST)]     = sizeof(struct ip_vs_svcdest_user),
2251         [CMDID(IP_VS_SO_SET_EDITDEST)]    = sizeof(struct ip_vs_svcdest_user),
2252         [CMDID(IP_VS_SO_SET_TIMEOUT)]     = sizeof(struct ip_vs_timeout_user),
2253         [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2254         [CMDID(IP_VS_SO_SET_STOPDAEMON)]  = sizeof(struct ip_vs_daemon_user),
2255         [CMDID(IP_VS_SO_SET_ZERO)]        = sizeof(struct ip_vs_service_user),
2256 };
2257
2258 union ip_vs_set_arglen {
2259         struct ip_vs_service_user       field_IP_VS_SO_SET_ADD;
2260         struct ip_vs_service_user       field_IP_VS_SO_SET_EDIT;
2261         struct ip_vs_service_user       field_IP_VS_SO_SET_DEL;
2262         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_ADDDEST;
2263         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_DELDEST;
2264         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_EDITDEST;
2265         struct ip_vs_timeout_user       field_IP_VS_SO_SET_TIMEOUT;
2266         struct ip_vs_daemon_user        field_IP_VS_SO_SET_STARTDAEMON;
2267         struct ip_vs_daemon_user        field_IP_VS_SO_SET_STOPDAEMON;
2268         struct ip_vs_service_user       field_IP_VS_SO_SET_ZERO;
2269 };
2270
2271 #define MAX_SET_ARGLEN  sizeof(union ip_vs_set_arglen)
2272
2273 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2274                                   struct ip_vs_service_user *usvc_compat)
2275 {
2276         memset(usvc, 0, sizeof(*usvc));
2277
2278         usvc->af                = AF_INET;
2279         usvc->protocol          = usvc_compat->protocol;
2280         usvc->addr.ip           = usvc_compat->addr;
2281         usvc->port              = usvc_compat->port;
2282         usvc->fwmark            = usvc_compat->fwmark;
2283
2284         /* Deep copy of sched_name is not needed here */
2285         usvc->sched_name        = usvc_compat->sched_name;
2286
2287         usvc->flags             = usvc_compat->flags;
2288         usvc->timeout           = usvc_compat->timeout;
2289         usvc->netmask           = usvc_compat->netmask;
2290 }
2291
2292 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2293                                    struct ip_vs_dest_user *udest_compat)
2294 {
2295         memset(udest, 0, sizeof(*udest));
2296
2297         udest->addr.ip          = udest_compat->addr;
2298         udest->port             = udest_compat->port;
2299         udest->conn_flags       = udest_compat->conn_flags;
2300         udest->weight           = udest_compat->weight;
2301         udest->u_threshold      = udest_compat->u_threshold;
2302         udest->l_threshold      = udest_compat->l_threshold;
2303         udest->af               = AF_INET;
2304 }
2305
2306 static int
2307 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2308 {
2309         struct net *net = sock_net(sk);
2310         int ret;
2311         unsigned char arg[MAX_SET_ARGLEN];
2312         struct ip_vs_service_user *usvc_compat;
2313         struct ip_vs_service_user_kern usvc;
2314         struct ip_vs_service *svc;
2315         struct ip_vs_dest_user *udest_compat;
2316         struct ip_vs_dest_user_kern udest;
2317         struct netns_ipvs *ipvs = net_ipvs(net);
2318
2319         BUILD_BUG_ON(sizeof(arg) > 255);
2320         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2321                 return -EPERM;
2322
2323         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2324                 return -EINVAL;
2325         if (len != set_arglen[CMDID(cmd)]) {
2326                 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2327                           len, set_arglen[CMDID(cmd)]);
2328                 return -EINVAL;
2329         }
2330
2331         if (copy_from_user(arg, user, len) != 0)
2332                 return -EFAULT;
2333
2334         /* increase the module use count */
2335         ip_vs_use_count_inc();
2336
2337         /* Handle daemons since they have another lock */
2338         if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2339             cmd == IP_VS_SO_SET_STOPDAEMON) {
2340                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2341
2342                 if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2343                         struct ipvs_sync_daemon_cfg cfg;
2344
2345                         memset(&cfg, 0, sizeof(cfg));
2346                         strlcpy(cfg.mcast_ifn, dm->mcast_ifn,
2347                                 sizeof(cfg.mcast_ifn));
2348                         cfg.syncid = dm->syncid;
2349                         rtnl_lock();
2350                         mutex_lock(&ipvs->sync_mutex);
2351                         ret = start_sync_thread(ipvs, &cfg, dm->state);
2352                         mutex_unlock(&ipvs->sync_mutex);
2353                         rtnl_unlock();
2354                 } else {
2355                         mutex_lock(&ipvs->sync_mutex);
2356                         ret = stop_sync_thread(ipvs, dm->state);
2357                         mutex_unlock(&ipvs->sync_mutex);
2358                 }
2359                 goto out_dec;
2360         }
2361
2362         mutex_lock(&__ip_vs_mutex);
2363         if (cmd == IP_VS_SO_SET_FLUSH) {
2364                 /* Flush the virtual service */
2365                 ret = ip_vs_flush(ipvs, false);
2366                 goto out_unlock;
2367         } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2368                 /* Set timeout values for (tcp tcpfin udp) */
2369                 ret = ip_vs_set_timeout(ipvs, (struct ip_vs_timeout_user *)arg);
2370                 goto out_unlock;
2371         }
2372
2373         usvc_compat = (struct ip_vs_service_user *)arg;
2374         udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2375
2376         /* We only use the new structs internally, so copy userspace compat
2377          * structs to extended internal versions */
2378         ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2379         ip_vs_copy_udest_compat(&udest, udest_compat);
2380
2381         if (cmd == IP_VS_SO_SET_ZERO) {
2382                 /* if no service address is set, zero counters in all */
2383                 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2384                         ret = ip_vs_zero_all(ipvs);
2385                         goto out_unlock;
2386                 }
2387         }
2388
2389         /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2390         if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2391             usvc.protocol != IPPROTO_SCTP) {
2392                 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2393                        usvc.protocol, &usvc.addr.ip,
2394                        ntohs(usvc.port), usvc.sched_name);
2395                 ret = -EFAULT;
2396                 goto out_unlock;
2397         }
2398
2399         /* Lookup the exact service by <protocol, addr, port> or fwmark */
2400         rcu_read_lock();
2401         if (usvc.fwmark == 0)
2402                 svc = __ip_vs_service_find(ipvs, usvc.af, usvc.protocol,
2403                                            &usvc.addr, usvc.port);
2404         else
2405                 svc = __ip_vs_svc_fwm_find(ipvs, usvc.af, usvc.fwmark);
2406         rcu_read_unlock();
2407
2408         if (cmd != IP_VS_SO_SET_ADD
2409             && (svc == NULL || svc->protocol != usvc.protocol)) {
2410                 ret = -ESRCH;
2411                 goto out_unlock;
2412         }
2413
2414         switch (cmd) {
2415         case IP_VS_SO_SET_ADD:
2416                 if (svc != NULL)
2417                         ret = -EEXIST;
2418                 else
2419                         ret = ip_vs_add_service(ipvs, &usvc, &svc);
2420                 break;
2421         case IP_VS_SO_SET_EDIT:
2422                 ret = ip_vs_edit_service(svc, &usvc);
2423                 break;
2424         case IP_VS_SO_SET_DEL:
2425                 ret = ip_vs_del_service(svc);
2426                 if (!ret)
2427                         goto out_unlock;
2428                 break;
2429         case IP_VS_SO_SET_ZERO:
2430                 ret = ip_vs_zero_service(svc);
2431                 break;
2432         case IP_VS_SO_SET_ADDDEST:
2433                 ret = ip_vs_add_dest(svc, &udest);
2434                 break;
2435         case IP_VS_SO_SET_EDITDEST:
2436                 ret = ip_vs_edit_dest(svc, &udest);
2437                 break;
2438         case IP_VS_SO_SET_DELDEST:
2439                 ret = ip_vs_del_dest(svc, &udest);
2440                 break;
2441         default:
2442                 ret = -EINVAL;
2443         }
2444
2445   out_unlock:
2446         mutex_unlock(&__ip_vs_mutex);
2447   out_dec:
2448         /* decrease the module use count */
2449         ip_vs_use_count_dec();
2450
2451         return ret;
2452 }
2453
2454
2455 static void
2456 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2457 {
2458         struct ip_vs_scheduler *sched;
2459         struct ip_vs_kstats kstats;
2460         char *sched_name;
2461
2462         sched = rcu_dereference_protected(src->scheduler, 1);
2463         sched_name = sched ? sched->name : "none";
2464         dst->protocol = src->protocol;
2465         dst->addr = src->addr.ip;
2466         dst->port = src->port;
2467         dst->fwmark = src->fwmark;
2468         strlcpy(dst->sched_name, sched_name, sizeof(dst->sched_name));
2469         dst->flags = src->flags;
2470         dst->timeout = src->timeout / HZ;
2471         dst->netmask = src->netmask;
2472         dst->num_dests = src->num_dests;
2473         ip_vs_copy_stats(&kstats, &src->stats);
2474         ip_vs_export_stats_user(&dst->stats, &kstats);
2475 }
2476
2477 static inline int
2478 __ip_vs_get_service_entries(struct netns_ipvs *ipvs,
2479                             const struct ip_vs_get_services *get,
2480                             struct ip_vs_get_services __user *uptr)
2481 {
2482         int idx, count=0;
2483         struct ip_vs_service *svc;
2484         struct ip_vs_service_entry entry;
2485         int ret = 0;
2486
2487         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2488                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2489                         /* Only expose IPv4 entries to old interface */
2490                         if (svc->af != AF_INET || (svc->ipvs != ipvs))
2491                                 continue;
2492
2493                         if (count >= get->num_services)
2494                                 goto out;
2495                         memset(&entry, 0, sizeof(entry));
2496                         ip_vs_copy_service(&entry, svc);
2497                         if (copy_to_user(&uptr->entrytable[count],
2498                                          &entry, sizeof(entry))) {
2499                                 ret = -EFAULT;
2500                                 goto out;
2501                         }
2502                         count++;
2503                 }
2504         }
2505
2506         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2507                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2508                         /* Only expose IPv4 entries to old interface */
2509                         if (svc->af != AF_INET || (svc->ipvs != ipvs))
2510                                 continue;
2511
2512                         if (count >= get->num_services)
2513                                 goto out;
2514                         memset(&entry, 0, sizeof(entry));
2515                         ip_vs_copy_service(&entry, svc);
2516                         if (copy_to_user(&uptr->entrytable[count],
2517                                          &entry, sizeof(entry))) {
2518                                 ret = -EFAULT;
2519                                 goto out;
2520                         }
2521                         count++;
2522                 }
2523         }
2524 out:
2525         return ret;
2526 }
2527
2528 static inline int
2529 __ip_vs_get_dest_entries(struct netns_ipvs *ipvs, const struct ip_vs_get_dests *get,
2530                          struct ip_vs_get_dests __user *uptr)
2531 {
2532         struct ip_vs_service *svc;
2533         union nf_inet_addr addr = { .ip = get->addr };
2534         int ret = 0;
2535
2536         rcu_read_lock();
2537         if (get->fwmark)
2538                 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, get->fwmark);
2539         else
2540                 svc = __ip_vs_service_find(ipvs, AF_INET, get->protocol, &addr,
2541                                            get->port);
2542         rcu_read_unlock();
2543
2544         if (svc) {
2545                 int count = 0;
2546                 struct ip_vs_dest *dest;
2547                 struct ip_vs_dest_entry entry;
2548                 struct ip_vs_kstats kstats;
2549
2550                 memset(&entry, 0, sizeof(entry));
2551                 list_for_each_entry(dest, &svc->destinations, n_list) {
2552                         if (count >= get->num_dests)
2553                                 break;
2554
2555                         /* Cannot expose heterogeneous members via sockopt
2556                          * interface
2557                          */
2558                         if (dest->af != svc->af)
2559                                 continue;
2560
2561                         entry.addr = dest->addr.ip;
2562                         entry.port = dest->port;
2563                         entry.conn_flags = atomic_read(&dest->conn_flags);
2564                         entry.weight = atomic_read(&dest->weight);
2565                         entry.u_threshold = dest->u_threshold;
2566                         entry.l_threshold = dest->l_threshold;
2567                         entry.activeconns = atomic_read(&dest->activeconns);
2568                         entry.inactconns = atomic_read(&dest->inactconns);
2569                         entry.persistconns = atomic_read(&dest->persistconns);
2570                         ip_vs_copy_stats(&kstats, &dest->stats);
2571                         ip_vs_export_stats_user(&entry.stats, &kstats);
2572                         if (copy_to_user(&uptr->entrytable[count],
2573                                          &entry, sizeof(entry))) {
2574                                 ret = -EFAULT;
2575                                 break;
2576                         }
2577                         count++;
2578                 }
2579         } else
2580                 ret = -ESRCH;
2581         return ret;
2582 }
2583
2584 static inline void
2585 __ip_vs_get_timeouts(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2586 {
2587 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2588         struct ip_vs_proto_data *pd;
2589 #endif
2590
2591         memset(u, 0, sizeof (*u));
2592
2593 #ifdef CONFIG_IP_VS_PROTO_TCP
2594         pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2595         u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2596         u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2597 #endif
2598 #ifdef CONFIG_IP_VS_PROTO_UDP
2599         pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2600         u->udp_timeout =
2601                         pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2602 #endif
2603 }
2604
2605 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2606         [CMDID(IP_VS_SO_GET_VERSION)]  = 64,
2607         [CMDID(IP_VS_SO_GET_INFO)]     = sizeof(struct ip_vs_getinfo),
2608         [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2609         [CMDID(IP_VS_SO_GET_SERVICE)]  = sizeof(struct ip_vs_service_entry),
2610         [CMDID(IP_VS_SO_GET_DESTS)]    = sizeof(struct ip_vs_get_dests),
2611         [CMDID(IP_VS_SO_GET_TIMEOUT)]  = sizeof(struct ip_vs_timeout_user),
2612         [CMDID(IP_VS_SO_GET_DAEMON)]   = 2 * sizeof(struct ip_vs_daemon_user),
2613 };
2614
2615 union ip_vs_get_arglen {
2616         char                            field_IP_VS_SO_GET_VERSION[64];
2617         struct ip_vs_getinfo            field_IP_VS_SO_GET_INFO;
2618         struct ip_vs_get_services       field_IP_VS_SO_GET_SERVICES;
2619         struct ip_vs_service_entry      field_IP_VS_SO_GET_SERVICE;
2620         struct ip_vs_get_dests          field_IP_VS_SO_GET_DESTS;
2621         struct ip_vs_timeout_user       field_IP_VS_SO_GET_TIMEOUT;
2622         struct ip_vs_daemon_user        field_IP_VS_SO_GET_DAEMON[2];
2623 };
2624
2625 #define MAX_GET_ARGLEN  sizeof(union ip_vs_get_arglen)
2626
2627 static int
2628 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2629 {
2630         unsigned char arg[MAX_GET_ARGLEN];
2631         int ret = 0;
2632         unsigned int copylen;
2633         struct net *net = sock_net(sk);
2634         struct netns_ipvs *ipvs = net_ipvs(net);
2635
2636         BUG_ON(!net);
2637         BUILD_BUG_ON(sizeof(arg) > 255);
2638         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2639                 return -EPERM;
2640
2641         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2642                 return -EINVAL;
2643
2644         copylen = get_arglen[CMDID(cmd)];
2645         if (*len < (int) copylen) {
2646                 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
2647                 return -EINVAL;
2648         }
2649
2650         if (copy_from_user(arg, user, copylen) != 0)
2651                 return -EFAULT;
2652         /*
2653          * Handle daemons first since it has its own locking
2654          */
2655         if (cmd == IP_VS_SO_GET_DAEMON) {
2656                 struct ip_vs_daemon_user d[2];
2657
2658                 memset(&d, 0, sizeof(d));
2659                 mutex_lock(&ipvs->sync_mutex);
2660                 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2661                         d[0].state = IP_VS_STATE_MASTER;
2662                         strlcpy(d[0].mcast_ifn, ipvs->mcfg.mcast_ifn,
2663                                 sizeof(d[0].mcast_ifn));
2664                         d[0].syncid = ipvs->mcfg.syncid;
2665                 }
2666                 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2667                         d[1].state = IP_VS_STATE_BACKUP;
2668                         strlcpy(d[1].mcast_ifn, ipvs->bcfg.mcast_ifn,
2669                                 sizeof(d[1].mcast_ifn));
2670                         d[1].syncid = ipvs->bcfg.syncid;
2671                 }
2672                 if (copy_to_user(user, &d, sizeof(d)) != 0)
2673                         ret = -EFAULT;
2674                 mutex_unlock(&ipvs->sync_mutex);
2675                 return ret;
2676         }
2677
2678         mutex_lock(&__ip_vs_mutex);
2679         switch (cmd) {
2680         case IP_VS_SO_GET_VERSION:
2681         {
2682                 char buf[64];
2683
2684                 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2685                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2686                 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2687                         ret = -EFAULT;
2688                         goto out;
2689                 }
2690                 *len = strlen(buf)+1;
2691         }
2692         break;
2693
2694         case IP_VS_SO_GET_INFO:
2695         {
2696                 struct ip_vs_getinfo info;
2697                 info.version = IP_VS_VERSION_CODE;
2698                 info.size = ip_vs_conn_tab_size;
2699                 info.num_services = ipvs->num_services;
2700                 if (copy_to_user(user, &info, sizeof(info)) != 0)
2701                         ret = -EFAULT;
2702         }
2703         break;
2704
2705         case IP_VS_SO_GET_SERVICES:
2706         {
2707                 struct ip_vs_get_services *get;
2708                 int size;
2709
2710                 get = (struct ip_vs_get_services *)arg;
2711                 size = sizeof(*get) +
2712                         sizeof(struct ip_vs_service_entry) * get->num_services;
2713                 if (*len != size) {
2714                         pr_err("length: %u != %u\n", *len, size);
2715                         ret = -EINVAL;
2716                         goto out;
2717                 }
2718                 ret = __ip_vs_get_service_entries(ipvs, get, user);
2719         }
2720         break;
2721
2722         case IP_VS_SO_GET_SERVICE:
2723         {
2724                 struct ip_vs_service_entry *entry;
2725                 struct ip_vs_service *svc;
2726                 union nf_inet_addr addr;
2727
2728                 entry = (struct ip_vs_service_entry *)arg;
2729                 addr.ip = entry->addr;
2730                 rcu_read_lock();
2731                 if (entry->fwmark)
2732                         svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, entry->fwmark);
2733                 else
2734                         svc = __ip_vs_service_find(ipvs, AF_INET,
2735                                                    entry->protocol, &addr,
2736                                                    entry->port);
2737                 rcu_read_unlock();
2738                 if (svc) {
2739                         ip_vs_copy_service(entry, svc);
2740                         if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2741                                 ret = -EFAULT;
2742                 } else
2743                         ret = -ESRCH;
2744         }
2745         break;
2746
2747         case IP_VS_SO_GET_DESTS:
2748         {
2749                 struct ip_vs_get_dests *get;
2750                 int size;
2751
2752                 get = (struct ip_vs_get_dests *)arg;
2753                 size = sizeof(*get) +
2754                         sizeof(struct ip_vs_dest_entry) * get->num_dests;
2755                 if (*len != size) {
2756                         pr_err("length: %u != %u\n", *len, size);
2757                         ret = -EINVAL;
2758                         goto out;
2759                 }
2760                 ret = __ip_vs_get_dest_entries(ipvs, get, user);
2761         }
2762         break;
2763
2764         case IP_VS_SO_GET_TIMEOUT:
2765         {
2766                 struct ip_vs_timeout_user t;
2767
2768                 __ip_vs_get_timeouts(ipvs, &t);
2769                 if (copy_to_user(user, &t, sizeof(t)) != 0)
2770                         ret = -EFAULT;
2771         }
2772         break;
2773
2774         default:
2775                 ret = -EINVAL;
2776         }
2777
2778 out:
2779         mutex_unlock(&__ip_vs_mutex);
2780         return ret;
2781 }
2782
2783
2784 static struct nf_sockopt_ops ip_vs_sockopts = {
2785         .pf             = PF_INET,
2786         .set_optmin     = IP_VS_BASE_CTL,
2787         .set_optmax     = IP_VS_SO_SET_MAX+1,
2788         .set            = do_ip_vs_set_ctl,
2789         .get_optmin     = IP_VS_BASE_CTL,
2790         .get_optmax     = IP_VS_SO_GET_MAX+1,
2791         .get            = do_ip_vs_get_ctl,
2792         .owner          = THIS_MODULE,
2793 };
2794
2795 /*
2796  * Generic Netlink interface
2797  */
2798
2799 /* IPVS genetlink family */
2800 static struct genl_family ip_vs_genl_family = {
2801         .id             = GENL_ID_GENERATE,
2802         .hdrsize        = 0,
2803         .name           = IPVS_GENL_NAME,
2804         .version        = IPVS_GENL_VERSION,
2805         .maxattr        = IPVS_CMD_MAX,
2806         .netnsok        = true,         /* Make ipvsadm to work on netns */
2807 };
2808
2809 /* Policy used for first-level command attributes */
2810 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2811         [IPVS_CMD_ATTR_SERVICE]         = { .type = NLA_NESTED },
2812         [IPVS_CMD_ATTR_DEST]            = { .type = NLA_NESTED },
2813         [IPVS_CMD_ATTR_DAEMON]          = { .type = NLA_NESTED },
2814         [IPVS_CMD_ATTR_TIMEOUT_TCP]     = { .type = NLA_U32 },
2815         [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2816         [IPVS_CMD_ATTR_TIMEOUT_UDP]     = { .type = NLA_U32 },
2817 };
2818
2819 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2820 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2821         [IPVS_DAEMON_ATTR_STATE]        = { .type = NLA_U32 },
2822         [IPVS_DAEMON_ATTR_MCAST_IFN]    = { .type = NLA_NUL_STRING,
2823                                             .len = IP_VS_IFNAME_MAXLEN },
2824         [IPVS_DAEMON_ATTR_SYNC_ID]      = { .type = NLA_U32 },
2825         [IPVS_DAEMON_ATTR_SYNC_MAXLEN]  = { .type = NLA_U16 },
2826         [IPVS_DAEMON_ATTR_MCAST_GROUP]  = { .type = NLA_U32 },
2827         [IPVS_DAEMON_ATTR_MCAST_GROUP6] = { .len = sizeof(struct in6_addr) },
2828         [IPVS_DAEMON_ATTR_MCAST_PORT]   = { .type = NLA_U16 },
2829         [IPVS_DAEMON_ATTR_MCAST_TTL]    = { .type = NLA_U8 },
2830 };
2831
2832 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2833 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2834         [IPVS_SVC_ATTR_AF]              = { .type = NLA_U16 },
2835         [IPVS_SVC_ATTR_PROTOCOL]        = { .type = NLA_U16 },
2836         [IPVS_SVC_ATTR_ADDR]            = { .type = NLA_BINARY,
2837                                             .len = sizeof(union nf_inet_addr) },
2838         [IPVS_SVC_ATTR_PORT]            = { .type = NLA_U16 },
2839         [IPVS_SVC_ATTR_FWMARK]          = { .type = NLA_U32 },
2840         [IPVS_SVC_ATTR_SCHED_NAME]      = { .type = NLA_NUL_STRING,
2841                                             .len = IP_VS_SCHEDNAME_MAXLEN },
2842         [IPVS_SVC_ATTR_PE_NAME]         = { .type = NLA_NUL_STRING,
2843                                             .len = IP_VS_PENAME_MAXLEN },
2844         [IPVS_SVC_ATTR_FLAGS]           = { .type = NLA_BINARY,
2845                                             .len = sizeof(struct ip_vs_flags) },
2846         [IPVS_SVC_ATTR_TIMEOUT]         = { .type = NLA_U32 },
2847         [IPVS_SVC_ATTR_NETMASK]         = { .type = NLA_U32 },
2848         [IPVS_SVC_ATTR_STATS]           = { .type = NLA_NESTED },
2849 };
2850
2851 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2852 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2853         [IPVS_DEST_ATTR_ADDR]           = { .type = NLA_BINARY,
2854                                             .len = sizeof(union nf_inet_addr) },
2855         [IPVS_DEST_ATTR_PORT]           = { .type = NLA_U16 },
2856         [IPVS_DEST_ATTR_FWD_METHOD]     = { .type = NLA_U32 },
2857         [IPVS_DEST_ATTR_WEIGHT]         = { .type = NLA_U32 },
2858         [IPVS_DEST_ATTR_U_THRESH]       = { .type = NLA_U32 },
2859         [IPVS_DEST_ATTR_L_THRESH]       = { .type = NLA_U32 },
2860         [IPVS_DEST_ATTR_ACTIVE_CONNS]   = { .type = NLA_U32 },
2861         [IPVS_DEST_ATTR_INACT_CONNS]    = { .type = NLA_U32 },
2862         [IPVS_DEST_ATTR_PERSIST_CONNS]  = { .type = NLA_U32 },
2863         [IPVS_DEST_ATTR_STATS]          = { .type = NLA_NESTED },
2864         [IPVS_DEST_ATTR_ADDR_FAMILY]    = { .type = NLA_U16 },
2865 };
2866
2867 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2868                                  struct ip_vs_kstats *kstats)
2869 {
2870         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2871
2872         if (!nl_stats)
2873                 return -EMSGSIZE;
2874
2875         if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) ||
2876             nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) ||
2877             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) ||
2878             nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes) ||
2879             nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes) ||
2880             nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) ||
2881             nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) ||
2882             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) ||
2883             nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) ||
2884             nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps))
2885                 goto nla_put_failure;
2886         nla_nest_end(skb, nl_stats);
2887
2888         return 0;
2889
2890 nla_put_failure:
2891         nla_nest_cancel(skb, nl_stats);
2892         return -EMSGSIZE;
2893 }
2894
2895 static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type,
2896                                    struct ip_vs_kstats *kstats)
2897 {
2898         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2899
2900         if (!nl_stats)
2901                 return -EMSGSIZE;
2902
2903         if (nla_put_u64(skb, IPVS_STATS_ATTR_CONNS, kstats->conns) ||
2904             nla_put_u64(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts) ||
2905             nla_put_u64(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts) ||
2906             nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes) ||
2907             nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes) ||
2908             nla_put_u64(skb, IPVS_STATS_ATTR_CPS, kstats->cps) ||
2909             nla_put_u64(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps) ||
2910             nla_put_u64(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps) ||
2911             nla_put_u64(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps) ||
2912             nla_put_u64(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps))
2913                 goto nla_put_failure;
2914         nla_nest_end(skb, nl_stats);
2915
2916         return 0;
2917
2918 nla_put_failure:
2919         nla_nest_cancel(skb, nl_stats);
2920         return -EMSGSIZE;
2921 }
2922
2923 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2924                                    struct ip_vs_service *svc)
2925 {
2926         struct ip_vs_scheduler *sched;
2927         struct ip_vs_pe *pe;
2928         struct nlattr *nl_service;
2929         struct ip_vs_flags flags = { .flags = svc->flags,
2930                                      .mask = ~0 };
2931         struct ip_vs_kstats kstats;
2932         char *sched_name;
2933
2934         nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2935         if (!nl_service)
2936                 return -EMSGSIZE;
2937
2938         if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2939                 goto nla_put_failure;
2940         if (svc->fwmark) {
2941                 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2942                         goto nla_put_failure;
2943         } else {
2944                 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2945                     nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2946                     nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2947                         goto nla_put_failure;
2948         }
2949
2950         sched = rcu_dereference_protected(svc->scheduler, 1);
2951         sched_name = sched ? sched->name : "none";
2952         pe = rcu_dereference_protected(svc->pe, 1);
2953         if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) ||
2954             (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
2955             nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2956             nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2957             nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2958                 goto nla_put_failure;
2959         ip_vs_copy_stats(&kstats, &svc->stats);
2960         if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats))
2961                 goto nla_put_failure;
2962         if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats))
2963                 goto nla_put_failure;
2964
2965         nla_nest_end(skb, nl_service);
2966
2967         return 0;
2968
2969 nla_put_failure:
2970         nla_nest_cancel(skb, nl_service);
2971         return -EMSGSIZE;
2972 }
2973
2974 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2975                                    struct ip_vs_service *svc,
2976                                    struct netlink_callback *cb)
2977 {
2978         void *hdr;
2979
2980         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2981                           &ip_vs_genl_family, NLM_F_MULTI,
2982                           IPVS_CMD_NEW_SERVICE);
2983         if (!hdr)
2984                 return -EMSGSIZE;
2985
2986         if (ip_vs_genl_fill_service(skb, svc) < 0)
2987                 goto nla_put_failure;
2988
2989         genlmsg_end(skb, hdr);
2990         return 0;
2991
2992 nla_put_failure:
2993         genlmsg_cancel(skb, hdr);
2994         return -EMSGSIZE;
2995 }
2996
2997 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2998                                     struct netlink_callback *cb)
2999 {
3000         int idx = 0, i;
3001         int start = cb->args[0];
3002         struct ip_vs_service *svc;
3003         struct net *net = sock_net(skb->sk);
3004         struct netns_ipvs *ipvs = net_ipvs(net);
3005
3006         mutex_lock(&__ip_vs_mutex);
3007         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3008                 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
3009                         if (++idx <= start || (svc->ipvs != ipvs))
3010                                 continue;
3011                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3012                                 idx--;
3013                                 goto nla_put_failure;
3014                         }
3015                 }
3016         }
3017
3018         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3019                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3020                         if (++idx <= start || (svc->ipvs != ipvs))
3021                                 continue;
3022                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3023                                 idx--;
3024                                 goto nla_put_failure;
3025                         }
3026                 }
3027         }
3028
3029 nla_put_failure:
3030         mutex_unlock(&__ip_vs_mutex);
3031         cb->args[0] = idx;
3032
3033         return skb->len;
3034 }
3035
3036 static int ip_vs_genl_parse_service(struct netns_ipvs *ipvs,
3037                                     struct ip_vs_service_user_kern *usvc,
3038                                     struct nlattr *nla, int full_entry,
3039                                     struct ip_vs_service **ret_svc)
3040 {
3041         struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3042         struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3043         struct ip_vs_service *svc;
3044
3045         /* Parse mandatory identifying service fields first */
3046         if (nla == NULL ||
3047             nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
3048                 return -EINVAL;
3049
3050         nla_af          = attrs[IPVS_SVC_ATTR_AF];
3051         nla_protocol    = attrs[IPVS_SVC_ATTR_PROTOCOL];
3052         nla_addr        = attrs[IPVS_SVC_ATTR_ADDR];
3053         nla_port        = attrs[IPVS_SVC_ATTR_PORT];
3054         nla_fwmark      = attrs[IPVS_SVC_ATTR_FWMARK];
3055
3056         if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3057                 return -EINVAL;
3058
3059         memset(usvc, 0, sizeof(*usvc));
3060
3061         usvc->af = nla_get_u16(nla_af);
3062 #ifdef CONFIG_IP_VS_IPV6
3063         if (usvc->af != AF_INET && usvc->af != AF_INET6)
3064 #else
3065         if (usvc->af != AF_INET)
3066 #endif
3067                 return -EAFNOSUPPORT;
3068
3069         if (nla_fwmark) {
3070                 usvc->protocol = IPPROTO_TCP;
3071                 usvc->fwmark = nla_get_u32(nla_fwmark);
3072         } else {
3073                 usvc->protocol = nla_get_u16(nla_protocol);
3074                 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3075                 usvc->port = nla_get_be16(nla_port);
3076                 usvc->fwmark = 0;
3077         }
3078
3079         rcu_read_lock();
3080         if (usvc->fwmark)
3081                 svc = __ip_vs_svc_fwm_find(ipvs, usvc->af, usvc->fwmark);
3082         else
3083                 svc = __ip_vs_service_find(ipvs, usvc->af, usvc->protocol,
3084                                            &usvc->addr, usvc->port);
3085         rcu_read_unlock();
3086         *ret_svc = svc;
3087
3088         /* If a full entry was requested, check for the additional fields */
3089         if (full_entry) {
3090                 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3091                               *nla_netmask;
3092                 struct ip_vs_flags flags;
3093
3094                 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3095                 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3096                 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3097                 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3098                 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3099
3100                 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3101                         return -EINVAL;
3102
3103                 nla_memcpy(&flags, nla_flags, sizeof(flags));
3104
3105                 /* prefill flags from service if it already exists */
3106                 if (svc)
3107                         usvc->flags = svc->flags;
3108
3109                 /* set new flags from userland */
3110                 usvc->flags = (usvc->flags & ~flags.mask) |
3111                               (flags.flags & flags.mask);
3112                 usvc->sched_name = nla_data(nla_sched);
3113                 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3114                 usvc->timeout = nla_get_u32(nla_timeout);
3115                 usvc->netmask = nla_get_be32(nla_netmask);
3116         }
3117
3118         return 0;
3119 }
3120
3121 static struct ip_vs_service *ip_vs_genl_find_service(struct netns_ipvs *ipvs,
3122                                                      struct nlattr *nla)
3123 {
3124         struct ip_vs_service_user_kern usvc;
3125         struct ip_vs_service *svc;
3126         int ret;
3127
3128         ret = ip_vs_genl_parse_service(ipvs, &usvc, nla, 0, &svc);
3129         return ret ? ERR_PTR(ret) : svc;
3130 }
3131
3132 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3133 {
3134         struct nlattr *nl_dest;
3135         struct ip_vs_kstats kstats;
3136
3137         nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3138         if (!nl_dest)
3139                 return -EMSGSIZE;
3140
3141         if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3142             nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3143             nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3144                         (atomic_read(&dest->conn_flags) &
3145                          IP_VS_CONN_F_FWD_MASK)) ||
3146             nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3147                         atomic_read(&dest->weight)) ||
3148             nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3149             nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3150             nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3151                         atomic_read(&dest->activeconns)) ||
3152             nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3153                         atomic_read(&dest->inactconns)) ||
3154             nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3155                         atomic_read(&dest->persistconns)) ||
3156             nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3157                 goto nla_put_failure;
3158         ip_vs_copy_stats(&kstats, &dest->stats);
3159         if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats))
3160                 goto nla_put_failure;
3161         if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats))
3162                 goto nla_put_failure;
3163
3164         nla_nest_end(skb, nl_dest);
3165
3166         return 0;
3167
3168 nla_put_failure:
3169         nla_nest_cancel(skb, nl_dest);
3170         return -EMSGSIZE;
3171 }
3172
3173 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3174                                 struct netlink_callback *cb)
3175 {
3176         void *hdr;
3177
3178         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3179                           &ip_vs_genl_family, NLM_F_MULTI,
3180                           IPVS_CMD_NEW_DEST);
3181         if (!hdr)
3182                 return -EMSGSIZE;
3183
3184         if (ip_vs_genl_fill_dest(skb, dest) < 0)
3185                 goto nla_put_failure;
3186
3187         genlmsg_end(skb, hdr);
3188         return 0;
3189
3190 nla_put_failure:
3191         genlmsg_cancel(skb, hdr);
3192         return -EMSGSIZE;
3193 }
3194
3195 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3196                                  struct netlink_callback *cb)
3197 {
3198         int idx = 0;
3199         int start = cb->args[0];
3200         struct ip_vs_service *svc;
3201         struct ip_vs_dest *dest;
3202         struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3203         struct net *net = sock_net(skb->sk);
3204         struct netns_ipvs *ipvs = net_ipvs(net);
3205
3206         mutex_lock(&__ip_vs_mutex);
3207
3208         /* Try to find the service for which to dump destinations */
3209         if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3210                         IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3211                 goto out_err;
3212
3213
3214         svc = ip_vs_genl_find_service(ipvs, attrs[IPVS_CMD_ATTR_SERVICE]);
3215         if (IS_ERR(svc) || svc == NULL)
3216                 goto out_err;
3217
3218         /* Dump the destinations */
3219         list_for_each_entry(dest, &svc->destinations, n_list) {
3220                 if (++idx <= start)
3221                         continue;
3222                 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3223                         idx--;
3224                         goto nla_put_failure;
3225                 }
3226         }
3227
3228 nla_put_failure:
3229         cb->args[0] = idx;
3230
3231 out_err:
3232         mutex_unlock(&__ip_vs_mutex);
3233
3234         return skb->len;
3235 }
3236
3237 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3238                                  struct nlattr *nla, int full_entry)
3239 {
3240         struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3241         struct nlattr *nla_addr, *nla_port;
3242         struct nlattr *nla_addr_family;
3243
3244         /* Parse mandatory identifying destination fields first */
3245         if (nla == NULL ||
3246             nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3247                 return -EINVAL;
3248
3249         nla_addr        = attrs[IPVS_DEST_ATTR_ADDR];
3250         nla_port        = attrs[IPVS_DEST_ATTR_PORT];
3251         nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3252
3253         if (!(nla_addr && nla_port))
3254                 return -EINVAL;
3255
3256         memset(udest, 0, sizeof(*udest));
3257
3258         nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3259         udest->port = nla_get_be16(nla_port);
3260
3261         if (nla_addr_family)
3262                 udest->af = nla_get_u16(nla_addr_family);
3263         else
3264                 udest->af = 0;
3265
3266         /* If a full entry was requested, check for the additional fields */
3267         if (full_entry) {
3268                 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3269                               *nla_l_thresh;
3270
3271                 nla_fwd         = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3272                 nla_weight      = attrs[IPVS_DEST_ATTR_WEIGHT];
3273                 nla_u_thresh    = attrs[IPVS_DEST_ATTR_U_THRESH];
3274                 nla_l_thresh    = attrs[IPVS_DEST_ATTR_L_THRESH];
3275
3276                 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3277                         return -EINVAL;
3278
3279                 udest->conn_flags = nla_get_u32(nla_fwd)
3280                                     & IP_VS_CONN_F_FWD_MASK;
3281                 udest->weight = nla_get_u32(nla_weight);
3282                 udest->u_threshold = nla_get_u32(nla_u_thresh);
3283                 udest->l_threshold = nla_get_u32(nla_l_thresh);
3284         }
3285
3286         return 0;
3287 }
3288
3289 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3290                                   struct ipvs_sync_daemon_cfg *c)
3291 {
3292         struct nlattr *nl_daemon;
3293
3294         nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3295         if (!nl_daemon)
3296                 return -EMSGSIZE;
3297
3298         if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3299             nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, c->mcast_ifn) ||
3300             nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, c->syncid) ||
3301             nla_put_u16(skb, IPVS_DAEMON_ATTR_SYNC_MAXLEN, c->sync_maxlen) ||
3302             nla_put_u16(skb, IPVS_DAEMON_ATTR_MCAST_PORT, c->mcast_port) ||
3303             nla_put_u8(skb, IPVS_DAEMON_ATTR_MCAST_TTL, c->mcast_ttl))
3304                 goto nla_put_failure;
3305 #ifdef CONFIG_IP_VS_IPV6
3306         if (c->mcast_af == AF_INET6) {
3307                 if (nla_put_in6_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP6,
3308                                      &c->mcast_group.in6))
3309                         goto nla_put_failure;
3310         } else
3311 #endif
3312                 if (c->mcast_af == AF_INET &&
3313                     nla_put_in_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP,
3314                                     c->mcast_group.ip))
3315                         goto nla_put_failure;
3316         nla_nest_end(skb, nl_daemon);
3317
3318         return 0;
3319
3320 nla_put_failure:
3321         nla_nest_cancel(skb, nl_daemon);
3322         return -EMSGSIZE;
3323 }
3324
3325 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3326                                   struct ipvs_sync_daemon_cfg *c,
3327                                   struct netlink_callback *cb)
3328 {
3329         void *hdr;
3330         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3331                           &ip_vs_genl_family, NLM_F_MULTI,
3332                           IPVS_CMD_NEW_DAEMON);
3333         if (!hdr)
3334                 return -EMSGSIZE;
3335
3336         if (ip_vs_genl_fill_daemon(skb, state, c))
3337                 goto nla_put_failure;
3338
3339         genlmsg_end(skb, hdr);
3340         return 0;
3341
3342 nla_put_failure:
3343         genlmsg_cancel(skb, hdr);
3344         return -EMSGSIZE;
3345 }
3346
3347 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3348                                    struct netlink_callback *cb)
3349 {
3350         struct net *net = sock_net(skb->sk);
3351         struct netns_ipvs *ipvs = net_ipvs(net);
3352
3353         mutex_lock(&ipvs->sync_mutex);
3354         if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3355                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3356                                            &ipvs->mcfg, cb) < 0)
3357                         goto nla_put_failure;
3358
3359                 cb->args[0] = 1;
3360         }
3361
3362         if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3363                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3364                                            &ipvs->bcfg, cb) < 0)
3365                         goto nla_put_failure;
3366
3367                 cb->args[1] = 1;
3368         }
3369
3370 nla_put_failure:
3371         mutex_unlock(&ipvs->sync_mutex);
3372
3373         return skb->len;
3374 }
3375
3376 static int ip_vs_genl_new_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3377 {
3378         struct ipvs_sync_daemon_cfg c;
3379         struct nlattr *a;
3380         int ret;
3381
3382         memset(&c, 0, sizeof(c));
3383         if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3384               attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3385               attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3386                 return -EINVAL;
3387         strlcpy(c.mcast_ifn, nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3388                 sizeof(c.mcast_ifn));
3389         c.syncid = nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]);
3390
3391         a = attrs[IPVS_DAEMON_ATTR_SYNC_MAXLEN];
3392         if (a)
3393                 c.sync_maxlen = nla_get_u16(a);
3394
3395         a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP];
3396         if (a) {
3397                 c.mcast_af = AF_INET;
3398                 c.mcast_group.ip = nla_get_in_addr(a);
3399                 if (!ipv4_is_multicast(c.mcast_group.ip))
3400                         return -EINVAL;
3401         } else {
3402                 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP6];
3403                 if (a) {
3404 #ifdef CONFIG_IP_VS_IPV6
3405                         int addr_type;
3406
3407                         c.mcast_af = AF_INET6;
3408                         c.mcast_group.in6 = nla_get_in6_addr(a);
3409                         addr_type = ipv6_addr_type(&c.mcast_group.in6);
3410                         if (!(addr_type & IPV6_ADDR_MULTICAST))
3411                                 return -EINVAL;
3412 #else
3413                         return -EAFNOSUPPORT;
3414 #endif
3415                 }
3416         }
3417
3418         a = attrs[IPVS_DAEMON_ATTR_MCAST_PORT];
3419         if (a)
3420                 c.mcast_port = nla_get_u16(a);
3421
3422         a = attrs[IPVS_DAEMON_ATTR_MCAST_TTL];
3423         if (a)
3424                 c.mcast_ttl = nla_get_u8(a);
3425
3426         /* The synchronization protocol is incompatible with mixed family
3427          * services
3428          */
3429         if (ipvs->mixed_address_family_dests > 0)
3430                 return -EINVAL;
3431
3432         rtnl_lock();
3433         mutex_lock(&ipvs->sync_mutex);
3434         ret = start_sync_thread(ipvs, &c,
3435                                 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3436         mutex_unlock(&ipvs->sync_mutex);
3437         rtnl_unlock();
3438         return ret;
3439 }
3440
3441 static int ip_vs_genl_del_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3442 {
3443         int ret;
3444
3445         if (!attrs[IPVS_DAEMON_ATTR_STATE])
3446                 return -EINVAL;
3447
3448         mutex_lock(&ipvs->sync_mutex);
3449         ret = stop_sync_thread(ipvs,
3450                                nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3451         mutex_unlock(&ipvs->sync_mutex);
3452         return ret;
3453 }
3454
3455 static int ip_vs_genl_set_config(struct netns_ipvs *ipvs, struct nlattr **attrs)
3456 {
3457         struct ip_vs_timeout_user t;
3458
3459         __ip_vs_get_timeouts(ipvs, &t);
3460
3461         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3462                 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3463
3464         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3465                 t.tcp_fin_timeout =
3466                         nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3467
3468         if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3469                 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3470
3471         return ip_vs_set_timeout(ipvs, &t);
3472 }
3473
3474 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3475 {
3476         int ret = -EINVAL, cmd;
3477         struct net *net = sock_net(skb->sk);
3478         struct netns_ipvs *ipvs = net_ipvs(net);
3479
3480         cmd = info->genlhdr->cmd;
3481
3482         if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3483                 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3484
3485                 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3486                     nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3487                                      info->attrs[IPVS_CMD_ATTR_DAEMON],
3488                                      ip_vs_daemon_policy))
3489                         goto out;
3490
3491                 if (cmd == IPVS_CMD_NEW_DAEMON)
3492                         ret = ip_vs_genl_new_daemon(ipvs, daemon_attrs);
3493                 else
3494                         ret = ip_vs_genl_del_daemon(ipvs, daemon_attrs);
3495         }
3496
3497 out:
3498         return ret;
3499 }
3500
3501 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3502 {
3503         struct ip_vs_service *svc = NULL;
3504         struct ip_vs_service_user_kern usvc;
3505         struct ip_vs_dest_user_kern udest;
3506         int ret = 0, cmd;
3507         int need_full_svc = 0, need_full_dest = 0;
3508         struct net *net = sock_net(skb->sk);
3509         struct netns_ipvs *ipvs = net_ipvs(net);
3510
3511         cmd = info->genlhdr->cmd;
3512
3513         mutex_lock(&__ip_vs_mutex);
3514
3515         if (cmd == IPVS_CMD_FLUSH) {
3516                 ret = ip_vs_flush(ipvs, false);
3517                 goto out;
3518         } else if (cmd == IPVS_CMD_SET_CONFIG) {
3519                 ret = ip_vs_genl_set_config(ipvs, info->attrs);
3520                 goto out;
3521         } else if (cmd == IPVS_CMD_ZERO &&
3522                    !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3523                 ret = ip_vs_zero_all(ipvs);
3524                 goto out;
3525         }
3526
3527         /* All following commands require a service argument, so check if we
3528          * received a valid one. We need a full service specification when
3529          * adding / editing a service. Only identifying members otherwise. */
3530         if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3531                 need_full_svc = 1;
3532
3533         ret = ip_vs_genl_parse_service(ipvs, &usvc,
3534                                        info->attrs[IPVS_CMD_ATTR_SERVICE],
3535                                        need_full_svc, &svc);
3536         if (ret)
3537                 goto out;
3538
3539         /* Unless we're adding a new service, the service must already exist */
3540         if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3541                 ret = -ESRCH;
3542                 goto out;
3543         }
3544
3545         /* Destination commands require a valid destination argument. For
3546          * adding / editing a destination, we need a full destination
3547          * specification. */
3548         if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3549             cmd == IPVS_CMD_DEL_DEST) {
3550                 if (cmd != IPVS_CMD_DEL_DEST)
3551                         need_full_dest = 1;
3552
3553                 ret = ip_vs_genl_parse_dest(&udest,
3554                                             info->attrs[IPVS_CMD_ATTR_DEST],
3555                                             need_full_dest);
3556                 if (ret)
3557                         goto out;
3558
3559                 /* Old protocols did not allow the user to specify address
3560                  * family, so we set it to zero instead.  We also didn't
3561                  * allow heterogeneous pools in the old code, so it's safe
3562                  * to assume that this will have the same address family as
3563                  * the service.
3564                  */
3565                 if (udest.af == 0)
3566                         udest.af = svc->af;
3567
3568                 if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) {
3569                         /* The synchronization protocol is incompatible
3570                          * with mixed family services
3571                          */
3572                         if (ipvs->sync_state) {
3573                                 ret = -EINVAL;
3574                                 goto out;
3575                         }
3576
3577                         /* Which connection types do we support? */
3578                         switch (udest.conn_flags) {
3579                         case IP_VS_CONN_F_TUNNEL:
3580                                 /* We are able to forward this */
3581                                 break;
3582                         default:
3583                                 ret = -EINVAL;
3584                                 goto out;
3585                         }
3586                 }
3587         }
3588
3589         switch (cmd) {
3590         case IPVS_CMD_NEW_SERVICE:
3591                 if (svc == NULL)
3592                         ret = ip_vs_add_service(ipvs, &usvc, &svc);
3593                 else
3594                         ret = -EEXIST;
3595                 break;
3596         case IPVS_CMD_SET_SERVICE:
3597                 ret = ip_vs_edit_service(svc, &usvc);
3598                 break;
3599         case IPVS_CMD_DEL_SERVICE:
3600                 ret = ip_vs_del_service(svc);
3601                 /* do not use svc, it can be freed */
3602                 break;
3603         case IPVS_CMD_NEW_DEST:
3604                 ret = ip_vs_add_dest(svc, &udest);
3605                 break;
3606         case IPVS_CMD_SET_DEST:
3607                 ret = ip_vs_edit_dest(svc, &udest);
3608                 break;
3609         case IPVS_CMD_DEL_DEST:
3610                 ret = ip_vs_del_dest(svc, &udest);
3611                 break;
3612         case IPVS_CMD_ZERO:
3613                 ret = ip_vs_zero_service(svc);
3614                 break;
3615         default:
3616                 ret = -EINVAL;
3617         }
3618
3619 out:
3620         mutex_unlock(&__ip_vs_mutex);
3621
3622         return ret;
3623 }
3624
3625 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3626 {
3627         struct sk_buff *msg;
3628         void *reply;
3629         int ret, cmd, reply_cmd;
3630         struct net *net = sock_net(skb->sk);
3631         struct netns_ipvs *ipvs = net_ipvs(net);
3632
3633         cmd = info->genlhdr->cmd;
3634
3635         if (cmd == IPVS_CMD_GET_SERVICE)
3636                 reply_cmd = IPVS_CMD_NEW_SERVICE;
3637         else if (cmd == IPVS_CMD_GET_INFO)
3638                 reply_cmd = IPVS_CMD_SET_INFO;
3639         else if (cmd == IPVS_CMD_GET_CONFIG)
3640                 reply_cmd = IPVS_CMD_SET_CONFIG;
3641         else {
3642                 pr_err("unknown Generic Netlink command\n");
3643                 return -EINVAL;
3644         }
3645
3646         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3647         if (!msg)
3648                 return -ENOMEM;
3649
3650         mutex_lock(&__ip_vs_mutex);
3651
3652         reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3653         if (reply == NULL)
3654                 goto nla_put_failure;
3655
3656         switch (cmd) {
3657         case IPVS_CMD_GET_SERVICE:
3658         {
3659                 struct ip_vs_service *svc;
3660
3661                 svc = ip_vs_genl_find_service(ipvs,
3662                                               info->attrs[IPVS_CMD_ATTR_SERVICE]);
3663                 if (IS_ERR(svc)) {
3664                         ret = PTR_ERR(svc);
3665                         goto out_err;
3666                 } else if (svc) {
3667                         ret = ip_vs_genl_fill_service(msg, svc);
3668                         if (ret)
3669                                 goto nla_put_failure;
3670                 } else {
3671                         ret = -ESRCH;
3672                         goto out_err;
3673                 }
3674
3675                 break;
3676         }
3677
3678         case IPVS_CMD_GET_CONFIG:
3679         {
3680                 struct ip_vs_timeout_user t;
3681
3682                 __ip_vs_get_timeouts(ipvs, &t);
3683 #ifdef CONFIG_IP_VS_PROTO_TCP
3684                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3685                                 t.tcp_timeout) ||
3686                     nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3687                                 t.tcp_fin_timeout))
3688                         goto nla_put_failure;
3689 #endif
3690 #ifdef CONFIG_IP_VS_PROTO_UDP
3691                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3692                         goto nla_put_failure;
3693 #endif
3694
3695                 break;
3696         }
3697
3698         case IPVS_CMD_GET_INFO:
3699                 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3700                                 IP_VS_VERSION_CODE) ||
3701                     nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3702                                 ip_vs_conn_tab_size))
3703                         goto nla_put_failure;
3704                 break;
3705         }
3706
3707         genlmsg_end(msg, reply);
3708         ret = genlmsg_reply(msg, info);
3709         goto out;
3710
3711 nla_put_failure:
3712         pr_err("not enough space in Netlink message\n");
3713         ret = -EMSGSIZE;
3714
3715 out_err:
3716         nlmsg_free(msg);
3717 out:
3718         mutex_unlock(&__ip_vs_mutex);
3719
3720         return ret;
3721 }
3722
3723
3724 static const struct genl_ops ip_vs_genl_ops[] = {
3725         {
3726                 .cmd    = IPVS_CMD_NEW_SERVICE,
3727                 .flags  = GENL_ADMIN_PERM,
3728                 .policy = ip_vs_cmd_policy,
3729                 .doit   = ip_vs_genl_set_cmd,
3730         },
3731         {
3732                 .cmd    = IPVS_CMD_SET_SERVICE,
3733                 .flags  = GENL_ADMIN_PERM,
3734                 .policy = ip_vs_cmd_policy,
3735                 .doit   = ip_vs_genl_set_cmd,
3736         },
3737         {
3738                 .cmd    = IPVS_CMD_DEL_SERVICE,
3739                 .flags  = GENL_ADMIN_PERM,
3740                 .policy = ip_vs_cmd_policy,
3741                 .doit   = ip_vs_genl_set_cmd,
3742         },
3743         {
3744                 .cmd    = IPVS_CMD_GET_SERVICE,
3745                 .flags  = GENL_ADMIN_PERM,
3746                 .doit   = ip_vs_genl_get_cmd,
3747                 .dumpit = ip_vs_genl_dump_services,
3748                 .policy = ip_vs_cmd_policy,
3749         },
3750         {
3751                 .cmd    = IPVS_CMD_NEW_DEST,
3752                 .flags  = GENL_ADMIN_PERM,
3753                 .policy = ip_vs_cmd_policy,
3754                 .doit   = ip_vs_genl_set_cmd,
3755         },
3756         {
3757                 .cmd    = IPVS_CMD_SET_DEST,
3758                 .flags  = GENL_ADMIN_PERM,
3759                 .policy = ip_vs_cmd_policy,
3760                 .doit   = ip_vs_genl_set_cmd,
3761         },
3762         {
3763                 .cmd    = IPVS_CMD_DEL_DEST,
3764                 .flags  = GENL_ADMIN_PERM,
3765                 .policy = ip_vs_cmd_policy,
3766                 .doit   = ip_vs_genl_set_cmd,
3767         },
3768         {
3769                 .cmd    = IPVS_CMD_GET_DEST,
3770                 .flags  = GENL_ADMIN_PERM,
3771                 .policy = ip_vs_cmd_policy,
3772                 .dumpit = ip_vs_genl_dump_dests,
3773         },
3774         {
3775                 .cmd    = IPVS_CMD_NEW_DAEMON,
3776                 .flags  = GENL_ADMIN_PERM,
3777                 .policy = ip_vs_cmd_policy,
3778                 .doit   = ip_vs_genl_set_daemon,
3779         },
3780         {
3781                 .cmd    = IPVS_CMD_DEL_DAEMON,
3782                 .flags  = GENL_ADMIN_PERM,
3783                 .policy = ip_vs_cmd_policy,
3784                 .doit   = ip_vs_genl_set_daemon,
3785         },
3786         {
3787                 .cmd    = IPVS_CMD_GET_DAEMON,
3788                 .flags  = GENL_ADMIN_PERM,
3789                 .dumpit = ip_vs_genl_dump_daemons,
3790         },
3791         {
3792                 .cmd    = IPVS_CMD_SET_CONFIG,
3793                 .flags  = GENL_ADMIN_PERM,
3794                 .policy = ip_vs_cmd_policy,
3795                 .doit   = ip_vs_genl_set_cmd,
3796         },
3797         {
3798                 .cmd    = IPVS_CMD_GET_CONFIG,
3799                 .flags  = GENL_ADMIN_PERM,
3800                 .doit   = ip_vs_genl_get_cmd,
3801         },
3802         {
3803                 .cmd    = IPVS_CMD_GET_INFO,
3804                 .flags  = GENL_ADMIN_PERM,
3805                 .doit   = ip_vs_genl_get_cmd,
3806         },
3807         {
3808                 .cmd    = IPVS_CMD_ZERO,
3809                 .flags  = GENL_ADMIN_PERM,
3810                 .policy = ip_vs_cmd_policy,
3811                 .doit   = ip_vs_genl_set_cmd,
3812         },
3813         {
3814                 .cmd    = IPVS_CMD_FLUSH,
3815                 .flags  = GENL_ADMIN_PERM,
3816                 .doit   = ip_vs_genl_set_cmd,
3817         },
3818 };
3819
3820 static int __init ip_vs_genl_register(void)
3821 {
3822         return genl_register_family_with_ops(&ip_vs_genl_family,
3823                                              ip_vs_genl_ops);
3824 }
3825
3826 static void ip_vs_genl_unregister(void)
3827 {
3828         genl_unregister_family(&ip_vs_genl_family);
3829 }
3830
3831 /* End of Generic Netlink interface definitions */
3832
3833 /*
3834  * per netns intit/exit func.
3835  */
3836 #ifdef CONFIG_SYSCTL
3837 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs)
3838 {
3839         struct net *net = ipvs->net;
3840         int idx;
3841         struct ctl_table *tbl;
3842
3843         atomic_set(&ipvs->dropentry, 0);
3844         spin_lock_init(&ipvs->dropentry_lock);
3845         spin_lock_init(&ipvs->droppacket_lock);
3846         spin_lock_init(&ipvs->securetcp_lock);
3847
3848         if (!net_eq(net, &init_net)) {
3849                 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3850                 if (tbl == NULL)
3851                         return -ENOMEM;
3852
3853                 /* Don't export sysctls to unprivileged users */
3854                 if (net->user_ns != &init_user_ns)
3855                         tbl[0].procname = NULL;
3856         } else
3857                 tbl = vs_vars;
3858         /* Initialize sysctl defaults */
3859         for (idx = 0; idx < ARRAY_SIZE(vs_vars); idx++) {
3860                 if (tbl[idx].proc_handler == proc_do_defense_mode)
3861                         tbl[idx].extra2 = ipvs;
3862         }
3863         idx = 0;
3864         ipvs->sysctl_amemthresh = 1024;
3865         tbl[idx++].data = &ipvs->sysctl_amemthresh;
3866         ipvs->sysctl_am_droprate = 10;
3867         tbl[idx++].data = &ipvs->sysctl_am_droprate;
3868         tbl[idx++].data = &ipvs->sysctl_drop_entry;
3869         tbl[idx++].data = &ipvs->sysctl_drop_packet;
3870 #ifdef CONFIG_IP_VS_NFCT
3871         tbl[idx++].data = &ipvs->sysctl_conntrack;
3872 #endif
3873         tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3874         ipvs->sysctl_snat_reroute = 1;
3875         tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3876         ipvs->sysctl_sync_ver = 1;
3877         tbl[idx++].data = &ipvs->sysctl_sync_ver;
3878         ipvs->sysctl_sync_ports = 1;
3879         tbl[idx++].data = &ipvs->sysctl_sync_ports;
3880         tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3881         ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3882         tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3883         ipvs->sysctl_sync_sock_size = 0;
3884         tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3885         tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3886         tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3887         tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3888         tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3889         tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3890         ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3891         ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3892         tbl[idx].data = &ipvs->sysctl_sync_threshold;
3893         tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3894         ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3895         tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3896         ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3897         tbl[idx++].data = &ipvs->sysctl_sync_retries;
3898         tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3899         ipvs->sysctl_pmtu_disc = 1;
3900         tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3901         tbl[idx++].data = &ipvs->sysctl_backup_only;
3902         ipvs->sysctl_conn_reuse_mode = 1;
3903         tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
3904         tbl[idx++].data = &ipvs->sysctl_schedule_icmp;
3905         tbl[idx++].data = &ipvs->sysctl_ignore_tunneled;
3906
3907         ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3908         if (ipvs->sysctl_hdr == NULL) {
3909                 if (!net_eq(net, &init_net))
3910                         kfree(tbl);
3911                 return -ENOMEM;
3912         }
3913         ip_vs_start_estimator(ipvs, &ipvs->tot_stats);
3914         ipvs->sysctl_tbl = tbl;
3915         /* Schedule defense work */
3916         INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3917         schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3918
3919         return 0;
3920 }
3921
3922 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs)
3923 {
3924         struct net *net = ipvs->net;
3925
3926         cancel_delayed_work_sync(&ipvs->defense_work);
3927         cancel_work_sync(&ipvs->defense_work.work);
3928         unregister_net_sysctl_table(ipvs->sysctl_hdr);
3929         ip_vs_stop_estimator(ipvs, &ipvs->tot_stats);
3930
3931         if (!net_eq(net, &init_net))
3932                 kfree(ipvs->sysctl_tbl);
3933 }
3934
3935 #else
3936
3937 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) { return 0; }
3938 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs) { }
3939
3940 #endif
3941
3942 static struct notifier_block ip_vs_dst_notifier = {
3943         .notifier_call = ip_vs_dst_event,
3944 };
3945
3946 int __net_init ip_vs_control_net_init(struct netns_ipvs *ipvs)
3947 {
3948         int i, idx;
3949
3950         /* Initialize rs_table */
3951         for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3952                 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
3953
3954         INIT_LIST_HEAD(&ipvs->dest_trash);
3955         spin_lock_init(&ipvs->dest_trash_lock);
3956         setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
3957                     (unsigned long) ipvs);
3958         atomic_set(&ipvs->ftpsvc_counter, 0);
3959         atomic_set(&ipvs->nullsvc_counter, 0);
3960
3961         /* procfs stats */
3962         ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3963         if (!ipvs->tot_stats.cpustats)
3964                 return -ENOMEM;
3965
3966         for_each_possible_cpu(i) {
3967                 struct ip_vs_cpu_stats *ipvs_tot_stats;
3968                 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
3969                 u64_stats_init(&ipvs_tot_stats->syncp);
3970         }
3971
3972         spin_lock_init(&ipvs->tot_stats.lock);
3973
3974         proc_create("ip_vs", 0, ipvs->net->proc_net, &ip_vs_info_fops);
3975         proc_create("ip_vs_stats", 0, ipvs->net->proc_net, &ip_vs_stats_fops);
3976         proc_create("ip_vs_stats_percpu", 0, ipvs->net->proc_net,
3977                     &ip_vs_stats_percpu_fops);
3978
3979         if (ip_vs_control_net_init_sysctl(ipvs))
3980                 goto err;
3981
3982         return 0;
3983
3984 err:
3985         free_percpu(ipvs->tot_stats.cpustats);
3986         return -ENOMEM;
3987 }
3988
3989 void __net_exit ip_vs_control_net_cleanup(struct netns_ipvs *ipvs)
3990 {
3991         ip_vs_trash_cleanup(ipvs);
3992         ip_vs_control_net_cleanup_sysctl(ipvs);
3993         remove_proc_entry("ip_vs_stats_percpu", ipvs->net->proc_net);
3994         remove_proc_entry("ip_vs_stats", ipvs->net->proc_net);
3995         remove_proc_entry("ip_vs", ipvs->net->proc_net);
3996         free_percpu(ipvs->tot_stats.cpustats);
3997 }
3998
3999 int __init ip_vs_register_nl_ioctl(void)
4000 {
4001         int ret;
4002
4003         ret = nf_register_sockopt(&ip_vs_sockopts);
4004         if (ret) {
4005                 pr_err("cannot register sockopt.\n");
4006                 goto err_sock;
4007         }
4008
4009         ret = ip_vs_genl_register();
4010         if (ret) {
4011                 pr_err("cannot register Generic Netlink interface.\n");
4012                 goto err_genl;
4013         }
4014         return 0;
4015
4016 err_genl:
4017         nf_unregister_sockopt(&ip_vs_sockopts);
4018 err_sock:
4019         return ret;
4020 }
4021
4022 void ip_vs_unregister_nl_ioctl(void)
4023 {
4024         ip_vs_genl_unregister();
4025         nf_unregister_sockopt(&ip_vs_sockopts);
4026 }
4027
4028 int __init ip_vs_control_init(void)
4029 {
4030         int idx;
4031         int ret;
4032
4033         EnterFunction(2);
4034
4035         /* Initialize svc_table, ip_vs_svc_fwm_table */
4036         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
4037                 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
4038                 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
4039         }
4040
4041         smp_wmb();      /* Do we really need it now ? */
4042
4043         ret = register_netdevice_notifier(&ip_vs_dst_notifier);
4044         if (ret < 0)
4045                 return ret;
4046
4047         LeaveFunction(2);
4048         return 0;
4049 }
4050
4051
4052 void ip_vs_control_cleanup(void)
4053 {
4054         EnterFunction(2);
4055         unregister_netdevice_notifier(&ip_vs_dst_notifier);
4056         LeaveFunction(2);
4057 }